Dedicated C‐fibre viscerosensory pathways to central nucleus of the amygdala

Abstract

Key points

• Emotions are accompanied by concordant changes in visceral function, including cardiac output, respiration and digestion. One major forebrain integrator of emotional responses, the amygdala, is considered to rely on embedded visceral afferent information, although few details are known.

• In the present study, we retrogradely transported dye from the central nucleus of the amygdala (CeA) to identify CeA‐projecting nucleus of the solitary tract (NTS) neurons for synaptic characterization and compared them with unlabelled, near‐neighboor NTS neurons.

• Solitary tract (ST) afferents converged onto NTS‐CeA second‐order sensory neurons in greater numbers, as well as indirectly via polysynaptic pathways.

• Unexpectedly, all mono‐ and polysynaptic ST afferent pathways to NTS‐CeA neurons were organized exclusively as either transient receptor potential cation channel subfamily V member 1 (TRPV1)‐sensitive or TRPV1‐resistant, regardless of whether intervening neurons were excitatory or inhibitory.

• This strict sorting provides viscerosensory signals to CeA about visceral conditions with respect to being either ‘normal’ via A‐fibres or ‘alarm’ via TRPV1 expressing C‐fibres and, accordingly, this pathway organization probably encodes interoceptive status.

Abstract

Emotional state is impacted by changes in visceral function, including blood pressure, breathing and digestion. A main line of viscerosensory information processing occurs first in the nucleus of the solitary tract (NTS). In the present study conducted in rats, we examined the synaptic characteristics of visceral afferent pathways to the central nucleus of the amygdala (CeA) in brainstem slices by recording from retrogradely labelled NTS projection neurons. We simultaneously recorded neuron pairs: one dye positive (i.e. NTS‐CeA) and a second unlabelled neighbour. Graded shocks to the solitary tract (ST) always (93%) triggered EPSCs at CeA projecting NTS neurons. Half of the NTS‐CeA neurons received at least one primary afferent input (classed ‘second order’) indicating that viscerosensory information arrives at the CeA conveyed via a pathway involving as few as two synapses. The remaining NTS‐CeA neurons received viscerosensory input only via polysynaptic pathways. By contrast, ∼3/4 of unlabelled neighbouring neurons were directly connected to ST. NTS‐CeA neurons received greater numbers of ST‐related inputs compared to unlabelled NTS neurons, indicating that highly convergent viscerosensory signals reach the CeA. Remarkably, despite multifibre convergence, all single NTS‐CeA neurons received inputs derived from only unmyelinated afferents [transient receptor potential cation channel subfamily V member 1 (TRPV1) expressing C‐fibres] or only non‐TRPV1 ST afferent inputs, and never a combination of both. Such segregation means that visceral afferent information followed separate lines to reach the CeA. Their very different physiological activation profiles mean that these parallel visceral afferent pathways encode viscerosensory signals to the amygdala that may provide interoceptive assessments to impact on behaviours.

Key points

• Emotions are accompanied by concordant changes in visceral function, including cardiac output, respiration and digestion. One major forebrain integrator of emotional responses, the amygdala, is considered to rely on embedded visceral afferent information, although few details are known.

• In the present study, we retrogradely transported dye from the central nucleus of the amygdala (CeA) to identify CeA‐projecting nucleus of the solitary tract (NTS) neurons for synaptic characterization and compared them with unlabelled, near‐neighboor NTS neurons.

• Solitary tract (ST) afferents converged onto NTS‐CeA second‐order sensory neurons in greater numbers, as well as indirectly via polysynaptic pathways.

• Unexpectedly, all mono‐ and polysynaptic ST afferent pathways to NTS‐CeA neurons were organized exclusively as either transient receptor potential cation channel subfamily V member 1 (TRPV1)‐sensitive or TRPV1‐resistant, regardless of whether intervening neurons were excitatory or inhibitory.

• This strict sorting provides viscerosensory signals to CeA about visceral conditions with respect to being either ‘normal’ via A‐fibres or ‘alarm’ via TRPV1 expressing C‐fibres and, accordingly, this pathway organization probably encodes interoceptive status.

Abbreviations

aCSF

artificial cerebrospinal fluid

CAP

capsaicin

CeA

central nucleus of the amygdala

NTS

solitary tract nucleus

PSC

postsynaptic current

ST

solitary tract

TRPV1

transient receptor potential cation channel subfamily V member 1

Introduction

Fear responses increase vigilance and arousal and contain imbedded viscerosensory surveillance of internal organ status (Garfinkel et al. 2014; Klarer et al. 2014; Pena et al. 2014). Although the autonomic components of fear responses are well known, the underlying mechanisms, structure and organization of visceral afferent contributions responsible for co‐ordinated, emotion‐appropriate visceral states are less clear (Parsons & Ressler, 2013; Garfinkel & Critchley, 2016). The amygdaloid complex deep within the mid‐temporal lobe is a key integrative centre for emotional processing of cognitive and visceral information (LeDoux, 2000; Maren & Quirk, 2004). Intense anatomical interconnections link the amygdala with cortical structures, although the central nucleus of the amygdala (CeA) is most closely associated with neurohumoral and brainstem autonomic aspects of conditioned fear responses (Sah et al. 2003; Ulrich‐Lai & Herman, 2009). The CeA critically orchestrates shifts in visceral homeostatic states via efferent projections to the brainstem at the periaqueductal grey, parabrachial nucleus and the nucleus of the solitary tract (NTS) and may help to distinguish between ‘physiological’ and ‘psychological’ stressors (Dayas et al. 2001; Ulrich‐Lai & Herman, 2009).

Primary afferents of the VII, IX and X cranial nerves convey the physiological status of visceral conditions routed along their solitary tract (ST) axons to the ST nucleus (NTS) (Andresen & Kunze, 1994). NTS neurons include reciprocal connections to and from specific CNS regions, including the amygdala (Loewy, 1990; Saper, 2002; Travagli et al. 2006), although little is known about the nature of the functional connections. Viscerosensory afferent excitation results in activation of NTS projection neurons that directly target the CeA (Ricardo & Koh, 1978; Danielsen et al. 1989; Kapp et al. 1989; Geerling & Loewy, 2006). This NTS to CeA pathway impacts on behavioural responses ranging from stress to satiety (Williams et al. 1998; Viltart et al. 2006). Viscerosensory information arrives via two broad phenotypic classes of afferents: the majority via C‐fibre primary afferent axons and the remainder along myelinated afferents. The C‐fibre cranial afferent neurons express transient receptor potential cation channel subfamily V member 1 (TRPV1), generally have high physiological activation thresholds and are often silent under normal conditions (Andresen et al. 2012). Thus, as a group, C‐fibres are generally most active in supraphysiological states of visceral function common in pathophysiology. A‐fibres generally lack TRPV1 and their low thresholds and high sensitivity mean that they are often quite active normally even during resting or basal conditions (Andresen et al. 2012). Despite this dichotomy, very little is known about the organization and qualities of information processing and flow beyond NTS to other brain destinations. We hypothesized that the direct NTS‐CeA pathway would be driven by convergent C‐fibre afferents. In the present study, we aimed to characterize cranial afferent inputs and also their fibre type and the connecting pathways to NTS neurons that project to the CeA.

Unexpectedly, our results indicate that cranial afferent information to the CeA follows two separated pathways segregated by the originating TRPV1‐based afferent phenotype. Incorporating well established differences in the physiological discharge characteristics of these phenotypic classes, this segregation means that the pathway lacking TRPV1 delivers information reporting largely resting/basal physiological conditions, whereas the TRPV1+ C‐fibre pathway signals primarily during extraordinary visceral circumstances that are common during stress or in pathophysiological states. This organization of segregated interoceptive afferent information probably helps to shape central integration at CeA during autonomic and stress‐like behaviours.

