This study was approved by the Institutional Review Board at Ohio University (IRB# 18-F-13) and written informed consent was obtained from each respondent prior to entering the study. Participants were recruited with a University-wide email and flyers hung on campus and in the community, then pre-screened for inclusion/exclusion criteria via an online survey and eligibility was confirmed at the baseline visit. Inclusion criteria included being aged 18–50 years, not pregnant, not diabetic, with no unresolved infections or diseases (diabetes, cardiovascular disease, inflammatory or autoimmune disease), and non-smokers. Participants were also free from prescribed anti-inflammatory and corticosteroid use for at least 2 months and had not taken antibiotics within the last year. See Figure 1 for the CONSORT diagram of recruitment and retention and Table 1 for participant demographics.

This study was a double-blind randomized control study. Participants completed five total visits for this study. The first visit was to obtain informed consent and explain the procedures thoroughly. The remaining four visits were scheduled for blood draw, blood pressure assessment (Omron HEM-711), body composition assessment (Bioelectrical impedance, InBody USA), and fecal sample collection. These occurred in the morning after a 10 h fast at baseline, and after 7, 14, and 30 days of supplementation. See Figure 2 for testing schematic.

MTC juice was prepared by diluting 1 fluid ounce of concentrate (King Orchards, Traverse City, MI) with 7 fluid ounces of filtered water in accordance with manufacturer instructions. The placebo was a visually similar carbohydrate- and calorie-matched placebo beverage. Participants were provided with 14, 240 ml (8 oz.) bottles of juice per week for 4 weeks. They were instructed to keep the juice refrigerated until consumption, to shake well at consumption, and to drink two bottles per day, ~8 h apart and not within an hour of exercise. Each bottle of MTC juice contained 77 kcal, 18 g carbohydrate and 0 g fiber and the equivalent of 0.5 lb of MTC per serving, while each bottle of placebo juice contained 67 kcal, 17 g carbohydrate and 0 g fiber. MTC capsules (King Orchards, Traverse City, MI) contained 500 mg of freeze-dried tart cherries, while placebo capsules contained 460 mg cornstarch. Participants were provided with an undisclosed number of capsules and instructed to take two capsules with breakfast each day and to return unused capsules at their next visit. Each MTC capsule contained 1.7 kcal, 0.4 g carbohydrate and 0.2 g fiber and contained 0.92 ounces of cherry skins and pulp. Each placebo capsule contained 1.8 kcal, 0.4 g carbohydrate and 0 g fiber. Independent lab analysis of anthocyanins via high-performance liquid chromatography and total polyphenols via Ultraviolet-Visible Spectroscopy (Certified Laboratories, Mellville, NY) showed MTC juice provided 227 mg anthocyanins and 793 mg total polyphenols per 240 ml bottle while MTC capsules provided 330 mg anthocyanins and 760 mg total polyphenols per capsule. These values are within the ranges reported by previous studies using MTC products (29). Doses were in accordance with previous literature and to ensure similar levels of polyphenols and anthocyanins between formulations. Two capsules were taken at the same time of day rather than spread out for easier compliance. Compliance for both juices was 100%, while it was 92% for placebo capsules and 94% for MTC capsules.

Following the consent visit and prior to their baseline blood draw, participants completed a 12 month food frequency questionnaire (FFQ) that included portion sizes (Dietary History Questionnaire (DHQ) III, National Cancer Institute) to determine typical dietary patterns (i.e., high, low, or normal intake of carbohydrate, protein, fat, and fiber based on established dietary recommendations). In addition to the FFQ, between the 14 and 30 day visits, participants completed a 3 day food record for 2 weekdays and 1 weekend day, tracking food and beverage intake along with portion sizes. Participants recorded their data in Food Prodigy, a companion program to the Food Processor Nutrition Analysis software (ESHA Research). Data from the 3 days was then exported from the Food Processor Nutrition Analysis program as an excel file for each participant. Finally, every 7 days during the course of the study participants completed an online survey regarding their exercise and dietary habits for the previous 7 days as well as the frequency of their intake of alcohol, anti-inflammatory medications, and the top 100 polyphenol containing foods.

Fecal samples were collected by participants into a toilet hat (Protocult 100, Ability Building Center, Minneapolis, MN) and aliquoted into three Polypropylene microcentrifuge tubes that were free from DNase/RNase and pyrogens (Eppendorf™ Biopur®). Samples were immediately frozen at −20°C until delivery to the laboratory where they were stored at −80°C. Samples were transported to the lab on ice (Utek, Sonoco ThermoSafe, Arlington Heights, IL) in a thermal insulated tote (Hopkins Medical Products Caledonia, MI).

