Metabolomic profiling of oxalate-degrading probiotic Lactobacillus acidophilus and Lactobacillus gasseri

Oxalate, a ubiquitous compound in many plant-based foods, is absorbed through the intestine and precipitates with calcium in the kidneys to form stones. Over 80% of diagnosed kidney stones are found to be calcium oxalate. People who form these stones often experience a high rate of recurrence and treatment options remain limited despite decades of dedicated research. Recently, the intestinal microbiome has become a new focus for novel therapies. Studies have shown that select species of Lactobacillus, the most commonly included genus in modern probiotic supplements, can degrade oxalate in vitro and even decrease urinary oxalate in animal models of Primary Hyperoxaluria. Although the purported health benefits of Lactobacillus probiotics vary significantly between species, there is supporting evidence for their potential use as probiotics for oxalate diseases. Defining the unique metabolic properties of Lactobacillus is essential to define how these bacteria interact with the host intestine and influence overall health. We addressed this need by characterizing and comparing the metabolome and lipidome of the oxalate-degrading Lactobacillus acidophilus and Lactobacillus gasseri using ultra-high-performance liquid chromatography-high resolution mass spectrometry. We report many species-specific differences in the metabolic profiles of these Lactobacillus species and discuss potential probiotic relevance and function resulting from their differential expression. Also described is our validation of the oxalate-degrading ability of Lactobacillus acidophilus and Lactobacillus gasseri, even in the presence of other preferred carbon sources, measuring in vitro 14C-oxalate consumption via liquid scintillation counting.


Introduction
Probiotics, defined by the World Health Organization (WHO) as "live microorganisms which when administered in adequate amounts, confer a health benefit on the host" [1,2], have become widespread in the global health market. Often sold as foods or dietary supplements [3], many different probiotics exist that contain various cocktails of bacteria formulated to deliver specific health benefits ranging from immune system support [4], gastrointestinal regularity [5], serum cholesterol control [6], management of allergic diseases [7], and even relief of mental ailments

Cell Culture, harvest, and Lysis
Pure cultures of L. acidophilus (ATCC TM 4357) and L. gasseri (ATCC TM 33323), the same isolates used by Hatch et al [16] obtained from the American Type Culture Collection (ATCC TM ), were grown anaerobically from frozen 10% glycerol stocks at 37˚C in deMan, Rogosa and Sharpe (MRS) medium [15] supplemented with 20 mM oxalate and 1% glucose. Using a sterile syringe and needle, 8 anaerobic bottles containing 75 mL medium were each inoculated with 150 μL glycerol stock for each species. Cultures were briefly shaken and allowed to incubate overnight at 37˚C for 24 hours. After incubation, cultures were harvested as individual biological replicates (n = 8 per species) using a process similar to our previously reported harvest and lysis method [38,39], which we describe here. Cultures were removed from the 37˚C incubator and centrifuged at 15,180×g, 4˚C for 5 min to isolate bacterial pellets by discarding the conditioned medium supernatants. Pellets were washed 3 times by repeated resuspension in 6 mL 100 mM KH 2 PO 4 -based lysis buffer [40] followed by centrifugation. After the third wash, pellets were dried, weighed, resuspended in lysis buffer to a normalized concentration of 75 mg/mL, and transferred to 15 mL polypropylene (PP) vials. Cells were lysed by sonication while chilled in an ice bath using a Sonic Dismembrator Model 500 with a Branson Sonicator Probe (Thermo Fisher Scientific, Waltham, MA, USA) by the following method: 30% amplitude for 30 sec, 1 min cool-down, 60% amplitude for 30 sec, 2 min cool-down, 60% amplitude for 15 sec. Cell lysates were immediately frozen at −80˚C to ensure their stability and were briefly held frozen (approximately 1 month) until needed for extraction, all samples being stored for an equal period of time. In our experience, lysates of this nature suspended in KH 2 PO 4 -based lysis buffer are stable for metabolomics analyses for several years.

