A Calcium-Rich Multimineral Intervention to Modulate Colonic Microbial Communities and Metabolomic Profiles in Humans: Results from a 90-Day Trial

Epidemiologic studies have shown an inverse relationship between calcium intake and colon cancer incidence (1). Experimental studies in animals have substantiated antitumor efficacy in the colon (2), and in vitro studies have provided mechanistic insight into how calcium influences epithelial cell proliferation and differentiation (3). In spite of these data, interventional trials with calcium have had only modest success in reducing colon polyp formation, with some trials demonstrating a reduction in incidence (4, 5), while others showing essentially no protection (6, 7) or an increase in incidence of colon polyps with a more-aggressive sessile-serrated phenotype (8). This complicated picture supports the idea that adequate or optimal calcium intake throughout life may be beneficial, but in a realistic dietary setting, the efficacy of supplementation by calcium alone may be limited by dietary complexity. Potentially adverse effects at higher dose levels further complicate the picture (9).

Recent evidence suggests that the antineoplastic activity of calcium supplementation may be enhanced by concomitant inclusion of additional trace minerals along with calcium. This has been demonstrated in our own studies using Aquamin, a calcium- and magnesium-rich multimineral product obtained from mineralized red marine algae. In two long-term (15–18 months) studies in mice, Aquamin more effectively suppressed colon polyp formation than calcium alone (10, 11). Further, Aquamin was more effective than calcium alone at suppressing proliferation and inducing differentiation in human colon carcinoma cells in monolayer culture (12, 13) and human adenoma–derived colonoids (14). Thus, Aquamin (as a source of calcium along with magnesium and additional trace minerals) may, ultimately, prove to be a more effective dietary colon polyp chemopreventive agent than calcium alone.

Exact mechanisms by which calcium alone or multimineral supplementation exerts antineoplastic activity are currently unknown. In addition to direct antiproliferative and prodifferentiating effects on colonic epithelium, beneficial activity may be mediated indirectly through effects on the gut microbial population and/or changes in gut microbial metabolic activity. The bacterial community in the gastrointestinal tract plays several important roles that contribute to health. A “healthy” microbiome is important for food digestion and production of important metabolites. In addition, a healthy microbial community helps maintain the tissue barrier, regulates the host immune response, and provides protection against pathogen overgrowth (15). Dysbiosis, especially in the colon, can lead to barrier breakdown and initiate a chronic inflammatory response. Dysbiosis has been directly linked to inflammatory bowel diseases and indirectly to the formation of premalignant colon polyps (15–17). In previous studies in mice, dietary calcium supplementation has been shown to cause a shift in the gut microbial community in comparison with control (18, 19). In another study, Aquamin itself induced gut microbial changes (20).

Metabolic changes may also be important. At high concentrations, bile acids are membrane-active agents and can be cytotoxic (21). In addition, certain bacterially derived secondary bile acids have carcinogenic properties (22). Alterations in the gut microbial population can affect bile acid profiles (23). In our own murine study (19), calcium supplementation altered the gut microbiome and reduced the level of total bile acids and certain microbially derived secondary bile acids. In addition to bile acids, gut microbes also produce short-chain fatty acids (SCFA; e.g., acetate, propionate, and butyrate) in the colon. SCFAs have demonstrable protective effect against colonic inflammation and carcinogenesis (24). Thus, a shift in the gut microbial community may affect colon health through alterations in bacterially derived metabolites.

Whether similar microbial/metabolic changes might also be seen in human participants is not known. To test the feasibility of using dietary Aquamin as an interventional strategy in humans, we conducted a 90-day, FDA-approved pilot study, comparing daily supplementation with Aquamin with calcium alone and placebo in healthy human participants. The main purpose of the study was to assess Aquamin's effect on gut microbial community structure and bile acid, SFCA, and eicosanoid profiles at endpoint in comparison with baseline values in participants from all three groups. Elucidating the microbial and metabolic changes resulting from intervention is a first step toward identifying potential mechanisms of action with Aquamin. Identifying measurable alterations in microbial and metabolomic parameters and determining effect-size with intervention will also be helpful for determining the necessary sample size going forward in a larger-scale clinical trial. As part of the study, we also obtained information on tolerability and safety of Aquamin. Initial results from the current study are reported herein.

