Construction of a Bacterial Lipidomics Analytical Platform: Pilot Validation with Bovine Paratuberculosis Serum

Lipidomics analyses of bacteria offer the potential to detect and monitor infections in a host since many bacterial lipids are not present in mammals. To evaluate this omics approach, we first built a database of bacterial lipids for representative Gram-positive and Gram-negative bacteria. Our lipidomics analysis of the reference bacteria involved high-resolution mass spectrometry and electrospray ionization with less than a 1.0 ppm mass error. The lipidomics profiles of bacterial cultures clearly distinguished between Gram-positive and Gram-negative bacteria. In the case of bovine paratuberculosis (PTB) serum, we monitored two unique bacterial lipids that we also monitored in Mycobacterium avian subspecies PTB. These were PDIM-B C82, a phthiodiolone dimycocerosate, and the trehalose monomycolate hTMM 28:1, constituents of the bacterial cell envelope in mycolic-containing bacteria. The next step will be to determine if lipidomics can detect subclinical PTB infections which can last 2-to-4 years in bovine PTB. Our data further suggest that it will be worthwhile to continue building our bacterial lipidomics database and investigate the further utility of this approach in other infections of veterinary and human clinical interest.


Introduction
Lipidomics is a rapidly evolving "omics" platform that provides valuable information regarding structural, energy source/reserve, and signal-transduction lipid pools. Bacteria possess a number of unique lipids that are not present in their mammalian hosts. This provides the opportunity of lipidomics to obtain valuable non-mammalian lipid data that can (i) detect bacterial infection in a host, (ii) monitor the progression of an infection, (iii) monitor the efficacy of treatments on an infection, and (iv) potentially define new targets in the design of targeted antimicrobial therapeutics.
While the individual lipids of a given lipid family for a bacterial strain will alter with development and with environmental stresses, lipid families will be preserved and can be monitored. Our first high-level overview is a comparison of our current knowledge base for Gram-positive vs. Gram-negative bacterial lipidomics.
The diversity of amino acyl lipidomes between different bacterial species is demonstrated by the detection of lysyl-PG in only 5 of 24 clostridia species examined [9] and the detection of alanyl-PG in only 3 of 24 clostridia species examined [9].

Gram-Positive Bacteria: Mycolic Acids
A very unique family of glycolipids is also present in the outer wall of a number of bacteria in the Actinomycetes taxonomic group. These are the mycolic acids present in mycolic-acid-containing bacteria (MACB) [1,27] which include Mycobacteria (M. tuberculosis, M. leprae, M. bovis, Tskamurella pulmonis, Rhodococcus erythropolis, R. opacus, and R. equi) and Corynebacteria (C. glutamicum) [27,28]. Long-chain mycolic acids are covalently bound in the inner layer of the cell wall but are present as free acids in the outer domain. Lipids in this lipid family are diverse (Table 3). Table 3. Gram-positive bacterial mycolic acids.

Gram-Positive Bacteria: Mycolic Acids
A very unique family of glycolipids is also present in the outer wall of a number of bacteria in the Actinomycetes taxonomic group. These are the mycolic acids present in mycolic-acid-containing bacteria (MACB) [1,27] which include Mycobacteria (M. tuberculosis, M. leprae, M. bovis, Tskamurella pulmonis, Rhodococcus erythropolis, R. opacus, and R. equi) and Corynebacteria (C. glutamicum) [27,28]. Long-chain mycolic acids are covalently bound in the inner layer of the cell wall but are present as free acids in the outer domain. Lipids in this lipid family are diverse (Table 3). Table 3. Gram-positive bacterial mycolic acids.
Fatty acyls of hydroxy fatty acids (FAHFAs) [49] are present at high concentrations in Gram-negative bacteria, and both the glycosylated and aminoacyl forms are critical membrane constituents. The glycosylation of hydroxy fatty acids yields rhamnolipids, which act as biosurfactant antimicrobials. Representative glycolipids in Gram-negative bacteria are presented in Table 6. Table 6. Gram-negative bacterial glycosyl hydroxy fatty acids (HFAs) and fatty acyls of hydroxy fatty acids (FAHFAs).

