Cingulate cortex stimulation drives distinct pupillary responses in rat via recruitment of noradrenergic neurons in the locus coeruleus

Abstract

The organization of the cingulate cortex has been the subject of intensive studies, concluding to its central role in motor control, cognition, and arousal. One of the key anatomical pathways through which the cingulate cortex influences behavior is its efferent connection to the locus coeruleus (LC). This brainstem region is responsible for noradrenaline (NA) release and is critical for various cognitive and behavioral functions. However, the specific impact of cingulate subregions on the LC-NA system remains unexplored. This study investigated how the different cingulate cortex areas affect LC-NA activity by measuring pupil-evoked responses (PERs) as an index of LC-NA activity. Using intra-cortical stimulation across the eight cingulate areas in rats, we found that anterior cingulate cortex and midcingulate cortex subregions evoked rare autonomic responses but significant pupil dilations whose amplitude increased along the caudo-rostral and dorso-ventral axes. By using the DSP-4, a neurotoxin-selective ablation of the LC-NA system, we suppressed PER and confirmed the role of LC-NA activity in this response. The differential influence of cortical areas on the PER demonstrates that each subregion of the rat cingulate cortex has the potential to differentially activate the LC-NA system, suggesting a clear parcellation of the rodent cingulate cortex, likely corresponding to functional specialization.

Introduction

The anterior division of the cingulate cortex (CC) is a complex brain section occupying a large part of the mesial surface of the frontal cortex and composed of the anterior (ACC) and midcingulate (MCC) cortices (Vogt and Paxinos 2014). Functionally, studies in rodents, monkeys, and humans have proven that this cortical region occupies a central role in behavioral adaptation (Michelet et al. 2009, 2016; Sheth et al. 2012; Brockett et al. 2020). It integrates information originating from brain regions responsible for emotional, cognitive, and motor control and sends back projections to other cortical as well as subcortical regions (Dalley et al. 2004; Vogt 2016; van Heukelum et al. 2020). Among the cerebral targets of the CC, the locus coeruleus (LC) plays a role in adaptive behavior by releasing noradrenaline (Aston-Jones and Waterhouse 2016). This release supports increased flexibility (Bouret and Sara 2004) or change in arousal states (Berridge 2008; Samuels and Szabadi 2008a, 2008b). Interestingly, LC-NA activity has been closely correlated with changes in pupil size (Rajkowski et al. 1993; Aston-Jones and Cohen 2005a; but see also Joshi et al. 2016) which has also been considered for a long time as a measure of the arousal level (Berridge 2008). Recent studies based on optogenetic strategies in rodents (Breton-Provencher and Sur 2019) confirmed this result, thereby qualifying pupil diameter measures as a proxy providing access to the LC-NA neuromodulatory system in several species (Murphy et al. 2014; Joshi et al. 2016; Reimer et al. 2016).

It has been proposed that prefrontal cortical inputs, clearly evidenced in rodents (Jodo et al. 1998) or monkeys (Arnsten and Goldman-Rakic 1984), may drive and regulate LC-NA activity (Jodo and Aston-Jones 1997; Jodo et al. 1998; Breton-Provencher and Sur 2019) and are responsible for initiating a switch between an LC phasic mode favoring the exploitation of stable sources of reward and a tonic mode favoring the exploration of alternative behaviors (Aston-Jones and Cohen 2005b).

However, a global understanding of cortical influences on LC activity is still difficult since contradictory evidence from electrophysiological, lesion, or anatomical studies persists. For example, anterograde labeling demonstrated a very low density of area 25 (the former infralimbic or IL) inputs (Takagishi and Chiba 1991; Vertes 2004), a low (Sesack et al. 1989) or even an absence (Vertes 2004) of area 32 (the former prelimbic or PL) inputs to the LC. Functionally, electrical stimulation of these regions provided a clear excitatory influence on LC activity (Jodo et al. 1998), while chemical inactivation of the same or adjacent regions is also accompanied by an increase in LC unit activity (Sara and Hervé-Minvielle 1995). A discrepancy in anatomical definition or delineation of prefrontal cortex subregions or areas (van Heukelum et al. 2020) could, in part, explain such differences and also explain why prior knowledge about afferent and efferent projections of the LC was considered inaccurate (Poe et al. 2020).

