Non-invasive vagus nerve stimulation prevents ruptures and improves outcomes in a model of intracranial aneurysm in mice

Abstract

Background and Purpose -

Inflammation is a critical determinant of aneurysmal wall destabilization, growth and rupture risk. Targeting inflammation may suppress aneurysm rupture. Vagus nerve stimulation (VNS) has been shown to suppress inflammation both systemically and in the central nervous system. Therefore, we tested the effect of a novel non-invasive transcutaneous VNS approach on aneurysm rupture and outcome in a mouse model of intracranial aneurysm formation with wall inflammation.

Methods -

Aneurysms were induced by a single stereotaxic injection of elastase into the cerebrospinal fluid at the skull base, combined with systemic deoxycorticosterone-salt hypertension, without or with high-salt diet, for mild or severe outcomes, respectively. Cervical VNS (two 2-minute stimulations 5 minutes apart) was delivered once a day starting the day after elastase injection for the duration of follow up. Transcutaneous stimulation of the femoral nerve (FNS) served as control. Multiple aneurysms developed in the circle of Willis and its major branches, resulting in spontaneous ruptures and subarachnoid hemorrhage, neurological deficits and mortality.

Results -

In the milder model, VNS significantly reduced aneurysm rupture rate compared with FNS (29% vs. 80%, respectively). Subarachnoid hemorrhage grades were also lower in the VNS group. In the more severe model, both VNS and FNS arms developed very high rupture rates (77% and 85%, respectively). However, VNS significantly improved the survival rate compared with FNS after rupture (median survival 13 vs. 6 days, respectively), without diminishing the subarachnoid hemorrhage grades. Chronic daily VNS reduced matrix metalloprotease-9 expression compared with FNS, providing a potential mechanism of action. As an important control, chronic daily VNS did not alter systemic arterial blood pressure compared with FNS.

Conclusion -

VNS can reduce aneurysm rupture rates and improve the outcome from ruptured aneurysms.

Introduction

Aneurysmal subarachnoid hemorrhage (SAH) is a devastating disease with very high acute mortality rates and persistent neurological deficits in survivors. Inflammation plays a key role in intracranial aneurysm formation, growth, destabilization and rupture, as well as outcome after rupture, both in humans and in experimental models.^1–5 Therefore, targeting inflammation may slow down or prevent aneurysm progression. Anti-inflammatory interventions have been efficacious in reducing aneurysm rupture in some experimental models,^{6, 7} although none has yet entered clinical practice, in part due to the difficulty in translating experimental pharmacological interventions to human use.

Vagus nerve stimulation (VNS), a neuromodulation technique in clinical use for intractable epilepsy and depression,⁸ has potent anti-inflammatory properties.⁹ However, the need for surgical implantation of electrodes around the vagus has thus far limited the clinical applications of VNS. Recently, non-invasive VNS techniques with excellent safety and tolerability profile have been developed and approved for clinical use for neurological indications.^{10, 11}

We, therefore, tested a novel non-invasive cervical transcutaneous VNS approach in a mouse model of intracranial aneurysm formation and rupture that recapitulates the clinical features as well as the inflammatory mechanisms.^{5, 12–14} We examined clinically relevant acute outcome endpoints including rupture rate and survival after rupture using two different levels of model severity. Data strongly suggest that chronic daily VNS inhibits aneurysm rupture, and implicate reduced matrix metalloprotease-9 (MMP-9) expression as one mechanism. Post-rupture survival is also improved by VNS regardless of the degree of SAH.

Methods

Experiments were conducted in accordance with the Guide for Care and Use of Laboratory Animals (NIH Publication No. 85–23, 1996), and approved by the Massachusetts General Hospital Institutional Animal Care and Use Committee. Experimental protocol and timeline are summarized in Figure 1A,B.

Figure 1. Experimental protocols and morphological endpoints.

(A) Experimental timeline shows the timing of aneurysm induction model procedures and VNS or FNS (red triangles). Mild and severe model differences in DOCA dosing and high-salt diet are also shown. Transcutaneous electrical stimulus to achieve VNS or FNS consisted of 1 msec, 5 kHz sine wave pulses delivered at 25 Hz for 2 minutes, repeated once 5 minutes apart.

