Ketamine effects on default mode network activity and vigilance: A randomized, placebo‐controlled crossover simultaneous fMRI/EEG study

Abstract In resting‐state functional connectivity experiments, a steady state (of consciousness) is commonly supposed. However, recent research has shown that the resting state is a rather dynamic than a steady state. In particular, changes of vigilance appear to play a prominent role. Accordingly, it is critical to assess the state of vigilance when conducting pharmacodynamic studies with resting‐state functional magnetic resonance imaging (fMRI) using drugs that are known to affect vigilance such as (subanesthetic) ketamine. In this study, we sought to clarify whether the previously described ketamine‐induced prefrontal decrease of functional connectivity is related to diminished vigilance as assessed by electroencephalography (EEG). We conducted a randomized, double‐blind, placebo‐controlled crossover study with subanesthetic S‐Ketamine in N = 24 healthy, young subjects by simultaneous acquisition of resting‐state fMRI and EEG data. We conducted seed‐based default mode network functional connectivity and EEG power spectrum analyses. After ketamine administration, decreased functional connectivity was found in medial prefrontal cortex whereas increased connectivities were observed in intraparietal cortices. In EEG, a shift of energy to slow (delta, theta) and fast (gamma) wave frequencies was seen in the ketamine condition. Frontal connectivity is negatively related to EEG gamma and theta activity while a positive relationship is found for parietal connectivity and EEG delta power. Our results suggest a direct relationship between ketamine‐induced functional connectivity changes and the concomitant decrease of vigilance in EEG. The observed functional changes after ketamine administration may serve as surrogate end points and provide a neurophysiological framework, for example, for the antidepressant action of ketamine (trial name: 29JN1556, EudraCT Number: 2009‐012399‐28).