DNA was extracted from fecal samples using the Qiagen DNeasy PowerSoil kit (Qiagen, 12888-100) per manufacturer's instructions. Isolated genomic DNA was amplified using custom designed primers targeting the 16s rDNA V3-V4 regions with Kapa Biosystems HiFi HotStart ReadyMix (Roche, KK2601). Amplified products were checked for the correct size using the Agilent 2100 Bioanalyzer on a DNA 1000 chip (Agilent, 5067-1504). Each amplified product was dual indexed using Illumina Nextera XT v2 indices (Illumina, FC-131-200X) according to manufacturer's instructions. Prepared libraries were checked for correct sizing and overall quality using the Agilent 2100 Bioanalyzer on a DNA 1000 chip (Agilent, 5067-1504). Libraries were quantified using a Qubit 3.0 Fluorometer (Invitrogen, Q32850) and pooled equimolarly. Prepared libraries were mixed with 10% PhiX and sequenced on an Illumina MiSeq using a 2 × 300 bp (600 cycle) paired-end sequencing kit (Illumina, MS-102-3003). Sequencing reads were downloaded from the BaseSpace server in FASTQ format. Reads were demultiplexed and adaptors removed. Operational Taxonomic Units (OTUs) were generated using QIIME 2.0 and identified against the SILVA database. Species richness and diversity were calculated with the Shannon index and Pielou's evenness (Supplementary Figure). The datasets presented in this study can be found at https://www.ncbi.nlm.nih.gov/bioproject/PRJNA742936.

Venous blood samples were collected from an antecubital vein into Ethylenediaminetetraacetic acid (EDTA) and serum separator tubes (SST) immediately before and after 7, 14, and 30 days of supplementation. One milliliter of EDTA blood was used to quantify erythrocyte sedimentation rate (ESR) by the Westegren method (Sedi-Rate, Globe Scientific, Inc., Mahwah, NJ). The remaining EDTA blood was stored at 4°C until centrifugation. SST tubes were stored at room temperature for 30 min then centrifuged with EDTA tubes for 15 min (2,500 × g) and 4°C after which supernatants were collected and stored at −80°C for later analysis. Two milliliter serum were sent to an outside facility (Pathology Laboratories, Inc., Toledo, OH) for measurement of uric acid (UA) and C-reactive protein (CRP). Remaining serum samples were assayed in house in duplicate for TNF-a (BMS223HS, Invitrogen, ThermoFisher Scientific, Vienna, Austria), insulin (Catalog #90095, Crystal Chem, Elk Grove, IL), glucose (Item #120003100P, Eton Bioscience, San Diego, CA), and glycated albumin (Catalog # IT3979, G-Biosciences, St. Louis MO). HOMA-IR was calculated as (insulin x glucose)/405 (30). Reference range for CRP was 0.00–0.744 mg·dL⁻¹ and 3.5 and 7.2 mg·dL⁻¹ for males and females for UA.

A priori power analyses for between and within study design to attain 95% power and an alpha of 0.5 were conducted. For microbiota changes, 12 participants per group were needed for an effect size of 1.32 (10, 11). For UA, CRP, and ESR 12 participants per group were needed for an effect size of 0.72 (22), and 0.72 (18, 23), and 0.63 (23), respectively. For glucose a sample size of 12 was needed for an effect size of 0.75 and for insulin a sample size of 16 was needed for an effect size of 0.58 (28). Statistical analysis was completed using Statistical Package for the Social Sciences (SPSS; SPSS 26.0, Chicago, IL, USA). Linear mixed models were used to examine the main effects of time, and treatment, and the interaction effect (time x treatment). For microbiome composition, data was analyzed overall for change in OTUs between groups at baseline, 14 and 30 days. To overcome interindividual variability, a secondary analysis was conducted where participants were divided into groups based on their baseline levels of high (≥10%; n = 23) or low (<10%; n = 35) levels of Bacteroides because they have been shown to be the primary genus impacted by in vitro tart cherry treatment (11). Different covariance structures were systematically fit to the data, and the one that minimized the Hurvich and Tsai's criterion was chosen for the final model. Where a significant F ratio was observed, post-hoc comparisons with LSD-adjusted p-values were used to identify which pairs of means were significantly different. Normality and homogeneity of variance of the residuals were checked using Q-Q plots, and scatter plots, respectively, and deemed plausible in each instance. Two-tailed statistical significance was accepted as p < 0.05. Data are represented as mean ± SD.

There was no significant change in bacterial OTUs over time (F = 0.70, p = 0.50) or between groups (F = 0.19, p = 0.91). When participants were divided by high or low levels of Bacteroides at baseline, there was a significant main effect for time (F = 4.10, p = 0.02) with greater Bacteroides at baseline vs. 14 days (p = 0.05) and 30 days (p = 0.01) and a significant interaction for time and Bacteroides level (F = 7.68, p = 0.001). However, for Firmicutes after dividing between high or low levels of Bacteroides there was no significant main effect for time (F = 0.60, p = 0.55) or group (F = 2.00, p = 0.16). Finally, there were no significant changes in any bacteria phyla or species over time or between groups. OTU data can be seen in Figure 3.

