Cooked Broccoli Alters Cecal Microbiota and Impacts Microbial Metabolism of Glucoraphanin in Lean and Obese Mice

ABSTRACT

Scope

Brassica vegetables contain unique compounds known as glucosinolates (GSLs), which, when hydrolyzed by plant or microbial myrosinase, form bioactive isothiocyanates (ITCs) that offer health benefits to the host. The present study evaluated the impact of cooked broccoli (broccoli myrosinase inactivated) consumption on cecal microbial metabolism of glucoraphanin (GRP) in lean and obese mice and characterized the changes in cecal microbiota following broccoli‐containing diets.

Methods and results

Twenty lean and 20 diet‐induced obese (DIO) mice were randomized to consume control or cooked broccoli supplemented diets for 7 days. Cooked broccoli consumption increased ex vivo microbial GRP hydrolysis by cecal contents collected from lean and obese mice, led to increased production of sulforaphane (SF), sulforaphane‐cysteine (SF‐CYS), total ITC, and colonic NAD(P)H: Quinone Oxidoreductase (NQO1) activity. Further investigation revealed increased abundance of health‐promoting gut microbiota, including Lachnospiraceae NK4A136 group and Dubosiella newyorkensis, following broccoli‐containing diets. The Peptococcaseae family, the Blautia genus, and an amplicon sequence variation (ASV) from the Oscillospiraceae family exhibited negative correlation with total ITC production.

Conclusion

These finding suggest that cooked broccoli consumption enhances microbial GRP hydrolysis to produce more bioactive ITCs and inform future strategies toward altering microbial GSL metabolism to promote gut health in both lean and obese individuals.

Seven‐day cooked broccoli feeding increased the ability of the gut microbiota to metabolize glucoraphanin (GRP) into beneficial isothiocyanates (ITCs) leading to increased colonic NQO1 activity. Gut microbiota shifted, with Lachnospiraceae NK4A136 group and Dubosiella newyorkensis increasing on broccoli diet. Further, the abundance of Peptococcaceae, Blautia, and an Oscillospiraceae ASV negatively correlated with ITC production.

1. Introduction

Broccoli (Brassica oleracea L.), belonging to the brassica vegetable family, is a good source of vitamin C, β‐carotene, calcium, and fiber. Yet there is a unique group of compounds in brassica vegetables with antioxidant and anticancer activities–a family of sulfur‐containing phytochemicals called glucosinolates (GSLs) [1]. GSLs themselves are not bioactive, at least at the concentration found in brassica, but the key hydrolysis product, the isothiocyanates (ITCs), are highly bioactive [2]. Among the GSLs in broccoli, glucoraphanin (GRP) is the most abundant; other notable GSLs include glucoiberin, glucobrassicin, and glucoerucin, with content varying by cultivars and developmental stages of broccoli [3, 4]. The ITC metabolites of these GSLs, such as sulforaphane (SF), iberin, indole‐3‐carbinol, and erucin, have been extensively studied for their health benefits. These compounds are associated with reducing the risk of various cancers, mitigating inflammation, preventing degenerative diseases like Alzheimer's, and lowering the incidence of cardiovascular diseases [5]. In the plant, the formation of ITCs requires an enzyme called myrosinase (β‐thioglucoside glucohydrolase), which is stored in myrosin granules, separated from the GSLs. When there is tissue damage (i.e., cutting, chewing, and insects biting), myrosinase gains access to and hydrolyzes GSLs, removing glucose to form an ITC, indole‐3‐carbinol, nitrile (NIT), or epithionitrile, depending on the GSL side chain, pH, the presence of ferrous ions, and the presence of different specifier proteins [6]. Most people cook brassica vegetables before eating, which denatures the plant myrosinase, and the transformation of GSL to bioactive ITC is then dependent on those gut microbiota exhibiting myrosinase‐like activities [7, 8].

With only 5% of an oral dose absorbed intact, most GSLs consumed are hydrolyzed by plant and/or microbial myrosinase [9]. Once GSLs are converted to ITCs, they are absorbed by the enterocyte, conjugated to glutathione inside the cell, then enter circulation, and are further metabolized in the liver to form ITC conjugates of cysteine, cysteinylglycine, and glutathione [10]. A further ITC conjugate, N‐acetylcysteine, can be formed in the kidney [11]. Similarly, the microbial metabolism of GSLs is more complicated than just forming ITCs. There are multiple pathways by which microbiota metabolize GSL [12]. Instead of producing ITCs, other physiological inert compounds can also be formed, such as NIT and desulfo‐GSLs [12, 13]. The formation of these inert and nontoxic compounds is considered as a potential mechanism that bacteria use to avoid forming potentially toxic ITCs. Lactobacillus agilis R16 and Escherichia coli VL8, two human gut bacteria, have been shown mostly to hydrolyze GSLs to NIT (about 80% of the added GSLs) [14]. In addition, NIT production in the human gut from broccoli consumption was found to be influenced by gut microbiome composition [15]. The range in the extent of conversion of GSL to ITC, from less than 0.1% to more than 40% of the dose consumed, was observed to be similar in two populations (urban USA and rural China) given a GRP‐containing drink with no myrosinase [16]. Variations in the individual's gut microbiota may explain the interindividual variation in ITC formation and bioactivity observed in human studies [16, 17, 18, 19]. Understanding the alternative pathways of microbial GSL metabolism will inform the development of strategies improving the formation of bioactive ITCs, and thus increase the health benefits of consuming cooked brassica vegetables. Furthermore, it is unknown how frequent brassica consumption alters GSL metabolism by the gut microbiota. It is reasonable to suggest several competing affects: (1) increase in microbial NIT production as this liberates a glucose for the microbe without generating a toxic ITC; (2) death of microbial ITC producers that are sensitive to ITCs; and (3) increase in microbial resistance to ITCs, especially by ITC producers.

