Galectin-3 causes enteric neuronal loss in mice after left sided permanent middle cerebral artery occlusion, a model of stroke

In addition to brain injury stroke patients often suffer gastrointestinal complications. Neuroimmune interactions involving galectin-3, released from microglia in the brain, mediates the post-stroke pro-inflammatory response. We investigated possible consequences of stroke on the enteric nervous system and the involvement of galectin-3. We show that permanent middle cerebral artery occlusion (pMCAO) induces loss of enteric neurons in ileum and colon in galectin-3+/+, but not in galectin-3−/−, mice. In vitro we show that serum from galectin-3+/+, but not from galectin-3−/−, mice subjected to pMCAO, caused loss of C57BL/6J myenteric neurons, while myenteric neurons derived from TLR4−/− mice were unaffected. Further purified galectin-3 (10−6 M) caused loss of cultured C57BL/6J myenteric neurons. Inhibitors of transforming growth factor β-activated kinase 1 (TAK1) or AMP activated kinase (AMPK) counteracted both the purified galectin-3 and the galectin-3+/+ pMCAO serum-induced loss in vitro. Combined we show that stroke (pMCAO) triggers central and peripheral galectin-3 release causing enteric neuronal loss through a TLR4 mediated mechanism involving TAK1 and AMPK. Galectin-3 is suggested a target for treatment of post-stroke complications.

Ischemic cerebral stroke is a major cause of death worldwide and is at present a leading cause of adult mortality and disability 1 . In addition to the brain injury, stroke patients are at risk of a number of comorbidities including cardiac, pulmonary and gastrointestinal (GI) complications as well as infections and sepsis, all affecting recovery 2 . GI complications promote bacterial translocation and include dysphagia, dysmotility and colorectal dysfunction 2 , which rely on enteric nervous system (ENS) control. Experimental stroke models have revealed that rodents subjected to unilateral permanent middle cerebral artery occlusion (pMCAO) exhibit decreased GI motility, intestinal mucosa damage, increased levels of circulating ghrelin 3 , reduced T-and B cell counts in Peyer's patches 4 , and bacterial translocation 5 . Whether neurological manifestation in the ENS post stroke occurs in parallel to that in the central nervous system (CNS) is unknown. Post stroke neuroimmune interactions involving microglia and recruited peripheral macrophages determine neurological outcome in the CNS 6 . Galectin-3 (gal-3) is an evolutionarily preserved beta-galactoside-binding protein with both pro-and anti-inflammatory properties 7 . Recently it was shown that gal-3 secreted from IBA-1-(marker of microglia and recruited peripheral macrophages) positive cells is an endogenous toll-like receptor 4 (TLR4) agonist leading to a sustained pro-inflammatory response 8 . Gal-3 show ubiquitous subcellular distribution with multiple intra-and extracellular effects 9 . In the GI tract gal-3 stabilizes cell-cell junctions and enables polarization in intestinal epithelial cells 10,11 . Furthermore, it has been associated with GI cancers 12 , as well as with Crohn's disease and ulcerative colitis 13 . Elevated serum levels of gal-3 correlates with poor outcome after acute heart failure and high gal-3 values are observed in heart failure patients with a history of stroke 14 .
This study aims to investigate possible consequences of pMCAO on the ENS and the possible role of gal-3 in enteric neuropathy. Sera from pMCAO treated gal-3 +/+ , but not from gal-3 −/− , mice and purified gal-3 induce loss of myenteric neurons in vitro. To test if the enteric neuronal losses noted in vivo post pMCAO were mediated by way of circulating factors, sera from sham and pMCAO treated animals were added to primary cultures of myenteric neurons, derived from C57BL/6J mice. Control wells displayed 2.5 ± 0.2 neurons/mm 2 (n = 78) of evenly distributed neurons, with a varicose fibre network. No significant changes in neuronal survival were observed between controls and cultures treated with 3 ( Fig. 2A) or 7 (Fig. 2B) days sham sera at any concentration tested and results were pooled into ctrl, gal-3 +/+ sham and gal-3 −/− sham. Exposure with pMCAO serum from gal-3 −/− mice did not change neuronal survival in any of the concentrations tested. Exposure of serum from pMCAO treated gal-3 +/+ mice 3 (P < 0.001) and 7 (P < 0.001) days post pMCAO, caused a significant neuronal loss. Results are summarized in Fig. 2A,B. Since gal-3 in the CNS is able to mediate both pro-and anti-inflammatory responses in response to injury [16][17][18] , we investigated the effect of purified gal-3 (10 −6 M) on cultured enteric neurons, and found a significant loss (P < 0.05) (Fig. 2C).