Methods

Ethical approval

All animal procedures were performed with the approval of the Institutional Animal Care and Use Committee at Oregon Health & Science University and conform to the guidelines of the National Institutes of Health publication Guide for the Care and Use of Laboratory Animals, and were in accordance with the Prevention of Cruelty to Animals Act 1986 under the guidelines of the National Health and Medical Research Council Code of Practice for the Care and Use of Animals for Experimental Purposes in Australia, and were also approved by the Animal Ethics Committee at the Florey Institute of Neuroscience and Mental Health.

Amygdala injections: retrograde tracer

Adult Sprague–Dawley male rats (Charles River Laboratories, Inc., Wilmington, MA, USA; Florey Core Animal Services, University of Melbourne, Parkville, VIC, Australia) (161 ± 7 g, n = 64) were anaesthetized (ketamine 60 mg kg^–1, xylazene 6 mg kg^–1 and acepromazine 1 mg kg^–1; i.p. at Oregon Health & Science University or 5% induction and 1.5–2% maintenance isoflurane at Florey Institute). The level of anaesthesia was confirmed by an absence of both flexor withdrawal and the corneal reflex. In a stereotaxic procedure to inject Retrobeads™ (100 nl; Red IX, Lumafluor Inc., Naples, FL, USA), rats were placed in a stereotaxic frame and an incision was made to gain access to the skull, holes were hand drilled through the skull over the target sites, and the dura mater was pierced. A glass pipette was lowed into the brain and retrograde tracer injected into the CeA cell column within co‐ordinates ranging from –2.0 to –2.3 anteroposterior, 4.0 to 4.22 mediolateral and –8.9 to 8.6 dorsoventral to produce injection sites from –1.88 to –3.3 mm anteroposterior of bregma. The pipette was removed, skin sutured and Meloxicam (Boehringer‐Ingelheim, Ingelheim am Rhein, Germany) (1 mg kg⁻¹, s.c.) was administered. Rats were housed individually with standard rat chow and water ad libitum and monitored (including body weight measurements) daily for 7 days after stereotaxic surgery and weekly thereafter. Rats recovered for a minimum of 10 days (mean ± SEM: 26 ± 2 days) before slice electrophysiology experiments were carried out. Following removal of the brainstem (see below), the forebrains were sectioned (200 μm) and injection sites photographed on the day of electrophysiology experiments for compilation with NTS neuron labelling in relation to dye location in the amygdala (Fig. 1).

Figure 1. NTS projection neurons terminate at the CeA.

Cartoons depict stereotaxic placement of dye into amygdala and its retrograde transport to the caudal NTS. A, coronal section that contains the terminal injection site within the left CeA and in which the corresponding NTS contained 13 labelled neurons. Scale = 2 mm. B, dye injection sites are depicted overlaid onto diagrams of the nuclear divisions within the amygdala where the central nucleus is shaded red over the rostrocaudal axis relative to bregma. A total of 442 dye‐labelled NTS‐CeA neurons was counted across 64 brains. The yields of back‐labelled NTS projection neurons in corresponding brainstem slices are depicted by transparency levels from zero to > 15 neurons per slice. Note that two additional ‘zero’ injections were mapped caudal of –3.14 mm bregma and are not shown. C, NTS‐CeA neuron yields plotted relative to injections sites reveals that the most consistent termination site of NTS projection neurons to the amygdala was restricted to the CeA. [Colour figure can be viewed at wileyonlinelibrary.com]

Brainstem slice preparation

Horizontal brainstem slices were prepared as described previously (Doyle & Andresen, 2001) and used to study the amygdala tracer injected rats. Briefly, adult rats (344 ± 11 g) were deeply anaesthetized with isoflurane (5%) and the brainstem was removed 10–30 days following amygdala tracer injection. Anaesthesia level was confirmed by the absence of the flexor withdrawal reflex, and the chest compressed to stop the heart, the spinal cord severed at C1 and skull bones removed to access and remove the brain stem. The medulla was rapidly cooled and trimmed rostral and caudal to yield a brainstem block centreed on obex. A rostral–caudal wedge of ventral brainstem was removed to orient the remaining tissue to yield a single, 250 μm thick, horizontal slice which contained the greatest length of ST axons together with the medial NTS. Slices were cut with a sapphire knife (Delaware Diamond Knives, Wilmington, DE, USA) mounted in a vibrating microtome (VT1000S; Leica Microsystems Inc., Bannockburn, IL, USA). The external solution was an artificial cerebrospinal fluid (aCSF) containing (mm): 125 NaCl, 3 KCl, 1.2 KH₂PO₄, 1.2 MgSO₄, 25 NaHCO₃, 10 dextrose and 2 CaCl₂. Slices were secured with a nylon mesh in a tissue chamber and perfused with aCSF at 34°C, 300 mOsm, and bubbled with 95% O₂–5% CO₂.

NTS‐CeA neuron quantification and correlation with injection site

The first phase of working with each slice began with a survey and count of the total number of retrogradely labelled fluorescent cell bodies within the single horizontal slice containing NTS. This survey work was performed using the 40× objective together with the appropriate filter set and all NTS subnuclei were checked. The tallies in each slice are registered in greyscale and overlaid at the corresponding dye injection sites within the amygdala (Fig. 1). Once this procedure was completed, the paired patch recording was commenced.

Paired whole cell recordings

Recording pipettes (2.8–4.2 MΩ) were guided by fluorescence labelling localized to neurons in the caudal NTS using anatomical landmarks (Fig. 2). Neurons were visualized (Doyle et al. 2004) using infrared illumination with differential interference contrast optics (40× water immersion lens) and filter set 15 with mercury lamp on an Axioskop‐2 FS plus fixed stage microscope (Zeiss, Thornwood, NJ, USA) with a digital camera (Hamamatsu Photonic Systems, Bridgewater, NJ, USA). Paired recordings were attempted in most slices, where neighbouring, unlabelled neurons were selected near to dye positive cells for simultaneous recordings. Results from a total of 22 successful neuron pairs (NTS‐CeA with near neighbour unlabelled) are reported (Fig. 2). Pipettes were filled with a low Cl⁻ intracellular solution containing (mm): 6 NaCl, 4 NaOH, 130 K‐gluconate, 11 EGTA, 1 CaCl₂, 1 MgCl₂, 10 Hepes, 2 Na₂ATP and 0.2 Na₂GTP (pH 7.3 and 290 mOsm). As a consequence of E _Cl = –69 mV, IPSCs had small amplitudes at V _H = –60 mV, although more prominent outward current amplitudes were achieved by shifting to V _H = –40 mV in some cases. All recordings were made in open, whole cell patch configuration under voltage clamp using a Multiclamp 700B (Molecular Devices, Sunnyvale, CA, USA). Signals were sampled at 20 kHz and filtered at 10 kHz using p‐Clamp, version 10.2 (Molecular Devices). Reported voltages were corrected for the calculated liquid junction potential (–10.7 mV at 34°C).

Figure 2. NTS‐CeA projection and paired unlabelled neighbouring neurons were distributed throughout the ipsilateral medial NTS in horizontal slices of the caudal brainstem.

