Autonomic neurostimulation therapies aim to modulate neural circuits controlling inflammation, of which a significant component is the innervation of the spleen. Assessing the preclinical effectiveness of these therapies has relied on indirect measurements, i.e., levels of TNF after an LPS challenge, which are typically made hours after the stimulus has been delivered, oftentimes in terminal experiments [35]; such measurements cannot be used to assess engagement of the anti-inflammatory pathway in real-time. Our work indicates that voltammetry in the spleen is an inexpensive and reproducible method to directly assess engagement of the anti-inflammatory pathway in response to autonomic stimulation. It also demonstrates for the first time a quantitative relationship between a physiological biomarker that is registered in real-time, the NE voltammetry signal, and an immune-mediated response that is several steps downstream, namely TNF suppression in endotoxemia. Physiological markers of anti-inflammatory effectiveness could help in the implementation of precision autonomic neuromodulation therapies, by allowing calibration of stimulation parameters to maximize desired and minimize undesired, off-target, effects in individual subjects [17].
Norepinephrine (NE) is the main neurotransmitter of the sympathetic nervous system, and the neurotransmitter released by fibers of the splenic nerve [36]. It belongs to a group of electrically active compounds termed catecholamines, which include epinephrine and dopamine [37]. Due to its oxidation-reduction properties, changes in NE concentration can be detected almost in real-time by electrochemical methods, including cyclic voltammetry. Fast scan cyclic voltammetry (FSCV) can detect local changes in NE concentration almost in real time owing to its high cycling speed (400 V/s at 10 Hz), and at low concentrations with increased spatial selectivity due to the microscopic scale of the working electrode [38]. When measuring in vitro in blood, we observed an Eo of approximately ~ 0.7 V, which agrees with previous reports [25]. When measuring from the spleen during intravenous NE injections, Eo is on average 0.85V +/- 0.06 (Fig. 1b). One possible reason for the increase of Eo in tissue and blood compared to PBS (0.6 V), is biofouling, a common phenomenon when electrochemical probes are implanted in biological tissues [39].
During intravenous injection of NE, io rises within seconds and returns to baseline after several minutes. Furthermore, io and its derivative measure Qo are dose-responsive with regard to the amount of injected NE (Fig. 1c-d). During saline injection or sham stimulation conditions io either did not change or it slowly drifted monotonically akin to a baseline drift (Fig. S1c & S2c), but never exhibited transients changes seen during stimulation. Using intravenous injections of different amounts of NE, we found that 0.03 mM is the lowest (estimated) concentration that produces a voltammetry signal and at 0.3 mM a robust signal is always produced, indicating that detection threshold of our method is between 0.03–0.3 mM. To estimate the NE concentration in blood for a given amount of injected NE, we used the average blood volume of a mouse as the volume of distribution of NE; however, we did not directly measure the concentration of NE, this being a limitation of our study. Compared to previous reports, in which 0.03–0.06 mM of NE were detected after neurostimulation [8, 9, 40, 41], the detection threshold of our system is relatively high. However, NE concentrations in previous reports were determined in processed blood samples using ELISA or mass spectrometry, where sample collection and processing to collect plasma or serum is typically done 5–10 minutes after stimulation. For that reason, NE concentrations at the time we perform voltammetry is likely greater, and above the detection threshold of the voltammetry method. Further, the concentration of NE in the spleen parenchyma is likely higher than plasma levels, since only a small portion of it overflows to the circulation [42].
The motor arc of a well-studied neuroimmune pathway that controls inflammation starts with parasympathetic efferent vagal fibers and sympathetic efferent splanchnic nerve fibers synapsing on noradrenergic neurons of abdominal ganglia and continues with post-ganglionic noradrenergic fibers inside the splenic nerve [1, 3, 43–45]. Activation of this neuroimmune pathway via neurostimulation of the vagus, the splanchnic or the splenic nerve and its terminals results in activation of splenic nerve fibers, release of NE in the spleen and subsequent suppression of manifestations of acute inflammatory responses [6, 8, 46–48]. Release of NE from splenic nerve fibers has been measured using biochemical assays measuring NE in splenic homogenates or inferred by recording splenic nerve activity after neurostimulation [4, 29]. However, such techniques cannot assess the real-time release dynamics of NE. In this study, we used voltammetry in the spleen to record a transient NE signal that is responsive to autonomic neurostimulation of several nerve targets that activate the anti-inflammatory neuroimmune pathway and suppress inflammation. For that reason, FSCV may be used in mechanistic studies of the regulation of inflammation by different components of the autonomic nervous system.