Gal-3 is highly expressed in peripheral immune cells.
Western blot analysis of tissues from digestive organs and immune cells from peritoneal lavage were conducted to determine the possible source of gal-3. Western blot revealed high levels of gal-3 in peripheral immune cells and low levels in digestive organs (Fig. 2F).

Discussion
Stroke triggers an inflammatory cascade caused by e.g. reactive oxygen or nitrogen species and damage-associated molecular pattern (DAMP) signals released from neurons, glia, microglia and vascular cells in the affected area. DAMP signals mediate inflammatory responses utilizing the same pattern recognition receptor (PRR) system as pathogens 20 . This concept of sterile inflammation highlights the ability of the immune system to sense and react to internal as well as external clues 20,21 . In the hours, days and weeks after cerebral ischemia a peripheral immune response involving both innate and adaptive responses are present in blood with elevated levels of monocytes, activated-and regulatory T cells, and cytokines such as TNF, IL1β and IL6 20,22-24 . Possible enteric comorbidity or neuropathy in response to cerebral ischemia or the accompanying post stroke inflammation have not previously been investigated. In the current study we also investigated if gal-3 in the periphery, like in the central nervous system, is a mediator of the pro-inflammatory response post stroke.
pMCAO serum-induced neuronal losses could be prevented by the presence of either the TAK1 inhibitor or the AMPK inhibitor. We have previously shown that LPS-induced and TLR4 activated loss of myenteric neurons in vitro depend on this pathway 19 . Moreover, in contrast to C57BL/6J derived cultures, serum from gal-3 +/+ pMCAO mice did not affect neuronal survival in myenteric cultures from TLR4 −/− mice. These findings suggest that the mechanism behind pMCAO-induced enteric neuronal loss is by way of gal-3 triggered activation of TLR4. Together our data strongly suggest that focal ischemic stroke causes enteric neuronal loss, in parallel with central manifestations, and that the neuroimmune mechanism behind involves gal-3 and TLR4 receptor activation. This may explain the graded and extended (involving both submucous and myenteric neurons) neuronal susceptibility to pMCAO observed in vivo, as immunocytochemical analyses show that TLR4 is more abundantly expressed in the distal, compared to the proximal, bowel in both neurons and glia cells 26,27 . Astrocytes mediate a neurotoxic inflammatory response in the central nervous system through TLR4 activation causing oxidative stress 28 . As TLR4 is expressed in both neurons and glia in the ENS a toxic contribution through reactive oxygen species can't be entirely excluded. Regardless, gal-3 is revealed a key mediator of pMCAO-induced enteric neuropathy. This suggestion is further strengthened by our finding that the neuronal loss reached a plateau at 3 days, coinciding with the peak of gal-3 expression in microglia and recruited macrophages in the brain 29 . In addition analysis of gal-3 showed high levels of gal-3 in peritoneal immune cells.
Gal-3 is able to form stable complexes with LPS, potentiating LPS ability to activate immune cells such as neutrophils 9 . A similar mode of action could be at play in neurons, however analysis of sera from gal-3 +/+ and gal-3 −/− sham as well as pMCAO mice showed very low LPS levels across all groups and genotypes. Although unable to entirely dismiss LPS as a possible player in pMCAO-induced enteric neuronal loss in vivo, our data strongly suggests gal-3 to be the main mediator behind the enteric neuronal loss. Gal-3 acts as an endogenous TLR4 agonist able to, independent of LPS, mediate a cellular response on microglia 8 , and neutrophils 9 . Inhibition or genetic removal of TLR4 is neuroprotective in models of cerebral ischemia, highlighting a central role of TLR4 in neuroinflammation [30][31][32] . As shown in present study this extends also to enteric neurons, as myenteric cultures from TLR4 −/− mice were resistant to gal-3 +/+ pMCAO serum exposure. Interestingly both TAK1 and AMPK inhibition protect central 33,34 , and enteric neurons after cerebral ischemic damage suggesting common pathways activated.
In conclusion a cerebral ischemic event like stroke, modelled by pMCAO, causes neuroinflammation leading to a significant loss of enteric neurons. The enteric neuronal loss is targeted through a gal-3 mediated TLR4 activation independent of LPS. This finding highlights the presence of parallel manifestations in central and enteric nervous systems post stroke, involving neuroimmune interactions, gal-3 induced TLR4 activation and the TAK1/ AMPK pathway. This is the first report showing the existence of a neuroinflammatory response emanating in the central and transmitted to the peripheral i.e. enteric nervous system.