To avoid any anatomo-functional confound, we proposed to explore systematically the pupil response as a proxy of the LC-NA system, using a recent map of cingulate regions and areas based on an ACC/MCC nomenclature, allowing a direct comparison with human cortical subdivisions (Vogt and Paxinos 2014; Vogt 2015). Our hypothesis is that the different parts of the CC can be functionally differentiated on the basis of their influence on the LC-NA system. We stimulated in anesthetized rats each of the 8 cortical areas of the ACC/MCC while recording the pupil diameter (Murphy et al. 2014; Joshi et al. 2016; Reimer et al. 2016). We also recorded other autonomic responses such as heart rate and respiratory rate in order to ensure the specificity of the pupil evoked-response (PER) among autonomous responses. Using a combination of neurovegetative component measures, electrophysiological, and pharmacological approaches, our results clearly indicate a caudo-rostral organization of the ACC/MCC inputs within the LC, with anterior regions exhibiting the larger effect on LC-NA system.

Materials and methods

Animals

A total of 21 Long–Evans male rats (Janvier Lab, Paris, France) aged 8 to 10 weeks were used in this study (n = 9 for intracortical microstimulation, n = 9 for pharmacological experiment, n = 3 for extracellular recordings). Animals were housed under standard conditions, with food and water available ad libitum. All the procedures were performed following the European Communities Council Directive (2010/63/UE) and approved by the local ethical committee (Ethics committee of the University of Bordeaux and the minister of higher education of research and innovation; approval number 21134).

Surgery

Stereotaxic surgeries for micro-stimulations were performed under 1.0% to 2.0% isoflurane anesthesia (in 50% air/50% O2; 1 L/min). After animal placement in the stereotaxic frame, flat-skull position was verified, and five holes were made above the MCC and ACC subregions at the following AP and ML (in mm, relative to bregma) coordinates in order to stimulate the following areas: 24a (AP = 1.8, ML = 0.4, DV = −2.6); 24b (AP = 1.8, ML = 0.4, DV = −1.4); 24a′ (AP = −1, ML = 0.5, DV = −2.2); 24b′ (AP = −1, ML = 0.5, DV = −1.3); 25 (AP = 2.5, ML = 0.4, DV = −4.4); 32d (AP = 3.6, ML = 0.4, DV = −1.3) and 33 (AP = 0, ML = 0.4, DV = −2.8). These coordinates, including the DV measure, correspond to the approximate central position of each area, according to the rat stereotactic atlas (Vogt and Paxinos 2014). A heating pad (RWD ThermoStar Homeothermic Monitoring System) was used to monitor and maintain their body temperature at 37 °C during the whole experiment.

