Elemental iron modifies the redox environment of the gastrointestinal tract: a novel therapeutic target and test for metabolic syndrome

Abstract

Metabolic syndrome (MetS, i.e., type 2 diabetes and obesity) is often associated with dysbiosis, inflammation, and leaky gut syndrome, which increase the content of oxygen and reactive oxygen species (ROS) in the gastrointestinal (GI) tract. Using near-infrared fluorescent, in situ imaging of ROS, we evaluated the effects of oral administration of elemental iron powder (Fe⁰) on luminal ROS in the GI tract and related these changes to glucose metabolism and the gut microbiome. C57Bl/6J mice fed low-fat or high-fat diets and gavaged with Fe⁰ (2.5 g per kg), in both single- and repeat-doses, demonstrated decreased levels of luminal ROS. Fourteen days of repeated Fe⁰ administration reduced hyperglycemia and improved glucose tolerance in the obese and hyperglycemic animals compared to the untreated obese controls and reduced the relative amount of iron oxides in the feces, which indicated an increased redox environment of the GI tract. We determined that Fe⁰ administration can also be used as a diagnostic assay to assess the GI microenvironment. Improved metabolic outcomes and decreased gastrointestinal ROS in Fe⁰-treated, high-fat diet-fed animals correlated with the increase in a co-abundance group of beneficial bacteria, including Lactobacillus, and the suppression of detrimental populations, including Oscillibacter, Peptococcus, and Intestinimonas. Daily Fe⁰ treatment also increased the relative abundance of amplicon sequence variants that lacked functional enzymatic antioxidant systems, which is consistent with the ability of Fe⁰ to scavenge ROS and oxygen in the GI, thus favoring the growth of oxygen-sensitive bacteria. These findings delineate a functional role for antioxidants in modification of the GI microenvironment and subsequent reversal of metabolic dysfunction.

Graphical Abstract

Introduction

Metabolic syndrome (MetS) is a clustering of obesity-related risk factors including impaired glucose homeostasis, dyslipidemia, and hypertension, and is associated with increased risk for developing type 2 diabetes and cardiovascular disease. MetS affects an estimated 1.4 billion people globally, with diagnosis rates rising annually [1]. As a result, various therapeutic strategies for combating MetS have been explored in recent years, including both pharmaceutical and dietary interventions.

MetS has been shown to be associated with low-grade inflammation of the intestinal lining and markers of poor intestinal health including gut dysbiosis, or an imbalance of gut microorganisms, and intestinal barrier dysfunction, colloquially referred to as “leaky gut” [2–7]. First, gastrointestinal (GI) inflammation triggers the overproduction of ROS via NADPH oxidase enzymes in epithelial cells and lymphocytes of the intestinal barrier [8]. This is followed by intestinal permeability increasing the transport of water and dissolved oxygen from the mesenteric vasculature into the GI lumen, likely contributing to the concomitant increase in ROS observed in obese and metabolically compromised mice [9]. Studies have shown that diets rich in antioxidant and ROS-scavenging compounds, such as plant polyphenols, may prevent or reverse MetS [10], reduce dysbiosis and improve intestinal barrier function [11–13]. These findings, along with the tendency for polyphenolic antioxidants to have poor bioavailability [14,15], have led to investigation of mechanisms by which polyphenols and other dietary antioxidants may modify the microenvironment of the GI tract. Poorly bioavailable dietary antioxidants, including polyphenols, have been shown to effectively scavenge ROS in the gastrointestinal GI lumen [9].

The precise mechanism by which dietary antioxidants influence GI ROS and metabolic outcomes remains unclear. Obese and insulin-resistant C57Bl/6J mice demonstrate a marked increase in ROS in the GI lumen compared to metabolically healthy animals [9], which suggests the possibility that metabolic status is linked to microbial GI ecology through changes in the GI redox environment. A direct link between dietary antioxidants and the gut microbiome has been demonstrated by the administration of antioxidant grape and cranberry polyphenols to metabolically compromised, obese mice, resulting in a dramatic increase in the relative abundance of Akkermansia muciniphila, an anaerobic bacterial species which has been directly associated with metabolic health [11,13,16]. Interestingly, these beneficial metabolic effects are not exclusive to polyphenols, but have also been observed as a result of treatment with dietary nitrate, which can also be found in the human diet [17–20]. Nitrates are suspected to modulate the gut microbiome due to their role in the nitrate-nitrite-nitric oxide pathway, modifies the local redox environment of the gastrointestinal tract [21]. These similarities support the idea that the modification of the gut microbiome and metabolic benefits observed upon administration of dietary compounds such as polyphenols or nitrates are not necessarily a result of the specific compound administered, but rather, their ability to interact with the redox environment of the gut.

Gut oxygen levels and the GI redox environment have also been implicated in the modification of the gut microbiome in circumstances beyond dietary interventions. In an investigation of the impact of hyperbaric oxygen therapy on the microbiome, it was shown that increasing luminal oxygen levels resulted in a loss of oxygen intolerant microorganisms [22]. This suggests a significant role for luminal oxygen levels in regulating the composition of the gut microbiome. However, current methodologies used to measure GI oxygen levels are invasive and not always reliable. They include cannulation of the intestine with a Clark electrode [23–25] and electron paramagnetic resonance [26] or phosphorescence quenching [9,22]. Thus, a simple and non-invasive, but accurate procedure for measurement presents a need that should be addressed.

Despite a majority of studies on dietary antioxidants focusing on phytochemicals, poor bioavailability and the ability to interact with ROS is not exclusive to these compounds. Elemental iron powder (Fe⁰) is widely used in food packaging as an efficient, non-bioavailable, and non-toxic oxygen and ROS scavenger [27,28]. In the presence of moisture, Fe⁰ reacts with oxygen forming iron oxides containing Fe⁺² and Fe⁺³ (i.e. Fe₂O₃, Fe₃O₄). The safety of Fe⁰ for consumption has been characterized in both case studies tracking accidental ingestion of hand-warmers [29,30], which utilize the exothermal reaction between Fe⁰ and oxygen to generate heat [31], and in toxicity studies comparing the bioavailability of iron in different chemical forms, where elemental iron has been shown to have extremely limited absorption in the intestines of rodents even at doses which far exceed typical levels of intake [32,33].

