Applications of minimally invasive multimodal telemetry for continuous monitoring of brain function and intracranial pressure in macaques with acute viral encephalitis

Alphaviruses such as Venezuelan equine encephalitis virus (VEEV) and Eastern equine encephalitis virus (EEEV) are arboviruses that can cause severe zoonotic disease in humans. Both VEEV and EEEV are highly infectious when aerosolized and can be used as biological weapons. Vaccines and therapeutics are urgently needed, but efficacy determination requires animal models. The cynomolgus macaque (Macaca fascicularis) provides a relevant model of human disease, but questions remain whether vaccines or therapeutics can mitigate CNS infection or disease in this model. The documentation of alphavirus encephalitis in animals relies on traditional physiological biomarkers and behavioral/neurological observations by veterinary staff; quantitative measurements such as electroencephalography (EEG) and intracranial pressure (ICP) can recapitulate underlying encephalitic processes. We detail a telemetry implantation method suitable for continuous monitoring of both EEG and ICP in awake macaques, as well as methods for collection and analysis of such data. We sought to evaluate whether changes in EEG/ICP suggestive of CNS penetration by virus would be seen after aerosol exposure of naïve macaques to VEEV IC INH9813 or EEEV V105 strains compared to mock-infection in a cohort of twelve adult cynomolgus macaques. Data collection ran continuously from at least four days preceding aerosol exposure and up to 50 days thereafter. EEG signals were processed into frequency spectrum bands (delta: [0.4 – 4Hz); theta: [4 – 8Hz); alpha: [8-12Hz); beta: [12-30] Hz) and assessed for viral encephalitis-associated changes against robust background circadian variation while ICP data was assessed for signal fidelity, circadian variability, and for meaningful differences during encephalitis. Results indicated differences in delta, alpha, and beta band magnitude in infected macaques, disrupted circadian rhythm, and proportional increases in ICP in response to alphavirus infection. This novel enhancement of the cynomolgus macaque model offers utility for timely determination of onset, severity, and resolution of encephalitic disease and for the evaluation of vaccine and therapeutic candidates.


48
Abstract 49 Alphaviruses such as Venezuelan equine encephalitis virus (VEEV)  Introduction viruses, including fever, ECG abnormalities such as QRS complex and QT-Interval abnormalities, and clinical 99 signs that suggested neurological disease (14). Exposure of macaques to aerosolized EEEV resulted in fatal 00 disease at high doses, typically within 5-6 days. Elevated heart rates were also seen and a prominent neutrophilia 01 that predicted outcome. At lower doses, no febrile response was seen and macaques survived exposure (3). We 02 recently reported ECG abnormalities seen with both VEEV and EEEV infection including a loss of heart-rate 03 variability that coincided with the febrile response and might be useful in determining outcome, particularly for 04 EEEV (15). While these and other studies have provided important insights, they were often gained despite 05 technical limitations. In particular, determining the onset and resolution of encephalitis often depends on 06 subjective and intermittent veterinary staff observations, rather than continuous and objective measurements of 07 certain key variables that can capture and quantify the complex and multi-modal manifestation of the disease.

08
To address this limitation and to extract more data from costly NHP studies, we set out to develop a 09 surgical method for the implantation of a radiotelemetry device for continuous monitoring of EEG and intracranial 10 pressure (ICP) in awake, conscious macaques, and complementary methods for the collection and analysis of 11 EEG and ICP data suitable for detecting significant changes indicative of CNS disease. This method constituted 12 a (i) minimally invasive, (ii) multi-modal, and (iii) continuous monitoring of brain function during alphavirus 13 infection. In this context, we define 'minimally invasive' as leaving the dura mater intact. This was considered 14 important to prevent pathogens from entering the brain in a non-physiological way, as well as to mitigate the 15 possibility of generating artifactual data. We define 'continuous' as no more than ten minutes' interruption of data  Surgical procedure.