Stimulation protocols

Different stimulation protocols were used to evoke a pupillary response. Light stimulations of the left eye were performed with a customized pen light (Comed, ref: 24671 20) at 2.59 lumen for 2.5 s to test the pupil light reflex (Grozdanic et al. 2002). Nociceptive footshock (FS) stimulations known to recruit the LC (Chiang and Aston-Jones 1993; Passerin et al. 2000) were performed through subcutaneous needles inserted in the right paw according to two distinct protocols: The “FS protocol 1” (80 pulses delivered at 20 Hz; 5 mA, 1-ms duration pulses) used to induce a strong activation of the LC neurons. The protocol “FS protocol 2” (20 pulses delivered at 0.1 Hz; 5 mA, 1 ms duration pulses) was used to identify LC neurons with a frequency sufficiently low to reveal a specific response to single pulse stimulation. Intracortical micro-stimulations (ICMSs) were performed using a concentric bipolar electrode (PHYMEP CBCSG50), according to the following five protocols: The “ICMS protocol 1” (3 trains of 10 pulses delivered at 20 Hz, every 0.5 s; 1 to 1.5 mA, 5-ms-duration pulses) was used to assess the potential effect of cortical regions on pupil response. The “ICMS protocol 2” (100 pulses delivered at 0.5 Hz; 1 mA, 1-ms-duration pulses) was used to explore the potential effect of cortical stimulation on LC neurons. ICMS Protocol 1 uses trains of pulses to maximize its impact on the LC to induce NA release and observe the PER. On the other hand, ICMS Protocol 2 evaluates the functional connectivity between the cortex and LC by delivering isolated, low-frequency pulses that minimize cumulative effects and allow us to construct a PSTH based on numerous stimulation events. The “ICMS protocol 3” (15 trains of 2 pulses delivered at 50 Hz, every 10 s; 1 mA, 50 Hz, 1-ms-duration pulses) and “ICMS protocol 4” protocol (15 trains of 2 pulses delivered 100 Hz, every 10 s; 1 mA, 1-ms duration pulses) was used to control for possible antidromic recruitment of the LC neurons. High-frequency stimulation (ICMS protocol 4) stresses the system to reveal rapid conduction and test refractory limits, while lower-frequency stimulation (ICMS protocol 3) allows more complete recovery from any refractory period. If both protocols yield similar antidromic responses, these complementary approaches should increase confidence in the results. At the end of the experiment, electrolytic lesions were made by applying an anodal direct current (1 pulse, 100 μA, 30 s) to mark the last stimulated area. After the electrolytic lesion, the animals were deeply anesthetized in the induction box with isoflurane at 5%, 3.5 ml/min for 5 min and decapitated, and their brains were removed rapidly. The brains were frozen and stored at −80 °C until they were cut into slices of 40 μm (microtome LEICA CM3000) and stained with the Nissl method to confirm the stereotaxic coordinates of the stimulated areas based on a rat brain atlas (Paxinos and Watson 2013).

Pupillometry

Custom software developed in our lab coupled with an infrared camera (Basler, acA1920-150uc, 100 frames/s with a Fujifilm HF25SA-1 lens) was used for measuring pupillary size and control for eye position along the horizontal and vertical axes. All measurements were performed in a dimmed light room. A bolus of urethane (20%, 0.195 ml/100 g) was injected intraperitoneally 5 min before the stimulation experiment to prevent eye movements induced by isoflurane (Nair et al. 2011) that would compromise pupil diameter monitoring. Each trial lasted at least 90 s with a stimulation performed after a 30-s baseline period. At least five stimulations were performed in each region of interest for ICMS protocol 1, with an intertrial interval ≥60 s until the pupil diameter recovered its basal measure before the next trial. All the data were processed using Matlab. The raw pupil signal was smoothed and artifacts due to miss tracks were interpolated. Trials were kept only if no artifacts lasting for more than 1 s were found during the period of analysis (5 s before and 30 s after stimulation). The dynamics of the response were studied through several metrics including the latency of the beginning/extremum/half-recovery of the response after stimulation. For each trial, the pupil diameter was normalized by subtracting the mean pupil diameter recorded 5 s before stimulation. All pupillary measures are presented in arbitrary units (AUs), relative to baseline pupillary measures.

Heart rate variability and respiratory rate frequency analysis

Subdermal 12 mm needles (No MF3.OE.1F35.12, Comepa, Saint-Denis, France) were implanted in the abdomen of the animals to record the electrocardiogram. Signals were sampled at 1 kHz and digitalized through an interface analog/digital (PowerLab 35 Series). The heart rate variability was determined with the Poincaré plot technique (Hoshi et al. 2013). The respiratory rate was inferred from the abdominal artifact movements visible in the envelope of the electrocardiogram (see Fig. 1C) due to the abdominal subcutaneous placement of electrodes.

