Multi-Omic Analysis Reveals Microbiome-Related Relationships Between Cocaine Use and Metabolites

Abstract

Objective:

Over 19 million individuals globally have a cocaine use disorder, a significant public health crisis. Cocaine has also been associated with a pro-inflammatory state and recently with imbalances in the intestinal microbiota as compared to non-use. The objective of this pilot study was to characterize the gut microbiota and plasma metabolites in people living with HIV (PLWH) who use cocaine compared with those who do not.

Design:

Cross-sectional study

Methods:

A pilot study in PLWH was conducted on 25 cocaine users and 25 cocaine non-users from the Miami Adult Studies on HIV (MASH) cohort. Stool samples and blood plasma were collected. Bacterial composition was characterized using 16S rRNA sequencing. Metabolomics in plasma were determined using gas and liquid chromatography/mass spectrometry.

Results:

The relative abundances of the Lachnopspira genus, Oscillospira genus, Bifidobacterium adolescentis species, and Euryarchaeota phylum were significantly higher in the cocaine using PLWH compared to cocaine non-using PLWH. Cocaine use was associated with higher levels of several metabolites: products of dopamine catabolism (3-methoxytyrosine and 3-methoxytyramine sulfate), phenylacetate, benzoate, butyrate, and butyrylglycine.

Conclusions:

Cocaine use was associated with higher abundances of taxa and metabolites known to be associated with pathogenic states that include gastrointestinal conditions. Understanding key intestinal bacterial functional pathways that are altered due to cocaine use in PLWH will provide a better understanding of the relationships between the host intestinal microbiome and potentially provide novel treatments to improve health.

INTRODUCTION

Drug use disorders are among the costliest health problems in the United States [1]. Illicit drug use plays a major role in the HIV epidemic, with a high percentage of the HIV-infected population having a drug use disorder [2]. Survival rates of people living with HIV (PLWH) who use drugs are significantly lower than non-users, largely due to inequities in access to care [3,4]. According to the United Nations, over 19 million individuals globally used cocaine in 2018, a significant public health crisis [5]. During 2008–2017, the disability-adjusted life years globally for overall drug use disorders increased by 24% and for cocaine use disorders by 17% [6]. Cocaine was the second leading cause of death from illicit drugs in Florida in 2018 and the number of cocaine-related deaths have more than doubled in Miami-Dade County since 2004 [7], the location where this study was conducted.

Cocaine is commonly used by PLWH in Miami and among the participants of the Miami Adult Studies on HIV (MASH) cohort [8]. Cocaine has recently been associated with intestinal microbiota dysbiosis or an imbalance in the intestinal flora [9,10], and with systemic inflammation in HIV infection [11]. Disturbances in the gut microbial environment in HIV infection may be associated with immune deficiencies and inflammation [12]. Lasting inflammation during HIV infection may also impact the intestinal composition of the microbiota [13–15]. Cocaine users have higher levels of inflammation [16], lower Healthy Eating Index scores, and lower food security, which may also reduce the diversity of the dietary intake affecting the exposure to different substrates for the gut microbiota [9]. Cocaine use has been linked to elevated levels of Interleukin (IL)-6, a proinflammatory cytokine [17–19]. In cocaine use disorders, IL-6 may also be a biomarker of diminished clinical outcomes related to behavior and neuronal responses due to its influence on the central nervous system signaling [20,21]. Studies have shown differences in the relative abundance of Bacteroidetes in cocaine users and PLWH [9] and cocaine use in rats was associated with decreased community richness and diversity of the gut microbiome [22]. Cocaine provided to mice led to changes in the gut-barrier through alterations of the tight junction proteins and increased epithelial permeability [10]. However, the mechanisms of how cocaine affects intestinal microbiota composition and distribution and inflammation in HIV infection are not known. Studies on the effects of cocaine on the gut microbiome and metabolome are lacking in humans.

Communication between the host and the microbiome occurs through metabolites, which may have an effect on the host physiology [23] and contribute to the maintenance of intestinal homeostasis through their interaction with the inflammasomes and promote or suppress proinflammatory cytokine secretion [23]. Metabolites may also have immunomodulatory effects on the body through their effect on development, differentiation, and activity of the immune system [24]. Distinct plasma metabolite profiles have been demonstrated in HIV [25]; yet, data on microbiome-related metabolites due to cocaine use in PLWH is sparse.

