Dietary Salt Exacerbates Experimental Colitis

This article is featured in In This Issue, p.829

The Western diet, characterized by foods high in salt, sugar, protein, and fat, but low in fiber from fruits and vegetables, has long been understood to drive atherosclerosis, diabetes, inflammatory bowel disease (IBD), and other diseases of the developed world (1, 2). For example, the incidence of Crohn’s disease, one of the two primary IBDs, is four times higher in North America than in the Middle East or Asia, reflecting the fact that environmental stimuli like diet have driven enormous surges in IBD incidence in developed countries in the last 50 y (3–5). Further, numerous recent studies have linked elements of the Western diet to poor gastrointestinal health: diets high in sugar and fat lead to a dysbiotic, proinflammatory microbiota (6), whereas diets low in fiber lead to poor immunosuppressive activity of regulatory T cells and dysfunctional immunity (7).

One aspect of the Western diet that has not been studied in the context of IBD, however, is dietary salt (NaCl). Processed foods typical of a Western diet can contain 100 times the NaCl of similar home-cooked meals (8, 9), and many Americans consume more than twice the American Heart Association’s daily recommended NaCl content. High NaCl consumption is associated with increased risk of developing high blood pressure (10), correlates with increased disability and progression in multiple sclerosis (11), interferes with and damages normal kidney function (10, 12–14), and may be involved in atherosclerosis (15). These data are corroborated by studies in human volunteers, where a controlled high salt diet correlated with enhanced blood monocyte counts and proinflammatory serum cytokines (including IL-17, IL-6, and IL-23) (14, 16, 17).

Several recent studies have shown that a high salt diet (HSD) in laboratory mice is proinflammatory (18). Excess NaCl in vitro was shown to enhance Th17 differentiation and activity via a p38 MAPK→NFAT5→SGK1 pathway, and that an HSD plus 1% NaCl-supplemented water exacerbates experimental autoimmune encephalitis (EAE) in mice (19, 20). Moreover, macrophages cultured in hyperosmotic NaCl produced more IL-1β, and wild-type (WT), but not inflammasome-deficient, mice developed worse EAE on an HSD plus 1% NaCl-supplemented water (21). In contrast, regulatory T cells cultured in high salt media (HSM) adopted a proinflammatory phenotype (expressing Ifng, Il17a, and Rorc) and were unable to suppress inflammation when transferred into two different disease models exogenously (22). Similarly, M2 macrophages in HSM failed to suppress T cell proliferation, and an HSD plus 1% NaCl water inhibited wound healing in vivo (23). Finally, an HSD with 0.9% NaCl water led to hypernatremia in the skin, which drove antimicrobicidal activity against Leishmania major by macrophages (24). Therefore, an HSD has significant proinflammatory effects in the brain and skin.

Because NaCl is first encountered in the gastrointestinal tract, we hypothesized that an HSD would also potentiate intestinal inflammation caused by either immunologic dysfunction or bacterial infection.

NaCl was calculated using data from the SELF Nutrition database (25). Dry weight was calculated as the mass of a serving size minus the mass of water. Where the water mass was unavailable, the mass of all contents was summed to provide the dry weight. Where sodium mass was zero, we assumed 1 mg as an upper limit for calculations.

Animals were placed on either an HSD or a low salt diet (LSD) after day 0 sample collection. Weight, food weight, water volume, fecal, and urine samples were taken every 12 h for 4 d. Upon necropsy, serum, urine, feces, all alimentary contents, and tissues from the stomach to the colon were collected. Solid samples were weighed, dried in a fume hood for ≥72 h, weighed again, reconstituted in a known volume of 1 M HCl, and disrupted using a bead homogenizer.

All liquid samples were separated from debris and diluted 1:100 in 1.5 mM CsCl buffer for the flame photometer (Model 943 Flame Photometer; Instrumentation Laboratory). Briefly, the machine measures minute quantities of sodium and potassium by incinerating a laminar flow of the sample solution, creating an ionized gas from which spectra are observed. This is translated into a millimolar concentration readout of ions in the original sample. The original concentration of sodium and potassium is then determined by back-calculation: concentration is multiplied by the known volume of acid to determine millimoles, and then divided by the volume of sample water.

