Sprague-Dawley rats (10–12 weeks old, Animal Resource Centre, Western Australia) were used, and all animal procedures were approved by the Animal Research and Ethics Committee of the Bionics Institute and complied with the Australian Code for the Care and Use of Animals for Scientific Purposes (National Health and Medical Research Council of Australia). Approval was also obtained from the United States Army Medical Research and Material Command Animal Care and Use Review Office, protocol SSC-7486.02. Animals were kept on a 12 h light (7 a.m.–7 p.m.)/dark cycle (7 p.m.–7 a.m.) and allowed ad libitum access to fresh food, standard chow and water. For all surgical interventions, rats were anaesthetized (2–3% isoflurane using an oxygen flow rate of 1–1.5 L/min) and breathing rate remained between 45 and 60 breaths per minute (Payne et al., 2018c). All procedures were performed under aseptic conditions and following recovery surgery, rats were monitored carefully, given an analgesic (Carprofen 5 mg/kg sub-cutaneous) and housed separately. At the conclusion of the experiment, rats were anaesthetized (2% isoflurane using an oxygen flow rate of 1–1.5 L/min), then euthanized (300 mg/kg Lethabarb, intracardial injection).

The vagus nerve electrode array consisted of four platinum (99.95%) electrodes embedded in a medical grade silicone elastomer cuff. Each platinum electrode had an exposed surface area of 0.3 mm². The distance between adjacent electrodes (E1–E2, or E3–E4, center to center) was 1.2 mm, while the distance between electrode pairs (E1–E2 to E3–E4, center to center) was 4.7 mm (Figure 1B). A channel (0.55 mm wide × 0.2 mm deep) traversed the length of the array to position the vagus nerve in close contact with the electrodes without damaging the nerve. A silicone ‘lid’ completes the cuff, preventing the nerve from migrating from the channel. A Dacron embedded silicone tab adjacent to the electrode array was used to suture the array to the esophagus, when chronically implanted onto the abdominal vagus nerve, in order to provide mechanical stability (Figure 1B). Individually insulated 50 μm diameter platinum/iridium (90/10) wires were welded to each electrode and formed a helical cable which traversed to a percutaneous connector mounted on the lumbar region of the rat.

Rats were anaesthetized, the skin incised on the ventral abdominal midline and along the dorsal-lumbar aspect of the spine. The vagus nerve electrode array was tunneled subcutaneously from the dorsal-lumbar incision to exit through the ventral abdominal incision. The abdominal cavity was exposed and the liver retracted gently using sterile saline soaked gauze. Abdominal tissue was kept moist at all times using warm sterile saline. The sub-diaphragmatic anterior abdominal branch of the vagus nerve was dissected away from the esophagus and the array implanted rostral to the hepatic and celiac branches of the vagus (Figure 1A,C). The array was sutured (7-0 silk, Ethicon) to the esophagus to provide stabilization and the abdominal cavity and skin sutured closed. The rat was rotated to expose the dorsal aspect of the spine, the percutaneous connector was anchored to the connective tissue of the lumbar region of the spine, and the skin closed around it.

To test the functionality of electrodes, the impedance of electrodes was measured using biphasic current pulses passed between the electrode of interest and all other implanted electrodes (Fallon et al., 2009). The peak voltage at the end of the first phase (V_total) of the current pulse was measured following delivery of a 25 μs per phase current pulse and current of 931 μA (Richardson et al., 2009). The V_total value was then used to calculate total impedance (Z_total) using Ohm’s law (Z = voltage/current).

Electrically-evoked compound action potentials (ECAPs) were recorded in anaesthetized rats using bipolar vagus nerve electrodes. Either pair of electrodes (E1–E2 or E3–E4) could be used to stimulate or record neural responses. Two sets of evoked electrophysiological recordings (averaged from a total of 50 responses) were made at currents from 0 to 2 mA in 0.1 mA steps using a biphasic pulse (width = 200 μs, 50 μs interphase gap) presented at a rate of 10 pulses per second. Recordings were sampled at a rate of 100 kHz and filtered (high pass: 200 Hz; low pass: 2000 Hz; voltage gain 10³) (Payne et al., 2018a). The electrically-evoked neural response threshold was defined as the minimum stimulus intensity producing a response amplitude of at least 0.1 μV within a post-stimulus latency window of 4.0–7.0 ms (Payne et al., 2018a). All recorded neural responses had conduction velocities within the range of a C-fiber response (Castoro et al., 2011).