Our group previously showed that cooked broccoli consumption alters rat cecal microbiota to improve ITC production [20]. However, it is unclear whether this is also true in mice. Given the fact that mouse models have more available genetic data, including various knock‐out mice, confirming the ITC production‐promoting effect of cooked broccoli in mice will support future evaluation of the bioactivity of ITC and underlying mechanisms using different mouse models. Additionally, gut microbiota communities differ between lean and obese populations, with accumulating evidence suggesting that both metabolic health status and Western diet are associated with alterations in host gut microbiota [21, 22]. Studies have shown that a high‐fat diet (HFD) can disrupt gut microbiota composition by increasing Desulfovibrionaceae while reducing Lactobacillus and Bifidobacterium [23, 24, 25]. Some studies have observed an increased risk of intestinal pathogen colonization in mice fed with an HFD [23, 24]. These microbial shifts could potentially impair the microbiota's ability to hydrolyze GSLs into bioactive ITCs, limiting the health benefits of broccoli. Given these microbial imbalances, a crucial question is whether cooked broccoli supplementation can modify the microbiota in obese hosts to produce beneficial ITCs and deliver health benefits comparable to those in lean hosts. Although several studies have shown that dietary broccoli consumption ameliorates obesity in rodent models [26], the impact of brassica consumption on microbial GSL metabolism in obese population is unknown. Since the microbial communities in lean and obese hosts are different, it is likely that the microbiota's response to cooked broccoli will vary between these groups. It is important to understand the microbial role in the GSL metabolic fate in obese populations and thus develop strategies to improve the bioavailability of GSL that benefits the host, with or without obesity.

In this study, we investigated the impact of cooked broccoli on the cecal microbial metabolism of GRP in lean and obese mice and determined colonic NQO1 activity in the animals. We also characterized compositional changes in cecal microbiota after cooked broccoli feeding and evaluated the correlation between changes in microbiota and microbial GRP metabolites.

2. Materials and Methods

2.1. Animals and Diets

All animal study procedures were approved by the Institutional Animal Care and Use Committee (IACUC) of University of Illinois (Approval Reference# 21261). Twenty male C57BL/6J DIO Control mice (lean group) at 16 weeks of age and 20 male C57BL/6J DIO mice (obese group) at 16 weeks of age (Jackson Laboratories, Inc., ME, USA) were housed singly and maintained in light/dark cycles of 12:12 h with food and water ad libitum. The C57BL/6J DIO mice have been fed with an HFD (D12492) since 6 weeks of age at the Jackson Laboratories and were purchased at 15 weeks old for the experiment. Commercially available precooked, frozen broccoli was microwaved 3 min, freeze‐dried, and ground to a powder before incorporation into the broccoli‐supplemented diet. Four diets were prepared for the study: a low‐fat diet (LFD, D12450J, 10 kcal% fat diet, Research Diets), an LFD with cooked broccoli (LFCB, Research Diet), an HFD (D12492, 60 kcal% fat diet, Research Diet), and an HFD with cooked broccoli (HFCB, Research Diet). All diets were formulated to match macronutrient and micronutrient content, amount broccoli was matched by calorie in LFCB and HFCB (Table 1).

TABLE 1.

Diet formulations.

Ingredients (g)	LFD	LFCB	HFD	HFCB
Casein	200	167.5	200	167.5
l‐Cystein	3	2.5	3	2.5
Corn starch	506.2	456	0	0
Maltodextrin	125	125	125	125
Sucrose	68.8	68.8	68.8	18.4
Cellulose	50	23.5	50	23.5
Soybean oil	25	25	25	25
Lard	20	16.2	245	241.2
Mineral Mix S10026	10	10	10	10
Dicalcium phosphate	13	13	13	13
Calcium carbonate	5.5	5.5	5.5	5.5
Potassium citrate, 1H2O	16.5	16.5	16.5	16.5
Vitamin Mix V10001	10	10	10	10
Choline bitartrate	2	2	2	2
Cooked broccoli powder	0	109	0	109
Calculated glucoraphanin (µmol/kcal diet)	0	0.097	0	0.097
Total (g)	1055.05	1050.50	773.85	769.10
kcal from nutrients
Protein	708	708	708	708
Carbohydrate	2840	2841	815	815
Fat	405	405	2430	2430
Fiber	50	50	50	50
Total kcal	3953	3953	3953	3952

