Sphingomyelin d18:1/14:0 activates TLR4 in metaflammation

doi:10.21203/rs.3.rs-2792338/v1

Download PDF

Biological Sciences - Article

Sphingomyelin d18:1/14:0 activates TLR4 in metaflammation

https://doi.org/10.21203/rs.3.rs-2792338/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Chronic low-grade inflammation, also called metaflammation, is associated with prevalent non-communicable diseases, such as atherosclerosis. Anti-inflammatory therapy provides a clinical benefit in patients, but the triggers that incite metaflammation remain largely unknown. To uncover non-genetic inflammatory factors influencing atherosclerosis severity, we performed an unbiased in vivo screen measuring thousands of host- and microbe-derived molecules in a mouse model of atherosclerosis. Machine learning-supported analyses identified, next to known pro- and anti-atherogenic factors, the medium fatty acid-chain sphingomyelin d18:1/14:0 (S14) as highly positively associated with atherogenesis. S14 activated macrophage innate immune signalling, immune-metabolic reprogramming and pro-inflammatory gene transcription. The inflammatory activity of S14 was TLR4- and MD-2-dependent, consistent with biochemical and molecular dynamics simulation analyses showing that S14 promoted the formation of active TLR4/MD-2 dimers. Human interventional and observational trials demonstrated that dietary changes altered concentrations of S14 and that increased circulating S14 was associated with carotid plaque development in obese individuals. Collectively, these findings identify an inducible endogenous ligand for TLR4 that drives metaflammation, providing the rationale for preventative approaches and pharmacological interventions that may curb detrimental inflammation secondary to Western-type lifestyle habits.

Biological sciences/Immunology/Inflammation/Chronic inflammation

Biological sciences/Immunology/Innate immunity/Pattern recognition receptors/Toll-like receptors

Biological sciences/Immunology/Immunological disorders/Inflammatory diseases/Atherosclerosis

Biological sciences/Immunology/Innate immune cells/Monocytes and macrophages

The environment humans are exposed to in the developed Western world has drastically changed in the last decades. However, while infectious triggers of the immune system - outside the current SARS-CoV-2 pandemic - have diminished¹, chronic, low-grade inflammation is increasingly present in contemporary Western societies. This subclinical inflammatory process (also called ‘metaflammation’) is thought to be causally linked to the development of many non-communicable diseases such as atherosclerosis, obesity, type II diabetes and even cancer or neurodegeneration in the ageing modern society^2–5. Epidemiological and experimental studies suggest that a Western lifestyle, characterized by increased intake of energy-dense food combined with decreased energy expenditure due to a sedentary lifestyle, along with other factors such as pollution or ageing, contribute to metaflammation^6–9.

In this study, we exploited the variability of phenotypes in mouse models of inflammation as a discovery approach to identify non-genetic environmental factors that drive metaflammation. We analysed thousands of parameters in a mouse model of diet-induced atherosclerosis unbiasedly and integrated the data using machine learning algorithms. Hereby, we identified the endogenous lipid N-myristoyl-D-erythro-sphingosylphosphorylcholine d18:1/14:0 (S14) to be associated with metaflammation. S14 was subsequently determined to be a novel activator of the TLR4 (Toll-like receptor 4)/Myeloid differentiation factor 2 (MD-2) receptor complex. Human intervention studies corroborated the inducible nature of S14 upon increased energy intake and suggested a causal link between S14 abundance and the progression of metabolic inflammation.

Sphingomyelins are associated with atherosclerosis

To determine the non-genetic factors contributing to the phenotypic diversity in atherogenesis, we performed an unbiased screen in 159 genetically identical Ldlr-KO mice fed a high-fat, high-cholesterol Western-type diet (WD). After eight weeks on WD, the portal and systemic vein blood, as well as tissue samples, were collected. The atherosclerotic plaque lesion area varied from undetectable to around 40 % in the aortic route (Fig. 1a). We integrated the results from systemic serologic, portal vein metabolomic, and general inflammation parameters from all mice using a machine learning analysis, taking into account a total of 25,710 features (Fig. 1b). The overall model reached a high degree of accuracy in predicting plaque size (Fig. 1c). We then evaluated the contribution of individual measured features that correlated with plaque size. Using this approach, we identified factors that potentially either contribute to plaque progression or are protective in this disease model. Notably, this analysis confirmed known risk factors for atherosclerosis, such as interleukin-1 beta (IL-1β), which showed a positive correlation with plaque size (Fig. 1d). Conversely, serum high-density lipoprotein (HDL) was inversely correlated with plaque size, in line with its protective role in atherosclerosis (Fig. 1e). Moreover, using portal vein metabolomics we identified several sphingomyelins that positively correlated with plaque burden (Fig. 1f, left), confirming previous literature that has linked sphingolipids to human cardiovascular disease¹⁰. The sphingolipid S14 was amongst the top-most important metabolome-derived factors, positively correlated with plaque burden in our study (Fig. 1f, right), suggesting that S14 is pro-atherogenic.

S14 activates mouse macrophages

Since we identified several sphingomyelins positively associated with plaque development in mice, we tested whether these lipids have pro-inflammatory activity. We stimulated mouse bone marrow-derived macrophages (BMDMs) with sphingomyelins of different fatty acid chain lengths. Notably, S14 and, to a lesser extent, sphingomyelin d18:1/16:0 (S16), but not S18, S20 or S26, induced IL-6 and TNF production (Fig. 2a). Given this pro-inflammatory capacity of S14, we performed dose-response analyses and selected a non-toxic concentration (Extended Data Fig. 1a, 1b). Subsequently, we assessed S14-mediated gene induction via transcriptomics. Like LPS, S14 induced the expression of many pro-inflammatory genes, of which many overlapped with those induced by LPS (Fig. 2b, Extended Data Fig. 2a, 2b). Structurally, sphingomyelins consist of a zwitterionic phosphocholine head group, sphingosine, and a fatty acid chain of variable length (Fig. 2c). The acyl chains of the LPS Lipid A portion, which bind to TLR4/MD-2, primarily consist of 14 or 12 carbons (Fig. 2d). Based on the structural similarity between Lipid A and S14, and their similar gene induction in macrophages, we therefore speculated that S14 could also engage the TLR4/MD-2 complex. While direct TLR4 activation induces the rapid engagement of early signalling events, indirect activators lead to delayed activation that can take up to several hours¹¹. Similar to LPS, S14 induced early signalling events, including phosphorylation of mitogen-activated protein kinase (MAPK) family member c-Jun N-terminal kinases (JNK) and nuclear factor-κB (NF-κB) (Fig. 2e). To translate signalling into function, macrophages adapt their cellular metabolism upon TLR activation. Hence, we stimulated BMDMs with S14 or LPS+IFNg and observed a drastic decrease in oxidative phosphorylation (OXPHOS) after 18 but not after 3 hours of stimulation (Fig. 2f), together with an increase in glycolysis (Extended Data Fig. 2c, 2d). Together, these data demonstrate that S14 mediates the activation of macrophages comparable to LPS.

