Breathing is a complex, vital function that can be modulated to influence physical and mental well-being. However, the role of cortical and subcortical brain regions in voluntary control of human respiration is underexplored. Here we investigated the influence of damage to human frontal, temporal, or limbic regions on the sensation and regulation of breathing patterns. Participants performed a respiratory regulation task across regular and irregular frequencies ranging from 6 to 60 breaths per minute (bpm), with a counterbalanced hand motor control task. Interoceptive and affective states induced by each condition were assessed via questionnaire and autonomic signals were indexed via skin conductance. Participants with focal lesions to the bilateral frontal lobe, right insula/basal ganglia, and left medial temporal lobe demonstrated significantly lower performance than individually matched healthy comparisons during the breathing and motor tasks. They also reported significantly higher anxiety during the 60-bpm regular and irregular breathing trials than healthy participants. Correlations between task performance, anxiety, and difficulty varied across experimental conditions and groups. This study demonstrates that damage to frontal, temporal, or limbic regions is associated with abnormal voluntary respiratory and motor regulation and tachypnea-related anxiety, highlighting the role of the forebrain in affective and motor responses during breathing.

Measurements and Recording Devices

Demographic data (age, gender, handedness, ethnicity, and height) were gathered through a questionnaire. Body weight was physically measured to calculate body mass index (BMI).

Physiological data were gathered via a MP100 acquisition unit (Biopac Inc., Santa Barbara, CA). Breathing activity was monitored using an elastic thoracic breathing belt and amplifier (RSP100C) at the mid-sternal level. Cardiac activity was measured using two electrodes paired with an amplifier (ECG100C) designed for electrocardiogram (ECG) monitoring and recorded at 200 Hz with a high pass filter to stabilize the ECG baseline. The positive electrode was positioned on the left lower anterior abdomen and the negative electrode on the right anterior chest at the mid-clavicular line (lead II configuration). Galvanic skin response electrodes, placed on the thenar and hypothenar eminences of the hand contralateral to the dial hand, were collected using a single-channel amplifier (EDA100C) at 200 Hz.

A custom-built dial was employed to measure hand movement regulation, consisting of a rotating potentiometer with a continuous rating scale ranging from 0.000 to 5.000 Volts at a sampling rate of 200 Hz, as utilized in our previous studies (Khalsa et al., 2009b; 2020). Participants were asked to rotate the dial using their dominant hand following a visual cue, while breathing normally during each trial.

Statistical Analysis

Descriptive statistics for demographic data were compared for the lesion and matched-healthy comparison group. T-tests were used for metric variables (age, BMI), with normality checked using the Shapiro-Wilk test and histogram distributions.

To evaluate respiratory and motor regulation performance, maximum cross-correlation values were calculated between the instructed and measured signals (respiratory trace for breathing trials and dial trace for motor trials) using MatLab R2021a software (Mathworks Inc., Natick, MA) and transformed using Fisher's Z-transformation to address skewed sampling distributions of correlation coefficients (Silver & Dunlap, 1987). Cross-correlations were calculated using the built-in xcorr function. Linear mixed effects models (LME) were used to compare performance across the respiration and motor regulation tasks via separate models for the regular and irregular tasks using the lme4 package in R (Bates et al., 2015).

For LMEs, random intercepts were included for each subject, as represented by the term (1|ID) in the model formulations; see next paragraph. This allowed us to include individual baseline values for each subject to account for between-participant variability. Additionally, we controlled for age in our models to account for its potential influence on the outcomes. However, random slopes for individual subjects were not included in these models, as it was assumed that the variability in slopes across subjects was not essential for the dependent variables under investigation.

We used the respective interaction terms between group membership (healthy vs. lesion) and the other predictor variables (breathing and motor performance, anxiety, HRV) as fixed effects for all tests, while controlling for age. The full term is outlined here as a model: value ~ group*condition + age + (1|ID).

For the significance tests of the fixed effects in our Linear Mixed-Effects (LME) models, Wald chi-square tests were conducted. These tests aimed to determine the significance of the fixed effects and were carried out using the summary-function in conjunction with the lmerTest package in R (Kuznetsova, Brockhoff & Christensen, 2022). For the post-hoc analysis, the emmeans R-package (Lenth, 2018) was employed. The analysis excluded participants who were missing respiratory or motor regulation data for more than two trials per task. In the multilevel correlation analysis, Spearman's correlation coefficients were calculated using R.

For skin conductance response (SCR) analysis, we utilized the open-source software Ledalab (version 3.x; Benedek & Kaernbach, 2010) for Matlab. Data were first preprocessed using a 1st order low-pass Butterworth filter with a cutoff frequency of 5 Hz. Adaptive smoothing was performed using Gaussian windows with a width of 16 samples. CDA.ISCR (Cumulative Driver Amplitude Integrated with Skin Conductance Response) was the dependent variable calculated for in-depth analysis of the SCR signals. CDA.ISCR represents the area covered by the temporal integration of the phasic driver of the SCR with respect to the duration of the SCR. This is achieved by multiplying the SCR by the size of the response window. CDA.ISCR is expressed in microsiemens-seconds (μS*s) and accounts for both the intensity and duration of the SCR over time. This allows a detailed study of the temporal evolution and intensity of responses to different stimuli.