Cartoons depict stereotaxic placement of dye into amygdala and its retrograde transport to the caudal NTS. Aa and Ab, fluorescent beads identified NTS‐CeA projection neurons (green circle). Aa and Ac, neighbouring unlabelled neurons (blue square, located more ventral in this example) were recorded alongside NTS‐CeA neurons. Ad, paired recordings from medial NTS (mNTS) neurons were collected as the solitary tract was shocked with a concentric bipolar electrode (ST stim). p1 and 2, pipette 1 and 2; 4V, 4th ventricle; yellow box indicates the area depicted in (B). B, NTS‐CeA projection (n = 58) and paired unlabelled neurons (n = 22) were distributed throughout the ipsilateral medial NTS and were not observed elsewhere within any slice. [Colour figure can be viewed at wileyonlinelibrary.com]

ST shock intensity‐recruitment profiles

The concentric bipolar stimulating electrode (outer diameter 200 μm; Frederick Haer Co., Bowdoinham, ME, USA) was placed on the visible ST ∼1–3 mm rostral from recorded neuron cell bodies (Fig. 2). Minimal stimulus‐intensity protocols with incremental recruitment define discrete inputs to central neurons (Allen & Stevens, 1994; Rancz et al. 2007), where the delivery of shocks to activate afferents distant from the recorded cells produced discrete, unitary recruitment‐intensity profiles (Blitz & Regehr, 2005; Acuna‐Goycolea et al. 2008). Thus, passing current via the stimulating electrode activated ST primary afferents to trigger postsynaptic currents (PSCs). The remote placement of the electrode minimized the likelihood that electrical shocks would activate non‐ST axons or local neurons (Doyle & Andresen, 2001; Doyle et al. 2004; Bailey et al. 2008). Once stable recordings of each neuron were established, a series of test shocks graded in intensity was delivered to the ST. A finely graded synaptic recruitment routine (see Supporting Information; supplementary video and instruction file) uncovered the contributions of convergent individual inputs (McDougall et al. 2009). Individual inputs were discriminated on the basis of their characteristic latency, waveform and recruitment threshold. Multiple trials, together with the retesting of some ST shock intensities, indicated that these critical values were important to the characteristics of each input resulting in the recorded PSC. A Master‐8 isolated programmable stimulator (AMPI, Jerusalem, Israel) generated single shocks, as well as bursts of five ST shocks at 50 Hz every 6 s (shock duration 0.1 ms), for a minimum of 10 consecutive sweeps for each shock intensity trial. All neurons were tested with stimulus shock intensities from 0 to 1000 μA. From 0 to 100 μA, step changes were 10 μA (minimum) and, from 100 to 1000 μA, step changes were 100 μA (minimum). The order of shock intensity presentation varied. IPSCs were often recognized by their prolonged decaying phase and, to improve signal to noise ratios, the holding current was shifted to –40 mV to take advantage of an increased Cl⁻ driving force and shocks repeated. Plots of the event amplitudes (discriminated by arrival time and waveform) against stimulus intensity created stimulus recruitment profiles to compare across events and allowed the quantification of afferent input. The stimulus recruitment profile of each synaptic event at minimum intensity is considered to reflect the all‐or‐none axon activation characteristics of the ST fibre initiating the recorded EPSC or IPSC.

Discriminating mono‐ from polysynaptic and TRPV1+ from TRPV1– ST synaptic responses

Synaptic latency jitter and failure rates were considered as reliable indices that distinguished direct contacts from ST afferents (monosynaptic) vs. indirect (polysynaptic) afferent pathways to NTS neurons. Analysis of latency was based on the responses to the first shock in the train of five (PSC₁). Shock to shock variability in latency (synaptic jitter) for PSC₁ was calculated for > 30 events as the SD of latency and served as a critical indicator of synaptic order. PSC₁ jitters of < 200 μs were considered monosynaptic but jitters > 200 μs indicated a polysynaptic connection from the ST (Doyle & Andresen, 2001; Bailey et al. 2006). This synaptic jitter criterion matches the jitter characteristics of dye‐identified second‐order NTS neurons using anterogradely transported label from the aortic depressor nerve (Andresen & Peters, 2008). As expected, polysynaptic pathways activated by ST shocks were particularly prone to synaptic failures even with infrequent stimuli (Bailey et al. 2006; Andresen & Peters, 2008). Suprathreshold ST shocks that evoked no identifiable PSC within the latency time window characteristic of that PSC were counted as synaptic failures and failure rates were calculated for PSC₁ over 40 trials as a percentage of total ST shocks delivered at a constant, suprathreshold intensity. Neurons that received at least one excitatory monosynaptic input were defined as second‐order NTS neurons. Neurons that received only polysynaptic inputs to ST shocks were classified as higher‐order NTS neurons. Neurons that exhibited no ST‐evoked synaptic responses to high‐intensity shocks (up to 1000 μA) were considered ‘not connected’ to ST afferents but are included in compiled population totals. The final step in the synaptic interrogation was to expose slices to a continuous infusion of the TRPV1 agonist capsaicin (CAP) initially dissolved in ethanol and diluted with aCSF to 100 nm. CAP blockade of ST‐evoked synaptic events classified the initiating primary ST afferent as TRPV1+. The mechanism of CAP‐selective block appears to be depolarization of voltage‐dependent excitability in the ST afferent axons, which interrupts transmission from afferents expressing TRPV1 (Hofmann & Andresen, 2016).

Statistical analysis

ST evoked PSC amplitudes were based on 40 ‘successful’ iterative evoked events for each input. However, in neurons receiving converging inputs, compound synaptic events were excluded from amplitude analysis, despite their use in enumerating distinct inputs. Thus, we restricted this analysis to the minimal evoked response or those events sufficiently separated in onsets. Therefore, 60 of 138 input amplitudes in NTS‐CeA neurones and 15 of 42 input amplitudes in unlabelled NTS neurons were excluded from the overall means reported in the results. Failure rates were calculated from a minimum of 40 ST test shocks. Input variables, including latency, jitter, amplitude and failure rates for PSC₁, were not normally distributed and thus were compared between groups by rank values using a Kruskal–Wallis, one‐way ANOVA on ranks with a Dunn's post hoc test. Proportion data sets were compared using a one tail Z test and input convergence rates by one‐way ANOVA (SigmaStat, San Jose, CA, USA). All data are reported as the mean ± SEM. P > 0.05 was defined as statistically significant.

Results

NTS‐amygdala neurons project ipsilaterally to CeA

Stereotaxically directed injections resulted in dye deposited along the rostral–caudal extent but generally within the subnuclear borders of CeA (Fig. 1). To gauge the connections between injection sites in CeA (n = 64) and the yield of labelled neurons in the medial NTS, we mapped the location of dye injections and then registered the dye profile in a greyscale depiction of the number of fluorescence positive NTS neurons counted in corresponding slices (Fig. 1). Such yields ranged between zero and 26 neurons per injection in a single 250 μm corresponding NTS slice. All horizontal brainstem slices were inspected for labelled neurons in the contralateral medial NTS, ipsilateral lateral NTS, ipsilateral rostral NTS and, finally, the ipsilateral medial NTS. We could not observe the entire commissural NTS as a result of the tilted slice orientation needed to preserve ST axons connected with NTS neurons. The fluorescence labelled cell bodies of these NTS‐CeA projection neurons were found to be broadly distributed only ipsilaterally within the medial portions of NTS (Fig. 2). Cell bodies were robustly labelled but were generally isolated, with few adjacent NTS cells bearing retrograde dye (Fig. 2 A). No retrogradely labelled NTS‐CeA neurons were found lateral to the ST or contralateral across the midline (Fig. 2 B). Generally, injections made in the mid‐region of the CeA, from –2.2 to –2.56 mm relative to bregma, yielded the greatest numbers of labelled neurons in the NTS. This pattern of injection sites indicated that the central arbors of NTS‐CeA projection neurons were concentrated in the CeA division of the amygdala. Off‐centre injections (i.e. dye deposits that only partially overlapped with the borders of CeA) resulted in fewer retrogradely labelled NTS neurons. Dye injections that did not overlap CeA failed to label any neurons in NTS (n = 4 rats), suggesting that NTS neurons primarily project to the central rather than the lateral or basolateral subnuclei. These retrograde tracer findings suggest that NTS‐CeA projection neurons were broadly distributed in caudal portions of medial NTS and these neurons projected to the CeA with little evidence of contacting neighbouring subnuclei of the amygdala. Next, we determined whether such neurons could be activated by ST afferent pathways.

