Reciprocal connectivity between secondary auditory cortical field and amygdala in mice

Recent studies have examined the feedback pathway from the amygdala to the auditory cortex in conjunction with the feedforward pathway from the auditory cortex to the amygdala. However, these connections have not been fully characterized. Here, to visualize the comprehensive connectivity between the auditory cortex and amygdala, we injected cholera toxin subunit b (CTB), a bidirectional tracer, into multiple subfields in the mouse auditory cortex after identifying the location of these subfields using flavoprotein fluorescence imaging. After injecting CTB into the secondary auditory field (A2), we found densely innervated CTB-positive axon terminals that were mainly located in the lateral amygdala (La), and slight innervations in other divisions such as the basal amygdala. Moreover, we found a large number of retrogradely-stained CTB-positive neurons in La after injecting CTB into A2. When injecting CTB into the primary auditory cortex (A1), a small number of CTB-positive neurons and axons were visualized in the amygdala. Finally, we found a near complete absence of connections between the other auditory cortical fields and the amygdala. These data suggest that reciprocal connections between A2 and La are main conduits for communication between the auditory cortex and amygdala in mice.

www.nature.com/scientificreports www.nature.com/scientificreports/ auditory cortex 22,23 . Recent studies characterized a feedback pathway from La to the auditory cortex in mice 24,25 , which were consistent with the finding of older studies that used monkeys 26,27 and rats 28 . However, neuronal tracing is often conducted without clear segregation between the auditory cortical subregions. As a result, whether the whole auditory cortex or just a local area has reciprocal connections with the amygdala is unclear. This could be addressed via systematic projection mapping between the auditory cortex and the amygdala. In the current study, we visualized comprehensive connectivity between the auditory cortex and the amygdala by injecting cholera toxin subunit b (CTB), a bidirectional tracer, into four tonotopic regions of the auditory cortex, according to a map that was functionally determined using flavoprotein fluorescence imaging 4 . We demonstrated that the A2 is specifically connected with the amygdala, which is consistent with the idea that A2 is a higher-order region in the mouse auditory cortex.

Results
Auditory cortical region-specific communications with the amygdala. Although there is much debate over how to map the mouse auditory cortex 29 , it is generally accepted that the mouse auditory cortex includes at least four tonotopic regions 3,4 . Here, we refer to them as DM, AAF, A1, and A2 (Fig. 1a,b). After identifying the precise location of AAF, A1, A2, and DM using flavoprotein fluorescence imaging, we injected CTB into the center of the responsive region to examine connective patterns with the amygdala. Deposits of CTB localized within the injected regions were confirmed after preparing brain slice sections (Fig. 1c).
We examined the distribution of CTB-positive axon terminals projected from the auditory cortex in the amygdala. The images shown in Fig. 2 were obtained by injecting CTB into A2. We found that densely distributed CTB-positive axon terminals were located mainly in La, and identified slight innervations in other divisions such as the central amygdala (CeA) and basal amygdala (BA), including the anterior basolateral amygdala (BLA), posterior basolateral amygdala (BLP), anterior basomedial amygdala (BMA), and posterior basomedial amygdala (BMP) (Fig. 2a-d). In the high-magnification images, we confirmed that the axon terminals projecting from A2 form many buttons and synaptic varicosities in La, which suggests that the projections from A2 neurons terminate in La (Fig. 2e). As for the three intercalated nuclei, which are a cluster of GABAergic neurons, the two dorsal nuclei the sides of La included some CTB-positive terminals. This is consistent with a previous study 15 . The most well-known cluster, which was located ventrally, included no CTB-positive axon terminals. These data suggest that La mainly receives input from the auditory cortex, and this result is consistent with the well-known view that La is an entrance in the amygdala for sensory input from sensory cortices 11 . Moreover, we found many CTB-positive somata, located mainly in La, after injecting CTB into A2 (Figs. 2 and 3). We quantitatively analyzed the number of CTB-positive somata in each amygdala division (Fig. 4a). Numerous CTB-positive neurons were