Abstract

   graphic file with name pt3c00261_0006.jpg

   Human gut microbiota are recognized as critical players in both
   metabolic disease and drug metabolism. However, medication–microbiota
   interactions in cardiometabolic diseases are not well understood. To
   gain a comprehensive understanding of how medication intake impacts the
   gut microbiota, we investigated the association of microbial structure
   with the use of single or multiple medications in a cohort of 134
   middle-aged adults diagnosed with cardiometabolic disease, recruited
   from Alberta’s Tomorrow Project. Predominant cardiometabolic
   prescription medication classes (12 total) were included in our
   analysis. Multivariate Association with Linear Model (MaAsLin2) was
   employed and results were corrected for age, BMI, sex, and diet to
   evaluate the relationship between microbial features and single- or
   multimedication use. Highly individualized microbiota profiles were
   observed across participants, and increasing medication use was
   negatively correlated with α-diversity. A total of 46 associations were
   identified between microbial composition and single medications,
   exemplified by the depletion of Akkermansia muciniphila by β-blockers
   and statins, and the enrichment of Escherichia/Shigella and depletion
   of Bacteroides xylanisolvens by metformin. Metagenomics prediction
   further indicated alterations in microbial functions associated with
   single medications such as the depletion of enzymes involved in energy
   metabolism encoded by Eggerthella lenta due to β-blocker use. Specific
   dual medication combinations also had profound impacts, including the
   depletion of Romboutsia and Butyriciocccus by statin plus metformin.
   Together, these results show reductions in bacterial diversity as well
   as species and microbial functional potential associated with both
   single- and multimedication use in cardiometabolic disease.

   Keywords: gut microbiota, microbial structure, medications,
   cardiometabolic disease, pharmacology
     __________________________________________________________________

   Cardiometabolic disease (CMD) is a leading cause of death
   worldwide.^[34]1 Emerging evidence suggests that the gut microbiota may
   play a pivotal role in influencing the susceptibility and progression
   of CMD, although the precise mechanisms and causal connections are
   still the subject of ongoing research. Disruptions in the diversity of
   the gut microbiota, accompanied by a decline in its functional
   capacity, have been observed in various cardiometabolic conditions,
   including coronary artery disease, diabetic cardiomyopathy, and heart
   failure.^[35]2 Notably, aberrant gut microbiota may serve as a
   formidable catalyst in CMD, elevating risk by exacerbating underlying
   inflammation and oxidative stress.^[36]3 Causal links have been
   established between gut-derived bacterial metabolites and CMD,
   including trimethylamine N-oxide and phenylacetylglutamine.^[37]4

   Research exploring the interactions between CMD medication and the gut
   microbiota is a relatively new and evolving field. Relevant to the
   present study, the gut microbiota is widely recognized as a key player
   in the metabolism of drugs.^[38]5 Extensive interactions between gut
   microbes and medications have been shown in vitro([39]6,[40]7) with
   microbes possessing capacities to affect drug metabolism via gene,
   enzyme, and metabolite modification.^[41]8 For example, Eggerthella
   lenta produces glycoside reductase that inactivates the cardiac
   medication digoxin in the absence of arginine.^[42]9 At the same time,
   medication use can also affect the gut microbiota and its diversity in
   CMD.^[43]10

   To address the complex nature of CMD, multiple medications are often
   used to manage various aspects of the disease, including hypertension,
   dyslipidemia, glucose intolerance, and heart failure. The impacts of
   single- and multimedication use on the gut microbiota are a topic of
   intense investigation, yet our understanding of the relationships
   between medications and the microbiota remains limited. Employing an
   established longitudinal cohort, the present study examined
   associations between gut microbiota and prescription medication use in
   CMD. These findings reveal the possible regulation of medication use on
   the gut microbiota from a population-based and clinical context.