S14 activates the TLR4/MD-2 complex

We next determined whether TLR4/MD-2 was responsible for S14-mediated macrophage activation. To exclude endotoxin contamination in the S14 preparation, we performed a sensitive Limulus assay (Extended Data Fig. 3a) and detected no endotoxin contamination. Pharmacological inhibition of TLR4 using the small-molecule inhibitor TAK-242 revealed that intact TLR4 signalling was indeed required for cytokine release from BMDMs (Fig. 3a). We further confirmed this TLR4-dependence using a genetic approach, demonstrating that S14 induced TNF and IL-6 release from wildtype (WT), but not TLR4-deficient (Tlr4-KO) BMDMs (Fig. 3b). Similar to LPS, other features of cell activation by S14, including increased inducible nitric oxide synthase (iNOS)-mediated nitric oxide production (Extended Data Fig. 3b) or upregulation of co-stimulatory surface molecules (Extended Data Fig. 3c), were also entirely TLR4-dependent. Finally, we ruled out that the observed pro-inflammatory effects of S14 were mediated by its catabolites by demonstrating that neither blocking sphingomyelin hydrolysis and subsequent ceramide synthesis via sphingomyelinase inhibition (Extended Data Fig. 4a) nor stimulating BMDMs with ceramide (Extended Data Fig. 4b) affected TNF release.

TLR4/MD-2 activation by different types of LPS can have species-specific effects¹². We next analyzed cytokine release in human whole blood in response to S14 stimulation, to validate our findings in human cells. We detected significant amounts of MCP-1, IL-8, and IL-6 after 8 and 24 h of stimulation (Fig. 3c). Of note, the addition of a human TLR4-neutralizing antibody confirmed that S14-induced cytokine release was entirely TLR4-dependent (Fig. 3d). We next asked whether S14 can activate the NLRP3 inflammasome, which has been assigned an important role in the context of metaflammation^13,14. Hence, we tested whether S14 can engage NLRP3 in human whole blood and found that IL-1β was released following both S14- and LPS-treatment. We confirmed the engagement of the NLRP3 inflammasome using the NLRP3-selective inhibitor MCC950 (CRID3) and the caspase-1 inhibitor VX-765 (Fig. 3e). Together, these results show that S14 can stimulate TLR4 in human cells and is capable of fueling inflammation via activation of the NLRP3 inflammasome.

S14 directly interacts with TLR4/MD-2

We next sought to understand whether S14 can directly engage TLR4. Direct lipid agonists of TLR4 bind MD-2, which interacts with the extracellular domain of TLR4, forming a heterodimer of TLR4 and MD-2 (TLR4/MD-2). Upon ligand binding, the TLR4/MD-2 complex forms an active dimer of the heterodimer (TLR4/MD-2)₂. We used a series of molecular dynamics simulations to examine a possible interaction between S14 and TLR4/MD-2. We performed four 1 µs simulations, modelled using the X-ray structure of hexa-acylated LPS-bound-(TLR4/MD-2)₂¹⁵, but with three diacylated molecules of either S14, S16, or their ceramide equivalents in complex with MD-2 (Fig. 3f and Extended Data Fig. 5a). Calculation of the root-mean-square deviation (RMSD) of the TLR4 ectodomains revealed that the S14-bound system exhibited the least structural drift compared to the X-ray structure (Extended Data Fig. 5b). MD-2 was also considerably more stable for sphingolipids compared to ceramides, suggesting stronger binding (Extended Data Fig. 5c). Indeed, sphingomyelins formed more hydrogen-bonds with the surrounding protein (Extended Data Fig. 5d) and salt bridge interactions were evident between the phosphate/choline groups and ionizable residues in TLR4 (Fig. 3g and Extended Data 5e). We also analysed the stability of the MD-2 loop containing F126, which is essential in maintaining the active (TLR4/MD-2)₂ complex¹¹. While the loop was more dynamic compared to the LPS-bound state, sphingomyelins were again more stabilizing than ceramides (Extended Data Fig. 5f). Interestingly, a comparison of the final simulation snapshots revealed that F126 in both MD-2 chains resembled an ‘active-like’ conformation. In the presence of S14, but not S16, F126 pointed into the MD-2 cavity, such that the (TLR4/MD-2)₂ dimerization interface is not disrupted (Fig. 3h and Extended Data Fig. 5g). Consistently, the amount of solvent accessible surface area (SASA) buried between lipid and the surrounding protein in each complex was greatest for S14 (Extended Data Fig. 5h). Collectively, these data indicate that the zwitterionic headgroups of the sphingomyelins may better anchor them to the (TLR4/MD-2)₂ complex compared to ceramides to facilitate favourable binding, with S14 potentially more ‘activating’ than S16.

To complement the simulation data, we used purified mouse TLR4/MD-2 and analysed the formation of TLR4/MD-2-S14 complexes via native polyacrylamide-gel electrophoresis (PAGE) mobility shift assay. Confirming previous reports, the purified extracellular domain of the TLR4/MD-2 complex migrated as a monomer and the addition of LPS induced dimerization of TLR4/MD-2 (Fig. 3i, lanes 1+2)^15,16. The addition of S14 to TLR4/MD-2 increased the presence of dimer bands in a dose-dependent manner, indicating that S14 bound purified TLR4/MD-2 and induced its dimerization (Fig. 3i, lanes 3-6). Conversely, S16 resulted only in a slight upward shift of the monomer band, but no dimer bands were observed (Fig. 3i, lanes 7-10). These results confirm that S14 can directly interact with the TLR4/MD-2 complex leading to its dimerization, whereas S16 binds the receptor but fails to induce effective dimerization.

To assess ligand-dependent dimerization of the TLR4 receptor complex and its subsequent endocytosis in cells, we used two conformation-specific antibodies that recognize either monomeric TLR4/MD-2 (MTS510) or both monomeric and dimeric TLR4/MD-2 (Sa15-21)^17–19. Dimerization of TLR4 can be tracked by loss of MTS510 surface staining, whereby the remaining cell surface expression after agonist addition represents un-activated TLR4/MD-2 heterodimers. In line with the cell-free binding experiments, BMDMs treated with either S14 or LPS exhibited rapid TLR4/MD-2 dimerization and ligand-induced endocytosis, albeit to a different extent and with distinct kinetics (Fig. 3j; Extended Data Fig. 6a). As TLR4 may also be activated in a MD-2-independent manner ²⁰, we next assessed whether MD-2 was required for the agonistic activity of S14. Using HEK293T cells expressing human TLR4 alone or in combination with MD-2, we found that the presence of MD-2 is necessary for S14-induced cytokine release (Extended Data Fig. 6b). The lipid A analogue Eritoran (also called E5564) antagonizes LPS activity by a competitive mechanism. Eritoran binds to a specific hydrophobic pocket in MD-2, thereby antagonizing LPS binding. This prevents (TLR4/MD-2)₂ complex formation and, thus, TLR4 signalling²¹. To address whether S14 engages the same hydrophobic pocket in MD-2, we treated BMDMs with S14 or LPS in the presence of Eritoran. Indeed, the TLR4/MD-2 inhibitor Eritoran completely abolished cytokine release following LPS or S14 stimulation (Fig. 3k), supporting that S14, like LPS, directly engages TLR4/MD-2.