The primary experimental groups consisted of rats that were implanted with an electrode array onto the abdominal vagus, allowed to recover for 7 days, habituated for 6 days and injected with TNBS to inflame the small intestine (see below). At 4 h following the TNBS injection, rats were randomly selected to receive abdominal VNS (TNBS+VNS: n = 7) or no stimulation (TNBS: n = 8). Blood and stool samples were collected, and ECAPs were recorded on days -14, 0, and 4 (Figure 1D). Rats were euthanized 4.5 days after TNBS injection (T = 4) and tissue taken for histological analysis. Similar to previous experiments (Pontell et al., 2009), control tissue (n = 9) was taken 5 cm oral to the ligation for the TNBS injection from animals in the TNBS (n = 5) and TNBS+VNS (n = 4) groups. An additional cohort of animals were euthanized 4 h after TNBS injection (4 H TNBS; n = 3) in order to evaluate the degree of intestinal inflammation at the onset of stimulation (Table 1).

At 14 days following implantation of the electrode array onto the abdominal vagus nerve, rats were anaesthetized and under aseptic conditions the abdominal midline incised and an 8 cm segment of jejunum, clear of intra-luminal content, was selected and ligated between two sutures (2-0 silk, Ethicon; Nurgali et al., 2007). Inflammation was induced within this ligated area by slowly injecting 1 mL of TNBS (2.5% dilution in 50% ethanol, Sigma) at the oral end of the ligated small intestine over a course of 2 min. After 5 min the ligatures were removed, the intestine returned to the abdominal cavity and the skin and abdominal wall muscle sutured closed in two layers (Pontell et al., 2009). The small intestine was kept moist with sterile saline solution during the whole procedure.

Rats were habituated for 6 consecutive days prior to the TNBS injection (T = -7 to -1) to ensure no additional stress to the animal during the testing period (T = 0–4). Animals were housed individually in Perspex boxes and percutaneous plugs connected to an external stimulator (Fallon et al., 2018), but no stimulation was delivered. Given that routine laboratory procedures can cause stress to the animal, every attempt to handle animals equally and as little as possible was made (Balcombe et al., 2004). Immediately prior to the TNBS injection (T = 0), ECAPs were generated in all implanted animals. At 4 h following TNBS injection, awake animals were randomly selected to receive VNS delivered at 1.6 mA and 200 μs/phase (i.e., 320 nC per phase) using a stimulus rate of 10 pulses/s with a 30 s ON 5 min OFF duty cycle for 3 h/day (1:30–4:30 p.m.) for 5 consecutive days (T = 0 to T = 4; 5 stimulation sessions in total). Unstimulated rats (TNBS group) were subjected to the same procedures as stimulated rats, but did not receive VNS. ECAPs were recorded on days -14, 0, and 4 (Figure 1D). Electrical stimulation was delivered using an external battery operated stimulator (Fallon et al., 2018) connected to the percutaneous connector. The stimulator delivered charge-balanced biphasic current pulses to the selected bipolar electrodes located on the nerve. Charge recovery was achieved via electrode shorting on completion of each current pulse.

Stool produced from implanted rats while being weighed (between 9 and 10 a.m. each day (T = 0 to T = 4) was assessed for consistency and signs of bleeding (Table 2 and Figure 1D) (Sun et al., 2013).

Blood was taken from the tail vein of implanted rats (Figure 1D). Blood taken prior to TNBS injection on day 0 served as a control. The final bleed was taken after the final round of stimulation. Whole blood (300 μL) was collected in K2-EDTA tubes (Starstedt) centrifuged (2000 g for 10 min) and plasma aliquoted and stored at -80°C. On the day of the assay, aliquots were thawed on ice and the C-reactive protein (CRP) ELISA conducted according to manufacturer instructions (Cusabio CSB-E07922r) and CRP levels determined via absorbance measurements using a Biorad BenchMark Plus microplate spectrophotometer.

At the conclusion of the experiment, implanted rats were euthanized and the TNBS-inflamed segment of small intestine tissue dissected and processed as previously described (Payne et al., 2018c). In brief, a 2 cm segment of control tissue was removed spanning 5–7 cm oral to where the ligation limiting the inflamed site had been. As TNBS-induced inflammation is patchy, the 8 cm length of inflamed intestine was divided equally into four 2 cm long segments, and cut longitudinally along the mesenteric border and pinned out onto balsa boards (mucosa side up). One half was placed in fixative (2% formaldehyde plus 0.2% picric acid in 0.1 M sodium phosphate buffer, pH 7.4) overnight, embedded in paraffin, sectioned (5 μm) and stained with hematoxylin and eosin (H&E) (Pontell et al., 2009) or immunohistochemically stained with anti-CD3, a cytotoxic T cell marker (1:200; Cytomation, Dako E0432) (Payne et al., 2018c). The other half of the tissue was processed for frozen sections (14 μm) and myeloperoxidase (MPO) staining (Payne et al., 2018c). All sections were mounted with DPX.