Methods

Study Setting and Population

   This study was approved by the Conjoint Health Research Ethics Board at
   the University of Calgary (REB17-1973). Written informed consent was
   received from all of the study participants prior to inclusion in the
   study. The trial complied with the protocols and clinical practice
   guidelines of the International Conference on Harmonization and the
   Declaration of Helsinki. Alberta’s Tomorrow Project is a
   population-based cohort started in 1999, with over 50,000 individuals
   enrolled to date. Details of the cohort have been published
   previously.^[44]11,[45]12 A small sample subset of participants (∼1000)
   from the cohort were selected to be recontacted for the present study.
   Briefly, participants within the cohort were selected and contacted via
   phone or email in the Calgary area (Calgary, AB, Canada). A random
   digit dialing method mapped to Alberta Regional Health Authorities was
   initially employed to select households with eligible residents.
   Participants responding to the call and between 35–69 years of age at
   the time of enrollment were screened for eligibility. In total, 443
   individuals (males: 28.2%; females: 71.8%) responded. Individuals were
   evaluated for suitability according to predetermined inclusion and
   exclusion criteria. Inclusion criteria for the current study included
   residence within the study area, the ability to complete the
   questionnaires (speak and read English), available medical records
   (including current medication use), provision of specimens (blood and
   feces) within a specified time frame, and the willingness to have
   samples analyzed for the measures of interest. Exclusion criteria
   included known confounders of the gut microbiota: antibiotic use in
   past 3 months, pregnant women and cancer patients. A study CONSORT
   flowchart is shown in [46]Figure S1.

   Sample size was not calculated prior to recruitment but was set to
   cover all participants with cardiometabolic diseases among the 443
   recruits. The cardiovascular subcohort consisted of 134 patients from
   whom the diagnosis was confirmed by a physician. CMD was defined
   following recognized, international definitions of disease according to
   the American College of Cardiology and American Heart Association
   including hypertension,^[47]13 heart failure,^[48]14 atrial
   fibrillation,^[49]15 valvular heart disease,^[50]16 overweight and
   obesity,^[51]17 and myocardial infarction.^[52]18

Metadata Collection

   Participants answered questionnaires relating to demographics,
   lifestyle, diet, eating/drinking behaviors, smoking, medication, and
   sleeping habits as previously described.^[53]19 Questionnaires were
   completed independently by the participants. Medication information was
   curated referring to DrugBank^[54]20 and Drugs.com.^[55]21 Medication
   was classified into categories based on the mechanism of action
   ([56]Table [57]1
   ). Cluster analysis of metadata was performed using Euclidean for
   distance-optimized data transformation^[58]22 as well as Hierarchical
   Ward’s linkage clustering.^[59]23

Table 1. Cardiometabolic Medication Use (n = 134).

     medication class number (%)
   major medications angII_receptor_antagonist 41 (30.6)
     statin 41 (30.6)
     angiotensin-converting enzyme (ACE) inhibitors 33 (24.6)
     NSAID 31 (23.1)
     diuretic 28 (20.9)
     Ca_channel_blocker 24 (17.9)
     beta_blockers 17 (12.7)
     metformin 17 (12.7)
     proton pump inhibitors (PPIs) 15 (11.2)
     levothyroxine 15 (11.2)
     5-aminosalicylic acid (5-ASA) 12 (9.0)
     selective serotonin reuptake inhibitors (SSRIs) antidepressant 10
   (7.5)