99
On the day of surgery, each macaque was given atropine (0.04 mg/kg, intramuscularly (IM)) to reduce 00 mucus secretions. Anesthesia was induced with ketamine hydrochloride (20 mg/kg, IM). In preparation for 01 surgery, an intravenous (IV) catheter for 0.09% normal saline was inserted into the greater saphenous vein, and 02 the macaques were intubated with a tracheal tube for oxygen (1.0-1.5 L/min) and gas anesthesia. The 03 macaque's head, neck and upper back were shaved and the macaque was transferred to the surgical suite and 04 maintained on gas anesthesia (Isoflurane 0.5-3.0%) throughout surgery. The macaque was positioned in a five-05 point stereotax to ensure positional stability during surgery. The incision sites were aseptically prepared with 06 betadine and chlorhexidine scrubs, and draped with sterile drapes. Heart rate, respiratory rate, pulse oximetry, 07 and core temperature were monitored continuously until the end of surgery. Incisions were made to expose the 08 surgical sites on the scalp and on the upper back between the scapulae slightly off midline.

09
A ~7cm long incision was made along the midline of the head from ~5mm posterior of the ocular ridge to 10 the occipital ridge. Fascia and temporalis muscle were separated from the calvaria using sharp surgical spoons.

11
Residual tissue was thoroughly scraped from the bone, and the skull was dried with sterile gauze. A 5cm long 12 incision was then made between the scapulae on the macaque's upper back. Using blunt dissection, a subdermal 13 pocket superficial to the trapezius muscle was formed of sufficient size to contain the telemetry transmitter (Data 14 Sciences International (DSI) Model No. M01, M11, or L11). A tunneling rod was inserted in the caudal aspect of 15 the cranial cut and tunneled to the pocket between the scapula. The two EEG lead wires and the ICP pressure 16 sensor were threaded from the transmitter into the tunneling rod and pulled underneath the skin rostrally to the 17 cranial cut.

18
The locations for the EEG screws were identified relative to landmarks on the skull. The position of the 19 anterior EEG screw (F 4 ) was identified just anterior of the coronal suture, approximately 0.5cm to the right of  Holes for the EEG screws were drilled using a 1.2mm surgical hand-operated drill. The self-tapping 36 titanium EEG bone screws (Crist Instruments Co, Inc. 6-YXC-035) were fastened in place at a depth that would 37 allow them to touch but not dimple the underlying dura. A 1mm-diameter, 2mm-deep hole had been drilled in the 38 center of the internal hex-head of the screw prior to the surgery. The lead wires of the DSI transmitter were 39 affixed to the EEG screws by means of a 1mm amphenol pin that was subsequently soldered to the EEG leads. After protruding portions of the amphenol pin and excess solder were clipped with pliers, the screws and pins 41 were covered with dental acrylic to prevent damage and to preserve the integrity of the lead junctions. The 42 access hole for the ICP transducer was drilled using a 2.5mm surgical drill. An electric drill with a dental drill-bit 43 was then used to file down the ridge of the posterior aspect of the hole and to provide an approach parallel to 44 the dura for the ICP sensor wire. A small plastic probe was inserted to carefully loosen the attachment of the 45 dura to the skull. After removing the protective sleeve, the ICP sensor was carefully inserted into the space 46 between dura and skull. The hole and the ICP apparatus were likewise sealed and covered with dental acrylic.

07
We implanted 12 cynomolgus macaques with our multi-modal telemetry implants. Following at least 3 08 days of baseline data collection, the macaques were infected either with EEEV (6 macaques), VEEV (4 09 macaques) or mock (2 macaques). The data from the baseline period served as the basis for characterization 10 of normal cynomolgus macaque EEG and ICP. Data from the two mock infected macaques served as the basis 11 to evaluate the stability and consistency of the relevant measures in the absence of an infection. Data from the 12 experimental macaques served as the basis to identify pathological changes that are specific to the two diseases 13 and/or predictive of the outcome of the infection.
14 Electroencephalography. 16 To extract interpretable data from the raw EEG traces, they were analyzed in the frequency domain.