Lipid extraction
Lipids were extracted using a modified version of the Folch method [41] similarly to our previous work [38] using the following process which we describe in detail here. From each normalized cell lysate sample, 150 μL was transferred to a 12 mL glass vial. Extraction blanks were included for downstream data filtering using 150 μL lysis buffer and were treated identically to biological  (d17:1), PAzePC, CER(Glycosyl(β) C12), BMP(14:0 (S,R)), LSM (d17:1), 5 μg/mL each in 2:1 chloroform:methanol. Next, 400 μL methanol was added to each sample followed by vortexing, then 800 μL chloroform was added. Samples were vortexed and incubated on ice for 20 min, with vortexing at 10 and 20 min, followed by addition of 200 μL water. Samples were briefly vortexed and incubated on ice for 10 min with vortexing at 5 and 10 min. Separation of the organic and aqueous layers was achieved by centrifugation at 3260×g, 4˚C for 10 min. The organic (bottom) layer containing lipid content was transferred to a new 12 mL glass vial in two steps: first with removal of 800 μL of the original organic layer, followed by another removal of 400 μL after re-extracting the remaining aqueous layer with 400 μL 2:1 chloroform:methanol by incubating on ice for 10 min and centrifugation at 3260×g, 4˚C for 10 min. Lipid extracts were dried under nitrogen at 30˚C and reconstituted in 300 μL isopropanol. Samples were centrifuged at 3260×g, 4˚C for 10 min to pellet any residual protein, and 250 μL supernatants were transferred to glass LC vials for UHPLC-HRMS analysis.

Analytical instrumentation and methodology
Metabolomics and lipidomics analyses by UHPLC-HRMS were performed on a Thermo Q Exactive Orbitrap Mass Spectrometer with heated electrospray ionization source coupled to a Dionex Ultimate 3000 UHPLC system (Thermo Scientific). Parameters for the metabolomics analysis will be discussed first, followed by the lipidomics analysis. For the metabolomics analysis, reverse phase chromatography with gradient elution was employed using an ACE . Injection volume was 10 μL. Data were acquired in both positive and negative ion mode by full scan (70000 mass resolution), data-dependent MS/MS (35000 mass resolution), and all-ion fragmentation MS/MS (70000 mass resolution) analyses from m/z 200-2200. Using pooled samples for each species, we employed iterative exclusion analysis, consisting of repeated data-dependent MS/ MS analysis with successive exclusion of detected features, to allow detection of lower abundance lipid species in both positive (6 rounds) and negative (4 rounds) ion mode [42].

Data processing
For quality control purposes, the performance of spiked internal standards was assessed in all samples. Excellent reproducibility was verified with standards showing relative standard deviations < 10%. Parameters for the metabolomics data processing will be discussed first, followed by the lipidomics processing. Metabolomics data files were converted from .raw to . mzxml format using RawConverter [43]. MZmine 2 was employed for all processing involved in peak picking, chromatographic alignment, and metabolite identification [44]. Metabolites were identified by m/z (5 ppm) and elution time (±0.2 min) matching to our method-specific internal library produced from pure analytical standards. Non-detected signals were replaced with half the minimum signal intensity value in the dataset [45]. Data were filtered to remove features with �10% signal contributed from the background as determined by comparison to the extraction blanks [46]. Signal intensities were median-normalized [47] and autoscaled [48]. For the lipidomics data, all processing, including file format conversion, peak picking, chromatographic alignment, and lipid identification (MS/MS fragmentation spectral matching to in silico databases) was performed using LipidMatch software [46]. Missing value replacement, as well as data filtration, normalization, and autoscaling, were performed similarly to the metabolomics data.

Statistical analysis
MetaboAnalyst 4.0 was used for statistical analysis and figure generation [49]. All p-values were determined using the two-tailed, unpaired Student's t-test assuming equal variance on the normalized, scaled dataset and adjusted for the false discovery rate using the Bonferroni-Holm correction [50]. In this report, we define significance with a p-value threshold of �0.001.

Determination of oxalate degradation by liquid scintillation counting
Two 100 mL anaerobically sealed vials with 75 mL sterile MRS medium [15], supplemented with 20 mM oxalate and 1% glucose, were each spiked with 3 μCi of 14 C-oxalate (Vitrax, Placentia. CA, USA). Following a thorough mixing of the vial contents, two 100 μL aliquots were removed from each 100 mL vial for liquid scintillation counting prior to inoculating with either L. acidophilus or L. gasseri. These vials were incubated at 37˚C for 3 days, after which time duplicate 100 μL aliquots were removed from each culture and placed in liquid scintillation vials followed by acidification of the aliquot with HCl to induce volatilization of 14 CO 2 in the fume hood for a 24-hr period. Scintillation fluid was then added to the vials followed by liquid scintillation counting using a LS 6500 Multi-Purpose Scintillation Counter (Beckman Coulter, Brea, CA, USA). Final counts were compared to those pre-inoculation to quantify the remaining counts and calculate the percent-degradation of the original oxalate substrate by L. acidophilus and L. gasseri.