distributed in La, and we found a small number of neurons in the basolateral and basomedial amygdala. Importantly, we found no CTB-positive somata in CeA (Fig. 4a), despite the proposed role of CeA as the output center of the amygdala 8 . We also conducted anterograde tracing by injecting a tracer into LA and confirmed that a number of axon terminal buttons were visualized in A2 (Supplemental Fig. 1), supporting the presence of projections from LA to A2. These data confirm that A2 projects mainly to La and receives direct feedback projections from La but not from CeA. We remarked that the density of CTB-positive axon terminals and the number of CTB-positive neurons in the amygdala differed by injection site in the auditory cortex. As described above, A2 has strong connections with La ( Figs. 3 and 4a,b). After injecting CTB into A1, we found one third the number of CTB-positive neurons compared with that observed after injecting CTB into A2 (Figs. 3 and 4a,b). The amount of CTB-positive axons was proportional to the number of CTB-positive neurons, indicating that A1 also had few reciprocal connections with La. Furthermore, we did not find CTB-positive axon terminals or CTB-positive neurons inside the amygdala after injecting CTB into AAF and DM (Figs. 3 and 4a,b). Although few CTB-positive neurons or axons were visualized in the amygdala after injecting CTB into AAF and DM, it was not due to a failure in our injection. As a  Fig. 2), indicating a successful injection. Interestingly, the numbers of CTB-positive neurons in the amygdala increased in proportion with the hierarchical order of the injected regions. This was reflected in the tonal response latencies. The quantitative latency data of each subregion were calculated based on our previous study 30 , in which latency was measured as quarter-maximal latencies of sound-induced flavoprotein fluorescence responses (Fig. 4c). Although the latency of flavoprotein fluorescence imaging is longer than electrophysiological methods, the auditory subregions are activated largely in the same order between the two methods 1,30 . These data indicate that A2 is reciprocally connected with La and that it is a main conduit for communication with the amygdala, while A1 has connections that are similar but much weaker in strength.
Two independent streams between the auditory cortex and La. Many deep brain parts have initially been considered to be single structures and are then found to actually contain several substructures that www.nature.com/scientificreports www.nature.com/scientificreports/ can be classified according to relationships with cortical targets 31,32 . To determine whether the origins of the La-A2 and La-A1 pathways are the spatially similar or different inside La, we evaluated the locations of La neurons projecting to A2 and A1. We found that amygdala neurons projecting to A2 were widely distributed rostrocaudally, while those projecting to A1 were comparatively localized in a caudal region (Fig. 4d). However, the distributions of amygdala neurons projecting to A2 and A1 were largely overlapped rostrocaudally. Therefore, we conducted double-injection of two fluorescence CTBs into A2 (green) and A1 (red) to investigate whether La neurons projecting to A1 and A2 were intermingled or topographically-segregated in the coronal view (Fig. 5a). We found that the two neuronal groups were distributed in a salt-and-pepper fashion and exhibited no clear spatial separation (Fig. 5b,c). However, most neurons were labeled with either tracer, and only 3.9% neurons were double-labeled (Fig. 5d). These data indicate that, although La is a single structure containing neurons with different cortical targets that are intermingled, the La-A2 and La-A1 pathways are mutually exclusive and mediated by discrete populations of amygdala neurons. www.nature.com/scientificreports www.nature.com/scientificreports/ The auditory cortex has tonotopic organization based on tonal frequencies. It is already known that the auditory cortex is connected to several deep brain structures in a frequency-related, topographical fashion, as indicated by auditory cortico-striatum projections 33 . To examine whether La and A2 are connected topographically, we injected two fluorescent CTBs into low and high frequency areas of A2 (Fig. 6). Although the distribution of the two neuronal groups was wider and more intermingled, the averaged locations of the two neuron types were significantly shifted in the dorsoventral direction on all rostrocaudal levels (Fig. 6b,c). These data suggest that the feedback projections from the amygdala to the auditory cortex have a frequency-related, topographic organization.

