Results

Detection of Microbiome Glycine Lipids in Systemic Blood, Portal Blood, Intestinal Lymph, and Liver

Previous work from our group identified microbiome GLs as normal constituents of host feces, with detectable levels in serum and tissues of mice and humans.11,14,30 The structures of the dominant lipid species of the GL classes are shown in Figure 1A. To investigate their potential absorption pathways, we quantified these lipids in portal vein plasma, systemic serum, and small intestinal lymph of C57BL/6J mice. Microbial GLs were found in all compartments, with the highest concentrations in intestinal lymph. (Figure 1B–D). The microbiome GLs, particularly L1256 class species, were also detected in human hepatic tissue (Figure 1E), representing the first demonstration that these microbiome GLs are present in human liver.

Figure 1: Detection and quantification of microbiome glycine lipids. Structures of microbiome glycine lipids found in Bacteroidota depicting the dominant lipid species within each lipid class (A). Note C15:0 (pink) and 3-OH iso C17:0 fatty acids (blue) are present in many of these lipids. Microbiome glycine lipids were quantified by UPLC-MS/MS as follows: total bacterial glycine lipids (B), L654 (C), and L1256 (D) were summed for each lipid class and analyzed across three mouse sample types: systemic circulation serum, portal vein plasma, and intestinal lymph. Values are mean ± standard error of the mean for systemic serum (n = 5). Portal plasma (n = 5) and intestinal lymph (n = 2) glycine lipids were quantified after pooling samples from 3 individual animals. Microbiome glycine lipids in human liver tissues (E) were quantified as ng/g total mass. Values are mean ± standard error of the mean (n = 4). Statistical significance determined by one-way ANOVA with Fisher LSD (∗P < .05, ∗∗∗∗P < .0001). ANOVA, analysis of variance; LSD, Least Squares Difference; UPLC-MS/MS, ultra-performance liquid chromatography-tandem mass spectrometry.

Fermentable Fiber Is a Major Dietary Factor Regulating Gut Microbiome Glycine Lipids

We previously reported decreases in fecal, liver, and serum microbiome GLs (including L654) in low-density lipoprotein receptor (Ldlr)−/− mice when they were switched from a low-fat chow diet to a Western-type HFD (45% kcal fat).14 In addition to fat differences, chow (Teklad 2918) contains soluble fibers and ∼15% (w/w) insoluble fiber, while the HFD contains only 5% (w/w) insoluble cellulose.

To test whether fermentable fiber rescues microbiome GLs in feces and liver, male Ldlr−/− mice (C57BL/6J background) were fed a low-fat chow diet (Chow), a HFD with 5% w/w cellulose, or the same HFD but lacking cellulose and supplemented with fermentable fiber (HFD + fiber) for 8 weeks (Figure 2A). Fiber supplementation (for a total of ∼7.5% w/w dietary fiber and ∼5% w/w soluble fiber) consisted of a mixture of citrus peel pectin (mainly soluble fiber) and pea fiber (insoluble and soluble fiber) (Supplementary Table 1). These fiber types promoted Bacteroides growth previously.31,32 Although chow-fed animals consumed more food by weight, terminal body weight, percent body weight change from baseline, and liver weight were generally higher in the HFD-fed groups than in chow-fed controls (CTLs; Supplementary Table 2). HFD reduced fecal microbiome GLs by ∼87% vs chow, including reductions in L654 (−74%), L1256 (−76%), L430 (−100%), and L342 (−98%) species (Figure 2B). The supplementation of HFD with fermentable fiber increased fecal Bacteroidota relative abundance (Supplementary Figure 1). Furthermore, fermentable fiber markedly increased the fecal microbiome GLs compared to HFD alone, with ∼7-fold higher total microbiome lipids, ∼3.6-fold more L654, ∼2.9-fold more L1256, ∼79-fold more L342, and detectable levels of L430 species. Despite these fecal recoveries, hepatic microbiome lipids and hepatic triglycerides (TG) remained unchanged with fiber (Figure 2C and D). However, hepatic microbiome GLs were inversely associated with liver triglyceride content (Figure 2E). Thus, these results show these lipids are diet-modifiable and that their accumulation in the liver is relevant to hepatic health.