The therapeutic potential of reducing the accumulation of oxygen and ROS in the GI tract has not been explored, although these conditions can be associated with the development of MetS. Furthermore, there is a clinical need for a non-invasive method for measuring GI oxygen and ROS levels. This study aims to determine whether orally administered Fe⁰ can mitigate ROS in the GI tract of obese and hyperglycemic mice, thus benefiting their metabolic health and gut microbiome. Luminal ROS was measured in situ using near-infrared fluorescence (NIRF) and metabolic health was assessed by measuring body weight and oral glucose tolerance. We hypothesized that the oxygen and ROS-scavenging abilities of Fe⁰ would decrease luminal ROS and oxygen levels, thereby modifying the gut microbiome and improving the metabolic health of obese mice. In the process of testing this, we also demonstrated the use of Fe⁰ oxidation as a diagnostic tool for measuring gut oxygen status via the measurement of fecal iron excretion and calculation of the ratio of oxidized iron (Fe^2+/3+) to unoxidized iron, or Fe⁰ in the original state it was administered. Our data suggest a link between gut oxygen status and metabolic health and provide a rationale for using Fe⁰ as a diagnostic and therapeutic tool for mitigating MetS.

Results

Fe decreases GI ROS in obese and lean mice.

ROS-associated NIRF imaging of nine healthy male C57Bl/6J mice fed a low-fat diet (LFD) showed that Fe⁰ was effective in reducing GI ROS at 2.5 and 10 g Fe⁰ per kg body weight (BW) (Fig. 1A–B) 4 h after administration. A trend in GI ROS reduction was apparent only 1 h after administering 2.5 g Fe⁰ per kg BW (p = 0.3014; Fig. 1C–D), suggesting a rapid effect of Fe⁰ on the GI ROS in lean mice. A second study with 30 C57Bl/6J mice fed a high-fat diet (HFD) showed that 2.5 g Fe⁰ per kg BW was effective in reducing GI ROS, though a lower dose (0.5 g Fe⁰ per kg BW) was not effective (Fig. 1E–F). The safety of the Fe⁰ treatment was assessed based on behavioral observation and visual assessment; mice did not demonstrate any signs of pain or distress upon observation for 4 h after treatment, nor were any macroscopic lesions observed upon dissection. Based on these data, we chose a treatment dose of 2.5 g Fe⁰ per kg BW for all subsequent experiments.

Figure 1.

Visualization of the effects of Fe⁰ on ROS in the GI of mice. (A) ROS-associated NIRF in LFD-fed mice treated with 10 or 2.5 g Fe per kg BW, or a volumetrically equivalent dose of water overlaid over the brightfield images. Data were analyzed using one-way ANOVA with Dunnett’s multiple comparisons test relative to the control group and are reported as mean ± SD; N = 3 mice per group. (B) Representative overlay NIRF and corresponding brightfield images of LFD-fed lean mice. (C) ROS-associated NIRF in LFD-fed obese mice 1 h after treatment with 2.5 g Fe per kg BW or a volumetrically equivalent dose of water. Data were analyzed using a Student’s T-test and are reported as mean ± SD; N = 3 mice per group. (D) Representative overlay of NIRF and corresponding brightfield images of LFD-fed mice 1 h after treatment administration. (E) ROS-associated NIRF in HFD-fed obese mice treated with 2.5 or 0.5 g Fe⁰ per kg body weight or a volumetrically equivalent dose of water. Data were analyzed using one-way ANOVA with Tukey’s multiple comparisons test. Values not sharing common letters are significantly different from one another (P ≤ 0.05). (F) Representative overlay NIRF and corresponding brightfield images of HFD-fed mice. The NIRF intensity scale, shown at left of images, was normalized using Bruker Molecular Imaging Software.

A modified ferrozine assay uses ferromagnetic properties of Fe to determine the ratio between elemental and oxidized forms.

A modified version of the ferrozine assay [34–36] (Fig. 2A) was developed to measure the relative oxidative index (ROI) of Fe in fecal samples collected from Fe⁰-treated mice. The ferrozine reagent measures iron in the sample by converting it into ferrous iron (Fe²⁺), which forms a complex that strongly absorbs at 562 nm. As Fe⁰ is magnetic while Fe⁺² and Fe⁺³ are not, measuring the total Fe in a sample before and after removing Fe⁰ with a neodymium magnet can quantify elemental and oxidized Fe in a sample. These values are then used to calculate the ROI by dividing the absorbance of the magnet-treated (MT) sample by that of the untreated (UT) sample. An ROI value of 1 suggests that all Fe⁰ has been oxidized and thus remained in the sample after magnet treatment, whereas ROI values closer to zero indicate a greater proportion of the elemental, non-oxidized Fe. This assay was validated in a model system of Fe⁰ with increasing concentrations of H₂O₂ to force oxidation of the Fe⁰ (Fig. 2B). Increasing the addition of H₂O₂ to the assay resulted increased the relative amounts of Fe^2+/3+ in a sample (i.e., greater levels of total iron after magnet treatment); thus, a greater ROI for the sample. This assay was used to assess the ROI of mice after one treatment with Fe⁰ and after 14 days of treatment to compare changes in ROI across the experiment.

Figure 2.

Modified ferrozine assay used to measure relative oxidative state of iron. Ferrozine binds ferrous iron (Fe²⁺), but not ferric iron (Fe³⁺) or Fe⁰, forming a complex that absorbs strongly at 562 nm. (A) A schematic illustration of the assay as applied to fecal samples collected from C57Bl/6J mice treated with Fe. (B) Assay validation using H₂O₂ to oxidize Fe⁰. Determining A₅₆₂ of untreated and magnet-treated samples enables the calculation of relative oxidation index (ROI), a unitless measurement of iron oxidation.

Fe⁰ reverses obesity-associated increases in GI ROS, lowers blood glucose, and improves glucose tolerance.

The direct oxygen-scavenging activity of Fe⁰ in the GI tract of obese and lean mice was measured using in situ NIRF imaging 1 h after the administration of a single dose Fe⁰ (Fig. 3A–B). In agreement with a previous study, we observed greater amounts of ROS in the GI of the obese HFD-fed, obese mice compared to LFD-fed, lean mice [9]. Oral administration of Fe⁰ to obese mice decreased luminal ROS-associated GI florescence by 52.0% and lowered it to the levels observed in lean mice. Fe⁰ administration to lean animals also decreased ROS-associated GI florescence by 38.6%.

Figure 3.