Aquamin is a calcium- and magnesium-rich natural product obtained from the skeletal remains of red marine algae of the Lithothamnion genus. In addition to calcium and magnesium, Aquamin contains detectable levels of 72 additional trace minerals including trace elements from the lanthanide family (essentially all of the minerals accumulated by the algae from seawater). Aquamin is sold as a food supplement (GRAS 000028; Marigot Ltd.) and is used in various products for human consumption in Europe, Asia, Australia, and North America. A single batch of Aquamin-Food Grade was used for this study. Mineral composition was established via an independent laboratory (Advanced Laboratories). Supplementary Table S1 provides a complete list of elements detected in Aquamin and their relative amounts. Aquamin has been used in past clinical studies in human subjects with no serious safety or tolerability issues reported (25–27). Calcium carbonate was used for active comparison, and maltodextrin was used as a placebo.

This study was a pilot, double-blind, parallel assignment, randomized clinical interventional trial in which 30 participants were included. Participants were male or nonpregnant female, in general good health, but having “an increased risk for colon cancer” based on (i) a personal history of colorectal polyp, early stage (stage I or II) colon cancer treated by surgical removal without recommendation for adjuvant therapy, or stage III colon cancer treated with surgery > 5 years prior or (ii) a first-degree relative diagnosed under the age of 60 with colorectal cancer. Exclusion criteria included history of kidney disease or kidney stones, Crohn's disease or ulcerative colitis, gastrointestinal hemorrhagic disorders, or coagulopathy, hereditary nonpolyposis coli, or familial adenomatous polyposis. The participants were recruited through the Michigan Medicine web portal and by posting flyers in the hospital.

This clinical interventional trial was conducted with FDA approval of Aquamin as an Investigational New Drug (IND#118194) and with oversight by the Institutional Review Board at the University of Michigan Medical School (IRBMED)—IRB#HUM00076276. The study was registered as an interventional clinical trial with details at Clinicaltrials.gov (study identifier NCT02647671). All participants provided written informed consent prior to inclusion. This phase I trial involving human participants was carried out in accordance with recognized ethical guidelines, for example, Declaration of Helsinki, International Ethical Guidelines for Biomedical Research Involving Human Subjects (CIOMS), the Belmont Report, and the U.S. Common Rule.

Figure 1 summarizes the study design of this trial in a flowchart. Briefly, at screening participants were given the NIH Diet History Questionnaire II (DHQ II), a food frequency questionnaire which includes portion size and dietary supplement questions as a way to evaluate baseline calcium levels in the past year (28). Participants were also asked about use of dietary supplements, antibiotics, and nonsteroidal anti-inflammatory drugs. No participants were on antibiotics at or in close proximity to the time of screening. Individuals ingesting supplements containing calcium and/or vitamin D were required to undergo a 2-week “wash out” period prior to starting and to not use these supplements during the study participation.

Thirty participants underwent baseline flexible sigmoidoscopy (unprepped; i.e., without bowel cleansing procedure). Twelve 2.5 mm colonic biopsies were obtained along with two stool specimens from within the sigmoid colon (20 cm above the anus). Tissue and stool samples were saved in 10% formalin, cryopreserved for cultures or snap frozen in liquid nitrogen, and saved at −80°C. Blood was drawn for the complete metabolic panel including liver function/liver injury markers (total albumin, bilirubin, aspartate aminotransferase, alanine aminotransferase, and alkaline phosphatase) and renal function tests (blood urea nitrogen and creatinine).

After baseline sigmoidoscopy, participants were randomized to one of three groups. Ten participants were treated daily for 90 days with Aquamin providing 800 mg of calcium per day. Ten participants received 800 mg of calcium carbonate daily, and ten participants received maltodextrin as placebo. During the interventional period, participants were contacted by study coordinators on a monthly basis to assess study progress/adherence to the study protocol and to identify unwanted side effects. Compliance was assessed by capsule log entries and by counting unused capsules returned at the end of the study.

At the end of the 90-day intervention period (90 ± 5 days), participants again underwent unprepped flexible sigmoidoscopy and eight colonic biopsies along with two stool specimens were collected and stored as at baseline. Blood was also taken for the same serum markers as at baseline. For each of the two timepoints, one biopsy and one stool specimen from each participant were utilized for microbiome analysis, and one biopsy and one stool specimen were utilized for metabolomic analysis. The remaining tissue samples were retained for backup or were used for histologic, IHC, and proteomic analyses. The results of the latter analyses will be reported separately.