Figure 2: Fermentable fiber is a major dietary factor regulating gut microbiome glycine lipids. Male Ldlr−/− mice (C57BL/6J) were fed low-fat chow, Western HFD (5% cellulose), or HFD + fiber (pectin/pea fiber replacing cellulose; ∼7.5% total, ∼5% soluble fiber) for 8 weeks (A). Lipids were extracted and total microbiome glycine lipids (total GLs), L654, L1256, Lipid 1242 (L1242), L567, L430, and L342 in cecal feces (B) and liver (C) lipid extracts were quantified by UPLC-MS/MS (n = 6–9, mean ± standard error of the mean). Hepatic triglycerides (TG) (D) and associations between hepatic microbiome glycine lipids and TG via Spearman correlation (E). Statistical significance determined by Kruskal-Wallis nonparametric test (∗P < .05, ∗∗P < .01, ∗∗∗P < .001). Illustration from NIAID NIH BIOART Source (bioart.niaid.nih.gov/bioart/281). UPLC-MS/MS, ultra-performance liquid chromatography-tandem mass spectrometry.

Microbiome Glycine Lipids Attenuate Histopathological Features of Liver Disease Progression in a Mouse Model of Metabolic Dysfunction–Associated Steatohepatitis

Given the presence of L654 and L1256 in portal plasma and liver, we next evaluated their bioactivity in a diet-induced MASH model. Male C57BL/6J mice were fed a high-fat (35% w/w), high-sucrose (35% w/w), and high-cholesterol (2% w/w) diet for 22 weeks to induce MASH. During the final 8 weeks, mice received intraperitoneal injections every 48 h of either vehicle CTL or 1 μg of bacterial lipid extracts enriched in L654 species or L1256 species (Figure 3A). We used a two-step semipreparative HPLC purification method to prepare highly-enriched fractions of L654 (Supplementary Figure 2) and L1256 (Supplementary Figure 3) classes of species and will therefore refer to them as L654 and L1256 treatments, respectively.

Figure 3: Microbiome glycine lipids attenuate histopathological features of liver disease progression in a mouse model of MASH. The effects of chronic administration of L654, L1256, or vehicle control (CTL) were studied in C57BL/6J mice (A). Liver sections were hematoxylin and eosin stained (B) for visualizing hepatic steatosis (x200). Blinded histopathological scoring of liver sections for vacuolation (C), lobular inflammation (D), portal inflammation (E), fibrosis (F), and hepatocyte ballooning (G) (n = 13–15, mean ± standard error of the mean). Statistical significance determined by Kruskal-Wallis nonparametric test (∗P < .05, ∗∗P < .01, ∗∗∗P < .001, ∗∗∗∗P < .0001). Illustration from NIAID NIH BIOART Source (bioart.niaid.nih.gov/bioart/281).

After 8 weeks, treatments minimally impacted food intake, body/tissue weights, calorimetry, and wheel-running, though free body water decreased in L654- and L1256-treated mice (Supplementary Table 3, Supplementary Figure 4). Serum analyses revealed L654-mediated reductions in alanine aminotransferase (−23%) and soluble TLR2 (−35%) compared to CTLs (Supplementary Table 4).

Blinded histopathological assessment for MASH was conducted on hematoxylin and eosin-stained livers. Consumption of the high-fat/high-sucrose/high-cholesterol (HF/HS/HC) diet for 22 weeks resulted in substantial MASH development in the CTL group (Figure 3B). Except for lobular inflammation, the histologic scores for lipid vacuolation (steatosis), portal inflammation, fibrosis, and hepatocyte ballooning were each significantly lower in both microbiome GL-treated groups than in CTLs (Figure 3C–G). Although microbiome GLs improved multiple histologic features, they did not significantly affect the fibrotic area quantified in Picrosirius red-stained liver sections (Supplementary Figure 5). Hepatic lipids were reduced by L654 and L1256 treatments, consistent with the histopathological findings (Table 1). Despite these benefits, the levels of microbiome GLs in liver (Table 1) and feces (Supplementary Table 5) did not differ between groups.