| INTRODUCTION
The default mode network (DMN) during resting-state (rs) condition constitutes a network of brain regions, including the medial prefrontal cortex (mPFC), posterior cingulate cortex (PCC)/precuneus, medial, lateral, and inferior parietal cortex. These regions become simultaneously active when subjects are self-referential and not focused on the outside world with the brain being at wakeful rest (Buckner, Andrews-Hanna, & Schacter, 2008;Lemogne et al., 2010;Lemogne, Delaveau, Freton, Guionnet, & Fossati, 2012). Depressed patients tend to be self-referential (Bergouignan et al., 2008;Grimm et al., 2009) and several studies have consistently demonstrated increased DMN functional connectivity in depressed patients (Greicius et al., 2007;Sheline et al., 2009;Sheline, Price, Yan, & Mintun, 2010;Zhu et al., 2012). Accordingly, DMN functional activity is of particular interest as a potential mechanistic marker of depression and antidepressant treatment-response. Using resting-state fMRI (rsfMRI) measures such as functional connectivity in DMN as a mechanistic and treatment response marker, however, may come with a major drawback.
As recently pointed out by Tagliazucchi and Laufs (Tagliazucchi & Laufs, 2014) in a large multicentric study of 1,147 rsfMRIs from the "1000 Functional Connectomes Project", resting state is an uncontrolled condition and its heterogeneity is neither sufficiently understood nor accounted for. Based on a long-standing tradition in electroencephalography (EEG) research to use EEG as an objective measure of vigilance (Ott, McDonald, Fichte, & Herrmann, 1982) and using simultaneously acquired EEG during rsfMRI measurements, one-third of subjects were found to exhibit unstable wakefulness and a loss of wakefulness within 3 min. In their study, these dynamic changes of wakefulness were associated with fundamental changes in the associated BOLD responses (functional connectivity). The authors concluded from their findings that vigilance or vigilance monitoring is required when using rsfMRI as a functional biomarker.
For obvious reasons, this kind of monitoring is even more required when drugs are investigated which may affect the level of vigilance and consciousness such as the anesthetic ketamine or related drugs.
For instance, Bonhomme et al. (2016) recently reported a breakdown of DMN functional connectivity between PFC and PCC during stepwise increase of ketamine (or propofol) dosage when comparing the level of sedation during wake state and light and deep sedation (until unresponsiveness). Accordingly, what is needed is an integration of the fMRI findings on the action of ketamine on DMN functional connectivity into a mechanistic neurophysiological model, which accounts for the dynamics of vigilance since even dynamic restingstate analysis techniques with its temporal resolution in a subsecond scale (Brinkmeyer et al., 2010;Zalesky, Fornito, Cocchi, Gollo, & Breakspear, 2014) might not be sufficient enough. Thus by extension, simultaneous assessment of fMRI data and vigilance using a neurophysiological measure such as EEG may improve our understanding of the antidepressant ketamine action. McKinnon et al. (2018) recently reported an association of sleep disturbance and DMN functional connectivity in patients with a lifetime history of depression. They found an increased DMN functional connectivity in depressed patients, most notably in depressed patients with concomitant sleep disorder but increased DMN functional connectivity was also seen in nondepressed control subjects with disturbed sleep. In many cases disturbed sleep emerges even before the onset of clinical depression, up to 90% of people with depression complain about diminished sleep quality, which often includes self-referential, agonizing rumination while being awake during the night (German: "Nächtliches Grübeln") (Benjamins et al., 2017;Nolen-Hoeksema, Wisco, & Lyubomirsky, 2008). At the same time, this reduced sleep quality is associated with altered state of vigilance regulation during daytime-in particular hyperarousal (Riemann et al., 2010;Ulke et al., 2017). Also, many antidepressants decrease vigilance (Alberti, Chiesa, Andrisano, & Serretti, 2015;Hensch et al., 2015)-in addition to their effect on frontal functional connectivity (see above)-and it has been suggested that this effect on vigilance contributes to their antidepressant effects (Hegerl & Hensch, 2014). On the basis of these studies, one could expect that the counter-acting effect of (antidepressant) subanesthetic ketamine on DMN functional connectivity is related to a decrease of the level of vigilance. If this notion were correct, the antidepressant effects of ketamine on functional connectivity and vigilance may constitute two sides of the same coin. In such a scenario, the ketamine effect on vigilance within the framework of resting-state fMRI study is not a nuisance parameter that needs to be corrected for.
Rather, the parallelism of drug effects would (a) provide a more complete picture of the potential antidepressant ketamine effect and (b) improve our understanding of the pathophysiology of depression.
In an earlier study, we have already argued that an effect of subanesthetic doses of ketamine on vigilance is most likely (Musso et al., 2011). Accordingly, the notion that subanesthetic ketamine is decreasing vigilance, similar to light sleep, is consistent with previous resting EEG studies during subanesthetic ketamine challenge (Maksimow et al., 2006;Muthukumaraswamy et al., 2015). These two subanesthetic ketamine EEG studies reported an increase of slow wave power with a concomitant reduction of alpha power-an EEG pattern that is typically seen during light sleep according to the sleep staging of Rechtschaffen and Kales (1968). Similarly, an increase of slow wave activity is also seen during increasing levels of anesthesia (Ching & Brown, 2014;Purdon et al., 2013). However, both EEG studies during subanesthetic ketamine infusion also reported an increase of high-frequency EEG (gamma) activity which was also found by others but with no concomitant increase of slow wave power (de la Salle et al., 2016;Hong et al., 2010;Lazarewicz et al., 2009). These seemingly discrepant findings (theta vs. gamma activity) are not entirely unexpected. Light sleep or sleep stage 1 (according to Rechtschaffen & Kales, 1968) is a transitory and dynamic stage between wake and sleep, it is an unstable (arousable) sleep stage with relatively frequent so-called cyclic alternating patterns (CAPs) compared to deep sleep (Terzano et al., 2001). In EEG, CAPs are characterized by transient slow wave oscillations (i.e., decreased vigilance) but also intermittent low amplitude high frequency oscillations up the gamma frequency range, that is, desynchronized EEG vigilance (Parrino, Grassi, & Milioli, 2014;Simor, Gombos, Szakadát, Sándor, & Bódizs, 2016). Interestingly, Ferri, Rundo, Bruni, Terzano, and Stam (2007) could show that the slow wave CAPs possess small world properties with high cluster coefficient and small path length. Accordingly, one could expect that after subanesthetic ketamine administration, an increase of slow frequency theta/delta EEG activity (or an increase of EEG gamma activity) during resting state, which indicates increased small world properties in frontal brain regions (Ferri et al., 2007), would be paralleled by a frontal decrease of DMN functional connectivity to parietal brain regions.
In this study, we sought to clarify whether subanesthetic ketamine effects on DMN connectivity (reduction of connectivity) are related to ketamine effects on vigilance with an increase of slow-frequency and high-frequency EEG activity and a concomitant decrease of alpha EEG activity (sleep stage 1). For this purpose, we analyzed data from a randomized, placebo-controlled crossover study with subanesthetic S-Ketamine study applying simultaneous rsfMRI/EEG. This is of special interest when having in mind that the Federal Drug Administration (FDA) approved the first ketamine-based medicine (S-ketamine nasal spray) for the treatment of (otherwise treatment-resistant) depression (Carey, 2019).

