Oxygen Metabolism-induced Stress Response Underlies Heart–brain Interaction Governing Human Consciousness-breaking and Attention

Abstract

Neuroscientists have emphasized visceral influences on consciousness and attention, but the potential neurophysiological pathways remain under exploration. Here, we found two neurophysiological pathways of heart-brain interaction based on the relationship between oxygen-transport by red blood cells (RBCs) and consciousness/attention. To this end, we collected a dataset based on the routine physical examination, the breaking continuous flash suppression (b-CFS) paradigm, and an attention network test (ANT) in 140 immigrants under the hypoxic Tibetan environment. We combined electroencephalography and multilevel mediation analysis to investigate the relationship between RBC properties and consciousness/attention. The results showed that RBC function, via two independent neurophysiological pathways, not only triggered interoceptive re-representations in the insula and awareness connected to orienting attention but also induced an immune response corresponding to consciousness and executive control. Importantly, consciousness played a fundamental role in executive function which might be associated with the level of perceived stress. These results indicated the important role of oxygen-transport in heart-brain interactions, in which the related stress response affected consciousness and executive control. The findings provide new insights into the neurophysiological schema of heart-brain interactions.

Introduction

Many studies have demonstrated the relationship and communication between visceral signals and the brain [1, 2], in which heart–brain interactions have attracted great attention [1, 3]. Especially, the heartbeat evokes a particular neural response during the resting state [4] and the task state [5], and the heartbeat-evoked response further shapes and even predicts conscious detection and attention in humans [6, 7]. These findings suggest that the autonomic nervous system of the heart, as the center of blood circulation, influences the central nervous system as the site of information processing; however, the underlying neurophysiological pathways are still largely unclear.

Consciousness and attention are tightly related to cardiac properties. In the visual sensory domain, a weak stimulus is more likely to be consciously detected while the brain is intensely responding to heartbeats: the success of subjective conscious detection of a faint grating annulus is accompanied by heartbeat-evoked responses [6]. Correspondingly, the time in the cardiac cycle mediates the access of a weak somatosensory stimulus to consciousness [7]; specifically, the hit rate and the sensitivity are higher during diastole than systole – this is temporally accompanied by a pulse wave that is regarded as a predictive change and is inhibited by the brain. This can be seen as a result of a shift of an individual’s attention between external environmental and internal body signals. Of note, the cardiac cycle is usually recorded as an electrocardiogram which is typically composed of the P wave (atrial contraction), R peak (ventricular contraction), and T wave (ventricular repolarization) [8, 9]. The R peak [5, 10] and T wave [11, 12] define the period during which the heart contracts to pump blood up to the brain and other systems; this has been used to explore the relationship of cardiac function to consciousness and attention. As to cardiac function, the blood consists of 80% red blood cells (RBCs), in particular hemoglobin that carries oxygen from the heart to sustain brain function [13]. Thus, the oxygen-carrying capacity of RBCs might embody a tight mode of heart–brain interaction. In fact, cardiac resuscitation studies have confirmed that the recovery of heart function directly affects the level of consciousness: resuscitation allows the heart to continue to supply blood to the brain and the whole body, and to restore consciousness in patients [14, 15]. Most importantly, current studies show that RBCs are not only involved in carrying oxygen from the lungs to organs, but are also active members of the innate immune system [16, 17], which facilitates fighting off infections by capturing bacteria on the cell surface, neutralizing them, and transporting them to immune cells in the spleen and liver [18–20]. To some extent, the heart interacts dynamically with the liver and kidneys. These organs are immunological and can become significant driving factors in diseases associated with inflammation and immunity when dysregulated [21]. The innate immune system interacts with the neuropathy of the regulation of mental symptoms [22–24]. What is more, the magnitude of stress-associated immune dysregulation has implications for mental health [25]. At the same time, the stress response is associated with executive control of the attention system, including the important predictive or mediating role of perceived stress on executive function [26–28]. Thus, this immune response mode, to a large extent, determines the information processing mode of the human brain and has a direct impact on mental activity [24, 29, 30], especially consciousness [31, 32] and attention [33, 34]. Consciousness is closely linked to attention, specifically, attention is necessary for consciousness [35–37]. The existence of inattention and change-blindness have revealed that a lack of attention boosts the neglect of consciousness [38, 39]. However, others have argued that consciousness and attention are two distinct processes and that there is no causal relationship between them [40, 41]. Experiments on attention without consciousness using continuous flash suppression (CFS) and masked priming paradigms have demonstrated that invisible stimuli are able to direct the allocation of attention [42, 43]. Besides, researches on consciousness without attention have shown that rapid visual presentation tasks require very little attention [44]. It is worth noting that the attention mentioned above primarily refers to selective and endogenous attention. Attention is a multifaceted mental process that can be divided into three subsystems, alerting, orienting, and executive control [45]. Therefore, the relationship between different types of attention and consciousness deserves further exploration.

In the present study, we emphasized the relationship between the oxygen-transport capacity and consciousness/attention from the aspect of heart–brain interactions. We hypothesized that an individual's consciousness-breaking and attentional performance depend on the brain's response mode, which is affected by the ability of human RBCs to transport oxygen during heart–brain interactions. To test this hypothesis, using mathematical models, we explore the relationship between cardiac oxygen-transport capacity and consciousness/attention across individuals with low oxygen saturation in the hypoxic Tibetan environment.