Fe⁰ reduces GI ROS in lean, LFD (LF) and obese, HFD-fed mice (HF) after a single dose, and restores metabolic competence and GI redox environment of obese mice after 14-d treatment. (A) ROS-associated NIRF in LFD- and HFD-fed mice treated with a single dose of 2.5 g Fe⁰ per kg BW or a volumetrically equivalent dose of water. Data were analyzed using two-way ANOVA with Tukey’s multiple comparisons test and are reported as mean ± SD; N = 8 mice per group. Values not sharing common letters are significantly different from one another (P ≤ 0.05). (B) Representative overlay NIRF and corresponding brightfield images of mice from each treatment group. The NIRF intensity scale, shown at left, was normalized using Bruker Molecular Imaging Software. (C) ROS-associated NIRF in LFD- and HFD-fed mice treated with Fe⁰ or a volumetrically equivalent dose of water for 14 d (measurements performed on d 15). Data were analyzed using a two-way ANOVA with Tukey’s multiple comparisons test and are reported as mean ± SD; N = 8 mice per group. Values not sharing common letters are significantly different from one another (P ≤ 0.05). (D) Representative overlay NIRF and corresponding brightfield images of mice from each treatment group of a 14 d experiment. The NIRF intensity scale, shown at left, was normalized using Bruker Molecular Imaging Software. (E) Blood glucose concentrations (mg/dL) expressed as mean ± SD; N = 8 mice per group were measured at the indicated timepoints (0–240 min) after administration of 2 g glucose per kg BW to mice before and after 14 d treatment with Fe⁰ or a volumetrically equivalent dose of water. At each timepoint, a two-way ANOVA with Tukey’s multiple comparisons test was used to evaluate differences between each treatment group. Values not sharing common letters are significantly different from one another (P ≤ 0.05). (F) Oral glucose tolerance, calculated as AUC over the course of 4 h, before and after 14 d treatment with Fe⁰ or a volumetrically equivalent dose of water. Data were analyzed using three-way ANOVA with Šídák’s multiple comparisons test,and are reported as mean ± SD; N = 8 mice per group. Values not sharing common letters are significantly different from one another (P ≤ 0.05). (G) Fasting blood glucose of LFD- and HFD-fed mice before and after 14 d treatment with Fe⁰ or a volumetrically equivalent dose of water. Data were analyzed using three-way ANOVA with Šídák’s multiple comparisons test and are reported as mean ± SD; N = 8 mice per group. Values not sharing common letters are significantly different from one another (P ≤ 0.05). (H) Fe⁰ treatment decreases relative oxidation index after 1 and 14 d of treatment with Fe⁰. Day 1 data demonstrates an assessment of baseline Relative Oxidative Index (ROI), wherein fecal iron from mice on the HFD had a greater ROI value than mice fed a LFD, a result of greater levels of oxygen in the gastrointestinal lumen of mice on the HFD. After 14 d of treatment with Fe0, the elevated ROI of mice fed a HFD is ameliorated to levels not significantly different from mice fed a LFD. Data were analyzed using two-way ANOVA with Holm-Šídák’s multiple comparisons test, and are reported as mean ± SEM from three independent experiments; N = 8 fecal samples per group.

Over the course of the 14-day repeat dose study of Fe⁰ effects in lean and obese mice,Fe⁰ treatment did not result in weight loss in mice fed LFD or HFD. Mice in both the LFD and LFD-Fe groups consumed greater amounts of food on the first and last day of the study (Fig. S1), which may be related to fasting one day prior to measurements in order to accurately measure glucose tolerance. NIRF ROS imaging was performed 24 h after the last Fe⁰treatment (on day 15). While the Fe⁰ treatment did not have a significant effect on the ROS-associated GI florescence in LFD-fed mice, luminal ROS-associated florescence decreased by 57.5% in the HFD-fed group treated with Fe⁰ compared to HFD-fed control group. Furthermore, at the end of the experiment, ROS-associated fluorescence in the HFD-fed mice treated with Fe⁰ were not significantly different from Fe⁰-treated or control LFD-fed groups (Fig. 3C–D). This demonstrates that repeat doses of Fe⁰ effectively ameliorate luminal ROS in obese, insulin resistant mice. Beneficial effects of Fe⁰ on the carbohydrate metabolism in the HFD-fed animals was evident from the improvements in the fasting blood glucose (FBG) levels and oral glucose tolerance (OGT) (Fig. 3E–G), which are traditional markers of MetS. Treatment with Fe⁰ for 14 days resulted in a decrease of FBG by 18.3% for the HFD-fed mice compared to their baseline values. Similarly, OGT of the HFD-fed mice, as determined by the area under the curve (AUC), decreased by 31.6% following 14 days of Fe⁰ administration. Treatment with Fe⁰ did not significantly affect the glucose tolerance of LFD-fed groups.

Non-invasive assessment of GI oxygen content in Fe-treated HFD- and LFD-fed mice.

We hypothesized that the Fe oxidation-based ROI analysis of feces can be used to monitor increased oxygen content associated with dysbiosis and leaky gut syndrome in obese mice. The feces collected from repeat-dose LFD and HFD-fed obese mice 6 h after the initial and final Fe⁰ doses were tested for relative amounts of total and oxidized Fe as depicted in Fig. 2. ROI analysis of the fecal samples collected after the first dose showed that the feces of obese mice contained a greater relative proportion of oxidized iron (higher ROI) than the feces of lean mice (Fig. 3H). After 14 days of Fe⁰ treatment, the ROI of mice on the HFD decreased in comparison to day 1 and was not significantly different from the measured ROI of mice on the LFD. It is notable that the ROI of the LFD-Fe group increased over the course of 14 days. This may be due to increased stress on the mice from daily handling and oral administration of Fe⁰ [37,38].

Fourteen-day Fe⁰ treatment modulates the gut microbiota consistent with improved metabolic status.

A total of 186 mouse fecal samples were collected and the DNA was extracted and sequenced, resulting in ~9.4 million usable 16S rRNA gene V4 reads (average reads/sample = 50,656 ± 8,677). After denoising and abundance-based filtering, 1,367 reliable Amplicon sequence variants (ASVs) were retained for further analysis. On day 0, the HFD-fed mice had significantly higher gut microbial diversity than the LFD-fed mice (Fig. 4A and S2A). Fe⁰ administration significantly reduced the overall microbial diversity in both LFD and HFD groups, as evidenced by the lower Shannon index in the LFD-Fe and HFD-Fe mice at day 7 and day 15 as compared to their diet-matched controls (Fig. 4A). The minimal differences in the numbers of observed ASVs and Faith’s phylogenetic diversity between mice with and without Fe⁰ treatment (Fig. S2) suggest that Fe⁰ altered the gut microbiota composition primarily by changing the abundance of bacteria that were prevalent and phylogenetically related to each other.