(B) Severe model consistently yielded higher arterial blood pressures than the mild model (p<0.001) without a difference between VNS and FNS groups (two-way ANOVA).

(C) Left: Two representative brains showing gross morphological changes in large cerebral arteries. Presence or absence of tortuosity was recorded as a secondary endpoint. Right: Two representative brains showing the method of arterial (left) and aneurysmal (right) diameter measurements as readouts using an on-field millimeter scale. Largest visible diameter of each aneurysm was measured.

(D) Representative brains showing typical examples for SAH grades 0–3 used to quantify SAH severity. Details of grading are provided in the methods.

Intracranial aneurysm model

Intracranial aneurysms were induced using intracranial elastase injection combined with systemic deoxycorticosterone acetate (DOCA)-salt hypertension in mice (8–10 week-old, male, C57BL/6, Charles River Laboratories, Wilmington, MA), as described previously with minor modifications (Figure 1A).^{5, 12} All mice underwent unilateral nephrectomy. One week later, a DOCA pellet (66 mg, 28-day release) (Innovative Research of America, Sarasota, FL, USA) was implanted, 1% NaCl was added to drinking water to induce systemic hypertension. A single dose of elastase (35 mU) (E7885, porcine pancreatic elastase, lyophilized powder, 20mg; Sigma Aldrich, St.Louis, MO, USA) was stereotaxically injected into the right basal cistern (2.6 mm posterior, 1.5 mm lateral, 5.7 mm ventral to bregma) after a midline scalp incision under isoflurane anesthesia (3% induction, 1% maintenance, in 30% O₂ and 70% N₂O). A separate cohort of mice (severe model) additionally received high-salt diet (0.39% NaCl) and supplemental doses of DOCA powder (34 mg every other day subcutaneously, in 0.1 ml olive oil) for higher mean arterial pressures (Figure 1B) and worse outcomes.

In the mild model, blood pressures were measured via a femoral artery catheter under general anesthesia (isoflurane 2.5% induction, 1% maintenance in 70% N₂O and 30% O₂) at the time of euthanasia. In the severe model, mortality and morbidity were high, and significsntly differed between VNS and femoral nerve stimulation (FNS) groups, precluding measurements just before euthanasia. Therefore, in the severe model, we measured the blood pressures using the same method but on day 5 in a separate dedicated cohort that has undergone identical procedures.

Vagus or femoral nerve stimulation

The day after elastase injection, animals were randomly assigned to either right transcutaneous (cervical) vagus nerve stimulation (VNS) or transcutaneous (inguinal) femoral nerve stimulation (FNS) groups. We selected FNS because we aimed: (a) to control for any potential effects of cutaneous stimulation and muscle twitching, (b) to avoid an autonomic nerve that may have effects of its own, and (c) to stay away from the cervical region to avoid inadvertent vagus stimulation in this small species. The stimulus consisted of 1-millisecond pulses of 5-kHz sine waves, repeated at 25 Hz, for 2 minutes, delivered twice 5 minutes apart (Figure 1A) using custom made bipolar stimulation electrodes connected to the gammaCore device (electroCore, Basking Ridge, NJ). For each transcutaneous stimulation session, we titrated the stimulus intensity to be just above the muscle twitch threshold (~7–9V), similar to that used in patients. This non-invasive VNS protocol has previously been shown to successfully activate the vagus nerve,^15–18 and has been efficacious in other animal models such as focal cerebral ischemia, pain and spreading depression.^19–23 Stimulation was delivered once daily until euthanasia by placing the electrodes on the shaved skin on the right side under brief general anesthesia. In a separate cohort, we detected only a 12±3% decrease in BP and a 19±7% decrease in HR during VNS, both of which started to recover even before the end of the 2-minute stimulation, and completely resolved within 3 minutes. These mild and brief hemodynamic effects of VNS were unlikely to alter the course of aneurysm formation and rupture. More important for the latter, resting state blood pressures did not differ between VNS and FNS groups (Figure 1B).