Materials and Methods

Participants

We recruited 140 Han Chinese college students who were born and grew up in the plains area of central China and had been living at Tibet University for 3 years. Among them, 4 participants were excluded from the b-CFS behavioral experiment due to the low correct rate (approximately 50%); 6 were excluded from the EEG analysis, 3 because of excessive eye-movement artifacts, head motion, and muscle artifacts that contaminated > 50% of the trials and the other 3 because of the shedding of reference electrode (CPz), resulting in poor and unstable data quality; and 4 participants quit the physical examination due to blood phobia. Data from the remaining 126 participants (70 females) with an average age of 22 ± 1.27 years were fully analyzed. All participants were healthy without neurological or psychiatric disorders, brain injury, or drug/alcohol addiction. All participants gave written informed consent and were paid for their participation. The study was approved by the local Ethics Committee of Tibet University and followed the Declaration of Helsinki.

Experimental Procedures

At the beginning of the experiments, participants completed the SCL-90 [46] and the b-CFS behavioral experiment, and finally took part in the ANT experiment with EEG recording. Then, the participants were given a structural magnetic resonance imaging (MRI) scan and physical examination on the next day at Tibet General Hospital of the Chinese People's Armed Police Force.

Health Status Measures

The health status was measured using SCL-90, which is designed to evaluate a broad range of psychological problems and symptoms of psychopathology. It consists of 90 items to evaluate 9 scores along the primary symptom dimensions, i.e., somatization, obsessive-compulsive, interpersonal sensitivity, depression, anxiety, hostility, phobic anxiety, paranoid ideation, psychoticism, and a category of additional items that helps clinicians assess other aspects of symptoms. Each of the dimensions reflects the extent of a certain aspect of the patient's symptoms. Assessment is divided into 5 grades (from 0–4), 0 = never, 1 = mild, 2 = moderate, 3 = considerable, 4 = severe. The General Symptom Index is defined as the total score divided by the item number (n = 90); this is also the average score.

Tasks and Stimuli

The b-CFS Paradigm

Stimuli were generated with MatLab R2013a (The MathWorks, Inc., Natick, USA) and presented on a 19-inch Mitsubishi Diamond Pro monitor (1280 pixels × 1024 pixels at 100 Hz) using Psychotoolbox (http://psychtoolbox.org/). The images presented to the two eyes were displayed side by side on the monitor and were fused using a mirror stereoscope mounted on a chin rest. A frame (10.71 cm × 10.71 cm) that extended beyond the outer border of the stimulus and fixation point was presented to facilitate the stable convergence of the two images. The viewing distance was 75 cm. A central cross was always presented to each eye, serving as the fixation point. Briefly, at the beginning of each trial, a standard dynamic noise pattern was presented simultaneously to the left and right side at full contrast for 1 s–3 s, and then the test figure (grating with different tilt angles ranging from 70° to 110° at a step size of 0.5°) was presented to the other eye at a random location within the region corresponding to the location of the noise. The contrast of the test figure was ramped up gradually from 0 to 100% in 1 s and then remained constant until the observer made a response to indicate the side on which the test figure tilted or until 10 s passed without a response. The interval between trials was 2 s. In the experiment, the test images were gratings with different tilt angles, ranging from 70° to 110° at a step size of 0.5°. At the beginning of each trial, participants perceived the noise patch and were unaware of the side containing the test image. The reaction times (RTs) of consciousness-breaking were measured from when the test image overcame the suppression noise to become dominant. Then participants were instructed to press the left or right arrow key on a standard keyboard to indicate the tilt orientation of the test image. They were required to respond to the appearance of any part of the test image as soon as possible; they did not need to know the specific content of the image (Fig. 1A).

Fig. 1.

Schematic representation of the experimental paradigm. A The b-CFS paradigm. A standard dynamic noise pattern is presented simultaneously to the left and right side at full contrast for 1 s–3 s, and then the test figure is presented to the other eye at a random location within the region corresponding to the location of the noise. The contrast of the test figure is ramped up gradually from 0 to 100% in 1 s and then remains constant until the participant makes a response to indicate the side on which the test figure is tilted. B The ANT paradigm. A trial begins with a 1000 ms fixation period, then a warning cue is presented for 200 ms. There is a short fixation period for a randomly-variable duration after the warning cue and then the target and flankers appear simultaneously. The target and flankers are presented until the participant responds, but for no longer than 2000 ms.

ANT Paradigm

The ANT [47], combined with the spatial-cue task and the flanker task [48], can effectively measure different attentional network effects (i.e., alerting, orienting, and executive); it is a quantitative evaluation based on RTs. The version in the present study was identical to that in a previous study [47]. The target stimuli were presented via E-Prime, consisting of a row of 5 black leftward or rightward arrowheads either above or below the fixation cross, against a gray background. In all target stimuli, the horizontally centered arrow was flanked on either side by 2 arrows in the congruent or incongruent direction. The target stimuli were preceded by one of three cue conditions (i.e., no-cue, center-cue, and spatial-cue). In the no-cue condition, no asterisk was presented and only fixation for 200 ms was required. In the center-cue condition, an asterisk appeared in the fixation cross and the time course was the same as that in the no-cue condition. In spatial-cue trials, an asterisk was presented at the target position. Each trial began with 1000 ms fixation, then a warning cue was presented for 200 ms. There was a short fixation period for a randomly variable duration (300 ms–1098 ms) after the warning cue and then the target and flankers appeared simultaneously. The target and flankers were presented until the participant responded, but for no longer than 2000 ms. After participants made a response, the target and flankers disappeared immediately. After 1000 ms interval, the next trial began. The fixation cross remained at the center of the screen during the whole trial (Fig. 1B). RTs were recorded and the difference in RTs between the different conditions was used to calculate a performance score for each of the three components of attention [47].

