Oxidative Stress-Induced HMGB1 Translocation in Myenteric Neurons Contributes to Neuropathy in Colitis

High-mobility group box 1 (HMGB1) is a damage-associated molecular pattern released by dying cells to stimulate the immune response. During cell death, HMGB1 is translocated from the nucleus to the cytoplasm and passively released. High levels of secreted HMGB1 are observed in the faeces of inflammatory bowel disease (IBD) patients, indicating its role in IBD pathophysiology and potential as a non-invasive IBD biomarker. HMGB1 is important in regulating neuronal damage in the central nervous system; its pathological activity is intertwined with oxidative stress and inflammation. In this study, HMGB1 expression in the enteric nervous system and its relevance to intestinal neuroinflammation is explored in organotypic cultures of the myenteric plexus exposed to oxidative stimuli and in Winnie mice with spontaneous chronic colitis. Oxidative stimuli induced cytoplasmic translocation of HMGB1 in myenteric neurons in organotypic preparations. HMGB1 translocation correlated with enteric neuronal loss and oxidative stress in the myenteric ganglia of Winnie mice. Inhibition of HMGB1 by glycyrrhizic acid ameliorated HMGB1 translocation and myenteric neuronal loss in Winnie mice. These data highlight modulation of HMGB1 signalling as a therapeutic strategy to reduce the consequences of enteric neuroinflammation in colitis, warranting the exploration of therapeutics acting on the HMGB1 pathway as an adjunct treatment with current anti-inflammatory agents.


Introduction
High-mobility group box 1 (HMGB1) is a highly conserved nuclear protein that binds DNA to stabilise the nucleosome and regulate transcription in physiological conditions [1,2]. HMGB1 has emerged as an important alarmin to cellular stress involved in the pathophysiology of several diseases [3]. In conditions of cellular stress, the two HMGB1 nuclear localization sites are hyperacetylated, and HMGB1 translocates to the cellular cytoplasm, where it is released actively by leukocytes or passively by necrotic and necroptotic cells [4]. In the extracellular space, HMGB1 functions as a damage-associated molecular pattern (DAMP), acting as a direct chemoattractant via the C-X-C chemokine ligand 12 (CXCL12) and C-X-C chemokine receptor 4 (CXCR4) axis, as well as stimulating the release of proinflammatory mediators via toll-like receptors (TLR)-2 and TLR-4 expressed on nearby immune cells [2].

Organotypic Culture of Myenteric Ganglia
The distal colon was collected and pinned in silicon-lined Petri dishes containing Hank's balanced salt solution (Sigma-Aldrich, Sydney, NSW, Australia) to remove the mucosa and submucosa layers before being cut into 1.5 × 1.5 cm sheets. The organotypic sheet preparations were loosely pinned into 24-well cell culture plates modified to contain a silicon elastomer (Sylgard; Dow Corning, MI, USA) that covered the bottom of the wells with a depth of~5 mm. Preparations were incubated (37 • C, 5% CO 2 ) for 24 h in α-MEM supplemented with 100 U/mL penicillin/streptomycin, 1% glutaMAX and 5% (v/v) foetal bovine serum (FBS) (Gibco ® , Life Technologies, Melbourne, VIC, Australia, for all, unless stated otherwise). Hyperoxic oxygen tension, and the chemical H 2 O 2 , were utilised as oxidative stimuli in organotypic preparations. A hyperoxic environment (95% O 2 and 5% CO 2 ) was formed using a self-contained modular incubator (Billups-Rothenberg, Inc., San Diego, CA, USA). Organotypic cultures were exposed to H 2 O 2 diluted in α-MEM at a final concentration of 100 µM. An equal volume of the α-MEM vehicle was applied to control cultures. Lipopolysaccharide (LPS, Sigma-Aldrich) was applied at 20 µg/mL [1]. Organotypic preparations were cultured for 24 h before being fixed overnight at 4 • C in Zamboni's fixative (2% formaldehyde and 0.2% picric acid). Preparations were subsequently washed in DMSO (Sigma-Aldrich) (3 × 10 min) to permeabilise the tissue and phosphate buffered saline (PBS) (3 × 10 min) to remove DMSO and the fixative for immunohistochemical experiments.

