Capturing Actively Produced Microbial Volatile Organic Compounds from Human-Associated Samples with Vacuum-Assisted Sorbent Extraction

Mono- and co-cultures of S. aureus, P. aeruginosa, and A. baumannii
The mono- and co-cultures consisted of the bacterial species S. aureus, P. aeruginosa, and A. baumannii. These are common opportunistic pathogens found in human wounds and chronic infections. To identify the volatile molecules present in the mono- and co-cultures, a short 1-h extraction was performed at 70 °C with 200 rpm agitation. From the mono- and co-cultures at 24- and 48-h timepoints, 43 annotated volatile molecules were detected (Figure 2) among which were aldehydes, ketones, alcohols, sulfuric compounds, hydrocarbons, carboxylic acids or esters, and aromatics. There were a small number of volatile molecules that were only detected in certain mono- or co-cultures at certain timepoints. For example, acetoin and 3-hydroxy-2-butanone acetate were only detected in the S. aureus cultures at the 48-h timepoint (Figure 2).

Volatile 1-propanol 2-methyl was detected only in the P. aeruginosa and A. baumannii co-culture at 48 h (Figure 2). Ethyl acetate was present in A. baumannii co-cultures with either S. aureus or P. aeruginosa at 48 h (Figure 2). The metabolites heptane, 2,3-dimethyl and pentane, 2-methyl were only detected in the A. baumannii culture at 24 h (Figure 2). Acetaldehyde and ethanol had higher relative abundances in the A. baumannii and S. aureus co-culture at the 24-h timepoint compared to 48 h and either of the strains in culture alone (Figure 2). Some of the volatiles were more abundant in cultures at either the 24- or 48-h timepoint. Short-chain fatty acids, including acetic acid, butanoic acid, and propanoic acid, were at high relative abundances in cultures at 48 h, but were not detected in the 24-h cultures (Figure 2). Hexane was more abundant in the TH control at 24 h compared to 48 h (Figure 2).

Stable isotope labeling of fecal, sewage, and saliva samples
To identify active production of volatile molecules from a biological sample, a labeled nutrient source, ¹³C glucose or D₂O, and media were added to support the growth of the microbial community. One unique sample was analyzed from each of the different sample types of fecal, sewage, and saliva samples in triplicate. There was more incorporation of the ¹³C into fully labeled volatile molecules (Figure 3A–D) compared to incorporation with deuterium (Figure 3E). The ¹³C was incorporated into 2-butanone, 3-hydroxy; 2,3-butanedione; acetic acid; and phenol for all fecal, sewage, and saliva samples (Figure 3A).

The other labeled volatiles were detected in either two or one sample types. For example, acetone, butanoic acid, and propanoic acid were detected as labeled in saliva and sewage (Figure 3B). The labeled volatiles, dimethyl trisulfide and disulfide dimethyl, were enriched in both fecal and saliva samples (Figure 3C). Volatiles, 1-propanol, 2-butanone, benzophenone, ethanol, and methyl thiolacetate, were enriched only in sewage (Figure 3D). The labeled volatile, 2,3-pentanedoine, was enriched in saliva (Figure 3D). Deuterium was incorporated into the volatiles, acetic acid; benzaldehyde, 4-methyl; dimethyl trisulfide; and phenol, from either saliva or sewage samples (Figure 3E). In addition to the isotope-enriched volatiles, there were volatiles detected that did not contain incorporated stable isotopes. For example, pyrazine compounds, except for pyrazine, 2,5-dimethyl, were detected in fecal, sewage, and saliva samples, but were not fully enriched with ¹³C (Supplemental Figure S1).