Electrophysiology

Electrocorticogram

To control the brain state under anesthesia, we recorded an electrocorticogram (ECoG) for each animal from 1-mm-diameter steel skull screws in contact with the right frontal cortex (approximately 2 mm anterior and 2 mm lateral to bregma) and the ipsilateral cerebellar hemisphere (reference electrode; approximately 1 mm posterior and 2 mm lateral to lambda). A recording sampling rate of 1 kHz was used, and signals were digitalized through an interface analog/digital (PowerLab 35 Series). Each stimulation period (designated here as a “trial”) started only if the ECoG presented burst suppression patterns (see Fig. 1B), indicating deep anesthesia (Kenny et al. 2014). In order to ensure that analyses were performed under anesthetic stability, post hoc analyses of the ECoG were performed and we excluded trials that had more than 30% of bursts. The electrocorticogram was pretreated with a detrend filter, a notch filter, and a bandpass filter set between 0.6 and 50 Hz. Finally, a wavelet denoise filter was used before the binarization of the signal (Kenny et al. 2014). The signal above a threshold (burst) was considered 0 and the signal below the threshold (suppression) was considered 1. The threshold was set to the mean ECoG value +9 SD calculated based on the longest suppression pattern.

Event-triggered averaging

To explore the relation between cortical state and pupil dilation, we performed an event-triggered averaging of pupil diameter segments (n = 823) recorded 3 s before to 3 s after a burst onset in the ECoG.

Extracellular recordings and spike sorting

For the extracellular recordings of LC neurons, a 32 channels silicone neural probe (H6b from Cambridge NeuroTech) was lowered into the LC (LC: AP = −9.8 mm, L = −1.4 mm, P = 5.7 mm relative to bregma). Electrophysiological signals were recorded and sampled at 30 KHz using the OpenEphys system (Siegle et al. 2017). Spike sorting was made offline using the SpikeInterface software (Yger et al. 2018) with MountainSort 5 (Chung et al. 2017). Once the signal of a single LC neuron was isolated, we tested all the different stimulation protocols in order to compare a single-unit activity between all conditions for the same cell. Raster display is represented and peri-stimulus time histogram (PSTH) analyses were performed on successive 5-ms time bins. For population representation, PSTHs were computed for each neuron and then averaged and smoothed using a moving average filter (Matlab smooth function, with a 25 ms span). To confirm the position of the extracellular electrode within the LC after electrophysiological recordings, we used a fluorescent dye (Invitrogen Colorant CM-DiI CellTracker) directly applied on the silicone probe prior to insertion into the brain.

ICMS and NA dosage after administration of neurotoxin DSP4

To evaluate the role of NA in the PER, injections of the neurotoxin DSP-4 (N-[2-chloroethyl]-N-ethyl-2-bromobenzylamine hydrochloride, 50 mg/kg), known to destroy the synaptic terminals of LC neurons (Ross and Stenfors 2015), were made into 5 rats and compared to 4 rats injected with saline (0.9%). DSP-4 was dissolved in saline just before the single injection. Three days after, animals were prepared according to the above-described procedure. Only one craniotomy was performed in order to stimulate the area 25, chosen based on preliminary analyses indicating that the PER is more important in this region. During the experiment, 5 light stimulations, 5 FS stimulations, and 10 ICMS (protocol 1) in area 25 were made. After these stimulation protocols, a last ICMS protocol consisting of 30 pulses at 20 Hz every 10 s for 30 min was used to promote NA release in LC terminal fields before measuring the concentration of NA in a subset of LC efferent areas (amygdala, prefrontal cortex, hippocampus, and thalamus). The tissue level of NA was quantified by a reversed-phase high-performance liquid chromatography combined with electrochemical detection (HPLC/EC) as described below (but see also De Deurwaerdère et al. 1995 and supplementary text). Results were expressed in percentage of NA reduction in DSP-4 animals compared to saline.