DNA was isolated from colon and stool samples using the Qiagen MagAttract PowerMicrobiome DNA/RNA kit. Total bacterial DNA levels were estimated using qPCR with broad-range primers BSF8 (AGAGTTTGATCCTGGCTCAG) and BSR357 (CTGCTGCCTYCCGTA), targeting bacterial 16S rRNA genes (29). Each 10 μL qPCR reaction contained 5 μL PowerUp SYBR Green Master Mix (Applied Biosystems by Thermo Fisher Scientific), 4 μL sample DNA (1:40 dilution), and 0.4 μmol/L of primers. The following cycles were run on a Light Cycler 96 (Roche): 1x(2 minutes at 50°C, 10 minutes at 95°C), 40x(15 seconds at 95°C, 1 minute at 60°C). Each sample was run in duplicate.

Microbial community profiles were generated by Illumina MiSeq sequencing of the V4 region of 16S rRNA–encoding genes after amplifying the extracted colon tissue and stool DNA of all the participants as described previously (30). Samples were amplified, normalized, and sequenced on the MiSeq, and analysis was performed using the MiSeq SOP (31) for the software mothur (v.1.39.0 and v.1.39.5) as described in our earlier study (19). In case of low bacterial biomass, 3 μL of DNA were amplified by touchdown PCR method [1x(2 minutes at 95°C), 20x(20 seconds at 95°C, 15 seconds at annealing temperature (starts at 60°C, decreases 0.3°C/cycle), 5 minutes at 72°C), 20x(20 seconds at 95°C, 15 seconds at 55°C, 5 minutes at 72°C), 1x(10 minutes at 72°C)]. Thirteen samples, including 8 of 10 postinterventional Aquamin colon samples, required touchdown PCR (with 40 amplification cycles vs. standard PCR with 30 cycles) to achieve sufficient sample amplification for sequencing. After processing, sequences were binned into operational taxonomic units (OTU) based on 3% difference in sequence using the OptiClust method (32). Pre- and postcomparisons between groups included differences in community structure using OTU-based Yue and Clayton distance metric (θ_YC; ref. 33). θ_YC distances were visualized using principal coordinates analysis (PCoA). θ_YC is a beta diversity metric. Specific OTUs driving community differences were identified by linear discriminant analysis effect size (LEfSe; ref. 34), as was done in our previous study (19). LEfSe utilizes both statistical significance and effect size (linear discriminant analysis score, or LDA) to determine the features (OTUs, in this case) that are differentially abundant between groups (i.e., pre- and postdifferences). FDRs of the significant LEfSe data were corrected using the p.adjust() function with the Benjamini–Hochberg algorithm employing R, and stated as q values. Alpha diversity was assessed by change in the Shannon diversity index from baseline. To evaluate Aquamin-associated microbial community differences in major gut phyla, we sorted the 1,000 most abundant OTUs in each of the baseline and final visit samples into phyla and compared intergroup differences in each of these phyla at endpoint. The DNA isolation and 16S rRNA gene sequencing were done by the University of Michigan Microbial Systems Molecular Biology Laboratory. Sequence data generated in this project are accessible within the NCBI Sequence Read Archive database (BioProject accession number: PRJNA575562).

Bile acids, SCFAs, and eicosanoid composition were quantified in colon biopsies and stool specimens from a randomly chosen subset of participants (n = 6 per group) at each time point (pre- and postintervention). Bile acids were quantified using LC-MS in a two-step solvent extraction (35) as performed in the Regional Comprehensive Metabolomics Resource Cores (RCMRC) at the University of Michigan. Supernatants were combined, dried, and resuspended for LC-MS separation by reverse-phase liquid chromatography (RPLC) and measured by multiple reaction monitoring (MRM). Sample identification was performed by comparison of retention times and mass with an in-house library of bile acids. SCFA analysis was performed by the RCMRC by electron ionization-gas chromatography–mass spectrometry as described recently (36). Compound identification was performed against a library of known SCFAs. Colon tissue was also utilized to analyze a profile of eicosanoids. Eicosanoids were extracted and concentrated using solid phase extraction in colon biopsies. The eluent was dried and resuspended for LC-MS separation by RPLC and measurements by MRM methods (37). Stool specimens were not analyzed for eicosanoids because these metabolites are not bacterial products. All analytes were reported as pmol/mg, after normalization to the sample weight. Metabolomics data acquired in this study are available on the Metabolomics Workbench (Project ID: PR000822; doi: 10.21228/M8F69D) as part of National Metabolomics Data Repository.