Gram-Negative Bacteria: Modified Ceramides
Gram-negative bacteria possess several unique modified ceramides which are considered to contribute to membrane charge (Table 8). Table 8. Gram-negative-bacterial-modified ceramides.

Gram-Negative Bacteria: Sterols
Gram-negative bacteria utilize several unique cholesteryl acyl-glycosides as immunostimulants and hopanoids which order membrane lipids and regulate membrane permeability [97] (Table 9).   [12,13] The diversity of amino acyl lipidomes between different bacterial species is demonstrated by the detection of lysyl-PG in only 5 of 24 clostridia species examined [9] and the detection of alanyl-PG in only 3 of 24 clostridia species examined [9].

Gram-Positive Bacteria: Mycolic Acids
A very unique family of glycolipids is also present in the outer wall of a number of bacteria in the Actinomycetes taxonomic group. These are the mycolic acids present in mycolic-acid-containing bacteria (MACB) [1,27] which include Mycobacteria (M. tuberculosis, M. leprae, M. bovis, Tskamurella pulmonis, Rhodococcus erythropolis, R. opacus, and R. equi) and Corynebacteria (C. glutamicum) [27,28]. Long-chain mycolic acids are covalently bound in the inner layer of the cell wall but are present as free acids in the outer domain. Lipids in this lipid family are diverse (Table 3).

Lipid Class Bacterial Strains References
Precursor CDP-DG  [12,13] The diversity of amino acyl lipidomes between different bacterial species is demonstrated by the detection of lysyl-PG in only 5 of 24 clostridia species examined [9] and the detection of alanyl-PG in only 3 of 24 clostridia species examined [9].

Gram-Positive Bacteria: Mycolic Acids
A very unique family of glycolipids is also present in the outer wall of a number of bacteria in the Actinomycetes taxonomic group. These are the mycolic acids present in mycolic-acid-containing bacteria (MACB) [1,27] which include Mycobacteria (M. tuberculosis, M. leprae, M. bovis, Tskamurella pulmonis, Rhodococcus erythropolis, R. opacus, and R. equi) and Corynebacteria (C. glutamicum) [27,28]. Long-chain mycolic acids are covalently bound in the inner layer of the cell wall but are present as free acids in the outer domain. Lipids in this lipid family are diverse (Table 3). Gram-negative bacteria produce a number of secondary metabolites that they utilize to protect against other microbes (Table 10). Table 10. Gram-Negative Bacterial Secondary Metabolites.

Lipid Class Bacterial Strains References
Undecylprodigiosin metabolites Streptomyces spp., Serratia marcescens [102][103][104] Malleilactone Burkholderia pseudomallei [105] In summary, the wide diversity of bacterial lipids offers the potential to differentiate different bacterial species via lipidomics analyses. For example, previous studies of polar lipids in Clostridia spp. in four different groups of bacteria based on morphological and biochemical criteria demonstrated that three of the four groups possessed lipids that distinguished each group. All groups had high levels of PE and PG. However, Group I (C. sporogenes prototype) possessed PE-NAcGlu-DGs, Group II (C. butyricum prototype) possessed glycerol and PG acetals of ethanolamine plasmalogens, Group III (C. novyi prototype) possessed aminoacyl-PGs, and Group IV (C. subterminale prototype) had no distinguishing polar lipids [106,107]. Extending future lipidomics analyses across a broader scope than just polar lipids should further increase our ability to differentiate ongoing bacterial infections.
The objective of our study was to initiate building a bacterial lipidomics database that we could utilize to interrogate serum from cows infected with paratuberculosis and provide the groundwork required to continue building and expanding the database such that it will allow for the interrogation of other clinically relevant infections.