Overnutrition affects blood S14 levels

Our initial discovery approach identified an inducible endogenous TLR4 agonist in mice, and our in vitro data demonstrated that S14 can activate both murine and human TLR4/MD2. To assess whether a hypercaloric diet in humans could also affect blood sphingomyelin levels, we analysed serum samples from the DESIRE study (ClinicalTrials.gov ID: NCT02628301) in which healthy male participants with normal body mass index (BMI) were exposed to a 60 % surplus hypercaloric diet (HCD) for 30 days, followed by a 30-day low caloric diet (LCD)²². We found an increase in S14 concentration after the HCD which decreased again once the participants were switched to low caloric intake (Fig. 4a, left panel). Given the diet-induced changes in body weight and fat percentage observed by Emanuel et al., we further evaluated serum adipokine levels. The concentration of the pro-inflammatory hormone resistin, which in humans is primarily released from circulating myeloid cells²³, coincided with the measured serum S14 concentrations (Fig. 4a, right panel). As the regulatory mechanism of resistin is not fully understood²⁴, we sought to assess whether S14 modulates resistin release. Treatment of human whole blood with S14 for 24 h prompted a significant increase in resistin concentration, indicating de novo synthesis, presumably by blood monocytes (Fig. 4b). Taken together, we reason that S14 not only induces pro-inflammatory cytokines but also promotes resistin release, which may further fuel inflammation^25,26.

Weight loss reduces S14 levels in T2D

Obesity, concomitant with insulin resistance, plays a pivotal role in the rising prevalence of type 2 diabetes (T2D)²⁷. To evaluate whether weight loss and restoration of insulin sensitivity in T2D patients affect circulating S14 levels, we performed lipidomics on plasma samples from the FAIR cohort (ClinicalTrials.gov ID: NCT05295160). The cohort includes overweight and obese (BMI >28 kg/m²) patients with T2D, who were exposed to a very-low caloric diet (VLCD) until a reduction in body weight of at least 15 kg was achieved. In line with what we observed for the healthy volunteers of the DESIRE cohort, the overall plasma sphingomyelin pool was reduced after VLCD (Extended Data Fig. 7a). Specifically, the concentration of S14 was reduced (Fig. 4c), whereas the more common sphingomyelin species such as S16 or S18 were not affected (Extended Data Fig. 7b). Overall, these results suggest that even slight concentration shifts in S14 may result in receptor activation and chronic low-grade inflammation.

Plasma S14 is increased in CAD patients

Finally, we assessed whether S14 levels also coincide with carotid plaques in humans as we observed in our unbiased murine atherosclerosis study. Plasma lipidomics analysis from participants of the 300 obesity (300-OB) cohort²⁸ revealed that circulating S14 concentrations were significantly higher in individuals with carotid plaques compared to those with no carotid plaques (Fig. 4d).

Sphingomyelin is the most abundant sphingolipid found in lipoproteins, and approximately 30 % of the sphingomyelin is associated with HDL²⁹. Therefore, we further measured the amount of S14 in HDL extracted from the serum of healthy human volunteers and patients with coronary artery disease (CAD). In line with the data obtained from the 300-OB patients, our analysis revealed that CAD patients exhibit a higher circulating S14 concentration, and that the amount of S14 carried on HDL particles is significantly higher (Fig. 4e). Together, these interventional and observational studies demonstrate that S14 levels in the blood appear to be regulated by diet and are associated with cardiovascular disease.

We used an unbiased approach to identify new factors that link the consumption of a Western-type diet to the development of murine atherosclerosis. This approach confirmed known protective or detrimental factors, such as HDL³⁰ and IL-1ß³¹, respectively. Among the other associated factors, we identified an endogenous trigger of inflammation which is sensed by the TLR4-MD2 complex and is linked to atherogenesis. TLR4 has long been known as a mediator of metaflammation in models of diet-induced inflammation and T2D³², but the diet-induced trigger has been a matter of speculation.

Our work identifies S14 as a plausible link between diet and TLR4-mediated inflammation. Whereas S14 is only sparsely abundant under homeostatic conditions, its adipose tissue concentration has previously been shown to be increased during obesity³³. We could demonstrate that circulating S14 concentrations are modifiable by lifestyle intervention by showing that a high-calorie diet induces levels that would activate TLR4 and low-calorie diet in obese individuals would reduce S14 below the threshold of TLR4 activation. Strikingly, we noticed that the S14 concentration in individuals at study start (t0) compared to the minimum concentration required to induce cytokine release from macrophages in vitro. It is therefore tempting to speculate that TLR4 has evolved not only as a highly sensitive sensor for bacterial-derived lipopolysaccharides but also as a sensor of overabundant energy leading to increased production of medium fatty acid-chain sphingomyelins. Of note, the enzyme CerS6, which is responsible for the synthesis of C14 and C16 ceramide and thus the direct precursors of the respective sphingomyelins, is highly expressed in context of obesity, and CerS6 deletion protects from diet-induced obesity³⁴. Given the putative role of S14 as a biomarker of metaflammation with intrinsic inflammatory function, it will be interesting to test whether S14 also plays a role in other diseases in which chronic inflammation and metabolic disease represent a risk factor, such as neurodegenerative diseases. Of note, plasma sphingomyelins have been associated with Alzheimer’s disease and could therefore be sensed by TLR4 in microglia cells affecting brain inflammation³⁵.

Data Availability

The raw data files for the RNA sequencing reported in this paper have been deposited at NCBI GEO: GSE221910. Source data are provided with this paper. All other data are available from the corresponding author upon reasonable request.

Code Availability

Any detail on code not provided in the methods section is available from the corresponding author upon reasonable request.

Acknowledgements

We thank all volunteers in the DESIRE cohort, FAIR cohort, 300-Obesity cohort and “Sphingomyelins in CVD” cohort for their participation and the project staff for providing biospecimens for analysis in this study. Eritoran tetrasodium (E5564) was a kind gift from Eisai Co. and we would hereby like to express our gratitude. We acknowledge the Microscopy and Cytometry Core Facility at the Amsterdam UMC (VUmc) for providing assistance. We further thank Luise Gorki for setting up R Scripts for data analysis. We also want to thank Samuel Wright for comments. This work was funded in part by the European Union’s Horizon 2020 research and innovation program under grant agreement No. 848146 (To_Aition) (to E.L.), by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany’s Excellence Strategy – EXC2151 – 390873048 (to E.L.), by the DFG SFB1454 - 432325352 (to E.L., P.D and C.T.), SFB1402 – 414786233 (to E.L.), TRR237 – 369799452 (to E.L.), GRK2168 – 272482170 (to E.L.), KU 2374/3-1 (to M.H.Y via Lars Kuerschner). This work was further supported by the Helmholtz-Gemeinschaft, Zukunftsthema ‘Immunology and Inflammation’ (ZT-0027). P.J.B. and J.K.M. gratefully acknowledge BII core funds. J.V.d.B. is funded by a consortia grant from the European Research Area Network on Cardiovascular Diseases (ERA-CVD 2019T108), the Netherlands Heart Foundation senior fellowship (2017T048) and an Enw-Klein-1 grant from NOW (OCENW.KLEIN.268). N.P.R., J.H.W.R., and M.G.N. were supported by a grant of the Dutch Cardiovascular Alliance/Dutch Heart Foundation (CVON2018-27). NPR was supported by a grant of the ERA-CVD Joint Transnational Call 2018, which is supported by the Dutch Heart Foundation in the Hague (JTC2018, project MEMORY; 2018T093). MGN is further supported by an ERC Advanced Grant (FP/2007-2013/ERC grant 2012-322698), and a Spinoza Prize (NWO SPI 92-266).