Pre (baseline) versus post (endpoint) comparisons were conducted for each microbial or metabolomic endpoint for individual participants in each of the three treatment groups (placebo, calcium, Aquamin). Group means and SDs were then calculated for baseline and endpoint measures. Intergroup differences for normally distributed continuous data were made by ANOVA followed by pairwise group comparisons with Bonferroni corrections for multiple comparisons to generate an adjusted P value. The two-stage linear step-up procedure of Benjamini, Krieger, and Yekutieli was applied to correct for multiple comparisons by controlling the FDR. Microbial community distances were compared by analysis of molecular variance (AMOVA) within the program mothur (v.1.39.0 and v.1.39.5; ref. 38). For continuous data that failed normality testing, intergroup comparisons were made by Kruskal–Wallis ANOVA followed by Mann—Whitney–Wilcoxon pairwise comparison. Statistical analyses were performed using GraphPad Prism (version8) and R computing software (R version 3.5.3 and RStudio Version 1.1.463). Pearson correlation coefficients were applied to compute correlations among colon and stool samples for θ_YC metric. A value of <0.05 was considered significant for both P and q values for all the analyses. Due to the small cohort size, analyses were not adjusted to any baseline sociodemographic or clinical characteristics or dietary calcium intake.

Thirty-six participants were randomly enrolled. Six subjects were lost to follow-up prior to intervention, and a total of 30 participants (10 per arm) completed the study (Fig. 1). This included of 22 female and 8 male participants. Participants were randomized to study arms without regard to age or gender (placebo: 4 males and 6 females; calcium: 2 males and 8 females; Aquamin: 2 males and 8 females). Ages ranged from 20 to 66 years. Compliance (capsule intake) was estimated to be 96% across the three groups. Demographic characteristics are presented in Supplementary Table S2. Based on responses to DHQ II, the average dietary calcium intake (mg/day) values for the three groups at baseline were estimated to be: placebo = 817 ± 245; calcium = 964 ± 412; and Aquamin = 919 ± 545 (no significant differences found).

Self-reported adverse events over the course of study are shown in Table 1. All adverse events were minor (i.e., headache, gastrointestinal symptoms) and did not preclude any individual from completing the study. The number of individuals reporting adverse events in the Aquamin group was the same as in the placebo group (3 of 10), whereas 6 of 10 individuals in the calcium group reported one or more events. No serious adverse events (defined as necessitating cessation of study participation or medical intervention/hospitalization) occurred. Serum metabolic markers (liver function/liver injury and renal function) are presented in Table 2. No significant change in any individual marker was observed over the course of the study. Likewise, there was no difference among the treatment groups in a panel of metabolic markers and no change in serum calcium levels before and after intervention (Table 2).

qPCR was used to assess bacterial 16S rRNA gene levels in colon biopsy and stool specimens as a way to estimate total bacterial DNA. Figure 2 demonstrates that in both colon biopsy and stool specimens, posttreatment Aquamin samples had higher cycle quantification (Cq) values, indicating lower amounts of total bacterial DNA than at baseline (Colon: P = 0.0001/q < 0.0001; Stool: P < 0.0001/q < 0.0001). There was also a decrease in total bacterial DNA in specimens from calcium-treated participants (Colon: P = 0.032/q = 0.0021; Stool: P < 0.0001/q < 0.0001), although, on average, the decrease was not as great as that seen with Aquamin. Placebo specimens showed no (average) decline in DNA content at endpoint.

To determine if Aquamin altered the composition of the gut microbiota in the participants, we analyzed bacterial 16S rRNA gene sequences from colon and stool specimens to explore pre- postinterventional differences. After sequence processing and exclusion of 2 samples for low sequence counts (i.e., one placebo stool and one calcium stool), a total of 7,081,672 sequences from 118 samples [median: 59,233 ± 23,618 (SD) sequences/sample; range 2,338–11,4235 sequences/sample] were included in this analysis. When we analyzed the sequence data by samples types for both colon biopsy and stool specimens, there was a total of 3,051,764 sequences from 60 colon samples [median: 49,853 ± 17,198 (SD) sequences/sample; range 12,481–102,310 sequences/sample]. The total number of sequences from 58 stool samples was 4,029,908 [median: 76,797 ± 25,547 (SD) sequences/sample; range: 2,338–11,4235 sequences/sample].