2.2. Experimental Design

After 2 weeks of acclimation, animals in the lean group were randomly assigned to LFD or LFCB for 7 days (n = 10 for each diet). Animals in the obese group were randomly assigned to HFD or HFCB for 7 days (n = 10 for each diet). Water and diets were provided ad libitum. Body weight and food intake were monitored over the 7 days. On Day 7, animals were sacrificed by CO₂ asphyxiation. The cecum was isolated and ligated at the anterior and the distal ends, surgically removed, and immediately transferred to an anaerobic chamber to collect cecal contents for ex vivo metabolism of GRP (Cayman Chemical, Ann Abor, MI, USA) and microbiota sequencing. The colon was removed and rinsed with ice‐cold PBS. Colonic mucosa was scraped and processed immediately for measuring NQO1 activity.

2.3. Ex Vivo Metabolism of Glucoraphanin by Mouse Cecal Microbiota

Fresh cecal contents were processed in an anaerobic chamber and incubated with GRP for measuring microbial metabolites following our published method [20]. Briefly, about 50 mg cecal contents were weighted and diluted 1:25 with PBS, followed by mixing 1:1 (v/v) with reinforced clostridial medium (RCM; BD Difco, Franklin Lakes, NJ, USA). The bacterial slurry was then incubated with GRP solution (200 µM in water) or PBS control for 2 h at 37°C anaerobically. Samples were centrifuged for 1 min at room temperature, supernatant was collected and flash‐frozen in liquid nitrogen, and stored at −80°C until analysis.

2.4. Quantification of Ex Vivo GRP Microbial Metabolites

Metabolites from ex vivo GRP incubation with mouse cecal contents were extracted using ice‐cold 50% methanol in 0.1% formic acid water (250 µL culture added with 500 µL extraction solvent), vortexed at 4°C for 4 min, followed by centrifugation at 13 000 × g, 4°C, for 10 min. Supernatants were filtered (0.22 µM) and transferred to vials for LC‐MS injection. Standards of SF, sulforaphane cysteine (SF‐CYS), and sulforaphane nitrile (SF‐NIT) were purchased from LKT Labs (Saint Paul, MIN, USA) and dissolved in methanol. Standard stocks were stored at −20°C and diluted for calibration curve preparation.

Quantification of GRP microbial metabolites, including SF, SF‐NIT, and SF‐CYS, was carried out on a Waters LC‐MS system (Waters Synapt G2Si ESI/LC‐MS) following a method described by Beran et al. [12] with some modifications. Briefly, an HSS T3 column (Waters, 1.8 µm, 2.1 × 100 mm) was used for the separation of compounds with a mobile phase containing ultrapure water as Solvent A and acetonitrile as Solvent B at a flow rate of 0.4 mL/min, with the following gradient: 100% (v/v) A (0.5 min), 85% (v/v) A (2 min), 85%–15% (v/v) A (3 min), 15%–0% (v/v) A (4 min), 0%–100% (v/v) A (4.1 min), and 100% (v/v) A (6 min). Compounds were detected in positive ion mode at a mass range of m/z 50–1000 and quantified via external calibration curves. Data analysis was performed using masslynx software (v4.2).

Total ITC quantification was performed following the cyclocondensation method as we previously described [20]. Briefly, samples were incubated with potassium phosphate buffer (25 mM) and 1,2‐benzenedithiol (10 mM) in a water bath at 65°C, for 2 h. Samples were then centrifuged at 16 000 × g at room temperature for 10 min, and the supernatants were filtered (0.22 µm) before analysis. Samples were analyzed using a Waters HPLC‐DAD system connected to a C18 reverse‐phase column (ODS‐3, 5 µm, 250 × 4.6 mm). The solvent system was operated isocratically with 80% methanol and 20% water at a flow rate of 1.0 mL/min. 1,3‐Benzodithiole‐2‐thione, the cyclocondensation product, was detected by absorption at 365 nm. Quantification was performed using an external calibration curve prepared by reacting 1,2‐benzenedithiol with pure SF.

2.5. Colonic NAD(P)H: Quinone Oxidoreductase (NQO1) Activity

The colon tissue was rinsed with ice‐cold PBS immediately after collection. The colonic mucosa was scraped and homogenized with the extraction buffer included in the NQO1 activity assay kit (ab184867, Abcam, Cambridge, UK) with addition of phenylmethylsulfonyl fluoride (PMSF, 1 mM, Thermo Fisher Scientific, Waltham, MA, USA), followed by incubation on ice for 20 min and centrifugation at 18 000 × g for 20 min at 4°C. The supernatants were collected and stored at −80°C until analysis. Colonic NQO1 activity was measured with the kit per manufacturer's instructions.

