Power spectrum analysis of EEG in a translational nonhuman primate model after chronic exposure to low levels of the common marine neurotoxin, domoic acid

Abstract

Domoic acid (DA), the focus of this research, is a marine algal neurotoxin and epileptogen produced by species in the genus Pseudo-nitzschia. DA is found in finfish and shellfish across the globe. The current regulatory limit for DA consumption (20 ppm in shellfish) was set to protect humans from acute toxic effects, but there is a growing body of evidence suggesting that regular consumption of DA contaminated seafood at or below the regulatory limit may lead to subtle neurological effects in adults. The present research uses a translational nonhuman primate model to assess neurophysiological changes after chronic exposure to DA near the regulatory limit. Sedated electroencephalography (EEG) was used in 20 healthy adult female Macaca fascicularis, orally administered 0.075 and 0.15 mg DA/kg/day for at least 10 months. Paired video and EEG recordings were cleaned and a Fast Fourier Transformation was applied to EEG recordings to assess power differences in frequency bands from 1–20 Hz. When DA exposed animals were compared to controls, power was significantly decreased in the delta band (1–4 Hz, p<0.005) and significantly increased in the alpha band (5–8 Hz, p<0.005), theta band (9–12 Hz, p<0.01), and beta band (13–20 Hz, p<0.05). The power differences were not dose dependent or related to the duration of DA exposure, or subtle clinical symptoms of DA exposure (intentional tremors). Alterations of power in these bands have been associated with a host of clinical symptoms, such as deficits in memory and neurodegenerative diseases, and ultimately provide new insight into the subclinical toxicity of chronic, low-dose DA exposure on the adult primate brain.

1. Introduction

Over the past several decades, a rapidly changing global environment has led to increasing algal blooms worldwide (Wells et al., 2020). While some blooms may not impact human or wildlife health, other blooms contain algal toxins that can cause health effects in mammalian species (Backer and Miller, 2016; Berdalet et al., 2015; Grattan et al., 2016b). One algal toxin of emerging concern is domoic acid (DA), a small, excitatory amino acid produced by marine algae in the genus Pseudo-nitzschia.

DA is a glutamate receptor agonist that primarily binds with high affinity to 2-amino-3-hydroxy-5-methyl-isoxazole-4-propionic acid (AMPA) and kainic acid (KA) type receptors and causes typical glutamate excitotoxity (Debonnel et al., 1989; Stewart et al., 1990). Its toxicity was not documented until 1987, when a mass poisoning occurred, afflicting over 100 people with seizures, memory loss, and death (Cendes et al., 1995; Perl et al., 1990a, 1990b). At high levels, DA targets the temporal lobe, with gross neurotoxicity (necrosis and vacuolization) observed in the hippocampus, thalamus, and amygdala (Carpenter, 1990; Cendes et al., 1995). In subacute studies, DA induces status epilepticus in rats when administered 1 mg/kg/h ip for 3–5 hrs (Muha and Ramsdell, 2011; Tiedeken and Ramsdell, 2013). DA has also been associated with an elevated risk of developing adult epilepsy or spontaneous seizures in naturally exposed sea lions (Ramsdell and Gulland, 2014). In these models, there is extensive neuronal damage, possibly attributable to the injury during seizures (Kirkley et al., 2014), as well as chronic deficits in memory (Cook et al., 2015). Additionally, a number of reports detail that chronic exposure to very low levels (below the current regulatory limit) of DA is associated with memory deficits in humans (Grattan et al., 2018, 2016a). Despite this, the existing body of literature on the effects of prolonged exposure to levels of DA near the human regulatory limit lacks any investigation of the subclinical neurophysiological changes that are linked to seizures and epilepsy.

Current regulatory limits in North America close fisheries when DA levels in shellfish reach 20 ppm DA (Wekell et al., 2004). This threshold was established based on information regarding DA levels associated with clinical symptoms in humans (~200 ppm) and a ten-fold safety factor (Perl et al., 1990b). Estimates suggest that, with the 20 ppm limit for shellfish, an adult consumes 0.075–0.1 mg DA/kg bodyweight per meal of seafood (Mariën, 1996; Toyofuku, 2006). Since 1987, there has been no documentation of acute human DA toxicity, but environmental factors are changing today’s oceans. In the last 30 years, there has been an increase in the frequency, severity and duration of toxic algal blooms, including DA blooms (McKibben et al., 2017; Wells et al., 2015). These changing DA algal bloom dynamics could lead to more frequent exposures to levels of DA at or below the current regulatory limit.