The NE voltammetry signal is responsive to stimulation of the splenic nerve (SpNS). The signal is proportional to stimulus intensity (Fig. 2a) and is suppressed by manipulations that block nerve activation by stimuli (Fig. 2d) or release of NE upon nerve activation (Fig. 2c). The NE signal appears within seconds after the onset of SpNS and lasts from several seconds to several minutes, depending on stimulus intensity (Fig. 2a). This is consistent with previous studies in pigs and in human post-mortem specimens, in which splenic nerve evoked compound action potentials were found to increase with higher stimulation intensity or duration and were blocked by lidocaine [8, 9]. The amount of NE released into splenic venous outflow during SpNS was shown to increase with higher stimulation charge [9]. Furthermore, those studies showed that physiologic responses to SpNS, such as changes in blood pressure, last for more than a minute after stimulation [8, 9], resembling the kinetics of the elicited NE signal in our study. The NE signal in our study has a slower and longer time course than what is reported in voltammetry studies in the brain (e.g. [23, 49–51]). In the brain, due to the high density of neural processes, the voltammetry electrode has an intimate contact with nerve terminals and NE is almost immediately detected. In contrast, in the spleen, NE likely has to travel through the red pulp cords, where nerve fibers form a network [44], before sufficient concentration builds up at the voltammetry electrode. The majority of released NE in the spleen is cleared by reuptake via the high-affinity, low-capacity NE transporter (NET) [42, 52]. NET is inhibited by electrical stimulation [53], which could explain the relatively long duration of the NE voltammetry signal following SpNS.
We also found that the NE voltammetry signal is responsive to stimulation of the vagus nerve (Fig. 3a). This is consistent with previous observations that VNS exerts anti-inflammatory actions via a splenic nerve-dependent mechanism [4]. In addition, we found that the voltammetry signal is responsive to efferent but not afferent VNS (Fig. 3c-e). This is consistent with previous reports implicating efferent vagal fibers in relaying signals to the splenic nerve via the celiac-superior mesenteric ganglion complex [3, 7]. It is likely that afferent vagal fibers may also contribute to the anti-inflammatory effect, via delayed engagement of multi-synaptic, vagal-sympathetic and vagal-vagal reflexes [32, 33, 54]; however, the lack of a temporally precise volley of action potentials reaching the spleen in response to afferent VNS may be responsible for the lack of a clear NE voltammetry signal. The voltammetry signal after efferent VNS has consistently lower magnitude than that after splenic nerve stimulation, even after stimulation at high intensities that produce a significant drop in heart rate (Fig. 2b-3b). A possible explanation may be that VNS results in eliciting action potentials only on a subset of postganglionic splenic nerve fibers or in activating those fibers at a sub-maximal degree. This implies that VNS at relatively low intensities might elicit release of NE at levels below the limit of detection of the voltammetry method. Studies that demonstrated anti-inflammatory actions of VNS at levels below those that induce a reduction in heart rate support this notion [47].
Third, the NE voltammetry signal is responsive to direct stimulation of the splanchnic nerve (Fig. 3f). To demonstrate this, we used a ChAT-tdTomato mouse strain to visualize and isolate the celiac-superior mesenteric ganglion complex and the associated splanchnic nerve. Although this ganglion and the splanchnic nerve have been isolated under direct vision [33, 55], we report here, for the first time, that the use of fluorescence microscopy improves yield and accuracy. The finding that splanchnic nerve stimulation elicits a NE signal in the spleen is consistent with reports that implicate splanchnic nerve activity in the anti-inflammatory effect of VNS [31–34]. For example, administering VNS with the splanchnic nerve sectioned abolishes the TNF-lowering effect of stimulation in a model of LPS endotoxemia [33].
Notably, the magnitude of the NE voltammetry signal varies across animals, regardless of the stimulated nerve. This variability may arise because of differences in electrode placement directly affecting the number of activated nerve fibers, even at identical stimulus intensities. This underlines the need for using quantifiable markers of target engagement for calibration of the dose of bioelectronic therapies [17, 21]. Another source of signal variability may lie with the voltammetry technique itself. It is likely that the location of the voltammetry electrode in the spleen relative to nerve endings influences the signal. The noradrenergic innervation of the spleen has a mesh-like structure and is most dense towards the center of the organ [43, 44]. Although the working length of the voltammetry electrode was the same in all of our experiments (500 mm), the exact insertion depth into the spleen likely varied, affecting which anatomical compartment of the spleen was sampled. This source of variability is a limitation of single-electrode cyclic voltammetry; it will have to be quantified in future studies with multiple voltammetry electrodes, capturing NE voltammetry signals from several sites in the spleen.
