Identification of volatile metabolites produced from gut microbial levodopa metabolism using an untargeted metabolomics approach

doi:10.21203/rs.3.rs-3335592/v1

Interspecies pathways in the gut microbiome have been shown to metabolize levodopa, the primary treatment for Parkinson’s disease, and reduce its bioavailability. While the enzymatic reactions have been identified, the ability to establish the resulting macromolecules as biomarkers of microbial metabolism remains technically challenging. In this study, we leveraged an untargeted mass spectrometry-based approach to investigate volatile organic compounds (VOCs) produced during levodopa metabolism by Enterococcus faecalis, Clostridium sporogenes, and Eggerthella lenta. We cultured these organisms with and without their respective bioactive metabolites and detected levodopa-induced shifts in VOC profiles. We then utilized bioinformatics to identify significant differences in 2,6-dimethylpyrazine, 4,6-dimethylpyrimidine, and 4,5-dimethylpyrimidine associated with its biotransformation. Supplementing cultures with inhibitors of levodopa-metabolizing enzymes revealed specific modulation of levodopa-associated diazines, verifying their relationship to its metabolism. Furthermore, functional group analysis depicts strain-specific VOC profiles that reflect interspecies differences in metabolic activity that can be leveraged to assess microbiome functionality in individual patients. Collectively, this work identifies previously uncharacterized metabolites of microbiome-mediated levodopa metabolism with applications for point-of-care testing and offers an alternative approach to evaluate interindividual differences in treatment response.

Parkinson’s disease (PD) is a neurodegenerative disorder that predominantly affects dopamine-producing neurons in the brain, resulting in hallmark symptoms including tremors, rigidity, slowed movements, and impaired balance¹. The primary treatment for PD is levodopa, an orally ingested central nervous system agent that is metabolized by aromatic amino acid decarboxylase (AADC) enzymes in the brain to produce dopamine². First-pass metabolism outside the nervous system heavily impairs therapeutic efficacy, since dopamine generated in the periphery cannot cross the blood-brain barrier³. However, despite its co-prescription with peripheral AADC inhibitors, patients display highly variable treatment responses and reduced efficacy over time. Recent studies have identified bacteria strains in the gut microbiome as major factors that reduce levodopa bioavailability^4,5. Previous work from Rekdal et. al established an interspecies pathway that begins with the bioconversion of levodopa to dopamine by Enterococcus faecalis, followed by dopamine dehydroxylation to m-tyramine by Eggerthella lenta⁴. In an alternative pathway, the prevalent gut microbe Clostridium sporogenes converts levodopa to 3-(3,4-dihydroxyphenyl) propionic acid (DHPPA), which is further degraded into 3-(3-hydroxyphenyl) propionic acid (3-HPPA) by Eggerthella lenta⁵. While previous efforts have sought to directly detect levodopa metabolites from biological samples^6,7,8, there remains an inability to establish these macromolecules as reliable indicators of extracerebral metabolism due to discrepancies in sampling and analytical techniques. The lack of standardized sampling procedures, coupled with differences in the specificity and sensitivity of analytical instruments, introduces consequential variability in metabolite measurements that create difficulty deriving biological interpretations from results^9,10,11,12. Such challenges underscore the need for a more robust method to evaluate levodopa metabolism in Parkinson’s patients and assess interindividual variations in treatment response.

Metabolomics has been identified as a powerful tool to enhance our understanding of the gut microbial metabolic repertoire and the discovery of microbiome-drug interactions. Microbes have a distinct metabolism and produce an array of metabolites, including bacteria-specific volatile organic compounds (VOCs) that can cross the blood-brain barrier and enter host circulation^13,14,15. VOCs are small molecules that represent the volatile portion of the metabolome and have gained attraction as a non-invasive method for diagnosis and assessing treatment efficacy^14,15,16,17. In contrast to macromolecule-based biomarkers, VOCs can be sampled non-invasively and with minimal preparation, providing a direct detection method with potential for translational research. Recent studies have demonstrated that VOCs characterized from biological mediums can reflect microbiome composition and distinguish physiological states, essential in understanding the mechanisms underlying host-microbiome interactions^14,15,18,19. Facilitating the analysis of VOCs, gas chromatography-mass spectrometry (GC-MS) remains the gold standard for the untargeted analysis of VOCs extracted from biological systems^13,15. As a result, there are a growing number of studies using GC-MS to analyze the composition and chemical reaction space of the human microbiome. Expanding the scope of gut microbial-related metabolites enables the identification of diverse microbiome-drug interactions that can differentiate metabolic phenotypes. Consequently, volatile signatures of levodopa metabolism represent a potential strategy to directly evaluate heterogeneous responses and account for the microbiome’s effect on pharmacokinetics.

To identify the full spectrum of metabolites associated with levodopa metabolism, this study builds off foundational works to characterize VOCs from the strains involved in its biotransformation. Using an untargeted approach, we characterized the volatile metabolome from Enterococcus faecalis strain OG1RF, Clostridium sporogones (WT), and Eggerthella lenta strain MSMC 77 − 67 (DSM 15644) in response to levodopa metabolism. Statistical analysis comparing cultures with and without levodopa enabled the identification of VOCs uniquely produced during the breakdown of the drug. Inhibiting drug-microbe interactions with compounds that inactivated the corresponding enzymes resulted in the modulation of levodopa-related VOCs, further demonstrating their association with its metabolism. These results identify previously unknown metabolites of gut microbial levodopa metabolism that correlate with metabolic pathways involved in cell growth, offering additional insights into the cellular mechanisms impacting pharmacology. Furthermore, strain-specific VOCs corresponding to each pathway offer a robust strategy to inform personalized treatment strategies based on interpersonal microbiome compositions.