Variable | Control | L654 | L1256
Triglycerides (mg/g) | 269.6 ± 40.2 | 192.2 ± 15.1a | 193.2 ± 13.5a
Cholesterol (mg/g) | 21.8 ± 4.25 | 14.6 ± 1.99 | 12.3 ± 0.92a
Cholesteryl ester (mg/g) | 19.0 ± 2.99 | 12.2 ± 1.38a | 11.9 ± 0.46a
Free cholesterol (mg/g) | 10.4 ± 2.59 | 7.36 ± 1.64 | 5.13 ± 0.77b
Microbiome glycine lipids (ng/g) | 1290.8 ± 235.6 | 1341.8 ± 279.4 | 1257.8 ± 203.3
Lipid 1256 (ng/g) | 781.6 ± 196.8 | 736.6 ± 229.0 | 669.5 ± 181.5

Microbiome Glycine Lipids Induce Hepatic Messenger RNA Expression of Genes Involved in Fatty Acid Oxidation and Promote Hepatocyte Mitochondrial Function

TLR2 signaling has been previously linked to hepatic peroxisome proliferator-activated receptor-γ coactivator 1-α (PGC-1α) induction,33,34 so we next assessed hepatic messenger RNA (mRNA) expression of Ppargc1a (PGC-1α) and other genes involved in fatty acid oxidation (Figure 4). Ppargc1a (+45–95%) showed increased mRNA expression in L654- and L1256-treated mice (Figure 4A). PGC-1α is a known coactivator of peroxisome proliferator-activated receptor alpha (PPARα).35 While Ppara (PPARα) mRNA remained unchanged, microbiome GLs regulated several genes governing fatty acid oxidation (Figure 4C). Both treatments upregulated peroxisomal thioesterases Acot3 (+99–142%) and Acot4 (+28–37%), involved in fatty acid oxidation regulation.36 L1256 downregulated B cell lymphoma 6 mRNA (−46%), which acts as a suppressor of PPARα-regulated target genes, including Acot3/4 expression.37 Additionally, L654 increased hepatic mitochondrial DNA (mtDNA) copy number, as measured by the mitochondrial-to-nuclear DNA ratio (Figure 4B).

Figure 4: Microbiome glycine lipids induce hepatic mRNA expression of genes involved in fatty acid oxidation and hepatic mitochondrial DNA. Quantification of mRNA expression of PGC-1α (Ppargc1a) (A) and fatty acid oxidation genes (C) is shown (n = 12–15, mean ± standard error of the mean). Total liver DNA was isolated and mitochondrial DNA copy number (B) quantified by mitochondrial DNA (mtDNA) to nuclear DNA (nDNA) ratios (n = 13–15, mean ± standard error of the mean). Ratios were Loge transformed due to positively skewed distributions. Statistical significance determined by one-way ANOVA with Fisher LSD (∗P < .05, ∗∗P < .01, ∗∗∗∗P < .0001). ANOVA, analysis of variance; LSD, Least Squares Difference.

To evaluate further the effects of microbiome GLs on the bioenergetics and mitochondrial function of liver cells, we conducted experiments where we treated the human HepG2 cells with either L654 (0.654 μg/mL, 6.54 μg/mL) or L1256 (1.256 μg/mL, 12.56 μg/mL), corresponding to approximately equimolar concentrations (∼1 μM and ∼10 μM) of the dominant lipid species (ie, L654 and L1256, respectively), or vehicle CTL every 24 h for 3 consecutive days (Figure 5A). mtDNA was increased in this model by L1256 but not by L654 (Figure 5B). After 3 days of treatment, cells were subjected to a mitochondrial stress test to evaluate respiratory function. Results from 3 independent experiments were normalized to vehicle CTL and analyzed (Figure 5C–K; representative experiment shown in Supplementary Figure 6). Compared to vehicle CTL, the highest concentration of L1256 treatment significantly increased HepG2 cell basal respiration, glycolysis, adenosine triphosphate production, maximal respiration, and proton leak by approximately 2-fold. No statistically significant changes were observed in oxygen consumption rate/ extracellular acidification rate ratio, spare respiratory capacity, or nonmitochondrial respiration. At the lowest concentration, the L1256 treatment had more modest effects. Thus, L1256 appears to act as metabolic enhancers in hepatocytes, significantly increasing mitochondrial content and function while also enhancing glycolytic capacity.

