Pupil-linked arousal reflects intracranial aperiodic neural activity in the human auditory cortex

doi:10.21203/rs.3.rs-4290405/v1

Download PDF

Article

Pupil-linked arousal reflects intracranial aperiodic neural activity in the human auditory cortex

https://doi.org/10.21203/rs.3.rs-4290405/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

Disrupted cortical and peripheral neural responses to salient stimuli occur in neural disease states, which may be rooted in pathophysiological neuromodulatory system dynamics. Although low-frequency oscillatory activity is the canonical measure of cortical neuromodulatory state, aperiodic 1/f slope encapsulates the balance between low-frequency and high-frequency activity and could thus provide a more sensitive electrophysiological measure. Here, we simultaneously record pupil diameter as a noninvasive measure of brain state, and intracranial local field potentials (LFP) in the auditory cortex and association regions in humans during an auditory oddball task. We demonstrate a trial-by-trial relationship between pupil and auditory cortical responses that is specific to the high gamma frequency band. We find that pupillary and cortical responses show a state-dependent relationship with aperiodic 1/f slope that is similar to canonical low-frequency measures, but that only the 1/f measure shows a trial-by-trial relationship with tonic and phasic pupil activity. Finally, salient stimuli trigger prolonged shifts in aperiodic 1/f activity after stimulus playback, which associates with altered responses on subsequent trials. Taken together, aperiodic slope captures tonic and phasic components of cortical state at the single trial level, providing mechanistic insight into human saliency responses and a path for pupillometry as a non-invasive readout.

Aperiodic

1/f slope

adaptive modulation

arousal state

pupil diameter

human iEEG

Neurological and psychiatric disorders are often characterized by an inability to distinguish relevant cues from the environment and properly process incoming information^1–5. The biased perception or sub-optimal processing of information in these disorders is likely tied to the concomitant disruption in cortical⁶ and autonomic⁷ responses to salient stimuli. What drives the pathological variability in responses between and within clinical groups is not well understood. Recent research highlights that variability in reaction time and neural response amplitude can act as a stable trait-like characteristic, which is linked to differences in task performance, psychopathology, aging, and cognitive decline^8–10. The variability in auditory^11–14, visual^15,16, olfactory^17,18, and somatosensory^19,20 neural responses can be partially explained by changes in moment-to-moment excitability dynamics, or arousal ‘state’. Through the lens of adaptive gain theory, the ongoing arousal state provides a type of gain to the drive of a presented sensory stimulus^15,21,22 and may be modulated by the incoming stimulus to suit the demands of the task at hand²². Systemic disruptions in state or state modulation may exist in neural diseases and manifest as sub-optimal sensory, peripheral, and behavioral responses to incoming stimuli. Thus, understanding how different components of arousal state can be optimally measured will likely improve our ability to define disease states and to track them.

The Locus Coeruleus (LC) is a small nucleus in the reticular activating system of the brainstem, which is thought to drive state and state-modulation processes via a tonic mode of consistent low frequency (1–6 Hz) firing patterns, and a phasic mode of short bursty activity in higher frequencies (8–15 Hz)(Janitzky 2020; Aston-Jones and Cohen 2005; Fazlali et al. 2016), respectively. Both modes of LC activity can be tracked non-invasively by peripheral measures such as pupil diameter^25–28. In the sensory cortices, electrophysiological correlates of tonic, spontaneous arousal states are graded in terms of the ongoing low-frequency oscillatory activity^29,30: quiet wakefulness (low arousal, small pupil diameter) is defined by a large proportion of high amplitude, low-frequency oscillations, and active wakefulness (high arousal, large pupil diameter) is defined by a lower proportion of these oscillations^14,31,32. Low-frequency oscillatory activity is typically quantified as delta power or summative low-frequency power (2–10 Hz)^14,29,32. The corresponding electrophysiological measures of phasic arousal in sensory cortices are typically quantified as spiking activity²⁶. In the adaptive gain theory, phasic activity may be optimized by tonic measures of arousal state, where intermediate levels of tonic activity correspond to the largest response amplitudes and best behavioral performance²², and show an inverted-u shaped state-dependent relationship, which is well-established in the LC and sensory regions of non-human animal models^14,30,31. However, the relationship between tonic and phasic state remains under-explored in the human brain, particularly in sensory regions.

Since tonic and phasic state modes exhibit a bidirectional relationship with one another, finding a single, composite electrophysiological measure that can track both simultaneously could provide a more holistic view of arousal state and reduce analytical dimensionality. Some studies have defined arousal state as the ratio between low- and high-frequency activity^24,29, though what defines ‘low’ or ‘high’ frequency in the ratio is highly variable between studies. The aperiodic (1/f) slope of the power spectrum is an increasingly prominent signal of interest that may also capture this balance³³ in a single computation and has been proposed as an index for physiological arousal between anesthesia, sleep and wake states³⁴. A steeper, more negative 1/f slope corresponds to increased low-frequency, synchronous activity, whereas a flatter slope corresponds to increased high-frequency, asynchronous activity associated with increased cortical excitability³⁵. We posit that the aperiodic (1/f) slope encapsulates the tradeoff between the global, tonic nature of the slow oscillatory activity and the localized phasic high-frequency activity that modulates it³⁶. To our knowledge, aperiodic (1/f) activity has not yet been tested as a measure of state modulation or state-dependence of cortical or peripheral salience responses in any model system. Elucidating the direct relationship between phasic and tonic state activity in sensory regions, as well as the state dependence of sensory cortical and pupillary responses, is an essential step to understanding the mechanics of arousal state in human disease populations.

In the current study, we aim to address these gaps by simultaneously recording pupil diameter and intracranial local field potentials (LFPs) in auditory cortical regions of awake, behaving human patients undergoing intracranial monitoring for seizure focus mapping for treatment-resistant epilepsy. We correlate pupil diameter and auditory cortical responses in an auditory oddball paradigm to identify the spectral specificity of state-modulation related phasic responses in the auditory cortex. We hypothesize that phasic pupil (‘state-modulatory’) responses will show the strongest relationship with phasic high gamma, low beta, and alpha responses in the auditory cortex in line with previous findings recorded directly in the anterior insula³⁷ and scalp electroencephalography (EEG) P300^38–40 responses. Second, we compare aperiodic activity and low-frequency power (defined as delta power fraction^14,29) as measures of arousal state to examine the state-dependence of these phasic responses. We hypothesize that both measures of state will show an inverted-u-shaped relationship, as has been shown in previous animal studies^14,30,31, where intermediate values of cortical state will correspond to the largest phasic response amplitudes. Finally, we examine adaptive modulation of arousal state between trials by quantifying the shift in the aperiodic slope between pre- and post-stimulus windows and hypothesize that this momentary shift may be tied to the variability seen in trial responses.