Graded ST afferent activation identifies multiple synaptic inputs

Single horizontal brainstem slices were assayed electrophysiologically from each animal. Under fluorescence illumination, dye positive cell bodies (NTS‐CeA neurons) were approached with a patch pipette for recording (Fig. 2 A). A second nearby neuron that was dye negative (unlabelled) was recorded with a second pipette. Once both neurons were ready for study, shocks were delivered to the ipsilateral ST and repeated every 6 s (0.17 Hz). Such an interval produces stable responses over extended periods (Peters et al. 2010). The shock intensity was gradually increased until a ST‐synchronized synaptic event was successfully activated in either neuron (Fig. 3). The graded intensity recruitment process was repeated using progressively higher and/or lower shock intensities aiming to best characterize the inputs until no further changes occurred in the response (see Supporting Information; supplementary video and instruction file). Each newly recruited synaptic event had a unique shock intensity threshold beyond which that event did not change in latency, shape or amplitude (i.e. all non‐compound events were all‐or‐none and spontaneous events not synchronized to the ST shock were ignored in this analysis). In no case did we detect evidence in the recruitment protocols for synaptic events that were of common ST origin in paired neighbouring neurons (i.e. synaptic events activated in both neurons with an identical threshold intensity consistent with activation of a single ST primary afferent diverging to two target neurons) (McDougall & Andresen, 2013). High‐intensity ST shocks failed to evoke responses in four NTS‐CeA neurons and two unlabelled neighbouring NTS cells. These numbers of ‘unconnected’ neurons were similar or lower than commonly encountered in previous NTS preparations and these cells were not studied further but are considered in the population comparisons. Given the thin slices studied, the reports of our synaptic studies probably underestimate the total numbers of NTS‐CeA neurons that exist in the intact system.

Figure 3. NTS‐CeA second‐order sensory neurons receive greater convergent primary afferent input compared to unlabelled neighbouring second‐order neurons.

NTS neurons were characterized by testing their responses to ST activation using a concentric bipolar electrode. A key aspect of this assay was the pattern of synaptic response recruitment acquired by graded shock intensities used to distinguishable inputs via EPSC amplitude, synaptic jitter and failure rate. A, traces from a labelled NTS‐CeA second‐order sensory neuron. Each panel displays 10 trials overlaid (shock artefacts blanked) at increasing ST shock intensities (ST arrows). Evoked EPSCs became increasingly complex as the ST shock intensities increased. Repeated shocks at 19 μA evoked relatively low amplitude EPSCs with consistent onsets and indicate the recruitment of a single input (open diamond). At 20 μA, an additional current summated with the previous current, yet its onset was also very consistent, indicating the recruitment of an additional input (open square). Finally, at 50 μA, an early onset EPSC was evoked and the currents summated further, indicating the recruitment of a third input (open triangle). Aa, each of the traces in (A) overlaid and colour coded to illustrate the EPSC dynamics with increasing shock intensities. The grey box is expanded to illustrate the differences in onset latencies. Ab, EPSC latencies are highly consistent (low synaptic jitter) and indicate simple monosynaptic pathways from the site of axon activation (ST) to the recorded NTS‐CeA projection neuron. Ac, EPSC amplitude in relation to shock intensity illustrates the step‐like increments in line with all or nothing activation of afferent inputs to NTS‐CeA projection neurons. Failures (no response after ST shock) did not occur once the threshold was met in this NTS‐CeA second‐order sensory neuron, which is typical of the high probability of release from primary ST afferents. B, traces from a paired unlabelled neighbouring neuron where each panel displays 10 trials overlaid (shock artefact blanked) at equivalent ST shock intensities (ST arrows). At 50 μA, shocks evoked low‐jitter EPSCs, designating input 1 (open diamond). Each 100 μA shock triggered EPSCs with the same characteristics. Thus, super threshold shocks of any intensity successively activated the same, single, synaptic input where the resultant EPSCs exhibited low jitter (Ba and Bb), unique threshold and zero synaptic failures (Bc). [Colour figure can be viewed at wileyonlinelibrary.com]

NTS‐CeA second‐order sensory neurons receive multiple primary afferent inputs

Increasing the ST stimulus intensity activated first one and then often multiple unique EPSCs in individual NTS‐CeA neurons (Fig. 3 A). The consistent waveform and minimal shock‐to‐shock variation in the latency (jitter = SD of latency) indicated that many evoked EPSCs were monosynaptic unitary responses to ST afferent activation (Doyle & Andresen, 2001) and thus such neurons were identified as second‐order sensory neurons. Recruitment profiles for NTS‐CeA second‐order sensory neurons most commonly showed multiple convergent ST‐evoked EPSCs with unique thresholds and distinct latencies and waveforms (Fig. 3 A and Aa). Single, high‐intensity shocks activated multiple ST axons and generated complex compound events that directly converged on single NTS‐CeA second‐order sensory neurons. Such compound events from strong ST shocks often obscured smaller events that were so clearly delineated with weaker shocks. The onset of the initial event in the compound EPSCs could be used to calculate latencies and these met the monosynaptic criterion (< 200 μs). The jitter values were unique across EPSC subcomponents and reflect variations in glutamate release from the underlying individual ST afferents (Fig. 3 Ab). Recruitment thresholds were defined across multiple trials at the same shock intensity, and these monosynaptic EPSCs rarely, if ever, failed (Fig. 3 Ac, right). Combined, these findings indicate primary afferent inputs converging on single NTS‐CeA second‐order sensory neurons. Unlabelled neurons generally received inputs that met monosynaptic jitter criteria, where most often even high‐intensity (> 500 μA) ST shocks recruited only a single, low jitter EPSCs (Fig. 3 B). Thus, recruitment profiles in such neurons neighbouring NTS‐CeA neurons were simple for these unfailing EPSCs (Fig. 3 Bc). For these unlabelled second‐order NTS neurons, EPSC amplitudes and onsets were very consistent and jitters were < 200 μs (Fig. 3 B). There were no differences between NTS‐CeA and neighbouring unlabelled neurons with respect to mean monosynaptic EPSC characteristics; latencies (4.4 ± 0.2 ms vs. 4.9 ± 0.4 ms); jitter (137 ± 6 μs vs. 127 ± 9 μs); amplitudes (n = 22 inputs, 107 ± 18 pA vs. n = 15 inputs, 104 ± 12 pA); and failures (2 ± 1% vs. 3 ± 2%) (P > 0.05, one‐way ANOVA). Thus, as expected, ST primary afferent transmitter release properties are similar across second‐order NTS neurons, irrespective of their target.