   minor medications other_antidepressant 7 (5.2)
     anticonvulsant 6 (4.5)
     antihistamine 6 (4.5)
     serotonin and norepinephrine reuptake inhibitors 6 (4.5)
     tricyclic_antidepressant 6 (4.5)
     melatonin 5 (3.7)
     sodium-glucose cotransporter-2 (SGLT2) inhibitors 4 (3.0)
     beta2_adrenergic agonist 4 (3.0)
     estrogen 4 (3.0)
     non_opioid_analgesic 4 (3.0)
     opioid_analgesic 4 (3.0)
     anticoagulant 3 (2.2)
     non_statin_cholesterol_absorption_inhibitor 3 (2.2)
     sulfonylureas 3 (2.2)
     alpha_blocker 3 (2.2)
     cannabinoid 3 (2.2)
     corticosteroids 3 (2.2)
     progesterone 3 (2.2)
     antirheumatic 2 (1.5)
     thiazolidinedione 2 (1.5)
     dipeptidyl peptidase IV (DPP IV) inhibitors 2 (1.5)
     leukotriene receptor antagonists 2 (1.5)
     xanthine_oxidase_inhibitor 2 (1.5)
   [60]Open in a new tab

Sample Collection, DNA Extraction, and Processing

   Stool samples were self-collected using a Protocult stool collection
   device (Ability Building Center, Rochester, NY). Samples were frozen at
   −20 °C at home, transported to laboratory, aliquoted, and stored at −80
   °C within 48 h of sample collection, until further analysis. Fecal
   genomic DNA was extracted using QIAamp Fast DNA stool mini kit (Qiagen,
   Germany) as previously described.^[61]24 Blank extraction controls were
   included for subsequent sequencing. Stool samples were collected 2
   weeks following questionnaire administration, ensuring alignment of the
   sample with current medication use.

High-Throughput Sequencing and Analysis

   The DNA library construction and pair-end sequencing were performed on
   an Illumina MiSeq platform with a MiSeq V3 600 cycle sequencing kit.
   Raw sequences were demultiplexed with 0 mismatches in the barcode
   sequences. Data processing was conducted using the DADA2 version 1.10
   workflow, which gives rise to abundance tables of amplicon sequence
   variants (ASVs), which underwent chimera filtering. ASVs were
   taxonomically annotated using the Bayesian classifier provided in
   DADA2^[62]25 and using the Ribosomal Database Project (RDP) classifier.
   To map ASV to taxonomically closest relatives, representative sequences
   per ASV were blasted using NCBI Nucleotide BLAST with reference to 16S
   rRNA databases. Data of the blank controls were subtracted from data
   obtained for sample composition.

Associations of Shannon Diversity Index with Bacterial Taxa and Medication
Used

   Shannon diversity index was calculated using the function diversity in
   the R package vegan (version 2.6–4). The correlation between Shannon
   diversity index and bacterial phylum was analyzed by using Spearman’s
   rho correlation analysis in GraphPad Prism version 9.3.1 (GraphPad
   Software, San Diego). P < 0.05 was considered to be significant
   correlation. Linear regression between the number of medications used
   (categorical data) and Shannon diversity index was evaluated using lm()
   function in packge stats (version 4.2.2) and results were visualized
   using ggplot2 (version 3.4.0).

Associations of Bray–Curtis Distance with Variations Explained by Medication
Classes

   Bray–Curtis distance was calculated using the function vegdist in the R
   package vegan (version 2.6–4). The variations of Bray–Curtis distance
   explained by each medication class (single- or multimedication) were
   evaluated using the function adonis2 from the R package vegan (version
   2.6–4). Medication classes were used at this point and subsequent
   analyses if they were present in at least 10 participants. The P value
   was adjusted for multiple testing using the Benjamini and Hochberg
   method.^[63]26 FDR < 0.1 was considered significant.

Microbial Composition Associated with Medication Classes

   The data of genus and ASVs were separately transformed into centered
   log-ratio before further analysis, considering the compositional
   property of microbiota data.^[64]27 To evaluate the relationship
   between microbial features and medication class, the associations of
   individual taxa to each medication class (single or dual medications)
   were assessed by Multivariate Association with Linear Model using
   MaAsLin2.^[65]28 For identifying dual medication combinations and
   avoiding overlap in medication pairs, medications were sorted in
   frequency-based descending order. Only medication pairs with >10
   participants were included for analysis. The default settings in
   MaAsLin2 were used. The associations were corrected for age, sex, BMI,
   and diet. FDR < 0.25, as defaulted in MaAsLin2, was considered
   significant association with medication class.