Auditory cortical responses track phasic pupil activity. We examined the extent to which phasic pupillary responses tracked high gamma, low beta, or alpha responses in the auditory cortex, and whether these relationships were condition-specific. The average peak latency of phasic oddball high gamma responses ranged from 0.06 to 0.15 seconds post-stimulus between subjects, with an average peak amplitude between 57.91-199.27 percent change from baseline. Average pupil oddball response latency ranged from 0.88 to 1.64 seconds post-stimulus between subjects, with an average peak amplitude between 2.04 and 11.88 percent change of maximum pupil size (Fig. 2a-b). In line with our hypotheses, between-subjects correlational analyses (Fig. 2c) showed a significant positive relationship between pupil responses and high gamma responses (r = 0.57;p-adj = 0.046), and a significant negative relationship between pupil responses and low beta (r=-0.79;p-adj = 0.0002) or alpha (r=-0.56;p-adj = 0.046) responses in the oddball condition. No clear relationship was identified between pupil responses and cortical responses for the standard condition (r=-0.04-0.30;p-adj = 0.37–0.87, Fig. 2c).

Given prior research linking high gamma, beta, and alpha neural responses to phasic pupil responses in non-invasive research paradigms, we also examined these bands at the trial level. Similarly to our between-subjects analysis, a linear mixed-effects model (LME) revealed a significant condition-specific relationship between high gamma and pupil responses within subjects. The initial high gamma LME was constructed with response size and condition as interacting fixed effects (see model notations and summaries in Table 1). This model revealed a significant interaction effect (p-adj.=0.007) between cortical response amplitude and condition (i.e. standard or oddball) in predicting pupil responses, with a non-significant main effect of cortical response amplitude (p-adj.=0.486) and trial condition (p-adj.=0.263). The lack of a main effect of condition was unexpected given the robust salience response seen between standard and oddball responses in high gamma and pupil responses (Fig. 2a-b). We separated standard and oddball trials into separate LMEs (Table 1) to disentangle the significant interaction term found in the full model: we found that high gamma response amplitudes predicted pupil response amplitudes for oddball (p = 0.0013), but not standard (p = 0.390) trials, such that as high gamma response amplitudes were larger, pupil response amplitudes were also larger when considering oddballs. Model results were significant following Benjamini-Hochberg corrections for three high gamma models (Table 1). We found no significant relationship between low beta-frequency or alpha-frequency neural responses and pupil dilations at the trial level (Table 1).

Table 1

Trial-Level Pupil Tracking of Auditory Cortical High Gamma, Beta and Alpha Responses
Model	Model Coefficients	Fixed Effect	Estimate	Adj.-P-Value
High Gamma (Full)	P ~ HG*C + (1\|Sub)	(Intercept)	9.56e-03	0.976
		HG Response	2.61e-03	0.486
		Trial Condition	5.21e-01	0.263
		HG*Condition	1.69e-02	0.007**
High Gamma (Oddball)	P ~ HG + (1\|Sub)	(Intercept)	8.52e-01	0.155
High Gamma (Oddball)	P ~ HG + (1\|Sub)	HG Response	1.58e-02	0.004**
High Gamma (Standard)	P ~ HG + (1\|Sub)	(Intercept)	-2.28e-02	0.947
High Gamma (Standard)	P ~ HG + (1\|Sub)	HG Response	3.35e-03	0.486
Low Beta (Full)	P ~ LB*C + (1\|Sub)	(Intercept)	1.75e-01	0.547
		Low Beta	-3.33e-03	0.090.
		Trial Condition	1.58	p < 0.0001***
		Beta*Condition	-9.86e-03	0.105
Alpha (Full)	P ~ A*C + (1\|Sub)	(Intercept)	1.19e-01	0.678
		Alpha	-2.91e-04	0.744
		Trial Condition	1.81	p < 0.0001 ***
		Alpha*Condition	-3.71e-03	0.159
Statistical significance (after correction for multiple comparisons): P < 0.05, P < 0.01, **P < 0.001. Critical P-value after FDR correction for multiple comparisons (Benjamini-Hochberg) are shown. Asterisks have been removed if no longer significant after correction. Estimates reflect the modeled differences between factors within each fixed effect: HG = High Gamma Response Amplitude (% Change), LB = Low Beta Response Amplitude (% Change), A = Alpha Response Amplitude (% Change), C = Condition (Oddball – Standard), P = Pupil Response Amplitude (% of Max Change), Sub = subject identifier.

Aperiodic 1/f activity tracks tonic pupil activity. Next, we examined if the tonic arousal state (indexed via non-baseline normalized pupil size) was tracked by a canonical measure of electrophysiological state (delta power fraction), or our proposed measure of electrophysiological state (1/f activity) extracted from the auditory cortex (see Methods). Trial-by-trial LME analyses (Table 2) showed that delta power fraction extracted from the auditory cortex was not a significant predictor of tonic pupil size (p-adj.=0.674), but aperiodic 1/f activity was a significant predictor (p = 0.019), such that pupil diameter increases as the aperiodic slope becomes flatter (i.e. state is more excitable).

Table 2

Aperiodic 1/f activity tracks tonic pupil diameter.
Model	Model Coefficients	Fixed Effect	Estimate	Adj.-P-Value
Aperiodic Slope	RP ~ S*Condition + (1\|Sub)	(Intercept)	80.02	p < 0.0001 ***
		Slope in Response	2.01	0.019*
		Condition	-1.95	0.671
		Slope*Condition	-0.95	0.658
Delta Power Fraction	RP ~ D*Condition + (1\|Sub)	(Intercept)	72.06	p < 0.0001***
		Delta in Response	2.75	0.674
		Condition	2.77	0.474
		Slope*Condition	-4.34	0.658
Statistical significance (after correction for multiple comparisons): P < 0.05, P < 0.01, **P < 0.001. Adjusted-P-values after FDR correction for multiple correction (Benjamini-Hochberg) are shown. Asterisks have been removed if no longer significant after correction. Estimates reflect the modeled differences between factors within each fixed effect: RP = Non-baseline corrected pupil size (% of Max) in response window (0-750 ms), Condition (Oddball – Standard), in response window (0-750 ms), S = Aperiodic 1/f slope estimate in response window (0-750 ms), D = Delta Power Fraction in response window (0-750 ms) Sub = subject identifier.

Auditory Cortical Response State Dependence. We sought to identify whether the phasic cortical and peripheral responses followed the well-characterized¹⁴ inverted-u shaped state-dependent relationship with measures of ongoing tonic arousal state extracted from the auditory cortex. We compared the state-dependent relationships between a canonical electrophysiological measure tracking arousal state (delta power fraction^14,29), and our proposed measure of state (aperiodic 1/f activity) for high gamma responses. To do so, we binned the high gamma cortical response sizes by each measure of state (i.e. 1/f slope or delta power fraction) extracted immediately before the onset of the tone stimulus (Fig. 3a-b). High gamma exhibited an inverted-u shaped relationship with pre-stimulus arousal state for both delta fraction (adj-R-squared_std=0.698,p = 0.012; adj-R-squared_odd=0.728,p = 0.0085) and 1/f slope (adj-R-squared_std=0.974,p = 4.50e-05; adj-R-squared_odd=0.511,p = 0.072).

At the trial level for high gamma responses, LME models did not show a significant relationship with delta power fraction (Table 3) but did show a significant inverted quadratic (i.e. inverted-u) fit (p-adj._slope<0.0001,p-adj._{slope−squared}=p < 0.0001) for aperiodic 1/f slope predicting high gamma response amplitude. An additional main effect of stimulus condition (p-adj._condition<0.0001), and interaction effects between condition and the quadratic fit for slope (p-adj._{slope*condition}<0.0001,p-adj._{slope−squared*condition}=0.0061), suggest that the oddball condition shows a larger response and a steeper inverted-u shaped relationship with slope compared to the standard condition (Table 3).