Enterococcus faecalis produces diazine by-products during levodopa decarboxylation

To collect VOCs from bacteria cultures, strains were inoculated into a custom culture vessel previously developed in our lab to enable the continuous flow of gas through the headspace²⁰,effectively increasing overall sensitivity and reproducibility of collected VOCs (Fig. S1). In short, gas is designed to flow continuously from the vessel inlet to the outlet and continuously deplete analytes from the headspace, increasing the total available signal as compared to static headspace sampling methods. To investigate differences in VOC production during the bioconversion of levodopa to dopamine, we collected VOCs from the strain E. faecalis OG1RF cultured with and without 1mM levodopa. VOCs were recovered using solid-phase microextraction (SPME) and characterized by GC-MS analysis (Table 1). Comparison of volatile metabolomes from each experimental replicate shows differential VOC abundance in E. faecalis cultures containing levodopa (Fig. 1a). Compounds endogenously produced from E. faecalis (VOCs not detected in the growth media) and levodopa decarboxylation are depicted in the chromatograms generated from each experimental group (Fig. 1c, Fig. 1d, Fig. 1g). Statistical analysis performed in R identified significant changes in 2,6-dimethylpyrazine and 4,6-dimethylpyrimidine (adjusted p < 0.05) during levodopa decarboxylation, and a significant difference in sec-butylamine from E. faecalis replicates and the media control. Applying a multiple hypothesis test correction²¹ with a false discovery rate (FDR) set at 0.10 maintained significant differences (Fig. 1e, Fig. 1f, Table S1).

*Table 1 Filtered VOC list from E. faecalis* OG1RF**
Number	VOC name	Chemical Class	Retention Time (min)	RI	ID Level
1	sec-Butylamine	Amine	1.95	469	3
2	Analyte 1	Unknown	2.23	478	4
3	Methylpyrazine	Heteroaromatic	3.21	828	2
4	o-Xylene	Aromatic	4.18	901	2
5	2,5-Dimethylpyrazine	Heteroaromatic	4.42	914	1
6	2,6-Dimethylpyrazine	Heteroaromatic	4.5	918	1
7	4,6-Dimethylpyrimidine	Heteroaromatic	4.89	938	2
8	Benzaldehyde	Aldehyde	5.4	964	2
9	Benzeneacetaldehyde	Aldehyde	7.22	1048	2
10	Acetophenone	Ketone	7.77	1071	2
11	4-Hydroxy-3-methylbenzaldehyde	Other	8.03	1082	3
12	3-Methylbenzaldehyde	Aldehyde	8.14	1087	2
13	Phthalan	Other	8.58	1108	3
14	Dodecane	Hydrocarbon	9.83	1200	1
15	Tetradecane	Hydrocarbon	11.36	1399	1
16	Phenol, 2,4-bis(1,1-dimethylethyl)	Other	11.99	1522	2

The foundational work from Rekdal et. al shows that the L-tyrosine analog, (S)-α-fluoromethyltyrosine (AFMT), effectively prevents levodopa decarboxylation in E. faecalis cultures by selectively inhibiting the tyrDC enzyme⁴. Therefore, we hypothesized that supplementing cultures of E. faecalis with levodopa in combination with AFMT would inhibit the production of 2,6-dimethylpyrazine and 4,6-dimethylpyrimidine. We found that adding AFMT (250 µM) generated similar VOC profiles to E. faecalis cultures without levodopa. Principal component analysis (PCA) demonstrated the ability to separate the VOCs detected across all four experimental conditions into distinct clusters when considering the first and second principles, although it is evident that the native bacterial volatile profile overlaps with broth volatile profile – as is commonly reported^20,22,23 (Fig. 1b). The abundance of 2,6-dimethylpyrazine and 4,6-dimethylpyrimidine were completely and significantly inhibited following coadministration of AFMT and levodopa. Together, these findings demonstrate that AFMT significantly reduces diazine volatiles generated in the presence of levodopa indicating the presence of 2,6-dimethylpyrazine and 4,6-dimethylpyrimidine as a potential indicators of levodopa metabolism.

Increased production of fatty acid esters observed during levodopa deamination by Clostridium sporogenes

To investigate each pathway for gut bacterial levodopa metabolism, we next worked to determine whether C. sporogenes produced VOCs associated with levodopa deamination. Using the same methods described previously, we detected VOCs from cultures of C. sporgenes with and without levodopa to capture differences in VOC production during the biotransformation to DHPPA (Table 2). GC-MS analysis revealed changes in VOC profiles acquired from C. sporogenes (Fig. 2a). Furthermore, PCA shows three distinct clusters of VOC profiles corresponding to each experimental condition and reflects levodopa-induced shifts in VOC abundance (Fig. 2b). The results from statistical analysis identify C. sporogenes endogenous VOCs and revealed significant differences between C. sporogenes replicates and media controls, with adjusted p-values > 0.05 (Table S2).