Metagenomics Prediction of the Microbial Functional Profiles

   The potential function of the gut microbiota was predicted from the ASV
   data using the Tax4Fun^[66]29 metagenomics prediction module available
   at the MicrobiomeAnalyst Pipeline,^[67]30 which resulted in the
   abundances of KEGG organisms’ functional profiles. The KEGG orthologues
   (KOs) were annotated and assigned to the KEGG pathways using the KEGG
   mapping tool. The associations of individual KOs to each medication
   were assessed with MaAsLin2, using centered log-ratio-transformed data
   and correcting for the impact of age, sex, BMI, and diet. FDR < 0.25,
   as defaulted in MaAsLin2, was considered significant association with
   medication class. Discriminant KOs were enriched into KEGG pathways
   through the Shotgun Data Profiling module implemented within
   MicrobiomeAnalyst. The presence of KOs in the representative bacterial
   strains associated with medications was mapped by searching against
   UniProt, NCBI Genome, and the BRENDA database. The coefficient heatmap
   of discriminant KOs with medication classes was mapped using GraphPad
   Prism version 9.3.1.

Data Availability

   16S rRNA gene sequencing data has been deposited in NCBI’s Sequence
   Read Archive (SRA) database and are available through the accession
   number PRJNA 922681.

Results

Subject Characteristics and Metadata

   The 134 participants included 50 males and 84 females in age ranging
   from 45.6 to 65.9 years (mean ± SD: 58.6 ± 4.9) and an average BMI of
   28 ± 5.9 kg/m^2 (mean ± SD). Predominant prescription medications
   analyzed were used by >10 participants of the cohort and included
   angiotensin II receptor blockers (ARBs) and statins (n = 41, 30.6%)
   followed by angiotensin-converting enzyme (ACE) inhibitors (n = 33,
   24.6%), nonsteroidal anti-inflammatory medications (NSAID, n = 31,
   23.1%), diuretics (n = 28, 20.9%), calcium channel blockers (n = 24,
   17.9%), β-blockers (n = 17, 12.7%), metformin (n = 17, 12.7%), PPI (n =
   15, 11.2%), and levothyroxine (n = 15, 11.2%) ([68]Table [69]1
   ). Although other medication use was reported (e.g., opioid analgesic,
   sulfonylureas, cannabinoid, DPP_4_inhibitors, and xanthine oxidase
   inhibitors), these medications were excluded from the downstream
   association analysis due to their low prevalence of use (<10
   participants). Different frequencies of medication use were observed
   when participants were stratified by age, BMI, and sex ([70]Figure
   [71]1A). For example, statins, ACE inhibitors, and diuretics were more
   frequently used by older individuals (>60 years) compared to those in
   midlife (40–59 years). Likewise, ARBs and NSAIDs were more frequently
   used by obese vs normal BMI and overweight participants. High blood
   pressure (HBP) was the most diagnosed condition (108/134) in the cohort
   with the highest number of prescribed medications. Clustering of
   metadata indicates medication-specific effects as well as different
   effects from age, BMI, sex, and diet ([72]Figure [73]1B), which were
   included for statistical correction in the following analysis. For
   example, metadata cluster analysis showed that statins were used more
   by older individuals, but had a stronger correlation to total number of
   medications used ([74]Figure [75]1B). Levothyroxine was also more used
   by older individuals ([76]Figure [77]1A) and was clustered with age.

Figure 1.

   Figure 1
   [78]Open in a new tab

   Number of medications used based on different categories (A) and
   cluster of variables based on individual metadata (B).

