Magnesium treatment increases gut microbiome synthesizing vitamin D and inhibiting colorectal cancer: results from a double-blind precision-based randomized placebo-controlled trial

Abstract

Background

Carnobacterium maltaromaticum and Faecalibacterium prausnitzii induce de novo gut synthesis of vitamin D to inhibit colorectal carcinogenesis in mice. Magnesium (Mg) treatment increases circulating vitamin D, and Mg homeostasis is dependent on TRPM7 genotype.

Objectives

We hypothesize that Mg treatment increases gut C. maltaromaticum and F. prausnitzii, and the effect differs by TRPM7 polymorphism.

Methods

The Personalized Prevention of Colorectal Cancer Trial is a double-blind, precision-based randomized controlled trial with 240 participants randomly assigned to both treatment and TRPM7 genotype. Stool, rectal swabs, and rectal mucosa were collected.

Results

Of 239 participants who completed the trial, 226 with valid microbiome data were analyzed (treatment n = 112, placebo n = 114). The interaction between treatment and TRPM7 genotype was only significant for C. maltaromaticum (P = 0.001) and F. prausnitzii (P = 0.02) in rectal swabs. In a stratified analysis by TRPM7 genotype without the missense variant, Mg treatment compared with placebo significantly increased abundance of C. maltaromaticum (0.217 ± 0.615 (23.01%) compared with –0.065 ± 0.588 (–6.30%); P = 0.006) and F. prausnitzii (0.105 ± 0.817 (2.13%) compared with –0.095 ± 0.856 (–1.92%); P = 0.04) in rectal swabs. The effect on C. maltaromaticum remained after multiple comparisons (Q = 0.05 for C. maltaromaticum across all sample types and genotypes). In those with the TRPM7 missense variant, Mg decreased C. maltaromaticum, but not F. prausnitzii, compared with placebo in rectal swabs [–0.065 ± 0.511 (–6.54%) compared with 0.133 ± 0.503 (13.30%); adjusted P = 0.04]. The effect did not remain after false discovery rate correction. Mg treatment’s effect on C. maltaromaticum in rectal swabs primarily appeared in females, and the treatment-genotype interaction remained significant.

Conclusions

In individuals with adequate TRPM7 function, Mg supplementation increases the abundance of C. maltaromaticum and F. prausnitzii.

Clinical Trial Registry

This trial was registered at clinicaltrials.gov as NCT04229992 (https://clinicaltrials.gov/study/NCT04229992?term=NCT04229992&rank=1). The parent study is registered as NCT03265483, and another relevant study is registered as NCT01105169.

Introduction

Despite a reduction in the incidence of colorectal cancer (CRC) due to increased endoscopic surveillance, CRC remains the fourth most common incident cancer in the United States [1,2]. Growing evidence suggests that the gut microbiome plays a critical role in colorectal carcinogenesis [3,4]. Recently, Li et al. [5] reported that Carnobacterium maltaromaticum, a lactic acid bacterium found in the gastrointestinal tract, is depleted in females with CRC [6]. Although it is known that UV radiation via sunlight is essential to vitamin D synthesis in humans [7], Li et al. [5] found that in mice, vitamin D can be synthesized in the absence of sunlight by C. maltaromaticum and other microbes, notably Faecalibacterium prausnitzii, which worked synergistically to induce de novo gut synthesis of vitamin D metabolites, activate colonic mucosal vitamin D receptor signaling, and inhibit CRC development. Metabolic cross-feeding of C. maltaromaticum with F. prausnitzii converted vitamin D metabolites produced by C. maltaromaticum to active vitamin D [5]. F. prausnitzii has also been associated with inflammatory bowel disease [8,9], irritable bowel syndrome [10], metabolic disease [11], and depression prevention and treatment [12,13].

We previously reported that magnesium (Mg) supplementation significantly increases circulating levels of vitamin D when baseline levels are low [14,15]. This finding was confirmed in subsequent trials [[16], [17], [18], [19]]. Although the mechanism by which this occurs is not fully understood, 1 explanation includes Mg being a cofactor for all the enzymes involved in vitamin D synthesis and metabolism [14,19]. Previous studies suggest that gut concentrations of Mg also play a critical role in regulating the growth and activity of microbiota [[20], [21], [22], [23], [24], [25], [26]]. Thus, in addition to enhancing bodily vitamin D synthesis and metabolism enzymes [15,19], we hypothesize that Mg treatment increases the abundance of C. maltaromaticum and F. prausnitzii in the gut compared with levels in the placebo arm, which may, in turn, increase de novo vitamin D synthesis and inhibit colorectal carcinogenesis [5]. Other proposed mechanisms by which Mg may modulate CRC risk have included improved insulin sensitivity [27] and promotion of colorectal adenocarcinoma cell apoptosis [28].

Transient receptor potential cation channel, subfamily M, member 7 (TRPM7) encodes a ubiquitously expressed cation channel essential to Mg homeostasis [29]. The inverse association between higher Mg intake and risk of colorectal adenoma differs by TRPM7 polymorphism [30]. Therefore, we hypothesize that the effect of personalized Mg treatment on gut microbes differs by TRPM7 genetic polymorphism. Additionally, we hypothesize that C. maltaromaticum mediates the effect of Mg treatment on circulating levels of vitamin D. We test these hypotheses within the Personalized Prevention of Colorectal Cancer Trial (PPCCT), a double-blind, precision-based randomized controlled trial (RCT) which was designed to test the interaction between Mg treatment and TRPM7 genotype on gut carcinogenic biomarkers [14,15]. We also conducted an exploratory analysis to examine whether C. maltaromaticum and F. prausnitzii are associated with risk of developing metachronous polyps.