While seizures and other abnormal limb activities after DA exposure have been well documented in acutely exposed laboratory and wildlife animals alike (Dakshinamurti et al., 1991; Fujita et al., 1996; Iverson et al., 1989; Ramsdell and Gulland, 2014; Tryphonas et al., 1990), it remains unclear if lower level exposures to DA may have effects on neuroelectrophysiology. In the present investigation, we used electroencephalography (EEG) in a cohort of adult, female Macaca fascicularis chronically exposed to 0.075 and 0.15 mg DA/kg/day to investigate the effects of DA on the power spectrum of neuroelectric activity measured by EEG.

EEG signals reflect the total amount of neuroelectric activity generated in the cortex, with contributing noise from the skull, scalp, dura, and surrounding electrical interference. Power is a quantitative EEG (qEEG) measure that originates from the signaling of large, pyramidal neurons in the cortex, hippocampus and amygdala (Kirschstein and Kohling, 2009) and uses the Fast Fourier Transformation to describe the strength of the signal at a given frequency. Dramatic increases in power are frequently observed in epilepsy, but, in silent epilepsy and absence seizures (i.e. petit mal seizures), these increases can be mild and within normal parameters (Clemens et al., 2000). Alternatively, discordant changes in bands can be indicative of psychiatric disorders, neurodegenerative disease, or the engagement of specific tasks, such as in memory and learning challenges (Harmony, 2013; Newson and Thiagarajan, 2019). Thus, even subtle changes in EEG bandpower may be indicative of larger neuroelectrical aberrations. The goals of this study were to employ noninvasive EEG methods in sedated animals to evaluate differences in EEG bandpower after chronic exposure to DA near the human regulatory limit.

2. Methods

2.1. Animals and Study Protocol

For this study, a subset of 20 adult female Macaca fascicularis were selected from the cohort described in Burbacher et al., 2019. Age ranged between 6 and 11 years (mean: 7 years), and weight ranged between 2.8 and 4.8 kg (mean: 3.5 kg). Animals were singly housed in stainless steel cages, with visual and grooming contact access to an adjacent animal, in the Infant Primate Research Center at the Washington National Primate Research Center. Room temperature was maintained at approximately 24°C, with a 12-hour light-dark cycle. All animals were fed Purina Monkey Chow twice daily and were supplied with regular environmental enrichment that included fresh fruit and vegetables, frozen foraging treats, and cereals, grains, and other treats. All animal protocols strictly adhered to the Animal Welfare Act and the Guide for Care and Use of Laboratory Animals of the National Research Council and protocols were approved by the University of Washington Institutional Animal Care and Use Committee.

Original study procedures and exposure monitoring protocols are described in detail in Burbacher et al. (2019). In brief, animals were exposed to either 0 (n=7), 0.075 (n=6), or 0.15 (n=8) mg/kg DA/day via an oral solution of 5% sucrose water for 306 to 596 days (mean: 429.6 ± SEM: 22.7), during which all but one female conceived and delivered a single infant (see Supplement 1). Testers performed clinical assessments at least 3x/week to track animal’s ability to visually orient and track a small piece of fresh produce, and then fully extend an arm to retrieve the treat. During the reaching task, hand/arm tremors were recorded as either present or absent. Testers were blind to the animal’s assignment and maintained an 80% reliability with the primary tester.

2.2. Experimental Design

All adult females underwent three sedated EEGs, for a total of 60 recordings. EEGs were conducted in the morning. The first and second EEGs were recorded 1.5 hours post dosing, and the third EEG was recorded 24 hours post-dosing. Each EEG occurred a minimum of 3 days apart, to assure the animal was fully recovered before undergoing sedation again. Animals were sedated for EEGs with 4 mg/kg im Telozol®, a combination of tiletamine and zolazepam (Zoetis Services LLC, Parsippany, NJ, USA), and monitored until fully sedate. Once sedated, animals were removed from the homecage and brought to a quiet testing suite for EEG acquisition. Females were prepped and closely monitored throughout EEG acquisition, as described below.