Outcome measures

Two observers, one of which was blinded to the treatment arm, performed neurological examinations daily. Neurological signs were graded as: 0, normal function; 1, reduced activity or weight loss >2 grams of body weight (~10% weight loss) for 24 hours; 2, flexion of the torso and forelimbs upon lifting the animal by the tail; 3, circling to one side with normal posture at rest; 4, leaning to one side at rest; 5, no spontaneous activity; and 6, sudden spontaneous death. When mice developed severe signs (grades 4 or 5), they were euthanized. Because previous studies using this model showed that aneurysmal rupture occurs within 3 weeks of aneurysm induction, all remaining mice were euthanized 21 days after aneurysm induction. Prior to euthanasia, mean arterial pressures were measured via a femoral artery catheter under isoflurane anesthesia. After euthanasia, mice were transcardially perfused with 5 ml heparinized saline followed by 0.5 ml carbon black ink (0.1 ml/s). Brains were carefully removed to preserve the integrity of the circle of Willis and ventral surface imaged under a stereomicroscope for morphological changes. Majority of animals developed some degree of dolichoectasia and tortuosity (Figure 1C) with increased diameters of the circle of Willis arteries; the latter was measured using a scale in view (Figure 1C). Aneurysms were defined as localized outward bulging of the vascular wall with the largest diameter greater than that of the parent artery. Aneurysms were often not filled with the dye indicating some degree of thrombosis. The number of aneurysms, and the largest diameter of each aneurysm were quantified in each animal. SAH severity was graded as: 0, no hemorrhage; 1, localized hemorrhage, thin SAH; 2, multiple or broad hemorrhage, diffuse thick SAH; 3, massive hemorrhage (Figure 1D).

In a separate cohort, we studied pro-inflammatory marker expression within the circle of Willis in a blinded fashion (n=6 each, VNS and FNS). Four days after the induction of aneurysms using the severe model as described above, we performed transcardiac saline perfusion, and harvested the circle of Willis arteries. We measured mRNA expression of tumor necrosis factor-α (TNF-α), interleukin-1β (IL-1β), interleukin-6 (IL-6), inducible nitric oxide synthase (iNOS), matrix metalloproteinase-9 (MMP-9), and chemokine C-C motif ligand 2 (CCL2; also known as monocyte chemotactic protein-1, MCP-1), using RT-PCR. Tissues were frozen in liquid nitrogen and kept in −80°C freezer until RNA extraction. RNA was extracted using a commercial RNA extraction kit (Zymo Research, Irvine, CA) and converted to cDNA using a SuperScript® III First-Strand Synthesis System kit (Invitrogen, Carlsbad, CA) according to manufacturer’s instructions. Reverse transcriptase polymerase chain reaction (RT-PCR) was performed using TaqMan® Gene Expression Assays and TaqMan® Fast Advanced Master Mix (Applied Biosystems, Foster City, CA) with the following inflammatory marker primers: MMP-9, IL-1β, TNF-α, CCL2, IL-6, iNOS, and the housekeeping gene 18S. All primers were purchased from Applied Biosystems. RT-PCR was performed in a 7500 Fast Real-Time PCR system (Applied Biosystems, Foster City, CA) in triplicates. Relative mRNA expression levels were normalized to housekeeping gene.

Primary outcome endpoints were survival, deficit-free survival, deficit grade, rupture rate and SAH grade. Secondary outcome endpoints were the presence or absence of tortuosity, circle of Willis artery diameters, and aneurysm counts, sizes and distribution. Pro-inflammatory marker mRNA expression was an exploratory endpoint. Sample sizes were chosen empirically (n=16/group for the mild model and n=14/group for the severe model) in the absence of prior experience with elastase technique in our lab. Mice were excluded if they failed to attain a target arterial blood pressure of at least 95 mmHg by induced hypertension paradigm (n=6 FNS, 2 VNS in mild model), if they developed severe neurological deficits (grade 4–5) but without evidence of elastase effect on subsequent morphological examination (no tortuosity in any vessel, SAH or aneurysm) suggesting unrelated cause (n=1 FNS, 1 VNS), and technical failure during elastase injection (n=3, prior to being assigned to a treatment arm). Only one animal was excluded in FNS group in gene expression experiment due to complete absence of tortuosity in any of the circle of Willis arteries suggesting lack of elastase effect.