Metabolomics analysis
The refined metabolomics dataset consisted of a total of 2077 features detected between positive and negative ion mode. Significant differences between the metabolomes of these bacteria were observed at the global scale as well as the level of individual analytes. S1A  Table with the magnitude and significance of the difference in their relative intensities between species. Table 1 details the top-25 significant identified metabolites. Among the most significant metabolites were several components of the Krebs Cycle, including succinic acid (p = 1.85x10 -13 , 4.3-fold higher intensity in L. acidophilus), fumaric acid (p = 6.25x10 -12 , 623.2-fold higher intensity in L. acidophilus), citric acid (p = 6.70x10 -12 , 139.9-fold higher intensity in L. gasseri), and malic acid (p = 6.74x10 -10 , 3.3-fold higher intensity in L. gasseri). It is worth mentioning that although these bacteria were grown under anaerobic conditions, leading us to expect the Krebs Cycle was not functioning, our results indicate a more complex metabolic process that is different between these two anaerobes. Such differential expression of these key metabolic compounds could have a variety of biological implications. Fumaric acid showed the greatest difference between species by magnitude. Lactobacillus has been noted in the literature due to its potential for industrial mass production of fumaric acid by fermentation [51]. In addition, fumaric acid is known to have an antimicrobial effect, likely due to acidifying the extracellular pH and making the environment inhospitable to competing microorganisms [52][53][54]. High production of fumaric acid for this purpose could be one reason behind the association between intestinal colonization by Lactobacillus and protective host immunity from pathogenic gut bacteria. Additionally, differential expression of fumaric acid between Lactobacillus species as seen in this experiment could result in dissimilar protective capacity between such species. Fumaric acid has also been suggested to serve as a terminal electron acceptor in electron transport chains (ETC) of several species of anaerobic bacteria, including lactic acid bacteria, for enhanced growth in the absence of oxygen [55,56]. The significantly different expression of fumaric acid between L. acidophilus and L. gasseri could suggest a differential reliance on this metabolic pathway, although to our knowledge, the utility of this proposed ETC in these species has not been reported or characterized in the literature. The high relative expression of citric acid by L. gasseri, serving as the second greatest difference by magnitude, is also interesting. Citric acid is a ubiquitous compound able to be produced by many different species of bacteria, some even extensively, under specific growth parameters [57]. Although the production of citric acid by Lactobacillus is not well-documented in the literature, several studies have shown that, under certain conditions, several species of Lactobacillus can use citric acid as an energy source and that the presence of citric acid in culture media has profound effects on their overall growth rate and metabolism [58][59][60][61][62]. The observed differential expression of citric acid between L. acidophilus and L. gasseri deserves further investigation. Gluconic acid (p = 1.17×10 −12 , 56.2-fold higher intensity in L. acidophilus) and glucuronic acid (p = 1.41×10 −10 , 11.9-fold higher intensity in L. acidophilus) also showed a significant difference between species. Gluconic acid and glucuronic acid are oxidation products of glucose, formed from oxidation at C1 and C6, respectively [63]. Both glucose derivatives have been shown to serve defensive functions in Table 1. Top-25 identified metabolites of greatest significant difference in signal intensity between L. acidophilus and L. gasseri.

Metabolite Species Exp Mass Ion Fold-Difference p-value Elevated Expression
Adenosine bacteria. Gluconic acid is produced by Pseudomonas as a key antifungal metabolite [64], which would translate as a potentially important role for an intestinal Lactobacillus species in terms of providing host immunity. Glucuronic acid plays a detoxification role in humans by binding to hormones, drugs, and toxins, forming glucuronides to facilitate their transport and elimination from the body. This process, known as glucuronidation, involves glycosidic bond formation of glucuronic acid from uridine diphosphate-glucuronic acid with these compounds by UDP-glucuronosyltransferases and is an important method by which harmful substances are solubilized and cleared from the body [65,66]. The production of glucuronic acid by Lactobacillus is indicative of a potential detoxification role these bacteria may play in the intestine. Since both gluconic acid and glucuronic acid were found to be elevated in L. acidophilus, it could be assumed that this species may be more effective in delivering their proposed health benefits to the human host as compared to L. gasseri, but this hypothesis requires further investigation to validate. Further work is needed to confirm the biological functionality of these discussed metabolites, as well as all other metabolites identified in this report, and characterize the nature of their differential expression between L. acidophilus and L. gasseri.