Table 3

High Gamma Response State Dependence.
Model	Model Coefficients	Fixed Effect	Estimate	Adj.P-Value
High Gamma by Aperiodic Slope	HG ~ (S + I(S^2))*Condition +(1\|Sub/Electrode)	(Intercept)	17.65	0.024 *
		Slope	-8.79	p < 0.0001 ***
		Slope^2	-0.78	p < 0.0001 ***
		Condition	71.46	< p < 0.0001 ***
		Slope*Condition	14.67	p < 0.0001***
		Slope^2*Condition	1.21	0.0061 **
High Gamma by Delta Power Fraction	HG ~ (D + I(D^2))*Condition +(1\|Sub/Electrode)	(Intercept)	38.74	p < 0.0001***
		Delta	10.79	0.50
		Delta^2	-28.32	0.42
		Condition	32.22	< p < 0.0001 ***
		Delta*Condition	28.94	0.41
		Delta^2*Condition	-64.28	0.40
Statistical significance (after correction for multiple comparisons): P < 0.05, P < 0.01, **P < 0.001. Critical P-values after FDR correction for multiple comparisons (Benjamini-Hochberg) are shown for all models. Asterisks have been removed if no longer significant after correction. Estimates reflect the modeled differences between factors within each fixed effect: HG = High Gamma Response Amplitude (% Change), C = Condition (Oddball – Standard), D = Delta Power Fraction in response window (0-750 ms), S = Aperiodic 1/f slope estimate in response window (0-750 ms), Sub = subject identifier, Electrode = electrode within the pSTG.

Peripheral Response State Dependence. To examine if pupillary responses were reflective of these state-dependent relationships using a measure of arousal state extracted from the auditory cortex, we performed congruent analyses to those examining cortical responses (i.e. group-level binning followed by LME fitting). We saw no clear relationship between pupil response and pre-stimulus auditory cortical arousal state for either state metric (Supplementary Fig. 1a-b). While the group-level standard condition did show a significant fit (adj-R-squared = 0.974,p < 0.0001), this was not seen at the trial level (see Supplementary Table 2). Given that pupil responses are known to lag behind cortical responses, we hypothesized that the response window for cortical activity (0-750ms) in the auditory cortex may be more reflective of the ‘state’ that is driving pupillary response differences. To test this, we binned pupil responses by post-stimulus (0-750ms) state values and saw different fits for each state metric (Fig. 3c-d). While the aperiodic slope showed an inverted-u-shaped relationship with pupil responses for oddball (p-adj.=0.032), it showed a positive linear relationship in the standard condition (p-adj.=0.032), both of which generalized to the trial-level within subjects (See Table 4). Delta fraction showed a non-significant inverted-quadratic relationship with pupil responses for the standard condition (p-adj.=0.0760) and a u-shaped relationship in the oddball condition (p-adj.=0.0573).

Table 4

Pupil Response State Dependence (Late Window)
Model	Model Coefficients	Fixed Effect	Estimate	Adj-P-Value
Aperiodic Slope (Oddball)	P ~ (S + I(S^2)) + (1\|Sub)	(Intercept)	-6.35	0.154
		Slope	-5.24	0.030 *
		(Slope)^2	-0.78	0.019 *
Aperiodic Slope (Standard)	P ~ S + (1\|Sub)	(Intercept)	2.83	0.013 *
Aperiodic Slope (Standard)	P ~ S + (1\|Sub)	Slope	0.74	0.009 *
Statistical significance (after correction for multiple comparisons): P < 0.05, P < 0.01, **P < 0.001. Critical P-value after FDR correction for multiple correction (Benjamini-Hochberg) is shown. Asterisks have been removed if no longer significant after correction. Estimates reflect the modeled differences between factors within each fixed effect: P = Pupil Response Size (% of Max Change), Condition (Oddball – Standard), S = Aperiodic 1/f slope estimate, D = Delta Power Fraction, Sub = subject identifier, I = polynomial constant.

Aperiodic 1/f activity reflects adaptive modulation of state. To identify the spectral characteristics of saliency responses within the auditory cortex and identify how they may be reflected in aperiodic slope, we averaged standard (Fig. 4a) and oddball (Fig. 4b) trials within each run per patient. Two distinct response components were evident: an early broadband increase in high-frequency power during and slightly after stimulus playback (0-350 ms), and a later (relative) broadband suppression in low-frequency activity after stimulus offset (250–750 ms). We computed the averaged percent power change from baseline between test blocks (N = 18) for every frequency band (high gamma [70–150 Hz]; low gamma [32–70 Hz]; high beta [20–30 Hz]; low beta [12–20 Hz]; alpha [8–10 Hz]; theta [4–8 Hz]; delta [2–4 Hz]) in the early and late time windows to quantify the magnitude and significance of these response components and extract any feature specificity (Supplementary Fig. 2). For the early time window, there was a significant increase in power after the onset of the stimulus in both the standard and oddball conditions for each canonical frequency band, except for oddball responses in the low beta frequency band (Supplementary Table 3). For the late time window, (250–750 ms) power increased after stimulus onset for all frequency bands in the standard condition. The oddball condition was differentially affected; high gamma, low gamma, and delta showed an increase in power, while high beta, low beta, alpha, and theta showed a change that was not statistically different from zero (Supplementary Table 4). We saw statistically significant changes to the early high frequency and late low frequency in the early and late cortical response windows, respectively (Fig. 4c).

We examined the difference spectrum between the standard and oddball conditions to characterize the salience detection component of the auditory response (Fig. 4d). Whereas in Fig. 4a-c we examined standard and oddball responses averaged between subjects, in Fig. 4d, we computed the within-test block Oddball - Standard differences and averaged those differences between all subjects. As expected, the early increase in high-frequency activity and the late intra-to-post stimulus decrease in low-frequency activity were significantly greater for oddball compared to standard trials in every canonical frequency band (Fig. 4e). Seeing that the phasic auditory responses appeared to be broadband in nature, we sought to examine the shift in slope between pre-stimulus, post-early, and post-late trial periods (Fig. 4f). Pre-stimulus slope was not significantly different between standard and oddball trials (t(84) = -0.358, p = 0.943; Fig. 4f), and both conditions exhibited significant flattening of the slope in the early time window (STD: t(84) = 5.95, p-adj. < 0.0001; ODD: t(84) = 8.76, p-adj. <0.0001) which was not different between the two conditions (t(84) = 2.45, p-adj. = 0.098). The late post-stimulus period slope estimate had nearly returned to the pre-stimulus slope estimate for the standard condition (t(84) = 1.49, p-adj. = 0.424), but the late slope estimate in the oddball condition maintained a flatter slope than the pre-stimulus period (t(84) = 7.03, p-adj. <0.0001) and was not different from the early post-stimulus window slope (t(84) = 1.73, p-adj. = 0.351; Fig. 4f). To examine the effects of this prolonged slope shift on the following trial, we extracted the standard trial responses that occurred immediately after the onset of the oddball trials and compared those responses to standard responses after other standards. The average ‘After Oddball’ (AO) response size across all subjects showed a larger amplitude than standard responses that came after other standard responses for high gamma (M_AO−Std(SD) = 3.79(57.34), t(3737) = 44.50, p-adj. <0.0001; Fig. 4g).

