Association between Accelerated Biological Aging, Diet, and Gut Microbiome

The 10,000 Families Study (10KFS) is a new, family-based prospective cohort study in Minnesota that was established in 2017 with ongoing recruitment. A description of the cohort recruitment methods has been published previously [26]. Briefly, index participants were eligible for inclusion in the study if they were adults (≥18 years), lived in Minnesota, provided consent, and had at least one family member eligible to participate. Eligibility was assessed via a brief online eligibility screener. These ‘index participants’ were asked to invite family members to the study if eligible. Once at least one additional family member was enrolled in the study, the family became eligible for inclusion. The definition of “family” was determined by study participants. Participation in the study included completing an online questionnaire (self-completion for individuals 18 or older; parent completion for children <18 years), an invitation to attend a health visit, and providing a stool sample for microbiome analysis.

This study was approved by the Institutional Review Board at the University of Minnesota (IRB approval number: STUDY00000877). The current analysis includes the subset of 10KFS participants who enrolled in the study by the end of 2021, provided consent, questionnaire data, and a stool sample (n = 148). After further excluding people with missing diet information and accelerated KDM-BA (n = 31), 117 participants were included in the final analysis.

Biospecimens and other health measurements were collected during an in-person health visit conducted by trained study personnel. The study health visit included a collection of anthropometric measurements (height, weight, hip circumference, and waist circumference), blood pressure, pulmonary function, and biospecimens, including blood, urine, saliva, and samples as previously described [26]. In addition, participants were sent home with a stool collection Maxwell^® RSC Fecal Microbiome DNA Kit that was returned via mail.

Seated blood pressure and heart rate were measured using an Omron EVOLV (7000 series) upper arm blood pressure monitor. Total serum cholesterol was measured in serum using a cholesterol oxidase method (Roche Diagnostics, Indianapolis, IN 46250, USA) on a Roche Cobas 8000 Chemistry Analyzer (Roche Diagnostics Corporation). IgG antibodies to CMV were measured in EDTA plasma using the electrochemiluminescence immunoassay (on a Cobas 8000 Chemistry Analyzer, Roche Diagnostics Corporation). Spirometry for assessment of pulmonary function was performed with a FlowMIR Air Smart Spirometer (Pond Healthcare Innovation AB Lastmakargata 3, SE11144 Stockholm Sweden) and a disposable turbine with a cardboard mouthpiece (Medical International Research S.r.l. Via del Maggiolina 125 Rome Italy; USA MIR USA, Inc. 1900 Pewaukee Road, Suite D Waukesha, WI 53188, USA) and Air Smart Spirometer app (Pond Healthcare Innovation, Stockholm, Sweden).

The primary outcome for our analysis is accelerated biological age, measured as accelerated KDM biological age (KDM-BA), estimated as a continuous variable. We first calculated the biological age (KDM-BA) using chronological age and nine clinical biomarkers [4]: albumin, creatinine, fasting glucose, blood urea nitrogen (BUN), total cholesterol, systolic blood pressure, C-reactive protein, antibodies to cytomegalovirus (CMV) infection, alkaline phosphatase, and peak expiratory flow [4,27,28]. Subsequently, we determined “accelerated KDM-BA” by subtracting a person’s chronological age from their KDM-BA. If the KDM-BA is higher than their chronological age, it indicates that their biological age is accelerated, meaning their body shows signs of aging faster than expected for their chronological age. Conversely, if the KDM-BA is lower, it suggests that their biological age is decelerated, meaning their body is aging more slowly than expected for their chronological age.

Diet was assessed using the National Health and Nutrition Examination Survey (NHANES) diet screening questionnaire, which collected detailed information about food and nutrient intake [29].The 26-item Dietary Screener Questionnaire (DSQ) asks about the frequency of consumption of selected foods and drinks in the past month. The DSQ captures intakes of fruits and vegetables, dairy/calcium, added sugars, whole grains/fiber, red meat, and processed meat.

We used a SAS program provided by the NHANES [29] to calculate the predicted intake of fiber (gm) per day, calcium (mg) per day, whole grains (ounce equivalent) per day, total added sugar (tsp equivalent) per day, dairy (cup equivalent) per day, vegetables including legumes and French fries (cup equivalent) per day, vegetables including legumes and excluding French fries (cup equivalent) per day, and fruits (cup equivalent) per day. Meat variables were assessed using a self-reported questionnaire where participants reported the frequency of eating individual items in the following categories; “1-time last month”, “2–3 times last month”, “1 time per week”, “2 times per week”, “3–4 times per week”, “5–6 times per week”, “2–3 times per day”/“4–5 times per day”, “1 time per day” and “never”. For this analysis, the diet variables were converted to continuous variables ranging from 0 (never) to 7 (4–5 times/day).

