In addition to its broad use for treating intractable epilepsy (Ben-Menachem, 2002; Landy et al., 1993; Handforth et al., 1998; Penry and Dean, 1990) and depression (O’Reardon et al., 2006; Marangell et al., 2002; Rush et al., 2005; Sackeim et al., 2001), VNS is one of the most extensively studied neuromodulation therapies due to its potentially widespread effectiveness in treating disorders throughout the body. Specifically, it is thought to hold great potential for treating disorders of the CNS such as Parkinson’s Disease (Farrand et al., 2017), and for promoting recovery after injury (Pruitt et al., 2016; Khodaparast et al., 2017; Ganzer et al., 2018), but progress has been limited by a lack of understanding of the neurobiological mechanisms underlying its effects. More importantly, preclinical large animal studies have been performed primarily in canine and swine models (Settell et al., 2020; Blanz et al., 2023; Nicolai et al., 2020; Yoo et al., 2013; Yoo et al., 2016) which prevent the comprehensive study of VNS effects on higher order behavior and cognition because they cannot perform complex tasks. To that end, we adapted an approach to implant clinical electrodes on the left vagus nerve of Rhesus monkeys and verify appropriate target engagement in both awake and anesthetized experiments. This model, while not as similar to humans in the anatomy or the vagus nerve compared to swine (Settell et al., 2020), is able to perform cognitive tasks and is the closest animal model available to humans in structure and function of the brain (Wise, 2008). We demonstrated long term viability of the implant, as well as its use in behavioral and imaging experiments examining the physiological and neurobiological basis of VNS. These efforts clearly show that a large animal model of VNS based on the Rhesus monkey is viable and provides a foundation for future studies examining the effects of VNS on autonomic, motor, and cognitive function.
Two major factors contributed to the data collected: 1) the animal preparation and the combination of existing techniques used to instrument the subjects for the delivery of VNS, and 2) the training for the study of the effects of VNS under various experimental contexts. The Rhesus monkey was chosen because its behavioral capabilities and central nervous system are most analogous to that of humans (Wise, 2008) making it the animal model of choice for studies of the mechanisms underlying higher order function. It must be noted, however, that the number and organization of fascicles in the swine VN, not the Rhesus monkey, most closely resembles the properties of human VNS, making the swine a more logical choice for studies aimed at optimizing electrode design, stimulation parameters and characterizing off target effects/dose response properties to be used in humans (Settell et al., 2020; Nicolai et al., 2020; Blanz et al., 2023). The swine, however, lacks the behavioral repertoire of the Rhesus and cannot be easily imaged under awake conditions to study the effects of VNS on cognitive function. In contrast, monkeys can be trained to sit quietly in the bore of the scanner for long periods of time (Birn et al., 2019) and their cognitive performance can be readily measured using tests that are nearly identical to those employed in human studies. In addition, the neck/head anatomy of the Rhesus monkey is very similar to that of humans, which allowed us to apply Santos’ (2004) method with the modifications listed above for implanting cuff electrodes to deliver cervical VNS to humans.
Implantation Methodology and Anatomical Considerations
The electrode and technique for implantation and anchoring used in this work were adapted from methodology used in human patients suffering from epilepsy and depression (see Santos, 2004 for a full description of the human implant procedure) and is similar to previous work in non-human primates (Rembado et al., 2021). Clinical helical cuff electrodes (LivaNova PerenniaFLEX) with a 2mm inner diameter were well suited to the different sized rhesus monkeys used in this study. We saw no evidence of lead migration leading us to conclude that our anchoring/strain relief positioning was successful. Functionally, the chronic implants were effective as impedances remained stable and within the range specified by the manufacturer over long periods of time (Fig. 4D).
The need to deliver precisely timed VNS within behavioral epochs to carry out cognitive testing required us to modify the electrodes by replacing the standard connector with one suitable for the MRI environment that could be connected to the laboratory stimulation hardware. This methodology proved successful as the electrodes remained viable over several months as demonstrated by stable impedance measurements (Fig. 4D). Only one subject experienced complications due to the implant. Specifically, the animal developed a mild superficial infection between the implant site and the connector as fluid entered the outer casing of the electrode leads that were routed to the head cap. Inspection of the area lead us to conclude that the problem resulted most likely from failure in the sealant applied to the interface area between the electrode leads and the new connector. The problem was resolved by removing the electrode, the leads, and the connector, and replacing them with a new implant after the animal had recovered. No complications were encountered with the second implant.