Shifts in gut microbial community composition were assessed by calculating θ_YC distances between the pre- and postsupplementation communities (beta diversity) for each individual. Individual pre- postdistance values, grouped by intervention, are shown in Fig. 3A and B. Colonic microbial communities demonstrated a bigger within-participant shift in the Aquamin-supplemented group than was observed for either the placebo or calcium-supplemented group. The difference between Aquamin and calcium in the colon biopsy specimens reached the level of statistical significance (P = 0.0087/q = 0.0061). Both figure insets show that the majority of participants that received Aquamin (8 of 10 in colon and 7 of 10 in stool) were above the median value for within-participant bacterial community change from pre- to postsupplementation for all samples. In colon biopsies, when samples were pooled as shown in the Fig. 3A inset, θ_YC distance was significantly increased in Aquamin subjects (n = 10) as compared with both calcium and placebo subjects (n = 20; P = 0.019). Figure 3C demonstrates a strong correlation between the two specimen types (colon biopsy and stool) for within-participant microbial community shifts (θ_YC distances; P = 0.0041). The correlation data, thus, suggest that stool specimens could be used as an estimate of the bacterial community in the colonic mucosa. It should be noted, however, that the stool specimens used here were obtained during flexible sigmoidoscopy and were not shed specimens. In addition, as part of this analysis, intrasubject versus intersubject variability was determined. θ_YC distances between the pre- and postintervention samples for each of the 30 individuals were compared with the θ_YC distances between each preintervention sample and all other samples. As seen in Supplementary Fig. S1, intrasubject differences were smaller than intersubject differences in both colon and stool samples (P ≤ 0.0001).

Figure 4A–D presents PCoA of θ_YC distances between samples for all three treatment groups in both colon and stool specimens. At baseline, there were no significant differences in the colon bacterial communities (Fig. 4A), but following the 90-day intervention period, microbial communities in the colon biopsies from participants treated with Aquamin segregated from those in the placebo or calcium groups (Aquamin vs. placebo: AMOVA P = 0.005 and Aquamin vs. calcium: AMOVA P = 0.009; Fig. 4A). Further, colonic microbial communities from the posttreatment Aquamin biopsy samples segregated from their own baseline communities (AMOVA, P = 0.022; Fig. 4B), whereas there was no difference in pre- versus postsupplementation communities for placebo- or calcium-treated participants (Fig. 4A). In contrast to these results in colon biopsy material, stool sample microbial communities were not different between or within groups either at baseline or at endpoint (Fig. 4C and D).

Alpha diversity was assessed in both colon and stool samples using the Shannon diversity index. In Aquamin-treated participants, there was a decrease in Shannon diversity (Fig. 4E and F). For colon biopsy samples, Shannon diversity was significantly decreased (P = 0.0037/q = 0.0026) as compared with the placebo.

First, we determined the total number of phyla and genera identified in each sample type. In colon biopsies, there were 15 different phyla and 224 genera identified. In stool samples, there were 10 phyla and 213 genera found. Overall, there were 7,267 OTUs detected; out of those, 2,573 were present in colon biopsies, 6,642 were present in stool samples, and 1,948 were common in both. A list containing all the phyla and genera identified are presented in Supplementary Tables S3 and S4.

Next, we assessed alterations in the relative abundance of individual OTUs representing the major gut phyla to compare intergroup differences in each of these phyla at endpoint. Bacteroidetes, Firmicutes, and Verrucomicrobia were decreased with Aquamin, whereas Actinobacteria and Proteobacteria were increased (with P ≤ 0.0001/q ≤ 0.0001). Trends were similar in both colon and stool specimens (Fig. 4G and H). Firmicutes OTUs also showed a drop with calcium intervention (colon biopsies only; P < 0.0001/q < 0.0001), but there was little change in the other phyla. Little change in OTU relative abundance in any of the major phyla was seen with placebo.

Differentially abundant OTUs were identified by LEfSe to explain the pre- and postsupplementation differences with Aquamin treatment. In the Aquamin-supplemented group, 82 OTUs were identified in colon samples (Tables 3A and 3B), and only 9 OTUs were identified in stool samples (Tables 3C and 3D,Table 3C,Table 3D) with an LDA score >2 and significance P < 0.05 (q < 0.05). Overall, more OTUs decreased in relative abundance than increased after supplementation. Differences between pre- and post-Aquamin supplementation were driven principally by an increases in OTUs within the normally less abundant phyla Proteobacteria, and Actinobacteria, along with three OTUs of phylum Firmicutes (Table 3A) and a decrease in OTUs within the normally higher abundance phyla Firmicutes and Bacteroidetes (Table 3B). Similar trends were also observed in stool samples (Tables 3C and 3D).