2.3. EEG Procedures and Acquisition

EEGs were acquired using an Avatar EEG 4000 Series Recorder (Avatar EEG Solutions Inc., Calgary, AB, Canada), equipped with eight 6-mm gold electrodes. Electrode cables were braided together to minimize sources of noise and cleaned with an alcohol wipe before each use. After the subject was fully sedated, the head was shaved and cleaned with alcohol, before applying electrodes with Ten20 conductive paste (Weaver and Company, Aurora, CO, USA) in approximation to the international 10–20 system locations F3/F4, T3/T4, P3/P4. A reference electrode was applied in the center of the head, and a ground electrode was applied to the left mid-back. Electrodes application was typically achieved within 15 minutes of sedation. All EEG sessions were paired with video monitoring to ensure that any movement during the session would be removed during preprocessing. EEGs were recorded for a minimum of 20 minutes or until the animal was waking from sedation. All impedances were kept below <100 kΩ and were checked at the beginning of the session and at least once during the session after electrodes had settled. Once the recording was complete, all electrode paste was removed, and the animal was returned to her homecage and monitored until fully awake.

2.4. EEG Processing and Power Calculation

Recordings were processed and analyzed in Matlab using the FieldTrip toolkit and a standardized method to analyze power from EEG recordings (Oostenveld et al., 2011). To minimize differences in sedation state, a 2–3-minute window was identified in all recordings at the same time post administration of the sedative (24 minutes post-injection). Once this window was identified, EEG recordings were synced with the paired video, and any segment with noticeable head movement, large facial movement (e.g. yawning), or electrode wire movement were removed from the recording, using the ft_databrowser function. Two independent, blinded testers reviewed each recording for epileptic and spike activity. After checking for epileptic activity, recordings were converted into Matlab files and preprocessed with ft_preprocessing to remove higher order trends from the data. Recordings were then divided into 1-second segments, and the ft_rejectvisualartifact graphical user interface was used to visually inspect the variance of each segment. Any segment outside of the normal variance in that recording was removed from further analysis. Whole recordings were excluded if: 1) the animal was resistant to sedation or 2) if more than 25% of total time was removed from the EEG. For each animal, the first EEG recording that met these criteria was used for analysis. All recordings from a single animal in the control group was rejected due to a lack of sedation on all 3 EEG recordings. Thus, the animals included in the analysis were 6 control animals, 6 in the 0.075 mg/kg group, and 8 in the 0.15 mg/kg group.

After preprocessing, recordings were unblinded and then band pass filtered at 50 Hz to remove harmonics. Timeseries data were segmented into 1-second epochs. Absolute and relative power (% of total power observed in each frequency) was calculated with FieldTrip function ft_freqanalysis using a Hanning window and with multitapered Fast Fourier Transformation, yielding power values expressed in 1 Hz frequency bands from 1–50 Hz.

2.5. Statistical Analysis

To understand if exposure to DA affected the distribution of power, FieldTrip’s ft_freqstatistics function was used. This uses nonparametric Monte Carlo simulation with a two-tail correction to estimate the significant probabilities of difference in power distributions between dose groups. The two-tailed alpha was set at 0.05, and distributions were compared at each channel and frequency from 1–50 Hz.

To compare frequency bands of interest that have previously been shown to be affected by DA (Scallet et al., 2004), two multivariate analysis of covariances (MANCOVAs) were conducted in R (R Core Team, 2020), using data averaged across the entire primate head. The initial analysis examined differences between DA exposed animals and controls, using DA exposure duration and tremor scores as covariates. A second analysis examined whether or not the DA effects were dose dependent, again using DA exposure duration and tremor scores as covariates. Outcome variables included total power in the frequency bands of interest: delta (1–4 Hz), theta (5–8 Hz), alpha (9–12 Hz), and beta (13–20 Hz). Any significant dependent variables in the MANCOVAs were further assessed with post-hoc testing using a Bonferroni correction applied to the p-values from the four bands of interest. Group differences were considered significant when the post-hoc Bonferroni corrected p-value was less than 0.05.

