Limbic and paralimbic respiratory modulation: from inhibition to enhancement

Abstract

Objective:

Increased understanding of the role of cortical structures in respiratory control may help the understanding of seizure-induced respiratory dysfunction that leads to sudden death in epilepsy (SUDEP). The aim of this study was to characterize respiratory responses to electrical stimulation (ES), including inhibition and enhancement of respiration.

Methods:

We prospectively recruited 19 consecutive patients with intractable epilepsy undergoing stereotactic EEG evaluation from June 2015 to June 2018. Inclusion criteria were patients ≥18 years and in whom ES was indicated for clinical mapping of ictal onset or eloquent cortex as part of the presurgical evaluation. ES was carried out at 50 Hz, 0.2 ms and 1–10 mA current intensity. Common brain regions sampled across all patients were- amygdala (AMY), hippocampus (HG), anterior cingulate gyrus (CING), orbitofrontal cortex (OrbF), temporal neocortex (TNC), temporal pole (TP) and entorhinal cortex (ERC). 755 stimulations were conducted. Quantitative analysis of breathing signal i.e., changes in breathing rate (BR), depth (TV), and minute ventilation (MV) was carried out during ES using the BreathMetrics breathing waveform analysis toolbox. Electrocardiogram, arterial oxygen saturation, end-tidal and transcutaneous carbon dioxide, nasal airflow, and abdominal and thoracic plethysmography were continuously monitored during stimulations.

Results:

Electrical stimulation of TP and CING (at lower current strengths <3mA) increased TV and MV. At 7–10mA, CING decreased TV and MV. On the other hand, decreased TV and MV occurred with stimulation of mesial temporal structures such as AMY and HG. Breathing changes were dependent on stimulation intensity. Lateral temporal, entorhinal, and orbitofrontal cortices did not affect breathing either way.

Significance:

These findings suggest that breathing responses other than apnea can be induced by ES. Identification of two regions, the temporal pole and anterior cingulate gyrus, for enhancement of breathing may be important in paving the way to future development of strategies for prevention of SUDEP.

Introduction:

Cortical modulation of breathing is well known, and several structures underpinning these processes have been identified ^1–6. The role of limbic/paralimbic structures has been demonstrated through electrical stimulation (ES) studies, which have produced apneic changes ^7–12. Seizure propagation into these regions leads to apnea and hypopnea¹³, which when prolonged, is thought to contribute to fatal outcomes such as sudden unexpected death in epilepsy (SUDEP) ^{7, 14–19}. Several knowledge gaps remain. First, seizure-induced, as well as ES-induced, non-apneic breathing changes have not been well characterized. Second, whether specific ES parameters in breathing modulatory structures enhance rather than inhibit breathing is unknown. This is particularly important since breathing rescue in apneic-hypopneic patients in the peri-ictal state may contribute to a targeted SUDEP prevention strategy using neuromodulation. In this study, analyzed both inhibitory and enhanced respiratory changes induced by ES of several human brain regions by using quantitative analysis of breathing waveforms to measure breathing rate (BR), tidal volume (TV), and minute ventilation (MV). We aimed to identify and characterize specific non-apneic breathing changes. We hypothesized that forebrain nodes that modulate breathing rate, depth, and volume can be identified with this approach. Further, we hypothesized that specific nodes sub-serving breathing enhancement can be potential candidate neuromodulatory targets for future investigation of breathing rescue in cases of peri-ictal breathing compromise.

Methods:

Experimental design:

We analyzed ES data from 19 patients with drug-resistant epilepsy who underwent stereoelectroencephalography (SEEG) between June 2015 and June 2018, as a part of an IRB approved study. Inclusion criteria were patients aged ≥18 years and in whom ES was clinically indicated for mapping of ictal onset or eloquent cortex. Number and locations of electrodes were tailored according to the putative epileptogenic zone. Platinum iridium depth electrodes (1.1mm diameter, 2.5mm length, 5mm spacing), were implanted stereotactically under general anesthesia.

