Electroencephalography and respiratory variables were measured as previously described (Hudson et al., 2016). Briefly, participants breathed through a mouthpiece connected to a pneumotachograph (3700 series; Hans Rudolph) and two-way valve (2700 series; Hans Rudolph) and airflow and mouth pressure were measured. EEG was measured from central scalp locations (Cz and FCz), the earlobes and under the right eye to monitor electrooculographic activity using an active electrode system (ActiCap; Brain Products). EEG signals were time locked to respiratory signals with simultaneous digital trigger pulses. Surface electromyographic (EMG) recordings were made from the scalene muscles on the right side with adhesive electrodes to assess if premotor potentials were associated with increases in inspiratory muscle activity as previously observed (Georges et al., 2016). One electrode was placed on the posterior triangle of the neck, posterior to the border of the sternocleidomastoid muscle at the level of the cricoid cartilage, and the second electrode was placed ∼2 cm caudally over the muscle. Respiratory signals and EMG were sampled at 2 kHz and EEG at 2.5 kHz (PowerLab, AD Instruments).

Ten of the fifteen participants performed the big breath condition. Here, during a trial of resting unloaded breathing, participants were instructed to take larger breaths of approximately twice the tidal volume of resting breaths (Figure 1). They were instructed by an experimenter during the preceding expiration and did not have visual feedback of lung volume. Instructions to take big breaths were given randomly, every 2–3 breaths. Eight of the fifteen participants performed a low load condition during which an ITL set to ∼7 cmH₂O was attached to the inspiratory port of the two-way valve in the respiratory apparatus. At the end of this condition, subjects (n = 6) rated their level of respiratory discomfort during the condition using a visual-analog scale that ranged from zero: “no breathlessness or discomfort” to ten: “maximal breathlessness or discomfort.” The other respiratory conditions, i.e., the positive and negative controls of self-paced voluntary sniffs and resting breathing, are described in Hudson et al. (2016) and were performed by all participants in the same recording session as the big breath or low load conditions. For 12 participants, premotor potential data during sniffs and resting breathing were reported previously (Hudson et al., 2016), but 3 participants had not performed the sniff and resting breathing conditions.

Offline, EEG recordings were resampled at 250 Hz, referenced to linked-ear electrodes and band-pass filtered (0.05–10 Hz). EEG was divided into 3.5-s segments, with 2.5 s prior to and 1 s after the onset of inspiration, based on airflow (big breaths) or negative mouth pressure (ITL). In the big breath condition, breaths were identified as either ‘instructed big breaths’ (i.e., the breaths when participants were told to increase tidal volume) or ‘uninstructed standard breaths’ (i.e., breaths in the trial that were of usual tidal volume as participants were not given directive). EEG segments were visually inspected and rejected from the average if they exhibited artifact, e.g., large deviations from baseline or intense EOG activity and the remaining segments were averaged. For the big breaths condition, 59 ± 9 (mean ± SD, range 50–78) and 76 ± 22 (range 55–123) segments were averaged for the instructed big breaths and uninstructed standard breaths, respectively. This corresponded to 87% (range 70–97%) and 82% (range 60–99%) of all segments recorded for the instructed big breaths and uninstructed standard breaths, respectively. For the low load condition, 90 ± 13 segments (range 80–109), or 75% (range 63–87%) of all segments, were averaged. For the three participants who had not performed the sniff and resting breathing conditions previously, 89 ± 4 (range 86–93) and 119 ± 46 (range 68–156) segments were averaged, respectively. This was 93% (range 89–97%) and 77% (range 56–97%) of all segments recorded for sniffs and resting breathing, respectively.

Average EEG traces at Cz and FCz derivations were then de-identified and examined for the presence of an inspiratory premotor potential, i.e., a slow increasingly negative shift in the Cz or FCz EEG signal starting from ∼2–0.5 s before inspiration. The signals were judged by three assessors (authors AH, MR, and TS), all blinded to the condition and participant during assessment, with incidence determined by majority. If present, the latency of each premotor potential was measured as the duration (in ms) between the onset of EEG negativity and the start of inspiration and the amplitude (in μV) of the EEG negativity at the start of inspiration was measured. Respiratory variables were measured for all breaths using a semi-automated script and averaged for each condition. Predicted maximal inspiratory pressure was calculated (Harik-Khan et al., 1998). To measure scalene EMG, the root mean square (rms, time constant 50 ms) of the EMG signal was computed and ensemble averaged for all breaths in each condition (see Hug et al., 2006). The area and amplitude of phasic inspiratory EMG was measured between the onset of inspiratory activity (i.e., the increase in activity from any tonic activity during expiration) and the end of inspiration. To compare between subjects, the inspiratory EMG area and amplitude in instructed big breaths, uninstructed standard breaths and low load were normalized to these measures for the same subject during the trial of resting breathing. Respiratory variables and EMG were also determined for signals recorded during resting breathing for all subjects as these measures were not reported with the previously published EEG data (Hudson et al., 2016).

