Electrical stimulation of the splenic nerve bundle ameliorates dextran sulfate sodium-induced colitis in mice

Background Vagus nerve stimulation has been suggested to affect immune responses, partly through a neuronal circuit requiring sympathetic innervation of the splenic nerve bundle and norepinephrine (NE) release. Molecular and cellular mechanisms of action remain elusive. Here, we investigated the therapeutic value of this neuromodulation in inflammatory bowel disease (IBD) by applying electrical splenic nerve bundle stimulation (SpNS) in mice with dextran sulfate sodium (DSS)-induced colitis. Methods Cuff electrodes were implanted around the splenic nerve bundle in mice, whereupon mice received SpNS or sham stimulation. Stimulation was applied 6 times daily for 12 days during DSS-induced colitis. Colonic and splenic tissues were collected for transcriptional analyses by qPCR and RNA-sequencing (RNA-seq). In addition, murine and human splenocytes were stimulated with lipopolysaccharide (LPS) in the absence or presence of NE. Single-cell RNA-seq data from publicly available data sets were analyzed for expression of β-adrenergic receptors (β-ARs). Results Colitic mice undergoing SpNS displayed reduced colon weight/length ratios and showed improved Disease Activity Index scores with reduced Tumor Necrosis Factor α mRNA expression in the colon compared with sham stimulated mice. Analyses of splenocytes from SpNS mice using RNA-seq demonstrated specific immune metabolism transcriptome profile changes in myeloid cells. Splenocytes showed expression of β-ARs in myeloid and T cells. Cytokine production was reduced by NE in mouse and human LPS-stimulated splenocytes. Conclusions Together, our results demonstrate that SpNS reduces clinical features of colonic inflammation in mice with DSS-induced colitis possibly by inhibiting splenic myeloid cell activation. Our data further support exploration of the clinical use of SpNS for patients with IBD. Supplementary Information The online version contains supplementary material available at 10.1186/s12974-022-02504-z.

and has a high financial burden. In this light, novel therapies that can improve disease outcome of patients with IBD are warranted.
It is increasingly acknowledged that the autonomic nervous system possesses a regulatory activity towards the immune response. Vagus nerve stimulation (VNS) was found to improve clinical outcomes in small scale clinical trials in patients with rheumatoid arthritis and Crohn's disease [1][2][3]. Similarly, in experimental colitis rat models, VNS decreased disease parameters and colonic cytokines [4][5][6][7]. Despite a positive outlook for the use VNS as immunosuppressive treatment, it has become evident that VNS affects many organs, acts on immune cells indirectly [8], and, therefore, has off-target effects, such as cardiovascular changes restricting its clinical application [9].
It has been shown that the anti-inflammatory effect of VNS in experimental endotoxemia rests on innervation of the spleen through the splenic nerve bundle [10,11]. Accordingly, anti-inflammatory activity in experimental colitis of cholinergic agonists relies on splenic innervation [12][13][14]. Since the anatomical and functional connection between the vagus nerve and the sympathetic splenic nerve bundle remain topic of debate, therapeutic efficacy of nerve stimulation might be improved through directly targeting the splenic nerve bundle instead of the vagus nerve [15]. Interestingly, it has been reported that stimulation of the splenic nerve bundle could be as effective as VNS in murine models of endotoxemia and arthritis [9,16]. Moreover, ultrasound stimulation of the splenic nerve bundle improves disease outcome in a setting of dextran sulfate sodium (DSS)-induced colitis [17]. Yet, thus far the effect of electrical splenic nerve bundle stimulation (SpNS) in experimental colitis is not well understood.
In this study, we demonstrate that electrical SpNS using implanted cuff electrodes in mice with DSS-induced colitis ameliorated colitis. Furthermore, we investigated the potential role of adrenergic receptor (AR) activation in splenic myeloid cells that could be mediating this effect.

Animals
Female C57BL/6 N mice (8-12 weeks) were purchased from Charles River Laboratories (Maastricht, the Netherlands). The animals were housed under specific pathogen free conditions in individually ventilated cages in the animal facility at the Amsterdam UMC, location Academic Medical Center (AMC), in Amsterdam. Animals were maintained on a 12/12 h light/dark cycle under constant condition of temperature (20 °C ± 2 °C) and humidity (55%) with ad libitum food and water. Mice handling and experimental protocols were in accordance with the local guidelines and approved by the local Animal Research Ethics Committee.