According to manufacturer recommendations, genomic DNA was extracted from human fecal slurry via Maxwell^® RSC Fecal Microbiome DNA Kit (Promega, Fitchburg, WI, USA). Previously published primers (Supplementary Table S1) targeting the V4 region of the 16s rRNA gene were acquired from IDT (Integrated DNA Technologies, Coralville, IA, USA) [30].

The amount of 2 µL of fecal microbial genomic DNA was amplified in 25 µL reactions using 3 µL of nuclease-free water, 12.5 µL of Platinum Hot Start PCR 2X Master Mix (ThermoFisher Scientific, Waltham, MA, USA), 5 µL of Platinum GC Enhancer (ThermoFisher Scientific, Waltham, MA, USA), and 1.25 µL of 100 µM primer for both the forward and reverse primers. DNA amplification was performed on a PCR thermocycler (Applied Biosystems™ MiniAmp™ Thermal Cycler, Waltham, MA, USA, Catalog number: A37834) with the following protocol: first denaturation at 94 °C for 2 min, followed by 35 cycles of 94 °C for 45 s, 50 °C for 60 s, and 72 °C for 60 s. The PCR product was purified via QIAquick PCR purification Kit (Qiagen, Hilden, Germany) according to manufacturer recommendations, and 4 nM libraries were pooled, denatured with 0.2 N NaOH, and diluted to 8 pM following the MiSeq v2 sequencing protocol (Illumina, CA, USA). Dual-indexed paired-end sequencing was performed using a MiSeq Reagent Kit v2 (500-cycles) cartridge on the Illumina MiSeq platform (Illumina, CA, USA).

After demulitiplexing, sequence data were processed using mothur ver. 1.41.1 [31], using a modified version of a previously published pipeline for V4 data [32]. Sequence data were quality-trimmed and aligned against the SILVA database (ver. 138) [32] for further processing. Chimeras were identified and removed using UCHIME ver. 4.2.40 [33]. Amplicon sequence variants were binned at 99% similarity using the furthest-neighbor algorithm. Taxonomy was annotated using the Ribosomal Database Project database (ver. 18) [34].

Three hundred and ninety-one genera were identified from 10KFS biological fecal samples collected from 148 participants. For this analysis, we only considered genera with a relative abundance of at least 1% in at least 80% of the samples, resulting in 266 genera. Finally, we included only 117 participants with complete information on KDM-BA and microbial genera.

Self-reported age, race/ethnicity, sex (male, female), smoking status, and alcohol intake were self-reported using a questionnaire at baseline. Smoking status was analyzed as a binary variable (current/former vs. never). Current and former smokers were combined into a single category due to the small number of current smokers in this study. Alcohol was recorded and analyzed as a categorical variable (current vs. former vs. never drinkers). Body mass index was calculated based on height (meters) and weight (kg) measured during the health visit using the formula “Weight (in Kg)/Height [2] (in meters)”. Family relatedness was self-reported by study participants at baseline.

Data analysis was performed using SAS version 9.4. Descriptive statistics included means and standard deviations for continuous variables (accelerated KDM-BA, chronological age, BMI) and percentages/frequencies for categorical variables (sex, smoking, alcohol use). We used a univariate linear regression model (Model 1) to analyze the unadjusted association between accelerated KDM-BA and individual dietary components. Next, we evaluated the association between accelerated KDM-BA and individual dietary components after adjusting for covariates such as chronological age, sex, BMI, smoking status, alcohol use, and a random intercept for family relatedness using mixed effect regression (Model 2). These covariates were chosen based on their known associations with biological aging and dietary habits, supported by the existing literature [27,28]. Finally, we created a single model (Model 3) in which all dietary components significantly associated with accelerated KDM-BA in Model 2 (i.e., fiber intake, red meat intake, and processed meat intake) were further adjusted for bacterial genera associated with accelerated KDM-BA. This allowed us to evaluate whether adjustment for microbial genera associated with accelerated KDM-BA changed the association between dietary variables and accelerated KDM-BA.

To identify bacterial genera correlated with accelerated KDM-BA, we used Spearman correlation. Bacterial genera with significant Spearman correlation (r ≥ 0.2 or r ≤ −0.2 and p < 0.05) were further evaluated using a linear mixed effects regression model after adjustment for age, sex, BMI, alcohol intake, and smoking status, and using family relatedness as the random intercept. This helped us find bacterial genera significantly associated with accelerated KDM-BA. A similar method was used to identify the association between bacterial genera and dietary variables associated with accelerated KDM-BA. Given the large number of tests conducted, we applied the false discovery rate (FDR) correction to balance the discovery of true positives while controlling the expected proportion of false positives [35]. Bacterial genera with FDR-corrected p-values < 0.05 were considered significantly associated with accelerated KDM-BA.