Table 2 Filtered VOC list from C. sporogenes
Number	VOC name	Chemical Class	Retention Time (min)	RI	ID Level
1	Analyte 1	Unknown	1.50	453	4
2	Analyte 2	Unknown	1.53	454	4
3	Analyte 3	Unknown	1.67	460	4
4	Analyte 4	Unknown	1.71	465	4
5	Analyte 5	Unknown	1.85	468	4
6	sec-Butylamine	Amine	1.95	469	3
7	Analyte 6	Unknown	2.34	482	4
8	Analyte 7	Unknown	2.63	493	4
9	4-Methyl-1-pentanol	Alcohol	3.26	832	3
10	2,5-Dimethylpyrazine	Heteroaromatic	4.42	914	1
11	Benzaldehyde	Aldehyde	5.4	964	2
12	Butanoic acid, butyl ester	Ester	6.03	996	2
13	Trimethylpyrazine	Heteroaromatic	6.2	1004	3
14	Butyl 2-methylbutanoate	Ester	7.09	1042	2
15	Butanoic acid, 3-methyl-, butyl ester	Ester	7.19	1046	2
16	Pentanoic acid, butyl ester	Ester	7.43	1056	3
17	Acetophenone	Other	7.77	1071	2
18	Butanoic acid, pentyl ester	Ester	8.32	1094	3
19	Butanoic acid, 2-methyl-, 3-methylbutyl ester	Ester	8.46	1100	2
20	Butanoic acid, 3-methyl-, 3-methylbutyl ester	Ester	8.55	1107	2
21	n-Amyl isovalerate	Ester	8.61	1111	2
22	Phenylethyl alcohol	Alcohol	8.74	1120	2
23	Benzyl nitrile	Nitrile	9.16	1151	3
24	Hexanoic acid, 3-methylpropyl ester	Ester	9.33	1163	3
25	Dodecane	Hydrocarbon	9.84	1200	1
26	Hexyl n-valerate	Ester	9.89	1206	3
27	Isopentyl hexanoate	Ester	10	1219	2
28	Benzenepropanol	Alcohol	10.18	1239	2
29	Hexanoic acid, pentyl ester	Ester	10.33	1257	3
30	Analyte 8	Unknown	10.82	1317	4
31	Analyte 9	Unknown	11.07	1355	4
32	Tetradecane	Hydrocarbon	11.36	1399	1
33	Phenethyl butyrate	Ester	11.6	1445	3
34	Benzeneacetic acid, butyl ester	Ester	11.63	1450	2
35	Butanoic acid, 3-methyl-, 2-phenylethyl ester	Ester	11.91	1504	3
36	Phenol, 2-4-bis(1,1-dimethylethyl)	Other	11.99	1522	2
37	Analyte 10	Unknown	12.10	1545	4
38	2-Methylpropyl benzenepropanoate	Ester	12.17	1560	3
39	Butanoic acid, 3-phenyl propyl ester	Ester	12.22	1573	2
40	Hexadecane	Hydrocarbon	12.35	1599	1
41	Pivalic acid, 2-phenylethyl ester	Ester	12.44	1621	3

While significant differences in Analyte 2 and Analyte 5 were detected during levodopa deamination, no significant differences were seen between levodopa replicates and the media controls, indicating that additional work is needed to determine possible metabolic origins. While no VOCs were exclusively produced in the presence of levodopa, we detected increased abundances of VOCs corresponding to levodopa deamination and identified endogenous compounds of C. sporogenes (Fig. 2c-f).

Volatile pyrimidine indicators of Eggethella lenta dehydroxylase activity generated during the degradation of levodopa metabolites

Dopamine (dadh) and hydrocaffeic acid (hcdh) are the most prevalent catechol dehydroxylase enzymes that specifically metabolize dopamine and DHPPA, respectively²⁴. However, since a single nucleotide polymorphism (SNP) in the dadh gene distinguishing metabolizing and non-metabolizing E. lenta strains, we chose to investigate VOCs produced during DHPPA dehydroxylation to 3-HPPA by hcdh. The resulting VOCs exhibit differential VOC abundance in E. lenta DSM 15644 cultured with DHPPA (Fig. 3a). Interestingly, supplementing E. lenta cultures with 500 µM DHPPA resulted in a significant increase of 4,5-dimethylpyrimidine and an unknown compound labeled Analyte 1 (Fig. 3c-d, Table 3).

The unidentified compound has a similar retention index and the same parent peak as the other diazine compounds identified in this study (m/z = 102), likely indicating that the unidentified compound is a dimethylpyrimidine isomer.

*Table 3 Filtered VOC list from E. lenta* ATCC 43055**
Number	VOC name	Chemical Class	Retention Time (min)	RI	ID Level
1	Methylpyrazine	Heteroaromatic	3.21	828	2
2	2,5-Dimethylpyrazine	Heteroaromatic	4.41	913	1
3	4,5-Dimethylpyrimidine	Heteroaromatic	4.89	938	3
4	Analyte 1	Unknown	5.09	948	3
5	Benzaldehyde	Aldehyde	5.4	964	2
6	Trimethylpyrazine	Heteroaromatic	6.2	1004	3
7	Benzene, 1-methyl-2-(1-methylethyl)	Aromatic	6.75	1028	2
8	Benzeneacetaldehyde	Aldehyde	7.22	1047	2
9	Acetophenone	Other	7.77	1071	2
10	3-Ethyl-2,5-dimethylpyrazine	Heteroaromatic	8.02	1082	2
11	3-Methylbenzaldehyde	Aldehyde	8.14	1087	2
12	Nonanal	Aldehyde	8.54	1106	2
13	Benzyl nitrile	Nitrile	9.15	1150	2
14	Dodecane	Hydrocarbon	9.83	1200	1
15	2-Methoxy-4-methylphenol	Other	9.9	1208	2
16	Nonanoic acid	Acid	10.48	1274	2
17	Indole	Heteroaromatic	10.75	1306	2
18	Indolizine	Heteroaromatic	10.78	1311	3
19	Benzene, 3-cyclohexen-1-yl	Aromatic	11.19	1373	3
20	Tetradecane	Hydrocarbon	11.36	1399	1
21	Phenol, 2-4-bis(1,1-dimethylethyl)	Other	11.99	1522	2

To confirm that the dimethylpyrimidine compounds were a result of DHPPA dehydroxylation by hcdh enzymes, we supplemented cultures of E. lenta with sodium tungstate in combination with DHPPA. The inclusion of tungstate inactivates the dehydroxylase enzymes without hindering bacteria growth by substituting the necessary molybdenum cofactor⁴.