ST shocks also activate polysynaptic inputs to NTS‐CeA second‐order sensory neurons

In many NTS‐CeA second‐order sensory neurons, ST synaptic responses with highly variable latencies (jitter>200 μs) could also be recruited and produced much more complex synaptic patterns (Fig. 4 A). The high jitter of these synaptic events indicated that the input path to these neurons followed an indirect, therefore multineuronal or polysynaptic pathway from ST initiation. In some cases, these events were generated by intervening inhibitory interneurons activated by primary (ST) afferents and all had high jitter. In the case of ST‐evoked IPSCs, shifting the recording to a depolarized holding potential (V _H = –40 mV) produced outward synaptic currents that were difficult to discern at V _H = –60 mV (Fig. 4 A, left). Polysynaptic EPSCs could also be evoked in some neurons by ST shocks. However, there was no relationship between the absolute ST test shock intensity and the excitatory or inhibitory nature of the synaptic event (i.e. the lowest intensity shocks in some neurons activated polysynaptic IPSCs and greater intensity shocks added monosynaptic EPSCs) (Fig. 4). As expected for directly linked ST‐EPSCs, synaptic failures were generally near zero but polysynaptic events failed quite often, even across multiple ST‐evoked PSCs that were well separated in time (Fig. 4 D, right). Such mixed synaptic responses (mono‐ with polysynaptic EPSCs and/or IPSCs) were common in NTS‐CeA second‐order sensory neurons and indicated that such neurons received convergent information directly and indirectly from primary afferent activation. Note that, even in single second‐order neurons with highly complicated synaptic mixtures, the combination of graded ST intensity and multitrial testing reliably and reproducibly identified multiple distinct mono‐ and polysynaptic EPSC–IPSC combinations (Fig. 4). We found that mean polysynaptic EPSC characteristics were similar (P > 0.05, one‐way ANOVA) across second‐order neurons: NTS‐CeA (32 inputs across 20 neurons) and paired unlabelled neurons (12 inputs across eight neurons); latencies (6.5 ± 0.4 ms vs. 6.3 ± 0.6 ms); jitter (373 ± 36 μs vs. 499 ± 91 μs); amplitudes (n = 16 EPSCs, 59 ± 9 pA vs. n = 6 inputs, 60 ± 14 pA); and failures (21 ± 3% vs. 40 ± 8%). Overall, the characteristics of polysynaptic IPSCs in NTS‐CeA second‐order sensory neurons (six IPSCs across six neurons) were similar (P > 0.05, one‐way ANOVA) to paired unlabelled second‐order neurons (two IPSCs across two neurons) with similar latencies (8.5 ± 1.5 ms vs. 8.1 ± 0.2 ms); jitter (633 ± 172 μs vs. 603 ± 340 μs); amplitudes (n = 5 IPSCs, 36 ± 7 pA vs. n = 1 input, 37 pA); and failure rates (37 ± 12% vs. 49 ± 29%).

Figure 4. NTS‐CeA second‐order sensory neurons receive convergent ST‐derived polysynaptic input.

A, traces from a labelled NTS‐CeA projection neuron. Each panel displays 10 trials overlaid (shock artefacts blanked) at increasing ST shock intensities (ST arrows) with the neuron held at –40 mV and –60 mV to characterize evoked IPSCs and EPSCs, respectively. Repeated shocks at 16 μA evoked relatively long latency IPSCs with inconsistent onsets and failures, indicating the recruitment of a single polysynaptic pathway (open diamond). Shocks at 18 μA evoked an additional relatively long latency set of EPSCs also with inconsistent onsets and failures, indicating the recruitment of an additional excitatory polysynaptic pathway (open square). ST shocks ≥ 20 μA evoked shorter latency EPSCs with consistent onsets and no failures, indicating the recruitment of an additional three monosynaptic inputs (open triangle, open circle and closed hexagon). Note that compound EPSC dynamics did not change beyond 35 μA (see 100 μA), indicating that no additional afferent inputs were activated by ST shocks. although disynaptic pathways are indicated in the diagrams, the number of intervening interneurons per polysynaptic pathway is unknown. An animated version is provided in the supplementary video. B, traces in (A) overlaid, colour coded and base levels matched to illustrate evoked EPSC and IPSC dynamics. C, distribution of EPSC and IPSC latencies illustrates that long latency PSCs exhibited high jitters (>200 μs, open diamond and square), indicating complex polysynaptic pathways from the site of axon activation (ST) to the recorded NTS‐CeA projection neuron. By contrast, the relatively shorter latency EPSCs were more narrowly distributed and jitters < 200 μs (open triangle and circle and closed hexagon), indicating simple monosynaptic pathways from the site of axon activation (ST) to the same recorded NTS‐CeA projection neuron. D, EPSC and IPSC amplitude in relation to shock intensity illustrates the step‐like increments at distinct thresholds in line with all or nothing activation of ST axons to evoke responses; note input 1 (diamonds) not plotted beyond 18 μA for clarity. Evoked EPSC and IPSCs exhibiting high failure rates were indicative of polysynaptic pathways, whereas zero failure rates were typical in low jitter EPSCs. [Colour figure can be viewed at wileyonlinelibrary.com]

NTS‐CeA neurons are commonly higher‐order sensory neurons

Not all dye‐labelled NTS neurons responded to ST shocks with low jitter synaptic events (Fig. 5) but, instead, ST shocks activated multiple, high jitter responses consistent with the convergence of several ST activated, polysynaptic pathways onto these single neurons. The synaptic responses recorded in these NTS‐CeA higher‐order sensory neurons were often complex with multiple synaptic failures and variable amplitudes that took the form of EPSCs or IPSCs (Fig. 5). EPSCs were inward current synaptic events with fast kinetic waveforms (Fig. 5 A). However, we also observed IPSCs with slower kinetics and small amplitudes at V _H = –60 mV (Fig. 5 B) that became accentuated, outward currents by shifting V _H more positively to E _Cl. ST evoked IPSCs exhibited characteristically slow, long lasting kinetics, high jitter and frequent failures. In two neurons, high‐intensity ST shocks activated low jitter monosynaptic IPSCs consistent with a spreading stimulus that directly activated inhibitory interneurons (McDougall & Andresen, 2012) and these were excluded from group analysis. ST derived IPSCs were much less common than ST evoked mono‐ or polysynaptic EPSCs in NTS‐CeA neurons. However, increasing the ST shock intensity often recruited additional excitatory synaptic currents, which often obscured IPSCs, and so this approach probably underestimates both the number and characteristics of additional ST‐derived inhibitory inputs. Polysynaptic EPSC mean characteristics were similar (P > 0.05, one‐way ANOVA) between NTS‐CeA (35 inputs across 22 neurons) and paired unlabelled higher‐order neurons (five inputs across three neurons); latencies (7.2 ± 0.5 ms vs. 5.5 ± 0.7 ms); jitter (437 ± 41 μs vs. 493 ± 121 μs); amplitude (n = 27 inputs, 47 ± 6 pA vs. n = 3 inputs, 62 ± 27 pA); and failures (28 ± 4% vs. 33 ± 9%). Similarly, polysynaptic IPSC characteristics were also similar between NTS‐CeA (14 inputs across 10 neurons) and paired unlabelled higher‐order neurons (two inputs across two neurons); latencies (8.9 ± 0.5 ms vs. 6.0 ± 0.4 ms); jitter (669 ± 157 μs vs. 647 ± 88 μs); amplitude (no compound measurements, thus n = 8 inputs, 30 ± 5 pA vs. n = 2 inputs, 44 ± 7 pA); and failures (43 ± 7% vs. 34 ± 11%) (P > 0.05, one‐way ANOVA). Thus, higher‐order NTS neurons, regardless of whether they were projecting to CeA or not, received inputs from interneurons that had generally similar synaptic characteristics.

Figure 5. NTS‐CeA higher‐order sensory neurons receive convergent ST‐derived polysynaptic input.