17
Spectral analysis of raw EEG traces is important, because different frequency bands have been linked to different 18 aspects of brain function. For example, slow-wave frequencies such as those encompassing the delta band, play 19 a role in the regulation of and/or reflect the status of homeostasis maintenance, while higher frequency waves 20 are associated with higher brain functions as well as muscular coordination and feedback (25)(26)(27)(28)(29)(30) Initially, a frequency spectrum analysis was done to assess the normal pattern for EEG data across the 27 four power bands. The average EEG power spectrum of a representative example macaque is shown in Fig 2A. 28 The approximately linear decline of EEG power with frequency (when plotted on a log-log scale) is expected and 29 suggests that the leads were implanted correctly and record physiologically plausible EEG activity. While there 30 is some variability in the daily averages, there is no systematic drift over the entire recording period. This speaks 31 to the utility of the spectral analysis as a tool to identify pathological changes of brain function. Fig 2B shows   To test for the presence of circadian EEG rhythms, we averaged EEG spectra separately for all twelve 45 macaques during the baseline period over the light and dark conditions of the animal holding room in (Fig 3A).

46
This initial analysis suggested a high degree of circadian variability in the beta band (p<0.0001). To better understand the temporal dynamics of these circadian variations, we extracted time-series data for all four EEG 48 power bands in bins of 15 minutes. Fig 3B documents the averaged circadian modulation of these power bands 49 in a single mock-infected macaque, the same macaque as in Fig 2A. As expected, beta activity coincides roughly 50 with activity and body temperature and is highest during daytime while delta peaks at night when activity and 51 body temperature are lowest. Similar patterns are seen for pooled, averaged baseline data for all macaques in 52 the cohort across a similar 24-hour span is shown in Fig 3C. At the group level, a number of details of the 53 circadian pattern emerge. Specifically, the group data show a transition during the night from a delta peak in the 54 first half of the night (1800 to 2400) to a theta and an alpha peak in the second half of the night (0000 to 0600).

55
It is also interesting to note that beta power starts ramping up a few hours before the day (facility lights-on).

56
These transitions of the EEG power spectrum during the night suggest diurnal arousal because they cannot be 57 explained by motion artifacts, as activity stays flat during the entire night. and power band traces from one macaque that developed disease following VEEV infection. Note also slight 20 dips every other day in the temperature trace due to anesthesia administered for blood draws. Initially, the time 21 series analysis also shows a regular and robust circadian modulation much like the mock-infected macaques.

22
However, on 7 DPE, there was a reduction in both beta power and delta power, with the latter finding being part 23 of a trend ranging from 5-10 DPE, and a suppressed circadian index. These findings were also visible in the 24 PSD plots and subsequent statistical analyses (Fig 6B, 6C, and 6D). It is notable that both of these changes are 25 seen during the second febrile period, which has been hypothesized to be a period in which the virus penetrates 26 the CNS (36). Note the massive delta peak in night 12 post-infection. This may correspond to a rebound of slow-27 wave sleep after several days without any discernable slow wave sleep activity during the night. Across the 28 population of macaques exposed to alphaviruses EEEV or VEEV, the blunting of the circadian rhythms is the 29 most common pathological change. Its duration for VEEV-infected macaques is typically longer than for the 30 EEEV-infected macaques, though the heightened severity of disease in EEEV-infected macaques, whose 31 neurological signs of disease warrant euthanasia during the febrile period, may mask a similar effect.

32
Additionally, the EEG spectra of macaques infected by VEEV demonstrated fewer statistically significant 33 deviations from baseline data than EEEV-infected macaques. Intracranial pressure. 45 Increased ICP during alphavirus encephalitis has been reported in humans (37-41). The raised ICP may 46 reflect increased brain parenchymal volume due to inflammatory and immune responses such as brain swelling.