Analytical controls. Given the known tonotopy of the auditory cortex⁴¹, we examined the extent to which cortical response sizes for each frequency band reflected the pitch of the individual sounds used for standard and oddball stimuli (1000 Hz or 2000 Hz, counterbalanced condition assignment, see Methods), and found no significant differences in response size for any frequency band when trials were grouped by tone frequency instead of by condition. (Supplementary Fig. 3, Supplementary Table 5).

The goals of the study were (1) to examine if phasic salience responses in the auditory cortex and association areas are tracked by pupil responses; (2) to identify how the phasic responses in the cortex and the periphery are influenced by fluctuations in cortical arousal state measured in sensory regions; and (3) to understand how different phasic and tonic response components are reflected in the aperiodic slope. We used simultaneous pupil and intracranial recordings in the posterior supratemporal gyrus (pSTG) during an auditory oddball paradigm. In line with our hypotheses, high gamma responses in the auditory cortex are tracked by peripheral responses at the trial level. When we examined the state-dependence of the high gamma phasic responses across two different electrophysiological measures of tonic arousal state, we found an inverted-u-shaped relationship between our proposed measure of tonic arousal state (aperiodic slope) and auditory salience responses. Finally, we saw that a broadband shift in high- and low-frequency activity between pre- and post-stimulus periods was reflected in aperiodic 1/f activity, which may contribute to differences in the following stimulus response sizes. Taken together, these findings suggest that aperiodic slope can capture both tonic and phasic state-related activity in auditory sensory regions, exhibit state dependence similar to canonical low-frequency measures of state, and reflect adaptive modulation of arousal state on a trial-to-trial basis. Moreover, our results collectively align with the adaptive gain theory and may inform the understanding of salience processing disruptions in arousal-related disorders, providing a method to non-invasively track this symptomatology.

Previous experimental work has established that pupil dilation responses can track neural responses to salient stimuli across the fMRI⁴², scalp EEG³⁸, and iEEG³⁷ modalities, even at the single-trial level. The current study demonstrates that this trial-to-trial relationship between salient pupil and cortical responses is robust enough to be seen in sensory regions of the human brain, rather than nodes of the salience network³⁷ or in slow-wave P300 activity³⁸. Thus, we establish that phasic state-modulatory activity can be tracked in auditory sensory regions. High gamma responses were tracked by pupil responses in a condition-dependent manner such that as high gamma responses increased, pupil dilations also increased for oddball but not for standard trials (Table 1). Given the structure of the auditory oddball paradigm (see Methods), the condition-dependence of the relationship is not unexpected. The task inter-trial intervals (ITIs) range between 1.0-1.2 seconds after stimulus offset, meaning that pupil diameter response did not have the typical 6–9 second ITI to fully return to the tonic baseline diameter size, which greatly decreased the response amplitude variability between standard trials. For this reason, we did not expect to see a relationship between cortical response amplitudes and pupil response amplitudes in the standard condition.

Interestingly, fast, event-related phasic changes such as those reflected in broadband gamma activity^43–45, are suggested to drive feed-forward communication to higher processing areas. Prior research has shown a relationship between pupil responses and high gamma responses in the anterior insular node of the Affective-Salience Network (ASN)³⁷. Future research would ideally examine the relationship between response peak latency and amplitude across auditory and salience regions to identify if a feed-forward relationship is discernable at the trial level between the sensory and salience networks. Furthermore, mood disorders such as major depressive disorder and generalized anxiety disorder often show disrupted phasic salience activity in nodes of the ASN in response to emotional stimuli^1,3–5. Examining if a sensory-salience feed-forward relationship is different between clinical groups or across mood severity status may highlight a mechanism contributing to aberrant emotional processing in these disorders.

Foundational research on brain arousal state has identified a tonic mode of consistent low frequency (1–6 Hz) activity and a phasic mode of short higher frequency bursts related to the presentation of salient stimuli^22,23. Both modes are reflected in pupil diameter fluctuations^{25–27,30,46} across several established measures of state ^14,30,32,47. The current study showed no trial-level relationship between tonic pupil and low-frequency delta activity in the auditory cortex, but we did identify a statistically significant relationship between our measure of state (aperiodic slope) and tonic pupil diameter. In line with prior studies, as state is more excitable (1/f activity becomes flatter), pupil diameter size is larger (Table 2). In turn, this suggests that 1/f slope has a similar relationship with pupil diameter as well-established measures of electrophysiological state. These findings present an opportunity to examine arousal-related disorders through a new lens, facilitated by aperiodic 1/f activity. A growing body of work shows that 1/f activity may be related to treatment response for deep brain stimulation⁴⁸ or electroconvulsive therapy⁴⁹ in treatment-resistant depression cases. Taken together with the findings in the current study, this may suggest that disruptions seen in salience responses of the ASN may be driven by systemic dysfunction in tonic and phasic state-modulatory processes. Thus, the shift in 1/f activity in responders to treatment may indicate stimulation therapies work to alter the tonic state in these regions (which is reflected in altered aperiodic 1/f activity), thereby restoring the optimal balance for properly distinguishing incoming emotional information. Future studies should aim to investigate how alterations in cortical arousal state directly track mood symptom severity and should compare how each of the measures of state may track tonic pupil fluctuations in different networks and at different time scales to identify any regional or temporal specificity in state-tracking.

To investigate the aperiodic slope as a measure of state, we also examined auditory salience response state dependence. Auditory region response amplitudes, or auditory task performance, are known to show an inverted-u-shaped relationship between state and response^14,31, and a u-shaped relationship between state and response variability¹⁴. These findings suggest that arousal state may optimize auditory detection and performance by increasing response size and decreasing variability at intermediate levels of arousal. Our initial between-subjects analyses of state dependence with delta power fraction were well-aligned with previous studies and showed the characteristic inverted-u-shaped relationship with localized auditory high gamma power change for both standard and oddball stimuli (Fig. 3a). Aperiodic slope also shows this relationship between subjects (Fig. 3b). However, state dependence of high gamma responses was only recapitulated at the single-trial level within subjects for an aperiodic 1/f slope and was not seen for delta power fraction. The interaction effect between 1/f slope and condition for high gamma state dependence shows a larger response and steeper inverted-u relationship for oddball trials compared to standards, which may suggest that state is not optimizing sensory detection as was previously thought in non-human studies¹⁴ but is optimizing the salience detection responses mainly driven by LC or ASN phasic activity and is simply reflected in feed-forward or feed-back type sensory region responses. Future studies should examine this relationship in further detail by comparing state-dependent relationships across multiple types and intensities of salient stimuli, across different modalities, and different brain networks.