Stable isotope labeling of sputum samples
The stable isotope labeling strategy was implemented for identifying actively produced volatiles with sputum samples from seven human subjects with cystic fibrosis. The volatiles in the sample were compared with those that emerged from samples cultured with a stable isotope label. Each volatile component of each sample was analyzed twice: before and after stable isotope probing with ¹³C glucose and media. The samples collected from the subjects spanned three different timepoints or clinical states: baseline, exacerbation, and treatment²³. The volatiles detected as labeled in the cultured sputum samples had different relative abundances compared to the unlabeled volatiles from the uncultured sputum samples. Culturing conditions in the stable isotope probing experiments with sputum may favor the growth of certain microbes, leading to differences in relative abundances of volatiles compared to the uncultured sputum samples.

For example, acetic acid, dimethyl trisulfide, acetone, and propanal, 2-methyl were more abundant in the cultured sputum samples compared to the uncultured sputum samples (Figure 4). Detecting ¹³C-labeled ethanol, which can be present in variable amounts in the background room air, provides evidence that the ethanol was actively produced by microbial metabolism from ¹³C glucose. The amount of variation was explained by subject as assessed by Permutational Multivariate Analysis of Variance (PERMANOVA) and was also different for the two different volatile datasets (Table 1 and Supplemental Figure S2). For the ¹³C-labeled cultured sputum, 51% of the variation was explained by the subject, while 33% of the variation was explained by subject from the volatiles in the uncultured sputum samples (Table 1). The microbiome community composition as determined by 16S rRNA amplicon sequencing from the seven subjects was unique to each subject (Supplemental Figure S3), and the individual signatures were also reflected in both the cultured and uncultured sputum volatile molecules detected.

In cultured sputum, 23 volatiles were detected that were fully labeled with ¹³carbon. The isotope-enriched (active) volatiles detected from the sputum samples were different for each subject. The volatiles with isotope enrichment detected in the sputum samples from all seven subjects were 2,3-butanedione; acetic acid; acetone; dimethyl trisulfide; disulfide, dimethyl; and pyrazine, 2,5-dimethyl (Figure 5). Although those volatiles were detected in all subjects, the isotope enrichment for each subject varied. Samples from subject 7 had higher isotope enrichment of disulfide dimethyl compared to the other six subjects (Figure 5B). Acetone was higher in subjects 4 and 6 (Figure 5). Some volatiles were enriched with ¹³C only in certain subjects. For example, 1-butanol, 3-methyl and propanoic acid, 2-methyl were only enriched in a subset of samples from subject 2 (Figure 5). In addition to the isotope-enriched volatiles, there were volatiles also detected as unlabeled from the same cultured sputum (Supplemental Figure S4). Volatiles 2-piperidinone; benzaldehyde, 4-methyl; benzothiazole; butanoic acid, 3-methyl; hexanal; hexane; isopropyl alcohol; phenol; propanoic acid, 2-methyl; and pyrrolo 1,2-apyrazine-1,4-dione, hexahydro were detected in the sputum samples, but were not isotope-enriched (Supplemental Figure S4).

Figure 1: Protocol schematic. A biological sample is placed into a glass vial and assembled with the lid liner and Headspace Sorbent Pen. A vacuum is applied to the vial until a pressure of approximately 30 mmHg is reached. The vacuum source is removed, and the vials are placed in the sorbent pen extraction unit where a static extraction is performed with the aid of heat, agitation, and time. After extraction, vials are placed on a cold metal block to remove water from the headspace and HSP. The HSPs are collected and run via thermal desorption on the GC-MS. The data are analyzed with ChemStation, DExSI, and R. Abbreviations: HSP = Headspace Sorbent Pen; GC-MS = gas chromatography-mass spectrometry. Please click here to view a larger version of this figure.

Figure 2: Heatmap of mono- and co-cultures. VOCs detected from mono- and co-cultures at 24- and 48-h timepoints. The co-cultures are the combinations of the letters representing each strain. All samples were extracted for 1 h at 70 °C with 200 rpm agitation. Heatmap intensity values are column Z-scores, normalized by metabolite. The Z-score was calculated by the difference of value from the mean of values, divided by the standard deviation of values. The dendrogram was generated with the cluster_cols option in the pheatmap function of R. The dendrogram represents hierarchical clustering in which metabolites that cluster together have more similar Z-scores across samples. Abbreviations: A = A. baumanii; P = P. aeruginosa; S = S. aureus; TH = Todd Hewitt media (control). Please click here to view a larger version of this figure.