3. Results

After artifact removal and visual inspection of the EEG recordings, 13 of the 60 total EEGs were excluded, primarily due to an animal’s lack of sedation during the recording (see exclusion criteria above). Overall, recordings were clean, without many artifacts, and amplitudes were low, as expected for sedated recordings (Supplement 2). No epileptic or spike activity was identified on any recording. On two recordings from one control animal, we captured distinct electrooculographic (EOG) artifacts across multiple sessions, which, to the best of our knowledge, have never been documented in this species before (Supplement 3). When observed, the EOG artifact occurred simultaneously on all channels. Notably, paired video-capture did not show any signs of body movement or eye blinking, but only displayed subtle, closed-eyelid movements.

Unsurprisingly and likely due to volume conduction across the small primate head, power distributions across all animals were highly similar on all electrodes on the same side of the head, and somewhat similar between each hemisphere (Fig. 1). Using a Monte Carlo simulation to compare the absolute and relative power spectrum between exposed and control groups across all frequencies, we found that there were no significant differences in the overall distributions between dose groups (Fig. 1).

Figure 1:

The group mean power spectral density across all frequencies for each channel on the EEG recording. Inset shows the position of each channel on the head.

Preassessment with the bands of interest (delta (1–4 Hz), theta (5–8 Hz), alpha (9–12 Hz), and beta (13–20 Hz)) confirmed that variances in each band were similar between each group, and power was normally distributed, thus the assumptions for MANCOVA testing were met. The initial analysis examining the effects of any DA exposure indicated a significant effect of DA exposure (p < 0.05, Table 1) that was not related to the duration of exposure to DA or tremors (p >0.05). This relationship, however, was not dose dependent, even when accounting for duration of exposure and tremor scores (p > 0.05, Table 2). Post-hoc testing using a t-test and Bonferroni correction to compare differences between the exposed and control groups in all bands revealed that all bands were affected, but in different directions (Table 3, Fig. 2). Mean differences between the exposed and unexposed groups were: −0.04 V² for delta power (adjusted-p=0.004); 0.011 V² for theta (adjusted-p=0.002); 0.006 V² for alpha (adjusted-p=0.009); 0.0068 V² for beta (adjusted-p=0.024).

Table 1:

MANCOVA results for exposure only assessments

	Exposure Only
	Pillai	Approximate F-statistic	p
Exposure Status	0.557	3.78	0.033
Tremor Score	0.308	1.33	0.314
Days Exposed	0.462	2.58	0.091

Table 2:

MANCOVA results for dose-response assessments

	Dose-Response
	Pillai	Approximate F-statistic	p
0.15 mg/kg	0.314	1.26	0.34
0.075 mg/kg	0.478	2.52	0.10
Tremor Score	0.311	1.24	0.35
Days Exposed	0.465	2.39	0.11

Table 3:

Post-Hoc Testing of Exposure

	Exposed Mean	Unexposed Mean	Difference	p-value	adj.p-value
Delta	0.19	0.23	−0.04	0.001	0.004
Theta	0.0168	0.0055	0.0113	0.0006	0.0024
Alpha	0.0089	0.0026	0.0063	0.002	0.009
Beta	0.0105	0.0037	0.0068	0.006	0.024

Figure 2:

Comparison of the mean total absolute power spectral density over the given frequency band among exposure groups. Each point overlay represents a single individual in the group. (*adj. p-value<0.05,**adj. p-value <0.01, ***adj. p-value<0.005)

4. Discussion

Domoic acid (DA) is a common marine algal neurotoxin that is becoming more widespread under changing oceanic conditions (Wells et al., 2020). It is well documented that acute exposure to DA (>1 mg/kg) can cause seizures and epilepsy (Cendes et al., 1995; Fujita et al., 1996; Gulland et al., 2002; Teitelbaum et al., 1990) and increased electroencephalography (EEG) power in all frequencies have been documented in humans, sea lions, and rodents after acute DA exposure (Cendes et al., 1995; Fujita et al., 1996; Gulland et al., 2002; Teitelbaum et al., 1990). Despite these clear effects at high levels, there is presently little research regarding changes in EEG following long-term exposure to DA at levels near the human regulatory limit (0.075–0.1 mg/kg). Recent evidence from human and animal models suggests that chronic, low-level DA exposure may be related to other subtle neurological changes in motor responses (intention tremors in nonhuman primates, see Burbacher et al., 2019) and challenges with regular daily activities involving everyday memory and cognition in humans (Grattan et al., 2018, 2016a) and mouse models (Lefebvre et al., 2017).