Cortical reconstruction and volumetric segmentation of T1 structural MRI scans, including amygdala nuclear parcellation, were performed using FreeSurfer image analysis suite (http://surfer.nmr.mgh.harvard.edu). Electrode localization was performed by co-registering post-implantation CT scans to preimplant MRI, using FMRIB’s Linear Image Registration Tool (FLIRT) linear registration available in FMRIB Software Library (FSLv5.0.9) (https://fsl.fmrib.ox.ac.uk/fsl/). 3D reconstructions were performed using 3D slicer (http://www.slicer.org) by adjusting thresholds to remove regions of no interest. Target regions of SEEG contacts were obtained using Desikan-Killiany atlas²⁰. Common brain regions sampled across all patients were amygdala (AMY), hippocampus (HG), anterior cingulate (CING), orbitofrontal cortex (OrbF), temporal neocortex (TNC-combined superior, middle and inferior temporal gyri), temporal pole (TP) and entorhinal cortex (ERC). All these regions were stimulated (Figure 1). Amygdala was further subdivided into various subnuclei to look for effects of stimulating the different subnucle using Saygin’s & Kliemann’s atlas for amygdalar subparcellation adapted in FreeSurfer (https://surfer.nmr.mgh.harvard.edu/). Clinical standard of care orthogonal electrode placement allows sampling mainly of basal, accessory basal and lateral nuclei²¹.

Figure 1:

Study pipeline: (A) Anatomical localization of SEEG electrodes to Desikan Killiany atlas was performed (visualized with iElectrodes). (B) Electrode contacts in 7 regions were used in stimulation protocol. These are shown color coded based on the 7 regions. The electrodes are color coded according to the subject number. (C) Electrical stimulation was accompanied by simultaneous recording of airflow signal using Salter Thermistors. (D) MATLAB based in-house code was developed to simultaneously view and clip stimulation segments (red dotted box) and parse the airflow signal for analysis. (E) Quantitative analysis of airflow signal was performed using the BreathMetrics toolbox²² to calculate tidal volume (TV – red double-ended arrow), breathing rate (BR – blue vertical tags at the peak of each inspiration) and minute ventilation (MV – effective ventilation in one minute).

Electrical stimulation (ES):

Bedside ES of cortical regions listed above was done using Ojemann (Integra Life Sciences, Plainsboro, NJ), Grass S-88X (Astro- Med,West Warwick, RI), or MS-120BK-EEG stimulators (Nihon Kohden, Tokyo, Japan). We used bipolar biphasic ES with pulse width 0.2ms, frequency 50Hz, intensity from 1mA to 10mA, and duration from 10 to 40 seconds (mean: 24s). The mean inter-stimulus duration was 30s. Real-time monitoring of breathing and electrocardiogram (ECG) were performed throughout testing. Patients were asked to report any abnormal sensations or discomfort.

Physiological recordings:

SEEG and polygraphic data were obtained simultaneously during ES sessions. Airflow was monitored using thermistors (Thermocouple Airflow Sensor; Pro-Tech, Philadelphia, PA), peripheral oxygen saturation (SpO₂) using pulse oximeters (Nellcor OxiMax N-600x, Dublin, Ireland), end-tidal carbon dioxide (ETCO₂) using capnography (Model 7900; Philips, Netherlands) and transcutaneous CO₂ using digital transcutaneous CO₂ sensors (Digital V-Sign; SenTec, Therwil, Switzerland)]. Chest and abdominal excursions were recorded using inductance plethysmography (Ambu [Ballerup, Denmark] Sleepmate). EEG and ECG were acquired using a diagnostic system (EEG-1200; Nihon Kohden). Custom-made algorithms were prepared to annotate stimulation trials.