Alterations in bile acid levels were assessed in stool samples from Aquamin- and calcium-treated participants in comparison with placebo as shown in Fig. 5. Aquamin treatment resulted in a net decline in total bile acids (P = 0.0375/q = 0.0262) with the major portion of this change being in unconjugated forms (the major components of the fecal bile acid pool; Fig. 5A). The major unconjugated primary bile acids—i.e., cholic acid (CA) and, particularly, chenodeoxycholic acid [CDCA; measured concurrently with deoxycholic acid (DCA)]—were significantly lower in posttreatment Aquamin stool samples (P = 0.0074/q = 0.0052 and P = 0.0310/q = 0.0217 respectively; Fig. 5B). This was not observed with calcium.

Of the measurable secondary bile acids, taurine-conjugated ursodeoxycholic acid (TUDCA), measured concurrently with taurohyodeoxycholic acid (THDCA), was decreased (P = 0.0287/q = 0.0161), as were the α and ω muricholic acids (minor secondary bile acid components; P = 0.0012/q = 0.0009 and P = 0.0003/q < 0.0001 respectively; Fig. 5C, inset). Two other unconjugated secondary bile acids [hyocholic acid (HCA) and hyodeoxycholic acid (HDCA)] were significantly decreased with Aquamin as compared with placebo (P = 0.015/q = 0.010 and P < 0.0001/q < 0.0001 respectively; Fig. 5C). HDCA is a byproduct of gut microbial metabolism (39), utilizing HCA and muricholic acids after microbial deconjugation and enzymatic modification. Overall, these Aquamin-associated bile acid changes in stool are consistent with both a decreased bile acid pool and decreased bacterial conversion of primary to secondary bile acids. Of interest, these effects were unique to Aquamin. Calcium supplementation did not show a measurable effect on total fecal bile acids, and only HDCA was significantly decreased (Fig. 5C).

Bile acid levels were also assessed in colon biopsy samples from Aquamin- and calcium-treated groups in comparison with placebo (Supplementary Fig. S2). Similar trends to those seen in stool specimens were observed, but none of the changes with Aquamin or calcium intervention reached statistical significance in comparison with placebo except HDCA which was significantly reduced with Aquamin (P = 0.0456/q = 0.0319) as compared with placebo.

SCFA levels were assessed in both stool and colon samples. Apart from a modest (but statistically significant) increase in acetate in stool samples from Aquamin-treated individuals relative to calcium (P < 0.0001/q < 0.0001), no other statistically significant alterations were observed in either stool (Fig. 5D) or colon (Supplementary Fig. S2) specimens.

As part of the analyses, we attempted to correlate bile acid changes and changes in SCFAs with alterations in microbial parameters. A weak correlation between total bile acid levels and θYC was observed, but this was not statistically significant (due, in large part, to small sample size).

There was no significant change from baseline detected for any eicosanoid apart from an increase in 13S-hydroxy-octadecadienoic acid (13S-HODE) in calcium-supplemented colon samples (Supplementary Fig. S3).

This study demonstrated that 90-day dietary intervention with Aquamin was well-tolerated by healthy individuals. There were no changes in liver injury/function and kidney function markers; there were no serious adverse events; and minor adverse events were no higher with Aquamin than with calcium alone or placebo. These findings are consistent with the previous reports (25–27). Although a much larger database will be required to establish safety, all of the data to date suggest that safety and tolerability will not be issues with Aquamin. At the same time, intervention with Aquamin resulted in measurable changes in the colonic microbial community and the attendant bile acid profile. We are currently carrying out a phase I/II therapeutic trial with Aquamin in ulcerative colitis in remission. The same microbial and metabolomic endpoints assessed here are being measured in the ongoing trial. It is our hope that these microbial and metabolomic findings will prove useful as biomarkers going forward in the interventional trial.

The microbial population present in the colon of an individual is sensitive to environmental influences, including diet (40, 41). Past studies have demonstrated that calcium (18, 19) and Aquamin itself (20) can influence the colonic microbial community in mice. In the present study, we found that Aquamin supplementation in human participants resulted in a decrease in total colonic microbial DNA and an overall decrease in OTUs within the major gut phyla Firmicutes and Bacteroidetes, with a decrease in gut microbial diversity. Thus, the major impact may simply be a decrease in total gut bacteria. The impact of the decrease in bacteria on colonic health is unknown at this time but poses interesting questions for future exploration. Germ-free mice have shown increased resistance to chemical or dietary-induced colon cancers (42). Because our own previous studies have shown that Aquamin-treated mice have decreased incidence of colon polyps (10, 11) as well as fewer liver tumors (43), the possibility that Aquamin-associated microbial decreases may play a mechanistic role in these observations warrants further exploration. Of interest, many of the observations made with samples from Aquamin-treated participants were not seen in samples from individuals ingesting calcium alone. This argues that although calcium is the major mineral component in Aquamin, the effects of Aquamin on the gut microbial community cannot be attributed to calcium alone.