| METHODS
The study was conducted in compliance with the declaration of Hel- The study was approved by the local ethics committee (Ärztekammer Nordrhein, Düsseldorf, Germany) and by the German federal drug agency (Bundesinstitut für Arzneimittel and Medizinprodukte, BfArM).
Written informed consent was obtained from all participants.

| Subjects and study design
Inclusion and exclusion criteria of the subjects and the study design were explained in detail in an earlier publication (Musso et al., 2011).
In short, during this randomized, double-blind, placebo-controlled crossover trial, N = 24 healthy, young male subjects without drug treatment at least 4 weeks before study inclusion and with no prior history of a neuropsychiatric disorder were investigated twice at least 1 week apart. Timing between both measurements was chosen to avoid possible residual drug effects on the following crossover scan. As mentioned in our previous publication (Musso et al., 2011), all participants underwent a full examination (medical, neurological, and psychiatric) by a board certified psychiatrist and neurologist. Subjects with a positive history of clinically significant medical/neurological/psychiatric conditions were excluded. Sixteen subjects were nonsmokers, while eight subjects were current smokers. To minimize any smoking effects on vigilance, the study setup followed a standardized operating procedure and the subjects were not allowed to smoke at least 1 hr before start of the resting-state measurement. Additionally, possible drug use was tested with a urine drug screening for amphetamine, barbiturate, benzodiazepine, cannabinoids, cocaine, and opioids. Subjects with a positive test result were excluded from the study. Study subjects and the involved clinicians and researchers were "blinded." For both investigations, the subject stayed overnight in a clinical research unit (CRU) of the clinical research organization FOCUS Drug Development GmbH (Neuss, Germany), then underwent the drug-challenge investigation with simultaneously acquired fMRI/EEG during the following day and leave the CRU after 24 hr under medical supervision. The randomization was carried out prior by the statistical staff of the CRU using block randomization with blocks of size 4.
For details of drug application and fMRI/EEG study design, see Figure 1. In short, either subanesthetic S-ketamine (Ketanest ® S, Pfizer Pharma PFE, Berlin, Germany) in 0.9% NaCl or saline (0.9% NaCl) was intravenously administered before and during the MR scan. As of its pharmacokinetics, immediately before starting the MR measurement a bolus of 0.1 mg/kg S-ketamine (or equal volumes of saline) was administered over 5 min. For 1 min after MR scan initiation, the infusion was stopped to reach equilibrium of the ketamine plasma levels.
Afterward, the infusion was continued with 0.015625 mg kg −1 min −1 (max. 1 hr). Since ketamine slowly increases when it is constantly administered (Feng et al., 1995;Umbricht et al., 2000), a dosage reduction of 10%/10 min was used to maintain stable ketamine plasma levels during the experiment. Out of 24 recruited subjects, seven did abort the measurements before the resting-state measurements could be performed (for sociodemographic details and reasons for exclusion, see Table 1).
Simultaneously to the functional MR scans, an EEG recording was conducted. An MR compatible EEG setup (Brain Products, Gilching, Germany) was applied including a standard EEG cap (BrainCap MR, EasyCap GmbH, Breitbrunn, Germany) with 32 electrodes (distributed according to 10-20 system, including electrocardiogram attached to the subjects' back and electrooculogram [EOG], attached on the outer canthi of the left eye). EEG data were recorded with a sampling rate of 5 kHz and online bandpass filtered with 0.016-250 Hz. Overall impedances of the recording electrodes were <10 kΩ.