Data preprocessing:

Using in-house MATLAB 2020b (https://www.mathworks.com/) algorithms, SEEG and airflow signals were preprocessed (SEEG: bandpass: 1–150Hz and airflow signal: lowpass: 10Hz) simultaneously to visually review the ES and respiratory channels with appropriate filtering. For quantitative analysis of the airflow signals, ES epochs were band-stop filtered (60Hz [±0.5Hz]), lowpass filtered (<5Hz to get a smooth respiratory signal after eliminating non-respiratory high-frequency artifacts), and de-trended to remove signal drifts. Data were clipped into epochs of baseline and stimulation segments and parsed for further analysis using annotations as a guide. We extracted additional 4–8 baseline segments of quiet resting breathing for each patient. All data were manually scrutinized for discontinuous data and artifacts and excluded from analysis. Quantitative breathing waveform analysis was performed to measure BR, TV, and MV, time-locked to ES, using BreathMetrics. We preferred airflow analysis to inductance plethysmography since BreathMetrics was validated on airflow recordings ²² and is less influenced by factors such as chest circumference, bed position and non-breathing thoraco-abdominal movement artifact ^{10, 22, 23}. Though not a closed-system of measuring respiratory signals, we minimize the error of estimating volume changes by estimating a % change compared to baseline where the losses form a common denominator for both the stimulation and baseline time windows.

Respiratory cycle events were identified in individual segments. We then visually scrutinized the position of event-markers, i.e., inhale and exhale onsets/pauses. Event-markers were corrected when detected inaccurately. Quantification of breathing waveforms to estimate a relative change in the nasal airflow waveforms to assess the effects on BR, TV and MV. This relative change (% change compared to baseline) was used as a surrogate because nasal airflow waveforms are not recorded as a closed system. ES-induced apnea has been reported by our group in a previous study ⁸ and was thus excluded for further analysis in this study since our focus was on non-apneic breathing changes. Finally, we studied ES-induced changes in respiration as a percentage change (increase/decrease) from preceding inter-stimulus baseline windows (BL):

$$\text{e.g.,}\%\text{change}\text{in}\text{breathing}\text{rate}(\text{BR})\text{during}\text{stimulation}=\left(\frac{{BR}_{stim}-{BR}_{BL}}{{BR}_{BL}}\right)\mathrm{*}100$$ Eq 1

Similarly, a % change in TV and MV were calculated during stimulation.

Statistical analysis:

ES-induced changes (increases/decreases) in BR, TV and MV, were tested using a one sample t-test to determine significance (false discovery rate [FDR] corrected for multiple comparisons across regions). To evaluate the effect of intensity, ES was stratified into 3 groups (<3mA, 3–7mA, 7–10mA) for comparison. We then studied ES current, duration, region and amygdalar subparcellation for reliable prediction of discrete ES-induced breathing changes using logistic regression with a simple random sampling bootstrapping model while controlling for independent effects of inclusive variables.

Results:

Patients and clinical setting:

Nineteen patients were studied (10 female; mean age 41 [18–69] years), reported in a previous apnea study⁸. None had known cardiorespiratory illnesses except for essential hypertension in patient-12. After thorough manual quality-check of airflow signals, data from 10 subjects were suitable for analysis with continuous artifact-free respiratory waveform. Demographic and clinical details are listed in Table 1.

Table 1.

Demographic and clinical details of study patients.

Patient	Age (y)	Gender	Epilepsy Duration (y)	Handedness	EZ	ICA	Length of ICA (s)	Seizure Frequency	FBTCS Frequency	MRI Brain
2	36	F	4	R	L T	No	-	1/w	1/y	Normal
3	48	M	3	R	L mTLE	No	-	2/d	None	Normal
4	39	M	10	R	L T	No	-	3/w	2/m	Normal
6	66	M	30	R	L HG	No	-	2–3/w	1/y	Normal
7	20	F	14	R	L Hem	Yes	10	1/d	1/d	Normal
8	32	F	25	R	L Hem	No	-	1/m	1/m	R F Heterotopia
11	69	F	44	R	R mTLE	No	-	1/m	Once	Normal
13	38	M	2	R	L OrbF	No	-	2/w	None	Normal
14	33	M	10	R	R T	Yes	10	1–2/w	½ m	Normal
15	27	F	7	R	L T	Yes	10	1–2/m	Twice	Normal