Discussion
The amygdala complex is important for associating affective information with sensory information. On the anatomical ground that La receives afferents from a wide range of locations in the thalamus and sensory cortex, auditory cues and footshock information converge in La 21 and are associated with one another based on synaptic plasticity 34,35 . Therefore, La is considered to be the center to the establishment of fear conditioning 11 . The associated information in La is then sent to CeA, which is indispensable for expressing freezing behaviors 36 . The output from CeA is then sent to a wide range of brain regions including the bed nucleus, hippocampus, midbrain, pons, and medulla oblongata 11,18 . This general scheme illustrates La as the input center and CeA as the output center of the amygdala. However, sensory cortices, including the auditory cortex, have not been found to be output targets of CeA. Indeed, old tracing studies had already revealed that the auditory cortex receives dense projections from the lateral parts of the amygdala in rats 28 and primates 27 . Both our findings and those of a recent study 24 confirm the presence of direct projections from La to the auditory cortex in mice, as well as the absence of projections from CeA to the auditory cortex. A similar scheme is also applicable to the visual system. The postrhinal area (POR) is a higher order visual cortical area that receives direct feedback projections from La but not from CeA in mice, and is modulated by La during processing of motivationally relevant sensory cues 37 . Therefore, it is reasonable to classify La, and not the CeA, as the main center in the amygdala that outputs information to the sensory cortex.
The results of the current study indicate that A2 specifically outputs and inputs signals to/from La. Although previous studies have reported evidence of a direct feedback pathway from La to the auditory cortex both in mice 24,25 and rats 28 , whether the entire auditory cortex is uniformly innervated by La has not been determined. As for the feedforward pathway from the auditory cortex to the amygdala, a consensus has not been made regarding whether the projections originate throughout the entire auditory cortex or are localized. Interestingly, one tracing study suggested that the ventral parts in the rat auditory cortex give rise to projections towards the amygdala 14 . However, the stereotaxic identification of the auditory cortex in this study was based on the classical regional annotation of Te1, Te2, and Te3 38 , which is quite different from the modern classification using AAF, A1, and A2. In the current study, the combination of neuronal tracing with prior functional identification of the auditory subregions using optical imaging enabled us to reveal the precise projectional topography based on the modern map of the auditory cortex. Thus, this study is the first to show that both the feedforward and feedback pathways between the auditory cortex and amygdala do not uniformly originate or terminate across the entire auditory cortex or amygdala in mice, and that A2 and La are main conduits for communication between the auditory cortex and amygdala.
In the current study, injections were conducted according to the auditory cortex map that had been established by our previous reports 4,6,30,31,[39][40][41] . However, the exact definition of the mouse auditory cortical regions is still under debate 29,42 , and interpretation of the current results could be differentiated depending on how the auditory cortex is mapped. Disagreements regarding the macroscale auditory cortical structure are derived from two major sources. First, it has not been determined whether the tonotopic direction of AAF is dorsoventral or ventrodorsal. The original map draws AAF as it runs ventrodorsally 2 , which influenced several following studies using electrophysiology 1 and voltage-sensitive-dye imaging 43 . In contrast, other studies, including our observations, draw AAF as it runs dorsoventrally using flavoprotein fluorescence imaging 4,44 , GCaMP3 3 , GCaMP6 imaging 42 , and intrinsic signal imaging 45 . A recent study attempted to reconcile these opposite results by proposing that AAF has branched tonotopy with two high frequency targets 29 . Although the precise location of the high frequency part of AAF remains unclear, all the studies identify the same location as the low frequency part of AAF. In the current study, we injected CTB into the low frequency part of AAF; therefore, the current results regarding injection into AAF can be generalized across different auditory cortex maps. Second, it has not been determined whether the tonotopic