| rsfMRI analysis: Seed-based DMN functional connectivity
For data analysis, the acquired MR data were processed using Matlabbased CONN connectivity toolbox V17.f (Gabrieli Lab, Massachusetts; Whitfield-Gabrieli & Nieto-Castanon, 2012). To reach equilibrium of the spin history, the first five scans of each individual session were discarded. Since three rsfMRI/EEG measurements had to be terminated prematurely because subjects felt uneasy, only data of the first 8.5 min (150 volumes) of the resting-state session were used for analysis. This is at the lower edge of what has been recently recommended by Birn et al. (2013) as the best length for rsfMRI sessions. It takes into account that a longer duration of restingstate scanning per se can be associated with functional connectivity fluctuations related to changes in subjects state of vigilance (Tagliazucchi & Laufs, 2014). Preprocessing included the following steps in identical order: realignment, slice-timing correction, outlier identification via a scrubbing process (using Artifact Detection

| EEG analysis: Preprocessing
EEG preprocessing was done with BrainVision Analyzer 2.1 Professional.
The average referenced EEG data were MR and pulse artifact corrected using a sliding average method, down sampled to 500 Hz and filtered (zero phase shift Butterworth filters, 0.53-45 Hz, notch filter of 50 Hz, see Result 1.2 in Supporting Information). In line with rsfMRI processing, EEG data of the first 8.5 min were used for analysis in this article. The continuous EEG data were segmented into 2 s segments with an overlap of 1 s to minimize potential filter artifacts and to achieve a minimal frequency of 0.5 Hz. Since the EOG electrode data were not stable for all subjects, we introduce virtual vertical EOG and horizontal EOG channels by pooling electrode Fp1 and Fp2, or T7 and T8, respectively, and those were used for ocular correction via an ocular artifact-related subtraction method (Gratton, Coles, & Donchin, 1983) implemented in BrainVision Analyzer 2.1. For further artifact correction, a maximal difference of 210 ± 27 μV for a 1 s data interval was used as threshold.
This amounted to 352 ± 106 segments for EEG power analysis.
The experimental design including scanning sequences and information about drug application. During the neuroimaging investigation, subjects received in random order either a subanesthetic dose of S-Ketamine HCl (Esketaminehydrochlorid) [Ketanest ® S, Pfizer] or placebo (crossover design, 1 week apart), both administered intravenously in 0.9% NaCl. This administration was carried out as a bolus of S-Ketamine (0.1 mg/kg during 5 min) immediately before measurements in the MRI scanner and a continuous infusion of S-Ketamine (0.015625 mg kg −1 min −1 for the duration of the investigation, i.e., 1 hr maximum) during the measurement. To avoid a slow increase of ketamine plasma levels, a reduction by 10% every 10 min (Umbricht et al., 2000) was determined. The resting-state fMRI measurement started 34 min after the beginning of the ketamine infusion T A B L E 1 N = 17 study subjects with complete data sets Mean age (median) ± SD (years) 27.5 (28.0) ± 4.6 Mean height ± SD (cm) 178.2 ± 6.3 Mean weight ± SD (kg) 76.8 ± 9.1 Mean BMI ± SD (kg/m 2 ) 24.2 ± 2.8 Note: Out of 24 eligible healthy male volunteers, five subjects were excluded because of incomplete data sets. Furthermore, two subjects did not finish the measurements because of adverse events (one subject with tachycardia, one subject with panic attack). Of the remaining cohort, seven subjects were smokers and 16 subjects were right-handed. Abbreviation: BMI, body mass index.

| Statistical analyses
This is an exploratory study (placebo-controlled, randomized, crossover) to understand subanesthetic ketamine effects on brain function-in particular on the relationship between drug-induced changes of rsfMRI DMN connectivity and vigilance as assessed by EEG power during resting-state condition with simultaneous rsfMRI/EEG data acquisition. The major hypothesis in this study is that a ketamine-induced reduction of vigilance (comparable to sleep stage 1) is related to a decrease of frontal rsfMRI DMN connectivity.