Correlations between the gut microbiome and other variables were mostly weak. BMI was positively correlated with Lactobacillus OTUs (r = 0.21, p = 0.01), Bifidobacterium OTUs (r = 0.21, p = 0.01), relative abundance of Actinobacteria (r = 0.23, p < 0.01), and relative abundance of Bacteroidetes (r = −0.25, p < 0.01). Glucose levels were correlated with relative abundances of Bacteroidetes, Firmicutes, Proteobacteria (r = 0.16, p = 0.04 for all), and Faecalibacterium (r = 0.23, p < 0.01). Roseburia OTUs were correlated with CRP (r = 0.19, p = 0.02).

There was no significant change in ESR over time (F = 0.59, p = 0.63) or between groups (F = 1.82, p = 0.15). Similarly, there was no significant change in CRP over time (F = 0.94, p = 0.42) or between groups (F = 1.64, p = 0.19) nor did UA change over time (F = 1.26, p = 0.29) or between groups (F = 0.18, p = 0.91). For TNF-alpha, there was a significant change over time (F = 2.77, p = 0.04) with higher TNF-alpha at 14 vs. 30 days (p = 0.013; mean difference: 0.14 pg·ml⁻¹, 95% CI: 0.02–0.27 pg·ml⁻¹). However, there was no difference between groups (F = 2.37, p = 0.08). Inflammatory marker data can be seen in Figure 4.

Fasting blood glucose levels steadily rose over the course of the study (F = 9.70, p < 0.001), with greater glucose at 14 days vs. baseline (p = 0.02; mean difference: 5.03 mmol·L⁻¹, 95% CI: 0.75–9.30 mmol·L⁻¹) and 7 days (p = 0.02; mean difference: 4.16 mmol·L⁻¹, 95% CI: 0.52–7.80 mmol·L⁻¹). Similarly, 30 days post was significantly higher vs. baseline (mean difference: 12.20 mmol·L⁻¹, 95% CI: 7.18–17.21 mmol·L⁻¹), 7 days (mean difference: 11.32 mmol·L⁻¹, 95% CI: 6.89–15.76 mmol·L⁻¹), and 14 days (mean difference: 7.17 mmol·L⁻¹, 95% CI: 3.38–10.89 mmol·L⁻¹) (p < 0.01 for all). However, there was no difference between groups (F = 1.17, p = 0.33). Insulin did not change over time (F = 1.79, p = 0.16), nor was it different between groups (F = 1.45, p = 0.24). Similarly, glycated albumin was not different over time (F = 2.39, p = 0.08) or between groups (F = 0.71, p = 0.56). Finally, HOMA-IR was not different over time (F = 1.07, p = 0.37) or between groups (F = 0.58, p = 0.63). Glucose regulation data can be seen in Figure 5.

Completed FFQ and 3 day diet logs were compiled for 51/58 participants. Typical dietary pattern analysis found that all participants were classified as normal for protein intake, 39 participants were classified as normal, 4 classified as high, and 8 classified as low for carbohydrate intake. For fat, 36 were classified as normal and 15 were classified as high. For fiber, 16 participants were classified as normal, 11 were classified as high, and 24 were classified as low. There was no significant difference between groups in any of the macronutrients analyzed from the 3 day food log in this study (Table 2). Consumption of the top 100 polyphenol containing foods were tracked weekly, results can be found in Supplemental data. Total polyphenol consumption was 5,024 ± 4,599 for the MTC juice group, 4,744 ± 2,387 for the MTC capsule group, 2,519 ± 2,471 for the juice placebo group, and 3,424 ± 3,278 for the capsule placebo group and this was not significant between MTC formulations (X² = 5.43, p = 0.14), but was higher for MTC vs. placebo group (U = 193, p = 0.02).