Author Contributions

F.S.G., P.D., A.C., U.O. and E.G. designed and conducted experiments and performed analysis. K.d.G. and M.R. conducted experiments. J.Z., K.P., A.H., M.H.Y. and P.W. conducted and analysed lipidomic experiments. J.K.M. and P.J.B. planned, conducted and analysed molecular simulation experiments. S.K.S and M.L.F. performed serum lipoprotein measurements of the in vivo model. L.D. and C.A.T. performed bioinformatics analysis of the in vivo model. A.P. and B.L. contributed to sample collection and data generation for the CAD patient cohort. J.H.W.R, M.G.N and N.P.R as well as A.F.H.P contributed to study management, sample collection and data generation for the 300-OB or FAIR cohort, respectively. C.T., M.G.N., N.P.R. gave additional feedback for experimental design and analysis of experiments. E.L., J.V.d.B and P.D. jointly supervised and guided the study. F.S.G., J.V.d.B and E.L. reviewed data and wrote the manuscript.

Competing interest declaration

E.L. is co-founder of IFM Therapeutics, Odyssey Therapeutics, DiosCure Therapeutics and a Stealth Biotech. P.D. is co-founder and employee of a Stealth Biotech. The other authors declare no competing interests.

Additional information

Extended Data is available for this paper.

Supplementary Information is available for this paper.

Correspondence and requests for materials should be addressed to Eicke Latz ([email protected], [email protected]).

Reprints and permission information is available at www.nature.com/reprints.

Finch, C. E. & Crimmins, E. M. Inflammatory exposure and historical changes in human life-spans. Science 305, 1736–1739 (2004).
Danaei, G. et al. The global cardiovascular risk transition: associations of four metabolic risk factors with national income, urbanization, and Western diet in 1980 and 2008. Circulation 127, 1493–1502 (2013).
Hotamisligil, G. S. Inflammation, metaflammation and immunometabolic disorders. Nature 542, 177–185 (2017).
Kerr, J., Anderson, C. & Lippman, S. M. Physical activity, sedentary behaviour, diet, and cancer: an update and emerging new evidence. Lancet Oncol 18, e457–e471 (2017).
Khilazheva, E. D. et al. The Role of Metaflammation in the Development of Senescence-Associated Secretory Phenotype and Cognitive Dysfunction in Aging Mice. J Evol Biochem Physiol 58, 1523–1539 (2022).
Christ, A., Lauterbach, M. & Latz, E. Western Diet and the Immune System: An Inflammatory Connection. Immunity 51, 794–811 (2019).
Hosseini, Z., Whiting, S. J. & Vatanparast, H. Current evidence on the association of the metabolic syndrome and dietary patterns in a global perspective. Nutr Res Rev 29, 152–162 (2016).
Egger, G. & Dixon, J. Non-nutrient causes of low-grade, systemic inflammation: support for a ‘canary in the mineshaft’ view of obesity in chronic disease. Obes Rev 12, 339–345 (2011).
López-Otín, C., Galluzzi, L., Freije, J. M. P., Madeo, F. & Kroemer, G. Metabolic Control of Longevity. Cell 166, 802–821 (2016).
Edsfeldt, A. et al. Sphingolipids Contribute to Human Atherosclerotic Plaque Inflammation. Arterioscler Thromb Vasc Biol 36, 1132–1140 (2016).
Lancaster, G. I. et al. Evidence that TLR4 Is Not a Receptor for Saturated Fatty Acids but Mediates Lipid-Induced Inflammation by Reprogramming Macrophage Metabolism. Cell Metab 27, 1096-1110.e5 (2018).
Ernst, O. et al. Species-Specific Endotoxin Stimulus Determines Toll-Like Receptor 4- and Caspase 11-Mediated Pathway Activation Characteristics. mSystems 6, (2021).
Christ, A. et al. Western Diet Triggers NLRP3-Dependent Innate Immune Reprogramming. Cell 172, 162-175.e14 (2018).
Duewell, P. et al. NLRP3 inflammasomes are required for atherogenesis and activated by cholesterol crystals. Nature 464, 1357–1361 (2010).
Park, B. S. et al. The structural basis of lipopolysaccharide recognition by the TLR4-MD-2 complex. Nature 458, 1191–1195 (2009).
Ohto, U., Fukase, K., Miyake, K. & Shimizu, T. Structural basis of species-specific endotoxin sensing by innate immune receptor TLR4/MD-2. Proc Natl Acad Sci U S A 109, 7421–7426 (2012).
Tan, Y., Zanoni, I., Cullen, T. W., Goodman, A. L. & Kagan, J. C. Mechanisms of Toll-like Receptor 4 Endocytosis Reveal a Common Immune-Evasion Strategy Used by Pathogenic and Commensal Bacteria. Immunity 43, 909–922 (2015).
Zanoni, I., Tan, Y., Di Gioia, M., Springstead, J. R. & Kagan, J. C. By Capturing Inflammatory Lipids Released from Dying Cells, the Receptor CD14 Induces Inflammasome-Dependent Phagocyte Hyperactivation. Immunity 47, 697-709.e3 (2017).
Zanoni, I. et al. CD14 controls the LPS-induced endocytosis of Toll-like receptor 4. Cell 147, 868–880 (2011).
Raghavan, B., Martin, S. F., Esser, P. R., Goebeler, M. & Schmidt, M. Metal allergens nickel and cobalt facilitate TLR4 homodimerization independently of MD2. EMBO Rep 13, 1109–1115 (2012).
Mullarkey, M. et al. Inhibition of endotoxin response by e5564, a novel Toll-like receptor 4-directed endotoxin antagonist. J Pharmacol Exp Ther 304, 1093–1102 (2003).
Emanuel, A. L. et al. Effects of a Hypercaloric and Hypocaloric Diet on Insulin-Induced Microvascular Recruitment, Glucose Uptake, and Lipolysis in Healthy Lean Men. Arterioscler Thromb Vasc Biol 40, 1695–1704 (2020).
Kaser, S. et al. Resistin messenger-RNA expression is increased by proinflammatory cytokines in vitro. Biochem Biophys Res Commun 309, 286–290 (2003).
Li, Y. et al. Resistin, a Novel Host Defense Peptide of Innate Immunity. Front Immunol 12, (2021).
Steppan, C. M. et al. The hormone resistin links obesity to diabetes. Nature 409, 307–312 (2001).
Bokarewa, M., Nagaev, I., Dahlberg, L., Smith, U. & Tarkowski, A. Resistin, an adipokine with potent proinflammatory properties. J Immunol 174, 5789–5795 (2005).
DeFronzo, R. A. et al. Type 2 diabetes mellitus. Nat Rev Dis Primers 1, (2015).
ter Horst, R. et al. Sex-Specific Regulation of Inflammation and Metabolic Syndrome in Obesity. Arterioscler Thromb Vasc Biol 40, 1787–1800 (2020).
Hammad, S. M. et al. Blood sphingolipidomics in healthy humans: impact of sample collection methodology. J Lipid Res 51, 3074–3087 (2010).
Gordon, D. J. et al. High-density lipoprotein cholesterol and cardiovascular disease. Four prospective American studies. Circulation 79, 8–15 (1989).
Ridker, P. M. et al. Antiinflammatory Therapy with Canakinumab for Atherosclerotic Disease. New England Journal of Medicine 377, 1119–1131 (2017).
Könner, A. C. & Brüning, J. C. Toll-like receptors: Linking inflammation to metabolism. Trends in Endocrinology and Metabolism 22, 16–23 (2011).
Turner, N. et al. Distinct patterns of tissue-specific lipid accumulation during the induction of insulin resistance in mice by high-fat feeding. Diabetologia 56, 1638–1648 (2013).
Turpin, S. M. et al. Obesity-induced CerS6-dependent C16:0 ceramide production promotes weight gain and glucose intolerance. Cell Metab 20, 678–686 (2014).
Mielke, M. M. et al. Plasma sphingomyelins are associated with cognitive progression in alzheimer’s disease. Journal of Alzheimer’s Disease 27, 259–269 (2011).
Dobin, A. et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics 29, 15–21 (2013).
van der Spoel, D. et al. GROMACS: fast, flexible, and free. J Comput Chem 26, 1701–1718 (2005).
Huang, J. et al. CHARMM36m: an improved force field for folded and intrinsically disordered proteins. Nat Methods 14, 71–73 (2017).
Jorgensen, W. L., Chandrasekhar, J., Madura, J. D., Impey, R. W. & Klein, M. L. Comparison of simple potential functions for simulating liquid water. J Chem Phys 79, 926–935 (1983).
Jo, S., Kim, T., Iyer, V. G. & Im, W. CHARMM-GUI: a web-based graphical user interface for CHARMM. J Comput Chem 29, 1859–1865 (2008).
Parrinello, M. & Rahman, A. Polymorphic transitions in single crystals: A new molecular dynamics method. J Appl Phys 52, 7182–7190 (1981).
Essmann, U. et al. A smooth particle mesh Ewald method. J Chem Phys 103, 8577–8593 (1995).
Humphrey, W., Dalke, A. & Schulten, K. VMD: Visual molecular dynamics. J Mol Graph 14, 33–38 (1996).
Bindea, G., Galon, J. & Mlecnik, B. CluePedia Cytoscape plugin: pathway insights using integrated experimental and in silico data. Bioinformatics 29, 661–663 (2013).
Bindea, G. et al. ClueGO: a Cytoscape plug-in to decipher functionally grouped gene ontology and pathway annotation networks. Bioinformatics 25, 1091–1093 (2009).