Methods

Study population

The PPCCT is a double-blind 2 × 2 factorial precision-based RCT conducted from 2011 to 2016 at Vanderbilt University Medical Center (VUMC) in Nashville, Tennessee. The trial was approved by the VUMC Institutional Review Board (IRB #100106) and was registered on clinicaltrials.gov (NCT04229992). All participants provided written informed consent before participating in the study. Demographic information was collected. Race was self-reported by participants who were asked to select from the following on the initial questionnaire that best described them: White, Black, Asian, Alaskan/Indian, Pacific Islander, Other, Don’t Know, Decline to Answer. The detailed methods were previously reported [15]. In brief, eligible subjects were enrolled sequentially and randomly assigned in blocks of 2 or 4 with a 1:1 ratio to 2 treatment arms, Mg treatment or placebo, within 3 strata defined by the TRPM7 genotype (GG/GA/AA), a polymorphism at rs8042919 which results in a missense variant [Thr-1482 to isoleucine (Ile)]. Of note, no AA participants were identified despite significant efforts to recruit participants with the AA TRPM7 genotype. Therefore, only GG and GA genotypes were present in the analysis.

Precision nutrition is an approach to optimize nutrition based on making targeted nutritional recommendations based on an individual’s unique characteristics (i.e., genotype and background nutrition) [31]. In this precision-based trial, each participant in the treatment arm was assigned an individualized daily dose of oral magnesium glycinate for 12 wk that reduced their calcium-to-magnesium ratio to ∼2.3 based on baseline calcium and Mg intake obtained from 2 24-h dietary recalls [15]. Dietary recall assessments were interviewer-administered via telephone for each participant, with 1 on a weekday and 1 on a weekend day if possible. Participants also completed 2 dietary recalls at weeks 1–6 and 2 at weeks 7–12 of the study period. During the dietary assessment, information on participants’ use of medications, nutritional supplementation (vitamins, minerals, fiber, etc.), and other medical conditions was also collected. The Minnesota Nutrient Data System for Research was utilized to collect dietary data and calculate nutrient scores.

Participants, study investigators, and staff were blinded to the assigned interventions. Blinding was implemented through the Vanderbilt Investigational Drug Service (VIDS). A research pharmacist at VIDS maintained the randomization schedule and was the only person who was aware of the actual interventions. The primary aim of the PPCCT (parent trial) was to examine the effects of Mg supplementation and the interaction with the Mg-TRPM7 genotype on the expression of biomarkers in the colorectum. The current study, based on a funded R01 project, was separately planned in parallel to the parent study with its own primary aims. At the time of planning, the sample size (n = 240 for those who completed the parent trial and donated biospecimens before and after the treatment) was already determined based on the design of the parent study. Thus, the focus of the current study was to estimate the effect size as stated in the primary aims for the funded R01 project on the gut microbiome. It is implicitly understood that, if the effect size is larger than that used for sample size planning in the parent study, the current study will have greater statistical power than the parent study; otherwise, the power would be lower. Two hundred forty participants provided samples and 239 participants completed the study in its entirety. After quality control of the microbiome data at baseline and at the end of the trial, 14 participants were excluded, leaving 226 with valid microbiome data for analysis in the current study (Figure 1).

FIGURE 1.

Study population flow diagram (CONSORT). The study population in the Personalized Prevention of Colorectal Cancer Trial: 250 participants were randomly assigned and stared the intervention. Finally, 239 participants completed the study, 240 provided study samples at week 0 and week 12, and 226 had microbiome data suitable for analysis. Mg, magnesium; PPCCT, Personalized Prevention of Colorectal Cancer Trial.

Sample collection

Three different biospecimens were collected to measure the gut microbiome. The decision was made to collect stool, rectal mucosa, and rectal swab samples to best represent luminal and adherent microbiota populations, as it has previously been reported that microbial composition varies across sample types, reflecting the varying microhabitats within the digestive tract [32,33]. Stool was collected ≤3 d prior to a study visit by study participants at home using a white plastic collection container covering the toilet bowl, aliquoted by the participant into sterile cryovials, and stored in a home freezer until transport with an ice pack to the study visit. Rectal swabs and mucosal tissues were collected by the study physician at the study visits. Rectal swabs were collected by inserting a culturette swab through the anal canal, swabbing the distal rectal mucosa, and placing the swab into a cryovial. Rectal mucosal samples were collected through an anoscope using standard mucosal biopsy forceps, and these samples were placed into separate storage vials. All 3 biospecimen types were frozen at − 80 °C until use.

Blood samples were collected at each study visit, with participants having fasted ≥8 h. Serum and plasma were rapidly cooled and frozen at –80°C. 25-Hydroxyvitamin D₂ [25(OH)D₂] and 25-hydroxyvitamin D₃ [25(OH)D₃] were extracted from plasma via liquid–liquid extraction. Then, liquid chromatography–mass spectrometry analysis was conducted.

DNA was extracted from buffy coat fractions or cheek cells from mouthwash samples using a QIAamp DNA mini-kit (Qiagen Inc.). TRPM7 polymorphism was evaluated with the TaqMan genotyping assay (Assay ID: C_25756319_10; Applied Biosystems).