Figure 3: Percent conversion of ¹³C into volatile molecule mass in fecal, saliva, and sewage samples during 18 h of simultaneous incubation and extraction. The % conversion was calculated for fully labeled compounds by taking the mass of the fully labeled compound (M+N) and dividing it by (M+N) + the mass of the unlabeled volatile mass (M), where N is the maximum number of possible carbons (in A-D) or hydrogens (in E) that can be labeled in each volatile molecule. Compounds are considered to be fully labeled when all carbons of the volatile are replaced by ¹³C. Where data are missing, the volatile was not detected. For example, in (D), 1-propanol was not detected in fecal or saliva samples. Number of replicates per sample = 3. (A) The ¹³C-labeled volatiles detected in all sample types (feces, saliva, and sewage). (B) The ¹³C-labeled volatiles only detected in saliva and sewage samples. (C) The ¹³C-labeled volatiles detected in feces and saliva samples. (D) The ¹³C-labeled volatiles detected in one of the three different sample types. (E) The deuterium-labeled volatile molecules. Please click here to view a larger version of this figure.

Figure 4: Heatmap of ¹³C-labeled volatiles from cultured sputum and volatile molecules detected from uncultured sputum. The labeled volatiles come from the stable isotope probing experiments where ¹³C glucose and Brain Heart Infusion medium were added to sputum during the extraction step to cultivate microbial growth and capture active volatile production. The unlabeled volatile molecules were detected directly from sputum samples. The heatmap intensities are Z-scores as described in the caption of Figure 2. However, the Z-scores were calculated within each experiment for the cultured and uncultured sputum experiments. The dendrogram was generated as described in Figure 2. Please click here to view a larger version of this figure.

Figure 5: Percent conversion of ¹³C into volatile molecule mass in sputum samples from seven subjects with cystic fibrosis during 18 h of simultaneous incubation and extraction. The % conversion was calculated as described in the caption of Figure 3. Volatiles not detected in samples are indicated by the absence of data. N = 1-3. (A) The ¹³C-labeled volatiles detected at a higher percent conversion in the majority of sputum samples. (B) The ¹³C-labeled volatiles detected at a lower percent conversion in the majority of sputum samples. (C) The ¹³C-labeled volatiles detected at a lower percent conversion in a minority of the sputum samples. Abbreviations: B = baseline; E = exacerbation; T = treatment. Please click here to view a larger version of this figure.

Table 1: Permutated multivariate analysis of variance (PERMANOVA) of sputum samples. The PERMANOVA was generated using the adonis function from the vegan package in R.

Supplemental Figure S1: The relative abundances of labeled (M+N(max)) and unlabeled (M+0) volatiles across fecal, saliva, and sewage samples. Please click here to download this File.

Supplemental Figure S2: Non-metric multidimensional scaling of cultured sputum with stable isotope probing and uncultured sputum. (A) The NMDS of cultured sputum with ¹³C glucose and media was generated with k = 3 dimensions. The stress value was 0.07. (B) The NMDS of uncultured sputum was generated with k = 3 dimensions. The stress value was 0.13. Abbreviations: NMDS = non-metric multidimensional scaling; B = baseline; E = exacerbation; T = treatment. Please click here to download this File.

Supplemental Figure S3: Microbial community composition of sputum samples from subjects with cystic fibrosis. Assessed by 16S rRNA amplicon sequencing as part of a larger study, further information about approach found in Carmody et al. 2020¹⁹. from subjects with cystic fibrosis. Each stacked bar is a different timepoint. Abbreviations: B = baseline, E = exacerbation, T = treatment. Please click here to download this File.

Supplemental Figure S4: The relative abundances of labeled (M+N(max)) and unlabeled (M+0) volatiles across sputum samples from seven subjects with cystic fibrosis. Please click here to download this File.