Animal diets were obtained from Harlan Teklad in irradiated, sterile packaging. Animals were subjected to either an HSD (TD.92034) of 4% NaCl w/w, or an LSD (TD.96208) of 0.49% NaCl w/w. Animals were permitted standard cage water ad libitum during all experiments except where noted.

Mice of the 129 SvEv Il10^−/− strain and WT controls were obtained in germ-free conditions from the National Gnotobiotic Rodent Resource Center and transferred to a specific pathogen-free (SPF) housing facility. Animals were swabbed with slurry of feces from WT SPF mice on their mouths and anuses to induce colonization, and placed in cages with either their HSD or LSD.

Myeloperoxidase (MPO) and lipocalin 2 (LCN2) were assayed using ELISA kits from R&D Systems (DY3667 and DY1857, respectively). Fecal samples were homogenized in PBS containing 0.1% Tween, and ELISA values were then standardized against protein content as determined by bicinchoninic acid assay.

Bone marrow–derived dendritic cells from C57BL/6 mice were cultured from bone marrow in ultra-low binding polystyrene plates with 10ng/ml IL-4 and G-CSF. Cells were preconditioned at least 16 h prior to use either in normal salt media (NSM; DMEM + 10% FBS + 0 mM NaCl) or HSM (DMEM + 10% FBS + 40 mM NaCl). Cells were pretreated for 30 min in the indicated doses of inhibitor, then subjected to 1 ng/ml LPS (Escherichia coli, Cat# L3024; Sigma-Aldrich). Supernatants were then collected and analyzed by RnD DuoSet ELISAs. SGK1 inhibitor GSK 650394 (Tocris Bio-Techne); p38 MAPK inhibitor SB203580 (InvivoGen).

SPF C57BL/6 mice were transferred to either an HSD or an LSD 24 h prior to dextran sodium sulfate (DSS) exposure. Animals were initially fed 3% w/v DSS for 5 d before harvest. Further experiments used the initial water consumption rates to calculate DSS concentrations that would normalize the dose of DSS per day. Stool samples were taken and tested for fecal occult blood, which was then scored based on the degree of biochemical reaction.

SPF mice were placed on their diets 48 h prior to infection. The next day, mice were deprived of food and water for 4 h prior to gavage with 20 mg streptomycin, after which food and water were restored. Salmonella typhimurium SL1344 was grown overnight to an OD between two and six. Cultures were then diluted in PBS to 10⁵ CFU per ml. After 4 h without food and water, mice were gavaged with 100 μl of this inoculum, followed by reintroduction of water. After 2 h, food was restored. Feces were collected daily and plated on Luria–Bertani agar plates containing streptomycin and ampicillin.

Cecal tissues were cut from the alimentary tract; colons were cut from cecum to (and including) the anus. Tissues were then placed in 10% formalin for at least 24 h prior to embedding, cutting, and H&E staining by the Cell Services and Histology Core. Slides were then given to a blinded scorer for assessment.

For S. typhimurium infection, histology was assessed using a modified scheme adapted from Barthel et al. (26). Slides were blind scored using a four-point scale of severity for each of four criteria: epithelial hyperplasia and damage; immune infiltrate into the mucosa and lamina propria; goblet cell loss; and submucosal edema.

For the Il10^−/− model, blinded histologic scores were assessed using a modified scheme adapted from Kim et al. (27). Slides were blind scored using a four-point scale of severity for each of three criteria: epithelial hyperplasia and damage; immune (particularly T cell infiltrate) into the mucosa and lamina propria; and goblet cell loss.

Tissues for real-time quantitative PCR (qPCR) were snap-frozen in RNAlater (76106; Qiagen) during harvest and placed at −80°C until use. For extraction, tissues were thawed on ice, and homogenized for 10 min in a lysis solution containing 1% β-mercaptoethanol and RNaseZap (AM9780; Thermo Fisher)-treated beads. Extraction proceeded using a PureLink RNA mini kit (12183018A; Thermo Fisher). RNA was quantified by nanodrop and treated with Turbo DNase post elution, with subsequent heat inactivation and reverse transcription. cDNA was run on a Life Quantstudio six real-time PCR machine using iTaq Universal Sybr Green Supermix (Bio-Rad). Data were analyzed using the ΔΔ cycle threshold (ΔΔC_T) method with HPRT as the control gene and either streptomycin pretreated mice for Salmonella experiments, or cotransferred gnotobiotic WT 129 mice as the control sample.