Therefore, the objective of this pilot study was to characterize gut microbiota and plasma metabolites in PLWH who use cocaine compared with those who do not. The understanding of this relationship is crucial to the development of new therapies for reducing inflammation and disruption of microbiota homeostasis. We hypothesized that cocaine use compared to non-use significantly impacts intestinal composition.

METHODS

Design and Participants

The Miami Adult Studies on HIV (MASH) cohort was the source of participants. Participants are recruited and followed by National Institute on Drug Abuse (NIDA) grant “Cohort Studies in HIV/AIDS and Substance Abuse, 5U01DA040381,” a prospective, longitudinal observational cohort study that follows the MASH cohort. This study was cross-sectional, enrolling 25 PLWH who used cocaine and 25 PLWH who did not use cocaine that were randomly chosen from the MASH cohort, who mainly use non-injection crack/cocaine. Participants were eligible for the pilot study if they had data to show consistent cocaine use or non-drug use for at least one year, diagnosed with HIV, 35–66 years old, were not living with hepatitis B virus (HBV) or hepatitis C virus (HCV), and were not on antibiotic treatment within 3 months of stool sample collection.

Assessments

After obtaining informed consent, data on demographics, socioeconomic status, substance use, parameters of HIV disease progression (CD4 cell count and HIV viral load) and reported sex with men by male participants (MSM) were collected from the MASH cohort. Cocaine use and non-use were determined through consecutive self-reported questionnaires and/or positive urine toxicology within the previous year. Additionally, plasma metabolomics was used to confirm the self-report and urine toxicology groups. At the time of fecal sample collection questionnaires on antibiotic use and dietary intake were also completed. Dietary data was collected using at least 4 individual 24-hour recalls that were collected at three-month intervals with the last recall collected on the date the fecal sample was received at the research clinic. Diet quality was assessed using the USDA Healthy Eating Index (HEI)-2015, with higher scores indicating better adherence to the US Dietary Guidelines (total maximum score is 100) [26]. Fasting blood plasma was used from the MASH cohort’s specimen repository that was collected within 3 months of the follow-up visit to determine IL-6 levels using the Quantikine ELISA assay from R&D systems (Minneapolis, MN).

Fecal Sample Collection

Stool samples from one bowel movement were collected using the Norgen Biotek kit (Thorold, ON, Canada) provided to the participants during their assessment visits to the research clinic and returned the sample as part of their follow-up visit for the study, within 5 days. The kit included all the materials for the fecal collection including the Fe-Col® collection paper and nucleic acid collection and preservation tube that allows for the collection of up to 2 grams of specimen and room temperature storage. This collection system has been shown to be effective in minimizing microbiota compositional shifts that may reduce bias in samples collected at ambient temperatures [27]. The samples were sent in one batch to the processing laboratory overnight with ice packs.

DNA Extraction, PCR Amplification and High-throughput Sequencing

16Sr RNA sequencing was completed by the University of North Carolina Microbiome Core. All samples were processed within the same batch and using the same lot of reagents. Total DNA from the fecal samples, 12.5 nanograms, was used to amplify the V4 region of the bacterial 16S rRNA gene using universal primers [28,29]. Purification of each 16S rRNA gen amplicon using the AMPURE XP reagent (Beckman Coulter, Indianapolis, IN). A limited cycle Polymerase Chain Reaction (PCR) program was then applied to each sample being amplified and Illumina sequencing adapters and dual-index barcodes (index 1(i7) and index 2 (i5) (Illumina, San Diego, CA) were added to the amplicon target. AMPURE XP reagent was used again to purify the final libraries. The libraries were then quantified and normalized prior to pooling. The final libraries were again purified using the AMPure XP reagent (Beckman Coulter), quantified and normalized prior to pooling. The DNA library pool was then denatured with NaOH, diluted with hybridization buffer and heat-denatured before loading on the MiSeq reagent cartridge (Illumina) and on the MiSeq instrument (Illumina). Automated cluster generation and paired–end sequencing with dual reads were conducted according to the manufacturer’s instructions.