| Seed-based functional MRI connectivity analysis
For seed-based analysis, using the Harvard Oxford Atlas, brain maps of bivariate correlation coefficients were calculated, by correlating the time-course of the seed region, that is, the filtered data from PCC/precuneus area, a key region of the DMN (Uddin, Kelly, Biswal, Castellanos, & Milham, 2009;Utevsky, Smith, & Huettel, 2014) defined as a sphere of 10 mm radius centered at coordinate 1, −61, Fisher Z-transformed functional connectivities for placebo (black) and ketamine (red) condition for the different significant clusters. Of further interest, for ketamine condition, mPFC connectivities were close to zero. Same could be observed for left and right IPL connectivities for placebo condition (Gasser, Cher, & Mocks, 1982) and differ by scaling factors (Zacharias, Sieluzycki, Kordecki, König, & Heil, 2011). Therefore, geometric means (GM) over subjects were computed for both conditions and the resulting GM FFT P were plotted as heatmaps on a two-dimensional surface model of the electrode position. For the same reason, we divided GM FFT P for ketamine condition by GM FFT P for placebo condition and plotted them as heatmap, to get a distribution of scaling factors over the head surface. When analyzing individual data, we account for the non-normal distribution of FFT P with the use of a logarithmic transformation (ln) (Gasser et al., 1982). For descriptive comparison of ketamine versus placebo, FFT P across electrodes and EEG frequency bands were calculated (p = .05, uncorrected).

| Relationship between rsfMRI and EEG FFT P
We calculated linear regressions (Pearson's R) of the DMN functional connectivity against EEG power (logarithmic FFT P ) to show possible relationships between DMN (rsfMRI) activity and neuronal activation (EEG) both in the placebo and ketamine data sets. We restricted our regression analyses (rsfMRI vs. EEG) to comparable brain regions (e.g., frontal DMN connectivity vs. frontal electrode positions).
We performed regression analyses using (a) statistically significant functional connectivities from 2.7 and (b) a corresponding electrode position with the maximum scaling factor of EEG power from 2.8.
Besides that, we looked for correlating changes of DMN activity with changes of EEG activity when ketamine was applied. Therefore, the individual differences of ketamine condition minus placebo condition (Δ) (difference values both for EEG power and rsfMRI connectivity) between the two modalities were compared.

| Seed-based functional connectivity MRI
The connectivity patterns in the placebo and ketamine condition are shown in the Supporting Information (Result 1.1, Figure S1). Figure 2a F I G U R E 3 Row 1 and 2 show heatmaps of geometric mean FFT P for all 30 electrodes of placebo and ketamine condition for five frequency bands (delta: 0.53-4 Hz; theta: 4-8 Hz; alpha: 8-12 Hz; beta: 12-25 Hz; gamma: 30-50 Hz). Row 3 shows heatmaps of factors of how the GM FFT P of the two conditions scale to each other (red = higher GM FFT P for ketamine condition, blue = higher GM FFT P for placebo condition, green = equal GM FFT P ). Please note logarithmic scaling for row 1 to 3. Row 4 shows heatmaps of uncorrected p values <.1 (one-way repeated measure ANOVA, blue p < .1, green p < .05, red p < .001). Despite of some electrodes, in alpha and beta, no clustering of electrodes shows neither high scaling factors, nor significant differences between conditions. In delta and theta, cluster of electrodes with high scaling factors and significantly different (p < .05) FFT P over conditions were observable for parietal-temporal regions and fronto-central regions, respectively. Highest scaling factors could be seen in fronto-central electrodes for gamma but without significant difference (p < .05) between conditions shows FWE-corrected significant connectivity clusters of functional connectivity changes for the contrast ketamine versus placebo. In the mPFC Even so, especially frontal electrode positions (Fz) showed the highest scaling factor (2.7) of our analyses across frequency bands. Distant from the frontally pronounced numerical increase of gamma activity with high scaling factor, statistical significance was reached at a right F I G U R E 4 Left side shows b-spline smoothed mean logarithmic FFT P of placebo (black) and ketamine (red) condition of electrode Fz (a) and Pz (b) against frequency bins from 0 to 45 Hz (bin-size = 0.48) with regarding 95% confidence interval (light black and light red, respectively). For both electrodes, overlapping graphs could be seen for frequency ranges from 8 to 25 Hz indicating no ketamine related differences. For Fz electrode, the graphs start to spread open at frequencies higher 25 Hz with higher ln(FFT P ) for ketamine condition. Same could be seen for both electrodes for frequencies between 1 and 7 Hz. Please be aware that for both electrodes, the energy content of small frequencies is at least one magnitude higher than those from high frequencies. For illustrative reasons and to account for the higher energy content of small frequencies, we plotted on the right side of (a) and (b) for the corresponding electrodes the mean FFT P of both conditions against a logarithmically scaled frequency axis. This leads to an emphasis on the difference between ketamine and placebo conditions at lower frequencies parietal electrode (P4). Apart from a single electrode in the left temporal region with a statistically significant difference of beta FFT P , statistically significant differences for alpha and beta FFT P were not seen.
Furthermore, we observed an energy shift of FFT P to slower frequencies within the alpha frequency band in the ketamine condition (see Figure S4). (e, f) Analyzing ln(FFT P ) of gamma for electrode Fz against DMN functional connectivity of mPFC, comparable results with a significant negative correlation for Δ analysis and a trendwise negative correlation between ln(FFT P ) and functional connectivity for placebo condition could be seen. Because of the aforementioned floor effect of ketamine related DMN functional connectivity of mPFC, a correlation between ln(FFT P ) and DMN functional connectivity for ketamine condition could not be observed In Figure 4, the group mean spectral power distribution, including 95% confidence intervals, is displayed both for the placebo and ketamine condition. It is shown (a) how absolute EEG power values are distributed in the frontal and parietal region and (b) how the EEG frequency spectrum is shifted in these brain regions between placebo and ketamine. Overall, a shift of energy to slow and fast wave frequencies is seen in the ketamine condition. For a more in-depth EEG signal analysis of ketamine effects, see Results 1.3-1.5 in Supporting Information.
3.3 | Ketamine-induced electric activity changes correlate with corresponding BOLD activation during rest