The most obvious way an HSD could affect IBD is by elevating NaCl content in the gut, which could directly alter the immune system. Although sodium is quickly and efficiently absorbed in the stomach and upper small intestine, an HSD could increase gastrointestinal tract salinity locally, or portal blood and lymph to drive systemic effects (19, 20).

To begin, we considered whether the mouse chow HSD is similar in NaCl content to commonly available foods in grocery stores, as well as fast food restaurants. We calculated the content by dry weight for these foods using the SELF Nutrition database (25), which uses the Nutritional Facts label in the United States. In contrast, mouse chow containing 4.0% w/w NaCl is a common HSD, whereas 0.49% NaCl is a common LSD. Comparing these diets revealed that the mouse HSD has comparable salt content to many fast food items (Fig. 1A, Supplemental Table I). Conversely, the LSD was similar to fruits and vegetables. In many murine experimental models using an HSD, the water available to the mice also contains 0.9–1.0% w/v NaCl (19, 20, 22–24, 28, 29). However, the NaCl content of commonly consumed beverages was negligible (Fig. 1B, Supplemental Table I). Therefore, in the studies below, we provided mice with ad libitum tap water with no added NaCl.

Having established that the mouse HSD is comparable to elements of a Western diet, we sought to determine if the HSD alters sodium concentration in the gut. To assess this, animals were fed an HSD or an LSD for 4 d, and sodium content was determined in gastrointestinal tissues and contents using a flame photometer. Mice ate equivalent amounts of food and gained equal weight on both diets. We expected that sodium would be largely absorbed in the upper gastrointestinal (GI) tract, then excreted by the kidney into the urine. Indeed, mice on an HSD drank significantly more water and excreted significantly higher concentrations of sodium in their urine starting 12 h after administration of the diet, with output stabilizing after 2 d (Fig. 1C). Concomitantly, we observed lower potassium concentrations in the urine (Fig. 1D).

Surprisingly, mice on an HSD also had significantly higher concentrations of sodium in their feces (Fig. 1E), again with concomitantly lower potassium (Fig. 1F). However, the idea that high sodium would be present through the GI tract is at odds with the conventional wisdom that sodium should be largely absorbed in the upper GI tract (30).

We therefore measured the sodium concentration of the GI tissue and luminal contents for both sodium and potassium. Tissue content was largely similar between the diets throughout the GI tract (Fig. 1G). As expected, the stomach contents, largely composed of the HSD, contained significantly more sodium. This sodium was normalized within the small intestine, but was then paradoxically elevated in the colonic contents (Fig. 1G). As before, GI content potassium levels were inverted (Fig. 1H). Therefore, the colon contents of mice on an HSD experience a local high sodium concentration compared with those on an LSD.

How can we explain the apparent reappearance of sodium in the colon? The primary function of the colon is to absorb water from the feces by absorbing ions, including sodium. However, mice on an HSD are sodium overloaded. This represses aldosterone production, which in turn represses expression of the sodium absorption channel ENaC. Thus, on an HSD, the colon will absorb potassium rather than sodium (31). Indeed, sodium mass in the cecal and colonic content only dropped in LSD animals, whereas potassium mass dropped in both groups (Fig. 1I). Then, as water is removed, the concentration of sodium rises (Fig. 1I, 1G). This gives the illusion of sodium reappearing in the feces. Interestingly, although potassium is reduced in the HSD feces, the magnitude was not equivalent to the sodium elevation. However, because the normal osmolar gradient that exists from the villus tip to the villus base exceeds the osmoles of both sodium and potassium in the content, it is likely other osmolytes contribute to water absorption (32).