Lipidomics analysis
As with the metabolomics analysis, significant differences between the lipidomes of L. acidophilus and L. gasseri were observed. S1B  Table 2. We observed that diacylglycerols (DGs), digalactosyldiacylglycerol (DGDGs), and phosphatidylglycerols (PGs) showed the greatest representation among the significant lipids. Regarding DGs, the nature of their fatty acyl chains was observed to differ significantly between species. . Furthermore, L. acidophilus appeared to exhibit a greater tendency to produce odd-chain lipids as 17 of the 19 odd-chain lipids detected were found to be elevated in L. acidophilus. Although the biological implications behind this observation are not immediately clear, it is known that Gram-positive bacteria can exhibit differential enzymatic biochemistry involved in lipid synthesis which can influencing the balance of even-versus-odd-chain fatty acid production [67]. Examining the general lipidomes of L. acidophilus and L. gasseri, we observed highly-differential class-level distribution of identified lipids as measured by summing the signal intensity of individual lipid classes (Fig 2). Both species displayed the same 5 lipid classes as the primary constituents (>98%) of their lipidomes: DGs, DGDGs, PGs, bis (monoacylglycero)phosphates (BMPs), and cardiolipins (CLs). All other detected lipid classes comprised a minority of the total lipid signal (�2%) and are detailed in S2 Table. Although L. acidophilus and L. gasseri were found to share the same 5 core lipid classes, the ratios of these classes as a measure of the total lipid signal are significantly different between species. In both L. acidophilus and L. gasseri, DGs had the highest class representation at 41% and 59%, respectively. A difference of nearly 50% in the primary lipid class between two bacteria, especially within the same genus, is notable. In L. acidophilus, DGDGs represented 23% of the lipid signal, a sharp contrast to 5% in L. gasseri. PGs and BMPs were also very different between the species with PGs accounting for only 10% of the lipid signal in L. acidophilus and 25% in L. gasseri, and BMPs accounting for 17% of the lipid signal in L. acidophilus and 3% in L. gasseri. The only core lipid class that showed similar representation between species was CLs at 7% in L. acidophilus and 8% in L. gasseri. Gaining an understanding of the unique lipid profiles of these Lactobacillus species is important to understand their relationship with the host intestine.
The composition of the cell structure directly affects how a bacterium responds to its environment, particularly through surface expression or secretion of key signaling and metabolic factors [68]. Certain Lactobacillus species have been shown to modify the composition of their lipid membranes as a protective mechanism against oxidative and thermal stress [69], salt exposure [70], and low pH [71]. Additionally, the nature of a bacterium's surface lipid content allows for modulation of immune responses, specifically activation of the innate immune response, due to interaction between microbe-associated molecular patterns (including lipids) on the microbe surface and pattern recognition receptors on the mucosal surface, triggering production of a variety of effector molecules [72]. Hence, the lipid profiles of candidate probiotic bacteria should be considered a significant point of interest as some species, as a result of their unique lipid makeup, may exhibit more robust colonization and delivery of probiotic health benefits due to the extent at which they can adapt to environmental stressors and coexist within the host intestine. Further work is needed to confirm the biological functionality behind the differential expression observed in the L. acidophilus and L. gasseri lipidomes.

Oxalate degradation by L. acidophilus and L. gasseri
Liquid scintillation counting evaluation of oxalate degradation by L. acidophilus and L. gasseri demonstrated significant degradation by both species. L. acidophilus showed 100% degradation of the 14 C-oxalate with~44% of counts remaining representing 14 C-formate in the media from enzymatic 14 C-oxalate degradation via oxalate decarboxylase [73]. L. gasseri showed~72% of counts remaining, meaning it degraded~50% of the 14 C-oxalate in the media. Our findings are consistent with past experiments reporting oxalate degradation by both L. acidophilus and L. gasseri in the presence of other carbon sources with L. acidophilus being noted as a more efficient degrader [15,16]. These results provide further evidence supporting the evaluation of these Lactobacillus species as potential probiotic remedies for oxalate pathologies.

Conclusions
We conclude that the metabolomes and lipidomes of L. acidophilus and L. gasseri displayed appreciable differentiation both in terms of their general profiles and relative expression of individual compounds. Although we successfully identified 97 metabolites and 71 lipids between these species, we acknowledge that there are many factors yet to be characterized in the Lactobacillus metabolic pool. Among the metabolites we identified, several hold potential to provide immune support and other benefits to the host in a probiotic relationship. We tested and verified the ability of L. acidophilus and L. gasseri to degrade oxalate even with availability of other carbon sources, providing supporting evidence for the need to further evaluate these Lactobacillus species as probiotic treatments for oxalate conditions. Further work is needed to fully define and characterize the L. acidophilus and L. gasseri metabolic profiles and validate their performance as oxalate-targeting probiotics.