The present research used EEG to assess chronic, low-level neuroelectric power effects in the nonhuman primate model, Macaca fascicularis, exposed to 0.075–0.15 mg DA/kg/day. Results from a power spectrum analysis demonstrated that animals exposed to DA have decreased power in the delta band and increased power in the alpha, theta, and beta bands. These changes in electrophysiology were not dose dependent or related to clinical symptomology (intention tremors) previously reported in some of the DA exposed females. Proportionately, animals with the highest median power in delta had the lowest median power in all other bands and tended to be in the control group (n=5/10). Similarly, animals with the lowest median delta power had the highest median power in other bands, and most were exposed animals (n=8/10).

EEG power is a quantitative measure that reflects the action and signaling of large, pyramidal neurons in the cortex, hippocampus and amygdala (Kirschstein and Köhling, 2009). Power can be divided into several frequency bands, each related to many different cognitive and neurological functions. The delta band is comprised of brain oscillations from the slowest frequency band in the EEG assessment and typically ranges from 0–4 Hz, and dominates deep sleep and basic homeostatic resting states (Başar et al., 1999; Knyazev, 2012). Delta also reflects most inhibitory responses of functional processing (Herrmann et al., 2016). Theta is comprised of oscillations in the 5–8 Hz range, and alpha ranges from 9–12 Hz. These two closely connected bands are often connected to memory and cognition, but frequently operate in opposite directions during tasks (i.e. when one increases, the other decreases) (Herrmann et al., 2016; Klimesch, 1999). Beta rhythms are between 13 and 20 Hz and have been linked to sensory and motor tasks and processing. While each band is often known for specific processing tasks, broad spectral changes across several bands have been identified as collectively reflective of performance on memory and learning tests (Finnigan and Robertson, 2011) and many psychiatric disorders (Newson and Thiagarajan, 2019), even during resting states. DA exposed macaques had decreased delta power in EEGs, and increased power in the theta and alpha bands. Decreases in low frequency power have been observed in neurodegenerative diseases, including Parkinson and Alzheimer’s disease (Bosboom et al., 2009; Poza et al., 2007); whereas increases in higher frequencies have been linked to psychiatric disorders, such as depression (Hinrikus et al., 2009).

These findings bear careful interpretation, because it is not clearly evident how the different power bands are translated across individuals and across species. Previous EEG research in Macaca fascicularis suggests that most resting EEG activity lies in the delta band, which aligns with our findings (Authier et al., 2009). In humans, however, resting-state EEG power is much more broadly distributed across bands (Hooper, 2005). Additionally, we used a sedative that contains two active ingredients: tiletamine, a relative of ketamine, and zolazepam, a benzodiazepine. These two sedatives interact with the NMDA and GABA pathways, respectively, and are not known to cause changes in the lower frequency oscillations (Choi, 2017). However, benzodiazepines are used to treat seizures and, in high, sedative doses, benzodiazepines result in excessive activity in the beta band (Van Lier et al., 2004). While we did not observe any seizure-like spikes or discharges in DA exposed females, any seizure-like activity may have been suppressed under sedation. Still, all sedation effects were balanced between dosed and control animals; so, while sedation likely had some effects on the results reported here, the observed differences are notable and necessitate further investigation.

The results of the present study collectively suggest that chronic exposure to DA near the current human regulatory limit may result in subtle changes in EEG, possibly reflecting changes in pyramidal neurons. While the functional effects of these EEG changes are not evident from the current study, continued research efforts should focus on how chronic ingestion of real-world levels of DA may be related to neurophysiological and other neurological changes in human populations across the globe.

Supplementary Material

2

Supplement 2: Ten second sample of raw EEG recording from one control animal, midway through the session.

3

Supplement 3: Displays the distinct electrooculographic (EOG) artifacts from one animal. Each vertical line represents a 1-second interval.

Highlights.

• Domoic acid is a common marine neurotoxin and epileptogen

• Adult macaques were orally exposed to domoic acid near the human regulatory limit

• Sedated EEGs were administered to monkeys and power was measured

• Exposed animals had decreased delta power and increased alpha, theta, beta power

• Further research is needed to understand functional effects of power changes