Metabolomics Analysis

Fasting blood plasma, 200 μL, was sent to Metabolon (Metabolon Inc., Morrisville, NC) for non-targeted gas chromatography/mass spectrometry (GC-MS) and liquid chromatography/mass spectrometry (LC-MS)-based profiling of metabolomics. Our study had 1,043 metabolites of known identity detected.

Bioinformatic and Statistical Analyses

Sequencing output from the Illumina MiSeq platform were changed to fastq format and demultiplexed using Illumina Bcl2Fastq 2.18.0.12. The resulting paired-end reads were processed using QIIME 2 2018.11 [20]. Index and linker primer sequences were trimmed using the QIIME 2 invocation of cutadapt. The resulting paired-end reads were processed with DADA2 through QIIME 2 including merging paired ends, quality filtering, error correction, and chimera detection. Amplicon sequencing units from DADA2 were assigned taxonomic identifiers with respect to Green Genes release 13_08.

The read counts of amplicon sequence variants (ASVs) were retrieved from QIIME 2 for all samples using R package phyloseq v. 1.38.0 [31]. The unit was considered present if at least five reads were aligned to it. The ASVs with a prevalence of less than 20% of all samples were filtered out, resulting in 287 ASVs. The relative microbial abundance was extracted such that the sum of values for each sample adapts to one. At the same time, the abundance profile was normalized using centered log-ratio (CLR) transformation [32]. The CLR approach is widely used in microbiome studies and has been shown to reduce the effect of false feature dependencies arising from compositional nature abundance profiles [33–35]. The normalizations were performed with R package microbiomeMarker v. 1.1.2 [36]. Observed ASV richness, Shannon and Simpson alpha diversity indices, were calculated using phyloseq v. 1.38.0 R package [31]. Wilcoxon rank-sum test was used to compare indexes between cocaine use and non-use groups and differences in taxa, metabolites, and IL-6. The boxplots were constructed with the ggpubr v. 0.4.0 R package [37]. A stacked bar plot of the relative microbial abundance was constructed with ploty v. 4.10 R package [38]. Linear discriminant analyses with effect size measurements (LEfSe) analysis compared the relative abundance of bacterial taxa between groups using microbiomeMarker v. 1.1.2 [36,39].

Data from the metabolomics analyses used the original scale, and missing data were imputed using the values with the minimum. The log transformation and imputation of missing values with the minimum observed value were performed for each metabolite. Two-way ANOVA was used to identify metabolites exhibiting significant interaction and main effects for experimental parameters of cocaine use and sex. The p-values were adjusted by Benjamini-Hochberg false discovery rate (FDR) method and referred to as q-values. Differences with q-values < 0.10 were considered significant. Spearman’s rank correlations were conducted to calculate the correlation between IL-6 and bacterial taxa.

Permutational multivariate analysis of variance (PERMANOVA) was performed with Vegan R package version 2.6.2 [40]. Principal coordinates analyses (PCoA) were completed with stats R package v. 4.1.0. The UniFrac distances used to compute PCoA components were calculated using the mia v. 1.0.8 R package [41].

Descriptive statistics such as frequencies, percentages, medians, and interquartile ranges were used to characterize the participants. The relative abundance of OTUs for all analyses was retrieved from QIIME 2. The Dirichlet-Multinomial (DM) distributions of ASV counts were parameterized with HMP v. 2.0.1 R package [42]. The Method of Moments (MOM) estimation was used to obtain Likelihood-Ratio-Test Statistics and the corresponding p-value for comparing DM distributions of cocaine users and non-users [43]. The Monte-Carlo simulation procedure performed the post-hoc power analysis with 1000 experiments [42]. The simulations were executed with MC.Xdc.statistics function of HMP package, where inputs were estimated DM parameters and number of reads (10,000 per sample).