We sought to determine if the local sodium increase we observed in colon contents might alter colonic immune responses. To test this, we examined the effect of our diets in the Il10^−/− model of murine colitis.

Il10^−/− mice spontaneously develop inflammatory colitis driven by resident intestinal bacteria (33). Whereas colitis develops asynchronously under normal SPF housing conditions, colonizing adult germ free Il10^−/− mice with SPF fecal microbiota synchronizes disease. Mice develop progressive colitis beginning after 1 wk, plateauing after 4–6 wk in our facility, and lasting for the life of the animal (34, 35). The inflammatory response is characterized by transmural mononuclear cell infiltrate in the cecum and colon, as well as epithelial hyperplasia. We therefore colonized germ-free Il10^−/− and WT mice on a 129SvEv background and simultaneously transferred mice to either an HSD or an LSD. After 3 wk we examined intestinal inflammation by H&E staining and tissue cytokine expression (Figs. 2, 3).

Histologic examination showed that an HSD broadly exacerbates colitis in both the ceca (Fig. 2A–D) and colons (Fig. 2E–J). Both the HSD ceca and colons had increased immune infiltration, increased epithelial hyperplasia, and extensive goblet cell loss. In the ceca, this was accompanied by edema and immune cell infiltration of the submucosa and lamina propria, which was not observed in LSD mice (Fig. 2C). Qualitatively, the HSD exacerbated blinded histologic inflammatory scores in all segments of the colon. The HSD significantly increased colitis scores in the cecum, trended toward an increase in the proximal colon, which again became significant in the transverse and distal colonic segments (Fig. 2L–N). Germ-free WT animals colonized with the same SPF displayed no appreciable inflammation in either diet group (Fig. 2L, 2M), indicating that an HSD alone is not overtly inflammatory. Consistent with this scoring, Il10^−/− mice had higher fecal MPO and LCN2 levels (Fig. 2N), two luminal biomarkers of inflammation (36). Together, these data suggest a general enhancement of inflammation, rather than a specific effect on epithelial growth or immune cell infiltrate.

Previous reports have noted that an HSD is capable of driving enhanced cytokine production as well as immune cell recruitment and stronger antimicrobial barrier defenses. Therefore, to better understand how an HSD exacerbated inflammation in this model, we examined transcription of several key cytokines via real-time qPCR within the colonic tissues of LSD- and HSD-fed mice. Indeed, an HSD induced significantly enhanced expression of Tnf and Il12b (Fig. 3A, 3B), as well as the Th17-axis cytokines Il23a and Il1b, although Il6 trended but did not reach significance (Fig. 3B–E). Although we observed that an HSD increased Il17a expression in our first replicate, the statistical significance of this difference was not observed after experiments were combined (Fig. 3F). Likewise, the Th1-axis cytokines Il12a and Ifng trended but did not reach significant enhanced expression (Fig. 3G, 3H).

The strong increase in expression of the IL-12 family genes Il23a and Il12b, as well as the myeloid cytokine Il1b, suggested that macrophages and dendritic cells (DCs) might respond to high intestinal sodium. Indeed, macrophages have already been shown to respond to HSM in vitro, suggesting that myeloid cells are a key sodium-responsive population (24). In line with our observations in vivo, we found that LPS-stimulated DCs exposed to HSM produced significantly more pro–IL-1β, IL-12p40, IL-23p19, and to a lesser extent IL-6 (Fig. 3I–L) than cells in NSM.

In vitro work has described a p38→NFAT5→SGK1 pathway for enhanced cytokine secretion in response to stimulation and HSM (19–21, 24). We tested whether this pathway was responsible for the enhanced DC cytokine production using the SGK1 inhibitor GSK 650394 and the p38 inhibitor SB203580. As hypothesized, HSM potentiation of IL-23p19 production was eliminated by either inhibitor (Fig. 3M, 3N), suggesting that DCs may use the same pathway to detect and enhance inflammatory responses in the presence of increased NaCl concentrations.

These data add DCs to the list of salt-responsive cell types, which also includes T cells and macrophages (19, 20, 24). Any combination of these cell types could be driving the high salt response in our models.