Histopathologist (J.B.Furness), blinded to procedures, used H&E stained sections to evaluate the degree of inflammation in each segment of intestine. Histological changes were on a scale of 0–3 for the assessment of damage to the mucosa (villi architecture changes, including loss of height, pinching, clubbing, venous engorgement), and assessment of inflammatory changes (leukocyte infiltration) on a scale of 0–2 for assessment of the numbers of leukocytes within venules (adapted from Payne et al., 2018c; Table 3). Scores were out of a total of 9.

Eosinophils, T cells (CD3+ cells) and neutrophils (MPO+ cells) were quantified using a Zeiss Axioplan II microscope, positive cells were counted (×40 objective) across three consecutive fields of view across the external smooth muscle layers, submucosal and mucosal layers. Cells were counted within the most inflamed area of the tissue. Images of the total field of view were generated (Axiovision, Zeiss, Germany). For MPO, eosinophil and T cell counts, cells per mm length of intestine were quantified.

Immediately following dissection of small intestine tissue, implanted animals were perfused intracardially with 0.9% saline followed by fixative (4% paraformaldehyde in 0.1 M phosphate buffer, pH 7.4, room temperature). The esophagus and implanted vagus nerve array were dissected from the carcass. At the implanted region, the vagus nerve was dissected from the array and the region of the nerve adjacent to the electrodes (E1–E4) labeled using tissue dye (Davidson’s Marking system, Bradley Products, MN, United States) (Villalobos et al., 2013). Tissue proximal to the implanted site was also taken and processed as an intra-animal, non-implanted control. The esophagus and attached vagus nerve were embedded in paraffin and serial sections (5 μm) taken. The tissue dye marked the area of vagus nerve that was adjacent to electrodes. Sections were stained for H&E and mounted with DPX. Sections were examined by an observer (S.C Payne), blinded to procedures, for signs of histopathological damage. At each location light microscope images were taken using a Zeiss Axioplan II microscope and Axiovision software (Zeiss, Germany). Using ImageJ, total fascicle area was quantified across one section per electrode position, per animal. The cross-sectional area of the vagus nerve was not measured as the boundary of the epineurium was difficult to define.

Differences between normally distributed data were tested using a one- or two-way repeated measures (RM) ANOVA with Sidak or Tukey post hoc tests as appropriate. Differences between data that was not normally distributed was analyzed using a non-parametric Kruskal Wallis one-way ANOVA and Dunn’s post hoc test. Details of each statistical test are stated in the relevant results section. Statistically significant differences were accepted as P-values of < 0.05 and GraphPad Prism 4 (GraphPad Software, United States) was used for all analysis.

The average (standard error of mean, SEM) threshold for activation of C-fibers by cervical VNS was 0.25 ± 0.07 mA and abdominal VNS was 0.43 ± 0.11 mA, indicating the test stimulation of 1.6 mA was substantially suprathreshold for C-fibers at both stimulation sites in all animals. In the example shown in Figure 2A, cervical VNS resulted in an average heart rate drop of 43 beats per minute (bpm) from baseline (370 ± 15.4 bpm), a maximum decrease in blood pressure of 8.2 mmHg and an almost complete cessation of breathing (baseline respiration rate: 52 ± 7.0 cycles per minute, cpm). In contrast, the same level of abdominal VNS produced no change in heart rate (400 ± 31.4 bpm), repiration rate (54 ± 5.0 cpm) or blood pressure (Figure 2B).

Statistical analysis [two-way (Current × Location) RM ANOVA; n = 5] of the effects of VNS on heart rate revealed a significant effect of Current (P = 0.06), Location (P = 0.04) and a significant Interaction (P = 0.02). Tukey’s post hoc analysis showed the average 31 ± 9 bpm (mean ± SEM) drop in heart rate during suprathreshold cervical VNS was significantly (P = 0.02) greater than suprathreshold abdominal VNS, which was not different to no stimulation (P = 0.9). Respiration recordings were noisy and difficult to quantify, however, severe disruptions to the regular respiration pattern were observed in 4 of 5 rats during the suprathreshold cervical VNS, while no changes in respiration were observed with abdominal VNS. Blood pressure changes were only assessed in n = 2 animals and therefore no statistical comparisons of the data were performed.

