Cerebellar theta-frequency transcranial pulsed stimulation increases frontal theta oscillations in patients with schizophrenia

Abstract

Cognitive dysfunction is a pervasive and disabling aspect of schizophrenia without adequate treatments. A recognized correlate to cognitive dysfunction in schizophrenia is attenuated frontal theta oscillations. Neuromodulation to normalize these frontal rhythms represents a potential novel therapeutic strategy. Here, we evaluate whether noninvasive neuromodulation of the cerebellum in patients with schizophrenia can enhance frontal theta oscillations, with the future goal of targeting the cerebellum as a possible therapy for cognitive dysfunction in schizophrenia. We stimulated the midline cerebellum using transcranial pulsed current stimulation (tPCS), a non-invasive transcranial direct current that can be delivered in a frequency specific manner. A single 20-minute session of theta frequency stimulation was delivered in 9 patients with schizophrenia (cathode on right shoulder). Delta frequency tPCS was also delivered as a control to evaluate for frequency specific effects. EEG signals from midfrontal electrode Cz were analyzed before and after cerebellar tPCS while patients estimated the passage of 3 and 12 second intervals. Theta oscillations were significantly larger following theta frequency cerebellar tPCS in the midfrontal region, which was not seen with delta frequency stimulation. As previously reported, patients with schizophrenia showed a baseline reduction in accuracy estimating 3 and 12 second intervals relative to control subjects, which did not significantly improve following a single session theta or delta frequency cerebellar tPCS. These preliminary results suggest that single session theta frequency cerebellar tPCS may modulate task-related oscillatory activity in the frontal cortex in a frequency-specific manner. These preliminary findings warrant further investigation to evaluate whether multiple sessions delivered daily may have an impact on cognitive performance and have therapeutic implications for schizophrenia.

Introduction

Schizophrenia is a severe, chronic, and disabling mental illness that affects 1% of the U.S. population (McGrath et al., 2008). Deficits in cognitive and executive function include impairments in working memory, attention, planning and timing, which are currently untreatable and significantly decrease quality of life (Andreasen et al., 1998; Carroll et al., 2009; Green and Harvey, 2014; Ho et al., 1998; Ward et al., 2011). A biological correlate to the cognitive impairment seen in schizophrenia is electroencephalographic (EEG) abnormalities in the frontal cortex (Andreasen et al., 1997; Carter et al., 2014; Parker et al., 2017). Previous work demonstrated dysfunctional working memory and executive function in subjects with schizophrenia is associated with abnormal activity within a functionally connected network including the temporal lobe and cerebellum (Andreasen and Pierson, 2008; Repovs et al., 2011).

Work from our lab has shown that patients with schizophrenia have attenuated frontal low-frequency rhythms concurrent with impaired interval timing performance and the two are correlated, such that greater attenuation of theta rhythms corresponds with worse task performance (Parker et al., 2017). A similar profile of underestimating the passage of time and attenuated theta oscillations follows D1 dopamine antagonist infusions in a region of the rodent frontal cortex (Parker et al., 2017). This parallels a finding in human schizophrenia patients: there is a homologous area in the frontal cortex of patients with schizophrenia reported to have a decreased binding potential for D1 dopamine (Abi-Dargham et al., 2002), lending support to the importance of this mechanism for timing performance. Further, we showed that optogenetic stimulation of another node in the network recruited for timing performance, the cerebellum, could rescue theta frequency oscillations in the frontal cortex and normalize timing performance (Parker et al., 2017).

Currently, there are no therapies that reliably improve cognitive dysfunction in schizophrenia; new therapeutic options are desperately needed to mitigate the disabling burden of this disease. Although traditionally associated with motor function, there is also a prominent role of the cerebellum in cognitive function (Koziol et al., 2013; Schmahmann, 1996, 1998; Schutter and van Honk, 2006; Stoodley, 2011; Strick et al., 2009). Previous clinical and neuroimaging reports have provided evidence of cerebellar pathology in schizophrenia patients with cognitive deficits and psychosis (Jurjus et al., 1994; Nopoulos et al., 1999; Sandyk, 1993; Tavano et al., 2007), which may be involved in abnormal cerebello-thalamo-cortical circuitry concurrent with cognitive and motor deficits (Andreasen et al., 1996; Rüsch et al., 2007). According to the traditional view, discrete circuits between the basal ganglia and cerebellum with motor areas of the cerebral cortex underlie motor performance, while other circuits between the basal ganglia and cerebellum functionally linked with prefrontal cortices associate more strongly with cognitive functions (Middleton and Strick, 2000). There is emerging evidence that the cerebellum can be a neuromodulation target not only for motor control but also for cognitive abnormalities in patients (Ferrucci et al., 2016; Grimaldi et al., 2014).