Group data are presented as mean (SD) or median [IQR] for parametric and non-parametric data, respectively. The H-statistic (degrees of freedom; df) is shown for unpaired data and the F-statistic or Z value (df) is shown for repeated measures and paired data, respectively.

To compare respiratory variables between instructed big breaths and uninstructed standard breaths in the big breath condition and with resting breathing, a one-way repeated measures analysis of variance (ANOVA) was used, with pairwise post hoc testing using the Holm-Sidak test. Normalized EMG was compared between instructed big breaths and uninstructed standard breaths with a paired t-test. A Kruskal–Wallis one-way analysis of variance (ANOVA) on Ranks with pairwise post hoc testing using the Tukey’s test was performed to compare the incidence of inspiratory premotor potentials between conditions. Respiratory variables and scalene EMG in recordings with and without a premotor potential were compared using t-tests.

The event-related potential technique used here is sensitive enough to detect premotor potentials in these types of experimental conditions as evidenced by their detection in resting breathing in patient populations (Tremoureux et al., 2014a; Launois et al., 2015; Georges et al., 2016) and in ‘pre-phonatory’ breaths in speech with similar tidal volume to that in the larger breaths here (Tremoureux et al., 2014b). The number of segments available for the large breath analysis was less than 80, as is typically used to detect respiratory-related premotor potential from average EEG signals (e.g., Raux et al., 2007b). Combined with the presumably smaller signal to noise ratio in this task, i.e., compared to sniffs (Tremoureux et al., 2014b), we may have underestimated the incidence of premotor potentials in the instructed larger breaths and uninstructed standard breaths. However, for the low load condition, 80 or more artifact-free segments were available for each participant and the incidence in this condition was also similar to that in resting breathing. Our observations do not completely rule out some degree of cortical activation, but indicate that if present, it is too weak to be readily and consistently detected by an event-related approach. Alternative methodology to detect cortical activation such as time-frequency maps are not likely to offer better sensitivity given we previously demonstrated that premotor incidence was superior to time-frequency maps to discriminate between resting breathing and moderate inspiratory threshold loading (Hudson et al., 2016).

For the large breath condition, participants were instructed during expiration to increase tidal volume on the subsequent breath, but following this general instruction, participants could choose when and how they made the big breath. Therefore, our protocol in the large breaths condition was not true “volitionality” (Deecke and Soekadar, 2016), but as argued by Schultze-Kraft et al. (2016) who used a similar protocol to investigate the temporal relationship between the readiness potential and the cancelation of movements, nor was it a contingent negative variation (CNV) paradigm (Haynes and Schultze-Kraft, 2016). Thus, in the present study, it was reasonable to expect that a readiness potential would be present prior to the instructed large breaths and to use event related potential methodology to detect it.

Evaluation of the relationship between the presence of a premotor potential and respiratory discomfort is limited in the present study as participants who performed the big breath condition were not asked to rate respiratory discomfort. The current protocol, in which instructed big breaths were interspersed with uninstructed standard breaths, prohibited this assessment. A protocol of sustained voluntary hyperventilation could be used to investigate the relationship between premotor potentials and discomfort with voluntary increases in breath size, and indeed to determine if continuous (i.e., not transient) increases in tidal volume elicit premotor potentials in healthy participants. Compared to resting breathing, sustained voluntary increases in breathing frequency are associated with greater respiratory-EEG coherence from numerous brain areas, including the frontal, precentral, and insular cortices, as measured with intracranial EEG in the non-affected brain areas of epileptic patients (Herrero et al., 2017).