Behavioral Data Analysis

The b-CFS task performance was assessed by computing the average RT and accuracy (i.e., correct responses divided by total responses). For ANT, we calculated the attentional network scores (i.e., alerting, orienting, and executive) using the RTs under different experimental conditions. In detail, the alerting score was calculated as the mean RT for all no-cue conditions minus all center-cue conditions. The orienting score was the mean RT for the center-cue conditions minus the spatial-cue conditions. Finally, the executive score was measured as the mean RT of all incongruent conditions minus that of the congruent conditions. Incorrect responses were not included in the RT analysis, and the extreme RTs of trials within each condition (mean ± 3 SD) were removed for the calculation in each participant.

MRI Acquisition

All participants were scanned in a 3 T MRI scanner (MAGNETOM Spectra, Siemens, Germany) at Tibet General Hospital of the Chinese People's Armed Police Force. We acquired 3D structural images using a T1-weighted MP-RAGE sequence: 192 sagittal slices; TR = 3200 ms; TE = 419 ms; inversion time = 600 ms; slice thickness = 1 mm; no gap; FA = 120°; FOV = 256 mm × 256 mm.

EEG Recording and Pre-processing

The EEG signals were recorded continuously with a 64‐channel ANT Neuro system (Eemagine Medical Imaging Solutions GmbH, Enschede, Netherlands). One channel was used for vertical electrooculographic (EOG) recording by placing an electrode below the left eye, while all others were used for scalp EEG recording. During recordings, the EEG was referenced to the CPz electrode, grounded at FCz, and sampled at 500 Hz. The impedance of all electrodes was kept below 5 kΩ. The data were pre-processed using MatLab R2013a and the EEGLAB toolbox [49]. Signals were bandpass filtered between 0.1 and 40 Hz. The data were re-referenced offline to the average of the two mastoids. For stimulus-locked analyses, the critical epochs ranged from − 500 to 1000 ms relative to the onset of the stimulus, with − 500 to 0 ms serving as the baseline. In each epoch, visual inspection was carefully conducted to reject segments containing non-physiological artifacts. After that, independent component analysis was performed to remove possible eye blink, movement, and heartbeat artifacts. Furthermore, signals exceeding an amplitude threshold of ±100 μV were marked for rejection.

Time-Domain Analysis of EEG Data

The EEG data analyses were performed using MatLab R2013a and the EEGLAB toolbox [49]. Amplitudes of the P1, N1, P2, and N2 waves under cue conditions and P3 waves under target conditions were assessed by averaging single-trial amplitude values across time windows. The P1, N1, and P2 are exogenous components that are affected by the physical properties of stimuli [50]. P1 reflects inhibition in the sense of task-irrelevant processes [51]; the visual N1 component is involved in a process of discriminating the attended location [52]; and the N2 and P3, later ERP components, are mainly associated with response inhibition [53] but are also regarded as an index of processing capacity and mental workload [54]. To calculate the P1, N1, P2, and N2, data were averaged across time windows of 80 ms–150 ms, 150 ms–200 ms, 200 ms–250 ms, and 250 ms–300 ms at electrodes O1, Oz, O2, P3, Pz, and P4. To quantify P3, the mean amplitude across the time window 350 ms–700 ms was calculated at electrodes CFz, Cz, and Pz. The alerting effect was calculated from the EEG amplitude for all no-cue trials minus that for all center-cue trials. The orienting effect was calculated from the amplitudes of all center-cue and spatial-cue trials. The magnitude of the executive effect was measured by the amplitudes of all incongruent and congruent trials. The attentional subsystems effects based on amplitudes were computed using the same formulae as for the RTs.

Time-Frequency Domain Analysis of EEG Data

In order to distinguish the energy and oscillations of the attention system, we performed time-frequency analysis. At the sensor level, data were analyzed using the Fieldtrip toolbox [55]. Time-frequency representations of power were estimated for frequencies between 2 and 40 Hz using a sliding time window from − 200 to 800 ms in steps of a 20-ms fixed window with a Hanning taper. The frequency bands were partitioned as follows: delta (1 Hz–4 Hz), theta (4 Hz–8 Hz), alpha (8 Hz–13 Hz), beta (13 Hz–30 Hz), and gamma (30 Hz–40 Hz). Given the likelihood that the higher frequencies still contained muscle artifacts even after data pre-processing, we limited the maximum gamma-band frequency to 40 Hz. After acquiring the time-frequency representations of single trials, we averaged the power estimates over trials for each participant and each condition. The baseline for response-locked time-frequency analysis was the same as the setting in response-locked ERP analysis.

Source Localization Analysis

To localize the underlying source activity, we applied dynamic imaging of coherent sources (DICS) beamforming approaches [56]. We constructed a boundary element method (BEM) volume conduction model of the head (head model) based on each participant’s MRI [57]. Then the brain volume was discretized into a grid in which the lead field matrix was calculated for a grid at 1-cm resolution. DICS spatial filters were constructed from the lead field and a cross-spectral density matrix to maximize the activity of interest at each specific grid point while suppressing the contribution of all other grid points. Spatial filters were also applied to Fourier-transformed sensor level data to estimate the power level at the source.