Although neuromodulation (such as transcranial direct current stimulation; tDCS) of frontal cortical areas in healthy participants and patients may enhance cognition, stimulation of the cerebellum has received less attention (Fregni et al., 2005; Jo et al., 2009). In 2008, however, Ferrucci et al., reported changes in cognitive performance in healthy individuals after cerebellar tDCS (Ferrucci et al., 2008). Based on these findings, one could hypothesize that cerebellar stimulation may be able to rescue frontal theta brain rhythms essential for normal cognition, thus correcting physiologic abnormalities that impair function and performance. Indeed, emerging evidence suggests that cerebellar stimulation may hold promise for improving cognitive problems in schizophrenia (Demirtas-Tatlidede et al., 2010; Garg et al., 2016), though it is not known whether this is mediated by enhanced frontal theta rhythms. We aim to extend this work by evaluating whether noninvasive approaches of cerebellar neuromodulation can enhance cognitive function by modifying frontal brain oscillations in patients with schizophrenia.

Here, we investigate whether noninvasive transcranial cerebellar stimulation, specifically transcranial pulsed current stimulation (tPCS), influences frontal cortex theta oscillations. TPCS is a relatively safe and inexpensive way to pass electrical current non-invasively through the skull in a frequency specific manner that may influence neuronal activity and connected brain circuitry (Vasquez et al., 2016). Using this methodology, we tested the hypothesis that theta-frequency cerebellar tPCS stimulation would 1) augment midfrontal theta frequency oscillations in patients with schizophrenia and 2) influence timing performance. We also tested the effect of delta frequency cerebellar tPCS to evaluate whether any changes in frontal oscillations were frequency-specific relative to the stimulation. Results showed that theta but not delta frequency cerebellar tPCS rescued low-frequency activity in the midfrontal region, yet neither frequency of stimulation improved timing performance.

Materials and Methods

Human Subjects

Nine patients with a DSM-IV diagnosis of schizophrenia (7 men, 2 women) were recruited from the Iowa Longitudinal Database. Subjects were recruited from a database of patients that had diagnoses confirmed by a board-certified psychiatrist at the University of Iowa. Data from the nine age-, sex-, and education-matched healthy control subjects performing the interval timing task without stimulation, were included for comparison (Parker et al., 2017). All participants were determined to have the decisional capacity to provide informed consent, resided within 100 miles of Iowa City and were able to independently travel to the University of Iowa Hospitals and Clinics. Written informed consent was obtained from every subject and all research protocols were approved by the University of Iowa Human Subjects Review Board. Medication status was not altered for any of the patients. Subject demographics and scores on cognitive tasks are described in Table 1.

Table 1.

Demographics and cognitive scores for patients with schizophrenia and controls.

	Control	Schizophrenia
		Pre-tPCS	Post-delta tPCS#	Post-theta tPCS#
Sex	9 (3 F)	9 (3 F);	8 (3 F)	9 (3 F)
Age	47.9 (6.8)	45.8 (2.9)	45.2 (3.1)	46.1 (2.2)
Education (Years)	15 (0.6)	12.8 (0.9)	12.9 (0.76)	12.9 (0.7)
MOCA	28 (0.6)	23.8 (1.5)*	24.7 (0.8)	24.4 (1.2)
TMT	26.9 (3.4)	81.9 (28.7)*	90.0 (20.1)	59.2 (15.1)
VF	45.6 (3.1)	29.2 (4.9)*	29.9 (4.4)	29.6 (2.6)
Digit	20.4 (1.1)	14.7 (1.4)*	15.2 (1.1)	14.3 (0.8)

The data are presented as mean +/− SEM.

^*p<0.05 Control vs Schizophrenia Pre-tPCS.