As a result, supplementing 250 µM tungstate to E. lenta cultures in combination with 500 µM DHPPA significantly decreased the concentration of 4,5-dimethylpyrimdine and its putative isomer (Table S3). The inhibition of 4,5-dimethylpyrimidine and its putative isomer indicates the relationship of these VOCs with DHPPA dehydroxylation. PCA demonstrated the ability to separate the VOCs detected across all four experimental conditions into distinct clusters when considering the first and second principal components (Fig. 3b). Inhibiting the dehydroxylase enzymes with tungstate yielded a more similar VOC profile to the media control compared to E. lenta cultured with no additives, possibly indicating off-target tungstate inhibition of the other molybdenum-dependent enzymes. Further, benzene, 1-methyl-2-(1-methylethyl) was generated in all conditions except for the media control, confirming its endogenous origins to E. lenta metabolism. Collectively, these findings suggest that increased dehydroxylase activity during the conversion of DHPPA to 3-HPPA shifts the metabolic activities of E. lenta which can be observed in the volatile metabolome of this organism.

Functional group analysis identifies strain-specific VOC profiles that reflect differences in metabolism

To capture unique metabolic activities harnessed by microbiota involved in levodopa metabolism, we analyzed the chemical diversity of each strain’s VOC profile. Comparing the VOC profile of each strain reveals differential VOC abundance and frequencies (Fig. 4a). VOCs detected across all three strains include 2,5-dimethylpyrazine, benzaldehyde, acetophenone, dodecane, tetradecane, and phenol, 2,4-bis(1,1-dimethylethyl). The results from PCA reveal highly distinct VOC profiles captured from E. faecalis, C. sporogenes, and E. lenta (Fig. 4b). To better assess organism-specific volatile signatures, compounds were classified by functional group to further characterize the underlying biochemical behaviors. Functional group breakdown revealed unique proportions and categories of metabolites detected from each strain (Fig. 4c). The volatile metabolome of C. sporogenes was dominated by esters (43.90%), followed by unknown compounds (24.39%). Heteroaromatic compounds were the most prevalent VOCs detected from E. faecalis (25%), followed by aldehydes and VOCs classified as “other” (18.75% each). The most frequent chemical class detected from E. lenta was heteroaromatic VOCs (33.33%), followed by aldehydes (19.05%).

In this study, we use an untargeted mass spectrometry-based approach to identify VOCs produced during gut microbial levodopa metabolism. The discovery of gut microbial-derived VOCs has become a growing area of research that can provide major mechanistic insights into the biological processes underlying host-microbiome interactions with applications for personalized medicine. The main advantage of identifying characteristic VOCs over traditional biomarkers of drug metabolism is the reduction of analytical complexity. While the presence of levodopa-metabolizing enzymes can gauge potential gut microbial activities, the high implementation costs to extract and analyze these markers limits their routine utilization as indicators of off-target metabolism. Hence, VOC signatures of levodopa metabolism offer a robust, microbiome-centered approach to assess individual changes in its pharmacokinetics.

Studies have demonstrated a positive correlation between tyrDC gene expression, substrate concentration, and growth of E. faecalis, indicating that higher tyrDC activity enhances bacterial growth^25–28. Importantly, the VOCs associated with levodopa decarboxylation correspond to metabolic activities tied to cell growth. These effects are further supported by functional analyses at the transcriptional level, which reveal that an increased tyrDC substrate concentration leads to enhanced pathways for purine and pyrimidine catabolism for DNA biosynthesis, along with an activation of amino-sugars metabolism in E. faecalis²⁸. Furthermore, the accumulation of reactive α-hydroxycarbonyl and α-aminocarbonyl intermediates during the growth phase of various bacteria has been correlated with an increased production of pyrazine VOCs, which are widely associated with lactic acid bacteria^{29,30,31,32,33}. These findings suggest that biological pathways corresponding to energy metabolism and growth may potentially give rise to the increased production of 2,6-dimethylpyrazine and 4,6-dimethylpyrimidine. The presence of these levodopa-induced diazines could potentially indicate a shift in E. faecalis metabolic activity in response to increased tyrDC substrate abundance. Notably, the results show that inhibiting levodopa decarboxylation with AFMT prevents the production of 2,6-dimethylpyrazine and 4,6-dimethylpyrimidine. The simultaneous inhibition of these compounds with AFMT suggests that these VOCs indicate a shift in E. faecalis metabolic activity in response to levodopa treatment and increased activity towards tyrDC substrates.

We also detect a shift in the volatile metabolome of C. sporogenes during the bioconversion of levodopa to DHPPA that corresponds to an increased abundance of VOCs classified as fatty acid esters. These energy-rich compounds are associated with pathways of cell signaling, membrane formation, and fatty acid biosynthesis. Moreover, fatty acid esters are associated with the distinct metabolite profile detected from C. sporogenes during anaerobiosis, but not aerobiosis³⁴. The production of short and medium chain fatty acids is well implicated across Clostridia species^35,36,37, with various metabolic pathways producing a variety of fatty acid products via chain elongation mechanisms using coenzyme A^35,38. Similarly, the fldABC protein complex responsible for DHPPA production uses the same CoA-dependent transfer mechanism to deaminate levodopa⁵. These processes generate an accumulation of ester precursors that enables their production. Although we did not observe statistically significant differences in VOC abundances between C. sporogenes cultured with and without levodopa, we report previously uncharacterized compounds that reflect the strain’s unique metabolic activity. While these VOCs have been implicated in cellular processes underlying energy production, further investigations are needed to link the underlying biological pathways leading to their production.