Figure 4.

Fe⁰ treatment changes gut microbiota composition. (A) Alpha diversity plot based on the Shannon index. Boxes show the medians and the interquartile ranges (IQRs), the whiskers denote the lowest and highest values that were within 1.5 times the IQR from the first and third quartiles. Data from the same group at different time points were compared using the Wilcoxon matched-pairs signed-ranks test (two-tailed) and that from different groups at the same time point were compared using the Mann-Whitney test (two-tailed). ^###P < 0.001 vs day 0 in the same group. *P < 0.05 and ***P < 0.001 between groups at the same time point. N = 23–24/group at day 0, N = 15–16/group at day 7 and N = 7–8/group at day 15. (B) Principal coordinates analysis based on the Weighted UniFrac distance. Each data point represents the mean principal coordinate (PC) score and the error bar represents the SEM. (C) Co-abundance groups (CAGs). The heatmap shows the log2 transformed relative abundance of each CAG. At each time point, CAGs were compared among the groups using the Kruskal-Wallis test and post hoc Dunn’s test. Values not sharing common letters are significantly different from one another (P < 0.05).

We used weighted UniFrac distance, which considered membership, phylogenetic distance, and abundance, as the representative beta diversity distance metric to assess the global effects of Fe⁰ on gut microbiota structure. The Principal Coordinate Analysis based on the weighted UniFrac distance (Fig. 4B) shows that PC1 and PC2 together accounted for 49.78% of the variations in the gut microbial community. Over the course of the study, gut microbiota of the LFD- and HFD-fed mice changed following a similar trajectory and remained significantly different at all time points (PERMANOVA test; P < 0.05). Mice in the two sub-groups receiving the same diet had similar gut microbiota at day 0 (LFD vs LFD-Fe, p = 0.25; HFD vs HFD-Fe, P = 0.143). Fe⁰ treatment shifted the gut microbiota in patterns that were distinct from those in diet-matched controls. Specifically, 7 days of Fe⁰ treatment induced significant changes in the gut microbiota structure (LFD vs LFD-Fe, P = 0.01; HFD vs HFD-Fe, P = 0.01), whereas the Fe⁰ effect was observed primarily along PC2, with the LFD-Fe and HFD-Fe microbiota shifted in opposite directions (Fig. 4B). Prolonging the Fe⁰ treatment induced further changes in the LFD-Fe and HFD-Fe gut microbiota (P = 0.007 and P = 0.003, respectively) but at day 15 the microbiota composition, at the global scale, was not different from their diet-matched controls (LFD vs LFD-Fe, P = 0.213; HFD vs HFD-Fe, P = 0.433).

Bacteria in the gut form complex interaction networks with interdependent functional groups (guilds) [39,40], which denote groupings of members that exploit the same class of resources in a similar way. To identify potential guilds, we explored the co-abundance relationships among the 192 prevalent and dominant ASVs. Shared by >20% of samples, these ASVs were clustered into 23 co-abundance groups (CAGs; Supplementary Table 1), and together they accounted for ~93% of the total abundance of gut bacteria. Prior to vehicle control or Fe⁰ treatment, some CAGs differed in LFD- and HFD-fed mice but there was no difference between the sub-groups that received the same diet (Fig 4C). Fe⁰ treatment altered CAG abundance and its effect appeared to be largely diet-specific. At day 7, a total of 10 CAGs were modulated by Fe⁰ in either the LFD- or HFD-fed mice, and only two CAGs were similarly changed by Fe⁰ in both groups: CAG19 (all ASVs from Lachnospiraceae) and CAG20 (including ASVs from Peptococcus, Oscillibacter, Butyricicoccus, Intestinimonas, Peptococcaceae, Ruminococcaceae, Lachnospiraceae, and Clostridiales) were significantly down-regulated by Fe⁰ in both LFD and HFD-fed mice as compared to their diet-matched controls. Fe⁰ treatment significantly reduced the abundance of seven CAGs (CAG2, 6–7, 12, and 19–21) in the LFD-fed group. Fe⁰ possibly reversed some, but not all, microbial changes induced by HFD. For example, the abundance of CAG1 (including ASVs from Roseburia, Blautia, and Lachnospiraceae) and CAG13 (including ASVs from Anaeroplasma, Lachnospiraceae, Ruminococcaceae, and Muribaculaceae) was significantly higher in the HFD-fed mice (vs LFD) at day 7. However, their abundance in HFD-Fe mice was not different from the LFD and LFD-Fe groups. At day 15, the overall gut microbiota composition in the Fe-treated mice was similar to that in the diet-matched controls but some differences at the CAG level were observed. Specifically, the effect of HFD on CAG1 and CAG19 was completely abolished by the Fe⁰ treatment, and CAG20 was partially abolished (Fig. 4C). The effect of Fe⁰ on the LFD gut microbiota at this later time point appeared to be minimal.

Finally, we explored the relevance of gut microbiota in host phenotypes by examining the association between CAGs and markers of glucose homeostasis using the Random Forest regression model. Using data at day 0, CAGs were selected as predictors for the best model for FBG (Fig. 5A and S3A) and OGT (Fig. 5B and S3B), respectively. These models were then validated using CAG abundances and glucose homeostasis markers at day 15. This time point showed significant correlations between the measured and predicted values for both FBG (r = 0.646; P = 0.00203; Fig. 5C) and OGT (r = 0.543; P = 0.00285; Fig. 5D). Altogether, we identified 15 distinct CAGs (nine overlapped in the two prediction models) that were associated with glucose homeostasis; among them six (CAG1–3, CAG6, CAG12, and CAG20) were shown to be modulated by the Fe⁰ treatment.

Figure 5.