RESULTS

Demographics

After obtaining informed consent, 50 participants from the MASH cohort who met the eligibility criteria were enrolled into the pilot study. The characteristics of the participants enrolled are displayed in Table 1. There were no significant differences between the cocaine use and cocaine non-use groups in age, sex, education, % on antiretroviral therapy, CD4 cell count, HIV viral load, or % of MSM. Although the data is not shown, there were no significant differences between groups by ART regimen. Significant differences were seen between the groups in race/ethnicity, with a greater proportion of the participants in the cocaine use group being non-Hispanic Blacks (P=0.016). Cocaine users had less annual income (P=0.040) and a lower BMI (P=0.018) compared to the non-user group. Additionally, cocaine users had a significantly poorer diet quality (P=0.048), however both groups had overall low diet quality scores that do not align with the Dietary Guidelines for Americans [44]. There were no significant differences in the total calories, total fat, or whole grains consumed between the groups (data not shown).

Table 1:

Characteristics of the Population

Variables	PLWHa-Cocaine Users	PLWHa-Cocaine Non-Users	P-Value
Age, median years (IQR)	56 (52–59)	54 (50–62)	0.414
Sex, N (%)	10 (60)	14 (56)	0.774
Female	15 (40)	11 (44)
Male
Race/Ethnicity, N (%)
Non-Hispanic Black	22 (88)	13 (52)	0.016
Non-Hispanic White	0 (0)	3 (12)
Hispanic White	3 (12)	9 (36)
Education, median years (IQR)	12 (10–13.5)	13 (10.5–15)	0.287
Annual Income, median $ (IQR)	9,443 (8.358–10,564)	16,108 (8,646–21,907)	0.040
On Antiretrovirals, N (%)	25 (100)	24 (96)	0.312
CD4 cell count, median cells/μL (IQR)	517 (405.5–689)	585 (396–884)	0.244
Undetectable HIV viral load, N (%)	11 (42)	18 (72)	0.089
Men who have sex with men (MSM), N %	6 (24)	8 (32)	0.397
BMI, median kg/m2	26 (23.9–30.9)	31.4 (26.1–36.7)	0.018
Healthy Eating Index, median score (IQR)	41 (35–48)	45 (41–57.5)	0.048

^aPeople Living with HIV (PLWH)

Bacterial Diversity and Composition by Cocaine Use

The LEfSe tool was applied to identify the most differentially abundant taxa between PLWH who used cocaine and non-users (Figure 1). The relative abundances of the Lachnospira genus, Bifidobacterium adolescentis species, and Euryarchaeota phylum were significantly higher in the cocaine users relative to non-users. On the other hand, cocaine non-users had significantly higher abundance of Enterobacteriaceae and Muribaculaceae (S24–7) families compared to cocaine users. Stacked bar charts illustrating the relative abundance on the genus level by cocaine use and non-use groups are shown in Figure 2. In general, Lachnospira and Oscillospira were more abundant in the cocaine use groups compared to the non-use group.

Figure 1:

LEfSe identified the most differentially abundant taxon between cocaine users and non-users. Taxa relatively enriched in cocaine users are indicated as turquoise color, and taxa enriched in the non-users group are shown in red. Only taxa with an LDA significant threshold of >2 are shown. LDA=Linear discriminant analysis

Figure 2: Metagenomics profile.

Stacked bar plot showing the microbial composition (relative abundance) on genus level. Organisms with a relative abundance of less than 1% were combined with the “other” label.

Observed ASV richness, Shannon, and Simpson alpha diversity measures were calculated for the fecal bacterial community and compared by cocaine use and non-use groups (see Figure, Supplemental Digital Content 1, which illustrates the diversity measures by groups). The observed richness shows the counts of unique sequence variants [45]. The Shannon index reflects the weighted geometric mean of the proportional abundances [46]. Simpson alpha diversity measures the probability that two units randomly selected from a sample will belong to the same ASV [47]. The alpha diversity measures were not significantly different between the groups for the observed richness (P=0.31), Shannon (P=0.60) and Simpson indices (P=0.79) (see Figure, Supplemental Digital Content 1, which illustrates the diversity measures by groups). Furthermore, no significant differences in the Shannon (P=0.45) and Simpson indices (P=0.37) were seen on the basis of sex (data not shown). PCoA shows clustering of the cocaine use groups (see Figure, Supplemental Digital Content 2, which illustrates the PCoA plot by groups) and PERMANOVA (see Table, Supplemental Digital Content 3A, which demonstrates the PERMANOVA statistics) showed a trend towards significance in the difference between clustering of cocaine users and non-users (P=0.066). The Dirichlet-Multinomial Models of ASV frequencies were not significantly different between cocaine user and non-user groups (likelihood ratio test P=0.55). The ratio of Firmicutes/Bacteroidetes was not significantly different between groups (see Figure, Supplemental Digital Content 4, which demonstrates the boxplots of Bacteroidetes/Firmicutes ratio by groups) (P=0.92).