| DISCUSSION
This is the first simultaneous rsfMRI/EEG study to assess the effects of ketamine. In this randomized, double-blind, placebo-controlled, Overall, our findings suggest that subanesthetic ketamine effects on rsfMRI are related to the ketamine effect on vigilance. However, this ketamine effect on vigilance comes together with opposite effects on cortical network synchronization as indicated by the ketamine effects on rsfMRI functional connectivity. Thus, our findings are compatible with the notion that hyperarousal together with increased functional connectivity during clinical depression is counteracted by the antidepressant compound ketamine through its diminishing effect on EEG vigilance and associated functional connectivity in the PFC. DMN functional connectivity in the parietal cortex using PCC as the seed region was hardly studied in depression (Kaiser, Andrews-Hanna, Wager, & Pizzagalli, 2015). However, a very recent study of Evans et al. (2018) indicates that depressed patients may show decreased DMN functional connectivity in various brain regions, including parietal regions.
The findings obtained in this study are of quite some interest with regard to recent work highlighting the association of the antidepressant effect of ketamine, vigilance, and neuroplasticity. Duncan et al. (2014) reported that the effect of ketamine on depressive symptoms in treatment-responsive patients is accompanied by increased slow wave EEG activity during night sleep (increased slow wave activity during sleep is regarded as a marker for sleep depth according to Rechtschaffen & Kales, 1968). Duncan et al. (2014) further reported proportionally increased plasma levels of BDNF (peripheral marker of plasticity) and increases of EEG slow wave activity. The authors discussed these findings in the context of comparable findings in earlier studies in rats (Feinberg & Campbell, 1993) as well as rat and human studies (Esser, Hill, & Tononi, 2007;Riedner et al., 2007;Vyazovskiy, Cirelli, Pfister-Genskow, Faraguna, & Tononi, 2008) (Tüshaus et al., 2017). During sleep, they reported about a negative correlation in the PFC between (decreased) blood flow and (increased) slow wave EEG activity whereas in the posterior parts of the brain, a positive correlation with increased blood flow and increased slow wave EEG activity was found. Accordingly, it is conceivable that both during ketamine application and sleep corresponding neuroplastic changes occur in the brain involving deactivation of the PFC and activation of parietal (and occipital) association cortices.