Figure 5: Microbiome-derived glycine lipids enhance mitochondrial bioenergetics in HepG2 cells. Cells were treated with L654, L1256, or vehicle control for 72 h before analysis of mitochondrial DNA and mitochondrial function (A). Mitochondrial DNA copy number (B) quantified by mitochondrial DNA (mtDNA) to nuclear DNA (nDNA) ratios. Results are shown for oxygen consumption rate (OCR) during the Mito Stress test over time (C), basal respiration (D), basal glycolysis (E), OCR/extracellular acidification rate (ECAR) ratio (F), adenosine triphosphate production (G), maximal respiration (H), spare respiratory capacity (I), proton leak (J), and nonmitochondrial respiration (K). Data are normalized to control within each experiment (indicated by dotted line) and presented as mean ± standard error of the mean from three independent experiments. Statistical significance determined by one-way ANOVA with Fisher LSD (∗P < .05, ∗∗P < .01, ∗∗∗P < .001). Illustrations from NIAID NIH BIOART Source (bioart.niaid.nih.gov/bioart/4 and bioart.niaid.nih.gov/bioart/202). ANOVA, analysis of variance; LSD, Least Squares Difference.

Microbiome Glycine Lipids Affect Hepatic mRNA Expression of Genes Involved in Inflammation

To assess the impact of microbiome GLs on inflammation and fibrosis, we next analyzed the mRNA expression of a broad panel of genes related to macrophage activity, immune signaling, and fibrogenesis in mouse liver (Figure 6).

Figure 6: Microbiome glycine lipids affect hepatic mRNA expression of genes involved in inflammation. Gene expression related to inflammation and oxidative stress (A), antigen presentation (B), and fibrosis (C) (n = 9–15, mean ± standard error of the mean). Statistical significance determined by one-way ANOVA with Fisher LSD (∗P < .05, ∗∗P < .01). ANOVA, analysis of variance; LSD, Least Squares Difference.

The hepatic mRNA expression of triggering receptor expressed on myeloid cells 2 (Trem2), which encodes for TREM2, was upregulated by L1256 (+77%) (Figure 6A). The expression of Ccn1 mRNA, which is involved in activating TLR2/4 signaling,38 decreased by L1256 (−31%). The mRNA expression of the negative acute phase protein, paraoxonase 1, was also increased by both microbiome GLs by 37%–46%. Major histocompatibility complex class II genes H2-Aa and H2-DMb1 were significantly decreased by lipid treatments (Figure 6B). However, no significant changes were observed in other inflammation- or fibrosis-associated genes (Figure 6A–C).

To investigate whether these molecular changes translated to cellular immune adaptations, male C57BL/6J mice underwent four 6-week regimens: (1) chow; (2) HFD (high-fat/high-cholesterol/high-sucrose); (3) HFD + vehicle; or (4) HFD + L654 (1 μg/dose, 3x weekly). Flow cytometry of intrahepatic lymphocytes revealed HFD increased hepatic effector memory cluster of differentiation 4+ (CD4+) (CD44+) and CD8+ (CD44+CD69+) T cells but reduced naïve subsets. While L654 did not reverse hepatic T cell shifts, it reduced splenic effector memory CD4+ T cells (CD44+CD69-) and restored naïve CD4+ frequencies (Supplementary Figures 7-9), indicating splenocyte modulation was distinct from its hepatic gene effects.