Phasic LC responses are also known to exhibit an inverted-u-shaped relationship with tonic LC activity²², and phasic pupil responses are known to reflect phasic LC responses²². We examined whether pupil phasic responses showed similar inverted-u-shaped state dependence with the sensory-region-localized measures of state. Pupil phasic activity tracked the high gamma activity, which showed state dependence in the auditory cortical regions (Figs. 2,3), but pupil diameter responses did not show a clear state-dependent relationship with pre-stimulus measures of state for either delta or 1/f activity (Supplemental Fig. 1, Supplemental Table 2). Upon further investigation, we observed a between-subjects state-dependent relationship for pupil responses and post-stimulus measures of state (Fig. 3c-d). The delayed nature of the state dependence may indicate that phasic pupillary activity is a feed-back salience response which is potentiated by tonic state in a slightly delayed time window compared to the high-frequency phasic responses that are thought to represent feed-forward activity^43–45. Alpha⁴³, low beta-alpha⁴⁴, and delta-modulated beta⁴⁵ activity are thought to reflect feedback response activity in visual^43,44,50 and auditory⁴⁵ processing, and these frequency bands were correlated with pupil diameter size between subjects (Fig. 2), though these relationships did not generalize to the trial level. Thus, pupil diameter responses may be tracking both, the feed-forward (high gamma) and feedback (alpha-beta) activity relevant to state-modulation processes in the auditory sensory regions. Future research should expand on the current study to examine if pupil responses track any other frequency-specific phasic responses in the auditory cortex besides the ones studied here and clarify if the phasic state-modulation-related activity reflected in pupil diameter is driven by feed-forward or feed-back activity.

Recent work shows that aperiodic activity can shift in response to auditory stimuli on a trial-to-trial basis⁵¹. In the same vein, average salience responses in the current study showed a broadband increase in high-frequency activity and a relative broadband suppression in low-frequency activity, which corresponded to a flattening of the aperiodic 1/f slope between the pre- and post-stimulus time windows. While the magnitude of this shift was not different between standard and oddball conditions, the shift in slope for the oddball condition was prolonged, being presented also into the later time window, whereas the standard condition did not maintain this shift in slope (Fig. 4). We further show that the high gamma phasic responses to standard stimuli that come immediately after oddball stimuli are larger in magnitude than the standard stimuli responses after other standards, suggesting that this prolonged shift may be adjusting localized state and contributing to trial-to-trial response amplitude variability. Future studies should aim to examine this shift in slope over longer periods and in response to ecologically relevant stimuli, which may produce a larger and more prolonged effect than pure sine wave tones.

In conclusion, the current study shows a trial-level relationship between intracranial auditory high gamma responses and pupil diameter responses, which suggests that responses in the auditory sensory regions are highly related to state-modulatory processes. These high-gamma cortical responses exhibit an inverted-u-shaped state dependence with low-frequency delta oscillations, and with aperiodic 1/f activity which can track tonic measures of arousal state as indexed by pupil diameter. Averaged event-related changes exhibit a broadband increase in high-frequency power and a broadband decrease in low-frequency power which aligns with a shift in aperiodic slope that we suggest is indicative of trial-level state modulation. These findings provide further evidence that 1/f activity can capture both tonic and phasic state-related activity, and reflects adaptive modulation on a trial-to-trial basis. Taken together, our study may provide a new understanding of arousal-related pathologies, and the mechanism behind alterations of excitability state mediated by therapeutic interventions.

Participants. The study was approved by the Institutional Review Board at Baylor College of Medicine (IRB: H-18112), in compliance with the international standard of the Declaration of Helsinki. All participants provided written informed consent. Participants consisted of 10 patients (N_female=5; mean age = 41 years; Supplementary Table 1) undergoing intracranial stereo-EEG (sEEG) monitoring for medically refractory epilepsy evaluation. sEEG probes were placed based on clinical criteria.

Auditory Oddball Paradigm. Participants completed an active ‘count’ auditory oddball (AOB) task in which they mentally counted the number of ‘oddball’ tones they heard. The count was reported verbally after each experimental block. Stimuli were 250-ms long pure tones at two frequencies (1 and 2 kHz) representing the standard and oddball sounds (Fig. 1a). The assignment of tone frequency to condition (standard or oddball) was counterbalanced across blocks within each participant. Each block consisted of 160 total trials. A trial included the auditory presentation of the tone followed by inter-trial intervals (ITIs) randomly varying between 1.0 and 1.2 seconds. The first ten trials of each block were always standard stimuli. The remaining 150 trials consisted of 80–85% standard tones with 15–20% pseudo-randomly interleaved oddball tones, which resulted in approximately 128–136 standard trials and 24–32 oddball trials per test block. The pseudo-randomization for oddball stimuli was necessary to ensure that there would be enough time to allow the pupil to return to baseline before presenting another oddball trial (6.0–9.0 seconds inter-oddball interval). Each block lasted approximately 3–4 minutes and each participant completed 2–7 blocks per session. Sounds were played binaurally from speakers situated directly behind the patient at an intensity adjusted based on individual comfort (~ 68–82 dB).

Electrode localization. FSL was employed to align post-implantation CT brain scans from each patient, showing the location of the intracranial electrodes, to their pre-operative structural T1 MRI scans. We used FreeSurfer to reconstruct the brain surface from the MRI volume. BioImage Suite 35 and iELVis⁵² were used to obtain electrode coordinates and determine the position of the electrode with respect to anatomical landmarks. The anatomical assignment was based on the proximity to the cortical surface and labeled based on the Destrieux cortical parcellation atlas, which was then confirmed based on independent expert visual inspection. Electrodes were included if they were located within 5 mm of the grey matter boundary of the posterior Supratemporal Gyrus (pSTG). The auditory cortex was identified based on major sulci along the supratemporal gyrus (pSTG)^53–55. More specifically, ROIs were within Heschl’s Gyrus bounded by the anterior temporal sulcus, through the second transverse gyrus bounded posteriorly by Heschl’s sulcus (Fig. 1b). A total of 80 electrodes were identified in the ROI. After the exclusion of electrodes contaminated by noise or artifacts, 73 electrodes remained for analyses.

Pupil Recording and Pre-Processing. Pupil diameter was recorded using an EyeLink 1000 system (SR-Research, Osgoode, ON, Canada) while participants fixated on a cross presented at the center of a display monitor (Viewsonic VP150, 1920 x 1080) positioned at a comfortable distance (22–24 in) during the AOB task. Each block of data was sampled at 1,000 Hz (n = 1, Desktop Mode) or 500 Hz (n = 9, Remote Mode) after the successful completion of a 5-point calibration, validation, and single-point drift correction paradigm.

Blinks and artifacts were isolated using the findpeaks.m ⁵⁶ function in MATLAB (R2021a) and confirmed by visual inspection. Peaks were identified as maximal points in the pupil diameter Z-score time series with a minimum peak height of 10 S.D., and a minimum distance of 50 samples between peaks. Peak points were padded by 200 ms and linearly interpolated. Due to the minimum peak distance parameter, not all blinks or artifacts were detected in just one iteration of the pipeline. The findpeaks.m function was used until it could no longer detect any peaks, indicating that all artifacts had been eliminated. This was confirmed via visual inspection on each iteration, with a maximum of 7 iterations of the artifact elimination pipeline for a single patient. Two patients were eliminated due to excessive blinks and artifacts.