Multilevel Mediation Analysis

To investigate how the erythrocyte affects psychological health, we performed multilevel mediation analyses implemented in the Mplus Toolbox (https://www.statmodel.com/). Three-path mediation analysis indicates whether two mediators intervene in series between an independent and a dependent variable [58]. In the three-path mediation analysis models, the erythrocyte was the independent variable (X), and the participants' health status (SCL-90) was the dependent variable (Y). Renal function (creatinine) and information-processing mode (amplitude of the orienting function) were the mediator (M). Mediated effects were estimated by the product of the coefficients for each of the paths in the mediational chain. The three-path mediated effect involves passage through both mediators, and the two-path mediated effect is passage through only one of the mediators. The total mediated effect of X on Y is the sum of the three-path and the two-path mediated effects. The chained multilevel implementation (more than two mediators) is largely identical to the three-path mediation analysis. Significant mediation was assumed when each of the four relevant paths [erythrocyte index to white blood cell (WBC), WBC to b-CFS and executive function, and executive function to SCL-90] was significant by itself and the product of the respective path coefficients was significantly different from zero. The multilevel mediation model was generated with 1000 percentile and bias-corrected bootstrapped samples adjusted for heteroskedastic standard errors, and predictors were standardized for analysis. Indirect effects were significant if the 95% percentile or bias-corrected bootstrapped confidence intervals (CIs) for indirect effects did not include zero [59]. Model fit was assessed using χ² divided by degrees of freedom < 3.0, root mean square error of approximation < 0.08, comparative fit index and Tucker–Lewis index > 0.9, standardized root mean square residual < 0.08, and models that satisfied all thresholds were considered to provide an acceptable fit [60].

Statistical Analysis

The first stage was descriptive statistics for the physiological data to describe and summarize the basic situation of recorded variables. Next, Pearson's correlation was applied to estimate the relationships between behavior, ERP components, physiology, and mental states. Based on the descriptive statistics, most indicators were normal, except for unconjugated bilirubin (UBIL), uric acid (UA), lymphocyte ratio (LYMPH%), neutrophil ratio (NEUT%), mean corpuscular hemoglobin (MCH), and thrombocytocrit (PCT). To investigate whether blood indexes affected attentional function, we calculated two-way repeated-measures analyses of variance (ANOVAs) per condition with physiological indicators as a factor (two levels: low and high) and the attention function (amplitudes of the alerting effect) as dependent variables. Significant interaction effects (P < 0.05, two-sided) were followed up by post hoc Bonferroni-corrected dependent samples t-tests. Then, to further assess the causal relationships between them, multilevel mediation analyses were used.

Results

Descriptive Statistics of Biochemical Indexes and Relationships Between Neurophysiological Components and Behaviors

First, descriptive statistics were calculated for the physical examination data. Most indicators of the biochemical characteristics (i.e., routine blood tests, liver function, and renal function) of the participants were normal, except for UBIL, UA, LYMPH%, NEUT%, MCH, and PCT (Table 1).

Table 1.

Biochemical characteristics of study participants

Biochemical indexes	High group		Normal group		Low group
	Male	Female	Male	Female	Male	Female
Liver functions
ALT	4	2	48	61	4	7
TP	0	0	56	70	0	0
ALB	0	0	57	68	0	2
GLB	0	0	48	68	8	2
A/G	6	1	50	69	0	0
TBIL	8	4	48	66	0	0
DBIL	4	0	52	70	0	0
UBIL	34	26	22	44	0	0
Renal functions
Urea	2	0	54	68	0	2
CREA	1	0	50	61	5	9
UA	25	27	31	43	0	0
Blood routine
WBC	2	1	54	68	0	1
NEUT#	2	1	48	61	6	8
NEUT%	2	1	26	32	28	37
LYMPH#	1	1	55	69	0	0
LYMPH%	26	37	29	33	1	0
MONO#	0	0	56	70	0	0
MONO%	8	8	48	62	0	0
EO	1	0	55	70	0	0
EO%	2	1	50	68	4	1
BASO	0	0	56	70	0	0
BASO%	1	0	55	70	0	0
RBC	10	1	46	69	0	0
HGB	8	1	48	64	8	5
HCT	0	1	56	64	0	5
MCV	0	0	56	62	0	8
MCH	33	31	23	32	0	7
MCHC	13	6	43	59	0	5
RDW-SD	0	1	56	69	0	0
RDW-CV	0	6	56	64	0	0
PLT	7	13	49	55	0	2
PCT	15	28	41	42	0	0
MPV	0	0	56	69	0	1
PDW	3	0	53	69	0	1
P-LCR	0	0	56	70	0	0

ALT: alanine transaminase; TP: total protein; ALB: albumin; GLB: globulin; A/G: albumin/globulin; TBIL: total bilirubin; DBIL: direct bilirubin; UBIL: unconjugated bilirubin; CREA: creatinine; UA: uric acid; WBC: white blood cell count; NEUT#: neutrophil count; NEUT%: neutrophil ratio; LYMPH#: lymphocyte count; LYMPH%: lymphocyte ratio; MONO#: monocyte count; MONO%: monocyte ratio; EO: eosinophil count; EO%: eosinophil ratio; BASO: basophil count; BASO%: basophil ratio; RBC: red blood cell count; HGB: hemoglobin; HCT: hematocrit; MCV: mean corpuscular volume; MCH: mean corpuscular hemoglobin; MCHC: mean corpuscular hemoglobin concentration; RDW-SD: standard deviation in red cell distribution width; RDW-CV: coefficient variation of red cell distribution width; PLT: platelet; PCT: thrombocytocrit; MPV: mean platelet volume; PDW: platelet distribution width; P-LCR: platelet larger cell ratio; the numbers in boldface represent the number of subjects with higher or lower indicator.