Bacterial Processing
Bacterial pellets purchased from the ATTC (Manassas, VA, USA) were sonicated (Thermo Fisher FB50) in 1 mL of methanol and 1 mL of water containing 2 nanomoles of [ 13 C 3 ]DG 36:2 (Larodan, Monroe, MI, USA). Next 2 mL of tert-butylmethylether was added, and the samples were shaken at room temperature for 30 min (Thermo Fisher Multitube Vortexer, Waltham, MS, USA). Next, the samples were centrifuged at 4000× g for 30 min at room temperature. From the upper organic layer of these centrifuged samples, 1 mL aliquots were transferred to a deep-well microplate. The microplate samples were dried via vacuum centrifugation (Eppendorf Vacfuge Plus, Hamburg, Germany).
The Gram-positive bacterial pellets which we evaluated were Mycobacterium avium,

Lipidomics Analysis
We utilized published data and lipid databases for bacterial lipids and then incorporated them into our established lipidomics analytical platform [106,[108][109][110][111][112] such that now we can interrogate approximately 11,000 individual lipids. As a pilot to evaluate the utility of this platform to detect active bacterial infections, we utilized the platform to examine the lipidome of a number of representative Gram-positive and Gram-negative bacteria and plasma samples from cows with paratuberculosis [112].
For MS/MS analyses, parent ions were selected with a 0.4 amu window and collision energies of 15, 30, and 50 arbitrary units. Product ions were monitored with a resolution of 240,000. Product ions with a <1.0 ppm mass error are listed in Supplementary Table S3. We utilized Lincoln Memorial University, Metabolomics Unit, Flow Infusion Lipidomics Analytical Platform (Version 1.0).

Bovine PTB Serum Samples
Serum samples (100 µL) from our previous research [112] were used for this study and processed as described above. The cattle (n = 10) were 2-to-2.5-year-old angus. PTB infection was confirmed utilizing enzyme-linked immunosorbent assay (ELISA) (IDEXX MAP ELISA Ab Test kit, Westbrood, ME, USA). All testing was performed at the University of Kentucky Veterinary Diagnostic Laboratory (UKVDL), a fully accredited laboratory of the American Association of Veterinary Laboratory Diagnosticians (AAVLD). For each lipid in the Excel mass list, the imported data were searched for a matching mass with <1.0 ppm mass error. For positive hits, the extracted mass and the associated peak intensity were imported into a new active spreadsheet. The specific details for each lipid class, along with the associated ionization modes and MS/MS products, are presented in Supplementary Table S3, which details all the lipid classes included in our lipidomics analytical platform, along with citations for representative publications.

M. avium Specific Lipids: Phthiodiolone Dimycocerosates and Diacyltrehaloses
M. avium was unique in that it was the only Gram-positive species we examined that possessed phthiodiolone dimycocerosates and diacyltrehaloses (Figure 1 and Supplementary  Table S1). The phthiodiolone dimycocerosates (PDIMs) are long-chain β-diols esterified at the hydroxy groups with multimethyl-branched fatty acids (mycocerosic acids). We specifically monitored PDIM-B forms in which a position 2 of the diol is a keto group. The dominant member of this lipid family was PDIM-B C82 in the ATCC bacterial pellets and was detected in the serum of cattle with paratuberculosis but not in control cows (Figure 1 and Supplementary Table S1). In contrast, while we detected diacyltrehaloses in the M. avium bacterial pellet (Supplementary Table S1), these lipids were undetectable in the serum of infected cows. The diacyltrehaloses were in the DAT2 family which have a fatty acid (16:0 to 19:0) and a mycolipanolic fatty acid substituent. The mycolipanolic fatty acids were 3-hydroxy-2,4,6-methyl fatty acids of 24 to 28 carbons.
For each lipid in the Excel mass list, the imported data were searched for a matchi mass with <1.0 ppm mass error. For positive hits, the extracted mass and the associat peak intensity were imported into a new active spreadsheet. The specific details for ea lipid class, along with the associated ionization modes and MS/MS products, are p sented in Supplementary Table S3, which details all the lipid classes included in our l idomics analytical platform, along with citations for representative publications.