Statistical analysis

All data were examined for normality and homogeneity of variance using Shapiro–Wilk and one-way analysis of variance (ANOVA) tests, respectively. Because either homogeneity of variance or normality tests failed, nonparametric statistical tests were used. Wilcoxon signed rank test (z as a result for the test and r for size effect) has been used to replace paired t-test, Wilcoxon rank sum test (z as a result for the test and r for size effect) to replace 2-sample t-test and Kruskal–Wallis (H as the result of the test and η² for size effect) to replace ANOVA.

Results

ICMS 1 protocol effect on pupil diameter

Effects of cortical stimulation on the PER are shown in Fig. 2. ICMS protocol 1 was able to induce PER in all the eight areas (24b′ n = 32, 24b n = 44, 24a′ n = 23, 24a n = 41, 33 n = 10, 32d n = 33, 32v n = 39, 25 n = 28) of the cingulate cortex (Fig. 2). In all areas the pupil increased significantly compared to baseline (all P < 0.01 and all r > 0.58). The maximum pupil diameter obtained after stimulation is expressed in arbitrary units: 24b′ = 0.11 ± 0.013, 24b = 0.14 ± 0.012, 24a′ = 0.15 ± 0.011, 24a = 0.23 ± 0.031, 33 = 0.24 ± 0.021, 32d = 0.34 ± 0.023, 32v = 0.44 ± 0.042, 25 = 0.60 ± 0.049. A Kruskal–Wallis test was performed to compare the PER between the stimulated areas. We obtained a Chi² = 136.1 and a P = 3.33e⁻²⁶, allowing us to perform a Bonferroni-corrected post hoc test with a P < 0.05, which detected a significant difference in PER between each of the three most anterior areas (25/32D/32 V) and the other more caudal areas tested (Fig. 2B). The size effect η² = 0.53 indicates a large effect of the stimulated area on the PER. Superimposing these results on a map of the rat cingulate cortex, highlighting the eight studied areas (Fig. 2C), reveals an increased response amplitude along the caudo-rostral and dorso-ventral axes. Critically, direct stimulation of the LC elicits a greater response than that of the most responsive (area 25) cortical area (Fig. 2D).

Dynamics of the PER

To describe the dynamic of the ICMS 1 protocol on the PER we used three different parameters consisting of the onset latency, maximum amplitude latency, and half-recovery of the response (Fig. 3A). The Kruskal–Wallis test coupled with a Bonferroni corrected post hoc when necessary were used to compare all parameters. The onset latency ranged between 0.28 and 0.61 s. The statistical analysis resulted in Chi² = 28 and a P = 2e⁻⁴ highlighting a significantly slower onset latency of the area 24b′ compared to the areas 24a and 32d with a moderate size effect of η² = 0.088. The latency of the maximum response was between 4.30 and 6.28 s with a Chi² = 59 and a P = 2.68e⁻¹⁰. Significant higher latencies were found for areas 25/32v in comparison with the others (η² = 0.21). Critically, these differences cannot be integrally explained by the extra time needed to reach a higher PER amplitude because the rising slope of the signal was higher for areas 32d, 32v, and 25 (but see Fig. 3C). Finally, latencies of the half recovery varied between 9.32 and 14.62 s with a Chi² = 28 and a P = 2e⁻⁴ emphasizing the significant difference between areas 25/32v and 24b with a moderate size effect η² = 0.086. Here again, the slope of the decay is constant, except for areas 33 and 32v, confirming the prolonged influence of NA on pupil dilation after stimulation of these areas (Fig. 3C).