We calculated the Pearson's correlation coefficient to estimate the relationship between attentional functions and consciousness-breaking. The results showed a positive correlation between the RT of b-CFS and the executive score (Fig. 2A) (Pearson’s correlation: r = 0.258; P = 0.0036), which was not found for the alerting or orienting scores (Pearson’s correlation: r = − 0.023; P = 0.8; r = − 0.009; P = 0.923). Meanwhile, there were no statistical correlations among the three attentional scores (alerting and orienting: r = − 0.116; P = 0.196; alerting and executive: r = 0.131; P = 0.143; orienting and executive: r = − 0.08; P = 0.37). The mean RT and the mean accuracy under experimental conditions are shown in Table 2.

Fig. 2.

Correlation between the attention network scores and b-CFS RTs. A Executive function is significantly correlated with individual b-CFS RTs. B A significant executive score is correlated with the amplitude of the executive effect. C The orienting score is significantly positively correlated with its amplitude. Regression lines and 95% CI are shown. D, E Grand average ERPs stratified by center-cue (solid line) and no-cue (dashed line) conditions shown as interpolation of electrodes O1, Oz, O2, P3, Pz, and P4. Subtracting the two conditions in D gives the alerting effect and in E gives the orienting effect. F Grand average target ERPs stratified by target conditions at midline electrodes FCz, Cz, and Pz. Parietal P3 displays a significant modulation of amplitude as a function of the flanker incongruency effect (executive effect). b-CFS, breaking continuous flash suppression; RTs, reaction times; CI, confidence interval; ERP, event-related potential.

Table 2.

Mean RT (ms) and accuracy (ACC, %) of behavioral performance

Behavioral performance	b-CFS	No-cue	Center-cue	Spatial-cue	Congruent	Incongruent
RT Mean ± SD	1499 ± 449	600 ± 66.5	591 ± 66.5	533 ± 67.7	531 ± 61	620 ± 73
ACC Mean ± SD	95.7 ± 3.7	97.8 ± 2.2	97.3 ± 2.7	98.6 ± 1.8	99.4 ± 1.1	96.4 ± 3.4

RT, reaction time; ACC, accuracy.

Further analysis of the relationships between behavioral and ERP components found a negative correlation between the executive score and the P3 amplitude of executive function (Pearson’s correlation: r = − 0.25; P = 0.005) (Fig. 2B). The orienting score showed a significantly positive correlation to the P2 amplitude of the orienting function in the 200 ms–300 ms time window (Pearson’s correlation: r = 0.206; P = 0.021) (Fig. 2C). Meanwhile, the alerting score showed no significant correlation to the amplitude of the alerting function (Pearson’s correlation: r = − 0.042; P = 0.642).

Attention Functions as a Compensatory Response to a Physiological State

Based on the descriptive statistics, we investigated whether the erythrocyte index influenced attentional functions and consciousness-breaking (b-CFS). Two-way repeated-measures ANOVAs were calculated. At the time-frequency level, the results showed no significant main effect of MCH and LYMPH% [F_{(2, 123)} = 0.015, P = 0.901; F_{(2, 123)} = 2.043, P = 0.156], but a significant interaction effect between MCH and LYMPH% in the delta power (200 ms–300 ms) of alerting [F_{(2, 123)} = 4.11, P = 0.045]. Post hoc testing revealed that the delta power in participants with high LYMPH% was significantly greater than in those with normal LYMPH% and normal MCH (P = 0.02, partial eta = 0.047), whereas no significant difference was found in the delta power (200 ms–300 ms) in participants with normal or high LYMPH% and high MCH (Fig. 3A).

Fig. 3.

The alerting and orienting responses to erythrocyte function and the immune system. A Delta power (200 ms–300 ms) in the high LYMPH% group is greater than in the normal LYMPH% group in the normal MCH group under the alerting condition. B A significant main effect of MCH and a significant interaction effect in orienting function at the time-frequency domain level. Theta power of orienting (250 ms–300 ms) is higher in the normal than in the high MCH group; is significantly higher in the normal LYMPH% group than in the high LYMPH% group with high MCH, and is significantly higher in the normal MCH group than in the high MCH group with high LYMPH%. C The P1 amplitude (80 ms–150 ms) in the high LYMPH% group is greater in the normal LYMPH% group than in the normal MCH group in the alerting condition. D A significant interaction effect of orienting function at the ERP level. The P1 amplitude (80 ms–150 ms) in the normal LYMPH% group is significantly greater than in the high LYMPH% group in the normal MCH group; while in the normal MCH group, it is significantly greater than in the high MCH group in the normal LYMPH% group. E The relationship between physiological functions, cognitive functions, and mental states. The sizes of the circles above the diagonal represent the relative size, and the colors represent the relative direction (blue, positive; red, negative). The numbers below the diagonal are Pearson's correlation coefficients. MCH, mean corpuscular hemoglobin; LYMPH%, lymphocyte ratio; N, normal; H, high.

A significant main effect of MCH in the theta power of orienting (250 ms–300 ms) was found [F_{(2, 123)} = 3.954, P = 0.049], leading to higher power in the normal MCH than in the high MCH group (Fig. 3B). Moreover, there was a significant interaction effect [F_{(2, 123)} = 6.651, P = 0.011]. Further post hoc comparisons showed that the theta power of orienting (250 ms–300 ms) in the normal LYMPH% group was significantly greater than that in the high LYMPH% group with high MCH (P = 0.012, partial eta = 0.055), and was significantly greater in the normal MCH participants than that in the high MCH participants with high LYMPH% (P = 0.002, partial eta = 0.083).