^#Following stimulation in either delta or theta frequencies, all cognitive measures remained significantly different from controls, i.e., there was not an improvement in their cognitive function. All controls were age and sex matched to a patient.

MOCA = Montreal Cognitive Assessment, TMT = Trails Making Test, VF = Verbal Fluency.

Interval-Timing Task

Timing is a well-characterized capability of the cerebellum; however, it is typically associated with sub-second timing (Ivry and Spencer, 2004). Our lab has recently shown that the cerebellum and frontal cortex participate in supra-second processing and timing (Parker et al., 2017). Previous studies have demonstrated that estimating the passage of time is a reliable way of eliciting frontal theta activity (Parker et al., 2015). While EEG recordings were being acquired, participants performed an interval timing task before and after cerebellar tPCS. The interval timing task involved the appearance of a “3 sec” or “12 sec” text cue on the computer screen that indicated both the interval of time to estimate and the start of the trial (Fig. 1). The subject would make a motor indication of their estimation of the elapse of the specified timing duration by pressing the keyboard space bar. Participants received feedback about their response time after 20% of the trials. The task was self-paced, and the participants were instructed not to count time in their head during the task. The interval-timing task contained 80 total trials with 3 s and 12 s intervals presented in random order.

Fig. 1. Experimental Timeline.

Delta and theta stimulations were randomized in visit 1 and 2.

Each subject’s time estimates for the two intervals were fit with Gaussian distributions, using Matlab (fitdist.m), and performance was quantified two ways using the mean and standard deviation (Rakitin et al., 1998). First, we subtracted the mean from the actual instructed interval, providing a measure of timing accuracy. Larger and smaller values correspond to systematic under- and overestimation of time, respectively. While subjects can be highly accurate with respect to average response times, their estimates can vary substantially from trial-to-trial. Therefore, we also measured timing precision. Timing precision (taken as the standard deviation of response times) decreases linearly with the interval being timed, a form of Weber’s law often referred to as the ‘scalar property’ of interval timing (Gibbon et al., 1984). Therefore, we evaluated precision by dividing the standard deviation by the mean response time, defined as the ‘coefficient of variation’ (CV). Larger and smaller values correspond to lower and higher precision, respectively. Furthermore, in both humans and a variety of other species, the coefficient of variation remains constant, regardless of the interval being timed (Buhusi and Meek, 2005).

Previous findings in healthy controls indicate that midfrontal theta activity (4-8 Hz) following trial start is essential for accurate timing performance (Parker et al., 2014, 2015, 2017). This cue-locked theta activity is attenuated in patients with schizophrenia and Parkinson’s disease (Parker et al., 2015, 2017). Based on these results, a time-frequency Region of Interest (tf-ROI) was derived from our previous work and was constrained to 0 to 1.0 s following cue. In addition to this highly-implicated a priori tf-ROI, we applied cluster-based permutation correction on the full time-frequency analysis in order to identify any other consistent changes between groups. This method involved thresholding the size of the statistical cluster (for 1000 permutations of group labels) and picked the one-dimensional cluster mass (at the 95^th percentile) as the threshold for chance occurrence.

Cerebellar tPCS

Figure 1 displays the experimental setup. A single session of noninvasive brain stimulation was delivered via a battery-driven stimulator (MindAlive, Inc., Oasis Pro) with anodal pulsed current stimulation of the cerebellar vermis applied via surface conductive rubber electrodes. The vermal location was defined by the location 1 cm below the external occipital protuberance or highest and point of the largest projection in the occipital bone is referred to as the inion (Gülekon and Turgut, 2003; Figure 5). The cathodal electrode was placed on the right shoulder (Bocci et al., 2015, 2018; Ferrucci et al., 2012, 2013). Stimulation duration was 20 minutes, with 1 mA peak-to-peak amplitude at either delta (n=8) or theta (n=9) frequencies. Delta frequency cerebellar tPCS was used as a control to evaluate the frequency-specific effects of cerebellar modulation of the frontal cortical oscillations during timing performance. This may be optimal as an “active sham”; stimulation causes a “buzzing” or “tingling” sensation at the stimulation site, which would be absent with a purely sham stimulation condition (van Driel et al., 2015). Additionally, using delta frequency cerebellar tPCS as a control frequency confirmed that general characteristics of physical stimulation were comparable across all conditions. Delta and theta stimulation session order was randomized and separated by 1-3 months. Although we did not test the stimulator before we applied stimulation, we did record EEG activity during stimulation and could see the artifact induced by stimulation to verify that stimulation was being delivered at the correct frequency.