Covariates

Information on sociodemographic factors and lifestyle factors at baseline was obtained via questionnaire-based interviews and examination. Sociodemographic factors included age, sex, race/ethnicity, and education level. Race was categorized as White or Black. Education level was separated into 2 levels: less than high school or high school or equivalent, and college or above. Lifestyle factors included consumption of cigarettes and alcohol, physical activity level, and daily intakes of nutrients and foods. Smoking status was classified as never smoker, former smoker, or current smoker. Current alcohol use was defined as ≥1 drink a week over the past 12 mo. Former drinkers did not meet the criterion for current use in the past 12 mo or longer. Physically active was defined as nonoccupational exercise for ≥30 min/d for ≥2 d in the past 7 d. The mean recorded intake from the 2 24-h dietary recalls was used to estimate the baseline daily intake of total energy (kilocalories). Weight was taken on a digital scale and measured in kilograms, and standing height was measured in centimeters with a fixed stadiometer with a vertical backboard and a movable headboard at least twice at each clinic visit. BMI was calculated as weight (kilograms)/height (meters squared).

Whole genome shotgun (WGS) sequencing of stool, swab, and rectal biopsies

Samples were transferred to a 2 mL tube containing 200 mg ≤106 μm glass beads (Sigma) and 0.3 mL of Qiagen ATL buffer (Qiagen), supplemented with lysozyme (20 mg/mL) (Thermo Fisher Scientific). The suspension was incubated at 37°C for 1 h with occasional agitation and then supplemented with 600 IU of proteinase K and incubated at 60°C for 1 h. Finally, 0.3 mL of Qiagen AL (Qiagen) buffer was added and incubated at 70°C for 10 min, followed with 3 min bead beating in a Qiagen TissueLyser II (Qiagen) at 30 Hz. After a brief centrifugation, supernatants were transferred to a new tube containing 0.3 mL of ethanol. DNA was purified using a standard on-column purification method with Qiagen buffers AW1 and AW2 (Qiagen) as washing agents and eluted in 10 mM Tris (pH 8.0).

WGS sequencing was performed as previously described [32,33]. 10 ng of genomic DNA was processed using the Illumina Nextera XT DNA Sample Preparation Kit (Illumina). Fragmented and tagged DNA was amplified using a limited-cycle PCR program. In this step, index 1 (i7) and index 2 (i5) were added between the downstream bPCR adaptor and the core sequencing library adaptor, as well as primer sequences required for cluster formation. The DNA library was purified using Agencourt AMPure XP Reagent (Beckman Coulter). The DNA library pool was loaded on the Illumina platform reagent cartridge and on the Illumina HiSeq instrument (Illumina). ZymoBIOMICS Microbial Community DNA Standard (Zymo Research Corporation, Cat#D6305) and a library blank composed of library preparation reagents alone were used as controls. Unlike conventional assays that may depend on specific quantification thresholds, whole genome sequencing generally produces extensive data capable of analysis without strict detection limits.

To validate the DNA extraction process, a known bacterial community, the ZymoBIOMICS Microbial Community Standard (Cat# D6300), and reagent-only blanks were processed alongside samples during DNA extraction and library preparation.

Additionally, library blanks containing only library preparation reagents were included to ensure procedural integrity.

Metachronous colorectal polyp

For the study participants who provided consent for future studies, we reviewed their follow-up colonoscopy results via electronic medical records at VUMC from 2017 to 2018. Information on the colonoscopy or surgery, the diagnosis from the endoscopic mucosal resection procedure, and pathology reports were abstracted. Among 236 participants with a history of colorectal polyps, 124 underwent follow-up colonoscopies after completing the trial with a 3.5-y median follow-up time. The participants were classified as metachronous polyp cases or free of polyps (controls), based on their follow-up colonoscopy and pathology reports. The metachronous cases were further subdivided into metachronous adenomas (tubular, tubulovillous, or villous adenomas) and metachronous serrated polyps (sessile serrated lesions or hyperplastic polyps). The data were used for the exploratory analysis.

Analysis of shotgun metagenome sequencing

The metagenome sequences were processed with KneadData for quality control and removal of human DNA contamination. MetaPhlAn2 was used for analyzing the taxonomic profiles of shotgun metagenome sequences [34], whereas HUMAnN2 was used for functional profiling including the abundance of functional pathways [35]. Both taxonomy-stratified and unstratified pathway abundance were generated. In the stratified abundance table, the abundance of each pathway was shown as abundance in each taxon, whereas in the unstratified abundance table, the abundance of each pathway was not separated by taxa. The abundance of taxa and pathways was normalized as below to correct for different sequencing depth [32,33].

$$N_{ij\_\text{norm}}=\log 10\left(\frac{N_{ij}}{N_{\text{sample}\_j}}\ast N_{\text{avg}}+1\right)$$

N_ij is the raw abundance of taxa or pathway i in sample j. N_{sample_j} is the total number of sequences in sample j. N_avg is the average number of reads per sample. N_{ij_norm} is the normalized abundance of taxa or pathway i in sample j.

Statistical analyses

Baseline characteristics for continuous variables (mean ± SD) and categorical variables (count and percent) were reported (Table 1). To assess the effect of personalized Mg treatment on changes in the abundance of F. prausnitzii or C. maltaromaticum, generalized linear models were used. On the basis of the PPCCT’s original study design, the main effect and the effect stratified by the TRPM7 genotype were considered as primary analyses. Other stratified analyses were secondary. Unadjusted models and models adjusting for age, sex, race, and baseline levels were evaluated. Stratified analyses by TRPM7 genotype (GG, GA) and/or sex (male or female) were conducted. Mediation analyses were conducted to examine whether the effect of Mg treatment on circulating vitamin D metabolites was mediated by C. maltaromaticum and F. prausnitzii. In the exploratory analysis, to study the associations between the abundance of microbiota and risk of metachronous colorectal adenoma/serrated polyp, logistic regression models were employed to estimate the odds ratios (ORs) and 95% confidence intervals (CIs) without adjusting (model 1) or adjusting for age, sex, and BMI (model 2). The abundances of F. prausnitzii or C. maltaromaticum were categorized into tertiles based on the distribution of the controls in all analyses. Tests for trends were performed in the models using the lowest tertile level as the reference group and considering the tertile as a continuous variable in the model.