2.6. Cecal Microbial Sequencing

Total DNA was extracted from 30 to 40 mg cecal contents using a QIAamp PowerFecal Pro DNA kit per manufacturer's instructions (QIAGEN, Hilden, Germany). The cecal DNA concentration was measured using a Qubit dsDNA HS Assay kit (Invitrogen, Waltham, MA, USA). 16S rRNA full‐length amplificons were generated with barcoded full‐length 16S primers from Pacific Bioscience (PacBio) and the 2× KAPA HiFi HotStart Ready Mix (Forward sequence: AGRGTTYGATYMTGGCTCAG; Reverse sequence: RGYTACCTTGTTACGACTT). Amplicons were converted to a library with the SMRTBell Express Template Prep kit 3.0. The library was sequenced on SMRTcell 8 M on a PacBio Sequel IIe sequencing platform using the CCS sequencing mode, at the Roy J. Carver Biotechnology Center, University of Illinois.

2.7. Statistical Analysis

The unpaired parametric t test was used to compare differences in metabolites and colonic NQO1 activities between two groups (LFD vs. LFCB; HFD vs. HFCB). A p value < 0.05 was considered statistically significant. A p value of 0.05−0.1 was considered a trend of change. Statistical analysis was conducted using Prism 9 (GraphPad Software).

All microbiome sequencing data analysis was performed in R (v4.2.1). The sequencing data preprocessing and identification of amplicon sequence variations (ASVs) was conducted using DADA2 [27] (v1.26). Briefly, primers were removed, and the reads were filtered to retain reads with expected lengths (1000–1600 bp). After removing chimeras, taxonomy was assigned using the RDP Naive Bayesian Classifier algorithm (DADA2) and the Silva database v138.1. Species‐level taxonomic assignment was added when possible, using the exact matching approach implemented in DADA2. The reads were then normalized by rarefaction with a cutoff read depth of 20 000 (Figure S1) using vegan (R, vegan package, v2.6‐4) [28] and used for downstream analysis.

For cecal microbiota diversity analysis, alpha‐diversity (within‐sample diversity) was quantified by the Observed, Chao1, Shannon, and Simpson index (R, phyloseq package, v1.40.0) [29]. The difference in alpha‐diversity between treatment groups was analyzed using the Kruskal–Wallis test with post hoc pairwise Wilcoxon test. Bray–Curtis dissimilarity (beta‐diversity) was calculated and visualized using principal coordinate analysis (PCoA) in R using vegan [28] and phyloseq [29]. Permutational multivariate analysis of variance (MANOVA) was used to identify beta‐diversity differences between intervention groups (R, vegan package, adonis2; permutations = 5000). To identify bacterial taxa change at the phylum level, Kruskal–Wallis test with post hoc Dunn's multiple comparison test was performed (Prism, GraphPad software). The impact of background diets (low‐fat vs. high‐fat), cooked broccoli, and the interaction between these two factors on the abundance of individual ASVs were quantified using negative binomial generalized linear models (R, v4.2.1). To control the false discovery rate (FDR), we implemented a model selection procedure. Briefly, an analysis of variance (ANOVA) was used to compare the fit of a null model (ASV abundance ∼ cooked broccoli) and the alternative model (ASV abundance ∼ cooked broccoli * background diet). FDR was then calculated using the p.adjust (method = fdr) function applied to the set of ANOVA p values and used to control the FDR to <0.05. To further evaluate the impact of different intervention groups on the abundance of individual ASVs, pairwise multilevel comparison using pairwise.adonis was performed (R, pairwiseAdonis, v0.4). Correlations between taxa abundance and microbial GRP metabolite concentrations were quantified using the cor.test function (Pearson, stats, v4.2.1). Best fit line and standard error were calculated for each correlation plot using the geom_smooth function (ggplot2). Ggplot2 (v3.4.1) [30] was used for visualization.

3. Results

3.1. Ex Vivo Quantification of Microbial Metabolites of Glucoraphanin

Body weight and food intake were monitored over the 7‐day intervention. There was no difference between control and broccoli‐fed groups in body weight (p > 0.05; LFD vs. LFCB, 28.0 ± 0.6 g vs. 28.5 ± 0.5 g; HFD vs. HFCB, 39.7 ± 1.8 g vs. 39.4 ± 1.8 g) or food intake (p > 0.05). To evaluate the impact of cooked broccoli consumption on cecal microbial GSL metabolism, we measured GRP metabolites by ex vivo, anaerobic incubation of GRP with mouse cecal contents. Free SF, SF‐CYS, and total ITC microbial production were significantly increased in the cooked broccoli‐supplemented group compared to the control group, in both lean and obese mice (Figure 1). The concentrations of these ex vivo cecal microbial metabolites from the control group were lower than the limit of detection and labeled as not detected (Figure 1). In addition, the concentration of SF‐NIT was not statistically different between the broccoli‐supplemented group and control in either lean or obese mice (p > 0.05). The total ITC concentration of each sample is listed in Table S1.

FIGURE 1.

Microbial metabolites of glucoraphanin. Concentrations of SF, SF‐CYS, SF‐NIT, and total ITC after incubation of glucoraphanin with mouse cecal content were quantified using LC‐MS (for SF, SF‐CYS, and SF‐NIT) and HPLC‐DAD (for total ITC) (n = 5 per treatment group). Concentrations lower than the LOD of the analytical method were labeled as ND. The LOD of SF, SF‐CYS, and SF‐NIT with LC‐MS are 0.18, 0.10, and 0.20 µM, respectively. The LOD of total ITC with LC‐DAD is 1.00 µM. Data are mean ± SEM. LOD, limit of detection; ND, not detected; SF, sulforaphane; SF‐CYS, sulforaphane‐cysteine conjugate; SF‐NIT, sulforaphane‐nitrile; total ITC, total isothiocyanates. Unpaired parametric t test was performed for statistical analysis. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001; ns, not significant; dotted lines represent the LOD.