47
Hence, the monitoring of ICP serves as an intuitive, proximal indicator of the response to infection within the 48 brain, and thus the status of the acute encephalitis. To measure ICP in the macaques, we inserted a standard 49 blood-pressure sensor in between the dura and the skull. Because blood pressure is much higher than 50 intracranial pressure, blood-pressure sensors are tuned to a different range of pressures. Hence, we first wanted 51 to verify that we could detect meaningful changes of ICP using this blood-pressure sensor. It is well known that 52 ICP fluctuates with individual heartbeats, known as the cardiac pulsation of ICP. If the blood pressure sensor 53 were indeed sensitive to physiological variations of ICP, the collected telemetry data would presumably register 54 an ICP pulsation corresponding to the cardiac pulsation, within the range of physiological values of the macaque 55 heart rate (here defined as between 1.5 and 3.5 Hz, corresponding to 90 to 210 bpm). Fig 7A shows  infection. We next tested whether we could indeed observe pathological increases of ICP associated with acute 84 viral encephalitis. Fig 8B shows the ICP trace of an example macaque that was infected with EEEV and 85 developed a severe encephalitis. The ICP began to rise at about the time the fever peaked at 4 DPE. Note that 86 ICP remained at peak levels well after the fever peaked and until the macaque was deemed moribund and 87 qualified for the humane study endpoint. Fig 8C shows  were elevated during periods of increased ICP. This would be a strong independent confirmation that the 08 observed increases of ICP during acute encephalitis are real, and not caused by zero-drift. Representative data 09 from one example animal shows that the amplitude of cardiac pulsations indeed increase in line with mean ICP 10 during encephalitic disease ( Fig 9A). Moreover, as expected with acute viral encephalitis, the heart rate 11 increased in infected animals during the febrile period as previously documented (3,15). This shift was also 12 visualizable in the power spectral density plots (Fig 9A), providing further evidence for the functionality of the 13 ICP implant modality. The frequency spectra of data collected for a total of 33 days ( Fig 9B)  Electroencephalography.

42
EEG is a direct measure of brain function that is highly conserved across macaques and humans. The 43 continuous long-term monitoring in mock-infected animals allowed us to describe in great detail the normal range 44 and rhythms of healthy macaque brain function. This data established a baseline against which to evaluate 45 pathological brain function caused by infection with an encephalitic alphavirus. Pre-infection, all frequency bands 46 exhibited circadian variation. As may be expected, the slow-frequency wavebands delta and theta, exhibited higher magnitudes during the night and lower magnitudes during the day, while the high-frequency wavebands 48 alpha and beta exhibit lower magnitudes at night and higher magnitudes during the day. Subtle differences in 49 timing make themselves manifest; for instance, the delta magnitude peaks early during the night, whereas theta 50 and alpha magnitude tend to peak after midnight. The beta magnitude also begins ramping up after midnight, 51 with a sharp uptick after lights-on. The data from each PSD plot are lognormally distributed rather than normally 52 distributed, and differences between macaques' baseline data suggest that group-wise statistical analysis of 53 EEG data can be strengthened by normalizing each macaque's data to its own baseline period. to specific brain regions. In the current context, a larger number of electrodes might similarly enable the mapping 95 of pathological changes to specific brain regions. Such analyses would be particularly informative in combination 96 with post-mortem pathology in different brain regions.

97
EEG electrodes are easily affected by electric sources other than the brain, e.g., eye movements, blinks 98 and chewing (59, 60). Our pre-processing steps were aimed at excluding epochs with obvious artifacts, but it 99 was impossible to detect and exclude many of the more subtle artifacts. In future studies, artifact detection can 00 be improved by using multi-electrode EEG arrays and using independent component analysis to identify artifacts 01 based on their spatial distribution and temporal properties. Nevertheless, we are confident that many of the 02 reported EEG abnormalities reflect abnormal brain function in the disease rather than altered artifacts from non-03 cerebral sources (e.g., less chewing in sick animals).

04
The current study used a blood-pressure sensor to measure ICP. While the sensor clearly was able to 05 measure physiological changes of ICP, it is plausible that a dedicated ICP pressure sensor that is designed to 06 have maximal sensitivity in the range of ICP would have provided less noisy data. Furthermore, it is known that 07 ICP can be affected by body posture. Thus, it is plausible that accounting for body posture using an automated 08 posture detection mechanism could further improve the signal-to-noise of the ICP measurements.

10
The augmentation of this well-established macaque model to produce encephalitis holds utility towards 11 the development of improved vaccines against aerosol-induced disease. The macaques that exhibited courses 12 of EEEV or VEEV disease exhibited deviations from the typical downward baseline frequency spectra in either 13 or both EEG and ICP modalities; these deviations comprised local maxima or minima in requisite segments of 14 the PSD plot for a day's worth of data. With a larger pool of animals, these changes in morphology of the 15 frequency spectra can be better characterized such that a divergence from the baseline spectra can be treated 16 as a pathognomonic differentiator of sick and well states.

17
We conclude that the analysis of EEG spectra and the derived sleep indices may portend significant