Glycyrrhizic Acid Administration
HMGB1 was inhibited by glycyrrhizic acid (GA) (Sigma-Aldrich) (10 mg/kg), which was dissolved in 2% cremophor (Sigma-Aldrich): 2% EtOH and 96% sterile water and given to mice via i.p. injections twice daily for 14 days with no less than 8 h between injections. The dose of GA used in our study was effective in reducing oxidative stress, inflammation and neuronal apoptosis in a rat model of ischemic brain injury [2]. The vehicle group received 2% cremophor: 2% EtOH, and 96% sterile water via i.p. injections twice daily for 14 days with no less than 8 h between injections. The volumes for all injections were calculated to each animal's body weight with a maximum of 200 µL per injection. Mice were culled 24 h after the final treatment for tissue collection.

MitoSOX Red Fluorescent Staining
The production of O 2 − in the myenteric plexus was assessed in freshly excised colon segments. Tissues were collected in physiological saline (composition in mmol L −1 : NaCl, 117; NaH 2 PO 4 , 1.2; MgSO 4 , 1.2; CaCl 2 , 2.5; KCl, 4.7; NaHCO 3 , 25; and glucose, 11), which was gently bubbled with carbogen (95% O 2 -5% CO 2 ) throughout the entire procedure. Tissues were viewed under a dissection microscope, cut along the mesenteric border and pinned in a silicon-lined Petri dish containing physiological saline. The mucosa, submucosa and circular smooth muscle layers were removed to expose the LMMP. Tissues were incubated for 40 min at 37 • C in physiological saline containing MitoSOX Red (1:1000) (Molecular Probes ® , Thermofisher, Melbourne, VIC, Australia). Tissues were then washed in physiological saline (3 × 10 min) prior to being fixed in 4% paraformaldehyde overnight at 4 • C. Paraformaldehyde was removed by washing the tissues with PBS (3 × 10 min) before they were mounted onto glass slides with DAKO fluorescence mounting medium (Agilent Technologies, Melbourne, VIC, Australia) and visualised by confocal microscopy.

Imaging and Analysis
An Eclipse Ti confocal laser scanning system (Nikon, Tokyo, Japan) was used to visualise MAP-2, HMGB1, CD45, 8-OHdG immunofluorescence or the MitoSOX red fluorogenic probe. Identical acquisition settings were used between samples in all experiments. Images were collected as ND2 files containing all metadata, including fluorescence signals at all Z levels. Images were visualised using Image J v1.50b open source software (National Institute of Health, Bethesda, MD, USA) [9,10] with the Image J ND2 Reader plugin and were converted into maximum intensity projections in 16-bit TIFF format. All subsequent analysis was performed using Image J. For all analyses, average mean values were calculated from eight individual images per sample.
For HMGB1 and MAP-2 quantification in wholemount LMMPs and organotypic preparations, Z-series images were randomly acquired using the 40X objective at a thickness of 1 µm. The average MAP-2-immunoreactive (IR) neuronal density was calculated within a 0.1 mm 2 (316.23 µm × 316.23 µm = 100,000 µm 2 ) field of view per image as previously described [11]. In the same images, HMGB1 expression in the ganglia was quantified using the cell counter plugin of ImageJ software. HMGB1 was determined to be expressed by MAP-2-IR neurons or other non-neuronal cells. The intracellular location of HMGB1 in these cells was determined as localised in the nucleus, absent in the nucleus or translocated to the cytoplasm with the aid of the nuclear marker DAPI. HMGB1 expression in cells was reported as the number of cells per 0.01 mm 2 (100 µm × 100 µm = 10,000 µm 2 ) of the ganglionated area or expressed as a percentage of total neurons. HMGB1 expression was pseudo-coloured green for all imaging in wholemounts and cross sections for greater visual distinction.
Leukocytes immunoreactive for CD-45 were quantified in images of wholemount LMMP preparations. Z-series images were randomly acquired using the 40X objective at a thickness of 1 µm.
CD-45-IR cells were enumerated within a 0.1 mm 2 (316.23 µm × 316.23 µm = 100,000 µm 2 ) field of view using the cell counter plugin of ImageJ software. Leukocytes were further classified by their location in relation to the MAP-2-IR ganglia (intra-ganglionic, periphery of ganglia, extra-ganglionic). Values were expressed as the average number of CD45-IR cells per 0.1 mm 2 area from eight images per mouse.