Fig. 5. Cue-evoked activity at cerebellar electrodes is not modulated by cerebellar theta-frequency tPCS.

A) There was no effect on cue-related activity at the site of cerebellar stimulation as measured by the midline electrode following delta tPCS on 3 s and 12 s trials (left side). Theta tPCS significantly increased cue-related activity in theta frequencies (4-8 Hz) on 12s trials (black outlined; p<0.05) while 3 s trials only trended to increase (p>0.05) (right side). B) Similar to cerebellar midline electrode, delta frequency stimulation did alter cue-related theta frequency activity at the right cerebellar electrode. Theta-frequency tPCS significantly increased theta-frequency activity for 3 s trials (black outlined; p<0.05) while there was only a trend for 12 s trials. C) The electrode above and to the right of the site of cerebellar stimulation did not show changes in theta frequencies after delta or theta tPCS. Increased power in red with permutation-corrected statistical significance p<0.05 outlined in black lines.

EEG Recording and Analysis

EEG recording was collected before (“pre”) and after (“post”) cerebellar tPCS. EEG was recorded on a Nihon Kohden system with a sampling rate of 500 Hz (Cohen, 2014). EEG was recorded from 21 channels based on the 10-20 system (Fz, Cz, Pz, F3/4, C3/4, P3/4, F7/8, T3/4, T5/6, O1/2, M1/2), as well as left-eye vertical electrooculography (VEOG) and ground (forehead). The lead that is traditionally placed at the FP1 location was relocated to 1 cm below the inion bone on the cerebellar midline while FP2 was placed 1 cm to the right of FP1. Traditional 10 mm EEG type passive electrodes were used to collect signals from the scalp. We applied EEG conductive gel (Covidien 30806734 Ten20) to hold the electrodes in place to conduct the signal with higher signal to noise ratio. This approach was selected to match our previous EEG datasets that described differences in low-frequency rhythms between patients with Parkinson’s disease and controls (Parker et al., 2015). Impedance of all electrodes was below 5 kΩ. Continuous data were re-referenced to the mathematical average of the two mastoid channels, yielding a total of 21 scalp EEG channels. Signals were segmented on the basis of the stimulus (cue) onset (−2 to 6 s for 3 s trials; −2 to 18 s for 12 s trials), from which the cue-locked segments were isolated. Eye blinks and horizontal eye movements were removed by hand using independent component analysis and EEGLab (Delorme and Makeig, 2004). Afterwards, EEG signals were then re-referenced to an average reference. Previous studies have associated cognitive impairment with changes in midfrontal regions; therefore, we selected midfrontal Cz electrode for the main analysis (Parker et al., 2014, 2015, 2017). We also analyzed the midline cerebellar electrode, the right cerebellar electrode, and the electrode above the right cerebellar lead to evaluate how cerebellar delta\theta tPCS influenced activity at and around the site of stimulation.

Power spectral analysis of “pre” and “post” cerebellar tPCS EEG signals was computed from the “pwelch” method. Signals were transformed into the power spectrum domain (using pwelch method: 256 points window size). Furthermore, a frequency range of 1-50 Hz was selected to compute relative power spectrum to abate the inter-recording variation. Here, we exported the mean relative power at delta (1-4 Hz) and theta (4-8 Hz) frequency bands.