We next sought to characterize VOCs associated with the dehydroxylation of the levodopa intermediate DHPPA. E. lenta performs catechol dehydroxylation using distinct molybdenum-dependent enzymes that are variably distributed across individual strains^4,24. The prevalence of genes encoding these specialized enzymes correlates with the metabolism of their specific substrates, with hcdh activity being the most prevalent across closely related gut microbiota^24,40. As such, there is a growing interest to elucidate the chemical mechanisms exerted by these enzymes to mediate primary and secondary metabolism in the gut. To better understand the effects of catechol dehydroxylation, we identified significant changes in VOCs detected during DHPPA metabolism by hcdh. During the dehydroxylation of DHPPA to 3-HPPA, we detect a significant increase in 4,5-dimethylpyrazine and a putative dimethylpyrimidine isomer from E. lenta. With the mechanisms of catechol dehydroxylase enzymes remaining poorly understood, it is difficult to decipher their biological roles and pathway involvements. However, additional work from Rekdal et. al demonstrated that the dehydroxylation of dopamine, the other levodopa intermediate, provided a growth advantage to E. lenta and potentially serves as an alternative electron acceptor^4,24. Furthermore, metabolite profiles recently generated by Noecker et. al revealed that nucleotide and cell wall metabolites comprised a large proportion of the E. lenta metabolome, along with nucleic acid intermediates³⁹. Based on these findings, we suspect the production of 4,5-dimethylpyrimdine and the other DHPPA-associated VOC could be a result of the potential growth-promoting effects of DHPPA metabolism. These studies also show that the addition of tungstate to the growth medium blocks the molybdenum-dependent dehydroxylase activity and inhibits the growth increase of E. lenta⁴. Interestingly, supplementing tungstate to E. lenta cultures suppressed the production of 4,5-dimethylpyrimidine and the putative dimethylpyrimidine isomer, providing evidence for endogenous production and further suggesting these VOCs may arise from increased growth tied to catechol dehydroxylation. While further investigations are necessary to characterize the dimethylpyrimidine isomer and confirm the biological origins of these VOCs, previous works have demonstrated the role of pyrimidines and their derivatives for bacterial growth and sensing mechanisms, implicating metabolic pathways for cell growth and signal processing as potential sources for VOCs^41,42,43. Taken together, these findings open the door to improve our understanding of the relationship of dimethylpyrimidine VOCs and bacteria growth to improve our understanding of the effects of catechol dehydroxylation on gut microbiota metabolism.

VOCs can be generated across a wide range of biosynthetic pathways and are thought to diffuse through cell membranes- giving them the potential to modulate gene expression and influence physiology²⁹. Thus, an emerging area of research is focused on understanding the molecular mechanisms underlying VOC production. While little information is known regarding the genes and pathways that interact with VOCs, the main biological responses reported for microbial VOCs include biofilm formation, virulence, secondary metabolite production, and growth²⁹. In line with these findings, many of the VOCs identified in this study are implicated in those cellular processes. In this study, we detected unique volatile signatures from E. faecalis, C.sporogenes, and E. lenta. Functional group analyses revealed diverse chemical compositions within the VOC profile of each strain, whose relative abundances reflect different biological processes. The VOC profile of E. faecalis was mainly composed of heteroaromatic compounds, followed closely by aldehydes and VOCs classified as “other.” Additionally, we identified analyte 1 from E. faecalis as an endogenous VOC. We suspect that these compounds originate from primary metabolic pathways encoded in lactic acid bacteria and are likely products of carbohydrate fermentation and amino acid degradation. The VOC profile of E. lenta showed a similar chemical composition after performing functional group analyses. Little is known about the metabolic activity of E. lenta and the biosynthetic pathways leading to VOC production. Interestingly, the VOCs we detected from E. lenta are implicated in processes for amino acid and nucleic acid metabolism. These findings are in agreement with previous studies using systems biology-based approaches, which show that a large portion of E. lenta’s metabolic activity corresponds to cell growth and energy metabolism⁴⁰. Collectively, these findings support that the distinct volatile profiles capture differences in metabolic activity across bacteria strains. Furthermore, the chemical signals encoded in these signatures can elicit a range of processes to coordinate behavior and respond to environmental stimuli. Thus, it is likely that the organism-specific VOC profiles detected in this study have distinct biological effects.

Here, we report differences in the VOC profiles of E. faecalis, C. sporogenes, and E. lenta and identify VOCs associated with levodopa metabolism. Using an untargeted, metabolomics approach, we report previously uncharacterized VOCs from each strain that furthers our understanding of metabolic processes harnessed by gut microbiota and offers a foundation for the development of personalized diagnostics to assess variable responses to levodopa treatment across PD patients. Collectively, these findings give insight into the metabolic activities of different gut microbiota, identify reproducible changes that occur during the breakdown of levodopa, and link 4,6-dimethylpyrimidine and 2,6-dimethylpyrazine to the endogenous bioconversion of levodopa by E. faecalis. Our consideration of species-specific VOCs shows diverse chemical classes within each organism’s volatilome that reflect interspecies differences in metabolic activity. In the future, functional interpretations of these VOCs will enable us to decode the molecular mechanisms by which gut bacteria interfere with PD treatment, offering novel strategies to assess treatment efficacy and improve patient outcomes.

Bacterial growth and incubation

All bacteria culturing work was performed in a glove bag (Sigma-Aldrich, SKU #Z530212) under an atmosphere of 5% H₂, 20% CO₂, and 80% N₂. Plates were sealed in an anaerobic jar (Thomas No. 1177X01) containing an Anaerobe GasPak satchet (BD, SKU #260678) and incubated at 37 ℃. Hungate tubes (Chemglass, CLS-4208) were used for all liquid culturing and incubated at 37 ℃ in a tube shaker at 500 rpm.