To measure 8-OHdG adducts in MAP-2-IR myenteric neurons, Z-series images were randomly acquired using the 60X objective at a thickness of 1µm. Eight randomly captured 16-bit images were analysed with a field of view of 0.04 mm 2 (200 µm × 200 µm = 40,000 µm 2 ) per individual image. Regions of interest were set over MAP-2-IR neurons. Images of 8-OHdG immunofluorescence were converted to binary images by thresholding for highintensity labelling [12]. Regions of interest defining the location of MAP-2-IR neurons were overlayed onto binary 8-OHdG images. Binary particles were then analysed to obtain the percentage area of 8-OHdG immunoreactivity within the area of myenteric neurons.
Mitochondria-derived O 2 − was visualised using the MitoSOX fluorogenic probe in wholemount LMMP preparations as previously described [13]. Eight Z-series images were randomly acquired using the 40X objective at a thickness of 1 µm; 16-bit images with a field of view of 0.1 mm 2 (316.23 µm × 316.23 µm = 100,000 µm 2 ) per individual image were analysed. Regions of interest were set to determine MitoSOX fluorescence within the myenteric ganglia. The fluorescence intensity of O 2 − in the ganglia was measured as the mean grey value (mean fluorescence intensity) of the pixels within the regions of interest.

Statistical Analysis
Data analysis was performed using GraphPad Prism v7 (GraphPad Software Inc., San Diego, CA, USA). A one-way ANOVA was performed with a post hoc Holm-Sidak test for multiple comparisons. X and Y correlations were determined using a linear regression analysis with p values for significant slope relationships recorded. For all analyses, p ≤ 0.05 was considered significant. All data were presented as mean ± standard error of the mean (SEM).

HMGB1 Is an Endogenous TLR Agonist Highly Expressed in the ENS
Previous studies implicate TLR2/4 in neuronal loss induced by LPS in cultured enteric neurons in vitro [1]; however, the physiological relevance of these signalling pathways is uncertain. LPS-induced neuropathy was replicated in the organotypic culture of the colonic LMMP preparations. LPS caused a significant reduction in neuronal density (16.7 ± 1.8, n = 7, p < 0.001) compared to controls (28.0 ± 1.7, n = 6, Figure 1A-A"). To identify candidate ligands for TLR signalling in enteric neurons, the publicly available String protein interaction database and the mouse brain single-cell transcriptome database, which contains data on enteric neurons, was utilized [14]. For the protein interaction analysis, only proteins that were part of a physical complex and with validated experimental evidence were included (confidence 0.4). TLR2 and TLR4 have strong physical interactions with the endogenous ligand HMGB1, as reported in immune cells [15] ( Figure 1B). A survey of the mouse brain database revealed that HMGB1 expression is particularly high in enteric neurons (inhibitory and excitatory) and enteric glial cells ( Figure 1C), making this molecule a strong candidate for local TLR signalling in the ENS during inflammation.

HMGB1 Is an Endogenous TLR Agonist Highly Expressed in the ENS
Previous studies implicate TLR2/4 in neuronal loss induced by LPS in cultured enteric neurons in vitro [1]; however, the physiological relevance of these signalling pathways is uncertain. LPS-induced neuropathy was replicated in the organotypic culture of the colonic LMMP preparations. LPS caused a significant reduction in neuronal density (16.7 ± 1.8, n = 7, p < 0.001) compared to controls (28.0 ± 1.7, n = 6, Figure 1A-A''). To identify candidate ligands for TLR signalling in enteric neurons, the publicly available String protein interaction database and the mouse brain single-cell transcriptome database, which contains data on enteric neurons, was utilized [14]. For the protein interaction analysis, only proteins that were part of a physical complex and with validated experimental evidence were included (confidence 0.4). TLR2 and TLR4 have strong physical interactions with the endogenous ligand HMGB1, as reported in immune cells [15] ( Figure 1B). A survey of the mouse brain database revealed that HMGB1 expression is particularly high in enteric neurons (inhibitory and excitatory) and enteric glial cells ( Figure 1C), making this molecule a strong candidate for local TLR signalling in the ENS during inflammation.