Individual Microbiota Variation

   Microbiota composition was characterized by a highly individual
   variation, which was driven by the relative abundance of dominant
   phyla, particularly Firmicutes and Bacteroidetes ([79]Figure S2A). The
   relative abundance of Firmicutes ranged from 24.1 to 87.6% (median of
   50.8%), while Bacteroidetes varied from 3.7 to 71% (median 33.2%). The
   individual variation was further supported by the Shannon index
   measurement (ranging from 1.92 to 4.3 (median 3.66, [80]Figure S2B).
   The relative abundances of Firmicutes and Actinobacteria were
   positively correlated with Shannon index (Spearman rho 0.4 and 0.36,
   respectively, P < 0.0001, [81]Figure S2C,D). These results indicate the
   phylum-level variation as a major factor contributing to the
   interindividual variation. At the genus level, individual variation was
   also observed for the dominant taxa, such as Blautia, Dialister, and
   Bacteroides ([82]Figure S3).

Gut Microbiota Composition Associations with Single Medications

   To determine how individual medications related to the overall
   microbiota structure, we assessed the Bray–Curtis distance with
   variations explained by 12 medications that were taken in at least 10
   participants. At a false discovery rate of <0.1, selective serotonin
   reuptake inhibitor (SSRI) antidepressants and ACE inhibitors were
   associated with microbiota composition as measured by Bray–Curtis
   distance ([83]Figure [84]2
   A), while NSAIDs were significantly associated with Shannon diversity
   ([85]Figure S4).

Figure 2.

   [86]Figure 2
   [87]Open in a new tab

   Multivariate association analyses between bacterial taxa and single
   medications. (A) The explained variance in Bray–Curtis distance by
   different types of medications. *FDR P < 0.05, # FDR 0.05 < P < 0.1
   from adonis2 analysis. (B) Significant associations between medications
   and bacterial taxa at the phylum/genus level. (C) Significant
   associations between medications and bacterial taxa at the species
   level. The plus symbol represents significantly positive association,
   while the minus symbol represents significantly negative association.
   For (B),(C), significance was defined as FDR P < 0.25 by MaAsLin2.

   We then performed multivariate association analyses between bacterial
   taxa and medications. At phylum and genus levels, a total of 32
   associations were identified when corrected for age, sex, BMI, and diet
   (FDR < 0.25, [88]Figures [89]2
   B and [90]S5). ARB use was positively associated with Streptococcus and
   statin use was positively associated with Alistipes and Ruminococcus.
   Metformin use was positively associated with Esherichia/Shigella and
   negatively associated with Ruminococcus, Dorea, and Romboutsia.
   β-blocker use was negatively associated with Faecalibacterium and
   Collinsella. Levothyroxine use was negatively associated with
   Faecalibacterium and Bacteroides. PPI use was negatively associated
   with Eggerthella and Dorea.

   To understand the relationship between bacterial species and
   medications, we further analyzed the association between ASVs and
   medications, which was combined with a blast to enable species-level
   resolutions. A total of 14 associations were identified when corrected
   for age, sex, BMI, and diet (FDR < 0.25, [91]Figures [92]2
   C and [93]S5). At the species level, metformin use was negatively
   associated with Bacteroides xylanisolvens and Dialister invisus. PPI
   use was positively associated with unclassified Lachnospiraceae and
   Alistipes. 5-Aminosalicylic acid (5-ASA) use was positively associated
   with Bacteroides thetaiotaomicron and Barnesiella intestinihominis,
   while negatively associated with unclassified Clostridium XIVa.
   β-Blocker use was negatively associated with Akkermansia muciniphila
   and E. lenta. SSRI antidepressant use was positively associated with
   Bacteroides thetaiotaomicron and Faecalibacterium prausnitzii
   ([94]Figure [95]2C). These results indicate the distinct associations
   between a single medication and gut microbiota composition.