ECAPs were recorded to ensure stimulation levels were above neural threshold. The mean threshold of recorded ECAPs remained unchanged (one-way RM ANOVA (Time): P = 0.8) between day -14 (393 ± 74 μA; Figure 3A), day 0 (357 ± 65 μA; Figure 3B) and day 4 (325 ± 75 μA; Figure 3C) and substantially below the levels used to delivery therapeutic stimulation (1.6 mA, 200 μs/phase). The latency of ECAPs also remained unchanged (one-way RM ANOVA; P = 0.16) during the implantation period (T = -14: 5.43 ± 0.35 ms; T = 0: 6.69 ± 0.58 ms; T = 4: 5.52 ± 0.35 ms).

The mean electrode impedance (±SEM) in saline (prior to implantation) was 2308 ± 96 Ω (range: 1880–3340 Ω). Immediately following implantation, common ground impedance of electrodes increased to 5070 ± 246 Ω (range between 3485 and 7194 Ω). At the conclusion of the experiment the in vivo impedances had increased to 8379 ± 269 Ω (5807–9650 Ω). During the implantation period, there were no short circuits, and only 4 out of 60 electrodes (n = 4 electrodes per rat; n = 15 implanted rats in total) became open circuit. These short circuits did not compromise the delivery of VNS.

On day 0 (referred to as the “control”), prior to the TNBS injection, all implanted rats (n = 15) produced stools that were solid, dry pellets, with no signs of blood. Following TNBS injection, unstimulated rats (TNBS, n = 8) had significantly worse stool quality [two-way RM ANOVA (Time × Treatment): Time: P < 0.0001; Treatment P = 0.0009; Interaction: P = 0.02] than control (day 0) on day 1 (Sidak post hoc test: P < 0.0001), day 2 (P < 0.0001), day 3 (P = 0.002) and day 4 (P = 0.015; differences indicated by “^∗” in Figure 4A). The stool quality of stimulated rats (TNBS+VNS, n = 7) remained similar to control following TNBS injection (P ≥ 0.05). Furthermore, stimulated rats (TNBS+VNS) had significantly better stool quality than unstimulated rats (TNBS) on day 1 (P = 0.0002), day 2 (P = 0.012), day 3 (P = 0.04), and day 4 (P = 0.03; differences indicated by a ‘circle’ in Figure 4A).

Following TNBS injection, the presence of blood in feces was observed (two-way RM ANOVA: Time: P = 0.058; Treatment P < 0.0001; Interaction: P = 0.058) in unstimulated rats (TNBS; n = 5) significantly more often, compared to control (day 0), on days 1 (P = 0.0002) and day 2 (P = 0.006; differences indicated by “^∗” in Figure 4B).

Unstimulated rats (TNBS; n = 5) were observed to have higher presence of blood in feces than stimulated rats (TNBS+VNS) on day 1 (P = 0.001) and day 2 (P = 0.02; differences indicated by a “circle” in Figure 4B). Stimulated rats (TNBS+VNS; n = 4) were not observed to have blood in feces at any time point following TNBS injection (Figure 4B).

Following TNBS treatment, plasma CRP levels in unstimulated rats (TNBS) were significantly increased [two-way RM ANOVA (Stimulation × Time); Treatment: P = 0.005; Time, P < 0.0001, Interaction, P < 0.0001] at day 1 (Sidak post hoc: P < 0.0001) and day 2 (P = 0.0004), compared to day 0 (indicated by “^∗” in Figure 4C). By day 4, CRP levels returned to baseline levels. CRP levels of stimulated rats (TNBS+VNS) were significantly higher than baseline only at day 1 (P = 0.006), and no different from control (day 0) on days 2 and 4 (P ≥ 0.05). However, on day 1 CRP levels were still significantly lower (P < 0.0001) than unstimulated rats (TNBS, indicated by a “circle,” Figure 4C).

In small intestine tissue 5–7 cm oral to the site of TNBS induced inflammation (i.e., control tissue), there was no damage to the epithelial cells of the villi along the length of the analyzed tissue, with villi exhibiting the normal long finger-like projections (Figure 5A,Ai). Few leukocytes (less than 4) were observed within venules of the lamina propria and within the circular and longitudinal muscle layers, while there was mild infiltration of leukocyte populations within the mucosa and submucosa layers. In acute TNBS tissue (4 h post-injection, at the time that active VNS was applied) extensive epithelial cell loss (Figure 5B,Bi indicated by arrows) and leukocyte infiltration (Figure 5B,Bi, indicated by circle) was observed. Leukocyte infiltration into the mucosa/submucosal layers was moderate. At 4 days following TNBS injection, tissue from unstimulated (TNBS) rats had extensive irregularities to villus epithelial cells (Figure 5C,Ci, indicated by arrows). Villi were shorter and had a blunted appearance, and leukocyte infiltration was severe within the mucosa, submucosa and venules (Figure 5C,Ci). Tissue from stimulated (TNBS+VNS) animals exhibited reduced inflammation-induced damage. Less damage or abnormalities to villi surface epithelium was observed (Figure 5D,Di), while villi architecture was sometimes indistinguishable from control small intestine tissue. However, some inflammation-induced irregularities, such as shortening of villi remained within tissue from stimulated animals (TNBS+VNS; Figure 5D).