Chronic Intestinal Inflammation Associates with HMGB1 Translocation in Enteric Neurons
Expression of HMGB1 was assessed in control mice and Winnie mice with chronic colitis. In control mice, HMGB1 was expressed in the nuclei of virtually all cells in the LMMP, with strong staining in neurons and glia in the myenteric ganglia ( Figure 2A). In mice with chronic colitis, obvious variations in the expression pattern of HMGB1 were observed, with HMGB1 becoming translocated to the cytoplasm or absent in many neurons, which is characteristic of the secretory mechanism of HMGB1 in cells undergoing necroptosis ( Figure 2A). In Winnie mice, HMGB1 translocation in myenteric neurons was often observed in clusters, as opposed to sporadically throughout the ganglia, which was suggestive of a localised effect ( Figure 2A). In the myenteric ganglia, the number of HMGB1expressing cells was decreased in Winnie mice compared to C57BL/6 controls (p < 0.05, Figure 2B, Supplementary Table S1, n = 5 animals/group). The number of MAP2 + neurons expressing HMGB1 in the nuclei was decreased in Winnie mice compared to C57BL/6 controls (p < 0.05, Figure 2B', Table S1, n = 5 animals/group). Loss of nuclear HMGB1 expression appeared to be specific to the neuronal population, with no effects observed in non-neuronal MAP2 − cells within the myenteric ganglia in conditions of chronic inflammation ( Figure 2B", Table S1, n = 5 animals/group). Nevertheless, these results were potentially affected by alterations in the neuronal density, as previously reported in Winnie mice [16]. To account for this, the expression patterns of HMGB1 were further investigated as a percentage of neurons within each sample ( Figure 2C,D). The percentage of neurons without nuclear expression of HMGB1 was increased in Winnie mice compared to C57BL/6 mice (p < 0.05, Figure 2C, Table S1, n = 5 animals/group). Given that HMGB1 release is a marker of necroptosis [17] and HMGB1 is an active DAMP, the relationship between HMGB1 expression and myenteric neuronal loss was explored by linear regression analysis. A negative correlation was observed between the proportion of neurons without nuclear HMGB1 and the neuronal density of the myenteric ganglia in vivo (F(1,13) = 14.07, p < 0.05, Figure 2C'). Similar to these results, the percentage of neurons with HMGB1 translocation into the cytoplasm was increased in Winnie mice compared to C57BL/6 mice (p < 0.05, Figure 2D, Table S1, n = 5 animals/group). The percentage of neurons with cytoplasmic HMGB1 expression negatively correlated with neuronal density (F(1,13) = 9.437, p < 0.05, Figure 2D'). Notably, neuronal density did not correlate with HMGB1 expression patterns in non-neuronal cells (F(1,13) = 0.06467, p = 0.8).

HMGB1 Translocation in Enteric Neurons and Neuropathy Is Mediated by Oxidative Stimuli
Cytoplasmic HMGB1 translocation and release have previously been associated with oxidative stress in the CNS neurons [18]. Considering HMGB1 translocation can be redoxsensitive, we explored the effects of chronic intestinal inflammation on oxidative stress in the myenteric ganglia ( Figure 3). To determine the levels of oxidative stress in chronic inflammation, DNA/RNA adducts were assessed by immunohistochemistry in cross sections of the distal colon by labelling 8-OHdG ( Figure 3A-B"'). Localisation of 8-OHdG to myenteric neurons was determined by measuring the area of 8-OHdG-IR adducts in myenteric neurons labelled by MAP-2. In myenteric neurons, 8-OHdG was greatly elevated in Winnie mice (30.6 ± 5.5%, n = 7 animals) compared to C57BL/6 mice (12.7 ± 3.6%, p < 0.05, n = 6 animals) ( Figure 3C). High levels of O 2 − , as indicated by the intensity of fluorescence, were observed in the myenteric ganglia compared to the muscle layer suggesting that the myenteric plexus is − levels (mean fluorescence intensity) in the myenteric ganglia (127.5 ± 22.5 arb. units, n = 6 animals) were more than twice the levels of that observed in control C57BL/6 mice (50.4 ± 3.2 arb. units, p < 0.05, n = 5 animals, Figure 3E).   