The LC NA role in the PER

To test that the PERs were mediated by the recruitment of the LC-NA system, we used DSP-4, a neurotoxin preferentially destroying terminals of the LC-NA neurons (Ross and Stenfors 2015). Two groups of animals have been compared: one injected with saline and one injected with DSP-4. In DSP-4 injected animals, a significant decrease in tissue NA compared to saline has been found in the amygdala, hippocampus, and thalamus (areas known to have broad LC input (Lindvall et al. 1974)) with a P < 0.02 and r > 0.6 (Fig. 4A), and a nearly significant decrease has been found in the prefrontal cortex (P = 0.066). As expected, we did not find such a decrease in the hypothalamus (P = 0.71), used as a control site because it receives NA afferents from different NA cell groups (Cunningham and Sawchenko 1988; Aston 2004). Regarding the action of the different protocols of stimulation in the saline-injected rats, we have replicated the pupillary dilation obtained with ICMS protocol 1 in the noninjected animals. With the FS protocol 1, known to induce an enhanced activity of the LC neurons (Chiang and Aston-Jones 1993; Passerin et al. 2000), we obtained pupil dilation. In DPS-4 treated rats, however, we have shown a significant decrease in the dilation response in both protocols ICMS 1 and FS 1 (respectively, P = 1e⁻¹⁵ and 8e⁻⁶, i = 0.86 and 0.71). Finally, we did not show any significant impairment on the light reflex (P = 0.52, see Fig. 4B), the well-known pupil constriction found using the light stimulation protocol (Grozdanic et al. 2002) in control animals. These results showed that DSP-4 has a specific impact on LC terminals and almost entirely abolishes the PERs, underlying the critical role of LC recruitments after CC stimulation on the PER.

Histological analyses confirmed with systematic comparison to the reference atlas (Paxinos and Watson 2013) the location of electrodes within the LC. A total of 54 neurons (n = 3 rats) were recorded, and spike sorting yielded 5 LC putative neurons, identified based on their electrophysiological properties. Only 3 cells presented a particular pattern, enhancing their activity after both the FS protocol and the ICMS protocol 2 (Fig. 5). No significant change in the firing rate was found in the other protocol.

Autonomic responses

We also recorded the respiratory and heart rates as well as the change in the SD1/SD2 ratio pre- and poststimulation of the different areas in order to explore their potential influence on autonomic responses. As shown in Table 1, the only parameter that was significantly impacted by the ICMS 1 was the change in the SD1/SD2 ratio poststimulation of the area 25 (pre: 0.82 ± 0.05, post: 0.76 ± 0.05, P = 0.017, r = 0.45, Wilcoxon sign rank test with Bonferroni-adjusted P-value). This decrease suggests an increase in activity of the sympathetic nervous system (Hoshi et al. 2013).

ECoG mean burst pattern proportion

As expected, the continuous isoflurane anesthesia (mean 2.5%) induced in the ECoG a pattern of burst suppression. For the ICMS and FS protocol trials, we characterized this pattern by calculating the mean proportion of bursts, which were 11.67 ± 0.27%, n = 337 and 8.16 ± 1,88%, n = 39, respectively.

Interestingly, a significant difference has been found when comparing the mean pupil diameter during the 3 s before and after bursts onsets (mean pupillary diameter preburst onset = −4.3e⁻³ ± 0.73e⁻³, mean postburst onset = −0.95e⁻³ ± 1.73e⁻³, P = 5.1e⁻¹², r = 0.17), highlighting a link between the brain state and the pupil diameter (Fig. 6).

Discussion

It has been proposed that prefrontal cortical regions and more particularly the CC regulate LC activity (Aston-Jones and Cohen 2005b). Our study is the first that systematically evaluates a functional response dependent on the LC-NA neurons, the PER, after stimulation of each area of the ACC/MCC regions.

We found that, in a way comparable to direct stimulation of the LC (Fig. 2D), electrical stimulations of the eight areas (Fig. 2A–C) evoked systematically a pupil dilatation. Stimulation of the different areas considered in this study evoked differential pupil dilation, primarily in terms of amplitude, whereas stimulations performed along the same electrode tracks but outside the LC, or the studied cortical sites, failed to elicit any PER. The maximal PER was observed after the stimulation of areas 25 and 32, while a significantly lower PER is observed within area 24.

As a whole, we were able to describe a marked gradient amplitude of the PER increasing along the caudo-rostral and dorso-ventral axes, with greater responses in the ACC compared to the MCC (Fig. 2C).