Experimental Model and Subject Details

Animal experiments were approved by the Institutional Animal Care and Use Committee (IACUC; protocol ID 1945-14 and 1945-17) of the University of Massachusetts Medical School, Worcester, MA. LDLR-KO mouse colony was bred internally, maintained on a C57BL/6 background and frequently genotyped (Transnetyx Inc., Cordova, TN, USA). With an average age of 10 weeks, mice were fed a high fat and high cholesterol western diet ad libitum (Teklad TD.88137; 42% calories from fat, 0.2% cholesterol) for 8 weeks. Mice were kept on a regular day-night rhythm, which was maintained throughout the entire study duration. Further, animal experiments were approved by the Committee for Animal Welfare of the VU University Amsterdam and the University of Bonn. Male and female wildtype (WT) C57BL/6J mice were 8-16 weeks of age, and purchased from Charles River and housed under specific pathogen-free conditions. TLR4-KO mice were initially supplied by Prof. C. Kurts (University of Bonn) and subsequently bred in house at the University of Bonn.

Study Cohorts

All cohort data generated in this study originates from serum and plasma samples obtained by previously registered clinical studies from third parties, whereby two of the cohorts have already been described earlier^22,28. We were provided with available biospecimens that were previously obtained within the respective clinical study scope, independent of the present study. All study participants had formerly given their written consent to the respective clinical study coordinators for use of the material.

The DESIRE cohort (ClinicalTrials.gov ID: NCT02628301) consists of 15 healthy male volunteers (age 18-30 years) from the Netherlands with a BMI of 20-25 kg/m². In short, participants were exposed to a hypercaloric diet (HCD) consisting of a normal ad libitum diet with 60 % surplus kilocalories. The HCD intake was calculated as resting energy expenditure (REE)x1.3x1.6. Once a 5 % gain in body weight was achieved, participants were switched to a low-caloric diet (LCD) consisting of 1.0xREE until returning to their baseline body weight.

The 300-Obesity (300-OB) cohort, established by Radboud University Medical Center, Nijmegen (NL) between 2014-2016, consists in total of 302 overweight and obese individuals aged 55-81 years with a BMI of >27 kg/m². The study was approved by the IRB CMO Regio Arnhem-Nijmegen (nr. 46846.091.13) and further details can be found at the Human Functional Genomics Project (HFGP) website. For the present study, only samples of male participants were considered.

The FAIR cohort (ClinicalTrials.gov ID: NCT05295160), for which recruitment started in 2020, consists of individuals aged 28-73 with type 2 diabetes and a BMI of >27 kg/m². In brief, study subjects received a very low-calorie diet (VLCD; 600-800 kcal/day) over a period of three months or until successful loss of 15 kg body weight. For the present study, only samples of male participants were considered. Due to the study still being ongoing, some restrictions apply to the availability of the FAIR cohort data, which were used under license for the current study, and so are not publicly available.

The study “Sphingomyelins in CVD” as part of SYSTEMI cohort (ClinicalTrials.gov ID: NCT03539133) has been established by the Cardiovascular Research Institute Düsseldorf (CARID) University Duesseldorf, Medical Faculty, Duesseldorf (Germany) between 2019-2022 and consists of both male and female individuals with a mean age of 63.5±10.7 with coronary artery disease (CAD) and a mean BMI of 28.7±4.6 kg/m². The healthy control cohort consists of individuals with a mean age of 30.7±4.4 and a mean BMI of 24.8±2.8 kg/m².

Lipid Preparation

Sphingomyelin stocks (Avanti®) were initially prepared in 95 % Methanol. For cell culture experiments, the organic solvent was evaporated under vacuum before dissolving the dried lipid film in the appropriate cell culture medium containing fatty-acid free BSA (Sigma) to a final molar ratio of 1:2.75 (Sphingomyelin:BSA), filtering and incubating the mix for 45 min at 40 °C whilst shaking and with occasional vortexing.

Bone marrow isolation and BMDM culture

Bone marrow was isolated from femurs and tibias of mice of the indicated genotype. Bone marrow-derived macrophages (BMDMs) were generated by culturing bone marrow cells in DMEM (Life Technologies) supplemented with 10 % FCS (Gibco), 1 % Penicillin/Streptomycin (Gibco) and 15 % L929 cell-conditioned medium (LCM). On day 6 of differentiation, BMDMs were collected and plated for experiments. Experiments were performed in DMEM supplemented with 10 % FCS and 1 % Penicillin/Streptomycin, unless stated otherwise.