3.2. Colonic NQO1 Activity

To assess the bioactivity of ITCs produced by cecal microbiota after cooked broccoli consumption, we measured colonic NQO1 activity. In both lean and obese mice, colonic NQO1 activity was significantly higher after consumption of the cooked broccoli‐containing diet compared to control (p < 0.01) (Figure 2).

FIGURE 2.

Colonic NAD(P)H: quinone oxidoreductase (NQO1) activity (n = 5 per treatment group). Results are absorbance change at 440 nm/min/µg of protein. Data are mean ± SEM. Unpaired parametric t test was performed for statistical analysis. **p < 0.01.

3.3. Cooked Broccoli Impact Cecal Microbiota Composition

A total of 3 356 165 reads were generated from the 16S rRNA full‐length amplification, with an average read length of 1533 bp and a median of 110 242 reads per sample. Within‐sample diversity (alpha‐diversity) was quantified by the Observed ASVs, Chao1, Shannon, and Simpson index, among which cecal microbiota from the HFCB group had a trend of higher alpha‐diversity (Observed ASVs and Chao1 index, p = 0.054 and 0.069, respectively) compared to HFD group. No difference in alpha‐diversity was observed among LFD, LFCB, and HFD groups (Figure 3A). Between sample diversity (beta‐diversity) was significantly associated with diet type (HFD vs. LFD; PERMANOVA; R ² = 0.13, F = 6.5, p = 0.0002), and supplementation of cooked broccoli (broccoli‐containing diet vs. control, R ² = 0.10, F = 4.9, p = 0.0002). A significant interaction effect between diet × broccoli was also observed in association with microbial diversity (R ² = 0.04, F = 2.04, p = 0.02) (Figure 3B).

FIGURE 3.

Alpha‐ and beta‐diversity of cecal microbiota in HFCB, HFD, LFCB, and LFD fed mice (n = 9–10 per treatment group). (A) Alpha‐diversity, quantified by the Observed, Chao1, Shannon, and Simpson index; the boxes represent the interquartile range (IQR) of alpha‐diversity, with the horizontal line inside the box indicating the median. Whiskers extent to the smallest and largest values within 1.5 times the IQR, while points outside this range represent outliers. The difference between treatment groups was analyzed using the Kruskal–Wallis test with post hoc pairwise Wilcoxon test; (B) beta‐diversity (Bray–Curtis distances) calculated and visualized using principal coordinate analysis (PCoA). Permutational multivariate analysis of variance (MANOVA) was used to identify differences between intervention groups. HFCB, high‐fat diet with cooked broccoli; HFD, high‐fat diet; LFCB, low‐fat diet with cooked broccoli; LFD, low‐fat diet.

A total of 1416 ASVs were identified, with 1355, 852, and 159 ASVs assigned to the family, genus, and species level, respectively (Figure 4A). Cecal microbiota compositional change was evaluated at the phylum level (Figures 4B–E) and by ASVs (Figures 5A–C). The abundance of Firmicutes was not changed by the addition of cooked broccoli in either lean or obese mice. However, Firmicutes was more abundant in HFCB group compared to LFCB (p < 0.05), although no difference was observed between LFD and HFD. There was a trend toward an increase in the abundance of Bacteroidota in LFCB compared to HFD or HFCB (p = 0.06). As for the Firmicutes to Bacteroidota ratio (F/B ratio), a significantly higher F/B ratio was found in HFCB compared to LFCB, and a trend toward a reduced F/B ratio was observed in LFCB compared to LFD (p < 0.1) (Figures 4C–E).

FIGURE 4.

(A) Number of ASVs assigned to phylum, family, genus, and species level after PacBio sequence. (B–E) Cecal microbiota compositional changes in treatment groups (n = 9–10 per group) at the phylum level. Kruskal–Wallis test with post hoc Dunn's multiple comparison test was performed for identifying taxa change at the phylum level (Figure 4C–E). *p < 0.05. ASV, amplicon sequence variation; FB ratio: Firmicutes to Bacteroidota ratio; HFCB, high‐fat diet with cooked broccoli; HFD, high‐fat diet; LFCB, low‐fat diet with cooked broccoli; LFD, low‐fat diet.

FIGURE 5.