Our data support the emerging theory that in the absence of an automatic breathing defect (see below), for a cortical activation sufficient to give rise to premotor potentials breathing must probably be dyspneic rather than be just a conscious sensation or awareness of breathing. Respiratory sensation arises from feedback from multiple afferents sources in the periphery (i.e., reafference, from for example, the lungs and respiratory muscles), but also in the form of corollary discharge, i.e., an ascending copy of the neural drive to breathe that originates from the medulla and/or motor cortex (Parshall et al., 2012). Typically, with increases in ventilation, there is more corollary discharge, of either medullary or cortical origin, or both, as well as increased afferent feedback, e.g., from the lungs with greater expansion in inspiration. However, the magnitude of corollary discharge with increased drive to breathe is not necessarily matched to an increase in afferent feedback from the periphery, e.g., if there is impaired respiratory mechanics or muscle weakness in respiratory or neurological disease. Dyspnea, the sensation of breathlessness, can be classified as sensations of (i) work or effort, (ii) chest tightness, or (iii) air hunger/unsatisfied inspiration. It is thought that dyspnea occurs not only from an awareness that breathing has increased, but rather from the “perception that the drive to breathe is not being matched by adequate pulmonary ventilation” (Parshall et al., 2012), i.e., a ‘mismatch’ between corollary drive and afferent feedback.

In healthy participants, respiratory-related premotor potentials are typically absent (incidence of 0–10%) during resting breathing, but present (incidence of 50–100%) during threshold or resistive loads of 10 cmH₂O or more, as well as in large ballistic movements like voluntary sniffs (Macefield and Gandevia, 1991; Raux et al., 2007b; Morawiec et al., 2015; Hudson et al., 2016). Although the loading and ballistic conditions are typically associated with increases in ventilatory variables such as tidal volume and/or mouth pressure, respiratory-related premotor potentials in healthy participants are not interrelated with increases in ventilation. In addition to our findings, further evidence that increases in ventilation alone do not evoke premotor potentials in healthy participants, is the nil-to-negligible incidence of cortical activation during hypercapnic breathing and exercise, i.e., comparable to the incidence in resting breathing (Raux et al., 2007b; Jutand et al., 2012; Tremoureux et al., 2014a). Hypercapnic gas mixtures are likely to increase ventilation from stimulation of the automatic respiratory centers and thus a lack of cortical contribution would be expected. Of note, the dyspnea in the hypercapnic condition is described as ‘air hunger,’ rather than ‘effort’ as in inspiratory threshold loading (Raux et al., 2007b). Consistent with no direct relationship between premotor potentials and increases in ventilation in healthy participants, we have shown that voluntary large breaths and low loads also do not strongly elicit of respiratory-related premotor potentials. Furthermore, for these conditions, the respiratory variables were similar in trials with and without a premotor potential.

Along these lines, there can be an increase in the incidence of respiratory-related premotor potentials in healthy participants that is not associated with changes in ventilation. For example, during non-invasive pressure support ventilation in healthy participants adjusted incorrectly to “discomfort” to model “ventilator fighting,” there is 100% incidence of premotor potentials (Raux et al., 2007a). The respiratory variables during these recordings were similar to those when the support ventilation was correctly adjusted (i.e., “comfort”) when the incidence of premotor potentials was only ∼29%. Rather than altered ventilation, the different conditions of “discomfort” and “comfort” were associated with changes in ratings of the intensity of respiratory discomfort, of the work/effort type, and scalene muscle EMG (Raux et al., 2007a). Consistent with this, the incidence of premotor potentials in the low load condition was similar to resting breathing in the present study and the ratings of dyspnea were low. However, no difference in inspiratory scalene EMG in the trials with and without premotor potentials was observed in the larger breaths and low load conditions.

Data from patient groups support the concept that respiratory-related premotor potentials may be related to “discomfort,” rather than increased afferent feedback and an awareness of breathing because ventilation is higher. Approximately 60% of patients with ALS exhibit a premotor potential in resting breathing and respiratory discomfort is rated higher in these patients, compared to those who do not have a premotor potential (Georges et al., 2016). Furthermore, in the sub-group with a premotor potential, there are parallel reductions in respiratory discomfort and premotor potential incidence during non-invasive ventilation (Georges et al., 2016). Although from a separate cohort of ALS patients, non-invasive ventilation typically increases tidal volume (Magalhaes et al., 2016), which suggests the reduction in premotor potential observed by Georges et al. (2016) was not associated with decreases in ventilation (and the associated reduction in respiratory afferent feedback). Impaired automatic drive to breathe in CCHS is associated with a 100% incidence of premotor potentials to maintain ventilation during resting breathing in wakefulness (Tremoureux et al., 2014a). Although the corresponding ventilatory data during these EEG recordings was not reported, the tidal volume and ventilation during resting breathing in wakefulness has been reported to be similar for CCHS and healthy participants (Ranohavimparany, 2012). CCHS patients do not experience dyspnea during resting breathing in wakefulness (Similowski, unpublished observations) or even during exposure to high levels of CO₂ (Shea et al., 1993). However, as suggested by a case study in a CCHS patient, there may be other ‘negative’ effects of continuous cortical control of breathing, such as impaired performance in cognitive tasks (Sharman et al., 2014). Here, larger voluntary breaths that were almost three times the tidal volume of resting breathing resulted in a premotor potential in only 40% of healthy participants. Given the high incidence of premotor potentials in ALS and CCHS which were likely to be associated with breaths of usual or lower tidal volumes, our findings support the concept that the presence of a premotor potential in breathing may be due to a mismatch between central neural drive and afferent feedback from the respiratory system, rather than just due to an awareness of breathing because of increased afferent feedback. This mismatch may manifest as abnormal respiratory perception, i.e., dyspnea, or cognitive impairment, but it remains to be determined whether premotor potentials are a neurophysiological manifestation of the mismatch between one or both (i.e., medullary and cortical) forms of corollary discharge and reafference.