Human monocyte isolation and macrophage differentiation

Buffy coats from healthy donors were purchased from Sanquin Blood Bank (Amsterdam, Netherlands). PBMCs were isolated by density gradient centrifugation using Ficoll-Paque PLUS (GE Healthcare). Monocytes were obtained from PBMCs by CD14 positive selection using CD14 microbeads (Miltenyi) and were subsequently plated at 2x10⁶ cells/mL in complete RPMI (RPMI 1640 supplemented with 10 % FCS, 1 % Penicillin/Streptomycin,1 % GlutaMax and 1 % Na-pyruvate [all Gibco]). Differentiation was induced by addition of 50 U/mL recombinant human M-CSF (rhM-CSF; Immunotools). After 3 days of incubation, cells were harvested and plated for experiments in complete RPMI with 25 U/mL rhM-CSF. On day 4, experiments were performed in complete RPMI.

HEK cell culture

Human Embryonic Kidney (HEK) cells stably expressing either human TLR4 or both human TLR4 and MD2 protein were kindly provided by Prof. Sandra van Vliet (VUmc, Amsterdam, Netherlands). Cells were continuously cultured in DMEM supplemented with 10 % FCS and 1 % Penicillin/Streptomycin. For experiments, 40.000 cells were seeded per well into a Poly-L-Lysine coated 96-well plate and left to attach for 6 h before stimulation.

Human Whole Blood Stimulation

Whole Blood from healthy donors collected in ethylene-diamine-tetraacetate (EDTA) tubes was purchased from Sanquin Blood Bank (Amsterdam, Netherlands). Shortly after collection the blood was processed by diluting 100 µL of whole blood 1:5 with the indicated stimuli in RPMI 1640 without further supplements. Diluted blood was then incubated for either 8 h or 24 h in a thermal shaker at 37 °C. For cytokine assessment the diluted whole blood was centrifuged at 400 x g for 5 min and the supernatant was immediately stored at -70 °C.

Gene Expression Analysis by 3’ RNA Sequencing

For RNA Sequencing, 1x10⁶ BMDMs were stimulated for 3 h or 18 h. Cells were subsequently washed with cold PBS and collected in RLT Buffer (Qiagen). RNA was extracted using the RNeasy Mini Kit together with the RNase-free DNase Set for DNA digestion (both Qiagen) according to manufacturer instructions. RNA quality assessment, library preparation and sequencing were performed by Lexogen GmbH, Vienna, Austria. In short, libraries were prepared using the QuantSeq 3’ mRNA-Seq FVD Library Prep Kit according to manufacturer’s instructions. Briefly, 150 ng total RNA were reverse transcribed with oligo(dT) primer containing an Illumina-compatible 5’ end sequence (first strand) and random priming (second strand). Finally, individual sample barcodes were introduced via 15 cycles of PCR. All libraries were then analysed for adapter dimers, size distribution and concentration on a Fragment Analyzer (Agilent). Sequencing was performed on an Illumina NextSeq 2000 sequencer at Lexogen GmbH.

Sequencing Quality Control, Alignment, Read Quantification and Analysis

Quality control of the FastQ files was performed using cutadapt version 1.18 (https://cutadapt.readthedocs.io/en/stable/) for adapter trimming and reads of the samples were analysed using FastQC version v0.11.7 prior to and after adapter trimming (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/). The read alignment to the spike-in complemented Ensembl release 101 of the Mus musculus high coverage assembly GRCm38 was performed with splice-aware aligner STAR version 2.6.1a³⁶. The featureCounts software program version 1.6.4 was used for quantification of gene expression using the annotations of Ensembl GRCm38.101 and spike-in specific annotations of Lexogen as reference (http://subread.sourceforge.net). A differential gene expression analysis was conducted using DESeq2 version v1.18.1 using the counts of unique alignments.

Immunoblotting by WES^TMand JESS^TM

Samples and 12-230 kDa Separation Modules (ProteinSimple) were prepared according to manufacturer’s instructions. Shortly, four units of equally concentrated samples were mixed with one unit of fluorescent master mix and incubated for 10 min at 95°C. Samples were loaded twice per plate and analysed for SAPK/JNK (Cell Signaling #9252; 1:20) or phospho-SAPK/JNK (Cell Signaling #4668; 1:20) on the Wes. Expression levels of phosphorylated NF-κB p65 (Cell Signaling #3033; 1:20) and total NF-κB p65 (Cell Signaling #4764; 1:20) per sample were sequentially analyzed in the same capillaries using the Jess Replex module (Protein Simple) on the Jess. Proteins were detected using the anti-rabbit chemiluminescence detection module (ProteinSimple). Protein separation and analysis was performed using the compass simple western software (Biotechne) with the following program: 25 min separation at 375 V, 5 min washing, 90 min (Wes) or 60 min (Jess) primary antibodies, 30 min secondary antibody. Per sample, the ratio of phosphorylated to total protein expression was calculated using the AUC of the respective chemiluminescent signals.

Measurement of Cytokine Secretion

Concentrations of mouse and human cytokines were determined using commercial ELISA kits according to manufacturer instructions: mouse IL-6 and TNFα (both ThermoFisher) mouse IL-1β (R&D), mouse IP-10 (R&D), mouse IFNβ (R&D), mouse IL-10, human IL6 and TNFα (both ThermoFisher), human IL-1β and human IL-8 (both R&D).

Multiplex cytokine and adipokine analysis were performed using the following LEGENDplex^TM kits according to manufacturer instructions: Human Inflammation Panel 1; Human Metabolic Panel 1 (both V-bottom plate format). Analysis was performed using a X20 Fortessa flow cytometer (BD Biosciences) and obtained data was evaluated using the LEGENDplex^TMData Analysis Software Suite (Biolegend).

Measurement of Nitric Oxide secretion

Concentrations of nitrite in cell culture supernatants was determined using a standard Griess Assay. Briefly, equal amounts of cell culture supernatant or a sodium nitrite standard were incubated 1:1 with Griess Reagent and absorbance values measured at 540 nm.

Measurement of Oxygen Consumption Rate (OCR) and Extracellular Acidification Rate

OCR and ECAR were determined using a XFe-96 Extracellular Flux Analyzer (Agilent Technologies). 8x10⁴ BMDMs were plated a day prior to the assay in a XF-96 Cell Culture Microplate and stimulated either overnight or on the assay day. 1 h before the assay, cell culture media was replaced by Seahorse XF Base Medium (Agilent) supplemented with 2 mM glutamine and cells were incubated in a non-CO₂ incubator. Measurements were performed under basal conditions and upon sequential addition of glucose (final 10 mM), oligomycin A (final 1 µM), fluoro-carbonyl cyanide phenylhydrazone (FCCP, final 2 µM) with sodium pyruvate (final 1 mM) and finally rotenone and antimycin A (both 0.5 µM). Calculations performed: Glycolysis = ECAR_Glucose – ECAR_basal.

Flow Cytometry

BMDMs were harvested, washed in cold staining buffer (PBS + 0.5 % BSA + 0.02 % sodium azide) and stained for surface markers for 30 min at 4 C in the dark. Unspecific antibody was blocked using an anti-CD16/CD32 antibody (1:100, clone 93, BD Biosciences) and fixable viability dye was added to discriminate live cells (FVD-e780, 1:1000, eBiosciences). If applicable, cells were subsequently washed, fixed and permeabilized for intracellular staining. For the assessment of inflammatory macrophage markers, cells were stained with the following fluorescently-labeled antibodies, diluted in staining buffer: CD80 (clone 16-10A1, Biolegend), CD86 (clone GL-1, Biolegend) and CD40 (clone 3/23, Biolegend).