Comparable results were observed in organotypic cultures exposed to H2O2. Application of H2O2 (n = 9 independent cultures) decreased the total number of neurons with nuclear HMGB1 expression ( Figure 5B Table S2). A negative correlation was observed between the number of neurons with cytoplasmic HMBG1 and neuronal density (F(1,13) = 38.12, p < 0.0001, R 2 = 0.7457) ( Figure 5E, Supplementary Table S4). Similarly, the proportion of neurons with HMGB1 expression in the nucleus decreased (p < 0.05) and the proportion of neurons without nuclear HMGB1 (p < 0.05) or those with cytoplasmic HMBG1 (p < 0.05) increased ( Figure 5F-F'', Table S2). The proportion of neurons with HMGB1 translocation was negatively correlated to neuronal density (F(1,13) = 26.21, p < 0.001) with an R 2 of 0.6685 (Figure 5G, Table S4). Similar to organotypic cultures under hyperoxia, no difference was observed in the number of non-neuronal cells with nuclear or translocated HMGB1 in the myenteric ganglia after treatment with H2O2 ( Figure 5H,H', Table S2). Together, these data demonstrate that HMGB1 translocation is regulated by oxidative stress in myenteric neurons and highlights the relationship between HMGB1 and neuronal loss under oxidative conditions. To determine the effects of oxidative stress on HMGB1 expression in the myenteric ganglia, ex vivo organotypic cultures were subjected to hyperoxic (95% O 2 and 5% CO 2 incubation) ( Figure 4A Table S2). Punctate immunoreactivity of HMGB1 in the cytoplasm was observed in neurons with high levels of HMGB1 translocation. A negative correlation was found between the latter and the neuronal density in the myenteric ganglia (F(1,10) = 13.78, p < 0.01) with an R 2 of 0.5794 ( Figure 4E, Supplementary Table S3). To account for the decrease in neuronal density caused by hyperoxia, the proportions of neurons with HMGB1 expression were also investigated ( Figure 4F-F"). Similar to total counts, the proportion of neurons with HMGB1 expression in the nuclei decreased (p < 0.05), and the proportions of neurons without nuclear HMGB1 (p < 0.05) or with cytoplasmic translocation increased (p < 0.01) ( Figure 4F-F", Table S2). Furthermore, a negative correlation was observed between the proportion of neurons with HMGB1 translocation and neuronal density (F(1,10) = 17.5, p < 0.01) with an R 2 of 0.6364 ( Figure 4G, Table S3). No changes were observed in the total number of non-neuronal cells in the myenteric ganglia that expressed HMGB1 in the nucleus or the cytoplasm after hyperoxia ( Figure 4H,H', Table S2).
Comparable results were observed in organotypic cultures exposed to H 2 O 2 . Application of H 2 O 2 (n = 9 independent cultures) decreased the total number of neurons with nuclear HMGB1 expression ( Figure 5B-B"") and increased the total number of neurons with cytoplasmic HMGB1 translocation ( Figure 5C-C"") compared to controls (p < 0.05, for both) ( Figure 5D,D', Table S2). A negative correlation was observed between the number of neurons with cytoplasmic HMBG1 and neuronal density (F(1,13) = 38.12, p < 0.0001, R 2 = 0.7457) ( Figure 5E, Supplementary Table S4). Similarly, the proportion of neurons with HMGB1 expression in the nucleus decreased (p < 0.05) and the proportion of neurons without nuclear HMGB1 (p < 0.05) or those with cytoplasmic HMBG1 (p < 0.05) increased ( Figure 5F-F", Table S2). The proportion of neurons with HMGB1 translocation was negatively correlated to neuronal density (F(1,13) = 26.21, p < 0.001) with an R 2 of 0.6685 ( Figure 5G, Table S4). Similar to organotypic cultures under hyperoxia, no difference was observed in the number of non-neuronal cells with nuclear or translocated HMGB1 in the myenteric ganglia after treatment with H 2 O 2 ( Figure 5H,H', Table S2). Together, these data demonstrate that HMGB1 translocation is regulated by oxidative stress in myenteric neurons and highlights the relationship between HMGB1 and neuronal loss under oxidative conditions.