The changes we observed in the post-Aquamin treatment specimens are similar to the effects of broad-spectrum antibiotic treatment, which have been shown to cause a significant reduction in the proportions of Firmicutes and Bacteroidetes and an increase in Proteobacteria (44). As a follow-up to the observations reported here, it would be informative to assess the potential antimicrobial activity of Aquamin supplementation directly in a minimum inhibitory concentration assay. Metals, particularly cations, have previously been utilized as antimicrobial agents (45). For example, colloidal or nanoparticulate silver and mineral-rich clays are known to have broad-spectrum antimicrobial effects (46, 47). Notably, in our study, the administration of Aquamin did not result in diarrhea and caused only a small drop in bacterial diversity in colon samples; both common adverse effects of broad-spectrum antibiotic use. Thus, the antimicrobial effects of Aquamin appear to be mild in comparison with antibiotics and may allow some of the benefit of reduced gut microbial populations (e.g., reduced potentially carcinogenic secondary bile acids) without inducing symptoms that would render the intervention intolerable.

In parallel with microbial changes, Aquamin administration decreased the levels of total bile acids and selected primary and secondary bile acids. This finding is consistent with the decrease in total bacteria and may be potentially beneficial to colon health. At high concentrations, bile acids are membrane-damaging and cytotoxic (21). In addition, several secondary bile acids, most notably lithocholic acid (LCA) and DCA (22, 23), are known to be carcinogenic. Although these species could not be individually measured in this study for methodological reasons, the combination of DCA and CDCA (the primary bile acid precursor of LCA) was significantly lower in the Aquamin-supplemented group (Fig. 5). This suggests that one beneficial effect of Aquamin supplementation might be through effects on bile acid metabolism. A caveat in the interpretation of bile acid data is that it is not currently known whether the reduction in gut bile acid pools is due to altered bacterial composition or, potentially, to a direct effect on the bile acids, themselves. Previous studies have shown that calcium, the major component of Aquamin, has the capacity to precipitate bile acids (48). Although mineral binding and precipitation might interfere with detection, this could be beneficial if it prevented potentially harmful secondary bile acids from binding to colonic epithelium or entering the circulation. Arguing against direct binding and precipitation is the finding that calcium alone did not mimic the effects of Aquamin on the total bile acid pool.

A decrease in gut bacteria could, potentially, be harmful to colon health if it resulted in decreased concentration of protective bacterial metabolites, such as the colon-protective SCFAs (24). However, in this study there was no decrease in the colonic concentration of butyrate in the Aquamin-treated group and an actual increase in acetate. Thus, even with the apparent decline in total bacteria, the size and/or composition of the microbial pool was sufficient to maintain presupplementation levels of certain SFCAs.

Although the data strongly suggest an Aquamin-associated decrease in gut microbial bacteria, the specific shifts in gut microbial populations are harder to interpret. One complication of low microbial biomass samples during microbial sequencing is greater likelihood of detecting reagent or laboratory contaminant microbes in the amplification step. This has been documented in low-biomass samples such as glacier ice, air, rocks, etc. (49), but is not likely to be a major issue in the normal high biomass of the colon.

It should be noted that the antimicrobial effect of Aquamin might not be a direct consequence of the mineral components making up the supplement. Our recent studies employing colonoid culture technology demonstrated upregulation of proteins having antimicrobial activity upon treatment with Aquamin (50). These included lactotransferrin (51), natural resistance–associated macrophage protein-2 (52), and Ly6/PLAUR domain–containing protein 8 (53). Thus, it is possible that the antimicrobial effects seen with Aquamin in vivo are mediated, in part, through its effects on the colonic mucosa, itself. Finally, decreased bacteria, especially as it relates to colon biopsy specimens, might reflect altered bacterial adhesion. In our colonoid culture study (50), several mucins and CEACAMS were altered by Aquamin. Studies by others have noted that alterations in mucosal surface proteins affect microbial interactions with the mucosal wall (15, 16, 53).