Cecal microbiota compositional changes in treatment groups (n = 9–10 per group). (A) ASV abundance impacted by background diet (LFD vs. HFD), cooked broccoli supplementation (broccoli‐supplemented diets [LFCB + HFCB] vs. background diets [LFD + HFD]), and the interaction between the two factors (top 20% significant ASVs). Color gradients represent the direction and magnitude of the estimated effects on ASVs abundance, with blue indicating an increase and red indicating a decrease. The effects of background diets, cooked broccoli, and their interaction on the abundance of individual ASVs were analyzed using negative binomial generalized linear models (R, v4.2.1). (B, C) Comparison of abundance of individual ASV across four intervention groups. The boxes represent the interquartile range (IQR), with the horizontal line inside the box indicating the median abundance. Whiskers extent to the smallest and largest values within 1.5 times the IQR, while points outside this range represent outliers. Pairwise multilevel comparison using pairwise.adonis was performed (R, pairwiseAdonis, v0.4). ASV, amplicon sequence variation; CB, cooked broccoli; HFCB, high‐fat diet with cooked broccoli; HFD, high‐fat diet; LFCB, low‐fat diet with cooked broccoli; LFD, low‐fat diet. Asterisk (*) in panel A and letters in panels B–C denote statistically significant differences. *p < 0.05; **p < 0.01; ***p < 0.001.

Next, we investigated the impact of type of background diet (LFD vs. HFD), the addition of cooked broccoli, and their interaction on the abundance of individual ASVs. The top 20% most abundant ASVs that were significantly affected by at least one of these factors are shown in Figure 5A. These ASVs mainly belong to the families of Peptococcaceae, Muribaculaceae, Lachnospiraceae, and species such as Romboutsia ilealis and Blautia coccoides. Specifically, LFD showed a positive impact on the abundance of Peptococcaceae and R. ilealis and a negative impact on Lachnospiraceae and B. coccoides, compared to HFD (Figure 5A, the 1st column). The supplementation of cooked broccoli to the background diets showed a negative impact on the abundance of ASVs belonging to B. coccoides and Lachnospiraceae, and a positive impact on some of the R. ilealis ASVs (Figure 5A, the 2nd column). The impact of background diet or cooked broccoli on ASVs belonging to Muribaculaceae revealed both positive and negative effects. Interestingly, the interaction between background diet and cooked broccoli (Figure 5A, the 3rd column) showed opposite effects on the abundance of Peptococcaceae and R. ilealis compared to the impact of cooked broccoli alone.

In further investigation of the responses of individual ASVs in different intervention groups, we observed the discordant impact of broccoli supplementation on ASV abundance in lean and obese mice (Figure 5B, C). Ruminococcaceae incertae sedis (ASV32) and Lachnospiraceae UCG‐006 (ASV227) were reduced by the addition of cooked broccoli in lean mice, whereas Muribaculaceae (ASV46) and GCA‐900066575 (ASV135) were increased by the addition of cooked broccoli in both lean and obese mice (Figure 5B). Furthermore, several ASVs were found to be exclusively responsive to cooked broccoli in only one of either lean or obese mice, such as Colidextribacter (ASV59), Dubosiella newyorkensis (ASV34), Blautia (ASV60), and Lachnospiraceae NK4A136 group (ASV105) (Figure 5C).

3.4. Correlations Between Microbiota and Microbial GRP Metabolites

Correlations between microbiota and microbial GRP metabolites were evaluated for the most abundant taxa affected by the interventions, as listed in Figure 5A. The ASVs belonging to the same family or genus were grouped together for the correlation analysis (Figure 6). In total, we identified three significant associations between taxa abundance and total ITC. We found a negative correlation between total ITC and Peptococcaceae (family), Blautia (genus), and Oscillospiraceae (ASV16). A trend toward a positive correlation (p < 0.1) was found between total ITC and Muribaculaceae (family), as well as Lachnospiraceae NK4A136 group (genus). Furthermore, we observed a trend toward a negative correlation between SF and Peptococcaceae (family), a trend toward a positive correlation between NIT and Ruminococcaceae incertae sedis (genus), and a trend toward a negative correlation between CYS and Peptococcaceae (family).

FIGURE 6.

Correlations between cecal microbiota and GSL microbial metabolites. Pearson correlations of microbial ASV, genera, and families with GSL microbial metabolite concentrations (cor.test function; R, stats, v4.2.1) including best‐fit line (geom_smooth function; ggplot; linear regression method) and standard error of best fit (shaded portions). Pearson's product‐moment correlation coefficient, denoted as “correlation”, and p value are indicated for each taxon‐metabolite pair. ASV, amplicon sequence variation; GSL, glucosinolate; SF, sulforaphane; SF‐CYS, sulforaphane‐cysteine; SF‐NIT, sulforaphane‐nitrile; Total ITC, total isothiocyanates.