Surgical implantation and stimulation of cuff electrodes around the splenic neurovascular bundles and the cervical vagus nerve
First, an incision was made in the skin of the head and the skull was cleaned to attach the head mount, Preci-Dip (RS components, Haarlem, the Netherlands) with dental cement (Super-Bond C&B, Sun medical, Hofmeester Dental, Rotterdam, the Netherlands) according to the manufacturer's protocol. Second, an incision was made in the left flank to reach the spleen and to place the cuff electrode (100 µm micro cuff sling, CorTec GmbH, Freiburg, Germany) around the splenic artery, or an incision was made in the neck to expose the left cervical vagus nerve to place the cuff. The wires of the cuff were led subcutaneously to the skull and attached to the head mount. The whole procedure was performed in mice under anesthesia with 2-2.5% isoflurane/O 2 . Pre-operatively and 24 h post-operatively, meloxicam (Metacam) 1 mg/kg (Boehringer, Ingelheim am Rein, Germany) and enrofloxacin (Baytril) 10 mg/kg (Bayer Healthcare, Whippany, NJ, USA) were administered subcutaneously. Before the start of DSS treatment, mice had a recovery period of 10 days and were tethered and single housed 5 days before start of stimulation. Before start of stimulation, mice were paired based on weight and then randomly allocated (1:1) to the sham or stimulation group (Table 1). Stimulation started at the same day as the treatment with DSS and was performed 6 times per 24 h (every 4 h) for 2 min with a biphasic pulse (650 µAmp, 10 Hz, 100 µs per phase). Mice were observed during the first stimulations for behavioral changes and altered breathing patterns. Before, during and after the experiment, cuff functioning was ensured by measuring impedance using a Minirator MR Pro (NTI Audio, Essen, Germany). If the impedance was > 25 kΩ before start of stimulation, the mouse was allocated to the sham group. Animals that received sham stimulation underwent the same surgical procedure with cuff placement, were tethered to a wire but no actual stimulation was performed.

Surgical splenic denervation
Via a midline incision the spleen and splenic artery were located. Selective denervation was achieved by cutting the catecholaminergic nerve fibres running along the splenic artery. Sham-operated animals underwent a laparotomy without denervation. Surgery was performed on anesthetized mice by injecting intraperitoneally (i.p.) a mixture of fentanylcitrate/fluanisone (Janssen, Beerse, Belgium) and midazolam (Roche, Woerden, the Netherlands). Finadyne (Intervet, De Bilt, the Netherlands) was injected subcutaneously pre-and postoperatively. Before the start of the DSS-induced colitis experiment, mice had a recovery period of 2 weeks.

Acute SpNS and endotoxemic shock
Seven days following implantation, animals were injected i.p. with a lethal dose of lipopolysaccharide (LPS; 400 µg). Electrostimulation was applied using a PlexStim V2.3 (Plexon) starting at − 10, 0 and + 20 min relative to LPS injection. Mice were electrically stimulated with the same parameters as above at 16,20,24,30,34 and 38 h after LPS injection. Survival was followed over 4 days. Serum was collected at 90 min after LPS injection and assessed for Tumor Necrosis Factor (TNF)-α levels. Again, controls were fully Cortec implanted mice, which did not receive electrical stimulation (sham). Electrostimulation were rectangular charged-balanced biphasic pulses (650 μAmp, 10 Hz, 100 μs per phase) for 2 min. For TNF-α, retro-orbital blood sampling was performed under isoflurane anesthesia. TNF-α levels were measured by ELISA (Mouse TNF-alpha DuoSet, R&D Systems) following manufacturer instructions.