In addition to the microbial and metabolomic changes, there are other, direct, beneficial effects of Aquamin on the epithelium of the colonic mucosa. Specifically, Aquamin suppressed colon epithelial proliferation and induced differentiation in colonoid culture. Effects on proliferation were seen, primarily, in colonoids derived from large adenomas (14), whereas improved differentiation—including upregulation of multiple cell–cell and cell–matrix adhesion molecules and barrier proteins—was observed in normal tissue-derived colonoids (50) as well as in adenoma colonoids (14). Direct effects on the colonic mucosa and indirect effects resulting from changes in microbial/metabolomic profiles are, of course, not mutually exclusive. Further, as noted above, the indirect effects (i.e., antimicrobial activity) may reflect changes induced in the colonic mucosa.

The primary limitations of this study were the small sample size and short duration, which were by design, as this was an initial tolerability study rather than an efficacy study. Although the study was too short and too small to provide definitive results with several endpoints, it was sufficiently powered to detect significant differences in several microbial and metabolomic (bile acid) endpoints in Aquamin-treated participants relative to the other two interventions. To better define effects on efficacy-related endpoints, we are currently beginning a 180-day interventional trial with Aquamin in participants with ulcerative colitis in remission. Another limitation was the potential for intrasubject (as well as intersubject) variability that needs to be acknowledged. Pre and postinterventional differences in the microbial community were observed in all subjects, but intrasubject variability was less than variability between individuals (Supplementary Fig. S1).

In summary, this pilot study demonstrated that 90-day dietary intervention with a calcium- and magnesium-rich multimineral supplement was well-tolerated in healthy human volunteers. Adverse events were mild and largely consisted of gastrointestinal symptoms (constipation, diarrhea, etc.) that did not preclude completion of the study. Reports of adverse events were actually less frequent with Aquamin than with calcium alone. Consistent with this, liver function/liver injury and renal function markers and other serum biochemical markers did not vary significantly with intervention. With respect to microbial and metabolomic findings, Aquamin supplementation resulted in an overall decrease in gut microbial numbers and a decrease in bile acids, including potentially carcinogenic secondary bile acids or their precursors. Despite the antimicrobial-like effects, concentrations of the colon-protective SCFAs were maintained. These findings, along with previous in vitro and animal data (10–14, 50) showing a beneficial effect of Aquamin on colonic health, support future longer-term interventional studies in human participants. Finally, the observation that Aquamin had a more pronounced effect on gut microbial populations and bile acid levels than calcium alone supports the view that the beneficial activity of Aquamin (calcium in conjunction with additional trace elements) cannot be attributed to calcium alone. This conclusion supports findings from a recent epidemiologic study suggesting that calcium in combination of additional minerals may be linked with the lower risk of colorectal cancer (54).

No potential conflicts of interest were disclosed.

Conception and design: M.N. Aslam, C.M. Bassis, S.M. Zick, D.K. Turgeon, J. Varani

Development of methodology: M.N. Aslam, C.M. Bassis, S.M. Zick, D.K. Turgeon, J. Varani

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.N. Aslam, C.M. Bassis, I.L. Bergin, K. Knuver, D.K. Turgeon, J. Varani

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M.N. Aslam, C.M. Bassis, I.L. Bergin, K. Knuver, A. Sen, J. Varani

Writing, review, and/or revision of the manuscript: M.N. Aslam, C.M. Bassis, I.L. Bergin, K. Knuver, S.M. Zick, A. Sen, D.K. Turgeon, J. Varani

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M.N. Aslam, C.M. Bassis, J. Varani

Study supervision: M.N. Aslam, D.K. Turgeon, J. Varani

This study was supported by NIH grant CA201782 including supplemental funding through the Office of Dietary Supplements to J. Varani. Metabolomic work was supported by a Pilot/Feasibility Grant from Michigan Nutrition and Obesity Research Center to I.L. Bergin and M.N. Aslam (through 2 P30 DK089503-06 to MNORC). This study also used services at the University of Michigan that were supported by NIH funding (UL1TR002240; KL2TR002241; and TL1TR002242) to the Michigan Institute for Clinical and Health Research (MICHR). We thank MICHR, the Michigan Clinical Research Unit, the University of Michigan Microbial Systems Molecular Biology Laboratory, and the RCMRC at the University of Michigan for help with various aspects of the study. We thank our study coordinators (Elaine Brady and Deepa Chandhrasekhar) for their assistance with the study. We also thank Marigot LTD. for providing Aquamin as a gift. Most importantly, we are very grateful to the study participants who made it possible for us to carry out this study.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.