We found variability in the structure and anatomical organization of the rhesus VN (see Table 1), which complicates the search for standard parameters for the delivery of VNS across subjects. One half (3/6) of the VNs we examined histologically contained a single large fascicle whereas the others had a large fascicle with 1 or 2 smaller fascicles separated by epineurium. The identity of accessory fascicles is unknown as their source/target was not identified and detailed immunohistochemistry was not performed. In addition, in approximately two-thirds of the NHPs used in this study, the cervical sympathetic trunk was found lying alongside the VN and could be easily isolated prior to cuff implant. However, in the remaining, the sympathetic trunk was merged with the VN requiring delicate dissection to open the adventitia/epineurium to isolate them. Thus, it is possible that the smaller fascicles observed histologically were sympathetic in origin as the sympathetic truck follows the course of the VN through the neck and often hitchhikes or sends cross connections to the VN (Seki et al., 2013; Seki et al., 2014). Thus, it is unlikely that any strategy will eliminate all fibers of sympathetic origin from stimulation. Moreover, data from one of our early acute experiments support this interpretation: after initially placing the electrode around both the VN and the putative sympathetic branch, stimulation of the electrode resulted in tachycardia, which turned into bradycardia after the sympathetic trunk was excluded from the cuff. We cannot, however, discount the possibility that some accessory fascicles may be efferent in origin and branching off the VN to form the laryngeal nerve that innervates musculature in the larynx (Settell et al., 2020; Upadhye et al., 2022) as we did not measure EMG in neck/laryngeal musculature. More detailed descriptions of fascicular organization including identification of fiber type via functional or histological measures, as well as, fascicle origin, merging and termination should be topics of future research. Such studies have been valuable in optimizing effect/side effect considerations in swine models (Settell et al., 2020; Nicolai et al., 2020; Blanz et al., 2023) and would be beneficial for the NHP as well.
Characterization of VNS Target Engagement
Investigations of VNS target engagement were conducted in both anesthetized monkeys during intraoperative sessions and in awake monkeys in the laboratory. The primary outcome measures of target engagement in these experiments were changes in cardiovascular or respiratory function as bradycardia/tachycardia and depressed respiratory function are often reported as a consequence of VNS (Hatridge et al., 1989; Sturdy et al., 2017; Saku et al., 2014; Armstrong et al., 2004). During intraoperative sessions, VNS was accompanied by decreases in respiration rate that began approximately 15s following the start of the 30s stimulation epoch and recovered the baseline value approximately 20s after stimulation ceased. The current amplitude to evoke respiration drops greater than 20% with respect to baseline varied across monkeys ranging from 0.8–1.2mA. Below those levels, changes in respiration were variable and did not always cross the 20% threshold. We found no significant change (i.e., less than 20% deviation on average) in heart rate or SPO2 as a result of VNS at simulation amplitudes that caused large drops in respiration rate (see above for description of anatomical variation and its effects. This is likely due to two factors: the moderate stimulation amplitudes applied during VNS and the cuff electrode being placed on the left VN, as our results mirror those in humans implanted for left VNS (Krahl, 2012; De Ferrari et al., 2017). Invasive measurements of blood pressure were variable and prone to error due to issues with the placement and function of the catheter; thus, we did not analyze these data. Our goals in these intraoperative experiments were twofold. First, early terminal experiments were used primarily to refine surgical methodology and verify target engagement to prepare for later chronic studies. Second, later procedures were in monkeys who received chronic implants and we sought to minimize the time spent under anesthesia.
Similar to the intraoperative experiments, VNS in awake monkeys with chronic implants was accompanied by significant changes in respiratory but not cardiovascular function. Effects of stimulation were variable. Short 10s bouts of VNS were accompanied by transient depressions in respiration rate and the depth of respiration as measured by the envelope of changes in PCO2. Longer periods of stimulation (i.e., 30 seconds) often resulted in transient depressions in PCO2 early in the epoch that gave way to more normal function as stimulation persisted, likely due to the NHPs actively counteracting the reflex-related effects of stimulation and breathing voluntarily (Fig. 4A). Thus, we computed dose response curves for VNS using shorter stimulation epochs where the effects were more robust and consistent. There was a monotonic relationship between current amplitude and respiration rate/PCO2, with increases in current amplitude resulting in larger drops in these parameters. Thresholds to evoke a significant drop were consistent across NHPs. Unfortunately, we did not explore a large enough parameter space to quantify the region of the curve where saturation of the effect occurred. Interestingly, we found these effects to be more consistent for measures of PCO2 compared to respiration rate (Figure S1). This was likely due to long integration times in the computation of respiration rate that didn’t adequately capture the dynamics of this experimental paradigm. A significant limitation of these data is related to the absence of measurement for side effects of stimulation. Anecdotally, higher stimulation amplitudes, above 1.2mA, seemed to elicit some side effect as there was evidence of somatosensory effects of stimulation (i.e., NHPs touching or brushing the skin near the implant location). Future work should characterize the current level required to evoke measurable electromyograms in neck muscles.