The pupil time series was down-sampled to 500 Hz and low-pass filtered at 50 Hz. Trial onsets were identified using event markers sent from the computer generating the standard and oddball sequence to the EyeLink 1000 system. The pupil time series was epoched between 1,000 ms before and 3,000 ms after sound onset. Maximum pupil size values were identified for each run per patient, and time series were normalized to the percentage of the maximum pupil size (% of max) to standardize the time series across patients and runs. Responses were calculated as the difference between the average % of max in the response window (first 2,500 ms) versus the baseline window. Single trials were eliminated if they contained greater than 40% interpolated values, or response average values were greater than or equal to five median absolute deviations from other epochs at that sample location within the task block (see Supplementary Table 1 for final trial counts across all runs electrodes within each condition).

Intracranial EEG Recording and Preprocessing. Neural signals were recorded from stereo EEG probes (Ad-Tech Medical Instrument Corporation or PMT Corporation) connected to a Cerebus data acquisition system (Blackrock Neurotech). All recording signals were amplified, filtered (high-pass 0.3 Hz first-order Butterworth, low-pass 500 Hz fourth-order Butterworth), and digitized at 2,000 Hz. Acoustic stimulation was synchronized with neural signals using an audio analog input which copied the sound waveform to the Cerebus system.

Visual inspection was used to exclude recordings from electrodes with excessive noise, artifacts, and/or interictal activity. Recordings from included electrodes were then re-referenced using a common average reference of all valid electrodes, and notch filtered to remove line noise (60 Hz, first harmonic, second harmonic, and third harmonic). Finally, all recordings were downsampled to 500 Hz using a low-pass Chebyshev type 1 IIR filter of order 8.

The signals obtained were transformed from the time domain into the frequency domain using a wavelet transformation (Morlet, 7 cycles per wavelet; frequencies equally spaced on a linear scale from 2 to 200 Hz). This provided an electrode-by-time matrix of power values for each data run. The analog recorded acoustic signal was used to identify sound onsets. The neural signals of interest (from the electrode-by-time matrix) were obtained by extracting values for each trial (between 1,000 ms before and 2,000 ms after sound onset). Trial power values were normalized to a baseline period (-250 ms to 0 ms; preceding stimulus onset) by calculating % signal change from baseline for the early intra-stimulus (0 ms to 350 ms) and late (250 ms to 750 ms)⁵⁷ post-stimulus time windows. While the stimulus playback only occurred between 0-250 ms, the early intra-stimulus time window was expanded to 0-350 ms to capture high gamma salience-related activity, which has been reported to cluster between 200–300 ms in the human auditory cortex ⁵⁸. We compute percent power change across canonical neural frequency bands (High Gamma [70–150 Hz]; Low Gamma [32–70 Hz]; High Beta [20–30 Hz]; Low Beta [12–20 Hz]; Alpha [8–10 Hz]; Theta [4–8 Hz]; Delta [2–4 Hz]) when investigating the shift in aperiodic 1/f activity between pre-and post-stimulus periods. We focus on our apriori hypotheses surrounding high gamma, low beta, and alpha bands in the correlational and linear mixed-effects modeling analyses. Cortical excitability was quantified in the pre (-250–0 ms) and post (0–750 ms) stimulus time windows using the delta power fraction (delta power / total power), or the aperiodic (1/f) slope estimate, which was calculated using a linear fit (MATLAB function polyfit.m, degree 1, MathWorks™ 2023b) over the 20–45 Hz spectral range. The intermediate frequencies (20–45 Hz) reflect the region of transition from high-to-low frequencies, such that increased delta power and decreased gamma power will have a steep slope in the transition region, whereas low delta power and increased gamma power should have a flatter slope⁵⁹.

Statistical Analyses. Pupil response profiles were assessed at the group level between subjects, and the trial level within subjects. Pupil diameter responses and cortical responses for both conditions were calculated for each task block in each canonical frequency band. A Pearson correlation was conducted for each LFP response component-condition pair, and Benjamini-Hochberg correction was performed for a total of six correlational analyses. Linear mixed-effects (LME) modeling was used to examine the trial-by-trial relationship within subjects for the significant group-level relationships (high gamma, low beta, alpha). Average high gamma, low beta, and alpha component responses were extracted for each electrode per trial and then averaged over all electrodes per trial, such that there was one value per trial, per test block. The LME was constructed with pupil response as the dependent variable, cortical response (high gamma, low beta, or alpha) and condition (standard vs. oddball) as interacting fixed effects, and subject ID as a random effect. Benjamini-Hochberg correction was performed for a total of three high gamma LMEs (see Table 1 for model notations). All LME modeling was completed using the lmer package in R, notations provided in results tables reflect lmer model notations.

State dependence was assessed at the group level between subjects, and the trial level within subjects. Trials were averaged over all electrodes such that each task block only had one value per trial. Pre-stimulus slope or delta fraction (delta fraction = delta power/total power) was computed in the baseline (-250–0 ms) window for each trial. These values were then sorted into evenly spaced bins per condition, except for the extremes which were extended to ensure adequate bin size (N_bin > 5, bin cutoffs for 1/f slope = [-8.5, -5.5,-5.0,-4.5,-4.0,-3.5,-3.0,-2.5,-2,-2.5] and delta fraction = [0, 0.05, 0.1, 0.15, 0.20, 0.25, 0.3, 0.35, 0.40, 0.45, 0.5]). Average pre-stimulus slope values per task block were examined using the MATLAB isoutlier.m function. One subject (Subject 10) was identified as an outlier for both task blocks (falling greater than three median absolute deviations from the group mean) and was excluded from group analyses but included in within-subjects trial-level analyses. A regression line was fit to the binned dataset using the MATLAB lmfit.m function. A quadratic (high gamma responses, pupil-oddball responses) or linear (pupil-standard responses) depending on which fit was most appropriate for each relationship based on R-squared values. Cortical trial-level analyses were conducted using LME modeling within subjects per electrode per subject. Pupil trial-level analyses also utilized LME modeling, though (pre- or post-stimulus) slope values were averaged over all pSTG electrodes per task block for each trial and only fit per subject to avoid pupil response redundancy.

Cortical response profiles were assessed at the group level. Epoched spectra were averaged over all trials for each condition (standard vs. oddball) for a given task block, and early (0–350 ms) and late (250–750 ms) time windows were extracted for each canonical frequency band. Wilcoxon signed-rank tests were first used to assess if each condition-frequency band pair exhibited a significant change from zero, and Benjamini-Hochberg correction was used to adjust for 14 total tests per time window. Then average oddball and average standard responses for each time window were matched by task block and the difference was computed for each pair (Oddball amplitude – Standard amplitude) for each canonical frequency band. Wilcoxon signed-rank tests were again used to assess if each band amplitude was different from zero, and Benjamini-Hochberg correction was used for seven total tests for each time window. This method was also used for assessing the effect of tone frequency (1,000 vs 2,000 Hz) rather than salience condition, and a Kruskal-Wallis test was used to compare group means.

Pre-, post-early, and post-late stimulus slope estimates were averaged over all trials per condition within a given task block. An LME was fit with aperiodic slope estimate as the dependent variable, time window (pre, post-early, post-late) and condition as interacting fixed effects, and subject ID as a random effect. Estimated Marginal Means (EMMs) were computed using the emmeans⁶⁰ function in R, and pairwise comparisons were made between each condition-time window pair. Corrections were made using the Benjamini-Hochberg method. After-oddball responses were identified, and an LME was fit with high gamma responses as the dependent variable and condition (standard, oddball, after oddball) as the dependent variable, and subject ID as a random effect. EMMs were computed, and pairwise comparisons were made across condition.