M. avium Specific Lipids: Phthiodiolone Dimycocerosates and Diacyltrehaloses
M. avium was unique in that it was the only Gram-positive species we examined th possessed phthiodiolone dimycocerosates and diacyltrehaloses (Figure 1 and Suppleme tary Table S1). The phthiodiolone dimycocerosates (PDIMs) are long-chain β-diols este fied at the hydroxy groups with multimethyl-branched fatty acids (mycocerosic acid We specifically monitored PDIM-B forms in which a position 2 of the diol is a keto grou The dominant member of this lipid family was PDIM-B C82 in the ATCC bacterial pell and was detected in the serum of cattle with paratuberculosis but not in control cows (F ure 1 and Supplementary Table S1). In contrast, while we detected diacyltrehaloses in t M. avium bacterial pellet (Supplementary Table S1), these lipids were undetectable in t serum of infected cows. The diacyltrehaloses were in the DAT2 family which have a fa acid (16:0 to 19:0) and a mycolipanolic fatty acid substituent. The mycolipanolic fatty aci were 3-hydroxy-2,4,6-methyl fatty acids of 24 to 28 carbons.

Trehalose Mycolates
Hydroxy-trehalose monomycolates (hTMMs) were monitored in all of the examined bacteria except for S. aureus and E. faecalis (Supplementary Table S1). Each bacterial strain had a different dominant hTMM. In the case of M. avium, hTMM 28:1 was the dominant member of the lipid family and was also detected in the serum of PTB-positive cattle ( Figure 2). While acetylTMMs were monitored in M. avium and a number of other Grampositive bacteria (Supplementary Table S1), we did not detect any of this lipid family in the serum of infected cows.
bacteria except for S. aureus and E. faecalis (Supplementary Table S1). Each bacterial stra had a different dominant hTMM. In the case of M. avium, hTMM 28:1 was the domina member of the lipid family and was also detected in the serum of PTB-positive cattle (F ure 2). While acetylTMMs were monitored in M. avium and a number of other Gram-po itive bacteria (Supplementary Table S1), we did not detect any of this lipid family in t serum of infected cows.

Lipoteichoic Acid Precursors
Lipoteichoic acid precursors (LTAPs; dihexosyldiacylglycerol-glycerol phosphat along with the mono-alanine and di-alanine analogs, were not detected in the M. aviu bacterial pellet (Supplementary Table S1). LTAP 32:0 was monitored in S. aureus, wh LTAPs and Ala-LTAPs were monitored in the bacterial pellets from R. equi, E. faecalis, a C. glutamicum. As with other lipids, the dominant LTAP lipid family member was differe for each bacterial strain. Di-Ala-LTAPs were detected only in R. equi bacterial pellets.

Mannosyl Phosphoinositols (PIM1)
Acyl-PIM1 family members were only monitored in the C. glutamicum bacterial p lets (Supplementary Table S1), consistent with prior studies [16]. The acyl-PIM1 fam has also been reported for a number of Mycobacteria [41]; however, we did not detect a acyl-PIM1 in the Mycobacteria we studied. This may have resulted from low levels and/ ion suppression.

Lipoteichoic Acid Precursors
Lipoteichoic acid precursors (LTAPs; dihexosyldiacylglycerol-glycerol phosphate), along with the mono-alanine and di-alanine analogs, were not detected in the M. avium bacterial pellet (Supplementary Table S1). LTAP 32:0 was monitored in S. aureus, while LTAPs and Ala-LTAPs were monitored in the bacterial pellets from R. equi, E. faecalis, and C. glutamicum. As with other lipids, the dominant LTAP lipid family member was different for each bacterial strain. Di-Ala-LTAPs were detected only in R. equi bacterial pellets.