Dextran sulfate sodium (DSS)-induced colitis
Two percent (w/v) DSS (TdB Consultancy, Uppsala, Sweden) was added to the drinking water for 5 consecutive days. Daily replacement of drinking water with fresh DSS solutions was performed. After 5 days, DSS drinking water was replaced by normal drinking water and animals were followed for weight and behavior for the At the end of the study, mice were sacrificed, and the colon weight and length were measured, as parameter for colitis. Then, the tissue was snap-frozen in liquid N 2 and stored at − 80 °C or put in 10% formalin for further processing. The disease activity index (DAI), ranging from 0 to 9, was used to assess the clinical outcome of the DSS-induced colitis. DAI was determined by combining scores of stool consistency (0-3), occult blood in the stool (0-3), and macroscopic inflammation (0-3) [18]. Endoscopy was performed at day 8 after start of DSS. Endoscopy was performed under anesthesia with 3% isoflurane/O 2 to assess colonic inflammation. The Olympus URF type V endoscope (Zoeterwoude, the Netherlands) was rectally inserted for a maximum of 5 cm and videos of the endoscopy were recorded using a Medicap USB200 Medical Digital Video Recorder (Roermond, the Netherlands), while retracting the endoscope. A blinded and trained technician determined the murine endoscopic index of colitis severity (MEICS), consisting of wall thickening, vascularity, visible fibrin, granularity, and stool consistency, with each component scoring between 0 and 3 [19].

Histology
Swiss rolls of the distal colon were fixated in 10% formalin. Afterwards the tissue was embedded at the Pathology department of the Amsterdam UMC, location AMC, in paraffin for routine histology. A blinded and experienced pathologist evaluated formalin-fixed hematoxylin and eosin (HE) stained tissue sections microscopically. The pathologist scored the distal colon based on eight characteristics of inflammation [20]. This resulted in a total histology score ranging from 0 to 24.

cDNA synthesis and quantitative PCR analysis
For quantitative polymerase chain reaction (qPCR), RNA was further cleaned from DSS with the Bioline ISOLATE II RNA mini kit (GC biotech B.V., Alphen a/d Rijn, the Netherlands). cDNA was synthesized using dNTPs (ThermoFisher Scientific, Landsmeer, the Netherlands), Random primers (Promega, Leiden, the Netherlands), Oligo dT primers (Sigma, Zwijndrecht, the Netherlands), Revertaid and Ribolock (ThermoFisher Scientific) according to the manufacturer's instructions. PCR was performed using SensiFAST SYBR No-ROX (GC biotech B.V.) on a LightCycler 480 II (Roche Applied Science) to analyze expression levels of genes of interest using Lin-RegPCR software [32]. For normalization the reference genes hypoxanthine phosphoribosyltransferase (Hprt), cyclophilin, Non-POU Domain Containing Octamer Binding (Nono) and ribosomal protein lateral stalk subunit P0 (Rplp0) were selected, after analysis for stability in geNorm [33]. Primers (synthesized by Sigma) are listed in Table 2.

Adrenergic cell culture assays
Murine spleens were obtained from female C57BL/6 mice. Human splenic tissue was obtained from patients that underwent distal pancreatectomy for pancreatic cancer during which the spleen was taken out as part of the procedure. Splenic tissue was obtained in consultation with a pathologist to assure that tissue was not affected by any tumorous tissue. The Medical Ethics research Committees United (MEC-U, Nieuwegein, the Netherlands) approved the protocol (registration number W16.182) and patients provided informed consent for the use of their material. Spleens or splenic tissue were immediately homogenized on a 70 µm cell strainer and suspended in RPMI-1640 medium (supplemented with 10% fetal calf serum (FCS; Bodinco, Alkmaar, the Netherlands), 100 U/ml pen/strep (Lonza, Basel, Switzerland), 2 mM L-Glutamine (ThermoFisher Scientific) to obtain a single cell suspension. Cells were counted using the Coulter Counter (Beckman Coulter, Indianapolis, USA) and were plated using 1*10 6 cells per well. All conditions were performed in triplicate. Cells were pretreated with norepinephrine (NE) and/or propranolol (both 1 mM, Sigma-Aldrich, Saint Louis, USA) 30 min before exposure to LPS 100 ng/mL (Bio-Connect, Huissen, the Netherlands). Murine TNF-α was assessed after 4 h in supernatant with ELISA (R&D systems, Minneapolis, USA) according to the manufacturer's instructions. Human cytokines (TNF-α, IL-6, and IL-8) were assessed by means of BD cytometric bead assay (CBA; BD Bioscience, San Jose, CA, USA) according to manufacturer's protocol, with the exception that reagents were 10 times diluted.