Enterococcus faecalis OG1RF (ATCC 47077) and Clostridium sporogenes (ATCC 15579) were plated onto Brain Heart Infusion (BHI) agar (Sigma-Aldrich, SKU #70138-500G) and incubated for 24 and 48 hours, respectively. Isolated colonies were then inoculated into BHI broth (Sigma-Aldrich, SKU #53286-100G) and grown 24 hours for E. faecalis, and 48 hours for C. sporogenes. Eggerthella lenta DSM 15644 (ATCC 43055) was plated onto BHI agar containing 1% arginine (Sigma-Aldrich, #A5006-100G) and incubated for 48–72 hours. Isolated colonies were then inoculated into BHI broth containing 1% arginine and 10 mM sodium formate (Sigma-Aldrich, #71539-500G) and grown for 48 hours.

VOC Collection

OD₆₀₀ measurements were taken and used to dilute turbid starter cultures such that the initial OD₆₀₀ seeding concentration was approximately 0.1 at the start of VOC collection and subsequently used for statistical normalization. For E. faecalis and C. sporogenes, bacteria were seeded into the custom glass culture vessel containing a total 5 mL broth with and without 1 mM levodopa (Sigma-Aldrich, SKU #D9628-5G). E. lenta was seeded into the culture vessel containing 5 mL of broth with and without 500 µM DHPPA (Sigma-Aldrich, SKU #102601-2.5G).

Bacteria cultures were then connected to a custom flow system designed to collect VOCs from each bacteria strain using dynamic headspace sampling methodology (Fig. S1). Anaerobic gas (5% H₂, 20% CO₂, 80% N₂) was connected to the custom glass culture vessel inlet using Nalgene™ 890 FEP tubing in-line with a hydrocarbon trap, sterile filter, and flow meter. The gas flow was turned on 24 hours prior to all VOC collection in order to obtain equilibrium and maintain a flow rate of 18 mL/min for the duration of sampling. The flow rate was lowered to 8.6 mL/min for E. lenta the longer 48 hour sampling duration. A DVB/Carboxen/PDMS SPME fiber (Sigma-Aldrich, SKU #57298-U) was used to recover VOCs from our in-house¹⁷. The SPME fiber was connected to the system using a previously described adapter component¹⁷ attached directly to the outlet of the custom glass culture vessel. Each SPME fiber was replaced after approximately every 80 injections and conditioned at 270 ℃ prior to use. Fibers were heated at 270 ℃ for 5 minutes between sample injections to minimize carryover.

The sampling duration for each strain was determined based on preliminary data determining the metabolic rates of levodopa or DHPPA by the respective strains^4,5. To extract VOCs, the SPME fiber was exposed in the headspace of E. faecalis cultures for 10 hours, C. sporogenes cultures for 24 hours, and E. lenta cultures for 48 hours. All experiments were done in triplicate for each experimental group. Control samples were collected in triplicate for each strain to account for VOCs produced from the medium.

Gas Chromatography-Mass Spectrometry Analysis

VOCs were analyzed using an Agilent 6890 GC-MS and an HP-5MS column (30.0 m x 250 µm x 0.25 µm; Agilent #19091S-433). SPME fibers were injected into the inlet using a Gerstel MPS autosampler maintained at 250 ℃, using splitless injection. The column was heated to an initial temperature of 60 ℃, ramped at 5 ℃/min to 100 ℃, then ramped at 30 ℃/min to 270 ℃ and held for 2 minutes. The helium carrier gas flow rate was 1.0 mL/min (UHP Helium 99.999%). The transfer line was maintained at 280 ℃. For mass spectrometry, the source temperature was 230 ℃. Mass spectra were acquired over a mass range of 35–300 amu with an ionization energy of -70 mV. A PFTBA standard was run prior to each SPME injection to tune the MS in an effort to minimize instrument variability. A 1 µL volume of methanol was injected prior to each sample to monitor for contaminants and further minimize carryover.

Data Processing

MSD ChemStation software was used to analyze the chromatograms from each run. Quantitative values for signal abundance were obtained by integrating the area under each peak in the chromatogram. Full-width half maximum was used to define the peak integration parameters. Each peak was compared to the Wiley and National Institute of Standards and Technology (NIST) 2005 Mass Spectral Libraries, and tentative peak names were assigned using reference mass spectra for each compound if the spectral similarity was ≥ 70 (70%). All integrated peaks and library hits were exported into Excel and the dataset for each replicate was saved as a .csv file.

Each dataset was imported into R software, version 4.1.2 (The R Foundation for Statistical Computing, Vienna, Austria). Custom R code was designed to streamline post-processing, statistical analysis, and data visualization. Peaks were integrated using the total ion chromatogram and filtered using a minimum threshold value of 1,000,000 for the integrated abundance to focus on the reproducible features in each bacterial volatilome. To align peaks

across chromatograms, the retention time shift had to be ≤ 0.6 seconds between samples. Compound abundance was normalized using the OD₆₀₀ measurement recorded at the start of each replicate. The resulting dataset was then manually filtered to remove known contaminants (siloxanes) and poorly resolved peaks.

VOC identities were assigned according to the metabolomic reporting standards set previously⁴⁴. Compounds received an ID confidence level between 1 and 4, with 1 representing the highest confidence supported by two independent and orthogonal data sources⁴⁴. In brief, compounds with mass spectral match ≥ 80%, to the Wiley and NIST 2005 Mass Spectral Libraries, received an initial confidence level of 3. Compounds below this mass spectral threshold were assigned an ID level of 4 and labeled as numbered analytes (i.e. “Analyte 1”). A C8-C20 alkane standard was analyzed and used to assign retention indices (RI) for all VOCs. Compounds with an ID level of 2 were verified with a ≥ 80% mass spectral match and a retention index that is consistent with the non-polar stationary phase using the mean of published RIs. Analytical standards (≥ 98%) were characterized under identical conditions (Fig. S2) and were used to assign an ID level of 1 (TCI #D1527; TCI #D2171).