Mannosyl Phosphoinositols (PIM1)
Acyl-PIM1 family members were only monitored in the C. glutamicum bacterial pellets (Supplementary Table S1), consistent with prior studies [16]. The acyl-PIM1 family has also been reported for a number of Mycobacteria [41]; however, we did not detect any acyl-PIM1 in the Mycobacteria we studied. This may have resulted from low levels and/or ion suppression.

Mycolic Acids
All of the Gram-positive bacteria that we studied were found to contain mycolic acids (Figures 3 and 4; Supplementary Table S1). A diverse array of mycolic acids was monitored in the bacterial pellets. Most mycolic acids are tethered in the outer membrane, but there are small membrane levels of free mycolic acids [31][32][33][34]113], as demonstrated in Figures 3 and 4. For the unsaturated lipids, our data do not distinguish between a double bond or a cyclopropyl substitution [113]. Both M. bovis and M. smegmatis mycolic acids were skewed to a distribution of longer-chain fatty acyl substituents (Figure 4). Interestingly, only these two bacterial strains had measurable levels of epoxymycolic acids (Supplementary Table S1). It also needs to be noted that our analyses do not distinguish between the isobars of oxygenated lipids [113]. For example, epoxymycolic acid 77:1 = ketomycolic acid 77:1 = methoxymycolic acid 77:2.
bond or a cyclopropyl substitution [113]. Both M. bovis and M. smegmatis mycolic acids were skewed to a distribution of longer-chain fatty acyl substituents (Figure 4). Interestingly, only these two bacterial strains had measurable levels of epoxymycolic acids (Supplementary Table S1). It also needs to be noted that our analyses do not distinguish between the isobars of oxygenated lipids [113]. For example, epoxymycolic acid 77:1 = ketomycolic acid 77:1 = methoxymycolic acid 77:2.  Supplementary Table S1. With FIA, we cannot distinguish between double bonds and cyclopropane groups in the mycolic acids designated as unsaturated.
Dicarboxylic mycolic acids were only detected in M. avium and M. bovis (Supplementary Table S1).
The complexity of mycolic acids in bacteria was reflected in our analysis of the serum from cows infected with PTB. Four of the ten cows had levels of mycolic acid 50:2 (0.0011 ± 0.00064), five cows had dicarboxylic acid 82:1 (0.0053 ± 0.00065), two cows had dicarboxylic mycolic acid 84:1, one cow had dicarboxylic mycolic acid 82:2, and one cow had dicarboxylic mycolic acid 84:2. These lipids were not detected in the 10 control cows. This heterogeneity of detectable mycolic acids in the serum of infected cows may be reflective of different stages of the PTB infection, which is known to progress slowly over time [114].  Supplementary Table S1. With FIA, we cannot distinguish between double bonds and cyclopropane groups in the mycolic acids designated as unsaturated.   Supplementary Table S1. With FIA, we cannot distinguish between double bonds and cyclopropane groups in the mycolic acids designated as unsaturated.

Glycopeptidolipids (GPLs)
The cell walls of a number of Mycobacteria contain a family of unique GPLs that consist of a hydroxy fatty acid coupled to a peptide which in turn is coupled to rhamnose [115][116][117]. The hydroxy fatty acid has a deoxytalose (dTal) glycation which has 0-to-2 possible acetylations. The peptide is Phe-Thr-Ala-Alaninol, and the terminal rhamnose has 0to-3 possible O-methylations. This lipid family serves as cell-surface antigens.
We monitored an array of GPLs with the rank order of prevalence C. glutamicum > M. smegmatis > R. equi > M. bovis (Supplementary Table S1).