The effects of HMGB1 inhibition on plexitis in the myenteric ganglia of Winnie mice were evaluated to determine whether the chemotactic properties of HMGB1 on leukocytes could explain its effects on neuronal loss. Leukocytes visualised by anti-CD45 antibody in LMMP wholemount preparations were observed in proximity to the myenteric ganglia using the pan-neuronal marker MAP-2 ( Figure 8A-C"'). Compared to C57BL/6 mice, both sham-treated and GA-treated Winnie mice exhibited higher numbers of leukocytes in the intra-ganglionic region (Winnie-sham, p < 0.001 and Winnie + GA, p < 0.01), the periphery of the ganglia (p < 0.0001, for both), extra-ganglionic region (p < 0.0001, for both) and in total (p < 0.0001, for both) ( Figure 8D, Supplementary Table S5; n = 6 animals/group). No differences were observed between sham-treated and GA-treated Winnie mice. These results demonstrate that the neuroprotective properties of HMGB1 inhibition were not mediated by the attenuation of plexitis.
age of neurons with translocation of HMGB1 from the nucleus to the cytoplasm wa vated in Winnie mice (11.5 ± 4.5%) compared to control C57BL/6 mice (0.5 ± 0.3%, p < this was ameliorated by GA treatment, where virtually no cytoplasmic HMGB1 tra cation was observed in neurons (p < 0.05) ( Figure 7G; n = 5 animals/group).    sham-treated and GA-treated Winnie mice exhibited higher numbers of leukocytes in the intra-ganglionic region (Winnie-sham, p < 0.001 and Winnie + GA, p < 0.01), the periphery of the ganglia (p < 0.0001, for both), extra-ganglionic region (p < 0.0001, for both) and in total (p < 0.0001, for both) ( Figure 8D, Supplementary Table S5; n = 6 animals/group). No differences were observed between sham-treated and GA-treated Winnie mice. These results demonstrate that the neuroprotective properties of HMGB1 inhibition were not mediated by the attenuation of plexitis.

Discussion
In this study, we identify HMGB1 as an important endogenous cytokine acting locally in the myenteric ganglia to contribute to enteric neuroinflammation. HMGB1 was constitutively expressed in the nuclei of myenteric neurons. A high number of HMGB1 translocation events was observed in the myenteric neurons of Winnie mice with chronic inflammation. This directly correlated with the degree of neuronal loss and was associated with elevated oxidative stress in the ganglia. Using organotypic cultures, we demonstrate that translocation of HMGB1 in myenteric neurons is a process regulated by the redox environment and can be induced by oxidative insult. In Winnie mice and organotypic models of oxidative injury, neurons had a high specificity for HMGB1 translocation in the myenteric ganglia. Inhibition of HMGB1 translocation by GA averted neuronal loss indicates a direct role of HMGB1 in enteric neuropathies downstream of oxidative stress and local inflammation. HMGB1 has important physiological functions in the nucleus of cells, binding DNA and regulating transcription; however, it is also an alarmin to cellular stress that can be released extracellularly, eliciting an array of responses depending on the biological context [22]. HMGB1 responds to changes in the redox environment and oxidative stress [17]. These processes are evident in cell death which is closely associated with the pattern of HMGB1 expression [17,22,23]. In our study, HMGB1 was either absent in the nucleus or translocated into the cytoplasm of myenteric neurons. The cellular compartmentalisation of HMGB1 is redox-sensitive, and translocation is induced by intracellular ROS [24]. Hyperacetylation of lysines in the two HMGB1 nuclear localisation sites inhibits HMGB1 entry into the nucleus and results in its accumulation into the cytoplasm [25]. This is a feature of necrotic and necroptotic cell death pathways but not apoptosis, where conversely, HMGB1 irreversibly binds to nuclear DNA [17,22]. Considering the negative association between HMGB1 translocation and neuronal density, myenteric neuronal loss in chronic inflammation and oxidative stress is possibly mediated by a necrotic-like pathway of cell death. The strong relationship between HMGB1 expression and neuronal density suggests that HMGB1 can be a marker of myenteric neuropathy. However, HMGB1 has also been demonstrated to contribute directly to the pathophysiological process of cell death in a broad range of cell types, including renal cells, hepatocytes and neurons of the CNS [26][27][28][29]. GA inhibited HMGB1 translocation and dramatically attenuated myenteric neuronal loss, suggesting that HMGB1 also had a direct role in myenteric neuropathy and did not serve simply as a marker of cell death. Notably, GA exhibits very weak radical scavenging properties; therefore, a direct antioxidant effect is unlikely to explain these effects [30]. Rather, GA has been shown to bind directly to extracellularly released HMGB1 and can inhibit its pro-inflammatory and chemoattractant properties [31]. The effect of GA on the intranuclear binding function of HMGB1 is weak, which may explain why inhibition of HMGB1 by this compound has a low chance of causing deleterious effects and is approved for clinical use [31,32].