Time-frequency analysis was computed by multiplying the fast Fourier transformed (FFT) power spectrum of single trial EEG data with the FFT power spectrum of a set of complex Morlet wavelets. These complex Morlet wavelets are defined as a Gaussian-windowed complex sine wave: e^i2πtfe^{−t^2/(2xσ^2)}, where “t” is time and “f” is frequency (which increases from 1-50 Hz in 50 logarithmically spaced steps). This equation defines the cycle of each frequency band, increasing from 3 to 8 cycles between 1 and 50 Hz and taking the inverse FFT. Ultimately, this computational method converts signal into time-domain convolution and results in estimates of instantaneous power (the magnitude of the analytic signal). The power value was normalized by conversion to a decibel (dB) scale (10*logl0(power_t/power_baseline)), allowing a direct comparison of effects across frequency bands (Cohen, 2014). The baseline for each frequency was calculated based on average power from −0.3 to −0.2 s prior to the onset of the stimulus. Each epoch was then cut in different lengths for visualization purpose (3 s full trial: −2 to 5 s; 12 s full trial: −2 to 14 s; also, around 3 s and 12 s cue: −0.5 to 1.0 s time from cue). Time-frequency plots were analyzed from electrode Cz in delta (1-4 Hz) and theta (4-8 Hz) frequency bands in accordance with well-established prior hypotheses (Parker et al., 2014, 2015).

Statistical Analyses

For binary comparisons between “pre” and “post” tPCS, we used paired t-tests with an alpha level of 0.05. Paired t-tests were performed to compute the statistical differences before (“pre”) and after (“post”) the tPCS for behavioral, clinical data measures and task-related frequency bands power. We applied repeated measure analysis of variance (ANOVA) followed by pairwise comparison using SPSS for spectral power in the targeted ROIs (tf-ROI) of 3 s and 12 s data for delta and theta tPCS separately. Spearman correlation analysis was performed to analyze the correlation between clinical scores and reaction time/tf-ROIs.

Results

Midfrontal Activity and Behavioral Response in Control Subjects

Healthy control subjects have increased delta frequency power in the midfrontal electrode Cz at the onset of cue in the interval-timing cognitive task (Fig. 2A; Parker et al., 2017a(Parker et al., 2017)). However, patients with schizophrenia have attenuated midfrontal theta frequency power around the time of cue (Fig. 2B). Patients with schizophrenia also had significantly increased variation in the time estimation (3 s and 12 s interval-timing tasks) as compared to control subjects, specifically in timing efficiency and the coefficient of variation (Fig. 2C-F). These results are previously published and are the foundation for the hypothesis that midfrontal activity is necessary for timing performance on an interval timing task (Parker et al., 2017).

Fig. 2. Spectral analysis of electrode Cz signal in healthy control and schizophrenia patients during interval-timing tasks.

A and B) Time-frequency spectrograms show increased delta frequency power around cue in the interval-timing task in control subjects as compared to schizophrenia patients. C) Quantification of response histograms reveals variations in the measurements of timing, including D) less variations in average response time, E) more variations in timing efficiency as defined by the number of responses occurring around 3 s (2-3 s) and 12 s (11-12 s), and F) a larger coefficient of variation. These data have been described previously in (Parker et al., 2017). *p<0.05.

Midfrontal Activity in Patients

Spectral analysis of delta and theta EEG data was computed before and after tPCS to examine the effect of stimulation on midfrontal region Cz during performance of the timing task in patients with schizophrenia. Over the entire 12 second interval, delta (Fig. 3A and B) and theta frequency tPCS (Fig. 3E and F) did not significantly alter the relative power of delta and theta frequency bands in the midfrontal cortex at electrode lead Cz. This is further confirmed by time-frequency analyses of the midfrontal electrode that show no significant changes in power after delta tPCS subtracted from power before, across both the short (Fig. 3C) and long (Fig. 3D) intervals. Interestingly, theta tPCS significantly increased both delta and theta oscillatory activity around the time of the cue for both 3 s and 12 second trials (Fig. 3G and H - permutation-corrected statistical significance p<0.05 outlined in bold lines).

Fig. 3. Spectral analysis of electrode Cz signal before and after cerebellar tPCS during interval-timing tasks.

There are no significant differences in relative power delta and theta frequency bands at electrode Cz in the midfrontal cortex following delta (A and B) and theta frequency tPCS (E and F). C-H) Time-frequency spectrograms after theta-frequency tPCS show significant changes in power (increased in red with permutation-corrected statistical significance p<0.05 outlined in black lines) around 3 s (G) and 12 s (H) cue events at time 0 s (note that the x-axes are scaled and shortened to 5 seconds to include only the time of the trial so that the intertrial intervals and next trial starts are not included in the image for 3 s trials). Increased power was more prominent in lower frequency bands (1-8 Hz; delta/theta frequency band). This effect was specific for theta-frequency stimulation as delta-frequency tPCS did not significantly alter power at any timepoints throughout the interval on either the 3s (C) or 12s (D) task.