Potential Microbial Functionalities Associated with Single Medications

   To gain functional insights into the associations identified, a
   metagenomics function prediction was performed with the available ASV
   data from the present cohort. Tax4Fun found a total of 5293 KEGG
   Orthologies (KOs). Using the multivariate association algorithms in
   MaAsLin2, 745 KOs were found to be associated with β-blockers,
   levothyroxine, PPI, SSRI antidepressant, 5-ASA, and metformin
   ([96]Figure [97]3
   A), characterized by the significant enrichment of [98]K01997
   (branched-chain amino acid transporter permease) by β-blockers. A total
   of 697 KOs were higher in medication use that were enriched in pathways
   including the biosynthesis of branched-chain amino acids, pyrimidine
   metabolism, peptidoglycan biosynthesis, and drug metabolism—other
   enzymes ([99]Figure [100]3B).

Figure 3.

   Figure 3
   [101]Open in a new tab

   Metagenomics functional prediction of KEGG orthologies (KOs). (A)
   Multivariate association analysis of KOs with different medications.
   All medications with significant associations (FDR < 0.25) are
   visualized. (B) Pathway enrichment analysis of the upregulated (e.g.,
   positive association) or downregulated (e.g., negative association)
   KOs. (C) Associations between medications and encoded KOs.

   Next, we were interested in determining whether the discriminant KOs
   might be present in the genomes of strains that exhibited strong
   positive/negative association with specific medications as shown in
   [102]Figure [103]3
   C. Corresponding to the negative association between β-blockers and E.
   lenta, the enzymes that negatively associated with β-blockers were
   found in the representative strain E. lenta DFI.6.2, including
   formate-tetrahydrofolate ligase (EC:6.3.4.3), citrate synthase,
   aconitate hydratase (EC:4.2.1.3), and glutamate synthase (NADPH) large
   chain (EC:1.4.1.13) ([104]Figure [105]3C). Similar alterations were
   found for 5-ASA, which were positively associated with
   N-acetylneuraminate synthase (EC:2.5.1.56) that was present in B.
   thetaiotaomicron strain ATCC 29148. Metformin use was negatively
   associated with the NitT/TauT family transport system substrate-binding
   protein that was present in B. xylanisolvens 2789STDY5608839
   ([106]Figure [107]3C). These results further implicate the functional
   association between specific microbes and medication use.

Multimedication Use Associations with Microbial Composition

   To further understand the impact of multimedication use, we explored
   the frequency of different medication combinations. In the present
   cohort, dual medication combinations were most observed, with at least
   10 participants having 6 specific pairs. A total of 15, 14, 14, 13, 13,
   and 10 participants were described as users of ARBs + NSAIDs, statin +
   ACE inhibitor, statin + ARBs, statin + metformin, statin + NSAIDs, and
   statin + PPI, respectively, as visualized in [108]Figure [109]4
   A. To increase the statistical power using unbalanced replicates from
   medication users and nonusers, medication pairs present in at least 10
   participants were employed for the following multimedication
   association analysis. Triple medication use was also observed, but was
   not included in the analysis due to the low frequency of participants
   in each group. However, as increasing medication use may affect the
   overall microbial composition,^[110]31 we assessed the relationship
   between microbial α-diversity and number of medications used. Linear
   regression analysis identified a significant negative correlation
   between the number of medications used and Shannon index (linear
   regression coefficient −0.05, r^2 = 0.056, P = 0.005, [111]Figure
   [112]4B), indicating that the increased use of medication changes the
   microbial diversity. When comparing the effects of medication
   combination on β-diversity, statin + ARB was significantly associated
   with microbiota composition, as measured by Bray–Curtis distance
   ([113]Figure [114]5A).

Figure 4.

   [115]Figure 4
   [116]Open in a new tab

   Graphical representation of medication combinations and their
   association with the Shannon index. (A) Network visualization of dual
   medication combinations. Combinations of 3+ medications are not shown
   due to low frequency. Color intensity indicates frequency of specific,
   dual medication combinations. (B) Linear regression analysis of the
   relationship between number of medications used and Shannon index
   (linear regression coefficient −0.05, r^2 = 0.056, P = 0.005).