Single-cell RNA-sequencing database studies
Publicly available single-cell RNA-sequencing (scRNAseq) data sets from two C57BL/6JN mouse spleens were downloaded from the Gene Expression Omnibus (GEO; GSE109774) [34]. Alignment was done against GRCh38 using Cell Ranger Software (v3.1.0; 10 × Genomics, Inc., Pleasanton, CA, USA) and subsequently data sets were imported in Rstudio (v1.3.1093) [35]. Seurat was used to import, integrate, and cluster data, and plots were made with ggplot2 (v3.3.3). [31,36] Cells were filtered for dead cells that were identified by a low (500-5000) gene count. Clusters were identified by Louvain clustering method, and annotation was performed with use of known markers [34]. Thereafter, expression of Adrb2 (encoding for β2-AR) was assessed. The top 13 principal components were used to calculate the t-distributed stochastic neighbor embedding (t-SNE).

Data presentation and statistical analysis
Graphs and statistical analyses were made with Prism 8.3 (GraphPad Software, La Jolla, CA, USA). For all data, a Kolmogorov-Smirnov test was used to determine normality of distribution. Data are shown as mean (if distributed normally) plus standard deviation or median (if not distributed normally), and individual data points. A P value < 0.05 was considered significant. Data were compared with the independent t test or Mann-Whitney U test as appropriate. Survival was plotted using Kaplan-Meier's curves and differences between groups were estimated using the log-rank test. In case of multiple groups, a one-way ANOVA and Dunnett's multiple comparison test were used.