Our previous work reports the importance of increased activity immediately following the onset of the cue (0-0.5 s) for accurate timing performance (Parker et al., 2014, 2017). Here we find that delta frequency tPCS did not modulate the activity at 3 s and 12 s cue-related tf-ROIs at electrode Cz in patients with schizophrenia: 1-4 Hz, 0-0.5 s from cue, (see table 2; Fig 4 A and B) and 4-8 Hz, 0.5-1.0 s from cue (see table 2; Fig 4 A and C).

Table 2.

Repeated-measure ANOVA tests (within subject variables: 3 s and 12 s tasks) were applied on tf-ROI power values to compute the difference between pre- and post-delta or theta tPCS.

	tf-ROI: 0-0.5 s		tf-ROI: 0.5-1.0 s
	1-4 Hz	4-8 Hz	1-4 Hz	4-8 Hz
Delta tPCS	F = 0.01; p=0.99	F = 0.3; p= 0.6	F = 0.25; p= 0.6	F =1.9; p= 0.2
Theta tPCS	F =1.3; p= 0.3	F =1.3; p= 0.3	*F =10.9; p= 0.01	*F = 7.9; p= 0.02

^*p<0.05 represents statistical significance.

Fig. 4. Cue-evoked low-frequency activity increases significantly at tf-ROIs following theta-frequency tPCS.

A) tf-ROI 3 s and 12 s cue-related midfrontal activity (at Cz) did not modulate following delta tPCS around 3 s and 12 s cues (B) at 1-4 Hz, 0-0.5 s from cue, or (C) at 4-8 Hz, 0.5-1.0 s from cue. (D) However, tf-ROI 3 s and 12 s cue-related midfrontal activity (at Cz) increased around 3 s and 12 s cues after theta tPCS (Permutation-corrected statistical significance p<0.05 outlined in bold lines). E) Increased power was observed in delta frequencies (1-4Hz), 0-0.5 s from cue (non-significant; p>0.05), and (F) in theta frequencies (4-8 Hz), 0.5-1.0 s from cue (significant, *=p< 0.05). Although the analyses focused on electrode lead Cz, there were several significantly increased electrodes on the scalp topography of tf-ROIs following theta tPCS that could have been analyzed for both timepoints (E and F, left - Statistically significant electrodes p<0.05 indicated by large diamonds). Note, figure 4A and 4D are ‘zoom-ins’ of figure 3C, D, G and H.

Theta frequency tPCS also did not change the activity at 3 s and 12 s cue-related tf-ROIs 1-4 Hz, 0-0.5 s from cue (see table 2; Fig 4 D and E). However, theta frequency tPCS increased the activity significantly at 3 s and 12 s cue-related tf-ROIs 1-4 Hz and 4-8 Hz, 0.5-1.0 s from cue (see table 2; Fig 4 D and F). We also performed skewness tests on tf-ROIs to measure the normality. We found that the absolute values of skewness were <1m, which confirm data normality.

Notably the increased theta was not entirely consistent with that of previous reports from healthy controls as it was delayed in onset by 0.5 seconds. Although our analyses focused on electrode Cz, several electrodes showed significant power increases as indicated by the large black diamonds on the topo plots (p<0.05: Fig. 4 D and F). Interestingly, theta frequency tPCS reinstated theta oscillations (4-8 Hz) in the mid frontal region. Previous reports have confirmed the role of increased mid frontal activity at 4-8 Hz in the improvement of cognitive performance in patients with cognitive deficits (Cavanagh and Frank, 2014; Singh et al., 2018). Therefore, current results suggest the potential benefits of theta tPCS with improved stimulation parameters to induce behavioral changes in patients with schizophrenia and other diseases associated with cognitive impairment. This effect was specific to theta frequency as cerebellar delta tPCS did not induce any significant changes in frontal activity (Fig. 4 A-C).