In interpreting the physiological response to the stimulation applied through the LivaNova cuff in this study, it is important to understand recruitment order as it pertains to physiological functions mediated by the vagus. The vagus consists of different fiber types ranging from large diameter myelinated fibers to small diameter unmyelinated fibers, which is important because they have different current threshold for activation. The easiest fibers to activate (i.e. lowest current threshold) are large myelinated Aα motor nerve efferents (12–20 microns in diameter) within the vagus trunk which innervate the intrafusal motor fibers of the deep neck muscles (Dubois and Foley, 1936; Foley and Dubois, 1937) and Aβ fibers (6–12 microns in diameter), which are canonically somatic sensory afferents projecting from the deep neck muscles to the brainstem (Sampson and Eyzaguirre, 1964) This is followed by Aγ, sensory afferents from extrafusal motor nerves ranging (3–8 microns in diameter), and Aδ, sensory afferents fibers (1–6 microns in diameter), although the overall range of Aδ fiber diameter overlaps with that of parasympathetic B fibers diameters (1–3 microns in diameter), and therefore are thought to have similar thresholds for activation. In addition to nociceptive fibers, Aδ fibers are also a class of sensory afferents leading from baroreceptors and chemoreceptors linked to VNS induced heart rate and breathing rate changes. In contrast B fibers are either 1) the vagus parasympathetic efferent innervation to the heart that have been most consistently linked to decreases in heart rate effect, or 2) pre-ganglionic sympathetic B fibers embedded or connected from the nearby sympathetic trunk that cause countervailing increased heart rate. Activation of parasympathetic B fibers have been most the most consistently linked to the observed bradycardic effects of VNS (Rajendran et al., 2019; Qing et al., 2018). Finally, the hardest to activate C-fibers, in addition to functions in chronic pain/ inflammation, also provide sensory afferent information from mechanoreceptors/baroreceptors/chemoreceptors associated with VNS induced changes in heart rate and/or breathing.
The amplitudes used in this study were guided by seminal work demonstrating the impact of VNS on learning conducted by Kilgard and colleagues (Engineer et al., 2011; Morrison et al., 2019; Loerwald et al., 2018; Borland et al., 2016; Kilgard et al., 2018). This work began in rodents and was successfully translated to human trials leading to a recent FDA approval. In these studies 0.8 mA was phenomenologically determined to be the optimum applied current on an inverted U curve for improving performance in task based learning paradigms. However, currents needed for activation of a specific fiber types are highly dependent on overall charge density, distance from the electrode, and non-linearities such as fascicle size/epineural thickness, etc. Therefore, it is entirely possible in the NHP with a larger diameter vagus than the rodent and a more simple fascicular organization than the human would require higher current levels to engage fiber types responsible for learning/plasticity (Settell et al., 2020; Upadhye et al., 2022). Higher levels of current were attempted in multiple animals, but was often limited by visible irritation from the awake animal. This is not surprising given that the current amplitude in human patients receiving chronic, therapeutic VNS is slowly titrated up over the course of many weeks to ‘habituate’ off-target activation of neck muscles (Fisher, et al., 2021). It is impractical to implement a similar procedure in our methodology due to the externalized electrode connections and absence of an IPG. Consequently, the currents used in this study may not have been sufficient to activate B fibers necessary to induce consistent bradycardia in the animals, but may have been at the threshold for A-delta activation thereby initiating the Hering-Breuer reflex. Although possible, our data are not consistent with the breathing changes driven by activation of the neck muscles, as neck muscle activation is immediate upon stimulation. Instead, the observed breathing rate reduction often took 15 seconds or more into the 30 second pulse train, which is stereotypical of Hering-Breuer reflex activation (Holmes and Remmers, 1989; Bucksot et al., 2020) (See Fig. 1D). Our lack of heart rate response is consistent with the recent VNS NECTAR clinical trial for heart failure, where an evoked response rate was observed in only 13/106 individuals, without intolerable neck muscle activation, at the 6 and 12 month time points post-implant (De Ferrari et al., 2017; Binks et al., 2001).