The observed increase in expression of Il23a and Il1b suggests an effect on innate immune responses. Therefore, we examined the effect of an HSD in the DSS model of colitis. In this model, animals are fed DSS, which injures intestinal epithelial cells. This drives an inflammatory response to luminal bacteria, the early phase of which is primarily mediated by the innate immune system. Thus, by treating HSD or LSD mice with 3% DSS for 5 d, we sought to examine whether an HSD would exacerbate innate immune-driven intestinal inflammation.

Although HSD animals did exhibit greater weight loss (Fig. 4A), and elevated fecal Hemoccult scores (Fig. 4B), these data are confounded by the fact that mice on an HSD drank nearly twice as much DSS water as LSD animals. HSD mice thus received significantly more DSS mass than LSD animals (Fig. 4C). We then attempted to normalize DSS consumption by reducing the concentration of DSS provided to HSD animals to 1.9% DSS (Fig. 4D–F). This nearly normalized DSS consumption rates (Fig. 4F), but also normalized Hemoccult scores between HSD and LSD mice.

This led us to consider whether DSS itself might be salty. Indeed, we found that depending on the manufacturer, a 3–4% solution provides as many millimoles of sodium per ml as a 1% NaCl solution (Fig. 4G). Therefore, the HSD and LSD mice both experience high sodium intake, but originating from the DSS water, which prevents the study of dietary sodium effects in this model. We therefore sought an alternative model to test intestinal innate transcriptional responses to an HSD.

Another model of innate intestinal immune responses is oral infection with Salmonella enterica serovar typhimurium (Fig. 5A). Because this model rapidly induces extreme inflammation, we speculated that an HSD might generally enhance cytokine responses, but fail to further accelerate histologic inflammation. Indeed, S. typhimurium–infected mice had similar bacterial burdens and histopathology on both diets (Fig. 5B, Supplemental Fig. 1). Thus, this model allows us to examine the effect of an HSD on cytokine expression in isolation from these factors.

In agreement with the Il10^−/− model, HSD mice infected with S. typhimurium also had elevated cytokine responses. HSD mice had significantly increased expression of Tnf and Il12b (Fig. 5C, 5D), as well as the Th17-axis cytokines Il23a and Il1b, the latter of which trended toward significance, though only reaching significance on day 3 (Fig. 5E, 5F). Likewise, Il6 trended toward significance, but in this case was statistically significant on days 3 and 4 (Fig. 5G). Unlike the Il10^−/− model, Il17a was significantly increased during S. typhimurium infection (Fig. 5H). The S. typhimurium model may thus have a greater sensitivity to detect cytokine increases, and similarly, the Th1-axis cytokines Il12a and Ifng, which trended in the Il10^−/− model, were significantly increased by the HSD during S. typhimurium infection (Fig. 5I, 5J). These results are consistent with the conclusions from the Il10^−/− model that an HSD enhances innate cytokine responses.

Because the NLRP3 inflammasome responds to loss of cellular potassium (37), we hypothesized that caspase-1/11 knockout mice would fail to respond to an HSD in the S. typhimurium infection model. However, despite delayed inflammation from loss of these genes, there were still differences in colonic cytokine expression (Supplemental Fig. 2). Interestingly, HSD and LSD mice expressed equivalent levels of Ifng and Il17a, which is likely due to the loss of IL-18 and IL-1β processing, respectively.

It is clear that components of the Western diet have significant, negative impacts on human health. Evidence suggests the Western diet is linked to the incidence of several autoimmune diseases, including multiple sclerosis and diabetes (38). Animal studies have shown that diets high in protein, fat, and sugar, and low in fiber can alter immune cell function (7), and promote metabolic syndrome (39). Further, the Western diet is associated with an increasing incidence of IBD (1, 2). However, the high salt content of the Western diet has not been studied as a component that drives IBD.

Although the effect of dietary NaCl on immune function has been described outside the GI track, the mechanism by which dietary NaCl exerts systemic effects is difficult to explain. Dogma suggests that sodium is quickly stripped from food in the proximal intestine (30), and is then stored in the skin (24, 29) or reconcentrated in the kidney prior to excretion. Therefore, one would expect HSD effects to be limited to the upper GI tract, the urinary tract, and the skin.