The associations between gut microbiota and host phenotypes. (A) and (B) Random Forest (RF) model regressing fasting blood glucose and oral glucose tolerance (at day 0), respectively, on the abundance of CAGs (at day 0). The RF assigns a mean error rate or featureimportance score to each feature; this value indicates the extent to which each CAG contributes to the accuracy of the model. The subpanel shows the number of variables and mean squared error of corresponding model. (C) Scatter plot of the measured fasting blood glucose (at day 15) and the predicted values based on CAG abundance (at day 15) with the RF model constructed in (A). (D) Scatter plot of the measured oral glucose tolerance (at day 15) and predicted values based on CAG abundance (at day 15) with the RF model constructed in (B).

Fe-induced changes in gut microbiota composition correlate with GI redox environment.

The association between overall gut microbiota composition and intraluminal ROS (PERMANOVA; P < 0.05) is consistent with our hypothesis that Fe, by increasing the redox environment of the gut, creates a selective pressure that favors bacteria lacking ROS defense systems and thus is more sensitive to GI oxygen and concomitant ROS. We used PICRUSt2 [41] to stratify the 1,367 reliable ASVs based on the predicted presence or absence of a functional enzymatic antioxidant system (EAS), a key mechanism by which microbes are protected against ROS [42]. ASVs were considered to possess a functional EAS if they were predicted to harbor genes that encoded at least two antioxidant enzymes (e.g., superoxide dismutases, peroxiredoxins, catalases, and superoxide reductases) — one for superoxide anion radical detoxification and one for hydrogen peroxide decomposition [42]. We estimated that approximately 5% of all ASVs, which accounted for 12% of the total abundance, lacked a functional EAS (Fig. 6A). Compared with diet-matched controls, Fe⁰ treatment significantly increased the abundance of these ASVs in both LFD and HFD-fed animals, an effect associated with a concomitant reduction in GI ROS (Fig.6B). At the guild level, based on Masslin2 analyses adjusted by groups, the abundances of CAG10, CAG15, and CAG18 were positively correlated with ROS and CAG3 was negatively correlated (Masslin2; adjusted P < 0.25), suggesting that ROS scavenging may be a mechanism by which Fe⁰ modifies the gut microbiome.

Figure 6.

Fe⁰ treatment increased gut bacteria that lacked a functional enzymatic antioxidant system (EAS). (A) Stacked bar plots showing the average relative abundance of ASVs with or without a functional EAS. (B) Boxplot showing the relative abundance of the ASVs without functional EAS. Boxes show the medians and the interquartile ranges (IQRs), the whiskers denote the lowest and highest values that were within 1.5 times the IQR from the first and third quartiles. Data from the same group at different time points were compared using the Wilcoxon matched-pairs signed-ranks test (two-tailed) and that from different groups at the same time point were compared using the Mann-Whitney test (two-tailed). *P < 0.05 and **P < 0.01 between groups at the same time point. N = 23–24/group at day 0, N = 15–16/group at day 7 and N = 7–8/group at day 15.

Discussion

Previous studies have shown that grape polyphenols and other poorly bioavailable dietary antioxidants reduce intestinal ROS [9], improve carbohydrate metabolism, and promote bloom of A. muciniphila [12]. The ability for dietary compounds to interact with the redox environment of the gut in a way that modifies the microbiome and metabolic status of the host has also been demonstrated with dietary nitrate [21].In this study, we applied a non-bioavailable antioxidant to further understand the influence of ROS and oxygen-scavenging on amelioration of metabolic disorders. The main objective of this study was to determine whether metallic iron powder, an efficient, non-bioavailable and non-toxic oxygen/ROS scavenger, can reduce ROS in the GI of obese and hyperglycemic mice, improve their carbohydrate metabolism, and alter their gut microbiome. We also developed a strategy for assessing the oxidative index of the GI tract by measuring the ratio of oxidized and metallic iron in the feces of mice gavaged with Fe⁰. This ratio can be used as an indicator of increased in GI oxygen, which results from the MetS-associated inflammation and intestinal permeability.

ROS may be an important trigger in the development of insulin resistance, although intestinal ROS has never been measured directly in this context [43]. The increased levels of luminal ROS in obese mice fed HFD, visualized by near infrared florescence imaging, is consistent with earlier observations [9] and can be, at least partially, explained by intestinal permeability and low grade gut inflammation observed in obese and hyperglycemic animals and people [44]. Increased permeability of the intestinal barrier allows increased influx of water and oxygen, while low-grade inflammation of the GI lining is associated with increased ROS production by the epithelial cells and lymphocytes. This creates a microenvironment of the GI tract, which influences the profile of microorganisms present. Numerous studies have linked the gut microbiome with metabolic health, and modification of the gut microbiome by antioxidants has been associated with the reversal of metabolic dysfunction, though a precise mechanism by which this occurs is unclear. We hypothesized that one mechanism of this phenomenon may be the scavenging of ROS and oxygen by antioxidants, which modifies the gut microenvironment and therefore, the microbiome.

To test this hypothesis, we treated lean and obese C57Bl/6J mice with Fe⁰ and measured changes in GI ROS, glucose homeostasis, and the composition of the gut microbiome. The ability of Fe⁰ to scavenge GI ROS was measured using NIRF after treating mice with hICG, a fluorescent, non-toxic and membrane impermeable dye has been shown to sensitively and selectively react with ROS in biological systems [45–51], including the GI tract [9]. After determination that 2.5 g Fe⁰ per kg BW was safe and effective dose level to reduce GI ROS of obese mice (Figure 1), the efficacy of this treatment in the remediation of metabolic dysfunction was tested. We observed that 14 days of repeated oral administration of Fe⁰ effectively scavenged GI ROS, reduced fasting blood glucose levels and improved oral glucose tolerance in metabolically compromised mice to the levels observed in healthy, lean mice, and that the metabolic status of the lean mice was unchanged by the treatment (Figs. 1 and 3).

Our measurement of the ROI of the GI using Fe oxidation status showed that feces of obese, hyperglycemic mice gavaged with Fe⁰ had a greater ratio of oxidized to total iron (referred to as ROI) compared to healthy mice, and that Fe⁰ was able to lower the ROI to the levels observed in healthy mice (Fig. 3H). This observation is consistent with the hypothesis that the GI tract of metabolically compromised mice presents a stronger oxidative environment, including greater ROS and oxygen levels compared to lean mice. Further, our data demonstrates the ability of Fe⁰ reverse this phenomenon, lowering the ROI of obese mice over the course of 14 days of treatment, which corresponded to lowered fasting blood glucose and improved glucose tolerance (Fig. 3E–G).