Results
Recently, it was shown that SpNS could decrease disease severity in a mouse model of arthritis and that SpNS causes the release of splenic NE, which in turn can reduce monocyte and macrophage LPS-induced cytokine secretion via β2-AR [9]. A much debated issue is whether NE targets β2-AR on macrophages or T cells that produce acetylcholine [8]. Therefore, the expression of Adrb2  IL-1β  GCC CAT CCT CTG TGA CTC AT  AGG CCA CAG GTA TTT TGT CG   Il-6  GAG TTG TGC AAT GGC AAT TCTG  TGG TAG CAT CCA TCA TTT CTT TGT   Il-10  TGT CAA ATT CAT TCA TGG CCT  ATC GAT TTC TCC CCT GTG AA   Il-12  AGA CCC TGC CCA TTG AAC TG  CGG GTC TGG TTT GAT GAT GTC   TNF-α  TGG AAC TGG CAG AAG AGG CACT  CCA TAG AAC TGA TGA GAG  in murine splenocytes was examined using a publicly available scRNA-seq data sets. A total of 13,365 features were retrieved in 9568 individual cells. After quality control, the remaining 9407 cells were annotated to main cell populations (i.e., T, B, natural killer cells (NK), dendritic cells, and macrophages) using previously reported markers (Fig. 1A) [34]. Subsequent feature analysis showed a marked expression of Adrb2 in a cluster of cells belonging to the B cell population, when compared with other cell populations (log2FoldChange = 0.51, padj = 8.98E-18). In addition, Adrb2 expression by a cluster of cells belonging to the macrophage population was found to be significantly increased when compared to other cell populations (log2FoldChange = 0.31, padj = 0.009) (Fig. 1B).
Hereafter, LPS-induced inflammatory activation on both murine and human splenocyte cultures was investigated. In line with earlier investigations, NE decreased the release of TNF-α following LPS challenge in murine and human splenocytes. This effect was abrogated by pre-treatment with propranolol, demonstrating a mechanism involving β-ARs (Fig. 1C, D).
As earlier studies demonstrated that vagotomy aggravates colitis outcome in a DSS-induced animal model through splenic innervation, and VNS ameliorates Crohn's disease in patient trials, we hypothesized that disrupting the splenic nerve bundle (SplX) would worsen colitis [37]. As expected, mice receiving DSS in drinking water lost weight, showed an increased DAI and had a shortened colon due to edema compared to control mice (Additional file 1). The expression of inflammatory colonic cytokines was also increased. However, splenic nerve bundle denervation did not alter these outcomes significantly when compared to sham-operated animals. Hence, we could not demonstrate a tonic and intrinsic role for splenic innervation for regulation of immune responses in this model. Notably, VNS using chronic nerve cuffs around the left cervical vagal branch did not ameliorate disease in the setting of DSS-induced colitis For both murine and human data, protein levels were relative compared with LPS condition. * indicates a significant difference compared to LPS treated splenocytes. Statistical differences were assessed with a one-way ANOVA and Dunnett's multiple comparison test. t-SNE t-distributed stochastic neighbor embedding, nKT natural killer T cell, DC dendritic cell, Adrb2 adrenergic receptor β2, TNF tumor necrosis factor, IL interleukin, LPS lipopolysaccharide, NE norepinephrine (Additional file 2). To address the role of splenic nerve activity in the immune response in DSS-induced colitis, we next assessed the effect of SpNS by implanting a cuff around the splenic artery and surrounding nerve bundle (Additional file 3). Electrical stimulation of the nerve bundle can be varied in amplitude, current, and frequency. Using the stimulation parameters depicted in Additional file 3, SpNS led to a reduced induction of systemic TNF-α and increased survival in LPS-injected mice ( Fig. 2A, B). Using the stimulation parameters established, we next implanted cuff electrodes around the splenic nerve bundle and artery and tested the stimulation regiment in conscious, freely moving mice, in the setting of DSS-induced colitis. SpNS applied 6 times daily for 12 days reduced the colon weight/length ratio (40.4 vs. 51.8 mg/cm; P = 0.01) and DAI (0 vs. 1.5; P = 0.06) when compared with sham stimulated mice (Fig. 3A). No clear difference in the loss of bodyweight over time was observed (Fig. 3B). At day 8, during the peak of DSS-induced colitis, comparable endoscopic scores (4.3 vs. 6.1; P = 0.14) were found in the animals treated with SpNS vs. animals that underwent shamstimulation (Fig. 3C). At the end of the experiment, spleen weight (98.5 vs. 133.0; P = 0.07), histology scores (4 vs. 11.5; P = 0.18), and protein expression of cytokines in the colon were not significantly different between mice that received SpNS and sham stimulated mice, although trends towards amelioration of colitis in SpNS-treated mice were evident (Fig. 3D-F).
Because SpNS ameliorated clinical features of colitis, we aimed to identify molecular events following SpNS. Transcriptional profiling by means of RNA-seq was performed on spleens from mice with DSS-induced colitis that received sham stimulation and mice with DSS-induced colitis that received SpNS. The top 50 genes that were differentially expressed between sham stimulated mice and SpNS mice are listed in Fig. 4A. The downregulation of nitric oxide synthase 2 (Nos2) and gene signatures associated with immune metabolism prompted us to further focus on genes and pathways that are involved in cellular immune metabolism, which is known to be influenced by β2-AR activation [38]. For instance, Hypoxia-Inducible Factor (HIF)-1α marks the cellular response to systemic oxygen levels in various cells, including activated macrophages [39]. Expression of genes in the HIF-1α signaling pathway was significantly reduced in SpNS spleens (P = 0.007), while expression of genes encoding proteins relevant to oxidative phosphorylation was relatively increased in SpNS spleen cells (P < 0.001, Fig. 4B). This suggests that SNS favors the metabolic state of oxidative phosphorylation over glycolytic states in stimulated splenocytes, corresponding to our earlier observation of NE-simulated macrophages [38] and characteristics of macrophages found in tissue mucosa of colitis in remission [40].
Next, to explore the underlying mechanisms through which SpNS affected intestinal inflammation, we performed bulk RNA-seq. Differentially expressed genes in colon were assessed between mice that did not receive stimulation or DSS, sham stimulated mice with DSS-induced colitis, and SpNS-treated mice with DSS-induced colitis (Additional file 4). Comparing the transcriptomes of mice with DSS-induced colitis (without SpNS) and control mice indicated an expected  TNF-α and IFN-γ in colon homogenates, normalized for total protein levels. N = 15-16 per group. All mice (both sham and stim) were implanted with a cuff electrode, were allowed to recover for 10 days and then received DSS in drinking water for 5 days followed by a 7 day recovery period. Data are expressed as mean or median and individual data points. Statistical differences between sham and stim mice were assessed using an independent t-test or a Mann-Whitney U test. P < 0.05 was considered significant. DSS dextran sulfate sodium, SpNS splenic nerve plexus stimulation, MCP monocyte chemoattractant protein, IL interleukin, TNF tumor necrosis factor, IFN interferon predominantly found in the domains of leukocyte chemotaxis and migration. Noteworthy is the SpNS-induced downregulation of S100A8 and S100A9, which encode subunits of calprotectin, the clinical biomarker for IBD. Genes of interest for colonic inflammation were validated in a second independent experiment using qPCR. Indeed, TNF-α was significantly reduced in mice that received SpNS compared with sham stimulated animals (0.44 vs. 1.00, respectively, P = 0.009; Fig. 5B). As we showed in the spleen, genes in the HIF-1α signaling pathway were reduced, although this effect was less pronounced in the colon compared with the spleen (Fig. 5C).