In half‐labelled NTS‐CeA projection neurons, ST shocks activated only high jitter, polysynaptic inputs, which classifies them as higher‐order NTS‐CeA neurons. A and B, display responses to 10 ST shock trials overlaid (shock artefacts blanked). A, at V _H = –60 mV, ST shocks (ST arrows) activated EPSCs with high jitter (left) and increases in shock intensity recruited a second, earlier arriving high jitter EPSC (right). B, in the same neuron, shifting to V _H = –40 mV, 30 μA ST shocks activated outward IPSCs that did not change with higher ST shock intensities (100 μA). At V _H = –40 mV, inward EPSCs are also evoked as in (A). Although disynaptic pathways are indicated in the diagrams, the number of intervening interneurons per polysynaptic pathway is unknown. C, PSC latency distributions for all three recruited inputs were wide and exhibited high jitter, indicating complex polysynaptic pathways from the site of axon activation (ST) to the recorded NTS‐CeA higher‐order sensory neuron. D, PSC amplitudes increased in step‐like increments, reflective of all or nothing afferent pathway activation. Failure rates for high jitter EPSCs and IPSCs were ∼40% across the inputs. [Colour figure can be viewed at wileyonlinelibrary.com]

NTS‐CeA neurons are highly integrative

In aggregate, proportionally fewer NTS neurons that project to CeA were second order (n = 29 of 58) compared to the population of neighbouring unlabelled neurons in the same regions (Fig. 6 Aa) (P = 0.034, Z test). Almost 75% (n = 16 of 22) of unlabelled neighbouring neurons were second order, with most receiving single ST inputs (Fig. 6 Aa). The mix of synaptic responses to second‐order neurons was exclusively excitatory (mono‐ + polysynaptic) in most second‐order neurons, regardless of whether they were NTS‐CeA or unlabelled (Fig. 6 Ab). NTS‐CeA second‐order sensory neurons received a greater total number of ST activated synaptic inputs than unlabelled, near neighbour NTS neurons (Fig. 6 Ac) (Kruskal–Wallis H = 3.882, d.f. = 1, P = 0.049, one way ANOVA). Neurons lacking monosynaptic ST inputs (i.e. higher‐order neurons) accounted for a larger proportion of the NTS‐CeA population (n = 25 of 58) and received a significantly larger proportion of synaptic inputs activated by ST shocks than unlabelled higher‐order neurons (n = 4 of 22) (Fig. 6 B) (P = 0.019, Z test). Thus, whether second‐ or higher‐order in synaptic organization, NTS neurons projecting to the CeA received greater convergent and more complex ST‐activated synaptic inputs.

Figure 6. NTS‐CeA projection neurons receive greater convergent primary and polysynaptic afferent input as compared to paired unlabelled NTS neurons.

Aa, lower proportion of NTS‐CeA neurons were classified as second order (receiving at least one low jitter ST input) compared to paired unlabelled neurons (50% vs. 73%, respectively). Ab, within this cohort of second‐order NTS‐CeA neurons, a much greater proportion received additional polysynaptic input compared to paired unlabelled neurons. Ac, second‐order NTS‐CeA neurons received a greater number of convergent ST afferent inputs and ST‐derived polysynaptic inputs per neuron compared to paired unlabelled neurons. Thus, total ST‐derived convergence was also greater. B, a greater proportion of NTS‐CeA neurons were classed as higher order (receiving ST‐derived polysynaptic input only) compared to the paired unlabelled neuron population (43% vs. 18%, respectively). Within this specialized cohort NTS‐CeA neurons received greater convergent polysynaptic input per neuron compared to paired unlabelled neurons. Note that 7% of NTS‐CeA and 9% of paired unlabelled neurons were classified as ‘not connected’ and, combined with those in (Aa) and (B), make up the total respective populations (NTS‐CeA group, n = 58; paired unlabelled group, n = 22). [Colour figure can be viewed at wileyonlinelibrary.com]

Afferent information arrives sorted by TRPV1 phenotype at CeA

Although unmyelinated (C‐fibre) sensory axons constitute 75–90% of the mix of cranial visceral afferents entering NTS (Andresen et al. 2012), myelinated cranial afferents dominate most regulatory reflexes that activate at physiological thresholds in the range of normal conditions (Jones & Thoren, 1977). In the present study, we relied on TRPV1 expression as a marker of C‐fibre afferents and CAP exposure to determine the myelinated/unmyelinated phenotype of the evoked antecedent ST afferent, which is a procedure that is supported by numerous previous investigations (Doyle et al. 2002; Jin et al. 2004). In a subset of neurons, we recruited the maximum number of ST activated synaptic inputs (Fig. 7 A) and then applied CAP. CAP blocked all ST evoked transmission in neurons classified as having TRPV1+ ST inputs even at supramaximal shock intensities (Fig. 7 A). Neurons whose evoked transmission was unchanged by CAP were classified as belonging to TRPV1– pathways and CAP altered neither monosynaptic, nor polysynaptic ST evoked EPSCs (Fig. 7 B). Note that all synaptic responses to CAP were all‐or‐none and no partial responses were detected, including indirect, polysynaptic inputs to a given second‐order neuron (Fig. 7 C). If neurons had multiple ST derived inputs evident, then all ST evoked inputs were blocked by CAP in neurons with TRPV1+ primary afferents. Conversely, in TRPV1– cases, no ST evoked events were blocked during CAP, regardless of whether those inputs were monosynaptic or polysynaptic or glutamatergic or GABAergic (McDougall et al. 2008; McDougall & Andresen, 2012). Such results are only consistent with a pathway organization in which the full pathway through NTS was fully segregated by the initiating TRPV1 ST phenotype. In essence, this suggests a parallel pathway arrangement in which TRPV1+ ST inputs at single second‐order NTS neurons did not mix with TRPV1– inputs.

Figure 7. NTS‐CeA second‐order sensory neurons contacted by convergent monosynaptic ST afferents were either TRPV1+ or TRPV1– phenotype, never mixed.

A, repeated increasing ST shocks (ST arrows) evoked EPSCs with consistent onsets and zero failures, indicating the recruitment of two monosynaptic inputs (open diamond and square) to an unlabelled NTS neuron (each panel displays 10 trials overlaid, with shock artefacts blanked). Shocks repeated in the presence of CAP were blocked, indicating that converging afferents both expressed TRPV1. B, in a NTS‐CeA second‐order sensory neuron, repeated graded shocks (ST arrows) evoked EPSCs (10 trials overlaid, with shock artefacts blanked) with consistent onsets and zero failures, indicating the recruitment of two monosynaptic inputs (open diamond and square) and relatively late occurring EPSCs with highly variable onsets and high failure rates, indicating the recruitment of an additional polysynaptic input (open triangle, not plotted in recruitment graph). Shocks repeated in the presence of CAP did not block these EPSCs, indicating that all converging pathways (mono‐ as well as secondary polysynaptic) relied on activation of primary afferents that did not express TRPV1. C, NTS‐CeA second‐order sensory and unlabelled NTS neurons received only convergent TRPV1+ (red) or TRPV1– (blue) inputs (never mixed). Remarkably, even polysynaptic (hashed) ST initiated events conformed with this segregation by TRPV1 phenotypic classification. [Colour figure can be viewed at wileyonlinelibrary.com]

Afferent pathways to higher‐order NTS neurons remain sorted by TRPV1 ST origins

A high proportion of NTS‐CeA second‐order sensory neurons received no direct ST synaptic contacts, although high jitter responses indicated the arrival of exclusively higher‐order, ST afferent information via polysynaptic pathways (Fig. 8). Despite their indirect coupling to ST, these NTS‐CeA higher‐order sensory neurons received multiple distinct inputs activated by ST afferent shocks and their organizational position deep within NTS networks with multiple inputs would appear to make them probable candidates for receiving a mixed combination of TRPV1+ and TRPV1– ST afferent drive. In both labelled and unlabelled higher‐order neurons, the introduction of CAP blocked all ST‐related synaptic events in half the neurons, typing them as TRPV1+ connected (Fig. 8 A). The remainder were unresponsive and considered as initiated by TRPV1– ST afferent axons (Fig. 8 B). No partial responses to CAP were detected. Taken together, these results demonstrate that, even for higher‐order NTS neurons and despite the presence of intervening neurons, all ST‐evoked events originated from a single TRPV1 phenotypic class of ST afferents (Fig. 8 C). Taken together, our results suggest that ST afferents activate pathways within the NTS that are fully segregated by the initiating TRPV1 phenotype, with this separation extending within the network to higher‐order output neurons such as those targeting CeA. The overall image is one of parallel pathways of information sorted by the ST afferent TRPV1 phenotype. Thus, information arising from unmyelinated cranial visceral afferents is held separate throughout the NTS network from information arising from myelinated afferents.