To assess ROI, we developed a simple and reliable method of measuring the ratio of oxidized to unoxidized iron in the feces. This method is based on the ferrozine assay, a colorimetric assay where ferrozine reacts with Fe²⁺ to form a complex that absorbs light at 562nm. In a typical ferrozine assay, two measurements are made- one to quantify Fe²⁺ and the other to quantify total iron by reducing Fe³⁺ in samples to Fe²⁺. The difference in these measurements is based on the buffer used; the total iron buffer (TFeB) contains ferrozine, HEPES and hydroxylamine hydrochloride. In the present study, fecal samples from Fe⁰-treated mice were suspended in hydrochloric acid and divided into two aliquots. Total iron was measured in one aliquot immediately (untreated; UT), and the other was treated with a magnet for removal of unoxidized, ferromagnetic Fe⁰ prior to measurement of total iron (magnet-treated; MT). The workflow of this assay is outlined in Figure 2. The design of this assay allows the calculation of a ratio of total iron in each sample based on absorbance and eliminates inter-sample inconsistencies such as background interference, the total mass of feces excreted from individual mice over the 6 h collection period and the total iron content of the collected fecal samples. This assay may have wide applications in the studies of the GI oxygen content. However, further development of the quantitative relationship between iron oxidation and the gut redox environment is needed before widespread application of this method including calibration curves and sensitivity and methods for eliminating interference from other dietary sources of iron, which presents a limitation for populations with diets less tightly controlled than those of the mice in this study. With further development, this assay has potential for use as a diagnostic clinical assay for leaky gut syndrome and gut inflammation.

Characterization of the gut microbiome demonstrated changes over the course of each week of the two-week study. The observed Fe⁰-induced improvements in glucose homeostasis correlated with changes in six distinct guilds of gut bacteria. Among the changes observed were an increase in the abundance of CAG3 and a decrease in abundance of CAG1, CAG2, CAG6, CAG12, and CAG20. CAG3 contained five ASVs, two from Clostridiales vadinBB60 group, one from Prevotella, and two from Lactobacillus. Members of Prevotella have been reported to correlate with insulin sensitivity [52], and several member of Lactobacillus have been reported to reduce obesity and improve glucose homeostasis in HFD-fed mice [53,54]. While our previous studies have shown the enrichment of Akkermansia muciniphila upon treatment with poorly bioavailable polyphenolic antioxidants, the increase in Akkermansia as a member of CAG9 observed in this study was not significant. Oscillibacter, Peptococcus, and Intestinimonaswere the ASVs noted in CAGs that decreased in abundance upon treatment with Fe⁰. Oscillibacter is negatively correlated with transepithelial resistance [55], while Peptococcus and Intestinimonas are associated with intestinal inflammation and the obese phenotype, respectively [56].

One overarching trend observed in our data was an overall Fe⁰-mediated increase in the abundance of gut bacteria without functional EAS (Fig. 6). This increase may be explained by the ability of Fe⁰ to remove GI oxygen and associated ROS from the GI lumen, thus creating a more hypoxic environment, even in the obese animals with intestinal permeability. In turn, Fe⁰-mediated reduction in the intestinal ROS decreased the need for the protective enzymatic systems that can deactivate ROS. Earlier human studies have documented a greater proportion of aerotolerant bacteria in fecal samples from individuals with chronic malnutrition. The proportion of aerotolerant to strict anaerobic bacteria was shown to correlate to the redox potential (mV) of the fecal samples as measured by a redox meter [57]. Notably, no guilds exclusively consisted of ASVs that lacked a functional EAS, which suggests that the difference in EAS occurs independently from co-abundance patterns among bacterial populations. Our findings delineate a mechanism Fe⁰ influencing metabolic health and the gut microbiome as a result of ROS and oxygen-scavenging in the gut, however, it is likely the direct scavenging ability of Fe⁰ in the microenvironment of the gut is not the only mechanism by which Fe⁰ modifies the gut microbiome. The influence of iron supplementation on changes to the microbiome has been explored with interest towards the idea of iron being toxic to certain groups of bacteria. Iron is often a rate-limiting nutrient for the growth of specific bacterial species, and long-term iron fortification has been shown to increase Enterobacteriaceae and decrease Bifidobacterium in human infants [58]; a similar trend was observed after long-term consumption of electrolytic iron where an increase in enterobacteria was accompanied by a decrease in Lactobacilli. Whether these results are from a targeted decrease or toxic effect on lactic acid bacteria, or from the creation of a more supportive environment for other taxa was not elucidated [59]. Iron is thought to influence the gut microbiome through a variety of mechanisms including changes in gene expression of certain bacterial species [60] and alteration of host immune signaling [61]. These mechanisms, in addition to our proposed radical scavenging mechanism, likely all contribute to the observed changes in the gut microbiome in this study, and insight towards the interplay between all potential mechanisms provides an exciting target for future studies.

Overall, our data suggest that Fe⁰, a poorly-absorbed antioxidant, improves metabolic health, possibly by modifying the redox environment of the GI tract. The administration of Fe⁰ has been shown to decrease ROS and oxygen levels in the GI lumen, which may alter the gut microbiome. We also demonstrated that measuring Fe⁰ oxidation in the GI may be a diagnostic tool for GI oxidative stress. The assay presented here is a simple, rapid and reliable method for measuring indices which were previously expensive and invasive to characterize. These results indicate that luminal ROS and oxygen scavenging is a novel therapeutic target for MetS and gut inflammation, which warrants further exploration in pre-clinical and clinical settings.

Materials and Methods

Chemicals and reagents

Indocyanine green (Cardiogreen) was purchased from Sigma Aldrich (St. Louis, MO) and converted to hydroindocyanine green (hICG) as previously described using methanol and sodium borohydride (Sigma Aldrich, St. Louis, MO) [9,47,50]. Elemental iron (Fe⁰; 325 mesh) was purchased from Beantown Chemical (Hudson, NH). Ferrozine hydrate (Thermo Fisher Scientific, Waltham, MA), hydroxylamine hydrochloride (Sigma Aldrich, St. Louis, MO), and hydrochloric acid were used to prepare the reagents for the modified ferrozine assay.