TABLE 1.

Descriptive characteristics of the study population at baseline.

	Magnesium (N = 120)	Placebo (N = 120)
Age	60.2 ± 7.8	61.2 ± 8.1
Sex: males (%)	65 (54)	62 (52)
Race: White (%)	118 (98.3)	119 (99.2)
Race: African American (%)	2 (1.7)	1 (0.8)
BMI	29.9 ± 6.1	30.6 ± 6.6
Education: college and above	108 (90.0)	108 (90.0)
TRPM7 genotype (%)
GG	77 (64)	75 (62)
GA	43 (36)	45 (38)
Smoking status (%)
Never	59 (49)	72 (60)
Former	51 (42)	40 (33)
Current	10 (8)	8 (7)
Drinking status (%)
Never	39 (32)	51 (42)
Former	25 (21)	24 (20)
Current	56 (47)	45 (38)
Physically active ≥2 d/wk (%)	93 (77)	102 (85)
Less than a college education (%)	12 (10)	12 (10)
Total energy intake (kcal/d)	2084 ± 547	2108 ± 604
Total magnesium intake (mg/d)	364 ± 97	338 ± 99
Total vitamin D intake (μg/d)	22 ± 21	42 ± 162
Ca:Mg ratio at baseline	3.7 ± 0.9	3.9 ± 1.5
Ca:Mg ratio during 12-wk trial	2.1 ± 0.7	3.5 ± 1.3
Plasma vitamin D level
25(OH)D (ng/mL)	33.4 ± 10.2	33.0 ± 12.7
25(OH)D3 (ng/mL)	32.3 ± 10.4	30.1 ± 11.3
25(OH)D2 (ng/mL)	1.1 ± 2.3	2.8 ± 10.9
24,25(OH)2D3 (ng/mL)	4.7 ± 3.7	3.8 ± 2.8
Baseline relative abundance
C. maltaromaticum
Rectal mucosa	0.101 ± 0.200	0.139 ± 0.220
Rectal swab	0.959 ± 0.537	1.021 ± 0.443
Stool	1.185 ± 0.448	1.204 ± 0.411
F. prausnitzii
Rectal mucosa	3.441 ± 0.400	3.448 ± 0.330
Rectal swab	4.954 ± 0.783	4.991 ± 0.702
Stool	5.384 ± 0.529	5.328 ± 0.573

Continuous variables are presented as mean ± SD.

Abbreviations: 25(OH)D, 25-hydroxyvitamin D; 25(OH)D₃, 25-hydroxyvitamin D₃; 25(OH)D₂; 25-Hydroxyvitamin D₂; 24,25(OH)₂D₃, 24,25-dihydroxyvitamin D3; Ca, calcium; C. maltaromaticum, Carnobacterium maltaromaticum; F. prausnitzii, Faecalibacterium prausnitzii; Mg, magnesium; TRPM7, transient receptor potential cation channel, subfamily M, member 7.

All P values are 2 sided and statistical significance was determined using an α level of 0.05. To account for multiple comparisons, false discovery rate (FDR) correction was applied to primary analyses in TABLE 2, TABLE 3, respectively, comparing the efficacy of Mg treatment on every bacterium across different sample types and different genotypes. Statistical significance for FDR correction was determined using a Q level of 0.1, given the exploratory nature of the investigation. The data analyses were performed with software SAS Enterprise Guide 7.1 (version 9.4; SAS Institute).

TABLE 2.

Changes in C. maltaromaticum relative abundance by magnesium compared with placebo in rectal mucosa, swab, and stool.

Participant category and sample type	Change in C. maltaromaticum relative abundance from baseline						FDR correction (Q)
	Mg treatment		Placebo		P1	P2
	Relative abundance (mean ± SD)	%	Relative abundance (mean ± SD)	%
All participants (n = 226)
Rectal mucosa	0.001 ± 0.268	0.64	–0.010 ± 0.294	–7.04	0.78	0.50	0.56
Rectal swab	0.110 ± 0.592	11.49	0.013 ± 0.562	1.24	0.23	0.31	0.56
Stool	0.051 ± 0.397	4.32	0.009 ± 0.459	0.73	0.46	0.46	0.56
TRPM7 genotype: GG (n = 143)
Rectal mucosa	0.007 ± 0.249	7.92	–0.027 ± 0.286	–19.30	0.44	0.93	0.93
Rectal swab	0.217 ± 0.615	23.01	–0.065 ± 0.588	–6.30	0.009	0.006	0.05
Stool	0.102 ± 0.416	8.83	–0.018 ± 0.446	–1.48	0.11	0.12	0.36
TRPM7 genotype: GA (n = 83)
Rectal mucosa	–0.011 ± 0.301	–9.41	0.021 ± 0.308	15.53	0.64	0.36	0.56
Rectal swab	–0.065 ± 0.511	–6.54	0.133 ± 0.503	13.30	0.09	0.04	0.18
Stool	–0.042 ± 0.346	–3.33	0.050 ± 0.481	4.02	0.33	0.48	0.56

Generalized linear models were used: P1 not adjusted; P2 adjusted for age, sex, BMI, and baseline level.

P for interaction between treatment and genotype of TRPM7 for C. maltaromaticum: 0.55 in mucosa; 0.001 in swab; and 0.11 in stool.