| Ketamine affects slow wave EEG power and DMN functional connectivity
In our study, we could show increased slow wave delta and theta activity and a marginal increase of fast wave gamma activity during ketamine administration. Overall, this is in line with recently reported findings of Muthukumaraswamy et al. (2015) who observed comparable changes in frequency content after ketamine administration, with increases in frontal theta (slow wave) and overall gamma activity (fast wave). This seemingly contradicting findings of increased slow wave activity on one hand and fast wave activity on the other hand is best explained in the context of ketamine effects on vigilance regulation, that is, increased slow and fast wave frequency activity is seen during sleep stage 1-a dynamic transition state between wake and sleep which is characterized by CAPs (Parrino et al., 2014). In the present study, we showed decreased functional connectivity for frontal parts of the DMN. This is in line with recently published rsfMRI studies on the effects of (sub-)anesthetic ketamine on human brain function. The study of Scheidegger et al. (2012)  of functional connectivity appears to normalize transiently 1 hr after drug application, decreasing again after 24 hr, which in turn is associated with higher Gln/Glu changes (Li et al., 2018). Interestingly, Bonhomme et al.
(2016) found a significant negative linear relationship between the depth of sedation (with increasing ketamine doses) and mPFC DMN connectivity. In this context, a study of Purdon et al. (2013) needs to be mentioned showing that with increasing depth of anesthesia, the power of low frequency EEG power is increasing in the frontal brain region. This again is in line with our observations suggesting a relationship between decreased functional connectivity and a reduced level of vigilance following ketamine administration.

| Slow wave EEG power and network connectivity are related
To our knowledge, we could show for the first time a negative relationship between frontal DMN functional connectivity and frontal theta power. The negative correlation of frontal DMN functional connectivity with frontal theta (and gamma) power for placebo condition gets lost for ketamine condition (see Figure 5). This is of some interest with regard to pharmacodynamic measurements of ketamine effects (or related drug compounds). The decrease of prefrontal functional connectivity after ketamine administration results in a floor effect of the functional connectivity, which means that the ketamine related effect levels out because functional connectivities are close to zero.
Accordingly, one would prefer EEG rather than fMRI measurements when it is the aim to measure ketamine effects. On the other hand, we also could show a positive relationship of bilateral IPL connectivity to PCC and slow wave delta activity. Here, a less restricted range of values is seen for parietal functional connectivities. In any case, for pharmacodynamic studies, it appears preferable to use both EEG and fMRI (ideally with simultaneous data acquisition). In some cases, EEG is the more sensitive tool with a wide range of measurement values.
In fact, EEG-informed fMRI analysis may even help to improve the sensitivity of fMRI to detect drug effects as recently demonstrated by us for nicotinic compounds (Warbrick et al., 2012). In addition, using both data sets together (EEG and fMRI) may help to build confidence in any observed drug effect when a consistent picture is emerging across modalities-most notably in small studies with limited statistical power such as in clinical Phase-I or Phase-II studies.
As a limitation of this study, one has to mention, that our study sample-as typical in the framework of a clinical Phase-0/Phase-I study (proof-of-concept)-only included healthy young men. Thus, the relevance of the bolus-infusion protocol to antidepressant action could not be investigated and might be of interest for future studies.
Furthermore, ketamine affects other physiological parameters like heart rate variability or blood pressure, too. For example, Komatsu et al. (1995) showed that ketamine reduces heart rate variability.
However, recently Chang et al. (2013) showed that heart rate variability is associated with functional connectivity of particular brain regions. This is accompanied by a study of Jennings, Sheu, Kuan, Manuck, and Gianaros (2016) who could show that a reduced heart rate variability is correlated with reduced DMN connectivity of mPFC.
Thus, based on the known association of vigilance reduction and decreased heart rate variability (Penzel, Kantelhardt, Lo, Voigt, & Vogelmeier, 2003), these results support our findings that ketamine effects on vigilance can be measured with simultaneously measured fMRI/EEG. Another difficulty in rsfMRI measurements with continuous ketamine infusion is the fact that with longer measurement time, subjects tend to move or start to feel unwell. Both might be related to the lack of a task and an individual increased focus on the inner self and one's own emotional state (Andrews-Hanna, 2012) after ketamine administration. We have accounted for this by reducing our analysis on the first part of our 20 min lasting resting-state experiment, although a comparison analysis of first and second half experimental data shows no reliable difference.

| CONCLUSION
We could show a relationship of altered EEG activity with DMN connectivity changes consistent with the notion of a ketamine-induced state of decreased vigilance (sleep stage 1 like). Our findings can be used as a mechanistic neurophysiological model to explain the antidepressant action of ketamine and related drug compounds. By extension, this neurophysiological model based on simultaneous fMRI/EEG may also provide a heuristic framework to improve our understanding of the effects of ketamine and related drugs on brain function in comparable pharmacofMRI studies (Mehta et al., 2018).

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.