Acknowledgments

The authors would like to thank the patients and their families for their participation in the research. We would also like to thank the intracranial monitoring unit staff (Luis Escudero, Katrina Reichardt), as well as the following individuals for their support of the project: Chengjun Huang, Joshua Adkinson.

Author contributions

Writing: M.M., E.B., K.R.B., J.W.D. Review and editing: M.M., K.R.B., J.W.D., M.J.M., J.M., E.B. Data analysis: M.M., J.M., E.B., M.J.M., J.W.D., B.A.M., R.K.M., B.P., and S.P. Methodology: M.M, M.J.M., J.W.D., and B.A.M. Conceptualization: M.M., M.J.M., J.W.D., K.R.B., B.A.M. Funding acquisition: K.R.B., S.S.,W.G. Data collection: M.M., B.P., and R.M. Project administration: R.K.M., B.P., S.P.

Data availability statement

All data, code, and materials used in the analysis are available to any interested researcher for the purpose of reproducing or extending the analysis. Please contact the corresponding author at [email protected] to request data from this study. Primary data are also hosted on the National Institutes of Mental Health Data Archive (NDA), and analysis data frames are posted on the Data Archive of the BRAIN Initiative (DABI).

Additional Information

M.M. and K.B. are inventors of a planned patent on the electrophysiology measure of state (1/f activity reported in the current manuscript) as a biomarker for MDD and mood state. KB is an inventor of an issued patent on an unrelated method of electrical stimulation to treat depression, anxiety, and pain. SS has consulting agreements with Boston Scientific, Neuropace, Abbott, and Zimmer Biomet. WG has received donated devices from Medtronic and has consulting agreements with Biohaven Pharmaceuticals. All other authors do not have a conflict of interest.