Due to the important role of short-chain fatty acids such as butyrate on gut health, we also examined the level of taxa associated with producing butyrate. Butyrate producers that included the genus Clostridium, Roseburia, Ruminococcus, Faecalibacerium, Eubacterium and Fusobacterium were not significantly more abundant within cocaine users relative to non-users. Nevertheless, Lachnospira was significantly overabundant in cocaine users compared to non-users (P=0.03) (see Figure, Supplemental Digital Content 5, which illustrates the boxplots of CLR values for butyrate-producing bacteria genera by groups), which were also selected as the biomarker for cocaine usage by the LefSE analysis (Figure 1).

While this is a pilot study, we performed the post-hoc power analysis to estimate the number of samples needed in a full-scale study to observe significant differences in the microbiome between cocaine usage groups (Table 2). Parameters of the Dirichlet-Multinomial Model for observed ASV counts were used as the basis for test comparisons [42]. In the settings of the current study (50 individuals), the observed power was as small as 0.004 (Table 2, column 2). We show that by increasing the number of participants to 110, the power would improve up to 0.86 (Table 2, column 4), and 140 participants would lead to a significant power of 0.99 (Table 2, column 7).

Table 2.

Post-hoc power calculations based on Dirichlet-multinomial parameters of metagenomic profiles to observe differences in the fecal microbiome between cocaine users and non-users. Each column represents the number of samples (N) needed for a full-scale study to achieve the power listed in the table.

	N=50	N=100	N=110	N=120	N=130	N=140
power (Type II error)	0.004	0.662	0.859	0.948	0.985	0.997

Inflammation

The nonparametric Wilcoxon rank-sum test showed, Interleukin-6 (IL-6), a measure of systemic inflammation, was higher in the cocaine users relative to non-users, albeit not statistically significant (P=0.15) (see figure, Supplemental Digital Content 6, which demonstrates the difference in IL-6 by groups). Eubacterium (r=0.36, P<0.05) and Clostridium (r=0.30, P<0.05) genera found in cocaine users were significantly correlated with IL-6.

Cocaine Use and Metabolites

Cocaine use compared to non-use by sex (female vs. male) with two-way ANOVA was associated with higher levels of the cocaine-derived metabolites, including benzoylecgonine and norbenzoylecgonine (P<0.05) (Figure 3). Benzoylecgonine is a primary product of cocaine, which is metabolized by tissue esterases and spontaneous conversion [48]. Additionally, cocaine users had higher levels of 3-methoxytyrosine and 3-methoxytyramine sulfate (P<0.05), products of dopamine catabolism (P<0.05) (see Figures, Supplemental Digital Content 7, which illustrate the differences in metabolites by groups and sex). Cortisone, a metabolite of cortisol associated with stress response, showed a trend towards significance (P<0.10) (data not shown).

Figure 3: Validation of Treatment Groups.

Metabolites of cocaine catabolism significantly higher by cocaine use compared to non-use in females and males (P<0.05 two-way ANOVA (Analysis of variance)). F=female and M=males

Several microbiome-associated metabolites were significantly elevated in the cocaine use groups (see Figures, Supplemental Digital Content 3, which illustrate the differences in metabolites by groups and sex). Cocaine users, compared to non-users, had significantly higher levels of amino acid derivatives such as phenylacetate (P<0.05), a product of phenylalanine catabolism. Benzoate was significantly elevated in cocaine users compared to non-users (P<0.05). The short-chain fatty acid butyrate and butyrylglycine, an intermediate formed in beta-oxidation reaction with butyryl-CoA, was higher in cocaine users compared to non-users, (P<0.05). The PERMANOVA analysis did not show a significant difference in metabolomics profile between the groups based on cocaine usage (P=0.66) (see Figure, Supplemental Digital Content 3B, which illustrates the PERMANOVA analysis for metabolites).