At the ERP level of the alerting effect, the results showed a significant interaction between MCH and LYMPH% in the P1 amplitude (80 ms–150 ms) of alerting [F_{(2, 123)} = 5.444, P = 0.021]. Neither the main effect of MCH [F_{(2, 123)} = 0.014, P = 0.905] nor that of LYMPH% [F_{(2, 123)} = 1.420, P = 0.236] was significant. Post hoc pair-wise comparisons confirmed that participants with a high LYMPH% index had a larger P1 amplitude of alerting than participants with a normal LYMPH% index in the normal MCH group (P = 0.017, partial eta = 0.049), while no significant difference was found in the amplitude of alerting between the normal or the high LYMPH% group with high MCH (P = 0.405, partial eta = 0.006) (Fig. 3C). As for the orienting function, the results showed a significant interaction effect between MCH and LYMPH% in the P1 amplitude (80 ms–150 ms) of orienting [F_{(2, 123)} = 4.027, P = 0.047]. Post hoc testing revealed an increased P1 amplitude of orienting in the normal group compared to that in the high LYMPH% group with normal MCH (P = 0.026, partial eta = 0.043), while no significant difference was found in the amplitude of alerting between the normal and the high LYMPH% groups with high MCH (P = 0.405, partial eta = 0.006). And the amplitude of orienting in the normal group was significantly greater than that in the high MCH group with normal LYMPH% (P = 0.028, partial eta = 0.04), while there was no significant difference in the amplitude of orienting between the normal and the high MCH groups with high LYMPH% (P = 0.534, partial eta = 0.003) (Fig. 3D).

Two Pathways of Erythrocyte Functional Effect on Consciousness-Breaking and Attention

Before structural equation analysis, Pearson's correlation coefficients were calculated to estimate the relationship between physiological functions (i.e., routine blood tests, liver function, and renal function), cognitive functions (i.e., consciousness and attention), and mental states (i.e., symptoms of psychopathologyms) and accuracy (ACC, %) of behavioral performance). The relations of multiple sets of data are detailed in Fig. 3E. We used mediation analysis to assess whether physiological functions and cognitive functions were involved in the effect of the routine blood index on individual mental states (Fig. 4). Three-path mediation analysis indicated mediation effects of UREA and orientating function on participants’ physical and mental state (indirect effect = 0.037, bootstrapped 95% CI = [0.003, 0.091], bias-corrected bootstrapped 95% CI = [0.004, 0.098]). Further multilevel mediation analysis showed that WBC, consciousness-breaking, and executive control score each partially mediated the relationship between erythrocytes and mental symptoms (indirect effect = 0.012, bootstrapped 95% CI = [0.000, 0.008], bias-corrected bootstrapped 95% CI = [0.004, 0.098]). Table 3 shows the path coefficients. The model fit is described in Fig. 4. At the same time, given that the general symptom index of SCL-90 consists of 10 symptom dimensions, we also analyzed the path based on the sub-items to investigate which dimensions were specifically influenced. However, the result showed no significant mediating effect. The path coefficients are list in Table 4.

Fig. 4.

Two neurophysiological pathways by which erythrocyte function affects mental state. Path diagrams and statistics for three-path mediation analyses between erythrocyte index, CREA of renal function, and the amplitudes of orienting function and psychological state (above); multi-level mediation analyses of the influence of erythrocyte index on the psychological state through the immune system of WBC, the RT of b-CFS, and executive score (below). Path coefficients are listed for each path with standard errors in parentheses. *P < 0.05, **P < 0.01, ***P < 0.001. CREA, creatinine; WBC, white blood cell; b-CFS, breaking continuous flash suppression; RT, reaction time; RMSEA, root mean square error of approximation; CFI, comparative fit index; TLI, Tucker–Lewis index; SRMR, standardized root mean square residual.

Table 3.

Path analyses results

Point estimate		SE	Est/SE	P value	Bootstrap 1000 times 95% CI
					Percentile		Bias corrected
					Lower	Upper	Lower	Upper
a1	0.618	0.045	13.624	0.000	0.529	0.703	0.512	0.695
d1	− 0.249	0.117	− 2.131	0.033	− 0.485	− 0.025	− 0.486	− 0.026
b1	− 0.242	0.077	− 3.165	0.002	− 0.398	− 0.096	− 0.382	− 0.080
c1	− 0.110	0.094	− 1.169	0.242	− 0.276	0.087	− 0.292	0.083
a2	0.157	0.048	3.268	0.001	0.083	0.28	0.068	0.243
d2	0.243	0.091	2.675	0.007	0.068	0.408	0.090	0.435
e2	0.264	0.105	2.522	0.012	0.047	0.453	0.066	0.473
b2	0.213	0.093	2.288	0.022	0.039	0.392	0.044	0.401
c2	− 0.097	0.082	− 1.185	0.232	− 0.253	0.015	− 0.251	0.068

CI, confidence interval; Est, estimate; values in boldface are significant.

Table 4.