4. Discussion

4.1. Cooked Broccoli Consumption Impacts Cecal Microbial GRP Metabolism and Bioactivity

In this study, we show that frequent cooked broccoli consumption increased the production of SF, SF‐CYS, and total ITC in ex vivo incubations with GRP by cecal microbiota, from lean and obese mice (Figure 1). To our knowledge, this is the first study that shows multiple GSL microbial metabolite changes in healthy and obese host mice following broccoli consumption. Although the concentration of SF‐NIT in cecal microbial metabolites was not different between control and broccoli‐supplemented groups (LFD vs. LFCB, HFD vs. HFCB), there is a trend toward an increase in its production by cecal contents from lean mice fed broccoli (p = 0.1). Previously, microbial metabolism of GSL in rodents or human has only been evaluated in the form of GSL degradation (measuring the amount of GSL being metabolized) [31] or total ITC production (measured by cyclocondensation) [31, 32]. However, the metabolism of GSL by microbiota is more complex than producing ITC. There are multiple alternative pathways that lead to production of non‐ITC metabolites, including SF‐NIT [12, 33]. It is important to note that these pathways may occur simultaneously in the gut and yet no studies have determined how these competing pathways may explain interindividual variation in microbially produced ITCs from cooked brassica consumption. Moreover, our results showed that only about 10%–20% of ex vivo GRP metabolites were detected, in the form of SF, SF‐NIT, and SF‐CYS. This observation is consistent with other group's findings [34] and supports the hypothesis that alternative GSL metabolism pathways are possibly undefined. The present study provides novel information for understanding the role of the gut microbiota in the metabolic fate of dietary broccoli in lean and obese mice. Further studies are needed to understand the impact of frequent broccoli consumption on non‐SF production pathways, such as the marginally increased NIT observed in this study, and the microbial functions involved. For instance, it remains to be explored whether frequent broccoli consumption also enriches SF‐resistant microbes and/or increases microbial sulfatase activities, which removes sulfur group of GSLs to form desulfo‐GSLs, one of the precursors of NIT. Despite the ability to control metabolism conditions (i.e., GRP concentration) in the ex vivo fermentation of GRP with cecal microbiota, a limitation of this approach is the lack of measurements of metabolites in plasma and/or tissues. However, since mice were fed ad libitum, the variability in GSL consumption before sample collection presents a challenge in accurately evaluating changes in microbial metabolites of GSLs in blood and tissues.

In addition to the increased capacity of cecal microbiota to hydrolyze GRP to form bioactive SF, colonic NQO1 activity was also increased by 7‐day cooked broccoli consumption, in both lean and obese mice (Figure 2). This suggests that increased SF production capacity in the colon leads to increased colonic bioactivity of SF. These findings are consistent with our previous studies in rats [20] and other groups’ studies in lean mice [32]; and for the first time, we show increased NQO1 activity in obese mice after broccoli consumption. NQO1 has been implicated in alleviating intestinal environment imbalance caused by oxidative stress in rodents fed with HFDs [35]. Our data suggest that frequent cooked broccoli consumption may protect the host with obesity from oxidative stress damage in the intestine caused by HFDs.

4.2. Cecal Microbial Composition

The characterization of the cecal microbial composition reveals changes in within‐sample and between‐sample diversity caused by HFD and/or cooked broccoli feeding (Figure 3). At the phylum level, no difference between control and broccoli‐supplemented groups was observed (LFD vs. LFCB, HFD vs. HFCB) (Figure 4C–E). Increased Firmicutes (p < 0.05, LFCB vs. HFCB), a trend toward reduced Bacteroidota (p < 0.1, LFCB vs. HFD, LFCB vs. HFCB), and increased F/B ratio (p < 0.05, LFCB vs. HFCB) were observed when comparing LFCB to HFD/HFCB, suggesting the obese mice may have an imbalanced microbiota community [36]. Seven‐day cooked broccoli feeding did not modulate the F/B ratio in obese mice, but longer intervention periods have been shown to mitigate the gut dysbiosis caused by obesity [26].

The evaluation on the impact of background diet, cooked broccoli, and their interaction on ASVs’ abundance revealed a negative effect of cooked broccoli on ASVs belonging to the families of Lachnospiraceae, Blautia, Oscillospiraceae, and Peptococcaceae, and a positive effect on some ASVs from the family of R. ilealis (Figure 5A). Several studies have reported changes in the Lachnospiraceae family abundance after broccoli/brassica vegetable consumption, including both increased and decreased abundance by brassica‐containing diet [31, 37–40]. The mixed results could be due to different intervention periods, treatment (i.e., brassica type, dose), host health status, and/or the specific genus, species, or strain of Lachnospiraceae. Also, strain of the host mouse. In the present study, broccoli consumption led to a decrease in Lachnospiraceae UCG‐006 (ASV227) in lean mice and an increase in Lachnospiraceae NK4A136 group (ASV105) in obese mice (Figures 5B, C). The latter is interesting because Lachnospiraceae NK4A136 group are SCFA producers and were reported to be negatively correlated with obesity and intestinal inflammation in obese rodents [41, 42, 43, 44]. The shift in Lachnospiraceae NK4A136 group in obese mice suggests the potential for cooked broccoli to improve the metabolic health of host with obesity, via promoting beneficial gut microbiota. The reduction in Blautia by cooked broccoli (Figure 5A, C) is in accordance with our previous findings in rats [20]. Currently, little is known about the families of Peptococcaceae, Oscillospiraceae, or R. ilealis, in terms of their physiological function and response to brassica diet. Interestingly, supplementation of R. ilealis in diet has been shown to impair glucose tolerance and reduce fasting insulin in mice [45]. In the present study, we observed a positive effect of the HFD alone and a negative effect of HFD with broccoli on the abundance of ASVs belong to R. ilealis (ASV73, 74, 79, 88). These data suggest that the interaction between cooked broccoli and diet may alleviate the potential impairment of glucose metabolism by reducing the abundance of R. ilealis.