Additionally, we estimated α diversity at the OTU (operational taxonomic unit) level using three indices: Shannon, Chao1, and Simpson’s. The Shannon index accounts for both OTU richness and evenness, Chao1 estimates OTU richness, including rare OTUs [36], and Simpson’s index assesses diversity with a focus on the dominance of common OTUs [37]. For β diversity, Bray–Curtis dissimilarities [37] were calculated in mothur and visualized by ordination with principal coordinate analysis. Analysis of similarity (ANOSIM) [38] was used to evaluate significant differences in β diversity. For γ diversity, we used the R statistical software’s vegan package, which measures the overall biodiversity of the gut microbiome across all participants.

The demographic characteristics of the participants included in this analysis are described in Table 1. This study included 117 individuals, with 78 women (53%) and a mean age of 54.34 +/− 17.49 years (age range: 19–91 years). Sixty-eight individuals (57%) showed decelerated KDM-BA (i.e., KDM-BA lower than chronological age). A vast majority of the study participants were white (94.87%) and nonsmokers (77.31%) (Table 1). There were 117 individuals in the study from 21 families; every individual was an adult. Each family is identified by a unique family identifier (FID).

Univariate analysis showed that processed meat, red meat, calcium (mg/day), and dairy (cup equivalent/day) intake were positively associated with accelerated KDM-BA (Table 2). There was a 2.58-year increase in accelerated KDM-BA with one standard deviation increase in processed meat consumption (p < 0.001) and a 1.92-year increase in accelerated KDM-BA with one standard deviation increase in red meat consumption (p = 0.01) (Table 2, Model 1). There was a 0.01-year increase in accelerated KDM-BA with one standard deviation increase in calcium intake (p = 0.04) and a 1.95-year increase in KDM-BA with one standard deviation increase in dairy intake (p = 0.04) (Table 2, Model 1). However, after adjustment for family relatedness, sex, chronological age, BMI, alcohol intake, and smoking status, only processed meat intake (2.21 years increase in KDM-BA; p < 0.001) and red meat intake (1.62 years increase in KDM-BA; p = 0.03) remained significantly associated with accelerated KDM-BA (Table 2, Model 2). Additionally, fiber intake (gm/day) showed a significant inverse association with accelerated KDM-BA (0.56 years decrease, p = 0.04) (Table 2, Model 2). Including all significant diet variables from Model 2 (processed meat, red meat, and dietary fiber) in the same model after adjusting for covariates attenuated the association between processed meat and accelerated KDM-BA (1.91 years of accelerated KDM-BA; p = 0.04). Fiber intake remained inversely significantly associated with accelerated KDM-BA (a decrease of 0.70 years in accelerated KDM-BA; p = 0.01), while red meat was no longer associated with accelerated KDM-BA after adjustment for fiber and processed meat intake (p = 0.51). The p-values reported here are adjusted for multiple comparisons; the false discovery rate (FDR) correction was applied in the analysis to balance the discovery of true positives while controlling the expected proportion of false positives.

Individual univariate linear regression analyses of the gut microbiome identified genera significantly correlated with accelerated KDM-BA. Streptococcus was positively associated with accelerated KDM-BA, while Faecalitalea, Subdoligranulum, and unclassified members of the Bacteroidetes (phylum) and Burkholderiales (order) were negatively associated with accelerated KDM-BA (FDR p-value < 0.05), suggesting an inverse relationship with age acceleration. Further adjustment for family relatedness, sex, chronological age, BMI, alcohol consumption, and smoking status did not change these associations (Model 2) except for Faecalitalea, which was no longer negatively associated with accelerated KDM-BA after multivariate FDR adjustment (Figure 1).

We found no association between α diversity indices (Shannon/Chao) and accelerated KDM-BA (Shannon: r = 0.15, p = 0.12; Chao1: r = 0.08, p = 0.41). Additionally, β diversity (differences in microbial community composition) did not differ significantly between individuals with accelerated KDM-BA and those with KDM-BA younger than their chronological age (ANOSIM R = 0.006, p = 0.296). Moreover, the γ diversity of 4.27 suggests that the gut microbiome is relatively diverse across all participants. This could imply that despite individual variations, the collective microbiome has a rich and varied composition.

We did not identify any microbial genera significantly associated with meat or fiber intake. After adjusting for the microbial genera Streptococcus, Subdoligranulum, Bacteroidetes (phylum), and Burkholderiales (order)—which were found to be associated with accelerated KDM-BA but not directly linked to dietary intake of processed meat or fiber—the association between processed meat and accelerated KDM-BA was no longer statistically significant (1.49 years of accelerated KDM-BA; p = 0.07) (Table 2, Model 3). However, the association between fiber intake and accelerated KDM-BA remained unchanged after adjustment for the microbial genera (−0.65 years decrease in the accelerated KDM-BA; p = 0.01) (Table 2, Model 3).

The data supporting this study’s findings are available upon request from the corresponding author, Bharat Thyagarajan.