Eye blinks are sensitive to arousal and attention (Maffei & Angrilli, 2019; Wood & Hassett, 1983), and are influenced by cognitive states (e.g. attentional allocation, transitions in information flow, etc.) (Ponder & Kennedy, 1927; Stern et al., 1984) and amenable to conscious modification. For instance, the state of consciousness of patients with persistent vegetative and minimally conscious states could be inferred by the posterior delta and alpha EEG frequency oscillations observed at blink moments (Bonfiglio et al., 2013; Bonfiglio et al., 2014).
Complementary findings emerge from pre-clinical studies documenting that blinking was associated with increases in arousal and with resetting of neural and vascular dynamics (Turner et al., 2022). The neuronal pathways that underlie blinking and their associations with the ascending arousal network (AAN) are critical to arousal (Singh et al., 2022). The AAN is comprised of various brainstem nuclei including serotonergic (dorsal, DR; and medial raphe, MR), noradrenergic (locus coeruleus, LC), dopaminergic (substantia nigra, SN; and ventral tegmental area, VTA) and cholinergic (pontis oralis, PO; pedunculopontine, PPN) nuclei that project to thalamus, striatum and cortex (Edlow et al., 2012). AAN nuclei through their extended connections can influence broad cortical areas through which they modulate arousal levels (Salhi et al., 2023). While the relationship between blinks, blink rates and the dopamine system has been widely investigated (Taylor et al., 1999) including clinical studies implicating D1 and D2 striatal receptors in blink rates (Demiral et al., 2022), the role of other AAN nuclei to blink events is limited. In humans, we recently showed that in AAN and thalamic nuclei and also in cerebellar nuclei peak BOLD activity correlated with blink moments, which were then followed by broad anti-correlations, including sensory-motor regions, insula, amygdala, parahippocamal regions, and temporal cortex, as BOLD activity decreased post blink (Demiral et al., 2023). In the same study we also showed that during performance in a gambling task, the button presses and blinks were highly synchronized for rewarding trials, revealing a link between blinks and arousal surges with attention and motivation.
Here we aim to replicate our initial results linking eye blinks to the arousal system, which were obtained with a 3 Tesla MRI, on a larger data set obtained with 7 Tesla MRI, which provides with better temporal and spatial resolution and to extend the analyses to the association with vigilance as estimated based on pupillary data. For this purpose we used the HCP dataset of participants scanned with a 7T MRI with eye-tracking and measured the lagged cross-correlations between eye-blink related fluctuations (canonical and temporal derivative of Hemodynamic Response Function, HRF, convolved blink time series, see Supplementary Figs. 1 and 2) and BOLD signal fluctuations in preselected regions of interest (ROI) extracted from AAN (Edlow et al., 2012) and thalamic nuclei (Behrens et al., 2003) from resting state fMRI images using temporal lags (TRs) of − 10s to + 10s. We also extracted ROI in cerebellar nuclei (Buckner et al., 2011), since they receive direct projections from brainstem (Singh et al., 2022) and are implicated in blinking (Dimitrova et al., 2002).
Our first hypothesis was that eye blinks would be in temporal correspondence with activation in AAN, thalamic and cerebellar nuclei such that their activity would peak in temporal association with eye blinks coincident with or preceding them.
Our second hypothesis was that the cortical vigilance state would influence arousal surges and blink related BOLD signals such that i) drowsiness would impose higher amplitude fluctuations and higher variance in the blink related BOLD response due to changes in cortical excitability (see below), and ii) delay in blink related neural response initiated by arousal surges (delayed time difference between blink associated neural activity and the blink itself) due to delays in neural communication.