TLR4 Dimerization Assay

TLR4 dimerization and endocytosis were assessed and calculated as published by Zanoni et al.¹⁸. Briefly, stimulated BMDMs were stained as described above with antibodies against TLR4 that allow analysis of either dimerization status (TLR4-PE-Cy7, clone MTS510, Biolegend) or receptor endocytosis (TLR4-APC, clone SA15-21, Biolegend).

All samples were acquired on the same day at the O|2 Core Facility at Amsterdam UMC (Netherlands) on a X20 Fortessa flow cytometer (BD Biosciences, Franklin Lakes, NJ, USA), which was calibrated daily using CS&T calibration beads (BD Biosciences). Data was analysed using FlowJo (TreeStar, v10) by gating on live, single cells and using single stains with UltraComp eBeads (ThermoFisher) as compensation controls.

Molecular Dynamics Simulation

All atom simulations were performed using the GROMACS 2018.3 simulation package³⁷ with the CHARMM36m³⁸ force field and TIP3P water model³⁹. The experimental structure (PDB: 3FXI¹⁵) of the (TLR4:MD-2)₂ dimer bound to E. coli ReLPS was used. All lipid structures and topologies were taken from CHARMM-GUI membrane builder⁴⁰. Given that E. coli ReLPS is typically composed of six acyl tails, three sets of double-tailed lipid molecules as studied in experiments were modelled bound to each MD-2 monomer. Four simulation systems were set up, composed of (TLR4:MD-2)₂ complexes bound to a given lipid type, namely: i) CER 18:1/14:0 ii) CER 18:1/16:0; iii) SM d18:1/14:0; and iv) SM d18:1/16:0. Lipids were aligned such that their headgroups overlaid with the crystallographically resolved ReLPS lipid A headgroup. Each dimeric construct was placed in a dodecahedron box with ~16 nm box edge. Approximately 90,000 TIP3P water molecules were added to the box, along with 150 mM NaCl salt whilst neutralizing the overall system charge. Energy minimization was performed using the steepest descent minimization algorithm with a 0.01 nm energy step size. The system was equilibrated in a step wise fashion: i) 1 ns with position restraints on protein alpha carbons and lipid headgroups in the NVT ensemble; ii) 10 ns with position restraints on protein alpha carbons and lipid headgroups in the NPT ensemble; and iii) 20 ns with position restraints on protein alpha carbons in the NPT ensemble. All position-restrained simulations were run with a force constant of 1000 kJ mol^-1 nm^-2. Pairs of replicas of unrestrained production runs were set to 1000 ns in the NPT ensemble for each system. A temperature of 310 K was maintained using the velocity rescaling thermostat with an additional stochastic term using a time constant of 1 ps. Pressure was maintained semi-isotropically at 1 atm using the Parrinello-Rahman barostat⁴¹ and a time constant of 5 ps. All bonds which involved hydrogens were constrained using the LINCS algorithm. Equations of motion were integrated using the leap-frog algorithm with a time step of 2 fs. Long-range electrostatic interactions were described using the particle mesh Ewald method⁴². The short-range electrostatics reals pace cut-off was 1.2 nm and the short-range van der Waals cut-off was also 1.2 nm. Periodic boundaries conditions were applied in all directions. Simulations were performed on an in-house Linux cluster composed of 8 nodes containing 2 GPUs (Nvidia GeForce RTX 2080 Ti) and 24 CPUs (Intel® Xeon® Gold 5118 CPU @ 2.3 GHz) each.

All simulation snapshots were generated using VMD⁴³. The root mean square deviation (RMSD) of the MD-2 backbone was calculated with respect to the experimental structure after alignment onto the TLR4 dimer backbone. The RMSD of the TLR4 dimer backbone was calculated with respect to the experimental structure after alignment onto the same set of atoms. The RMSD of the F126 loop encompassing backbone atoms of residues 120-129 was calculated with respect to the experimental structure after alignment onto each MD-2 monomer backbone. Protein-lipid hydrogen bonds were calculated based on a 0.35 nm donor-acceptor cut-off distance and 30^º hydrogen-donor-acceptor cut-off angle. The buried solvent accessible area (SASA) between MD-2/lipid and TLR4 was calculated by summing the constituent SASAs of each MD-2/lipid complex and TLR4 chain, before subtracting the SASA of the entire TLR4-MD-2-lipid complex. RMSDs, hydrogen bonds, and buried SASA values were calculated as a mean of two parts of the (TLR4:MD-2)₂ dimer and averaged over the final 200 ns of pairs of replicas for each system.

Native PAGE

Sphingomyelin d18:1/14:0 (S14) and d18:1/16:0 (S16) were initially dissolved at 2.5 mg/mL in water before mixing with purified protein as described below.

The de-glycosylated recombinant protein for the extracellular domain of mouse (m) TLR4 (residues 22-627) was prepared as described previously¹⁶. Purified extracellular domains of mTLR4/MD-2 (25 μM) were mixed with either Re-LPS (125 μM), S14 (0.25, 0.5, 1.0, and 1.5 mM), or S16 (0.25, 0.5, 1.0, and 1.5 mM) in 20 mM Tris-HCl pH 8.0, 0.15 M NaCl, and 0.1% Triton X-100 and incubated for 3 h at 37 ºC. After incubation, samples were subjected to 12.5% native PAGE and stained with Coomassie Brilliant Blue.

Serum cytokines and apolipoprotein measurements

Serum cytokines were measured with a ThermoFisher ProcartaPlex custom design Immunoassay on a Luminex Multiplex platform with xMAP bead-based technology, according to the manufacturer’s instructions. Samples were read and analyzed on a Luminex MagPix system.

Isolation of mouse plasma lipoproteins by FPLC

For lipoprotein separation by FPLC, total sera cholesterol levels were first determined with a standard enzymatic assay (Infinity^TM Cholesterol Liquid Stable Reagent, Thermo Fisher Scientific, #TR13421). For FPLC analysis, 30 μL sera aliquots were pre-warmed to 37 °C for 5 min., and centrifuged for 30 seconds at 15,000 x g. Subsequently, 20 μL of sample was taken by avoiding pelleted material and the floating chylomicron layer. 15 μL of the clarified sample was subsequently fractionated by fast performance liquid chromatography (FPLC) gel filtration on a Superose^® 6 PC 3.2 /300 column at 4°C (AKTA Purifier 10 System, GE Healthcare, Uppsala, Sweden). The elution fractions were monitored using absorbance at 280 nm, with a constant flow of 30 μL/min and fractions (80 μL) were collected beginning 18 min after sample injection. Cholesterol in each fraction was measured by the Infinity^TM Cholesterol Liquid Stable Reagent assay, and the area under the curve for VLDL, LDL and HDL was determined in comparison to a standard of known amounts of human VLDL, LDL and HDL run in parallel (HDL, Sigma Aldrich, #L8039; LDL, Sigma Aldrich, #L7914; VLDL, Kalen Biomedical LLC, #770100-7).