Sulfonolipids
Sulfonolipids are characterized by the replacement of serine in the sphingolipid base by the sulfonic acid capnine generating sulfobacins (monohydroxy) and sulfocristamides (di-hydroxy) [118]. These lipids are required for gliding motility and demonstrate pro-  Supplementary Table S1. With FIA, we cannot distinguish between double bonds and cyclopropane groups in the mycolic acids designated as unsaturated.
Dicarboxylic mycolic acids were only detected in M. avium and M. bovis (Supplementary Table S1).
The complexity of mycolic acids in bacteria was reflected in our analysis of the serum from cows infected with PTB. Four of the ten cows had levels of mycolic acid 50:2 (0.0011 ± 0.00064), five cows had dicarboxylic acid 82:1 (0.0053 ± 0.00065), two cows had dicarboxylic mycolic acid 84:1, one cow had dicarboxylic mycolic acid 82:2, and one cow had dicarboxylic mycolic acid 84:2. These lipids were not detected in the 10 control cows. This heterogeneity of detectable mycolic acids in the serum of infected cows may be reflective of different stages of the PTB infection, which is known to progress slowly over time [114].

Glycopeptidolipids (GPLs)
The cell walls of a number of Mycobacteria contain a family of unique GPLs that consist of a hydroxy fatty acid coupled to a peptide which in turn is coupled to rhamnose [115][116][117]. The hydroxy fatty acid has a deoxytalose (dTal) glycation which has 0-to-2 possible acetylations. The peptide is Phe-Thr-Ala-Alaninol, and the terminal rhamnose has 0-to-3 possible O-methylations. This lipid family serves as cell-surface antigens.
We monitored an array of GPLs with the rank order of prevalence C. glutamicum > M. smegmatis > R. equi > M. bovis (Supplementary Table S1).

Sulfonolipids
Sulfonolipids are characterized by the replacement of serine in the sphingolipid base by the sulfonic acid capnine generating sulfobacins (monohydroxy) and sulfocristamides (di-hydroxy) [118]. These lipids are required for gliding motility and demonstrate proinflammatory and cytotoxic activities [118]. Both C. glutamicum and M. bovis were found to possess these highly charged sphingolipids (Supplementary Table S1).

Alpha-Acyl Hydroxy Fatty Acids (AAHFAs)
AAHFAs are a unique family of FAHFA lipids in which case the acylation is at a hydroxy group on carbon 2, with the acyl substitution being butyric acid [119]. The functions of these newly discovered lipids remain to be elaborated. In our analyses, we found high levels of AAHFAs in M. avium and moderate levels in S. aureus and M. bovis (Supplementary Table S1).

Gram-Negative Bacteria
Gram-negative bacteria lack the cell wall characteristic of Gram-positive bacteria. Lipid A is a major membrane lipid in the cell envelope, comprising an inner and outer membrane with an intermediate peptidoglycan layer. While intact lipid A molecules are large and tethered, a number of lipid A precursors are easily analyzed via conventional lipid-extraction procedures. Modified fatty acyls of hydroxy fatty acids (FAHFAs) are one example of these lipid A constituents that are absent from Gram-positive bacteria.
Gly-Ser-FAHFAs are characteristic of some Gram-negative bacteria [73,77]. We monitored these unique dipeptide lipids in P. mirabilis and M. bovoculi (Supplementary Table S2). Gly-Ser-hydroxy-fatty acids were also monitored in these two bacterial strains, as well as in H. pylori.
Aminoacyl FAHFAs have long been conjectured to play a role in replacing glycerophospholipids in membranes, where they regulate membrane charge. Other studies have also demonstrated their roles in signal transduction. For example, ornithine lipids act at GPCRs involved in immune activation [65]. Similarly, Gly-Ser lipids act at Toll-like 2 receptors involved in immunostimulation [69,70]. bacteria we evaluated. Orn-FAHFA ( Figure 5) and Gly-FAHFA were monitored in all bacteria examined, while Ala-FAHFA was absent from H. pylori (Supplementary Table S2).
Gly-Ser-FAHFAs are characteristic of some Gram-negative bacteria [73,77]. We monitored these unique dipeptide lipids in P. mirabilis and M. bovoculi (Supplementary Table  S2). Gly-Ser-hydroxy-fatty acids were also monitored in these two bacterial strains, as well as in H. pylori.  Supplementary Table S1. Aminoacyl FAHFAs have long been conjectured to play a role in replacing glycerophospholipids in membranes, where they regulate membrane charge. Other studies have also demonstrated their roles in signal transduction. For example, ornithine lipids act at GPCRs involved in immune activation [65]. Similarly, Gly-Ser lipids act at Toll-like 2 receptors involved in immunostimulation [69,70].