In our paper we refer to ‘arousal’ as the physiological state driven by the ascending brainstem and thalamic arousal nuclei required to keep the organism attentive and conscious, and refer to 'drowsiness’ (somnolence) as the cortical transitional state between wakefulness and sleep, observed under non-engaging or routine environments or under conditions of sleep deficits. In particular, while conducting resting state MRI scans, participants lay supine under dimmed lights, which can make some individuals feel drowsy and disengaged with a concomitant shutting down of their external senses, a phenomenon known as micro-sleep (Poudel et al., 2014; Soon et al., 2021). These episodes can have partially overlapping features and behavioral presentation with fatigue or tiredness (i.e., sleepiness and fatigue often coexist as a consequence of sleep deprivation) (Shen et al., 2006) or merely emerge due to losing interest in and disengage from the setup in the scanner environment (since there is nothing to do in the resting state), or due to inconsistent activity in systems that regulate sleep; for instance orexin neurons contribute to the maintenance of arousal by inhibiting sleep-promoting neurons in the ventrolateral preoptic nucleus (De Luca et al., 2022). Both sleep deprivation and drowsiness interfere with performance (Makeig et al., 2000; Van Dongen et al., 2003) increasing missed or delaying responses to external task stimuli (i.e., psychomotor vigilance task, PVT).
Drowsiness and arousal can interact with each other, an interaction that has been most commonly described in the sleep literature, where arousal related activity has been found to influence sleep stage changes (Horner et al., 1997; Jones, 2020). Similarly, in the MRI environment, arousal surges can emerge to help the person stay awake and attentive and overcome drowsiness and sleep drive.
Cortical excitability as a function of vigilance has been assessed with Transcranial Magnetic Stimulation (TMS). Specifically TMS evoked EEG potentials (TEP) has been used to measure the propagation of activity from the site of stimulation, thereby providing information on the excitability of brain networks (Sulcova et al., 2022). For example, a recent study showed that prolonged wakefulness and sleep deprivation related drowsiness increased the immediate (< 30ms) TEP EEG response (both amplitude, theta power, and slope to peak-time) to TMS, which was rebalanced after one night of sleep (Huber et al., 2013). Another study suggested that sleep deprivation upscales cortical excitability due to enhanced glutamate-related cortical facilitation and decreases and/or reverses GABAergic cortical inhibition (Salehinejad et al., 2022), probably by engaging homeostatic factors that modulate these neurotransmitter systems (Del Cid-Pellitero et al., 2017). (However, see (Mroczek et al., 2024) for a comparison and contrast of the effects of different TMS techniques on a set of cortical excitability measures).
In this context, we hypothesized that blinking and its temporal proximity to neural activation in the arousal systems would be impacted by drowsiness, such that the tonic cortical responsiveness during states of decreased vigilance would lead to slower responses to the transient fluctuations of arousal surges. Indeed, sleep deprivation and states of drowsiness have been shown to delay responses to task stimuli (i.e., in PVT) (Makeig et al., 2000; Van Dongen et al., 2003). Considering the reported increased neuronal excitability to TMS in states of drowsiness, we also hypothesized enhanced blink-associated BOLD reactivity. Thus, we expected that blink-related neural excitability would increase, and responses would be delayed under drowsiness.
To determine the effects of vigilance on BOLD-blink associations we relied on the pupillary data to estimate the percentage of time the eyes were open during an fMRI run; defining ‘Vigilant’ when the eyes were open > 90% and ‘drowsy’ when the eyes were closed between 10%-75% of the time. We found that cross-correlations were strongest preceding or coincident with the blink within AAN and thalamic nuclei (also in cerebellum) replicating our prior findings, and that vigilance influenced these associations, affecting their amplitude and timing. Brain-wide analyses also identified strong positive cross-correlations in insula, lateral geniculate nucleus (LGN) and visual cortex, and wide cortical anti-correlations post blink that were driven by the state of drowsiness in sensory-motor and temporal cortices. Complementing these analyses, using the temporal derivative of the HRF, we show that time-to-peak of the blink-related BOLD activity occurred earlier in the drowsy than in the vigilant state as the HRF rise from the neural activity was delayed. The event-based analysis also showed that blinks were delayed after the initiation of arousal surges in the drowsy compared to the vigilant state. Our results revealed vigilance-dependent blink timing coincident with arousal and cortical networks in the brain.