Portal Vein Serum Metabolomics and Analysis

Metabolomics analysis was conducted from portal vein sera of mice fed a high for 8 weeks. During necropsy, blood from the portal vein was drawn to receive 100-200 mL of serum volume. Serum was sent to Metabolon (Morrisville, NC, USA) for metabolomics analysis on the mViewTM platform. Samples were analyzed on a Ultrahigh Performance Liquid Chromatography-Tandem Mass Spectroscopy (UPLC-MS/MS) system, using separate reverse phase (RP)/UPLC-MS/MS methods with positive and negative ion mode electrospray ionization (ESI). Raw data was extracted, peak-identified and QC processed using Metabolon’s hardware and software. Metabolon maintains a library based on authenticated standards that contains the retention time/index (RI), mass to charge ratio (m/z), and chromatographic data (including MS/MS spectral data) on all molecules present in the library.

Lipidomics

All solvents were HPLC grade or LC-MS grade purchased from VWR International GmbH (Darmstadt, Germany) and Merck KGaA (Darmstadt, Germany).

To 2 µL serum or plasma sample, 500 mL extraction mix (CHCl₃/MeOH 1/5 containing internal standards: 210 pmol PE(31:1), 396 pmol PC(31:1), 98 pmol PS(31:1), 84 pmol PI(34:0), 56 pmol PA(31:1), 51 pmol PG (28:0), 28 pmol CL(56:0), 39 pmol LPA (17:0), 35 pmol LPC(17:1), 38 pmol LPE (17:0), 32 pmolCer(17:0), 99 pmol SM(17:0), 55 pmolGlcCer(12:0), 14 pmol GM3 (18:0- D3), 359 pmol TG(50:1-d4), 111 pmol CE(17:1), 64 pmol DG(31:1), 103 pmol MG(17:1), 724 pmol Chol(d6), 45 pmol Car(15:0)) was added. Samples were sonicated in the bath sonicator for 10 s and subsequently lipids were collected. After centrifugation at 20,000 x g for 2 min, supernatants were retrieved and mixed with 200 µL CHCl₃and 800 µL 1 % acetic acid. The samples were briefly shaken and centrifuged for 2 min at 20,000 x g. The upper aqueous phase was removed, and the entire lower phase was transferred into a new tube and evaporated in a speed vacuum concentrator (45°C, 10 min). Spray buffer (8/5/1 2-propanol/MeOH/water, 10 mM ammonium acetate) was added and samples were sonicated for 5 min.

Lipidomics analysis

All samples were separately infused at 10 μL/min into a Thermo Q Exactive Plus spectrometer equipped with the HESI II ion source for shotgun lipidomics. MS1 spectra (resolution 280 000) were recorded in 100 m/z windows from 250 to 1200 m/z (positive mode) followed by recording MS/MS spectra (resolution 70 000) by data independent acquisition in 1 m/z windows from 250 to 1200 (positive mode). Raw files were converted to .mzml files and imported into and analysed by LipidXplorer software using custom mfql files to identify sample lipids and internal standards. For further data processing, absolute amounts were calculated using the internal standard intensities and normalized to the amount of extracted serum.

Analysis of sphingolipids on high-density lipoprotein (HDL)

Phospholipids were extracted by dilution of 25 µL Plasma/HDL with 275 µL MeOH, followed by 10 µL internal standards (30 pmol SM d18:1/d17:0 in MeOH, Avanti Polar Lipids Inc., Alabaster, AL). Samples were mixed, precipitated overnight at −80 °C and centrifugated (5 min, 21,300 x g at 4 °C). The supernatant was transferred into mass spectrometry sample vials and stored at −80 °C until measurement.

Chromatographic separation was performed on a LCMS-8050 triple-quadrupole mass spectrometer (Shimadzu Duisburg, Germany) interfaced with a Dual Ion Source and a Nexera X3 Front-End-System (Shimadzu Duisburg, Germany). HPLC is performed with a 2x60 mm MultoHigh-C8 RP column with 3 µm particle size at 40 °C. Mobile phases consisted of [A] 10 mm ammonium formate in LC-MS grade water and [B] 10 mm ammonium formate in MeOH. The following LC gradient was used: 1.5 min at 78% [B], linear increase of [B] from 78% to 93% from 1.5 to 2.0 min, linear increase of [B] from 93% to 99% [B] from 2.00 min to 10 min and equilibration from 10.01 min to 13.00 min prior next injection. Flow rate was 300 µL/min and injection volume of all samples was 5 µL. MS settings were the following: Interface: ESI, Nebulizing Gas flow: 2 L/min, Heating Gas Flow: 10 L/min, Interface Temperature: 300 °C, DL Temperature: 250 °C, Heat Block Temperature: 400 °C, Drying Gas Flow: 10 L/min. Data were collected using multiple reaction monitoring (MRM). Negative ionization was used for qualitative analysis and positive ionization was used for quantification. Standard curves were generated by measuring increased amounts (50 fmol to 25 pmol) of SM (d18:1/14:0) and SM (d18:1/16:0) together with internal standard (SM d18:1/d17:0, 0.3 µm final conc. in MeOH). Linearity of the standard curves and correlation coefficients were obtained by linear regression analysis. All MS analyses were performed using LabSolutions 5.99 SP2 and LabSolutions Insight (Shimadzu Duisburg, Germany).

Random Forest Analysis

Data obtained from mouse specimens were further analysed using the statistical computing environment R with packages: ggplot2, lattice, knitr, caTools, caret, SHAPforxgboost, xgboost, and randomForest (version R 4.0.3). In brief, dataset was split (split == 2/3) and scaled for random forest analysis using function randomForest(), and test data were used to predict the % plaques by using function predict(). Feature importance was evaluated by setting argument: importance = TRUE. Furthermore, data were trained and fit using functions xgb.DMatrix() and xgb.train(). Shap values used for individual feature contributions to model were calculated using function shap.prep(). Importance based on xgboost fitted data was evaluated by using function xgb.importance(). Data were plotted using plot(), shap.plot.summary() and xgb.plot.importance().

Software, Quantification and Statistical Analysis

Data was analysed using Microsoft Excel (Microsoft), R Studio version 1.4.1106 and Graph Pad Prism 9 (Graph Pad Software Inc.), or with the specific software stated under the according method section. All replicates represent biologically independent samples. Unless stated otherwise, two-tailed Student’s t-test, one-/two-way ANOVA with Bonferroni post-test correction for multiple testing (95 % confidence interval) was performed to determine significance of the difference between groups. P-Values of p<0.05 were considered to be statistically significant and are denoted by asterisks (*p<0.05, **p<0.01, ***p<0.001). Chemical structures were designed using ChemDraw® (Perkin Elmer) version 21.0.0.28. Biological Process GO-enrichment analysis (GOEA) was performed using Cytoscape v3.9.1 and ClueGO v2.5.8^44,45. Visualization networks were created by performing right sided hypergeometric test including Benjamini-Hochberg correction. Cluster names represent GO terms with highest significance inside the individual cluster.

Yes there is potential Competing Interest. Co-founder of IFM Therapeutics, Odyssey Therapeutics, DiosCure Therapeutics and a Stealth Biotech (Eicke Latz). Co-founder and employee of a Stealth Biotech. The other authors declare no competing interests (Peter Düwell)

ExtendedFigures.docx

Download PDF

Version 1

posted

You are reading this latest preprint version

Sphingomyelin d18:1/14:0 activates TLR4 in metaflammation

Status:

Version 1

Abstract

Figures

Introduction

Discussion

Declarations

References

Methods

Additional Declarations

Supplementary Files

Status:

Version 1