Figure 8. Convergent ST‐derived polysynaptic input to NTS‐CeA higher‐order neurons are segregated by TRPV1 phenotype.

A, repeated graded shocks evoked EPSCs (10 trials overlaid, with shock artefacts blanked) with inconsistent onsets and failures indicating the recruitment of two polysynaptic inputs (open diamonds and squares) to this NTS‐CeA neuron. Responses to ST shocks repeated in the presence of CAP were blocked, indicating that each pathway relied on the initial activation of primary afferents that expressed TRPV1. B, repeated graded shocks (ST arrows) evoked EPSCs (10 trials overlaid, with shock artefacts blanked) with inconsistent onsets and failures, indicating the recruitment of three polysynaptic inputs (open diamonds, squares and triangles) to a higher‐order NTS‐CeA neuron. Responses to ST shocks were unaltered by CAP and thus were considered as TRPV1– pathways. C, higher‐order NTS neurons projecting to CeA were evenly divided by TRPV1 phenotype and, on a single unlabelled higher‐order neuron, received convergent input initiated by TRPV1+ expressing primary afferents. In all cases, converging input was segregated by TRPV1+ (red) or TRPV1– (blue) input and indicates fully segregated viscerosensory pathways within the NTS and onto the CeA. [Colour figure can be viewed at wileyonlinelibrary.com]

Discussion

Stress and emotion impact on the physiological state of the visceral organs and vice versa. Considerable anatomical evidence supports bidirectional neural communication between the NTS and a number of forebrain regions, including the amygdala (Geerling and Loewy, 2006 b), although little is understood about how ascending lines of visceral afferent communication are organized. The present study examined this issue by retrogradely labelling NTS neurons that project to the CeA and then determined the performance of synaptic connections activated by ST cranial visceral afferents to these NTS‐CeA output neurons. We report five major, new findings: (1) NTS neurons projecting to CeA were activated by ST afferent initiated pathways almost without exception; (2) significantly more (∼half) CeA projecting NTS neurons were higher order than neighbouring unlabelled neurons, which were mostly (∼75%) second‐order sensory neurons; (3) NTS‐CeA neurons displayed greater numbers of convergent ST‐related inputs than unlabelled cells; (4) ST afferent pathways within the NTS were organized into separate lines of exclusively TRPV1+ or TRPV1– viscerosensory pathways of synaptic transmission; and (5) all higher‐order NTS neurons, whether projecting to CeA or not, received only TRPV1+ or TRPV1– ST afferent pathways, despite the lack of direct ST contacts. This remarkable and absolute separation means that CeA receives a highly convergent set of craniovisceral sensory information along communication lines dedicated to one phenotypic afferent class, either TRPV1+ or TRPV1–. This TRPV1 based segregation was true regardless of whether ST information arrived in CeA after crossing a single synapse within the NTS or followed multineuron paths within the NTS which could include GABAergic interneurons. Taken together, these findings suggest for the first time that ST afferent information courses through NTS via pathways that do not mix across TRPV1 phenotypes and that resembled parallel pathways sorted by afferent phenotypic class arriving in CeA.

Visceral afferent organization within the NTS delivers highly ordered information to CeA

Our results indicate that visceral organ information arrives at CeA (NTS‐CeA neurons) in patterns established within the NTS. Dye injected into CeA was transported retrogradely to fill neurons that were found primarily within caudal portions of medial NTS ipsilateral to the injections in CeA. This highly focused distribution within the NTS corresponds to the location of the central terminal fields of a broad mix of visceral afferents, such as baroreceptors, oropharyngeal afferents, airway afferents and gastrointestinal afferents (Mendelowitz et al. 1992; Sekizawa et al. 2003; Rinaman & Schwartz, 2004). Previous anatomical tracing has indicated that the NTS projections terminate specifically in the medial CeA (Bienkowski & Rinaman, 2013). Our results suggest that the total number of CeA projection neurons is relatively restricted and the lack of dye‐filled cells when injections missed their CeA target suggests a relatively delimited relationship between the caudal NTS and the amygdala. These NTS‐CeA neurons received strongly convergent, ST‐related inputs and clearly diverged from the more common pattern of neighbouring second‐order NTS neurons, which tended to receive more limited and, not uncommonly, single ST afferent inputs (McDougall et al. 2009; Peters et al. 2011; McDougall & Andresen, 2013). Because unlabelled NTS second‐order neurons were assayed simultaneously in the same slices as NTS‐CeA neurons, any systematic ST damage from slicing probably does not account for these differences. Thus, the high ST convergence to NTS‐CeA neurons suggests that CeA receives mostly consolidated, multi‐afferent signals. This high convergence is a prominent organizational feature for co‐incident information processing established before reaching CeA.

The high convergence of afferent inputs at NTS‐CeA neurons included polysynaptic ST‐driven inputs. Polysynaptic inputs may represent a stage in the local networks within NTS with a more integrated form of sensory input. This strong tendency to receive multi‐afferent ST inputs was even more unexpected given the absolute segregation of myelinated and unmyelinated afferent messages at NTS‐CeA neurons. In the physical layout within the NTS, myelinated and unmyelinated ST axons branch to spread their terminal fields broadly crossing anatomical subregions (Kubin et al. 1991; Kubin et al. 2006) and covering similar areas of NTS with no discernible spatial regionalization. Despite close proximity, adjacent second‐order neurons received only one ST TRPV1 phenotype (McDougall & Andresen, 2013). TRPV1+ expression coincides with conduction velocities within the unmyelinated axon range and, correspondingly, TRPV1– afferents conducted within the Aδ range (Jin et al. 2004). The conformance of ST‐polysynaptic inputs at NTS‐CeA neurons to the TRPV1 based segregation observed in second‐order NTS neurons challenged our expectation that this relationship would break down deeper within the NTS local circuits to produce mixed A/C inputs at higher‐order neurons. The separation or organizational sorting into two streams by TRPV1 phenotype pathways appears to be generalized because it was found consistently in CeA specific projection neurons, as well as unspecified neurons in the same regions (second‐ and higher‐order unlabelled neurons). Polysynaptic EPSCs and IPSCs conformed with this segregation across two major neurochemical phenotypes of NTS neurons via glutamatergic and GABAergic interneurons, neither of which themselves express TRPV1 (Cavanaugh et al. 2011). This unexpected strict separation of pathway structure should provide the amygdala with cranial visceral information segregated by TRPV1 phenotype. From the perspective of the afferent sensory endings and their physiological activation characteristics (Andresen et al. 2012), this phenotype sorting of myelinated and unmyelinated afferents will effectively drive two distinct pathways: the TRPV1 lacking myelinated line providing normal ‘physiological’ status signals and the TRPV1+ or C‐fibre afferents signalling potentially more ‘alarming’ signals. For example, during ‘rest’, it is mostly A‐fibre baroreceptors that convey information to the CNS because they are activated at lower pressure (stretch) thresholds compared to C‐fibre afferents (Coleridge et al. 1973; Yao & Thoren, 1983). Similarly, this is the case for pulmonary afferents, where at ‘rest’ A‐fibres are active and at higher thresholds (greater stretch, chemical) C‐fibre afferents begin to be recruited (Coleridge & Coleridge, 1984). In the gastrointestinal system, it is the presence of food that stimulates release of various mediators that activate C‐fibre sensory neurons but, otherwise, they are mostly silent (Page et al. 2002). Thus, the relative roles and activation characteristics of these two classes of afferent phenotypes are systematically different with respect to the physiological contexts and pathological states of visceral organs (Andresen et al. 2012). A limitation of our approach in the present study is that we have not identified the sensory modality (e.g. mechano‐ or chemoreceptor) or the visceral tissue origin (e.g. cardiac or airway) (Andresen et al. 2007; Andresen & Paton, 2011; Andresen et al. 2012). We do not know whether cardiac and respiratory information is combined with gastrointestinal input through this convergence at single NTS neurons, although our studies show that they are segregated by TRPV1 phenotype. In the intact system, it may be that the TRPV1– line to the CeA updates normal status, whereas the TRPV1+ line contacts CeA neurons involved in physiological stress alarming (Dayas et al. 2001). In addition, we are only considering the most direct NTS‐CeA pathway in the present study, although, clearly, visceral information may reach the amygdala via other intervening nuclei such as the parabrachial nucleus (Sah et al. 2003). Interestingly, although our methods focused on CeA, a subset of NTS neurons projecting to the amygdala also sends co‐lateral axons to the periaqueductal grey (27%) (Reyes & Van Bockstaele, 2006) and to the paraventricular nucleus of the hypothalamus (10%) (Petrov et al. 1993). Thus, visceral information from NTS to these areas is probably also similarly organized with respect to TRPV1 segregated pathways.