Statistical Analysis

Compounds variably present (> 30% missing observations) in each dataset were removed, unless otherwise indicated, to focus on the reproducible aspects of the bacterial volatile metabolome. Integrated peaks were log₁₀ transformed in R, Version 4.0.3 (The R Foundation for Statistical Computing, Vienna, Austria) and missing values were imputed to 0 abundance. A two-sided, unpaired students’ t-test assuming unequal variance was applied to compare VOC abundances between each experimental condition in a pairwise manner. The resulting p-values were then adjusted according to the Benjamini-Hochberg correction procedure²¹ using the “p.adjust” function in the base R “stats” library.

Enzyme Inhibition

To determine whether VOCs significant to levodopa metabolism were produced as a direct result of its degradation, bacteria cultures were treated with known inhibitors of levodopa metabolism⁴. E. faecalis cultures were supplemented with 1 mM levodopa and 250 µM (S)-𝛼-fluormethyltyrosine (AFMT) (Sigma Aldrich, SKU # SML3100-5MG) using the same procedures described previously to prevent levodopa decarboxylation by the tyrDC enzyme. To inhibit DHPPA dehydroxylation by E. lenta, 250 µM sodium tungstate (Sigma Aldrich, SKU # 72069-25G) was added to E. lenta cultures containing 500 µM DHPPA. Three experimental replicates were collected from each strain using the same procedures described previously to determine the resulting effects on VOCs identified as significant. All inhibition studies were performed in experimental triplicate using the methodology described previously.

Ethics Approval and Consent to Participate

Not applicable.

Consent for Publication

Not applicable.

Availability of Data and Materials

Data Availability

All raw chromatographic and mass spectral data have been made publicly available in the MetaboLights⁴⁵ repository at: https://www.ebi.ac.uk/metabolights/MTBLS8293

Code Availability

Custom R code designed to streamline post-processing, normalization, statistical analysis, and data visualization has been made publicly available in the Smith Lab Github repository (https://github.com/BSmithLab/MicrobiomeMetaboliteAnalysis).

Competing Interests

The authors declare no competing interests.

Funding

The supporting grant for this project was ADHS18-198858.

Authors’ Contributions

TRP and BSS designed the study. TRP and JE performed the experiments and acquired all data used in the study. Data analysis was performed by TRP and JE. TRP wrote the manuscript, and JE and BSS helped revise the manuscript. All authors read and approved the manuscript.

Acknowledgements

The authors would like to acknowledge Dr. Christopher Plaisier in the School of Biological Health Systems and Engineering at Arizona State University for his assistance in analyzing metabolic pathways from the microbial genome. The authors would also like to acknowledge Dr. Trenton Davis in the School of Life Sciences at Arizona State University for his general microbiological guidance.