Abbreviations: C. maltaromaticum, Carnobacterium maltaromaticum; FDR, false discovery rate; Mg, magnesium; TRPM7, transient receptor potential cation channel, subfamily M, member 7.

TABLE 3.

Changes in F. prausnitzii relative abundance by magnesium compared with placebo in rectal mucosa, swab, and stool.

Participant category and sample type	Change in F. prausnitzii relative abundance from baseline						FDR correction (Q)
	Mg treatment		Placebo		P1	P2
	(mean ± SD)	%	(mean ± SD)	%
All participants (n = 226)
Rectal mucosa	–0.017 ± 0.392	–0.49	0.046 ± 0.425	1.34	0.25	0.16	0.36
Rectal swab	0.044 ± 0.787	0.89	–0.028 ± 0.752	–0.56	0.48	0.48	0.86
Stool	–0.073 ± 0.469	–1.35	–0.047 ± 0.558	–1.88	0.71	0.91	0.91
TRPM7 genotype: GG (n = 143)
Rectal mucosa	0.006 ± 0.315	0.18	–0.012 ± 0.369	–0.34	0.75	0.79	0.89
Rectal swab	0.105 ± 0.817	2.13	–0.095 ± 0.856	–1.92	0.15	0.04	0.18
Stool	–0.082 ± 0.518	–1.52	–0.051 ± 0.565	–0.98	0.73	0.77	0.89
TRPM7 genotype: GA (n = 83)
Rectal mucosa	–0.058 ± 0.503	–1.66	0.145 ± 0.495	4.17	0.07	0.01	0.09
Rectal swab	–0.060 ± 0.730	–1.20	0.087 ± 0.516	1.72	0.28	0.12	0.36
Stool	–0.057 ± 0.375	–1.06	–0.040 ± 0.554	–0.74	0.87	0.67	0.89

Generalized linear models were used: P1 not adjusted; P2 adjusted for age, sex, BMI, and baseline level.

P for interaction between treatment and genotype of TRPM7 for F. prausnitzii: 0.01 in mucosa; 0.02 in swab; and 0.61 in stool.

Abbreviations: FDR, false discovery rate; F. prausnitzii, Faecalibacterium prausnitzii; Mg, magnesium; TRPM7, transient receptor potential cation channel, subfamily M, member 7.

Results

The current analysis includes 226 participants. The participants in the Mg treatment arm did not differ significantly compared with those in the placebo arm in terms of age, sex, race, BMI, TRPM7 genotype (GG or GA), smoking status, drinking status, physical activity, education level, and daily total energy intake (Table 1). The average ages were 60.2 and 61.2 y in the treatment and placebo arms, respectively. Of 108 female participants, 40.7% (n = 44) carried the GA genotype. Of 118 male participants, 33.0% (n = 39) carried the GA genotype.

Baseline relative abundances of C. maltaromaticum and F. prausnitzii for treatment and placebo groups are shown in Table 1. In rectal swab samples at baseline (Supplemental Table 1), females had a significantly higher relative abundance of C. maltaromaticum (1.103 ± 0.489, mean ± SD) compared with males (0.885 ± 0.471) (P = 0.001 for difference). There were no significant differences in baseline relative abundances of C. maltaromaticum or F. prausnitzii between males and females in rectal mucosa and stool samples. As primary analyses, pre- to posttreatment changes in C. maltaromaticum relative abundance are shown in Table 2. Overall, the changes in C. maltaromaticum abundance did not significantly differ between the Mg treatment and the placebo groups for any sample type. However, when stratified by TRPM7 genotype, those with the GG genotype (i.e., without the missense variant) in the treatment arm had a significant increase in C. maltaromaticum abundance compared with placebo in rectal swab samples [0.217 ± 0.615 (23.01%) compared with –0.065 ± 0.588 (–6.30%); adjusted P = 0.006], whereas those with the TRPM7 missense variant (GA) had a significant decrease in C. maltaromaticum abundance in the treatment arm compared with placebo in rectal swab samples [–0.065 ± 0.511 (–6.54%) compared with 0.133 ± 0.503 (13.30%); adjusted P = 0.04]. After FDR correction, the increased effect of personalized Mg treatment in those with the TRPM7 GG genotype remained significant (Q = 0.05), whereas the observed decrease in those with the TRPM7 GA genotype did not remain significant (Q = 0.18). The test for the interaction between the treatment and the TRPM7 genotype was statistically significant (P = 0.001) and remained significant in females (P = 0.01). As secondary analyses, changes in C. maltaromaticum relative abundance stratified by sex alone and sex and TRPM7 genotype are shown in Supplemental Table 2. When stratified by sex alone, there were no significant changes in C. maltaromaticum compared with placebo. However, when stratified by both sex and TRPM7 genotype, females with the GG genotype who had received Mg treatment had a significantly higher increase in abundance of C. maltaromaticum compared with placebo in rectal swab samples [0.239 ± 0.624 (24.06%) compared with –0.177 ± 0.588 (–14.70%); adjusted P = 0.02], and stool samples [0.152 ± 0.406 (13.52%) compared with –0.109 ± 0.513 (–9.27%); adjusted P = 0.04]. There were no significant changes in C. maltaromaticum among females with the TRPM7 GA or males with the GG and the GA genotypes when comparing Mg treatment compared with placebo groups. No significant effect was identified in the analysis of rectal mucosa.