CeA receives highly processed visceral afferent input

Half of the NTS‐CeA neurons were higher‐order and thus received ST driven local NTS network input. Our electrophysiological findings corroborate, with direct synaptic function, the implications obtained via anterograde viral tracer studies in which extended infection times suggested multisynaptic processing pathways within the medial NTS before delayed expression in the amygdala after peripheral injections (Rinaman & Schwartz, 2004; McGovern et al. 2012; McGovern et al. 2014). Other supramedullary targets of NTS projection neurons showed highly integrative, higher‐order NTS neurons projecting to, for example, the paraventricular nucleus of the hypothalamus (Bailey et al. 2006). By contrast, retrograde tracer injected into targets within the brainstem such as the caudal ventrolateral medulla yielded only second‐order NTS neurons (Bailey et al. 2006). Combined, these studies illustrate how initiating visceral afferents give rise to pathway organization and this may be associated with specific efferent targets. However, little is known about what such differences in afferent specificity might convey to ultimate circuit functions (e.g. single organ vs. multiple organ; high vs. low convergence; or myelinated vs. unmyelinated).

NTS‐CeA pathway function

The NTS‐CeA pathway is well known to impact on stress‐related behaviours with some hints about the expected general characteristics. Earlier immunohistochemical studies established that more than 50% of NTS‐CeA neurons express tyrosine hydroxylase as members of the A2/C2 catecholaminergic group (Riche et al. 1990; Petrov et al. 1993; Reyes & Van Bockstaele, 2006). Noradrenergic NTS neurons projecting to the hypothalamus promote synaptic plasticity during stress (Inoue et al. 2013) and activation of presynaptic α2 receptors decreases the probability of glutamate release from excitatory afferents in the CeA (Delaney et al. 2007). During a stress response, the release of norepinephrine within the amygdala enhances memory processes (Williams et al. 1998; Clayton & Williams, 2000). Subdiaphragmatic vagotomy disrupts many cranial visceral afferents, although extra care is needed to maintain the health of such animals to avoid misleading results (Romanovsky et al. 1997). However, when carefully controlled, this surgery in rats indicates important subdiaphragmatic vagal afferent contributions to conditioned fear responses and anxiety‐like behaviour (Klarer et al. 2014). The NTS circuits participate in satiety, nausea and energy balance (Mayer, 2011; Babic & Browning, 2014) and shapes behavioural responses in varying homeostatic states (Dallman, 2010), where each of these interactions may depend on NTS‐CeA neurons. Combined, multisynaptic tracing and functional studies suggest that a wide range of sensory modalities reach the CeA. Our tracer showed that NTS‐CeA neurons were located in regions known to receive afferent inputs from multiple visceral tissues (Blessing, 1997). This fits with the idea of integrative, convergent signal processing within the NTS and thus the CeA may receive a holistic view of the visceral function, ranging from baroreceptors (Garfinkel et al. 2014) to airway (McGovern et al. 2015) and gastrointestinal vagal afferents (Viltart et al. 2006). Because NTS signals terminate within the medial CeA (Bienkowski & Rinaman, 2013), this information may influence or gate output activity that executes the behavioural and autonomic changes associated with fear‐conditioned responses (Parsons & Ressler, 2013). Indeed, even at rest, the TRPV1– (Aδ‐fibre) line to the CeA may be influential. In support of this idea, Garfinkel et al. (2014) reported that emotional processing and fear outcomes differed based on heartbeat, such that specific fear ratings and amygdala responses (imaging) were greater during systole than during diastole.

Our most unexpected finding is that myelinated visceral afferent information is strictly sorted and isolated from that of C‐fibres. Our results link information processing of visceral information to the CeA and its behavioural associations. The direct NTS‐CeA relationship suggests that the amygdala has access to TRPV1‐sorted information essential for assessing homeostatic conditions. Information available on normal resting conditions at the visceral organs conveys an eu‐physiological signal line via the myelinated afferent driven pathways. However, the C‐fibres might signal an alarming state of vulnerability or disrupted homeostasis along a dys‐physiological signal line. We speculate that these interoceptive signals affect behavioural functions via activity received along these parallel pathways from the NTS.

Additional information

Competing interests

The authors declare that they have no competing interests.

Author contributions

SJM and MCA conceived and designed the work. SJM and HG were responsible for the acquisition, analysis or interpretation of data. All authors drafted the paper or revised it critically for important intellectual content. All authors have approved the final version of the manuscript and agree to be accountable for all aspects of the work. All persons designated as authors qualify for authorship, and all those who qualify for authorship are listed.

Funding

Supported by grants from the National Institutes of Health; HL‐41119 (MCA), HL‐105703 (MCA) and Oregon Health & Science University Presidential Bridge Funding (MCA). The National Health and Medical Research Council of Australia CJ Martin Fellowship #400405 (SJM) and the Victorian Government's Operational Infrastructure Support Program (Florey).

Translational perspective.

Changes in visceral function including heart rate, breathing and digestion are concordant with emotions. We speculate that the myelinated afferent line assures ‘all's well’ and the C‐fibre line signals ‘alarming’ states of visceral function. The very different physiological activation profiles mean that these two parallel visceral afferent pathways encode interoceptive signals to the amygdala that may provide for homeostatic valence assessments to impact on behaviours. Equally, this probably applies with external activation of the vagal afferents via vagal nerve stimulation (VNS). A‐fibres exhibit electrical properties such that this subgroup is activated by VNS at the lower intensities. The upper ceiling for shock intensity in VNS devices is limited by the coactivation of vagal efferents. Our data would indicate such VNS devices probably modulate those amygdala neurons receiving A‐fibre input via the NTS.

Supporting information

Disclaimer: Supporting information has been peer‐reviewed but not copyedited.

Supplementary Video file. Use QuickTime or VLC player to view as a movie. Drag the slider to the desired time points, advance one or multiple frames using the arrow key ( ← or → ) for full inspection.

Full recruitment protocols interrogated multiple convergent synaptic pathways to single recorded NTS

neurons. Finely graded shocks to the ST revealed key synaptic characteristics of the subcomponents of complex, compound synaptic volleys. This file shows results of a single NTS neurons identified by retrogradely transported dye from the central nucleus of the amygdala. The synaptic event latency, amplitude, shape and threshold intensity quantitatively defined the ST paths to the recorded NTS neuron. Iterative activation of events reveals consistent dynamics of the underlying paths to the recorded neuron. The movie presents raw traces and event analysis as it was recorded.