Postuma RB, et al. MDS clinical diagnostic criteria for Parkinson's disease. Mov Disord. 2015;30:1591–601.
LeWitt PA. Levodopa for the treatment of Parkinson's disease. N Engl J Med. 2008;359:2468–76.
Beckers M, Bloem BR, Verbeek MM. Mechanisms of peripheral levodopa resistance in Parkinson’s disease. npj Parkinson's Disease. 2022;8:56.
Rekdal VM, Bess EN, Bisanz JE, Turnbaugh PJ, Balskus EP. Discovery and inhibition of an interspecies gut bacterial pathway for Levodopa metabolism. Science. 2019;364:eaau6323.
van Kessel SP, et al. Gut bacterial deamination of residual levodopa medication for Parkinson’s disease. BMC Biol. 2020;18:1–14.
Goldstein DS, Holmes C, Sharabi Y. Cerebrospinal fluid biomarkers of central catecholamine deficiency in Parkinson’s disease and other synucleinopathies. Brain. 2012;135:1900–13.
Andersen AD, Blaabjerg M, Binzer M, Kamal A, Thagesen H, Kjaer TW, Stenager E, Gramsbergen JB. Cerebrospinal fluid levels of catecholamines and its metabolites in Parkinson's disease: effect of l-DOPA treatment and changes in levodopa‐induced dyskinesia. J Neurochem. 2017;141:614–25.
Zheng J, Mandal R, Wishart DS. A sensitive, high-throughput LC-MS/MS method for measuring catecholamines in low volume serum. Anal Chim Acta. 2018;1037:159–157.
Peng B, Li H, Peng XX. Functional metabolomics: from biomarker discovery to metabolome reprogramming. Protein Cell. 2015;6:628–37.
Chagoyen M, Pazos F. Tools for the functional interpretation of metabolomic experiments. Brief Bioinform. 2013;14:737–44.
Wishart DS. Advances in metabolite identification. Bioanalysis. 2011;3:1769–82.
Beckonert O, et al. Metabolic profiling, metabolomic and metabonomic procedures for NMR spectroscopy of urine, plasma, serum and tissue extracts. Nat Protoc. 2007;2:2692.
Bos LD, Sterk PJ, Schultz MJ. Volatile metabolites of pathogens: a systematic review. PLoS Pathog. 2013;9:e1003311.
Parekh PJ, Balart LA, Johnson DA. The influence of the gut microbiome on obesity, metabolic syndrome and gastrointestinal disease. Clin translational Gastroenterol. 2015;6:e91.
Haick H. Volatile Biomarkers for Human Health: From Nature to Artificial Senses, (2022).
Nyangale EP, Mottram DS, Gibson GR. Gut microbial activity, implications for health and disease: the potential role of metabolite analysis. J Proteome Res. 2012;11:5573–85.
Chen MX, Wang SY, Kuo CH, Tsai IL. Metabolome analysis for investigating host-gut microbiota interactions. J Formos Med Assoc. 2019;118:10–S22.
Papadimitropoulos ME, Vasilopoulou CG, Maga-Nteve C, Klapa MI, Untargeted GC-MS. Metabolomics Metabolic profiling: Methods and protocols, 133–47 (2018).
Zheng X, et al. Quantitative analysis of neurochemical panel in rat brain and plasma by liquid chromatography-tandem mass spectrometry. Anal Chem. 2012;84:10044–51.
Eshima J, et al. An engineered culture vessel and flow system to improve the in vitro analysis of volatile organic compounds. Preprint at. 2023. https://doi.org/10.1101/2023.08.05.552027.
Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J Roy Stat Soc: Ser B (Methodol). 1995;57:289–300.
Sousa M, Mulaosmanovic E, Erdei AL, Bengtsson M, Witzgall P, Alsanius BW. Volatilomes reveal specific signatures for contamination of leafy vegetables with Escherichia coli O157: H7. Food Control. 2023;146:109513.
Fitzgerald S, Duffy E, Holland L, Morrin A. Multi-strain volatile profiling of pathogenic and commensal cutaneous bacteria. Sci Rep. 2020;10:17971.
Rekdal VM, et al. A widely distributed metalloenzyme class enables gut microbial metabolism of host-and diet-derived catechols. Elife. 2020;9:e50845.
Bargossi E et al. Tyrosine decarboxylase activity of enterococci grown in media with different nutritional potential: tyramine and 2-phenylethylamine accumulation and tyrDC gene expression. Front Microbiol, 259 (2015).
van Kessel SP, et al. Gut bacterial tyrosine decarboxylases restrict levels of levodopa in the treatment of Parkinson’s disease. Nat Commun. 2019;10:310.
Connil N, et al. Identification of the Enterococcus faecalis tyrosine decarboxylase operon involved in tyramine production. Appl Environ Microbiol. 2002;68:3537–44.
Perez M, et al. The relationship among tyrosine decarboxylase and agmatine deiminase pathways in Enterococcus faecalis. Front Microbiol. 2017;8:2107.
Weisskopf L, Schulz S, Garbeva P. Microbial volatile organic compounds in intra-kingdom and inter-kingdom interactions. Nat Rev Microbiol. 2021;19:391–404.
Ratiu IA, et al. The effect of growth medium on an Escherichia coli pathway mirrored into GC/MS profiles. J Breath Res. 2017;11:036012.
Huang X, et al. Laboratory-scale fermentation and multidimensional screening of lactic acid bacteria from Daqu. Food Bioscience. 2021;40:100853.
Wang C, et al. Potential of lactic acid bacteria to modulate coffee volatiles and effect of glucose supplementation: fermentation of green coffee beans and impact of coffee roasting. J Sci Food Agric. 2019;99:409–20.
Zhao J, Fleet G. The effect of lactic acid bacteria on cocoa bean fermentation. Int J Food Microbiol. 2015;205:54–67.
Million M, Armstrong N, Khelaifia S, Guilhot E, Richez M, Lagier JC, Dubourg G, Chabriere E, Raoult D. The antioxidants glutathione, ascorbic acid and uric acid maintain butyrate production by human gut clostridia in the presence of oxygen in vitro. Sci Rep. 2020;10:7705.
Wang Q, et al. Developing clostridia as cell factories for short-and medium-chain ester production. Front Bioeng Biotechnol. 2021;9:661–94.
Liu Y, et al. Clostridium sporogenes uses reductive Stickland metabolism in the gut to generate ATP and produce circulating metabolites. Nat Microbiol. 2022;7:695–706.
Joshi S, Robles A, Aguiar S, Delgado AG. The occurrence and ecology of microbial chain elongation of carboxylates in soils. ISME J. 2021;15:1907–18.
Guo P, Zhang K, Ma X, He P. Clostridium species as probiotics: potentials and challenges. Journal of animal science and biotechnology 11, 1 – 0 (2020).
Dong X, et al. Genetic manipulation of the human gut bacterium Eggerthella lenta reveals a widespread family of transcriptional regulators. Nat Commun. 2022;13:7624.
Noecker C, et al. Systems biology elucidates the distinctive metabolic niche filled by the human gut microbe Eggerthella lenta. PLoS Biol. 2022;21:3002125.
Rogers HJ. Importance of pyrimidine derivatives in the growth of Group C (Lancefield) Streptococci upon a simplified medium. Nature. 1944;153:251–1.
Lauritsen I, et al. Temporal evolution of master regulator Crp identifies pyrimidines as catabolite modulator factors. Nat Commun. 2021;12:5880.
Watanabe S, et al. Intensive DNA replication and metabolism during the lag phase in cyanobacteria. PLoS ONE. 2015;10:e0136800.
Sumner LW, et al. Proposed minimum reporting standards for chemical analysis: chemical analysis working group (CAWG) metabolomics standards initiative (MSI). Metabolomics. 2007;3:211–21.
Haug K, et al. MetaboLights: a resource evolving in response to the needs of its scientific community. Nucleic Acids Res. 2020;48:D440–4.

No competing interests reported.

AdditionalInformation.pdf

Identification of volatile metabolites produced from gut microbial levodopa metabolism using an untargeted metabolomics approach

Status:

Version 1

Abstract

Figures

Introduction

Results

Discussion

Methods

Bacterial growth and incubation

VOC Collection

Gas Chromatography-Mass Spectrometry Analysis

Data Processing

Statistical Analysis

Enzyme Inhibition

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1