Statistical analysis

All results were expressed as whisker-box plots (whiskers, full range; box, interquartile range; horizontal line, median; +, mean), mean ± SEM, or as median [95% confidence interval]. Statistical testing was carried out using Prism 6 (GraphPad Software, San Diego, CA). Individual statistical tests for each dataset are indicated in figure or table legends, or in text. P<0.05 was taken as statistical significance cutoff.

Results

The elastase plus induced-hypertension model led to aneurysm formation and spontaneous ruptures with various severities of SAH and neurological deficits. In the milder model, most animals survived the 21-day follow up period (Figure 2A, upper left), although some developed neurological deficits during the follow up (Figure 2A, lower left). Deficit-free survival tended to be higher (p=0.055, Figure 2A, lower left) and neurological deficit grades milder (p=0.095; Figure 2B, left) in the VNS group compared with FNS (n=14 and 10, respectively). Maximum body weight loss did not differ between FNS and VNS groups in the mild model (6.1±0.9% vs. 6.4±0.5%, respectively; unpaired t-test). Post-mortem examinations revealed gross morphological evidence for SAH indicating aneurysmal rupture in 80% of animals in the FNS group (Figure 2C, upper left). VNS significantly reduced rupture rates (29%; p=0.036). The relative risk (RR) of rupture was more than 3-fold higher in the FNS group compared with VNS (RR=3.6, confidence interval (CI) [1.0–12.9]). SAH grades were also significantly lower in the VNS group (p=0.025; Figure 2C, lower left). In contrast, aneurysm counts (Figure 3A, upper left) did not differ between the groups. Although aneurysm diameters were statistically smaller in the VNS group (Figure 3A, lower left), the magnitude of this effect was of uncertain biological significance. The center of gravity of aneurysm distribution did not differ between FNS and VNS groups (median [95% CI]: 1.61 [2.34, 0.38] mm lateral, 3.10 [4.01, 1.12] mm posterior, and 1.58 [2.20, 0.87] mm lateral, 2.96 [3.38, 2.77] mm posterior from bregma, for FNS and VNS, respectively; Figure 3B, left). However, aneurysms developed farther from the injection site in FNS group compared with VNS (Figure 3C, left).

Figure 2. Primary outcomes.

(A) Survival (upper row) and deficit-free survival (lower row) curves for FNS and VNS are shown for mild (left) and severe (right) models (n=10 FNS and 14 VNS in the mild model, and 13 each in the severe model; Gehan-Breslow-Wilcoxon test).

(B) Neurological deficit grades for mild (left) and severe (right) models. Each data point reflects one animal (Mann-Whitney test).

(C) Rupture rates (upper row) and SAH grades (lower row) are shown for mild (left) and severe (right) models. Rupture rates were compared between FNS and VNS using Fisher’s exact test, and SAH grades using Mann-Whitney test.

Figure 3. Secondary outcomes.

(A) Aneurysm count per animal (upper row) and aneurysm diameters (lower row) are shown for mild (left) and severe (right) models (numbers of mice: 10 FNS, 14 VNS in the mild model, and 13 each in the severe model). Aneurysm counts were compared between FNS and VNS using unpaired t-test. Aneurysm diameters (number of aneurysms: 19 FNS, 22 VNS in the mild model, 30 FNS, 28 VNS in the severe model) were compared between FNS and VNS using Mann-Whitney test for non-normally distributed data.

(B) Aneurysm size and distribution maps for mild (left) and severe (right) models are shown for FNS (black) and VNS (red). Blue “+” indicates injection site.

(C) Aneurysm distance to the injection site for mild (left) and severe (right) models are shown (number of aneurysms as in B; unpaired t-test with Welch’s correction for unequal variances).