Methods and Material
Ethical statement. Procedures were approved by the regional Malmö/Lund committee for experimental animal ethics, Swedish board of Agriculture (M301-09 and M95-15). Animals were used in accordance with the European Community Council Directive (2010/63/EU) and the Swedish Animal Welfare Act (SFS 1988:534).
In separate experiments C57BL/6J mice were anesthetised (Ketalar, Rompun) and liver, fundus, ileum and colon were collected. Animals were killed by heart puncture. Muscularis propria was further separated from ileum and colon and samples were snap frozen in liquid nitrogen. Intestinal lavage was performed on C57BL/6J mice killed by cervical dislocation, 5 ml saline containing 5U heparin per ml were injected into the peritoneal cavity, followed by 5 min abdominal massage. The immune cell rich fluid (2.5 ml) was removed and centrifuged at 700 rcf for 3 min at room temperature, prior to protein extraction.
Endotoxin assay. Pooled serum samples 1-8 were analysed by Limulus amoebocyte lysate (LAL) assay for endotoxin levels with a commercially available kit (Pierce LAL chromogenic Endotoxin Quantification Kit, Thermo Fisher Scientific, SE) used according to the manufacturer's protocol.
Western blot. Total protein was extracted in RIPA buffer (Merck, SE) containing a soluble protease inhibitor cocktail (1 tablet per 50 mL RIPA buffer according to manufacturer's protocol, Thermo Fisher Scientific, SE). Protein concentrations were estimated using the Bradford Protein Quantification method (BCA Protein Assay KIT, Thermo Fisher Scientific, SE). Western blot analysis for gal-3 (M3-38) was performed using 10 μ g protein extract separated on pre-cast 4-20% SDS-PAGE gels (Stain-Free Gels, BioRad, SE). Precision Plus Protein Western Blot pre-stained standard (BioRad, SE) was used as molecular weight marker. Densitometry was by ChemiDoc XRS+ and Image Lab, both from BioRad analysis software.
Analysis. Submucous and myenteric neurons were counted in longitudinally cut sections from ileum and colon, using a computerized image analysing system (NDP view 2, Hamamatsu, JP). Sections comprised at least a total length of 15 mm cut at 3-5 different depths, per region and mouse. Results are expressed as numbers of HuC/D-immunoreactive (IR) submucous or myenteric neurons per mm section. Heights of mucosa and muscularis propria were measured on toluidine blue stained sections using mean values of 8-10 representative measurements from each region of each mouse. Survival of cultured neurons was calculated by counting the total number of PGP9.5-IR neurons within the wells using fluorescence microscopy (Olympus BX43, LRI, SE) with appropriate filter setting and expressed as percentage of the control well run in parallel.
LAL contents in serum samples 1-8 were plotted against a LPS standard curve using non-linearly fit.
Statistical analysis. Data are presented as means ± SEM and analysed by GraphPad Prism (GraphPad Software Inc, USA). In the experimental groups in vivo n = 3-11 animals and in the groups in vitro n = 3-18 repeats from a minimum of 3 different animals. Statistical significance was for in vivo data determined using two-way analysis of variance (ANOVA) and for in vitro data one-way ANOVA. These were followed by Dunnet's post hoc test towards the sham group (in vivo data) or untreated controls (in vitro data). T-test was used to determine differences between sham groups. A confidence level of 95% was considered significant.