Modified Ceramides
The addition of a polar phosphoethanolamine or phosphoglycerol group to ceramides has been shown to be another unique feature of a number of Gram-negative

Modified Ceramides
The addition of a polar phosphoethanolamine or phosphoglycerol group to ceramides has been shown to be another unique feature of a number of Gram-negative bacteria [64,67,[73][74][75][76][77]79]. We monitored a diverse array of these lipids in H. pylori, P. mirabilis, and M. bovocali but not in E. coli or P. aeruginosa (Supplementary Table S2).

Unique Sterols
Cholesteryl-acylphosphoglycosides (CPGs) have been detected in H. pylori [93,94] and Borella burgdorferi [95,96]. We confirm that H. pylori has these unique lipids and report for the first time that P. mirabilis also has these membrane lipids (Supplementary Table S2).

Phosphatidyltrehalose (PT)
Phosphatidyltrehaloses have been reported for Salmonella paratyphi and S. typhi [120]. We report for the first time that these immunostimulant lipids are also present in P. mirabilis and E. coli (Supplementary Table S2).

Discussion
Our data support previous studies demonstrating the stark contrast of the lipidomes of Gram-positive and Gram-negative bacteria. Furthermore, by utilizing a standard lipidextraction procedure, we were able to demonstrate the presence of both PDIM-B C82, a phthiodiolone dimycocerosate, and the trehalose monomycolate hTMM 28:1 in the plasma of cows with PTB. These specific constituents of the bacterial cell envelope in M. avium are the dominant family members we extracted from commercial bacterial pellets. Serum mycolic acids were also detected, but the levels were much more variable. Our data demonstrate the power and specificity of lipidomics to detect bacterial infections. Presumably, targeted assays to provide absolute lipid levels will provide even more specificity and sensitivity.
Lipid biomarkers have been utilized previously to demonstrate the presence of tuberculosis in archaeological samples [121][122][123][124][125] and to monitor Gram-negative bacterial infections in carotid atheroma (Gly-Ser-lipids) [73] and in oral samples from patients with periodontitis [77]. These and our current data support the idea of building a database of microbial lipids of interest to human and veterinary clinical medicine. Such a database will, in turn, yield the data required to determine which lipids might be of value to establish absolute quantitation clinical assays.

Study Limitations
This is the first step in building a comprehensive bacterial lipidomics database that will be expanded as we add the profiles of other bacteria to increase its applicability to bacterial research. Our FIA methodology has the strengths of covering a broader range of lipids and providing a stable and constant background, compared to hybrid chromatographic methods. However, issues with isobars are more prevalent with FIA. To reduce this risk, we utilized HRMS and only accepted lipids that were <1.0 ppm mass error. We also utilized MS 2 to validate the lipid identities. The MS 2 parameters for each lipid class are presented in Supplementary Table S3: Lincoln Memorial University, Metabolomics Unit, Flow Infusion Lipidomics Analytical Platform (Version 1.0).

Supplementary Materials:
The following supporting information, in a single file, can be downloaded at https://www.mdpi.com/article/10.3390/metabo13070809/s1. Table S1: Rank order of lipid families in Gram-positive bacteria. Table S2: Rank order of lipid families in Gram-negative bacteria. Table S3: Lincoln Memorial University, Metabolomics Unit, Flow Infusion Lipidomics Analytical Platform (Version 1.0).
Author Contributions: Both authors were responsible for the conceptualization and conduct of the study. P.L.W. was responsible for the methodology, data reduction software, validation and formal analysis, investigation, resources, data curation, and the original draft preparation. Both authors were responsible for the manuscript review and editing. All authors have read and agreed to the published version of the manuscript.