y-years, EZ-epileptogenic zone, ICA-ictal central apnea, FBTCS-focal to bilateral tonic clonic seizures, MRI-magnetic resonance imaging, F-female, M-male, R-right, L-left, T-temporal, F-frontal TLE-temporal lobe epilepsy, HG-hippocampal gyrus, Hem-hemisphere, OrbF-orbitofrontal cortex, w-week, d-day, mo-month

Stimulating electrodes:

A total of 1,510 electrode contacts were implanted. A total of 755 stimulations were conducted according to study protocol (AMY:266 stimulations, HG:171, OrbF:95, TP:77, TNC:66, CING:55 and ERC:25). Of these, clean respiratory waveforms were available from 246 (after excluding movement artifacts and discontinuous respiratory signal, and segments with less than five breaths to quantify respiratory indices).

Effect of ES on breathing measures:

ES of several cortical regions produced changes in BR, TV and MV (Figure 2).

Figure 2:

Effect of stimulation on breathing. All breathing metrics breathing rate (BR), tidal volume (TV) and minute ventilation (MV) are expressed as a percentage change from baseline. This is a composite summary of aggregated data across all stimulation intensities. (A) Overall effect of cortical stimulation on BR: CING and HG stimulation resulted in a decrease in BR. (B) Stimulating AMY and HG resulted in a significant decrease in TV. Stimulating TP resulted in a significant increase in TV. (C) Stimulating amygdala and hippocampus resulted in a significant decrease in MV. (D) The table summarizes the regions that show the main effects of stimulation. [*: Bonferroni corrected p value <0.05; AMY: Amygdala; CING: Anterior cingulate gyrus; HG: Hippocampus; OrbF: Orbitofrontal cortex; ERC: Entorhinal cortex; TNC: Temporal neocortex consisting of superior, middle and inferior temporal gyri; TP: temporal pole].

Overall, 62% of ES trials showed a reduction in BR, 73% showed a reduction in TV, and 75% showed a reduction in MV. The main effect of stimulating AMY was ~50% decrease in TV and consequently MV, compared to pre-stimulation baseline (T-values (Ts) < −10.8, P-values (Ps) < 0.001). AMY stimulation did not have a significant effect on BR. Stimulation of HG produced a significant decrease in all three breathing parameters (BR: ~7% decrease, TV: ~25% decrease, and MV: ~30% decrease. Ts<−2.9, Ps<0.005). Overall, the decrease in MV was associated with an increase in ETCO₂ levels (t=2.5, p=0.013) and a decrease in SpO₂ level (t=−3.9, p<0.001). Stimulation of AMY produced increased ETCO₂ levels: t=3.1, p=0.004 and increased SpO₂ levels: t=−2.5, p=0.01).

Stimulating CING produced an effect only on BR (~16% decrease. T<−4.4, p<0.001) but not on TV and MV (Figure 2 A, B & C). There was no significant change in SpO₂ (t=−1.86, p=0.1).

Stimulation of TP, resulted in ~42% increase in TV (T=2.7, P=0.017). There was no significant change in SpO₂ (t=−1.86, p=0.1) (Figure 2B). There was no significant change in SpO₂ (t=2, p=0.1) and ETCO₂ (t=−1, p=0.4). These changes depict averaged group level changes in breathing. However, individual stimulations can produce physiological variations in responses as depicted in Figure 5.

Figure 5:

(A) Stimulation of amygdala (8mA) resulted in a decrease in the depth of breathing. (B) Stimulation of cingulate (2mA) resulted in a sustained increase in the depth of breathing and a decrease in breathing rate. (C) Stimulation of TP (2mA) results in an increase in the depth of breathing.

The other regions, including OrbF, ERC, and TNC did not induce any significant changes.