Path coefficients of sub-items

Variable	Point estimate	SE	Est/SE	P value	Bootstrap 1000 times 95% CI
					Percentile		Bias corrected
					Lower	Upper	Lower	Upper
Ery → Crea → Ori → SOM	− 0.046	0.091	− 0.506	0.613	− 0.053	0.053	− 0.049	0.055
Ery → Crea → Ori → OC	− 0.028	0.070	− 0.402	0.688	− 0.166	0.099	− 0.163	0.086
Ery → Crea → Ori → IS	− 0.007	0.068	− 0.095	0.924	− 0.146	0.123	− 0.145	0.123
Ery → Crea → Ori → Dep	0.044	0.058	0.747	0.455	− 0.080	0.155	− 0.076	0.163
Ery → Crea → Ori → Anx	0.005	0.069	0.069	0.945	− 0.15	0.128	− 0.133	0.133
Ery → Crea → Ori → Hos	0.053	0.066	0.812	0.417	− 0.069	0.186	− 0.066	0.186
Ery → Crea → Ori → PA	− 0.060	0.072	− 0.836	0.403	− 0.210	0.084	− 0.197	0.062
Ery → Crea → Ori → PI	0.027	0.065	0.412	0.680	− 0.112	0.129	− 0.107	0.153
Ery → Crea → Ori → Psy	0.043	0.069	0.622	0.534	− 0.112	0.166	− 0.107	0.168
Ery → WBC → CB → Exc → SOM	− 0.001	0.026	− 0.031	0.975	− 0.053	0.053	− 0.049	0.055
Ery → WBC → CB → Exc → OC	− 0.013	0.027	− 0.475	0.635	− 0.061	0.047	− 0.061	0.046
Ery → WBC → CB → Exc → IS	0.003	0.021	0.155	0.877	− 0.038	0.045	− 0.034	0.047
Ery → WBC → CB → Exc → Dep	− 0.001	0.023	− 0.060	0.952	− 0.043	0.052	− 0.044	0.051
Ery → WBC → CB → Exc → Anx	0.015	0.029	0.539	0.590	− 0.045	0.069	− 0.043	0.071
Ery → WBC → CB → Exc → Hos	0.005	0.021	0.257	0.797	− 0.037	0.047	− 0.036	0.048
Ery → WBC → CB → Exc → Pa	0.005	0.021	0.257	0.797	− 0.037	0.047	− 0.036	0.048
Ery → WBC → CB → Exc → Pi	0.013	0.023	0.560	0.576	− 0.030	0.065	− 0.029	0.065
Ery → WBC → CB → Exc → Psy	0.014	0.028	0.500	0.617	− 0.041	0.064	− 0.041	0.064

Ery, erythrocyte; Crea, creatinine; Ori, orienting function; SOM, somatization; OC, obsessive-compulsive; IS, interpersonal sensitivity; Dep, depression; Anx, anxiety; Hos, hostility; PA, phobic anxiety; PI, paranoid ideation; Psy, psychoticism; WBC, white blood cell; CB, conscious breaking; Exc, executive function.

Sources Responsible for Attention Network Oscillatory Activity

We finally uncovered substantive issues and mechanisms of attention in the time-frequency domain and source reconstruction. The orienting effect evoked increased neuronal oscillations at frequencies below 8 Hz between 0 and 300 ms, whereas it increased the suppression of neuronal oscillations at alpha frequencies (8 Hz–13 Hz) and latencies between 400 ms–700 ms (Fig. 5A). Source analysis revealed that the strong positive power at low frequencies (< 8 Hz) was localized to the right insula (Fig. 5B) and the suppression of oscillations was localized to the inferior occipital gyrus. For executive function, there were positive neuronal oscillations in the theta (4 Hz–8 Hz) band during 300 ms–600 ms (Fig. 5C). Further reconstructed source analysis showed elevated power in the prefrontal cortex of the left superior frontal gyrus (Fig. 5D). Besides, time-frequency analysis confirmed that the alerting effect evoked suppression of neuronal oscillations at alpha and low beta frequencies (8 Hz–20 Hz) at latencies between 400 and 700 ms (Fig. 5E). Source reconstruction analysis revealed that the power was localized to the left middle occipital gyrus (Fig. 5F).

Fig. 5.

Distinct patterns of oscillatory rhythms and source activity of the three attentional networks. A, B The neural source of the orienting function at frequencies below 8 Hz between 0 and 300 ms is localized to the right insula (peak MNI coordinates: [46 13 − 3]). C, D For executive function, the neural source of theta (4 Hz–8 Hz) during 300 ms–600 ms is localized to the left prefrontal cortex of the superior frontal gyrus (peak MNI coordinates: [− 23 37 40]). E, F The left middle occipital gyrus is the neural source of alerting effect-evoked suppression of neuronal oscillations at alpha and low beta frequencies (peak MNI coordinates: [− 29 − 73 39]).

Discussion

For a long time, the physiological origin of psychological phenomena has been controversial [61–63]. Focusing on the important role of oxygen metabolism in consciousness and attention, in the present study, we used the structural equation method to construct the neurophysiological pathways between the oxygen supply capacity for physiological activities and the mental states of consciousness and attention. We provide experimental evidence that the properties of RBCs, as a source of the visceral signal, could affect mental states through two pathways: the unconscious visceral sensory information integration pathway and the conscious sensory integration pathway. The two pathways joined in the attentional system, and the stress response played a key role in the conscious awareness pathway. Interestingly, the neurophysiological pathways were more sensitive to the general health status which reflected the severity of individuals’ health status rather than the specific sub-dimensions. The present study provides new insights into the visceral impact on mental states.

We found two independent neurophysiological pathways corresponding to different attentional systems underlying the heart–brain interaction. One was the unconscious visceral sensory information integration pathway, in which the relationship between RBC properties and mental states was mediated by creatinine and orienting attention. Orienting to upcoming stimuli involved an increase of beta-band oscillations in the insular cortex, which contains one of the central representations of the visceral organs [64] and plays a fundamental role in human conscious awareness [65]. Traditionally, the insular cortex is a key node of the “salience network” which is thought to be involved in the detection of relevant stimuli and the intuitive processing of complex situations to identify salient stimuli [66–68]. Notably, activity in the insular cortex has been reported in most studies of the stimulus-triggered orienting of attention [69], mediating interactions between externally-oriented and internally-oriented attention. For instance, some studies have shown increased activation in the insula in a selective-attention task which is considered to be related to reorienting attention [69, 70]. In a way, the insula plays a crucial part in regulating the relationship between alerting and orienting. On the other hand, the insula has been regarded as the primary visceral cortex [71] and responds to the cardiovascular system [72]. It is important to note that the view of the cortical representation of the visceral signal or heart in the insula enriches the conventional opinion that regulation of the heart is restricted to stereotyped responses to external stimuli [1, 3, 73]. These results also support the hypothesis that the targets of visceral inputs include a cortical target (insula), and RBCs are a source of visceral information able to influence a subject's mental experience, which further verified the essence of the James–Lange theory that mental feelings originate in visceral or bodily physiological variations [63, 74]; more precisely, mental states seem to be the sum of the collective sensations caused by changes in internal/visceral organs.