DISCUSSION

We investigated the effect of cocaine use on the gut microbiome in PLWH using a multi-omics approach. This study has shown that PLWH who use cocaine compared to non-users had more abundant taxa that may have the potential for pathogenicity, especially gastrointestinal disorders. Moreover, Eubacterium and Clostridium genera were significantly associated with the pro-inflammatory cytokine, IL-6. However, neither of these bacteria were found to have a significant differential relative abundance between the cocaine use compared to non-use groups. Lachnospira genus, which is part of the Lachnospiraceae family, is associated with some health-promoting effects as well as intra- and extraintestinal diseases [49] was overabundant in cocaine users compared to non-users. Oscillospira genus and Euryarchaeota phylum were more abundant in cocaine users are linked to intestinal dysfunction within the literature [50–52]. In addition, cocaine users had higher levels of products of dopamine catabolism, phenylalanine catabolism, benzoate and the short-chain fatty acid butyrate compared to non-users. These products of microbiome metabolism have been previously associated with accumulation of lipids in the liver [53], small intestine inflammation, and abnormal gut permeability [54].

Our findings were different to those by Volpe et al. [9] conducted in 15 PLWH and 17 HIV un-infected; 7 people living with HIV who used cocaine compared to 7 HIV un-infected participants who also used cocaine. This was the first study conducted in humans and showed higher relative abundance of Bacteroidetes phyla in all cocaine users than non-users, and higher abundance of Proteobacteria phyla in PLWH than in HIV negatives [9]. We did not find a significant difference by major phyla by cocaine use; however, we found a greater abundance of Euryarchaeota, a major phylum of Archaea, in cocaine using PLWH. Much of the taxa and metabolites found to be enriched in cocaine users have a relationship with gastrointestinal problems, which is common in PLWH and with cocaine use [10,55,56]. Additionally, we also examined the influence of cocaine on microbiome-related metabolites. The clinical significance of our findings will need to be examined in a larger sample overtime.

A higher relative abundance of Bifidobacterium adolescentis species, Oscillospira genus, and Euryarchaeota phylum were present in the PLWH who used cocaine. Bifidobacterium adolescentis, a species within the Bifidobacteria genus and Actinobacteria phylum, which are anaerobic bacteria, is associated with the capability to stimulate Gamma-Aminobutyric acid (GABA) production, a neurotransmitter [57,58]. In the literature, Bifidobacterium adolescentis has been found in reduced levels in PLWH compared to individuals without HIV [59]. Conversely, inoculation of B. adolescentis, capable of inducing T-helper (Th)17 cells, in mice showed pathological relevance by worsening autoimmune arthritis [60]. Th17 cells are involved in the pathogenesis of several immune-mediated diseases such as inflammatory bowel disease [61]. Oscillospira is an anaerobic bacterial genus from the Firmicutes phylum and is known to produce butyrate and other short-chain fatty acids [51]. Women living with HIV had a higher abundance of Oscpillospira than women living without HIV and its abundance was also associated with leaner BMI in women living with HIV [62] and lean adults living without HIV [50]. Additionally, this genus has also been related to constipation, which may be due to the bacteria’s slow-growing property [50,51]. A novel finding of this study was the greater relative abundance of the Euryarchaeota phylum present in the cocaine use group. Archaea are found in less abundance within the gut of humans and Euryarchaeota is the most abundant phyla of Archaea [52]. Euryarchaeota include methanogens, which recently have been implicated in infectious diseases and intestinal disorders [52]. Methanogens have been detected in samples from patients with chronic sinusitis and paravertebral abscess [52,63]. Moreover, there is an association with the existence of methanogens and constipation as methane is known to slow intestinal transit contributing to constipation [52,64]. Irritable bowel syndrome with constipation was associated with the presence of methanogens and higher methane production [65]. Cocaine use may produce gastrointestinal disorders including colitis and constipation [55,66,67]. Taken together, the greater abundance of these taxa suggests a potential for pathogenicity and gastrointestinal disorder in PLWH who use cocaine.