regions of A1 and DM are different regions or a single region. The original study 2 , electrophysiology 1 , and voltage-sensitive-dye imaging studies 43 identify one tonotopic region of A1 in the caudal auditory cortex. Studies using GCaMP3 3 and GCaMP6 imaging 42 suggest the possibility that the A1 tonotopy is branched as it runs from one low frequency part towards two high frequency parts. Our previous studies suggest that two branches in A1 can be split into two distinct subregions and named the dorsal region DM and the ventral region A1 46 . Intrinsic signal imaging studies agree with our findings although name these two subregions A1 and the ventral auditory field (VAF), respectively 47 . If there are not two separate regions, but rather one single causal region, the current results would be unchanged. CTB-positive neurons for A1 and DM would be combined, and the mean number of CTB-positive neurons found in the amygdala per animal is changed to 319.6 (A2 injection), 42.3 ("enlarged A1" injection), and 3.0 neurons (AAF injection), from 319.6 (A2 injection), 81.0 (A1 injection), 3.0 (AAF injection), and 3.5 neurons (DM injection). Even after conducting this reanalysis, the number of CTB-positive axons and somata in the amygdala are significantly larger when injecting CTB into A2. Therefore, predominance of A2 over other auditory subregions in terms of connectivity with the amygdala is not changed depending on definition of the auditory cortex map. The current study was conducted based on the auditory cortex map that had been described in our previous studies, where A1 was assigned to the second most ventral part. However, it is more straightforward that the region annotated as A1 sits in the location with www.nature.com/scientificreports www.nature.com/scientificreports/ the shortest latency. The recent find-grained mapping using optical imaging has paved the way for the possibility that the auditory cortex map can be comparable between mice and rats; therefore, despite debate over defining auditory subregions, the terminology could be unified between mice and rats in future. www.nature.com/scientificreports www.nature.com/scientificreports/ Overall, our findings endorse the notion that A2 is a higher-order auditory region in the mouse auditory cortex. The auditory subregions of DM, AAF, A1, and A2 have been found to be tonotopic 4 , and all of these fields receive substantial projections from the lemniscal auditory thalamus of the ventral division of MGB (MGv) 4,40 . Because the general concept explains that cortical areas that receive afferents from MGv are primary auditory areas, a question arises regarding whether these fields, including A2, are all primary areas in the mouse cortex. However, the results of the current study clearly show the presence of dense connections between A2 and the amygdala, while other auditory fields have slight or no connections with the amygdala. Moreover, previous studies have reported that A2 neurons have long latencies 1,30,48 , wide bandwidth 3,40 , and a less-ordered tonotopic arrangement 3,40 . These data confirm that mouse A2 has properties of a higher-order region, and also support the idea that auditory cortical subregions are functionally specialized in mice 5,6,49 . Given that primary-like regions have few feedforward or feedback connections with the amygdala, they may work to analyze auditory information, as previously proposed. Although our data showed a relatively long latency in the region labeled A1, the connectivity with the amygdala was faint, indicating that A1 may function as a primary-like region. In contrast, A2 may work in cooperation with the amygdala to associate affective information with auditory information after it receives auditory information from primary fields via the corticocortical pathways 50,51 . A recent study using rats suggested that value-coding neurons exist in parts that correspond to those located near mouse A2 52,53 , which may be evidence of the La-A2 feedback pathways. It is possible that A2 works to provide biological meaning for emotion-arousing sounds such as conspecific courtship songs and predatory calls, and thus plays a more important role in signal processing than simple sound analysis.

Methods
Animals. The experimental procedures in the present study were approved by the Committee for Animal Care at Niigata University (Protocol Number: SA00235), and carried out in accordance with the approved guidelines. Data were obtained from male C57BL/6 mice at the age of 7-8 weeks (Charles River Japan, Kanagawa, Japan). The animals were housed in cages with ad libitum access to food pellets and water, and were kept on a 12-h light/ dark cycle.