The severe model resulted in much higher mortality (p<0.001) and lower deficit-free survival rates (p<0.001) compared with the milder model in FNS controls (Gehan-Breslow-Wilcoxon test; Figure 2A, left vs. right panels). In the severe model, compared with FNS, VNS significantly improved survival (median survival 13 vs. 6 days, respectively, n=13 each; p=0.003) and deficit-free survival rates (median deficit-free survival 6 vs. 4 days, respectively, p=0.029), without affecting neurological deficit grades (Figure 2B, right), rupture rates or SAH grades (Figure 2C, upper and lower right). Maximum body weight loss was similar between FNS and VNS groups (11.2±1.8% vs. 10.2±1.3%, respectively; unpaired t-test). Aneurysm counts or sizes (Figure 3A, upper and lower right) also did not differ between FNS and VNS groups. Neither the center of gravity of aneurysm distribution (median [95% CI]: 1.31 [2.09, 0.89] mm lateral, 3.28 [3.91, 2.76] mm posterior, and 1.99 [2.19, 1.61] mm lateral, 3.03 [3.60, 1.82] mm posterior from bregma, for FNS and VNS, respectively; Figure 3B, right), nor the average distance of aneurysms from the injection site differed between FNS and VNS groups (Figure 3C, right). Arterial diameters and presence of tortuosity in the circle of Willis also did not differ between FNS and VNS groups in either the mild or the severe model (Table 1), suggesting that VNS did not directly interfere with elastase effect.

Table 1.

Anatomical examination of the circle of Willis

	Mild model			Severe model
	FNS (n=10)	VNS (n=14)		FNS (n=13)	VNS (n=13)
Tortuosity (% present)
Ipsilateral anterior circulation	70%	79%	p=0.665	69%	77%	p=1.000
Contralateral anterior circulation	60%	21%	p=0.092	77%	77%	p=1.000
Basilar artery	90%	100%	p=0.417	100%	100%	p=1.000
Diameters (μm)
Ipsilateral ICA	189±13	212±13	p=0.462	166±14	174±12	p=0.175
Contralateral ICA	184±18	154±10		167±8	170±10
Ipsilateral ACA	162±11	159±13		141±10	159±11
Contralateral ACA	151±16	132±8		142±7	148±8
Ipsilateral MCA	169±11	160±9		134±9	158±9
Contralateral MCA	157±15	136±6		131±5	142±7
Basilar artery	221±19	217±8		197±14	211±10

Vessel diameters (two-way ANOVA for repeated measures) and presence of tortuosity (Fisher’s exact test) did not differ between FNS and VNS groups. ICA, internal carotid artery; ACA anterior cerebral artery; MCA, middle cerebral artery.

After finding reduced rupture rates and improved survival following rupture in the VNS group, we also explored whether the anti-inflammatory effects of VNS may play a role. To this end, we measured the expression of pro-inflammatory mediators within the circle of Willis 7 days after elastase injection in the severe model. Expression was very low in naïve circle of Willis (not shown). VNS generally suppressed the expression of all pro-inflammatory mediators compared to FNS (−34% MMP-9, −21% IL-1β, −44% TNF-α, −25% CCL2, −28% IL-6, −12% iNOS), although this effect reached statistical significance only for MMP-9 in this exploratory cohort (n=5 and 6, FNS and VNS, respectively) (Figure 4). Importantly, SAH grades in this cohort were mild and comparable between the groups (FNS: 1.20±0.20; VNS: 1.00±0.26), suggesting that VNS suppressed the inflammatory response rather than SAH severity.

Figure 4. Pro-inflammatory gene expression within the circle of Willis.

Exploratory data are shown on mRNA expression of pro-inflammatory genes within the circle of Willis in FNS- or VNS-treated mice 4 days after elastase injection (n=5 FNS, 6 VNS; unpaired t-test).