Animals and protocols

Male C75Bl/6J mice were purchased from Jackson Labs (Bar Harbor, ME). Mice were fed either a low-fat diet (LFD, 10% kcal from fat; D12450J; Research Diets, Inc., New Brunswick, NJ) or a high-fat diet (HFD, 60% kcal from fat; D12492; Research Diets, Inc., New Brunswick, NJ). All animal procedures were undertaken with the approval of Rutgers University’s Institutional Animal Care and Use Committee.

Near-infrared fluorescence imaging

In situ imaging of ROS was performed as previously reported [9]. Briefly, abdominal hair was removed from mice 24 h prior to imaging by shaving and applying Nair™ to the shaved area for 3 min, followed by rinsing with warm water. In all imaging experiments, animals were gavaged with hICG at a dose of 2 g per kg BW. One h after dye administration, mice were anesthetized with 3% isoflurane and placed ventrally on the imaging stage of the In-Vivo MS FX PRO imaging system (Bruker, Ettlingen, German) capable of capturing both brightfield and infrared images of the experimental animals at precise and reproducible positioning angles. Excitation illumination was carried out at 760 nm, followed by emission at 830 nm, which was recorded using a filter-equipped high-sensitivity camera. Fluorescent image acquisition was carried out for 30 s, followed by a brightfield light photograph with 0.5 s exposure time. The representative images shown in the results are a superimposition of the fluorescent image over the brightfield photograph for anatomical orientation.

Fluorescent intensity was quantified in units of photons/s/mm² using Bruker Molecular Imaging software. The background intensity of each image was set to zero and fluorescent range normalized to a range of 7.4×10⁴ to 9.0×10⁶ photons/s/mm², identical elliptical regions of interest were drawn on each fluorescent image (117.43 × 159 pixels; interior area = 14,627).

Determination of effective dose and time for Fe⁰ treatment

Nine C57Bl/6J mice fed the aforementioned LFD were housed three animals per cage and allowed two weeks to acclimate prior to experiments. The mice were randomly divided into three treatment groups: control, low dose (2.5 g Fe⁰ per kg BW), and high dose (10 g Fe⁰ per kg BW). Animals were fasted for 4 h prior to oral administration of Fe⁰ or a volumetrically equivalent dose of water, and the near infrared imaging was carried out as described above. To observe the timeframe for ROS scavenging, a low dose of Fe⁰ was administered to control mice immediately after their initial imaging, and a second image was taken after 1 h. Following Fe⁰ administration, animals were observed for any signs of pain or distress including changes in respiration pattern, responsiveness, posture, and rearing. Animals were euthanized via CO₂ asphyxiation and dissected for evidence of macroscopic lesions in the GI.

Effective dose was further investigated with 30 C57Bl/6J mice on a HFD, housed in groups of five mice per cage and randomly divided into three treatment groups: control, low dose (0.5 g Fe⁰ per kg BW), and high dose (2.5 g Fe⁰ per kg BW). Animals were fasted for 4 h prior to oral administration of iron or a volumetrically equivalent dose of water. Thereafter, near infrared imaging and quantitative analysis was carried out as described above.

Single- and 14-day repeat dose studies

Ninety-six 9-week-old C57Bl/6J mice were purchased from Jackson Laboratories (Bar Harbor, ME). Forty eight mice had been fed a LFD while the other 48 had been fed a HFD (60% kcal from fat; D12492; Research Diets, Inc., New Brunswick NJ). All mice were weighed prior to the start of experiments and randomly divided into four groups wherein animals on the LFD and HFD would continue receiving their previous diets, supplemented with either the 2.5 g Fe⁰ per kg BW or a volumetrically equivalent dose of water. The groups were assigned as follows: LFD, LFD-Fe, HFD and HFD-Fe. Each group comprised 24 animals. Animals were housed in groups of three based on their treatment assignment and acclimated for 3 weeks prior to the start of experiments. Within each cage of three mice, each mouse was randomly given a sub-assignment based on one of three endpoints for the experiment- single-dose, repeat-dose, and tissue collection.

Throughout the experiment, all animals were weighed every 3 d. Food consumption was measured by weight every three days and recorded as g of food consumed per mouse per day.

OGT was tested prior to the start of the experiment and after the duration of the experiment [11]. For each OGT test, mice were fasted for 6 h prior to measurement of FBG using Clarity BG1000 Blood Glucose Monitoring System (Clarity Diagnostics, Boca Raton, FL). Animals were then gavaged with 2 g glucose per kg BW, and blood glucose tested after 30 min and again after 1, 2, and 4 h. OGT was calculated as the area under the curve (AUC).

Two days prior to the start of experiments, each mouse was weighed and OGT was tested. Fecal samples for the microbiome were collected one day before treatment (day 0). On day 1 of the experiment, all animals were treated with 2.5 g Fe⁰ per kg bodyweight or the equivalent water volume after a 4 h fast. Mice assigned to the single-dose subgroup were gavaged with hydroindocyanine green (hICG) 3 h after Fe⁰ administration, and imaging was carried out as previously described. These mice were removed from the experiment after imaging. Food was returned to the mice assigned to the repeat-dose and tissue collection subgroups, and fecal samples were collected over the course of 6 h after Fe⁰ treatment for fecal iron redox assessment.

Repeat-dose and tissue collection mice were treated with 2.5 g Fe⁰ per kg bodyweight or the equivalent water volume daily for the following 13 days for a total treatment time of 14 days.

On day 14, OGT was tested in the repeat-dose mice using the previously described procedure. Six h after Fe⁰ administration, fecal samples were collected from both repeat-dose and tissue collection mice for fecal iron oxidation analysis.

On day 15, all mice were fasted for 5 h. The repeat-dose mice then underwent NIRF imaging as described above.

Ferrozine Assay for Relative Oxidative Index (ROI) of the gut

The ROI of the gastrointestinal lumen was assessed by measuring the amount of oxidized iron relative to total iron excreted over the course of 6 h post-gavage. This was done using a modified version of the ferrozine assay [34–36]. This assay is based on the ferromagnetic properties of elemental iron versus ferric oxide. In short, total iron is measured in a homogenized sample before and after treatment with a neodymium magnet (Fig. 2). As elemental iron is ferromagnetic and ferric oxide is not, the total iron measured in solution after treatment with the magnet is representative of iron that has been converted to iron oxide.