Further, we analyzed the electrodes which were located at the site of stimulation (1cm below the inion bone), 1 cm to the right of the stimulation site, and a site to the right and superior to the stimulation location to evaluate the effects of tPCS on cue-evoked theta frequencies. Cue-evoked power at theta frequencies at the site of stimulation and 1 cm to the right of stimulation tended to increase after theta but not delta tPCS (Fig. 5A and B). Analyses of the electrode above and to the right of the stimulation site did not show any changes in power after delta\theta tPCS (Fig. 5C). These results indicate that lasting effects of theta cerebellar tPCS activity were localized only to the frontal cortex and there were no local persistent changes at the cerebellar site of stimulation.

Interval-Timing Performance in Patients

To further explore how the cerebellum may modulate the frontal cortex to support timing (and possibly cognitive function), schizophrenia patients performed an interval-timing task before and after delta (n=8) or theta (n=9) tPCS. At baseline, patients with schizophrenia underestimated the short interval (p=0.043) and overestimated the long interval, although the latter effect only had a trend towards significance (p=0.09). This resembles a pattern of regression to the mean referred to as a ‘migration effect’, which is also observed in Parkinson’s patients (Malapani et al., 2002). Further consistent with Parkinson’s data, subjects showed larger coefficients of variation (CVs = standard deviation / mean of the response times) for the short interval than the long, indicating that they did not conform to the ‘scalar property’ of interval timing. In other words, the variability of the response times did not grow linearly with mean response time, as one would expect constant CVs for both durations if this were the case. However, neither delta nor theta stimulation significantly altered these patterns (Fig. 6A-D).

Fig. 6. Single session delta or theta frequency cerebellar tPCS does not influence interval timing performance.

A) Delta frequency cerebellar stimulation did not influence interval timing on 3 s or 12 sec trials. This is indicated by similar average response histograms and overlapping standard error bands of 40 trials before (black) and 40 trials after (red) 20 minutes of stimulation. Quantification of these response histograms reveals no significant differences for measures of timing, including B) average response time, C) timing efficiency as defined by the number responses occurring around 3 s (2-3 s) and 12 s (11-12 s), and D) coefficient of variation. E) Likewise, theta stimulation did not influence interval timing performance as shown by similar average response histogram curves and as quantified by insignificant F) average response times, G) response time accuracy, and H) coefficient of variation.

Cognition in Patients

Although unlikely that a single session of tPCS could alter performance on core cognitive tasks, we compared pre-post tPCS baseline scores for cognitive measures including the Montreal Cognitive Assessment, Trail Making Task, Verbal Fluency, and Digit Span on 7 of the schizophrenia patients included in this study. There were no significant changes in cognitive function following a single session of tPCS (see Table 1).

Discussion

We demonstrate that twenty minutes of 1mA theta-range tPCS of the midline cerebellum can alter power in time-frequency spectrograms in the lower frequency (delta and theta) range during an interval timing task. This effect was not seen with delta-range tPCS, suggesting that this is a frequency-specific effect of tPCS. These changes were most notable in the Cz (mid-frontal) and prefrontal tf-ROIs, which is consistent with our hypothesis that midline cerebellar stimulation would alter frontal cortex EEG activity. This is also consistent with previous literature utilizing other forms of low-frequency cerebellar stimulation to modulate frontal EEG activity during a timing task (Parker et al., 2017). Some literature suggests that pulsed stimulation protocols are effective at entraining the dominant brain rhythm, which is often task- and region-specific. Indeed, there is some evidence that in other tasks, such as associative learning tasks, behavior is associated with theta synchronization between cerebellum and prefrontal cortex regions (Schutter and van Honk, 2006). Thus, it is possible that theta-range tPCS was effective when delta-range stimulation was not due to driving already-dominant cerebellar theta rhythms that are active and synchronized in the execution of a timing task (Parker et al., 2017). Additionally, although theta-range tPCS altered low-frequency EEG activity, why this was not specific to theta-range activity remains unclear. The localization of induced EEG changes to the midfrontal regions is promising as this is a region previously implicated in cognitive timing tasks and abnormal in patients with cognitive impairment (Parker et al., 2014, 2015, 2017).