Discussion
Here, we demonstrate that SpNS induced transcriptional changes in the spleen, and reduced clinical signs of colitis such as colon oedema and histological parameters scored in the assessment of colitis, while having a limited effect on colon inflammation. The results of this study support both ex and in vivo studies in models of endotoxemia demonstrating SpNS can be used to reduce inflammatory activation of splenocytes [16,41,42]. Recently, the effect of both electrical and ultrasound SpNS was found to improve inflammation in two different animal models of arthritis [9,43]. Similarly, ultrasound stimulation of the spleen and splenic nerve bundle improved DSS-induced colitis [17]. Although it could not be excluded that other organs were affected by this stimulation, the immune modulatory effects were not observed in splenectomized animals, and therefore, ultrasound was thought to target the spleen and splenic nerve bundle directly. Our results largely corroborate the observations made by Nunes et al. regarding the weight loss and decrease in colon density and histology scoring after stimulation. In line, in our studies a negative regulation of expression of various cytokines by SpNS was noted, depending on the starting time of stimulation after the induction of DSS-induced colitis. This highlights the complexity of the DSS model, in which duration of DSS and moment of outcome measurement vary throughout literature [44,45]. The fact that  SpNS mice. B mRNA levels of TNF-α, IL-6, IL-1α, IL-1β, IL-10, IL-12, MMP-7, ACOD1, CXCL1, CXCL2, S100A8 and S100A9. mRNA levels are normalized against reference genes Nono and RPLP0. N = 15-16 per group. C GSEA plots showing depletion of HIF-1 signaling pathway associated genes among the SpNS mice. Data are expressed as mean (TNF-α) or median (other) and individual data points. Statistical differences between sham and stim mice were assessed using an independent t-test or a Mann-Whitney U test. P < 0.05 was considered significant. DSS dextran sulfate sodium, SpNS splenic nerve plexus stimulation, HIF hypoxia-inducible factor, Nono non-POU domain-containing octamer-binding protein, RPLP0 ribosomal protein lateral stalk subunit P0, TNF tumor necrosis factor, IL interleukin, MMP matrix metalloproteinase, ACOD cis-aconitate decarboxylase, CXCL chemokine (C-X-C motif ) ligand, S100 S100 calcium-binding protein, GSEA gene set enrichment analysis both electrical (this study) and ultrasound (Nunes et al.) SpNS reduce disease severity following DSS-induced colitis strengthens the idea that SpNS might hold potential as a therapy for patients with IBD, although further mechanistic studies are warranted.
The parameters used in this study were similar to the stimulation settings that were used in an earlier reported model of experimental arthritis [9]. Interestingly, the anti-inflammatory effects of SpNS in the arthritis experiments were more profound than what we have found in our study. It may well be that the pathology in the collagen-induced arthritis model is primarily dependent on systemic cells and cytokines, and is, therefore, better controlled through SpNS compared to DSS-induced colitis, which is a more mucosal inflammatory process. Noteworthy, Guyot et al. focused on stimulation of a nerve branch that innervated the cranial pole of the spleen, which consists of cholinergic and adrenergic fibers in contrast to the adrenergic, arterial branches. Conversely, in our present study we focused on stimulation of the nerves that run along the artery, because this bundle resembles human anatomy regarding splenic innervation more accurately and is surgically accessible in humans [46,47].
While VNS has been found to result in anti-inflammatory effects in various experimental conditions, such as sepsis, postoperative ileus, rheumatoid arthritis and kidney disease, we were unable to demonstrate any antiinflammatory effects of VNS in our DSS-induced colitis model [48][49][50][51]. Histologic examination of the cuff showed intact neural tissue and stimulation parameters used were similar to other studies, making it unlikely that VNS was taking place insufficiently, or vagus nerve bundles were damaged [15,52]. There could be several other explanations for this contradiction. In earlier studies VNS was shown efficacious in reducing colitis in rats under 2,4,6-trinitrobenzene sulfonic acid (TNBS)-induced colitis and only improved survival from oxazolone-induced colitis in mice, both representing more detrimental models of colitis [6,53]. For the DSS-induced colitis model no beneficial effect of stimulation has been reported previously, although vagotomy did worsen colitis [37]. Furthermore, VNS did not improve the histology score in all studies, which is the conventional outcome parameter for experimental colitis [4]. While we could not replicate the anti-inflammatory effect of VNS, it still holds potential as it seems to improve outcome of patients with Crohn's disease in a limited, non-randomized controlled clinical trial [2]. The off-target effects such as heart rate depression remain undesirable, but this might be prevented by application of subdiaphragmatic VNS [7].
Although we discuss indirect modulation of colitis through the splenic nerve plexus and vagus nerve, it should be noted that sympathetic nerve stimulation or stimulation of other nerves directly innervating the colon such as the superior mesenteric nerve and sacral nerves have all been shown to improve experimental colitis in earlier studies [54][55][56]. This underlines the potential of neuromodulation as a treatment for IBD patients.
Our analysis showed that inflammatory gene transcription during DSS-induced colitis is counteracted by SpNS in spleen and colon. Our approach is one of a preventive intervention into DSS-induced colitis, rather than a treatment of an established colitis, which is a limitation of this study. Another drawback of the current study is that the observed signal could well be the result of an interplay between the proportional representation of various cell types as well as changes in their transcriptional profiles. From earlier studies we know that SpNS causes transcriptional changes in the T and B cell populations present in the spleen [43]. However, we previously demonstrated that the AR activity affects the macrophage potential to adapt their metabolic profile in vitro, supporting our interpretations in the current study [38]. Inflammatory macrophage activation blunts oxidative phosphorylation, thereby preventing repolarization, strictly directing inflammatory cell expression [39]. Having identified SpNS-associated differences in the expression of HIF-1α and oxidative phosphorylation, it is enticing for us to suggest that SpNS affects macrophage activation quite specifically. This is also supported by the scRNA-seq results showing a large portion of macrophages expressing Adrb2. The latter could at least in part provide a mechanism for the beneficial action of splenic nerve bundle activation on immune driven pathology, such as in colitis or collagen-induced arthritis [9]. Nonetheless, future studies are necessary to disentangle which SpNS-associated differences are the result of cellular heterogeneity and the transcriptome thereof.