Of course the load-capacity imbalance evoked in healthy participants using external inspiratory threshold loads is not permanent like those in disease (although it may fluctuate in disease, e.g., with exacerbations in chronic obstructive pulmonary disease and asthma). However, premotor potentials in healthy participants do persist during ‘chronic’ or sustained exposure (1 h) of an inspiratory threshold load of 17 cmH₂O or more (Tremoureux et al., 2010). Using functional magnetic resonance imaging, although the network of brain areas associated with continuous inspiratory threshold loading was shown to be limited compared to single-breath load exposure, 3 min of continuously loaded breaths (∼10 cmH₂O inspiratory threshold load) was associated with persistent activation in numerous regions, including the supplementary area (Raux et al., 2013) (see also Gozal et al., 1995). Here, with ∼10 min of a ∼7 cmH₂O load, there may have been greater ‘automatization’ and thus no physiologically significant cortical activation (i.e., as a premotor potential) to detect in the majority of our participants. That there was a limited degree of cortical contribution to the breaths with a low load is corroborated by the fact that ventilation during this condition here (∼7 L/min) is similar to that in resting breathing (Hudson et al., 2016). Given that healthy participants typically over-ventilate in response to an external load in wakefulness, but that this response is abolished and even reversed, during sleep (Henke et al., 1992), the current data of a lower incidence in cortical activation and no change in ventilation are consistent with minimal cortical activation. Immediate changes to the breathing pattern and diaphragm EMG with unexpected presentation of loads near or below the perceptual threshold suggests the response to loading, at least for these very low loads, are not due to perceptual response, i.e., are independent of cortical networks (Daubenspeck and Bennett, 1983; Kobylarz and Daubenspeck, 1992). Even for loads above perceptual threshold, but still of low magnitude (i.e., similar to that used here), the ventilatory response is moderate and transient (Freedman and Weinstein, 1965; Yan and Bates, 1999) compared to ‘bigger loads,’ i.e., those associated with reliable evidence of cortical activation in the form of premotor potentials (Raux et al., 2007b). Given the cortical response to low loads is likely to be transient, or at least diminished in its magnitude (see Raux et al., 2013), a method with superior time resolution (see below) is required to detect any cortical changes associated smaller loads, especially given these are more relevant for pathophysiological conditions.

Compared to the low load, the larger breaths condition was different in that ‘instructed large breaths’ were interspersed with ‘uninstructed standard breaths.’ Therefore, there was breath-to-breath variation which is likely to be associated with less automatization and a different preparatory state to loading. Preparatory state has been identified as a factor that influences the premotor potential (Shibasaki and Hallett, 2006). Here, the incidence of premotor potentials was 40% for large breaths and 20% non-instructed standard breaths with the interspersed protocol compared to 25% in the low load. However, direct comparisons of the incidence between the large breaths and low load conditions is not appropriate due to differences in the ventilation changes, i.e., disparities in tidal volume changes with mouth pressure with and without an inspiratory threshold load. To confirm that breath-to-breath fluctuations in respiratory control evoke more cortical activation compared to continuous increases in cortical activation, future studies should compare cortical EEG between continuous- and interspersed- augmented breaths and between continuous- and single breath-loading (Raux et al., 2013). The ability to detect transient changes in respiratory-related cortical activation using EEG is critical for the continued development of the brain-ventilator interface (Navarro-Sune et al., 2016) that aims to use EEG to optimize mechanical ventilation in critically ill patients. However, given the load-capacity imbalance in critically ill patients can change quickly due to changes in respiratory mechanics or neural activity (e.g., due to changes in sedation or pain), the use of event-related potentials to detect changes in respiratory control is limited by poor time resolution and continuous connectivity approaches are preferable (Hudson et al., 2016; Navarro-Sune et al., 2016).