Fecal iron oxidation status was assessed by collecting fecal samples from each mouse over the course of 6 h, beginning immediately after gavage with the Fe⁰ suspension. Fecal pellets were stored on dry ice immediately after excretion and pooled for each mouse over time. Upon analysis, pooled fecal samples from each animal were suspended in 4 mL 0.5 M HCl and homogenized for two rounds of 5 min each using a Spex MiniG 1600 (Spex Sample Prep, Metuchen, NJ). After each round of homogenization with the MiniG 1600, samples were manually ground for 30 s using a plastic pestle. After homogenization, a 15 μL aliquot of the sample (labeled UT for untreated) was added to 900 μL of total iron buffer (TFeB; 0.2 mg/mL Ferrozine, 11.9 mg/mL HEPES, 10 mg/mL hydroxylamine HCl, adjusted to pH 7). A neodymium magnet was then wrapped in one layer of weighing paper to prevent fouling of the magnet and used to cover the opening of the sample vessel. The sample vessel was inverted three times to remove elemental iron from the fecal slurry. A 15 μL aliquot was then removed from this sample and added to a second preparation of 900 μL TFeB labeled MT for “magnet-treated.” After 1 h of incubation in the dark at room temperature, samples were measured for absorbance at 562 nm using a Biotek Synergy HT Multi-Detection Plate Reader (Winooski, VT). Relative oxidation index (ROI) was calculated as a ratio of total iron content in magnet-treated to untreated samples. A greater value demonstrates that a sample contains a greater proportion of oxidized iron. ROI was measured only for animals that received the Fe⁰ treatment throughout the course of the study. The use of ferromagnetic properties to distinguish oxidized iron from elemental iron was validated in a model system wherein samples of 125 mg Fe⁰ were suspended in 0.5 M HCl with increasing concentrations of H₂O₂ from 0–2% for 6 h, similar to the time between Fe⁰ gavage and fecal collection in the animal model.

Microbiome Analysis

A total of 186 fecal samples were collected at day 0 (N = 23–24 per group), day 7 (N = 15–16 per group) and day 15 (N = 7–8 per group). Genomic DNA was extracted using the QIAmp Power Fecal DNA kit (QIAGEN, Germantown, MD) as per the manufacturer’s instructions. The hypervariable region V4 of the 16S rRNA gene was amplified using the 515F and 806R primers modified by Parada et al. [62] and Apprill et al. [63], respectively, and sequenced using the Ion GeneStudio S5 (Thermo Fisher Scientific, Waltham, MA). Primers were trimmed from the raw reads using cutadapt [64] in QIIME 2 software [65]. Amplicon sequence variants (ASVs) [66] were obtained by denoising using the dada2 denoise-single command in QIIME 2 with parameters –-p-trim-left 0 –p-trunc-len 215. Spurious ASVs were further removed by abundance filtering [67]. A phylogenetic tree of ASVs was built using the QIIME 2 commands alignment mafft, alignment mask, phylogeny fastree, and phylogeny midpoint-root to generate the weighted UniFrac metric. Taxonomy assignment was performed by the q2-feature-classifier plugin [68] in QIIME 2 based on the silva database (release 132) [69]. The data were rarified to 12,000 reads/sample for subsequent analyses.

Overall gut microbiota structure was evaluated using alpha diversity indices (Shannon index, observed ASVs, and Faith’s phylogenetic diversity) and beta diversity distance metric (weighted UniFrac). Principal coordinates analysis (PCoA) was performed using the R package “ape” [70] to visualize community dissimilarities between treatment groups along the two principal coordinates that accounted for most of the variations. Random Forest analysis was performed and cross-validated using the R package “randomForest” e [71] and the “rfcv” function, respectively, to test for correlations between gut microbiota composition and host phenotypes. Figures were generated using the R “ggplot2” [72] and “pheatmap” packages [73].

ASVs shared by >20% of the samples were considered prevalent and selected for further analysis. Pairwise correlations among the ASVs were calculated using the method described by Bland and Altman [74]. The correlation values were converted to a correlation distance (1 –correlation value) and the ASVs were clustered using the Ward clustering algorithm. From the top of the clustering tree, permutational multivariate analysis of variance (PERMANOVA; 9999 permutations with a P < 0.001 cut-off) was used to sequentially determine whether the two clades were significantly different and accordingly clustered the prevalent ASVs into CAGs [40]. Bray-Curtis distance were used to compare community dissimilarities at the guild level. Functional prediction of ASVs was performed using PICRUSt2 [41]. Masslin2 was applied to find the correlations between CAG abundance (AST transformed) and ROS with adjustments made for the different groups.

Statistical analysis

All statistical analyses of the mouse phenotypic data were performed using Prism 9 (GraphPad Software, San Diego, CA). Two- or three-way ANOVAs or repeated measures ANOVAs were performed when appropriate. Tukey’s, Šídák’s and Šídák-Holms multiple comparison tests were used to examine differences in all variables among treatment conditions. All statistical analyses of the gut microbiota were performed in R. Specifically, alpha diversity at different time points within the same treatment group was compared using the Wilcoxon matched-pairs signed-ranks test (two-tailed) and alpha diversity of the same treatment group at the same time point was compared using the Mann-Whitney test (two-tailed). Comparisons of gut microbial community dissimilarity was performed using PERMANOVA. Kruskal–Wallis tests and post hoc Dunn’s tests were performed to compare CAGs among different groups at each time point. Results were considered significant at P < 0.05.

Data Availability

Raw gut microbiome sequencing data have been deposited to the sequence read archive at NCBI under the accession number BioProject ID PRJNA655513.

Highlights.

• Oral administration of Fe⁰ reduces ROS in the GI of obese and hyperglycemic mice and improves their glucose metabolism.

• Daily Fe⁰ treatment increased the relative abundance of gut microbes without functional antioxidant systems, suggesting that that the presence of antioxidants in the GI tract reduces the need for microbiota with the ability to neutralize ROS.

• Luminal oxygen scavenging is a potential mechanism by which poorly absorbed antioxidants may modify the gut microbiome and metabolic status of obese mice.

• Reducing the levels of GI ROS and oxygen may be a novel therapeutic target for mitigating MetS; additionally, non-bioavailable antioxidants, such as Fe⁰, can improve carbohydrate metabolism and gut integrity.

• The simple, rapid and non-invasive test developed for assessment of GI redox environment by simultaneously measuring total and oxidized Fe⁰ in feces, following oral administration of Fe⁰, may have significant impacts on diagnosis of MetS and dysbiosis if applied in a clinical setting.