Discussion

Our data in an experimental intracranial aneurysm model with two levels of severity show that VNS reduces rupture rate in the mild version and improves post-rupture survival and neurological deficits in the more severe version. We used two different model severities because neither alone allowed us to test all intended endpoints. Survival was nearly 100% in the mild model; therefore, mild model did not allow us to assess VNS effect on survival. Conversely, nearly all animals developed a ruptured aneurysm in the severe model; therefore, severe model did not allow us to assess rupture rates. Together, the two variations of the model complemented each other, and independently yielded outcomes consistently favoring VNS.

Given the role of inflammation in aneurysm formation and growth,^1–5 and the well-known anti-inflammatory effects of VNS,^{8, 9} suppression of inflammation is likely to be the mechanism of VNS action in this model. This is also supported by our exploratory data showing a significant reduction in MMP-9 and an overall trend for reduced expression of pro-inflammatory markers within the circle of Willis arteries. MMP-9, the main gelatinase produced by macrophages, is associated with cerebral aneurysm rupture, and may predict poor prognosis.^{5, 8, 24} MMP-9 also plays a critical role in outward vascular remodeling in a mouse common carotid artery ligation model.²⁵ VNS has been shown to reduce MMP-9 expression in macrophages,^{9, 26} and prevent abdominal aortic aneurysm formation by suppressing inflammatory cytokines and MMP activity.²⁷ Interestingly, VNS has reduced the expression of MMP-9 in models of post-myocardial infarction remodeling and cardiac dysfunction as well.²⁸ Although statistically not significant in our exploratory studies, suppression of other cytokines by VNS may also contribute. For example, TNF-α has been shown to play a critical role in aneurysm formation and rupture,^{4, 29} a reduction in CCL2 may suppress aneurysm growth and rupture by limiting macrophage infiltration. Of note, gammaCore device that we employed selectively stimulates the low-threshold, myelinated, Aβ afferents to the brain stem, rather than the high-threshold, non-myelinated, efferent C fibers. Vagal afferents project to nucleus tractus solitarius and from there to numerous subcortical neuromodulatory nuclei, including basal forebrain and locus ceruleus. Therefore, central mechanisms of VNS action may also play a role in this model. Clearly, more studies are needed to elucidate the mechanisms of action of VNS on aneurysm growth and rupture, and to dissect its molecular mediators.

Improved survival and deficits in the severe model despite comparable rupture rates and SAH grades strongly suggest that VNS has a protective effect on brain tissue as well. In massive ruptures, brain tissue suffers variable degrees of global ischemia due to a rise in intracranial pressure, as well as focal ischemia due to potential interruption of blood flow downstream to the ruptured vessel. VNS has uniformly improved outcomes in animal models of ischemic stroke,^{19, 23, 28, 30} and anti-inflammatory mechanisms have been implicated as one mechanism.^{23, 26} Therefore, it is possible that in addition to preventing aneurysm growth and rupture, VNS may improve outcome after SAH.

Although the elastase-hypertension model of intracranial aneurysms we employed here is among the most widely used experimental models to examine aneurysm development and rupture,^{5, 6, 12, 31–33} it does have caveats. One such caveat is the unpredictable timing of rupture, and survival and deficits after a rupture. For example, sudden death due to massive SAH precludes ink perfusion for anatomical examination for aneurysms. Moreover, after massive SAH ruptured aneurysms may clot or collapse, making them harder to identify. Therefore, it is likely that some aneurysms were undetected, which might explain the lack of differences in aneurysm counts or diameters between mild or severe models, or between FNS and VNS groups.

No doubt these data will have to be independently confirmed, preferably in more than one model, sex and species, and incorporating longer-term outcomes. However, given the outstanding safety and tolerability profile of non-invasive VNS, and easy bedside or ambulatory implementation, there is ample opportunity for rapid translation of these findings into clinical use. For example, chronic daily VNS can be used prophylactically to prevent aneurysm formation in patients with known propensity (e.g. fibromuscular dysplasia), or growth in patients with known unruptured aneurysms. Management of such patients is currently limited to follow up neuroimaging only. In addition, VNS can also be used after rupture to improve SAH outcomes, which would be easy to implement at the bedside. Of course, more work is needed to determine the optimal therapeutic paradigms for these applications.