In our study, GA inhibited the translocation of HMGB1 into the cytoplasm. Similarly, studies in the CNS have observed that GA inhibits HMGB1 translocation and secretion under conditions of cellular stress [27,28,33,34]. However, it is uncertain whether GA inhibited translocation by binding HMGB1 within the cell or by another mechanism. Considering that GA can weakly affect HMGB1 activity in the nucleus, there may be a degree of affinity for GA to bind HMGB1 intracellularly [31]. Alternatively, it has been postulated that released HMGB1 can induce HMGB1 translocation and further release in cells that are receptive to extracellular HMGB1, including neurons [33,35]. Myenteric neurons expressing translocated HMGB1 were often observed in clusters rather than sporadically throughout the ganglia. The proximity of neighbouring myenteric neurons with HMGB1 translocation may support the notion that released HMGB1 induces further translocation. Myenteric neurons express the major receptors for HMGB1, including the receptor for advanced glycation end-products (RAGE), TLR-2 and TLR-4 [36][37][38]. Given that the TLR-2 and 4 agonist LPS were shown to cause myenteric neuropathy, TLRs appear to be highly relevant to the mechanisms of enteric neuroinflammation. Future studies should clarify whether HMGB1 could also act via RAGE in enteric neurons. Notably, activation of TLR-4 is also known to induce HMGB1 translocation [35]. Thus, it could be predicted that TLR-induced HMGB1 translocation and release can drive further neuropathy.
Although HMGB1 is ubiquitously expressed, the CNS has been a focus for the role of HMGB1 in pathophysiological conditions, including ischemia, reperfusion, haemorrhage and physical trauma [27,28,33,34]. In the nervous system, alterations to HMGB1 expression and its release appear to be specific to neurons [18,33]. In in vitro and in vivo models of experimental subarachnoid haemorrhage, HMGB1 translocation was observed primarily in neurons, whilst HMGB1 translocation events were rare in the glia [18]. Similarly, in a model of traumatic brain injury, only MAP-2 positive neurons exhibited HMGB1 translocation, which was not observed in astrocytes or microglia [33]. This suggests that HMGB1 translocation in nervous system pathologies is relatively specific to neurons. In our study, changes in the number of HMGB1 translocation events were observed in neurons and not in other cell types of the myenteric ganglia in conditions of inflammation and oxidative stress. Therefore, like in the CNS, only neuronal HMGB1 appears to be involved in the pathology of damage to the ENS, particularly driving neuronal loss.
Mechanisms of neuronal death are highly diverse and can be interrelated. These include traditional caspase-dependent apoptosis pathways and various forms of programmed necrosis or necroptosis [39]. In response to oxidative stress and chronic colitis, myenteric neurons lost expression of HMGB1 in the nuclei and HMGB1 was translocated to the cytoplasm. This suggests that myenteric neurons might undergo necrotic or necroptotic cell death in these conditions. This is further supported by the correlation between the degree of HMGB1 translocation and neuronal loss, in addition to the neuroprotective effects of GA treatments. High levels of ROS are observed in necrosis [17,40]. Likewise, high levels of oxidised DNA/RNA adducts, a marker of oxidative stress-induced DNA damage, were observed specifically in myenteric neurons. Previously, it was observed in a model of dinitrobenzene sulfonic acid (DNBS)-induced colitis that cleaved caspase-3 can be present in 1.4% of neurons at 1.5 h after installation [41]. Comparatively, there was a 42% loss in myenteric neurons [41]. This suggests that a small number of myenteric neurons show signs of apoptosis in the acute stages of chemically-induced colitis; however, other caspase 3-independent or non-apoptotic cell death pathways may also be involved.