Dose-dependent neuromodulatory effects on the breathing:

The charge density for each of the stimulation strengths were <7.5μC/ cm2/phase for stimulations <3mA, 7.5–17.6μC/ cm2/phase for stimulations between 3–7mA, and 17.6–25.2μC/ cm2/phase for stimulations between 7–10mA.

Stimulating AMY at 7–10mA was associated with a 78% decrease in TV and a 77% reduction in MV. This was lower compared to stimulating at <3mA (TV: mean difference (MD)= −38.5%, _Bonferronip= 0.001; MV: MD = −34.5%, _Bonferronip= 0.002) and 3–7mA (TV: MD = −32.4%, _Bonferronip= 0.008; MV: MD = −30.3%, _Bonferronip= 0.011)(Figure 3A)

Figure 3:

Breathing changes (in breathing rate, tidal volume and minute ventilation) with increasing current intensity. A) Percentage changes are shown to be significant in two regions i.e., amygdala and anterior cingulate gyrus. Bar-graphs indicate response to increasing current intensity (3mA, 3–7mA, 7–10mA). (B) shows region-wise changes in BR, TV, and MV, with a summary of differences between stimulation at low (<3mA) and high currents (7–10mA). [** - fdr p value <0.05, * p uncorrected <0.05].

Stimulating CING at <3mA was associated with a 48% increase in TV and a 20% increase in MV. This was an increase in TV and MV compared to stimulating at <7–10mA (TV: MD= 77.6%, _Bonferronip= 0.03; MV: MD = 61%, _Uncp= 0.019) and 3–7mA (TV: MD = 48%, _Uncp= 0.04)(Figure 3B).

Other regions did not show any stimulation induced effects on the breathing parameters comparing different current intensities.

Effects of stimulating different amygdalar nuclei on BR, TV, and MV:

All electrode contacts within the amygdala were assigned to specific subnuclei. Quantitative analysis of respiratory signal could be performed on a total of 79 amygdalar stimulations [Accessory basal: 10, Basal (62) and lateral (7)](Figure 4A).

Figure 4:

(A) Glass brain showing electrodes implanted into the amygdala in seven patients. The amygdala was parcellated into three nuclei - accessory basal (ACC BAS), basal (BAS) and lateral (LAT). Electrodes are color coded by subject. (B) Changes in breathing rate (BR), (C) changes in tidal volume (TV), and (D) changes in minute ventilation (MV) while stimulating the three amygdala nuclei. Stars indicate significant change compared to baseline (https://surfer.nmr.mgh.harvard.edu/fswiki).

BR decreased while stimulating the accessory basal (t=−3.4, p=0.008) and lateral nuclei (t=−3.1, p=0.02) but not the basal nucleus (t=1.1, p=0.3). TV decreased while stimulating the accessory basal (t=−49.8, p=<0.001) and basal nuclei (t=−8.1, p<0.001) but not the lateral nucleus (t=−2.2, p=0.06)( Figure 4B). Consequently, the most prominent decrease in MV was noted with stimulation of the accessory basal nucleus (t=−87.5, p<0.001), followed by the basal (t=−8.2, p<0.001). Comparatively, the lateral nucleus produced the least decrease in MV (−3.5, p=0.01)(Figure 4C).

A one-way ANOVA comparison between the different subnuclei showed that decreases in TV (f=15.1, p=0.001) and MV (f=13.4, p=0.012) were most pronounced during stimulation of the accessory basal nucleus, followed by basal nucleus. Stimulating accessory basal and lateral nuclei was associated with decrease in BR compared to basal nucleus stimulation (f=8.2, p=0.001).

Do stimulation parameters influence changes in BR, TV and MV?