Our results thus confirm the anatomical differentiation between ACC and MCC regions (Paxinos and Watson 2013) and confirm the validity of the PER approach to map the different circuits of the cingulo-coeruleus system. This result is compatible with the separate functional roles of ACC and MCC, the latter being involved in phasic LC activation, while more rostral regions are responsible for tonic LC activation. It is then likely that efferent messages from different cingulate areas, involving, for example, differential input strength or different subregions will promote the shift from phasic to tonic activation of the LC, which is involved in many behavioral adaptations (Aston-Jones and Cohen 2005a). This result is also in accordance with a functional and anatomical dorso-ventral subdivision of the rat mPFC (Heidbreder and Groenewegen 2003; Gabbott et al. 2005). For example, we found a significant difference in the dynamics of the PER between areas 32v and 32d (see Fig. 3B), thereby functionally confirming the pertinence of the subdivision of area 32 based on anatomical studies.

We failed to find a significant modulation of other autonomic responses except a decrease of the SD1/SD2 ratio, associated with a change in the sympathetic tone (Hoshi et al. 2013) when we stimulated area 25. This area has major projections for autonomic centers in comparison with other ACC or MCC subregions (Vogt 2015). Interestingly, other studies reported an absence of blood pressure or other autonomic response modulation after electrical stimulation of the mPFC (Al Maskati and Zbrożyna 1989; Jodo et al. 1998). Chemical stimulation by contrast induced prolonged systolic blood pressure changes (Bacon and Smith, 1993) indicating that chemical stimulation induces a more prominent effect. While we found a clear relationship between ECoG bursts and pupil dilation (Fig. 6C and D), we also failed to find a clear effect of our stimulation on cortical desynchronization. Indeed, based on previous findings demonstrating changes in the proportion of burst suppression following direct NA release (Berridge and Foote 1991; Vazey and Aston-Jones 2014), we could have anticipated a similar change after stimulation of the CC. These results are probably due to isoflurane anesthesia but may also suggest that our stimulation intensity is moderate and that prolonged or combined inputs on LC-NA neurons are needed to fully activate the LC and induce effective physiological changes. This hypothesis is supported by the greater PER (Fig. 2D), as well as changes in autonomic response or in cortical state, observed after direct stimulation of the LC (Drolet and Gauthier 1985). Incidentally, this also indicates that PER is probably the most sensitive response to evaluate CC-LC functional connectivity.

In order to corroborate the implication of LC-NA neurons in the PER, we injected the DSP-4 neurotoxin, known to preferentially destroy LC-NA neuron terminals (Ross and Stenfors 2015). Our results showed that DSP-4 abolished the PER response. Moreover, after stimulation of area 25 in animals injected with DSP4, the tissue level of NA quantified by HPLC/EC revealed, as expected, a dramatic NA loss within LC-prominent efferences such as hippocampus or amygdala. The lower loss of NA in the hypothalamus of DSP4-injected rats is consistent with multiple NA sources in the hypothalamus beyond LC efferences (Cunningham and Sawchenko 1988; Aston 2004). Animals treated with DSP4 also showed significant suppression of the PER after FS stimulation, a procedure known to involve LC neurons (Cedarbaum and Aghajanian 1978) while preserving pupil light reflexes, thereby confirming the causal role of LC-NA in the PER.

Whether these PER gradients rely only on anatomic specificity is not addressed in this study. However, several reasons can account for this difference in amplitude or kinetics. In this respect, the precise location of cortical termination on LC neurons is critical but very difficult to estimate (see, for example, discrepancy regarding A32 or A25 labeling projection in Heidbreder and Groenewegen 2003; Vertes 2004; Lu et al. 2012; Jiménez-Sánchez et al. 2016) or even to confirm (eg for 24a′/24b′, see Paxinos and Watson 2013). Moreover, the topographical organization of the LC efferent could be localized in specific subregions of the nuclear core or distinct pericoerulear regions, potentially associated with different functions (Schwarz and Luo 2015). Furthermore, it is now clear that LC has a modular organization, based on spatial segmentation or at least on neuronal heterogeneity, allowing different LC neuronal populations to project toward separate brain regions and subserve diverse functions (Poe et al. 2020). It is then likely that distinct category of LC neurons should be more specifically related to the PER. The existence of direct projection to sympathetic autonomic regulation centers in the spinal cord controlling pupil dilation is in accordance with this hypothesis (Clark and Proudfit 1991; Szabadi 2018).