Inflammation may impact not only the immunity of individuals but also behavioral outcomes in substance use disorders [20]. Although IL-6 was elevated in cocaine users compared to non-users, we did not observe a significant difference in inflammation by cocaine use. Eubacterium and Clostridium genera, were significantly associated with higher IL- 6 levels. Substance use disorders, and particularly cocaine, have previously been shown to be associated with elevated inflammation, dysregulated immunity [68] and greater immune activation [68]. Chivero et al. [10] demonstrated in a mouse model that cocaine administration was associated with higher levels pro-inflammatory cytokines, chemokines and transcription factors that modulate inflammation in the gut. The use of cocaine in PLWH may further exacerbate HIV disease and lead to additional pathologies. IL-6 among other cytokines is known to have smaller and targeted relationships to the gut microbiota, therefore future studies should include additional measures of inflammation that may be strongly influenced by the gut microbiome such as tumor necrosis factor-alpha (TNF-α) and interferon-gamma (IFN-γ) [69].

Alpha-diversity measures are often examined in gut microbiome-related studies as diversity has been associated with health and lower diversity with disease and mortality [70]. We did not find significant differences in ASV richness or alpha-diversity (Shannon and Simpson indexes), which has been shown to vary by MSM status [71], CD4 cell count [72], and HIV viral load [72]. However, in our study, the groups did not differ by MSM status, CD4 cell count or HIV viral load and majority of the participants were on ART. The impact of ART on the diversity of the gut microbiota in PLWH has not been constant across studies [73]. Although sex has been associated with differences in alpha-diversity in PLWH [70], we did not find significant differences by sex. Volpe et al. [9] also did not find any differences in alpha diversity measures in PLWH by cocaine use. A study conducted by Scorza and colleagues [22] demonstrated a significantly lower Shannon index between rats exposed to volatized cocaine compared with control rats. Chivera et al.[10], also did not show a change in alpha diversity in mice administered cocaine compared with mice provided a saline solution. Controversy in the literature still exists on whether substance use is associated with higher or lower alpha-diversity and whether these measures are associated with additional health-related effects [74,75]. Limited studies in human subjects on the topic of cocaine use and HIV does not allow for additional comparisons of these measures and additional studies with larger sample sizes are needed to fill the gaps.

While majority of the literature on the effects of cocaine on the gut microbiota have focused on animal studies, our study also examined the influence of cocaine use on microbiome-related metabolites in PLWH. We found differences in microbiome-associated compounds between the PLWH who used or did not use cocaine. The amino acid derivative, phenylacetate, was significantly higher in the cocaine using PLWH and was previously found to be associated with the accumulation of lipids in the liver [53] and with small intestine inflammation and abnormal gut permeability [54]. Microbial phenylacetate also has toxic effects on mitochondria by reducing the rate of nicotinamide adenine dinucleotide-dependent respiration and the production of reactive oxygen species [76]. Another metabolite, benzoate, a carboxylic acid [77], was significantly elevated in the cocaine using PLWH and has been linked to greater bacterial virulence and adrenergic stress in animal models [78]. While the evidence on the potential role of these metabolites is sparse, our study indicates the likelihood that cocaine may induce pathogenic microbiome changes. The bacteria and archaea identified to be more abundant in the cocaine use groups did not seem to be associated with the production of the metabolites discussed above. Nonetheless, Lachnospira and Oscillospira genera, which were more abundant in the cocaine use group are linked to butyrate production, and butyrate was found in greater amounts in the cocaine use group.

The present study found that cocaine users compared to non-users had higher plasma butyrate and butyrylglycine levels. Although butyrate is often considered to be associated with favorable health outcomes [79], it has been shown to have pleiotropic effects on health and disease [80]. Kaiko and colleagues [81] demonstrated that butyrate reduced intestinal epithelial stem/progenitor cell growth and reduced wound care in vivo. Other studies have shown that high levels of butyrate in animal models of the intestinal mucosa layer may affect the intestinal barrier and permeability of the colon [82,83]. Interestingly, the taxa associated with Lachnospiraceae, which are butyrate producers, are also known to impact host physiology in positive and negative manners and are associated with pro-inflammatory conditions and anxiety-related behavior in mice [49,84]. Our study showed a higher abundance of Lachnospira genus in cocaine users, contrary to cocaine administered to mice, which showed a depletion in Lachnospiracea bacteria in mice [10]. Oscillospira genus, also a butyrate producer [51], was found in greater abundance in the cocaine use group. The interaction of short-chain fatty acids in the context of HIV infection and its role in the gut microbiome is still mainly unknown.