ES parameters (current and duration) and stimulation site were studied. We found that the regions stimulated (MV: Wald’s: 15.2, p=0.03; TV: Wald’s: 21.2, p=0.004) and stimulation duration (MV: beta: −0.07, p=0.02; TV: beta: −0.07, p=0.02) were associated with decrease in TV and MV (Model Accuracy: 91%). Among the regions stimulated, AMY (Wald’s:13.6, p=<0.001) and HG (Wald’s:13.6, p=<0.001) had greater odds of being associated with decreases in TV. Within the amygdala, accessory basal (BR: r=−0.7, p=0.03) and basal (TV: r=−0.31, p=0.04), had greater odds of being associated with decreases in TV.

Discussion:

Normal breathing metrics comprise breathing rate (BR), tidal volume (TV), and minute ventilation (MV). Using quantitative breathing waveform analysis, we inferred changes in respiratory rate and volume from nasal airflow waveforms as a surrogate of changes in BR, TV and MV, we identified several non-apneic breathing changes produced by ES. Here, we report the novel finding of breathing enhancement (increased TV and/or MV), produced by ES of temporal pole (TP) and anterior cingulate gyrus (CING). We also found that non-apneic breathing changes are commonly produced by ES of amygdala [AMY] and hippocampus [HG], and are likely also encountered during spontaneous seizures, although in practice, these are neither looked for nor reported. Temporal (combined superior, middle and inferior temporal gyri), entorhinal and orbitofrontal cortices did not produce breathing changes.

Several cortical structures are involved in breathing modulation ^{1, 7–10, 12}. While respiration primarily concerns metabolism and homeostasis, some cognitive states modify rate and depth of breathing ^24–26. In addition to diaphragmatic control²⁷, evidence of breathing modulation by limbic structures ^{1–3, 28}, and intracranial-EEG demonstration of respiratory influences on cognitive function via multiple sensory pathways^{6, 29}, suggest complex cortical breathing networks above brainstem level. The extent of this network is unmapped. Limited previous studies have centered mainly on amygdala^{7, 10}. Alteration of respiratory movements have been induced by ES of other regions, ⁷ although inconsistently ^{8, 11}.

Here, using quantitative analysis, we show that sites within the limbic/paralimbic (AMY, HG, TP, and CING) system, not just limited to amygdala, have the potential to modulate breathing, confirming human limbic/paralimbic breathing control. Quantitative breathing analysis during stimulation periods allowed identification of additional, subtle breathing changes beyond apnea and thus, additional cortical regions that modulate breathing. Limitations of visual analysis may explain why many studies using cortical stimulation of sites outside AMY, failed to show significance ¹¹.

Enhancement of breathing - temporal polar and anterior cingulate neocortices.

Previous studies reported involvement of regions outside the amygdalo-hippocampal complex, including insula, CING and TP, as having inhibitory influences on breathing ^{1, 30–33}. In fMRI studies, similar regions, were identified as breathing modulatory sites ³⁴. Visual analysis of ES in regions outside the temporal lobe did not show changes in prior studies⁸. Here, using grouped quantitative analysis of respiration, we identified TP and CING as modulatory sites for breathing enhancement. Their stimulation increased TV. In CING, breathing responses were found to be stimulation intensity-dependent with lower current strengths enhancing breathing and higher current strengths decreasing breathing.

Temporal pole increases TV.

Whereas previous observations suggested a TP role in breathing inhibition ^{1, 32}, this is the first report of TP stimulation inducing increased TV. TP has connections with AMY and HG ^{35, 36}. The former, through the fasciculus amygdalo-temporalis, has rich intermodal connectivity with TP³⁵, which in turn has heavy projections to HG. Conversely, TP receives hippocampal afferents via the subiculum³⁵. Thus, there is rich connectivity between several mesial temporal network nodes, with likely outputs to brainstem nuclei via amygdala efferents. The anatomo-functional relationship between TP and brainstem is not clearly defined. The hippocampus, amygdala and cingulate cortex are shown to have 5HT1A receptors, but selective enhancement of breathing in TP which also has 5HT1A receptors is not clearly explained by serotonergic connectivity alone. This remains speculative and additional work will be needed to study this pathway of respiratory enhancement.

Anterior cingulate gyrus increases TV and MV.