HMGB1 is well-established in driving inflammatory and autoimmune diseases [42,43]. In IBD, HMGB1 is implicated in the pathophysiology of intestinal inflammation and has become of interest as a useful faecal biomarker of paediatric and adult IBD [44,45]. In animal models of colitis, increased HMGB1 expression in colon tissues has been observed in mouse models of 2,4,6-trinitrobenzenesulfonic acid (TNBS), DSS chemically-induced colitis and interleukin (IL)-10 deficient mice [46][47][48]. In TNBS-induced colitis and IL-10 −/− mice, administration of ethyl pyruvate attenuated colitis and reduced HMGB1 expression [46]. Concomitantly, the antioxidant enzyme haem oxygenase 1 (HO-1) was upregulated, which may indicate that inhibition of oxidative stress contributed to the decrease in HMGB1 [46]. In DSS-induced colitis, the efficacy of direct HMGB1 inhibitors has been evaluated. Daily treatment with dipotassium glycyrrhizate reduced HMGB1 release and attenuated colitis [47]. In a similar model, treatment with an anti-HMGB1 antibody before administration of DSS improved histological scores but had modest effects on clinical scores [48]. Similar results were observed using GA in our study, with a significant but modest reduction in disease activity. However, in our study, GA was applied after chronic inflammation was established; thus, GA was not investigated as a preventative treatment like previous studies and may offer higher efficacy if employed before the onset of inflammation. In studies of colitis, HMGB1 release was associated with a pro-inflammatory role. Nevertheless, in our study, GA treatments did not appear to yield an anti-inflammatory effect near the myenteric ganglia, with high numbers of leukocytes observed in all regions proximal to myenteric neurons despite the inhibition of neuronal HMGB1 translocation. This may suggest that plexitis is upstream of HMGB1 translocation in myenteric neurons.
In Winnie mice, GA treatments paralleled the attenuation of diarrhoea and crypt hyperplasia. Considering that secretion and motility are neurally-regulated processes and dysfunction in the ENS is associated with altered stool consistencies in human pathologies, and experimental models [19][20][21], the prevention of diarrhoea could be a result of reversing neuronal damage. Likewise, ablation of the ENS causes epithelial hyperplasia indicating an important role for mucosal maintenance by the ENS [49], which may explain the correlation between the neuroprotective effects of GA and reduction in epithelial hyperplasia in colitis. Taken together, these results indicate that GA prevents enteric neuropathy mediated by HMGB1 translocation and extracellular release, and this property could be responsible for mitigating diarrhoeal symptoms. Nevertheless, the administration of GA appeared to have a limited effect on inflammation. Signs of active inflammation remained high, and levels of myenteric immune cell infiltrate were not attenuated by GA. These data suggest that GA does not inhibit plexitis and acts via downstream inhibition of HMGB1 signalling in enteric neurons, which prevents further HMGB1 translocation and neuropathy in neighbouring cells. However, it should be considered that high levels of ROS and proinflammatory cytokines modify the electrochemical properties and promote dysfunction in enteric neurons [50][51][52][53][54][55]. Therefore, despite GA promoting neuronal survival, perturbations in their function are still possible.
Considering GA had limited efficacy on leukocyte infiltration, treatment efficacy may be increased in combination with other anti-inflammatory agents to target multiple mechanisms of disease. Therapeutics that act upstream of the HMGB1 pathway may also warrant attention. As HMGB1 translocation was found to be driven by oxidative stress, a combination with antioxidant compounds could be explored to suppress enteric neuroinflammation. However, the limitations of antioxidant compounds, such as low bioavailability and difficulty in achieving therapeutic concentrations in vivo, must also be considered [56,57]. Alternatively, downstream targets of the HMGB1 pathway could also offer a therapeutic avenue. For example, the TLR4 antagonist C34 has already demonstrated efficacy in necrotizing enterocolitis [58].

Conclusions
Translocation of the HMGB1 protein in myenteric neurons was associated directly with neuronal loss in chronic colitis. HMGB1 is ubiquitously expressed by most nucleated cells; however, the specificity for HMGB1 translocation in neurons is a feature of CNS pathologies associated with inflammation and oxidative stress. Likewise, our study demonstrates that translocation of HMGB1 within cells of the myenteric ganglia was only increased in neurons in chronic inflammation conditions. The expression pattern of HMGB1 in myenteric neurons suggests that neuronal loss was possibly caused by necroptotic-like mechanisms. High levels of oxidative stress and ROS production are observed in myenteric neurons in chronic colitis. Oxidative stimuli were demonstrated to directly induce HMGB1 translocation in myenteric neurons; therefore, oxidative stress appeared to instigate HMGB1 translocation and neuronal loss. Inhibition of HMGB1 by GA validated its importance as a key regulator of neuronal loss in vivo. GA did not reduce inflammation or plexitis in the myenteric ganglia, which appear to be upstream events of HMGB1 translocation and neuronal loss. Notwithstanding, therapeutic avenues to modulate HMGB1 signalling warrant exploration as adjunct treatments for colitis to suppress the neuroinflammatory component.