More abundant taxa in the cocaine non-use group compared to the cocaine use group included Enterobacteriaceae and Muribaculaceae/S24–7 families. The Enterobacteriaceae family from the Proteobacteria phylum is considered to be pathogenic, associated with inflammation, and reported to be elevated in HIV infection [85,86]. Volpe et al. [9] did find a higher relative abundance of Proteobacteria phylum in PLWH compared to participants living without HIV. Muribaculaceae or S24–7 family is part of the Bacteroidetes phylum and has not been cultured until recently [87]. We did not find additional studies conducted in the context of HIV and/or cocaine use and Muribaculaceae and although it has been found in the gut of humans most of the studies are in animal models [87]. Conversely, humanized mice treated with morphine and infected with HIV did show a reduction in Muribaculacaea [88]. In mice, it has been associated with short-chain fatty acid production [89].

The diet quality or HEI score was significantly lower for the cocaine use group compared with those who did not use cocaine in this study, however both groups were below the average score of 59 for Americans indicating that the overall diet for both groups do not adhere to the Dietary Guidelines for Americans [44]. These results are similar to Volpe et al. [9] in that cocaine users had lower HEI scores than non-users, yet our cohort HEI scores were much lower than reported by Volpe and colleagues. Cocaine users also had lower BMI consistent with previous reports [90]; yet, there were no significant differences in total calories consumed between the groups. Some research suggests that cocaine users may lose weight through metabolic effects as opposed to changes in appetite [90,91]. Of note, the cocaine use group had significantly lower annual income than the cocaine non-use group, which may also influence the diet quality [92,93]. Lower quality diets tend to be higher in sugars and fats and more energy-dense diets tend to lack fruits and vegetables and can be less expensive [94]. The impact of diet in the context of HIV and substance use should be further explored.

We present the first multi-omic study comparing the fecal bacterial microbiome of PLWH who use and do not use cocaine. Limitations of our study include a small sample size and its cross-sectional design. Additionally, we were not able to observe the viral and fungal components of the fecal microbiome. Due to the small sample and focus on the primary objective of this study further exploration of the other factors that may affect the gut microbiome such as ART, ethnicity, and diet were not able to be executed. Strengths of the study included objective measures of cocaine use and measurement of other metabolites associated with cocaine use in humans. This pilot study also examined and accounted for sex, a possible confounder within microbiome-related studies. Although sex was not significantly different by cocaine use due to the impact of sex on metabolism in the context of the intestinal microbiome, we decided to analyze the metabolite data by sex and cocaine use group [95,96]. Together, these findings demonstrate that cocaine use modulates the host gut microbiome in an adverse manner that may exacerbate gastrointestinal conditions. Improved understanding of the shift in bacterial composition and distribution in the presence of cocaine can be used to improve the management of substance-using populations. This study lays the foundation for future studies that may address important public health challenges such as substance use to improve health and decrease disability. Future longitudinal studies that incorporate omic approaches to fully capture the potential of the gut microbiome that include all types of microorganisms are needed to devise therapeutic strategies to modulate the gut for disease and substance use disorder management.

Supplementary Material

SDC - Clean

Supplemental Digital Content 1. Figure that illustrates the difference in ASV Richness and Shannon and Simpson indices by groups.docx

Supplemental Digital Content 2. Figure that illustrates the PCoA plot by groups.docx

Supplemental Digital Content 3. Table that demonstrates the PERMANOVA statistics for metagenomics and metabolomics.docx

Supplemental Digital Content 4: Figure that illustrates the Boxplot of Bacteroidetes/Firmicutes (B/F) ratio by groups.docx

Supplemental Digital Content 5: Figures that illustrate the Boxplots of centered log-ratio (CLR) values for butyrate-producing bacteria genera by groups.docx

Supplemental Digital Content 6: Figure that illustrates the Boxplot of Interleukin-6 by groups.docx

Supplemental Digital Content 7: Figures that illustrate the differences in metabolites by groups and sex.docx