The broccoli diet led to similar changes in the abundance of several ASVs among lean and obese mice (Figure 5B), whereas several other ASVs were only changed in lean or obese mice (Figure 5C). For example, the abundance of GCA‐900066575 (ASV135) was increased by the cooked broccoli diet in both lean and obese mice, and Lachnospiraceae UCG‐006 (ASV227) abundance was decreased by cooked broccoli in both. Although GCA‐900066575 and Lachnospiraceae UCG‐006 were found to be positively correlated with obesity [46], their abundance changes (increase or decrease) caused by broccoli were not impacted by the background diet type in the present study. On the other hand, D. newyorkensis was only identified in obese mice, and the abundance was increased by cooked broccoli (p < 0.05) (Figure 5C). D. newyorkensis, a novel member of the family Erysipelotrichaceae, was identified from murine intestine in 2017 [47]. It has been associated with antiaging activities and effects in reducing epithelial oxidative stress; its abundance was also negatively correlated with Alzheimer's disease biomarkers in mouse models [48, 49]. Increased abundance of D. newyorkensis in obese mice suggests the prebiotic effect of cooked broccoli, which could lead to potential benefits to the host metabolic health.

4.3. Correlation of Microbial GSL Metabolites Formation With Microbiota Abundance

Correlation between the microbiota abundance and different GSL metabolites provides information for understanding microbial GSL metabolism, which may explain interindividual variation in intestinal ITC production and therefore inform future prebiotic strategies to direct the metabolism of GSL toward more bioactive ITC production. In the present study, the family Peptococcaceae was found to be negatively correlated with total ITC levels, with trends toward negative correlations with SF and CYS. In contrast, the abundance of Muribaculaceae (family) and Lachnospiraceae NK4A136 group (genus) were positively correlated with total ITC (Figure 6). Wu et al. reported similar correlation between Lachnospiraceae NK4A136 and the metabolites. Furthermore, they found that Lachnospiraceae NK4A136 was negatively correlated with proinflammatory cytokines (i.e., TNF‐α and IL‐6) [50], which may be partially explained by the potential of Lachnospiraceae NK4A136 in improving ITC production. Currently, little is known about the metabolism of GSL by Muribaculaceae. However, its potential in promoting total ITC production, as well as being strongly correlated with complex carbohydrate degradation and propionate concentration [51, 52], makes the Muribaculaceae family a promising candidate for studying microbial metabolism of GSL and health impacts of broccoli diet‐caused alteration in host microbiota. ITCs, a group of health‐promoting compounds for the host, can be toxic to bacteria when they reach to sufficiently high concentration, which would lead to loss in bacterial abundance or development of a detoxification pathways (i.e., production of NIT) [12]. Unclassified Oscillospiraceae (ASV16) and Blautia exhibited a negative correlation with total ITC, which was in accordance with their lowered abundance after cooked broccoli consumption. Similarly, Ruminococcaceae incertae sedis was decreased by cooked broccoli and positively correlated with NIT production. Altogether, data suggest these gut bacteria can be sensitive to ITCs released by broccoli consumption and possibly develop alternative pathways to produce inert NIT. Consequently, when the relative abundance of the NIT producers is high in the host gut, the physiological benefits of dietary SF may be compromised, as more microbial metabolic flux of GSLs will be directed to inactive NIT and not bioactive ITCs. Although we did not identify taxa at the species level that correlated with GSL metabolite formation, the current data provide information for understanding and predicting interindividual variation in gut microbial GSL metabolites.

5. Conclusions

Here, we show that 7‐day cooked broccoli consumption improved cecal microbial hydrolysis of GRP to form bioactive SF in lean and obese mice. In addition, we characterized the changes in other microbial GRP metabolites, including SF‐NIT, SF‐CYS, and total ITC, after frequent consumption of cooked broccoli. Improved capacity in SF formation also led to increased colonic NQO1 activity. Lastly, cooked broccoli supplementation altered microbiota composition and boosted the abundance of several microorganisms. Taxa associated with SF‐, SF‐NIT‐, and SF‐CYS‐production were identified. Findings in this study suggest cooked broccoli consumption can alter cecal microbiota to improve the production and bioactivity of SF in both lean and obese mice; the correlation between microbiota and metabolites provides information for understanding microbial GSL metabolic fate.

Conflicts of Interest

The authors declare no conflicts of interest.

Peer Review

The peer review history for this article is available at https://publons.com/publon/10.1002/mnfr.202400813.

Supporting information

Supporting information

Zhao A., Li J., Peterson M., Black M., Gaulke C. A., Jeffery E. H., Miller M. J., Cooked Broccoli Alters Cecal Microbiota and Impacts Microbial Metabolism of Glucoraphanin in Lean and Obese Mice. Mol. Nutr. Food Res. 2025, 69, e202400813. 10.1002/mnfr.202400813

Data Availability Statement

The raw sequence files generated in this project are available at the NCBI Sequence Read Achieve (SRA) project number PRJNA1000605.