The histological score measured from control tissue of unstimulated (TNBS; n = 5) and stimulated (TNBS+VNS; n = 4) rats were no different from each other (P ≥ 0.05; unpaired T-test). Therefore the control groups were combined (n = 9). Statistical analysis of the histological score showed there was significant (Kruskal–Wallis one-way ANOVA: P = 0.001) histological damage in tissue from 4 h post-TNBS injection (Dunn’s post hoc: P = 0.002) and in unstimulated rats (TNBS; P = 0.0006; Figure 5E). However, the histological score of tissue taken from stimulated animals (TNBS+VNS) was no different from control (P = 0.17), suggesting that VNS improved recovery from inflammation-induced damage (Figure 5E).

The numbers of leukocytes measured from the control mucosal, submucosal and muscle layers were no different from each other (P ≥ 0.05; unpaired T-test). Thus the control groups were combined (n = 9). Tissue taken 4 h (4 H TNBS; n = 3) following TNBS is shown in Figure 5B,Bi, but was not included in the statistical cell count analysis, nor included in subsequent graphs (Figure 6), due to the low number of animals in this group. Four days following the TNBS injection, there was a significant increase in eosinophils within the mucosal layer (Kruskall–Wallis: one-way ANOVA: Treatment P = 0.0002), and a significant increase in the number of CD3+ cells (T cells) within both the mucosa (P = 0.001) and submucosa (P = 0.01) in unstimulated animals (TNBS, Figure 6A–C). In contrast, there was no significant elevation in the number of eosinophils, MPO+ cells or CD3+ cells in stimulated animals (TNBS+VNS; Figure 6A–C). Furthermore, eosinophil (Dunn’s post hoc: P = 0.03) and T cells (P = 0.02) populations were significantly reduced in the mucosal layer of stimulated animals (TNBS+VNS), compared to unstimulated animals (TNBS; Figure 6A,C).

Tissue taken 4 h (4 H TNBS; n = 3) following TNBS was not included in the statistical cell count analysis, nor included in subsequent graphs (Figure 7), due to the low number of animals in this group. In the circular and longitudinal muscle layers, there was an increase in eosinophils within tissue taken from unstimulated (TNBS; Kruskal–Wallis: P = 0.001) and stimulated rats (TNBS+VNS; P = 0.012), compared to control (Figure 7D). Numbers of MPO+ cells (representative images for control, TNBS and TNBS+VNS tissue shown in Figure 7A–C) (neutrophils) increased in tissue taken from unstimulated rats (TNBS; P < 0.0001), but was significantly less in tissue taken from stimulated rats (TNBS+VNS; P = 0.048; Figure 7E). Numbers of CD3+ cells (T cells) within the muscle layers of tissue taken from TNBS treated rats were greater than control (P < 0.0001). However, there were significantly fewer T cells in the muscle layer of tissue taken from stimulated rats (P = 0.04), compared to unstimulated rats treated with TNBS (Figure 7F).

Chronically implanted abdominal vagus nerves (n = 9) were assessed for histopathological changes (Figure 8A,Ai) and changes in fascicle area (Figure 8B). We observed no tissue granulation or infiltration of acute inflammatory cells, however, a foreign body tissue response (Figure 8Ai, indicated by ^∗∗) was seen within the electrode channel, surrounding the nerve (Figure 8Ai). No inflammatory cells were observed suggesting the foreign body response was likely benign. Furthermore, blood vessels (Figure 8Ai, indicated as “BV”) were observed within the nerve. There were no observed differences between stimulated and unstimulated nerve histology.

The fascicle area (indicated by dotted lines, Figure 8Ai) of the nerve adjacent to the electrodes E1-E4 was analyzed and compared to non-implanted vagus nerve tissue (taken proximal to the implantation site). There were no significant changes in fascicle area between implanted (E1–E4) and non-implanted (proximal) tissue (one-way RM ANOVA: P = 0.9; n = 9) (Figure 8B).