Figure 5.

   Figure 5
   [117]Open in a new tab

   Multivariate association analyses between bacterial taxa and dual
   medication use. (A) The explained variance in Bray–Curtis distance by
   different combinations of medications. *FDR P < 0.05 from adonis2
   analysis. (B) Significant associations between multimedication and
   bacterial taxa. A plus symbol represents a significant positive
   association, while the minus symbol represents a significant negative
   association. Significance was defined as FDR P < 0.25 by MaAsLin2..

   Next, we performed the multimedication association analysis with the
   microbial composition. A total of 20 associations were identified when
   corrected for age, sex, BMI, and diet (FDR < 0.25, [118]Figures [119]5
   B and [120]S5). Among the associations identified, ARB + NSAID was
   associated with seven features, including negative associations with
   Ruminococcus, Bacteroidetes, Alistipes, and B. intestinihominis. Statin
   + metformin was positively associated with Eshcherichia/Shigella, and
   Proteobacteria, and negatively associated with Butyricicoccus,
   Romboutsia, and D. invisus. Statin + NSAID were positively associated
   with A. muciniphila and negatively associated with Eggerthella.
   Overall, the combination of statin + metformin and ARB + NSAID imposed
   a more profound impact on the microbial composition compared to other
   dual medication combinations.

Discussion

   Effective management of CMD often necessitates the utilization of
   multiple medications. For this reason, we examined the impact of both
   individual and multimedication use on gut microbiota structure,
   composition, and function in CMD. Employing multivariable association
   analysis strictly controlled for age, sex, BMI, and diet, we found many
   novel associations between bacterial taxa and medications.

   To gain insight into differences in the gut microbial community
   structure with individual medications, the Bray–Curtis distances were
   calculated. Results showed variable impacts of medications, with the
   largest changes observed with SSRI antidepressants. This finding is in
   line with studies showing SSRI to exert antibiotic effects, disrupting
   both the integrity and stability of the gut microbiota.^[121]32,[122]33
   Although SSRIs are not commonly classified as medications for CMD,
   there exists a firmly established, two-way connection between
   depression and CMD with individuals with CMD being 1.5 times more
   likely to be diagnosed with depression compared to healthy
   controls.^[123]34,[124]35

   In the present study, many novel associations between medications and
   gut microbiota at both the genus and species levels were identified.
   Those of particular interest included the use of β-blockers and statins
   that were negatively correlated with A. muciniphila. A. muciniphila is
   a keystone mucin-degrading bacteria with metabolic and immunological
   benefits in the gut.^[125]36 Likewise, administration of pasteurized
   (but not live) A. muciniphila for 3 months improves insulin sensitivity
   and dyslipidemia in obese individuals with metabolic syndrome.^[126]37
   Given these findings, it is worthwhile for future studies to test the
   protective effect of A. muciniphila supplementation when these
   medications are prescribed. Results also found a positive association
   between metformin and Escherichia/Shigella. This finding is consistent
   with reports of gastrointestinal side effects, reduced diversity, and
   elevated Escherichia species with metformin use.^[127]38,[128]39
   Nevertheless, in Type 2 Diabetes Mellitus, metformin may exhibit
   positive impact on the gut microbiome by enhancing the growth of
   short-chain fatty acid-producing bacteria, including Bacteroides,
   Butyricoccus, and Bifidobacterium.^[129]40 To this end, the impact of
   metformin needs to be evaluated in disease-specific and multimedication
   contexts.