Only low intensity CING stimulation produced increased TV and MV. These findings are unsurprising; although human stimulation studies had exclusively described breathing suppression, animal studies reported a variety of responses, including increased depth and/or rate with CING and AMY stimulation ^{32, 33, 37}. In recent studies using visual analysis, CING stimulation failed to induce any breathing effect ^{8, 11}, emphasizing the limitations of visual approaches. However, our results are in consonance with previous observations of both apnea and increased BR reported in humans by Pool and Ransohoff, in 1949 ³⁰. CING single neuron discharges show relationships with respiratory patterning ³⁸. fMRI signals at breathing onset provide further support; activity of CING is synchronized to onset of inspiration in periodic breathing ³⁹. Event-related fMRI has been used to examine limbic/paralimbic-bulbar circuitry underlying human respiratory-related neural activity during unlabored spontaneous breathing at rest and during cognitive tasks ³⁴. A simple cognitive task produced an increase in breathing frequency and strongly modulated neural activity in the pontine raphe magnus, amygdala, and anterior cingulate cortex ³⁴. It is likely that variability in type and magnitude of respiratory effects obtained from CING ES, reflect the strength and characteristics of the stimulus, but may also be an indication that these regions by themselves are divided in their representation of respiratory function.

Non-apneic inhibition of breathing-the amygdalo-hippocampal paleocortex and archicortex.

ES of AMY and HG decreased TV, and MV. In addition, ES of HG and accessory basal and lateral nucleus of the AMY also produced decreased BR, suggesting that non-apneic breathing changes occur in these structures. These areas are consistent with previous studies using visual analysis of breathing signal; AMY and HG ES-induced hypopnea or apnea^7–10. In this study, higher current intensities decreased TV and MV more prominently. This is in concordance with our previous work⁹. We noted that lower current intensities produce subtle decrease breathing changes whereas higher intensities produce frank apnea.⁹

In AMY, we found the more medial the stimulated nuclei, the greater the drop in TV and MV. This is consistent with our previous finding, that apnea is elicited with stimulation of medial amygdalar contacts. In contrast, lateral amygdala required higher frequency and intensity to induce apnea ⁸. Other studies found ES-induced apnea to occur only with medial most amygdalar stimulation^{10, 11}. Rhone et al used a machine learning model to localize the region of the amygdala that inhibits respiration and induces apnea (AIR site). This AIR site was localized to the medial amygdala. While their study highlighted apnea, we found consistent inhibitory breathing patterns (greater decrease in the % change in TV and MV) with the stimulation of the medial amygdala stimulation compared to stimulating lateral amygdala¹¹. Nobis et al have reported that stimulation-induced apnea occurred only when stimulating the medial-most amygdalar contacts located in the central nucleus¹⁰. Amygdala nuclei projections may underlie these findings; in animal studies, the lateral amygdala lacks afferent connections to brainstem respiratory centers, unlike medial subregions ⁴⁰. However, lateral amygdala sends projections to HG, TP, and other cortical areas ³⁵. It is also reciprocally connected with amygdalar basal and accessory basal nuclei⁴¹. Thus, lateral amygdala response influences may be exerted through these connections to medial amygdala nuclei or through hippocampal projections. In addition, Totola et al 2019, using an amygdala rapid kindling model have demonstrated a strong coupling between amygdala and respiratory neurons in the brainstem⁴². Mesial temporal modulation of breathing is evidenced by increased HG activity before apnea termination in cats ⁴³. Some HG and AMY neurons phase-lock with the respiratory cycle, suggesting that these structures are involved in breathing regulation ^{2, 29}. To date most of this information about how various cortical regions modulate breathing is based on both direct and indirect evidence on the understanding of a widespread functional connectivity of these regions with the brainstem.