Our data surprisingly argue that an HSD actually creates a high local concentration of sodium in the colon. Even though sodium levels from food are rapidly normalized in the small intestine, presumably by a combination of ion absorption and intestinal and pancreatic secretion, we observed an apparent re-enrichment of sodium in the feces. Nevertheless, this observation can be explained by comparison with the reciprocal absorption of potassium.

Like sodium, potassium levels are normalized in the small intestine. As the contents pass through the cecum and colon, both groups of mice absorb potassium, decreasing its mass. However, although LSD mice also absorb sodium, HSD mice do not, leaving it behind. This leads to an increase in concentration of sodium in the colons of HSD mice relative to the LSD mice as water is removed. Therefore, we speculate that humans consuming a Western diet may also experience high sodium concentrations in their colons.

How does luminal salt concentration cause an immunologic response in the seeming absence of unchanged tissue salt concentrations? We can imagine numerous possible mechanisms. For example, the high sodium concentration in the colonic lumen caused by the HSD likely creates a sodium microgradient across the epithelium that is experienced primarily by the epithelial cells. These epithelial cells could experience this as a static, elevated concentration gradient, or as increased sodium flux across their cytosol and membranes. Both the concentration and flux changes could alter epithelial signaling cascades to promote inflammation. However, because the blood rapidly washes this sodium away, it is undetected in our examination of total tissue sodium content.

Additionally, immune cells in the lamina propria may also be affected by altered microgradients or flux of sodium from the colon. In particular, DCs that extend pseudopodia into the colonic lumen to sample its contents might also directly experience the new high sodium environment. These effects would be exacerbated during inflammation when the epithelial barrier becomes more permeable to luminal contents or is effaced.

How exactly does an HSD exacerbate colitis, or the other immune phenotypes previously observed? The first reports demonstrated a role for sodium on the T cell compartment, but subsequent studies have also implicated macrophages. Our data now show that DCs respond to HSM by expressing additional pro–IL-1β, IL-6, IL-12, and IL-23, which would be expected to drive effector T cell responses in IBD. Indeed, colitis was exacerbated 3 wk after colonization in the Th1/Th17-driven Il10^−/− model, although we did not observe consistent elevation in Il17a. This was in contrast to the effect of an HSD upon EAE, which is predominantly Th17 driven (19–21, 28). Nevertheless, we examined a time point where clinical severity has not yet reached a plateau, so it may be that an HSD will also potentiate Th17 responses later. Enhanced expression of the myeloid cytokines Il1b as well as Il23a also suggests additional effects on innate immune cells. The S. typhimurium model, by comparison, displayed enhanced Il17a expression, perhaps originating from ILC3s, along with enhancement of the Il1b and Il23a myeloid cytokines. Thus, the data we have presented highlights a novel role for an HSD in the DC compartment.

Placing our data in context with published reports leads one to believe that an HSD initiates numerous proinflammatory effects on the immune system, which may create a series of positive feedback loops. Although the high sodium concentration in the colons may have direct effects on T cells via the SGK1 mechanism, they may also be driven by the enhanced cytokine production and costimulatory molecule expression of myeloid cells. Simultaneous HSD-mediated impairment of regulatory T cell function could fuel a proinflammatory environment in the colon. Future studies could also look at indirect effects of an HSD, which have not currently been ruled out. An HSD could have important effects on the composition of the colonic microbiota, or alter the function of certain microbes, as has already been described with the cag gene island of Helicobacter pylori (40). Finally, an HSD strongly affects the renin-angiotensin-aldosterone system, inhibition of which has been shown to ameliorate murine EAE (41).

In conclusion, an HSD generates a high sodium concentration in the colon, which exacerbates a murine model of chronic, immune-mediated colitis. Our data may inspire clinical studies in humans to determine if NaCl has detrimental effects on IBD.

We thank M. Goy and R. Fellner for assistance with the flame photometer and discussions, C. Suitt for sectioning and staining, and B. Grubb for early discussions.

The authors have no financial conflicts of interest.