Previous studies of LPS endotoxemia report considerable variability in TNF inhibition in response to VNS or SpNS [12, 35, 54]. Accordingly, we found that the same intensity of SpNS produces a wide range of TNF responses in animals injected with LPS (Fig. 4b); out of 17 stimulated animals, at least 3 have TNF values similar to those of sham-stimulated animals. The magnitude of the NE voltammetry signal during SpNS explains a significant portion of this variability, about 40%. In addition, TNF suppression is more likely to occur when stimulation results in relatively small Qo values but is often minimal with greater values of Qo (Fig. 4). This relationship has not been previously described, and we can only speculate of the mechanism behind it. Monocytes/macrophages are known to express both a- and b-adrenergic receptors (ARs); a-ARs are typically associated with mediating pro-inflammatory signaling and bind NE with high affinity at low concentrations, whereas b-ARs mediate anti-inflammatory effects and only bind NE at high concentrations [56–59]. Upon SpNS, NE concentration is highest at nerve terminals, where it is released, and lower away from nerve terminals [57]. Therefore, the location of immune cells relative to nerve terminals may affect the anti-inflammatory response to SpNS. It is possible to consider that relatively small NE release may only affect ChAT+ T cells that lie close to nerve terminals by binding b2-ARs and causing acetylcholine release, which then acts on macrophages to inhibit TNF release. In contrast, greater NE release may result in monocytes/macrophages located further away, in the red pulp of the spleen, to be exposed to small but not zero concentrations of NE, binding high-affinity a-ARs and favoring TNF release [44, 57]. In fact, in a similar model of endotoxemia, one study found that co-treating animals with both a- and b -AR agonists resulted in TNF suppression similar to that in vehicle treated animals, while treating with an a- or b-AR agonist alone increased or decreased TNF, respectively [59]. Furthermore, large stimulus intensities or pulsing frequencies might induce the release of co-transmitters, such as NPY and ATP, that are also immune-modulators and might negate the NE effect [57, 60]. Therefore, it is conceivable that non-responders to neurostimulation, reported frequently in previous studies, could be in fact animals that were under- or over-stimulated. In our study, which produced a range of Qo values, animals with high Qo values showed smaller TNF suppression (Fig. 4C). In a study by Brinkman et al. [12], most of the animals receiving splenic nerve stimulation had TNF values close to those of the sham-stimulated group (~ 6000 pg/mL vs. ~7500 pg/mL, respectively). The results from our study, which used similar stimulation parameters, suggest that in the Brinkman et al. study it is likely that in a subset of animals that showed smaller TNF suppression, Qo may have been similarly high. These findings underscore the potential usefulness of FSCV as a tool to calibrate stimulation dose and optimize intensity and other stimulation parameters to achieve a predictable anti-inflammatory response within and across subjects. Although recent studies in pigs suggest using changes in splenic artery flow as a marker of effective splenic nerve stimulation [8], this approach is indirect and does not reflect the actual NE content in the spleen. Changes in splenic artery blood flow could be mediated by the direct effects of stimulation on the vessel innervation itself [61], and not necessarily reflect activation of intrinsic splenic nerve fibers.
The FSCV method has several limitations. First, the method is invasive, as it requires puncture and repair of the spleen, which might alter the splenic response and introduce an additional source of variability. This variability is likely to be minor, as the range of TNF responses we observed is similar to previously reported data using the same LPS and concentration [47]. Second, our configuration has a relatively high threshold for detection of NE, which may limit its use in some neuromodulation therapies, for example, auricular or low-level VNS [62–64]. Third, due to variations in the voltammetry electrode placement, measurement of Qo has low spatial resolution and differences in signal magnitude may partially represent variations in electrode location relative to intrinsic nerves or splenic blood supply. Finally, FSCV using our methodology cannot distinguish between different catecholamines (e.g. NE and dopamine) due to their similar chemical structures [65]. However, since dopamine is typically co-released with NE in very small amounts [66], its contribution to the signal is likely minimal.
Recent advances in voltammetry electrode fabrication allow integration onto highly flexible biocompatible materials and its implantation into various organs to record catecholamine transients. For example, the use of platinum wires allowed the detection of real-time NE release in a beating heart overcoming the fragile nature of carbon fibers [26]. Further, Li et al. developed a flexible and stretchable multichannel interface integrated on a graphene-elastomer composite, which they used to measure monoamine transients chronically in the brain and gut [67]. These technological advances, along with progress in the processing and analysis of voltammograms, may facilitate the development of implanted devices that continuously monitor catecholamine release in the spleen as part of integrated autonomic stimulation systems [68].