Despite the tPCS frequency-dependent changes in time-frequency spectrograms noted in this study, there were no notable changes in task performance behavior. There are several possible explanations for the lack of behavioral modifications. First, it is possible that the stimulation parameters were not robust enough to drive a change in brain activity significant enough to modify behavior on a task that probes a core cognitive process like timing. Studies of tPCS show a duration-dependent effect of the stimulation, with 20 minutes of stimulation often leading to the longest-lasting effects (Vasquez et al., 2016). However, whether this is optimal for a clinical population of patients with schizophrenia with abnormal low-frequency frontal EEG rhythms is unknown. Although some studies will show EEG changes from a single session of stimulation lasting up to 50 minutes (Vasquez et al., 2016), other studies, especially those focusing on entrainment of specific EEG rhythms with brain stimulation, have shown that effects may last only minutes after the stimulation is stopped, or cease with stimulation offset (Ozen et al., 2010). It is possible that repeated stimulation sessions, or sessions with different stimulation parameters, could have induced longer-lasting or more robust EEG changes that may have ultimately led to behavioral changes. Future studies could explore this question with longer stimulation protocols.

The EEG findings identified in this study beg the question of the specific mechanism of action through which the cerebellar tPCS is altering frontal regions. The neurophysiologic data for cerebellar tPCS is limited. Transcranial magnetic stimulation (TMS) studies using paired associative stimulation paradigms have detected changes in motor cortex plasticity after cerebellar TMS, with pharmacological manipulations demonstrating that these effects are dependent on N-methyl-D-aspartate (NMDA) receptor activity. Transcranial direct current stimulation (tDCS) studies with similar anodal and cathodal placement to this study have modeled the electrical fields induced and suggest the posterior aspect of the cerebellum can be affected by the stimulation (Parazzini et al., 2014). Despite this evidence, some research has drawn into question whether a low electrical current, such as that utilized in this study (1mA) can actually induce electrical changes in the brain that would be clinically or physiologically relevant (Voroslakos et al., 2018). Although this remains an open question, it invites other potential physiologic explanations for the changes noted here, as well as in volumes of previous noninvasive electrical stimulation research. Another potential mechanism may include direct stimulation of peripheral nerves triggering a “bottom-up” central nervous system response to stimulation (Underwood, 2016). Whether peripheral or cranial nerve stimulation at a rhythmic frequency could entrain neuronal elements remains to be seen. Based on the organization of the cerebellum where the hemispheres slightly cover our targeted vermis, it is likely that stimulation influenced the hemispheres in addition to the vermis. The vermis was targeted based on previous reports of improvement of cognitive function using cerebellar transcranial magnetic vermal stimulation. Going forward, we will investigate the spread of stimulation and changes in the deep nuclei using electrophysiology in rodents to clarify the spread and pathways influenced by midline cerebellar stimulation.

Despite several unanswered questions, this study demonstrates that it is possible to use frequency-specific tPCS over the cerebellum to alter low-frequency brain rhythms during a timing task in a population of patients with schizophrenia. Although this did not lead to behaviorally-relevant changes in this patient sample, it should prompt future studies looking at methods for optimizing this modulatory effect. As other studies have demonstrated compromised cerebello-frontal activity in schizophrenia and that low-frequency stimulation of the cerebellum can “rescue” low-frequency deficits in the frontal cortex with associated behavioral improvements, this is a line of research worthy of further pursuit and exploration. Transcranial PCS itself is a highly understudied technology, and the research of its application to the cerebellum in healthy or diseased populations is sparse.

Future studies should include looking at different stimulation parameters (e.g. different current intensities, polarities, and/or electrode locations) or attempting to replicate these findings with other stimulation modalities, such as transcranial magnetic stimulation, which can also be pulsed in frequency-specific patterns (e.g. theta burst stimulation) and may induce a more robust neuromodulatory effect with increased spatial precision. Indeed, there have been two promising studies of using theta-range transcranial magnetic stimulation of the midline cerebellum to improve cognitive and executive functioning in populations of patients with schizophrenia, suggesting some potential clinical benefit to this technique (Demirtas-Tatlidede et al., 2010; Garg et al., 2016). This effect may be facilitated by cerebellar transcranial magnetic stimulation increasing functional connectivity between the cerebellum and non-motor cortical nodes within the default mode network, as has been previously reported (Halko et al., 2014). Schizophrenia remains a major source of morbidity worldwide, and research searching for new treatments and therapies to address the disabling cognitive impairments of the disease should be a priority for clinician scientists and mental health researchers.