The relationship between polyphenols and the gut microbiome has been established, whereby the degradation of most polyphenols requires host microbes and these microbes in turn utilize the products produced from polyphenol degradation for energy. This degradation of polyphenols often leads to greater bioavailability and biological activity (31) however this is dependent on the host microbiota composition (32). Tart cherries are well-known to be high in polyphenols and two previous investigations have examined the impact of tart cherry supplementation on the composition of the gut microbiome with disparate findings. Mayta-Apaza (11) investigated ingestion of 8 oz. MTC juice for 5 days in 10 participants. When comparing pre-to post-intervention microbiota, very little change was detected. However, the authors determined the need to divide participants into groups based on the baseline relative abundance of Bacteroides, either high or low. In the high Bacteroides group, ingestion of tart cherry juice resulted in a sharp decline in Bacteroides and an increase in Firmicutes (such as Ruminococcus, Clostridium, Streptococcus and Lactobacillus), while the opposite was seen in the low Bacteroides group, with an increase in Bacteroides and decreases in Firmicutes (such as Streptococcus and Lachnospiraceae). These changes may have been due to the underlying diets of the participants, as those in the low Bacteroides group consumed more carbohydrates, sugars, and fibers and the high polyphenol intake from the tart cherry juice resulted in increases in Bacteroides to facilitate breakdown of these polyphenols. When we divided our participants into groups based on high or low baseline levels of Bacteroides, we noted significantly higher Bacteroides in the high vs. low group at baseline and over the 30 days and no change in Firmicute levels, in contrast to the finding of Mayta-Apaza. However, we did not find any significant differences or changes in the other bacteria phyla or species, even if the groups were divided by their baseline Bacteroides levels.

Contrary to the findings of Mayta-Apaza, Lear (25) found no significant impact of 4 weeks supplementing with tart cherry concentrate on gut microbial composition. These authors noted their samples contained very low abundance of Lactobacillus and Bifidobacterium, which should have been more abundant and found their collection and storage methods may have resulted in alterations in their abundances. Nevertheless, our results are very similar to theirs despite our samples having an average 4% relative abundance of Bifidobacterium and 1% for Lactobacillus, well within the ranges of the healthy lower limits for these taxa (33). Early data suggests that the microbiome can be altered with dietary intake strategies, such as exclusively plant or animal based diets (6). For example, 20 days of red wine consumption increased Firmicutes and Bacteroidetes (8) and 30 days consumption of a high polyphenol cocoa drink increased bifidobacterial and lactobacilli populations. However, further analysis suggests these changes are transient, only lasting for 24 h after the consumption of high-fat/low fiber or low-fat/high fiber intake (34). In our study, it is possible that MTC supplementation brought change in microbial composition that might have not been measured with our fecal sample collection times (i.e., we missed the most opportune time for bacterial changes, although admittedly, we do not know what the most opportune timing for microbiome sampling following polyphenol consumption is since the majority of work is after very short (≤ 5 days) or long (>30 days) interventions). Further, Leeming et al. (35) note that the gut microbiota typically reverts back to its baseline state post-intervention, suggesting the need for long term dietary intervention. Therefore, future research should focus on detecting the sufficient time that bring the effective change in microbial composition from the intake of tart cherry supplementation.

It is also interesting to note that dietary intake of our participants did not seem to be a determining factor in whether or not the MTC supplementation altered microbial diversity, in contrast to Mayta-Apaza. Nearly 25% of our sample were considered to have high fat diets, while 40% had low fiber intake. While this likely determined whether a participant had high or low levels of Bacteroidetes at baseline, either the supplements did not provide enough of a stimulus for change, or the participant's dietary patterns were too influential to bring about change. This might indicate a need for greater changes in diet, along with MTC supplementation, to see significant alterations in gut microbiome composition.

Cell-line work demonstrates MTC extract treatment leads to changes in key enzymes related to glucose regulation in diabetes, including alpha amylase (26). This would result in slower peak glucose levels and a longer time for carbohydrate digestion, which may decrease average blood glucose over time. Glucose regulation in the current study was assessed via measurement of glycated albumin, which is reflective of glycemia over a 2–3 week period (51), as well as blood glucose and insulin levels. Glycated albumin and insulin were not significantly changed over time nor were they different between groups, however blood glucose levels steadily rose over the course of the study but were not different between groups. This data is in line with the work by Lear (25) who found a decrease in insulin sensitivity (decreased Matsuda index), with an increase in insulin and no decrease in blood glucose in those supplementing with MTC juice. Additionally, Chai et al. (28) found an increase in blood glucose following 12 weeks of MTC concentrate ingestion, while Kimble et al. (52) found no impact of 12 weeks MTC concentrate ingestion on blood glucose levels. Interestingly, a seven day intervention in patients with metabolic syndrome found a significant decline in blood glucose and an increase in insulin (53), which suggests in the short-term, supplementation may improve insulin sensitivity, but this does not appear to translate to long-term benefits. Our findings show an elevation in glucose without a rise in insulin, which might indicate presence or development of insulin insensitivity, indeed 16 individuals (28%) had a fasting glucose >100 mg·dL⁻¹ (but <126 mg·dL⁻¹) at baseline and an additional 5 participants had elevated glucose levels from 7 days onward. Finally, the difference in sugar composition between the juice and capsule groups were significantly different (17.7 vs. 0.4 g), however there was no significant difference between groups in blood glucose regulation, it appears unlikely the results are due to greater sugar intake from the supplements and further human intervention work with MTC supplements and blood glucose regulation are needed.