Conclusions
We demonstrated the effects of SpNS in a murine model of colitis and showed that SpNS ameliorated clinical features of colitis and caused differential expression of genes in the spleen and colon involved in inflammation. Future experimental studies should focus on the connection between the spleen and colon to gain more insight in the therapeutic mechanisms of SpNS. Meanwhile, a clinical trial in patients undergoing esophagectomy will investigate safety and feasibility of SpNS for application in humans (www. clini caltr ials. gov; NCT04171011).
Additional file 1. SplX did not affect outcome of DSS-induced colitis.
[A] Bodyweight loss of mice over time, compared with day 0. Data are expressed as mean and standard deviation (SD).
[B] Disease activity index.
[D] mRNA levels of IL-1β, IL-6 and TNF-α. mRNA levels are normalized against the reference genes Cyclophilin and RPLP0. N = 8-10 per group. Data are expressed as mean or median and individual data points. * indicates a significant difference compared to control. Statistical differences between control and DSS-treated mice and between sham mice and SplX were assessed using an independent t-test or a Mann-Whitney U test. P < 0.05 was considered significant. DSS: dextran sulfate sodium; Splx: absence of splenic innervation; TNF: tumor necrosis factor; IL: interleukin; LPS: lipopolysaccharide; RPLP0: Ribosomal Protein Lateral Stalk Subunit P0. Additional file 4. Schematic overview of experimental setup for SpNS. Differentially expressed genes were assessed between colonic samples of mice that did not receive stimulation or DSS, sham stimulated mice with DSS-induced colitis, and SpNS-treated mice with DSS-induced colitis. SpNS: splenic nerve bundle stimulation; DSS: dextran sulfate sodium.
Additional file 5. Transcriptomic changes in the colons from colitic mice. Volcano plot of the -log 10 (-P) on the Y-axis vs. the mean log2 fold change on the X-axis of mice that received DSS vs. control mice. Genes that were significantly different are indicated in blue (downregulated) and red (upregulated). DSS: dextran sulfate sodium; TNF: tumor necrosis factor; IL: interleukin; MMP: matrix metalloproteinase; ACOD: cis-aconitate decarboxylase; CXCL: chemokine (C-X-C motif ) ligand; S100: S100 calciumbinding protein.