Overall, from the results of our current study and from various human^{9–12, 29} and preclinical animal studies^{1, 32, 37}, it is evident that breathing changes are specific to the stimulation site and not to the seizure network or the epilepsy type^{10, 11}. These responses are specific to the structural brain substrates. In human studies, most of these responses have been reliably and reproducibly elicited with the stimulation of the limbic and paralimbic structures suggesting that aside from the involuntary brainstem control and the voluntary motor network of breathing control, there also exists a limbic-paralimbic breathing network which modulate the respiratory patterns. Some of the most frequently reported structures among these are amygdala, hippocampus and in animal studies the cingulate cortex^{1, 8, 9, 32}. However, what remains to be studied in finer detail, is how these different limbic regions network with the primary mechanisms of breathing control to establish the definitive circuits that modulate breathing. Most human studies have looked at respiratory inhibition leading to apnea as the primary respiratory response. Our current results are the first to quantify that more subtle changes in breathing can be mapped to a widespread cortical network.

However, it is essential to note that there has been a heterogeneity in the findings across various teams in terms of the anatomical substrates that are causing these changes. It is essential to understand the factors contributing to these responses. It should be noted that anatomical substrates underlying these changes have not been consistent across observers. Differences in stimulation current intensities likely explain these discrepancies. Nobis et al’s highest current intensity used was 4mA in all subjects. In our study population, we used up to 10mA for hippocampal stimulation¹⁰. Higher current intensities (walds: 7.4, p=0.006) were highly predictive of apnea induction. Only 34 of 342 stimulations in 5 of 8 patients of Rhone et al’s stimulations were in hippocampus, gyrus rectus and cingulate. No apnea was seen in these structures although details on maximum voltages used in these cases were not available¹¹

Translational significance:

The main finding of ES-induced breathing enhancement provides an opportunity for future investigation of neuromodulatory therapies that have the potential to rescue breathing dysfunction in peri-ictal states in high SUDEP-risk patients.

Evidence from pre-clinical and clinical studies suggest respiratory mechanisms in SUDEP, and that therapeutic focus on leveraging breathing mechanisms may be appropriate ^{14–16, 19, 44}. The MORTEMUS study highlighted a three-minute window for lifesaving intervention, where such an intervention may have impact ¹⁹. Since many SUDEP cases are unwitnessed, systems that detect and rescue apnea hold promise for targeted intervention. Identification of stimulation targets is the first step towards this goal. Several steps remain, including better characterization of stimulation parameters for maximal efficacy and utility. In this study, we found 50Hz stimulation to produce significant breathing enhancement; other stimulation paradigms remain to be characterized.

Several limitations to our study exist. All patients had refractory epilepsy, and electrode sampling depended on clinical diagnosis and the surgical hypothesis for implantation. Thus, not all regions of interest are possible to study, in particular the central and medial amygdalar nuclei which were difficult to implant. Whether ES of cortical regions identified in this study will actually prove to benefit breathing compromise in the peri-ictal state is unknown. It is also necessary to note that the parcels in the atlas used were rather large to map out a more precise anatomical localization of the stimulation effects. This will require a more finely resolved atlas and larger patient population in the future. These data must be reproduced in a larger cohort of patients, and include the peri-ictal state. The amygdaloid complex is structurally diverse and comprises several nuclei and our electrode implantation in the amygdala could not cover all regions. It is pertinent to isolate the effects of stimulation to distinct anatomical structures. For example, high current stimulation of hippocampus is likely to simultaneously spread/stimulate the surrounding amygdala resulting in very similar effects. Hence electrical stimulation based models to study breathing modulation should also estimate the volume of tissue activated and determine if surrounding structures are being simultaneously stimulated. Opportunities for exploring these, using different stimulating parameters, and optimizing stimulation paradigms, are a priority.

We confirm that we have read the journal’s position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.

Key Points:

Quantitative analysis of airflow signal provides a better understanding of electrical stimulation (ES)-induced breathing changes.

Cortical ES can elicit inhibitory breathing responses other than apnea.

ES of temporal pole and anterior cingulate gyrus can increase tidal volume and minute ventilation, paramount to breathing enhancement.