   Next, we used functional metagenomics prediction via KEGG orthologies
   to examine how microbial function was impacted by CMD medication use.
   As expected, medications resulted in the alteration of several
   microbial functions and pathways. For pathways, CMD as a collective
   group had the most profound impact on the biosynthesis of amino acids.
   Disturbances in branched-chain amino acids and amino acid profiles are
   known to be associated with CMD,^[130]41 and this finding prompts the
   question of whether medication-induced changes in microbiota also play
   a role. Examination of genes and proteins affected and their relation
   to specific microbes uncovered changes in the representative taxa E.
   lenta with β-blocker use. In particular, β-blockers selectively reduced
   the relative abundance of predicted enzymes, including citrate
   synthase, glutamate synthase, and aconitate hydratase, indicating that
   energy metabolism was likely affected. Of note, these functions were
   unaffected by other medications. E. lenta is an important microbe
   exerting diverse metabolic activities toward bile acids,
   medications,^[131]42 and short-chain fatty acids, particularly acetate
   for energy metabolism.^[132]43 Taken together, these results suggest
   that E. lenta survival and function may be compromised by β-blocker
   use.

   As polypharmacy is common in CMD, we also examined the impact of
   multimedication use on the gut microbiota. Increasing medication use
   (up to 6) resulted in reductions in microbial α-diversity as assessed
   by the Shannon index. Similar findings have also been reported in
   patients with irritable bowel syndrome.^[133]31 Collectively, it is
   thought that increasing medication use alters the luminal
   microenvironment and favors the selective colonization of microbes with
   high adaptability. As high diversity is linked to greater microbial
   stability and resilience,^[134]44 future research could consider adding
   a multistrain probiotic to those taking multiple medications for CMD.
   There is emerging evidence to suggest that specific probiotics can
   improve symptomology related to CMD.^[135]45,[136]46

   Beyond the previously discussed medications, the current data set also
   presented an opportunity to investigate specific combinations of dual
   medications. This type of analysis has rarely been undertaken in the
   literature. Predictably, statins were the most prescribed medication in
   our cohort, and as such, most combinations involved statins with
   another medication. Results show that the combination of statins and
   ARBs was the most disruptive in terms of bacterial community structure
   as determined by the Bray–Curtis distance. Despite the observed gut
   microbiota disturbance, this combination is highly effective with
   protective effects on both cardiovascular mortality and
   morbidity.^[137]47−[138]49

   Examination of significant medication-microbe associations with dual
   medication use highlights the combination of statins and metformin as
   noteworthy. This combination displays several important relationships,
   including negative associations with Romboutsia and Butyriciocccus.
   Reduced Romboutsia abundance in hypertension has been previously
   demonstrated in the HELIUS study,^[139]50 but it is unclear whether
   underlying condition(s) and/or medication were to blame as the
   association existed with and without correction for antihypertensive
   medication use. This medication combination also reduced Butyriciocccus
   likely indicating reduced butyrate production, a microbe known to have
   several beneficial properties that are essential to maintaining
   gastrointestinal health.^[140]51

   This study boasts several notable strengths, such as meticulous control
   over various confounding factors that are recognized for their
   influence on the gut microbiota. These encompass age, BMI, sex, and
   diet. Additionally, results explore the effects of single, dual, and
   multimedication use. Limitations include the correlational nature of
   the analysis and a lack of detailed information regarding the
   medication, specifically the prescription duration, dosage, and
   formulation. For instance, quick-release or extended-release formats
   could potentially exert a significant influence on how a drug affects
   the gut microbiota. Another limitation is the functional prediction of
   gut microbiota. In the future, fecal metabolomics profiling could be
   used to confirm functional changes including alterations to short-chain
   fatty acids, branched-chain fatty acids, and other metabolites.

   In conclusion, the results reveal several novel medication–microbiota
   interactions as well as the influence of many common dual medication
   combinations. These discoveries enhance our understanding of the
   intricate interplay between pharmaceuticals and the gut microbiota in
   the context of CMD. Given the profound impact of the gut microbiota on
   overall health, finding a balance between medication usage and
   maintaining a healthy microbial structure could offer valuable insights
   for more effective disease management.

Acknowledgments