As primary analyses, the changes in abundance of F. prausnitzii did not change significantly in the Mg treatment group compared with the placebo group (Table 3). In a stratified analysis by TRPM7 genotype, participants with the GG genotype in the Mg treatment group had a significant increase in F. prausnitzii abundance in rectal swab samples compared with the placebo group [0.105 ± 0.817 (2.13%) compared with –0.095 ± 0.856 (–1.92%); adjusted P = 0.04], whereas participants with the GA genotype in the Mg treatment group tended to have a decrease in F. prausnitzii abundance in rectal swab samples compared with the placebo group [–0.060 ± 0.730 (–1.20%) compared with –0.087 ± 0.516 (1.72%); borderline significance with adjusted P = 0.12]. Notably, the observed increase in F. prausnitzii abundance in rectal swab samples did not remain significant after FDR correction was applied (Q = 0.18). The test for the interaction between the treatment and the TRPM7 genotype was statistically significant (P = 0.02) in the analysis of rectal swabs. The participants with the GA genotype in the Mg treatment group had a significant decrease in F. prausnitzii abundance in rectal mucosa compared with the placebo group [–0.058 ± 0.503 (–1.66%) compared with 0.145 ± 0.495 (4.17%); adjusted P = 0.01, Q = 0.09] whereas those with the GG genotype in the Mg treatment arm did not have an apparent change in F. prausnitzii abundance in rectal mucosa compared with the placebo group [0.006 ± 0.315 (0.18%) compared with –0.012 ± 0.369 (–0.34%); adjusted P = 0.79]. The test for the interaction between the treatment and the TRPM7 genotype was statistically significant (P = 0.01) in the analysis of rectal mucosa. As secondary analyses, no significant changes in F. prausnitzii abundance were observed in the Mg treatment compared with the placebo groups when stratified by sex alone, or TRPM7 genotype combined with sex (Supplemental Table 3). No significant effect was identified in the analysis of stool samples.

A mediation analysis was performed to investigate whether the changes in abundance of C. maltaromaticum or F. prausnitzii mediate the effect of Mg treatment on alterations in circulating levels of vitamin D. No significant mediation effect was identified. Additional stratified analyses by baseline and changes of 25(OH)D levels showed no significant microbe-mediated effects of Mg treatment on vitamin D metabolites. We also did not find any relationship between plasma 25(OH)D metabolites and changes in abundance in C. maltaromaticum and F. prausnitzii (data not shown).

An exploratory data analysis was also performed to examine the associations between levels of C. maltaromaticum and F. prausnitzii and risk of polyp development. In rectal mucosa, a higher abundance of F. prausnitzii was significantly associated with an almost 3-fold increased risk of developing metachronous polyps [OR = 2.84 (1.05, 7.65), adjusted P = 0.04]. Meanwhile, in rectal swabs, higher levels of C. maltaromaticum were marginally associated with a decreased risk of developing serrated polyps [OR = 0.15 (0.02, 1.38), adjusted P = 0.10]. In all other sample types, C. maltaromaticum and F. prausnitzii abundance were not associated with risk of polyp development (Table 4).

TABLE 4.

Odds ratios (ORs) and 95% CIs for risk of metachronous colorectal polyps by relative abundance of bacteria in mucosal tissue and swab, results from the PPCCT.