In addition to direct projection from the cortex to the LC, there are multiple indirect routes between the cortical subregion and the LC (Jodo and Aston-Jones 1997; Jodo et al. 1998) that could exert a differential influence on the PER. Among them, intracingulate connection is of particular importance. Area 32, for example, sends input both directly to the LC and to all the other LC-projecting cortical areas from ACC and MCC. Conversely, area 24a′ projects only to areas 24b′ and 24b additionally to LC while area 24a seems in an intermediate position (Vogt 2015). In this context, the net functional effect should result in a progressively increased and longer PER from stimulation of areas 24a′, 24a, and 32. Our results favor this hypothesis of a cumulative effect on PER when ACC subregions are stimulated in comparison with MCC (Fig. 2).

Finally, it is well known that LC-NA neurons also project mono-synaptically to forebrain regions, controlling back cortical activation, thanks to these reciprocal connections (Chandler et al. 2013, 2014). These bidirectional circuits are of great functional importance (Berridge and Waterhouse 2003; Uematsu et al. 2015; Poe et al. 2020), but imply that LC neurons can consequently be antidromically activated after ICMS (Jodo and Aston-Jones 1997; Jodo et al. 1998) and potentially contribute to the PER. When we recorded LC neurons during stimulation of area 25 receiving more input from LC than areas 24a and 24b (Chandler et al. 2013)), we did not find LC neurons antidromically activated. Moreover, it has been shown that antidromic activation of LC neurons induced inhibitory response via inhibitory recurrent LC collaterals (Aghajanian et al. 1977). While we cannot exclude the possibility that our cortical stimulations also induce antidromic conduction, they mainly activated synaptically LC-NA neurons, resulting in a global net dilatation. This confirms that inhibition when it exists is overridden by orthodromic activation (Jodo et al. 1998).

Numerous ongoing studies continue to detail the complex anatomical pathways existing between the cingulate cortex and LC-NA neurons, and there is still a need for further exploration of this circuitry (Poe et al. 2020). In the meantime, the PER provides a comprehensive functional readout of CC-LC pathways and confirms that pupillometry is a sensitive approach to apprehending CC-LC interaction.

In this respect, PER is a good indicator of the functional heterogeneity of rodent cingulate regions and our study provides a first functional map that should be useful in absence of a complete anatomical description of the very complex cingulate-LC interactions. Taking into account the homology between human and rodent cingulate regions, this study could then be helpful to guide the targeting of specific cortical regions in order to modulate efficiently the NA system for clinical and fundamental research purposes (Torres-Sanchez et al. 2018).

Acknowledgments

We thank Maria-Carmen Medrano for comments on the manuscript and for very helpful discussions.

Author contributions

V.S.: Investigation, Data curation, Formal analysis, Writing—original draft; G.C.: Software, Methodology; P.D.D.: Investigation, Methodology; S.V.: Investigation, Resources; S.D.: Investigation, Resources; L.C.: Investigation; F.G.: Methodology, Investigation, Writing—original draft; T.M.: Conceptualization, Methodology, Data curation, Formal analysis, Writing—original draft, Writing—review & editing.

Funding

This work was supported by the French government in the framework of the University of Bordeaux’s IdEx “Investments for the Future” program/GPR BRAIN_2030. V.S. was supported by the Fondation pour la Recherche Médicale (grant number FDT202304016345).

Conflict of interest statement: None declared.

References

Author notes