Schimmelpfennig, J., Topczewski, J., Zajkowski, W. & Jankowiak-Siuda, K. The role of the salience network in cognitive and affective deficits. Front. Hum. Neurosci.17, 1133367 (2023).
Pugliese, V. et al. Aberrant salience correlates with psychotic dimensions in outpatients with schizophrenia spectrum disorders. Ann. Gen. Psychiatry21, 25 (2022).
Yuan, J., Tian, Y., Huang, X., Fan, H. & Wei, X. Emotional bias varies with stimulus type, arousal and task setting: Meta-analytic evidences. Neurosci. Biobehav. Rev.107, 461–472 (2019).
Bradley, B. P., Mogg, K., White, J., Groom, C. & de Bono, J. Attentional bias for emotional faces in generalized anxiety disorder. Br. J. Clin. Psychol.38, 267–278 (1999).
Leppänen, J. M. Emotional information processing in mood disorders: a review of behavioral and neuroimaging findings. Curr. Opin. Psychiatry19, 34–39 (2006).
Klawohn, J., Santopetro, N. J., Meyer, A. & Hajcak, G. Reduced P300 in depression: Evidence from a flanker task and impact on ERN, CRN, and Pe. Psychophysiology57, e13520 (2020).
Silk, J. S. et al. Pupillary reactivity to emotional information in child and adolescent depression: links to clinical and ecological measures. Am. J. Psychiatry164, 1873–1880 (2007).
Myerson, J., Robertson, S. & Hale, S. Aging and intraindividual variability in performance: analyses of response time distributions. J. Exp. Anal. Behav.88, 319–337 (2007).
Arazi, A., Gonen-Yaacovi, G. & Dinstein, I. The Magnitude of Trial-By-Trial Neural Variability Is Reproducible over Time and across Tasks in Humans. eNeuro4, (2017).
Haigh, S. M., Heeger, D. J., Dinstein, I., Minshew, N. & Behrmann, M. Cortical variability in the sensory-evoked response in autism. J. Autism Dev. Disord.45, 1176–1190 (2015).
Curto, C., Sakata, S., Marguet, S., Itskov, V. & Harris, K. D. A simple model of cortical dynamics explains variability and state dependence of sensory responses in urethane-anesthetized auditory cortex. J. Neurosci.29, 10600–10612 (2009).
Kisley, M. A. & Gerstein, G. L. Trial-to-trial variability and state-dependent modulation of auditory-evoked responses in cortex. J. Neurosci.19, 10451–10460 (1999).
Edeline, J. M., Manunta, Y. & Hennevin, E. Auditory thalamus neurons during sleep: changes in frequency selectivity, threshold, and receptive field size. J. Neurophysiol.84, 934–952 (2000).
McGinley, M. J., David, S. V. & McCormick, D. A. Cortical membrane potential signature of optimal states for sensory signal detection. Neuron87, 179–192 (2015).
Goris, R. L. T., Movshon, J. A. & Simoncelli, E. P. Partitioning neuronal variability. Nat. Neurosci.17, 858–865 (2014).
Cui, Y., Liu, L. D., McFarland, J. M., Pack, C. C. & Butts, D. A. Inferring Cortical Variability from Local Field Potentials. J. Neurosci.36, 4121–4135 (2016).
Murakami, M., Kashiwadani, H., Kirino, Y. & Mori, K. State-dependent sensory gating in olfactory cortex. Neuron46, 285–296 (2005).
Ma, M. & Luo, M. Optogenetic activation of basal forebrain cholinergic neurons modulates neuronal excitability and sensory responses in the main olfactory bulb. J. Neurosci.32, 10105–10116 (2012).
Castro-Alamancos, M. A. & Gulati, T. Neuromodulators produce distinct activated states in neocortex. J. Neurosci.34, 12353–12367 (2014).
Rosanova, M. & Timofeev, I. Neuronal mechanisms mediating the variability of somatosensory evoked potentials during sleep oscillations in cats. J Physiol (Lond)562, 569–582 (2005).
Fontanini, A. & Katz, D. B. Behavioral states, network states, and sensory response variability. J. Neurophysiol.100, 1160–1168 (2008).
Aston-Jones, G. & Cohen, J. D. An integrative theory of locus coeruleus-norepinephrine function: adaptive gain and optimal performance. Annu. Rev. Neurosci.28, 403–450 (2005).
Janitzky, K. Impaired Phasic Discharge of Locus Coeruleus Neurons Based on Persistent High Tonic Discharge-A New Hypothesis With Potential Implications for Neurodegenerative Diseases. Front. Neurol.11, 371 (2020).
Fazlali, Z., Ranjbar-Slamloo, Y., Adibi, M. & Arabzadeh, E. Correlation between Cortical State and Locus Coeruleus Activity: Implications for Sensory Coding in Rat Barrel Cortex. Front. Neural Circuits10, 14 (2016).
Murphy, P. R., O’Connell, R. G., O’Sullivan, M., Robertson, I. H. & Balsters, J. H. Pupil diameter covaries with BOLD activity in human locus coeruleus. Hum. Brain Mapp.35, 4140–4154 (2014).
Joshi, S., Li, Y., Kalwani, R. M. & Gold, J. I. Relationships between Pupil Diameter and Neuronal Activity in the Locus Coeruleus, Colliculi, and Cingulate Cortex. Neuron89, 221–234 (2016).
Reimer, J. et al. Pupil fluctuations track rapid changes in adrenergic and cholinergic activity in cortex. Nat. Commun.7, 13289 (2016).
de Gee, J. W. et al. Dynamic modulation of decision biases by brainstem arousal systems. eLife6, (2017).
Poulet, J. F. A. & Crochet, S. The cortical states of wakefulness. Front. Syst. Neurosci.12, 64 (2018).
McGinley, M. J. et al. Waking state: rapid variations modulate neural and behavioral responses. Neuron87, 1143–1161 (2015).
de Gee, J. W. et al. Pupil-linked phasic arousal predicts a reduction of choice bias across species and decision domains. eLife9, (2020).
Reimer, J. et al. Pupil fluctuations track fast switching of cortical states during quiet wakefulness. Neuron84, 355–362 (2014).
Gao, R., Peterson, E. J. & Voytek, B. Inferring synaptic excitation/inhibition balance from field potentials. Neuroimage158, 70–78 (2017).
Lendner, J. D. et al. An electrophysiological marker of arousal level in humans. eLife9, (2020).
McCormick, D. A., McGinley, M. J. & Salkoff, D. B. Brain state dependent activity in the cortex and thalamus. Curr. Opin. Neurobiol.31, 133–140 (2015).
Harris, K. D. & Thiele, A. Cortical state and attention. Nat. Rev. Neurosci.12, 509–523 (2011).
Kucyi, A. & Parvizi, J. Pupillary Dynamics Link Spontaneous and Task-Evoked Activations Recorded Directly from Human Insula. J. Neurosci.40, 6207–6218 (2020).
Chang, Y.-H. et al. Linking tonic and phasic pupil responses to P300 amplitude in an emotional face-word Stroop task. Psychophysiology e14479 (2023) doi:10.1111/psyp.14479.
Studenova, A. et al. Event-related modulation of alpha rhythm explains the auditory P300-evoked response in EEG. eLife12, (2023).
Fabio, R. A., Suriano, R. & Gangemi, A. Effects of Transcranial Direct Current Stimulation on Potential P300-Related Events and Alpha and Beta EEG Band Rhythms in Parkinson’s Disease. J. Integr. Neurosci.23, 25 (2024).
Saenz, M. & Langers, D. R. M. Tonotopic mapping of human auditory cortex. Hear. Res.307, 42–52 (2014).
He, H., Hong, L. & Sajda, P. Pupillary response is associated with the reset and switching of functional brain networks during salience processing. PLoS Comput. Biol.19, e1011081 (2023).
van Kerkoerle, T. et al. Alpha and gamma oscillations characterize feedback and feedforward processing in monkey visual cortex. Proc Natl Acad Sci USA111, 14332–14341 (2014).
Michalareas, G. et al. Alpha-Beta and Gamma Rhythms Subserve Feedback and Feedforward Influences among Human Visual Cortical Areas. Neuron89, 384–397 (2016).
Fontolan, L., Morillon, B., Liegeois-Chauvel, C. & Giraud, A.-L. The contribution of frequency-specific activity to hierarchical information processing in the human auditory cortex. Nat. Commun.5, 4694 (2014).
Yang, H., Bari, B. A., Cohen, J. Y. & O’Connor, D. H. Locus coeruleus spiking differently correlates with S1 cortex activity and pupil diameter in a tactile detection task. eLife10, (2021).
Vinck, M., Batista-Brito, R., Knoblich, U. & Cardin, J. A. Arousal and locomotion make distinct contributions to cortical activity patterns and visual encoding. Neuron86, 740–754 (2015).
Veerakumar, A. et al. Field potential 1/f activity in the subcallosal cingulate region as a candidate signal for monitoring deep brain stimulation for treatment-resistant depression. J. Neurophysiol.122, 1023–1035 (2019).
Smith, S. E. et al. Clinical EEG slowing induced by electroconvulsive therapy is better described by increased frontal aperiodic activity. Transl. Psychiatry13, 348 (2023).
Bastos, A. M. et al. Visual areas exert feedforward and feedback influences through distinct frequency channels. Neuron85, 390–401 (2015).
Gyurkovics, M., Clements, G. M., Low, K. A., Fabiani, M. & Gratton, G. Stimulus-induced changes in 1/f-like background activity in EEG. J. Neurosci.42, 7144–7151 (2022).
Groppe, D. M. et al. iELVis: An open source MATLAB toolbox for localizing and visualizing human intracranial electrode data. J. Neurosci. Methods281, 40–48 (2017).
Nourski, K. V. Auditory processing in the human cortex: An intracranial electrophysiology perspective. Laryngoscope Investig. Otolaryngol.2, 147–156 (2017).
Nourski, K. V. et al. Functional organization of human auditory cortex: investigation of response latencies through direct recordings. Neuroimage101, 598–609 (2014).
Ozker, M., Schepers, I. M., Magnotti, J. F., Yoshor, D. & Beauchamp, M. S. A Double Dissociation between Anterior and Posterior Superior Temporal Gyrus for Processing Audiovisual Speech Demonstrated by Electrocorticography. J. Cogn. Neurosci.29, 1044–1060 (2017).
MathWorks Inc. Signal Processing Toolbox Reference. 2024a, 1013–1039 (2024).
Lindenbaum, L., Steppacher, I., Mehlmann, A. & Kissler, J. M. The effect of neural pre-stimulus oscillations on post-stimulus somatosensory event-related potentials in disorders of consciousness. Front. Neurosci.17, 1179228 (2023).
Todorovic, A., van Ede, F., Maris, E. & de Lange, F. P. Prior expectation mediates neural adaptation to repeated sounds in the auditory cortex: an MEG study. J. Neurosci.31, 9118–9123 (2011).
Hacker, C. et al. Aperiodic neural activity is a biomarker for depression severity. medRxiv (2023) doi:10.1101/2023.11.07.23298040.
Searle, S. R., Speed, F. M. & Milliken, G. A. Population marginal means in the linear model: an alternative to least squares means. The American Statistician34, 216–221 (1980).

Competing interest reported. M.M. and K.B. are inventors of a planned patent on the electrophysiology measure of state (1/f activity reported in the current manuscript) as a biomarker for MDD and mood state. KB is an inventor of an issued patent on an unrelated method of electrical stimulation to treat depression, anxiety, and pain. SS has consulting agreements with Boston Scientific, Neuropace, Abbott, and Zimmer Biomet. WG has received donated devices from Medtronic and has consulting agreements with Biohaven Pharmaceuticals. All other authors do not have a conflict of interest.

SupplementalMaterials04252024Fixed.docx

Download PDF

Editorial decision: Revision requested
08 Jul, 2024
Reviews received at journal
06 Jul, 2024
Reviews received at journal
16 Jun, 2024
Reviewers agreed at journal
08 Jun, 2024
Reviews received at journal
11 May, 2024
Reviewers agreed at journal
05 May, 2024
Reviewers agreed at journal
03 May, 2024
Reviewers invited by journal
03 May, 2024
Editor assigned by journal
03 May, 2024
Editor invited by journal
29 Apr, 2024
Submission checks completed at journal
25 Apr, 2024
First submitted to journal
18 Apr, 2024

You are reading this latest preprint version

Pupil-linked arousal reflects intracranial aperiodic neural activity in the human auditory cortex

Status:

Version 1

Abstract

Figures

Introduction

Results

Discussion

Methods

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1