Bacteria	Cases/controls	Model	Tertile relative abundance			P trend
			Ref	OR (95% CI)	OR (95% CI) (high)
Rectal mucosa
Metachronous polyp
C. maltaromaticum	68/53	Model 1	1.00	0.71 (0.32, 1.60)		0.41
C. maltaromaticum	68/53	Model 2	1.00	0.83 (0.35, 1.96)		0.66
F. prausnitzii	68/53	Model 1	1.00	2.00 (0.77, 5.18)	2.82 (1.11, 7.21)	0.03
F. prausnitzii	68/53	Model 2	1.00	1.71 (0.62, 4.71)	2.84 (1.05, 7.65)	0.04
Metachronous adenoma
C. maltaromaticum	41/53	Model 1	1.00	0.56 (0.21, 1.48)		0.24
C. maltaromaticum	41/53	Model 2	1.00	0.61 (0.21, 1.73)		0.35
F. prausnitzii	41/53	Model 1	1.00	2.11 (0.76, 5.90)	1.53 (0.52, 4.49)	0.47
F. prausnitzii	41/53	Model 2	1.00	1.88 (0.62, 5.64)	1.38 (0.42, 4.46)	0.62
Serrated polyp
C. maltaromaticum	14/53	Model 1	1.00	0.93 (0.25, 3.39)		0.91
C. maltaromaticum	14/53	Model 2	1.00	0.97 (0.26, 3.67)		0.96
F. prausnitzii	14/53	Model 1	1.00	2.50 (0.43, 14.60)	3.71 (0.67, 20.40)	0.13
F. prausnitzii	14/53	Model 2	1.00	2.11 (0.34, 12.96)	3.41 (0.60, 19.50)	0.16
Rectal swab
Metachronous polyp
C. maltaromaticum	69/53	Model 1	1.00	0.77 (0.33, 1.79)	0.56 (0.23, 1.39)	0.21
C. maltaromaticum	69/53	Model 2	1.00	0.75 (0.30, 1.85)	0.68 (0.26, 1.79)	0.42
F. prausnitzii	69/53	Model 1	1.00	0.71 (0.29, 1.74)	1.24 (0.52, 2.91)	0.62
F. prausnitzii	69/53	Model 2	1.00	0.70 (0.26, 1.86)	1.21 (0.48, 3.02)	0.67
Metachronous adenoma
C. maltaromaticum	42/53	Model 1	1.00	0.63 (0.24, 1.67)	0.61 (0.23, 1.66)	0.31
C. maltaromaticum	42/53	Model 2	1.00	0.65 (0.23, 1.87)	0.78 (0.26, 2.35)	0.61
F. prausnitzii	42/53	Model 1	1.00	0.56 (0.20, 1.60)	1.12 (0.43, 2.91)	0.81
F. prausnitzii	42/53	Model 2	1.00	0.63 (0.20, 2.00)	1.31 (0.46, 3.75)	0.61
Serrated polyp
C. maltaromaticum	14/53	Model 1	1.00	0.86 (0.24, 3.06)	0.15 (0.02, 1.36)	0.09
C. maltaromaticum	14/53	Model 2	1.00	0.83 (0.20, 3.38)	0.15 (0.02, 1.38)	0.10
F. prausnitzii	14/53	Model 1	1.00	0.80 (0.18, 3.47)	1.06 (0.26, 4.32)	0.94
F. prausnitzii	14/53	Model 2	1.00	0.84 (0.18, 3.94)	0.95 (0.22, 4.03)	0.94
Stool
Metachronous polyp
C. maltaromaticum	68/50	Model 1	1.00	1.02 (0.41, 2.54)	1.75 (0.73, 4.22)	0.21
C. maltaromaticum	68/50	Model 2	1.00	0.81 (0.30, 2.17)	1.56 (0.62, 3.97)	0.35
F. prausnitzii	68/50	Model 1	1.00	0.91 (0.37, 2.25)	1.26 (0.52, 3.05)	0.61
F. prausnitzii	68/50	Model 2	1.00	0.78 (0.29, 2.08)	1.27 (0.49, 3.30)	0.60
Metachronous adenoma
C. maltaromaticum	41/50	Model 1	1.00	0.85 (0.30, 2.42)	1.54 (0.58, 4.09)	0.40
C. maltaromaticum	41/50	Model 2	1.00	0.66 (0.21, 2.09)	1.42 (0.49, 4.12)	0.54
F. prausnitzii	41/50	Model 1	1.00	0.93 (0.34, 2.55)	1.06 (0.39, 2.91)	0.91
F. prausnitzii	41/50	Model 2	1.00	0.75 (0.25, 2.29)	1.23 (0.41, 3.72)	0.73
Serrated polyp
C. maltaromaticum	14/50	Model 1	1.00	0.34 (0.06, 1.87)	0.90 (0.24, 3.43)	0.81
C. maltaromaticum	14/50	Model 2	1.00	0.28 (0.05, 1.67)	0.74 (0.18, 2.98)	0.63
F. prausnitzii	14/50	Model 1	1.00	0.40 (0.07, 2.35)	1.49 (0.39, 5.65)	0.52
F. prausnitzii	14/50	Model 2	1.00	0.30 (0.05, 2.04)	1.32 (0.30, 5.80)	0.57

Unconditional logistic regression models, model 1, unadjusted; model 2, adjusted for age (continuous), sex, and BMI (continuous).

Metachronous polyps are further classified into metachronous adenomas or metachronous serrated polyps.

Abbreviations: C. maltaromaticum, Carnobacterium maltaromaticum; CI, confidence interval; F. prausnitzii, Faecalibacterium prausnitzii; PPCCT, Personalized Prevention of Colorectal Cancer Trial; ref, reference.

Discussion

Our findings demonstrate that Mg treatment increases the abundance of C. maltaromaticum and likely F. prausnitzii in participants without the TRPM7 missense variant. Li et al. [5] noted increased intestinal vitamin D production in C. maltaromaticum-treated mice. In additional experiments, C. maltaromaticum led to the increased levels of 7-dehydrocholesterol whereas F. prausnitzii converted 7-dehydrocholesterol to 25(OH)D₃ and 1,25(OH)₂D₃ and, in turn, decreased CRC incidence. Our findings in this RCT suggest that Mg treatment increased both C. maltaromaticum and F. prausnitzii, which can convert vitamin D precursors to 25(OH)D₃ and 1,25(OH)2D3 in the gut [5].

We previously demonstrated that Mg treatment increases 25(OH)D₃ when the baseline 25(OH)D level is lower [15]. The mechanism by which this occurs is likely multifactorial. Here, C. maltaromaticum and F. prausnitzii were not mediators of Mg treatment’s effect on circulating vitamin D levels, suggesting that Mg increases gut microbial abundance and vitamin D levels through independent mechanisms. Thus, Mg supplementation likely increases vitamin D synthesis and metabolism enzymes in the body [15]. Although we cannot eliminate the possibility that statistical power was insufficient to observe an association, we found that Mg treatment significantly increases circulating levels of 25(OH)D₃ and C. maltaromaticum and likely F. prausnitzii (borderline significance) in rectal swabs compared with placebo. Thus, based on Li et al.’s [5] findings, we postulate that local de novo synthesis of gut vitamin D caused by increased C. maltaromaticum and F. prausnitzii may only lead to a local effect, including inhibition of colorectal carcinogenesis.

The observed effect of Mg treatment on these microbes is biologically plausible. Mg is essential for many cellular functions in gut microbiota [20,21] and is maintained at high levels (∼0.5–2.0 mM) in microbiota, unlike most divalent cations [21]. Several bacterial metabolic enzymes are Mg-dependent [[22], [23], [24], [25], [26],36,37], and Mg concentration in culture media is essential for bacterial growth [[38], [39], [40]]. Meanwhile, TRPM7 is a key ion channel that regulates Mg homeostasis [41,42]. An in vitro study discovered that the heterologously expressed Thr1482Ile missense variant in the TRPM7 gene caused an elevated sensitivity to inhibition by intracellular Mg²⁺ [43]. We observed that Mg treatment only significantly increased the abundance of C. maltaromaticum and F. prausnitzii in rectal swabs for those without the missense TRPM7 genotype. Meanwhile, Mg treatment significantly reduced the abundance of C. maltaromaticum in those with the missense variant. Taken together, these findings suggest that Mg treatment in the context of a functional polymorphism in the TRPM7 gene is key to these microbes’ ability to locally synthesize vitamin D in the gut and may inhibit colorectal carcinogenesis [5].