Flavoprotein fluorescence imaging. Locations of regions in right auditory cortex were identified according to responses revealed using flavoprotein fluorescence imaging 54 . Cortical images were recorded by a CCD camera system (AQUACOSMOS with ORCA-R2, Hamamatsu Photonics, Shizuoka, Japan) via an epifluorescence microscope (Ex, 500-550 nm; Em, 470-490 nm; M651 combined with MZ FL II, Leica Microsystems, Wetzlar, Germany). The area covered by one pixel was 20.4 × 20.4 µm. Mice were anesthetized with urethane (1.65 g/kg, i.p.; Wako, Osaka, Japan). The rectal temperature was maintained at ~37 °C. A craniotomy (~3 × 3 mm) was performed over the right auditory cortex. The auditory cortex was activated by presentation of sound waves that were made using a LabVIEW program (National Instruments, Austin, TX). Tones were amplitude modulated (20 Hz, 100% modulation) and set to ~60 dBSPL. The location of auditory cortical subregions was identified according to the tonotopic shift that were visualized using 5 and 30 kHz tones. neural tracer injection. A glass pipette (tip diameter ~30 μm) filled with cholera toxin subunit b (CTB) solution was introduced into the center of auditory regions visualized and identified using optical imaging, to ~450 μm below the cortical surface so that tracer solutions spread to all the cortical layers. Here, we used a low-salt type CTB (#104; List Biological Laboratory, Campbell, CA) that is suitable for iontophoretic injection 55 . CTB was injected by carrying 70 pulses, 5 s on/5 s off anodal currents at an intensity of 4 μA. Partly, Alexa Fluor 488-and 555-conjugated CTBs (0.5% in 0.1 M phosphate buffer, Thermo Fisher Scientific, Waltham, MA) were iontophoretically injected in A1 and A2 each in the same animals by applying 70 pulses. After finishing injections, a glass pipette was slowly withdrawn. The cranial hole was covered using 2% agarose (1-B, Sigma-Aldrich, MO) and the skin was sutured. Mice were placed in a warm place for recovery, and after awaking they were reared in their home cages. In several experiments, we injected a 10% biotinylated dextran amine (BDA, mw: 3,000, Thermo Fisher Scientific) solution by carrying ~40 pulses, 7 s on/7 s off anodal currents at an intensity of 5 μA via a ~20 μm-thick pipette.
To visualize CTB, sections were initially rinsed in 20 mM PBS and incubated in PBS containing 3% hydrogen peroxide and 0.1% Triton X-100 for 15 min at room temperature. After incubating in 20 mM PBS containing 0.1% Triton X-100 (PBST) for 60 min, slices were incubated overnight at room temperature with goat anti-CTB antibody (List Biological Laboratories) diluted to 1:30,000 with 20 mM PBS containing 0.5% skim milk. The next day, slices were rinsed in 20 mM PBS, and incubated at room temperature for 2 h with HRP-conjugated rabbit anti-goat IgG antibody (MBL, Nagoya, Japan) diluted to 1:200 using 20 mM PBS containing 0.5% skim milk. Sections were rinsed in 20 mM PBS, and the immunoreactions were visualized in a Tris-HCl buffer containing 0.05% diaminobenzidine tetrahydrochloride and 0.003% hydrogen peroxide for 5 min at room temperature. After visualization, slices were Nissl-stained using 0.1% cresyl violet (Chroma Gesellschaft, Kongen, Germany), and they were dehydrated in ethanol, cleared in xylene, and cover-slipped using the covering reagent Bioleit (Okenshoji, Tokyo, Japan). When coverslipping sections with fluorescent CTB, Fluoromount (Cosmo Bio, Tokyo, Japan) was used as a covering reagent instead. To visualize BDA, slices were incubated in a Tris-HCl buffer containing 0.05% diaminobenzidine tetrahydrochloride and 0.003% hydrogen peroxide for 20 min.