The other is the conscious sensory integration pathway, in which the immune system, the consciousness-breaking, and the executive function are involved in the tight relation between RBCs and mental states. Our results showed that healthy RBCs mediated the immune system, a tightly regulated network that is able to maintain physiological immunological homeostasis [75]. The physiological homeostasis further affects consciousness with psychological homeostasis as the main driving force [76], which is important for the organization of attentional function. In the past, neuroscience claimed that the brain was nothing but an information-processing machine that for the most part could be understood in isolation from the physiological state of the body [77]. However, we should note that interoception is always present in the perception and integration of signals within the body [78], including autonomic, visceral, and immune functions. That is, physiology and mental state are dynamically coupled, and the central nervous system and the autonomic nervous system interact. In an embodied cognitive framework, cognitive processes are deeply rooted in the body’s interactions with the world [79].

There has always been controversy about the relationship between attention and consciousness. Previous researchers have argued that an understanding of consciousness is based on the mechanisms of attention [80]. Nevertheless, some cognitive scientists dispute that attention and consciousness must be distinct processes [40, 41]. Our findings demonstrated that attentional function is dependent on consciousness, and the essence of consciousness forms the basis for an understanding of attentional function. Consequently, the essence of consciousness could be the ability to integrate higher-order cognitive functions, such as attention and emotion, with lower-order body functions, such as the immune system, to form a complete experience.

The stress response plays an important role in the physiological origin of mental states. Executive function has been counted as a promoter of an individual’s plasticity in response to stress and is closely relevant to fewer mental complaints that are likely to weaken the sense of perceived stress based on the degree of stressful, unpredictable, and uncontrollable events in daily life [81, 82]. Besides, for executive function, the neural source of theta (4 Hz–8 Hz) during 300 ms–600 ms was localized to the left superior frontal gyrus, consistent with the perspective that a higher level of the perception of stressor severity is directly relevant to declining prefrontal cortex volume and greater fractional amplitudes of low-frequency fluctuations in the left superior frontal gyrus, both of which are crucial regions for executive control [83, 84]. Beyond that, our results from ANOVA analysis indicated that the immune system interacted with RBCs regulated an individual's cognitive processes, i.e., orienting, and alerting. It is generally accepted that human behavior and mental health, to a great extent, depend on the interaction and dynamic balance between the organism and external environmental stimuli. Meanwhile, the stress response is a protective response to adverse environmental factors that alter equilibrium/homeostasis. Accordingly, in the present study, we demonstrated that the body and the environment together constitute a cognitive system, and mental states may be rooted in the interaction between the physiological homeostasis of the immune system and the psychological homeostasis of consciousness.

The main contribution of the present study is that we demonstrated the neurophysiological pathways connecting physiological indices and mental states through consciousness and attentional performance. RBCs first interacted dynamically with the immune system and renal function to maintain physiological homeostasis, which further together affected individual’s orienting, consciousness, and executive function to maintain psychological homeostasis, to form two neurophysiological pathways dominated by unconscious visceral integration and conscious sensory integration. Furthermore, the orienting function triggered in the insular cortex was not only regarded as a visceral signal but also played a crucial part in consciousness in the unconscious visceral integration path, and the executive control function related to perceived stress was based on consciousness-breaking (b-CFS) in the conscious sensory integration path, providing new insights into the relationship between consciousness and attention. Finally, in the conscious sensory integration path, the stress response mode played a major role in homeostasis. These findings may have implications for promoting heart–brain communication and weakening the sense of perceived stress by balancing the physiological homeostasis and psychological homeostasis and emphasize the basic processes in shaping our experience of the world.

There are several considerations for interpreting the present findings. First, the most frequent index adopted by the other studies of heart–brain interactions was heartbeat rhythm, and these studies harvested several exciting findings [5–7, 10, 11]. In the present study that used RBCs as the physical indicator for the heart, we considered that the pumping ability of the heart is essential to not only activity and oxygen circulation, but also cognitive processes that are sensitive to oxygen saturation in the environment. Therefore, further research is needed on higher-order cognitive processes that depend on cardiac output. Second, although we did find two neurophysiological pathways of heart-brain interaction regulating individuals’ psychological states, we mainly used correlation analysis, consequently further studies are needed to confirm the causal relationship of heart–brain interaction so that more effective interventions, such as aerobic exercise [85], can be put into effect.

Together, the present study investigated how the two neurophysiological pathways of the impact of functional attributes of RBCs, serving as a source of visceral information, affect participants’ consciousness-breaking and attention. Moreover, consciousness-breaking was more basic to attention, specifically executive control, as an indicator of perceived stress which plays a crucial part in the physiological origin of mental states. Therefore, the present study, using multilevel mediation analysis, provides valuable and novel insights into the processing of oxygen transport and the relation of the stress response to consciousness-breaking and attention from the aspect of heart–brain communication.

Conflict of interest

The authors have no conflicts of interest to declare that are relevant to the content of this article.