In our exploratory analysis investigating risk of polyp development, we found that a higher abundance of C. maltaromaticum in rectal swabs was marginally associated with a reduced risk of developing serrated polyps, whereas a higher abundance of F. prausnitzii in rectal mucosa was significantly associated with an ∼3-fold increased risk of developing metachronous polyps. It has been found that a higher abundance of F. prausnitzii in rectal mucosa indicates a loss of gut integrity, which is linked to increased levels of inflammation and progression of colorectal polyps [44]. In our study, Mg treatment reduced F. prausnitzii in the rectal mucosa in those with the missense variant. These findings suggest that Mg treatment in missense TRPM7 individuals may decrease CRC risk and could explain the previous finding that higher intakes of Mg were associated with a reduced risk of colorectal polyps among those with the Thr1482Ile missense variant in the TRPM7 gene [30].

Li et al. [5] reported that C. maltaromaticum was higher in abundance in females compared with males in an in-house cohort and in a public inflammatory bowel disease dataset, and further, C. maltaromaticum was depleted in females with CRC [6]. Furthermore, the finding that C. maltaromaticum and F. prausnitzii could synthesize vitamin D and inhibit colorectal carcinogenesis was shown in female mice. Consistent with this, we found that the abundance of C. maltaromaticum was higher in females than in males at baseline. Mg supplementation did not have a significant effect for males, regardless of TRPM7 genotype, suggesting a sex effect leading to the increased bacterial abundance observed in females. This observed difference could be explained by the known role of estrogen in shifting Mg from circulation into cells, leading to increased cellular uptake in females [45]. Thus, estrogen may play a mediating role in Mg supplementation and its downstream effects.

Interestingly, the only significant increases caused by Mg treatment were observed in rectal swabs in females. This may be due to differences in oxygenation of the sample environment, as the colon is known to have the most drastic oxygen gradient in the body [33]. Anoxia increases markedly from colorectal mucosa to rectal swabs to stool samples [33]. Li et al. [5] reported that C. maltaromaticum and F. prausnitzii operate synergistically to produce active vitamin D via fermentation. As such, to conduct fermentation in the gut, C. maltaromaticum and F. prausnitzii may be of higher abundance in low oxygenation environments (i.e., rectal swab and stool samples) and in lower abundance in higher oxygenation environments (i.e., colorectal mucosa).

This study has several strengths. It features a double-blind, precision-based RCT in which participants demonstrated strong adherence to the study protocol, and attrition rates were low. We collected a variety of samples that represented the gastrointestinal microbiome in different niches, giving a better representation of participants’ microbiomes [46]. Notably, our findings align with recent research, including a sex effect associated with increased C. maltaromaticum abundance [5]. Furthermore, the PPCCT was specifically designed to test the interaction between Mg treatment and TRPM7 genotype on gut biomarkers related to colorectal carcinogenesis. Our findings provide support for the TRPM7 genotype-Mg intake interaction affecting risk of colorectal polyp development [30].

This study also has some limitations. Although the benefits of an increase in C. maltaromaticum and F. prausnitzii relative abundances are hypothesized to increase de novo synthesize vitamin D and reduce colorectal carcinogenesis, we did not specifically investigate the downstream effects of increased C. maltaromaticum and F. prausnitzii abundance. Also, although our adjusted P values showed significant increases in both C. maltaromaticum and F. prausnitzii, the significant effect did not remain for F. prausnitzii after FDR correction. Specific strains of C. maltaromaticum and F. prausnitzii were not distinguished in our study, warranting future investigation into Mg’s effects on vitamin D production among specific strains. Additionally, given that changes in microbe abundance were quantified in relative abundances, changes could have occurred due to other bacteria not studied. Finally, our study population is limited in race, ethnicity, and geographic diversity, limiting the generalizability of our findings.

In conclusion (Figure 2), we found that Mg supplementation increases the abundance of C. maltaromaticum and F. prausnitzii in individuals with adequate TRPM7 function, specifically in females. In those with a missense variant in the TRPM7 gene, Mg treatment reduces the abundance of F. prausnitzii in colorectal mucosa. Li et al. [5] previously revealed that F. prausnitzii converted vitamin D precursors produced by C. maltaromaticum into active metabolites and inhibited CRC development in females. Together, these findings lay a foundation for a precision-based strategy for the prevention of CRC in high-risk populations.

FIGURE 2.

Mechanistic figure: magnesium supplementation increases levels of C. maltaromaticum and F. prausnitzii. TRPM7, transient receptor potential cation channel, subfamily M, member 7.

Author contributions

The authors’ responsibilities were as follows – MJS, CY, QD: designed research; XZ, MJS, RN, HJM, CY, QD: conducted the research; MAA-P: conducted the assay; ES, XZ, SS: analyzed data; ES, QD: drafted the manuscript; and all authors: contributed to the data interpretation and manuscript revision and approved the final version of this manuscript.

Data availability

Data described in the manuscript, code book, and analytic code will be made available on request.

Funding

The authors reported no funding received for this study.

Conflict of interest

The authors report no conflicts of interest.

Appendix A. Supplementary data

The following is the Supplementary data to this article: