Inhibition of Δ24-dehydrocholesterol reductase activates pro-resolving lipid mediator biosynthesis and resolution

Andreas Körnera, Enchen Zhoub, Christoph Müllerc, Yassene Mohammedd, Sandra Hercegc, Franz Bracherc, Patrick C. N. Rensenb, Yanan Wangb, Valbona Mirakaja, and Martin Gierad,1

aDepartment of Anesthesiology and Intensive Care Medicine, Molecular Intensive Care Medicine, Eberhard Karls University Tübingen, 72072 Tübingen, Germany; bDepartment of Medicine, Division of Endocrinology, and Einthoven Laboratory for Experimental Vascular Medicine, Leiden University Medical Center, 2333ZA Leiden, The Netherlands; cDepartment of Pharmacy-Center for Drug Research, Ludwig Maximilians University Munich, 81377 Munich, Germany; and dCenter for Proteomics and Metabolomics, Leiden University Medical Center, 2333ZA Leiden, The Netherlands

Edited by Christopher K. Glass, University of California San Diego, La Jolla, CA, and approved September 3, 2019 (received for review July 27, 2019)

Targeting metabolism through bioactive key metabolites is an desmosterol to cholesterol (Fig. 1A). Of the 10 enzymes involved in upcoming future therapeutic strategy. We questioned how modi- distal cholesterol biosynthesis, starting with squalene, DHCR24 has fying intracellular lipid metabolism could be a possible means for recently taken center stage in several diseases. This enzyme has been alleviating inflammation. Using a recently developed chemical probe linked to Alzheimer’s disease (AD), oncogenic and oxidative stress (SH42), we inhibited distal cholesterol biosynthesis through selective (10), hepatitis C virus (HCV) infections (11), differentiation of T 24 inhibition of Δ -dehydrocholesterol reductase (DHCR24). Inhibition helper-17 cells (12), development of foam cells (13), and prostate of DHCR24 led to an antiinflammatory/proresolving phenotype in a cancer (14). While the role of DHCR24 in AD is controversially murine peritonitis model. Subsequently, we investigated several omics discussed, it has been postulated as a possible drug target for HCV layers in order to link our phenotypic observations with key metabolic infections (11) as well as arteriosclerosis (8). Interestingly, this sug- alterations. Lipidomic analysis revealed a significant increase in endog- gests a pronounced involvement of DHCR24 in inflammatory pro- enous polyunsaturated fatty acid (PUFA) biosynthesis. These data in- cesses, as can be further underlined by the actions of its substrate, tegrated with gene expression analysis, revealing increased expression INFLAMMATION desmosterol. Desmosterol has been shown to interact with the liver IMMUNOLOGY AND of the desaturase Fads6 and the key proresolving enzyme Alox-12/15. X receptors (LXRs), the adenosine 5′-triphosphate–binding cassette Protein array analysis, as well as immune cell phenotype and func- transporter A1 (ABCA1), the sterol response element-binding pro- tional analysis, substantiated these results confirming the antiinflam- γ matory/proresolving phenotype. Ultimately, lipid mediator (LM) analysis tein (SREBP), and ROR (8, 12, 15, 16). In addition, it has been revealed the increased production of bioactive lipids, channeling proven that activation of LXRs has a marked influence on immu- the observed metabolic alterations into a key class of metabolites nological functions and PUFA biosynthesis, particularly within Φ known for their capacity to change the inflammatory phenotype. (M ) (15, 17). Importantly, PUFAs are substrates for the synthesis of immunologically relevant lipid mediators (LMs) inflammation resolution | desmosterol | cholesterol | lipid mediator | PUFA involved in onset and offset of inflammation (7).

n recent years, our knowledge of human lipid metabolism has Significance Isignificantly increased, particularly its role in controlling and defining biological phenotypes. The most important example is Lipid metabolism is crucial to many (patho-)physiological pro- possibly the role of immune cell metabolism in cancer (1). Con- cesses, including inflammation. In particular, cholesterol bio- sequently, this has triggered scientists to develop novel technolo- synthesis has emerged as an exciting novel therapeutic target in gies for studying the interactions between metabolome and numerous recent studies. Much is known about the early steps in phenotype (2). Moreover, metabolic reprogramming has become cholesterol biosynthesis; however, the later steps have hitherto accepted as a possible therapeutic strategy (3, 4). In particular, largely been neglected. Here, we investigated the druggability lipid metabolism has long been recognized for its important role in of distal cholesterol biosynthesis using a selective and potent inflammatory processes (5, 6). However, to date, most therapeutic chemical probe (SH42). This inhibition leads to the accumulation strategies target specific enzymes or receptors rather than attempting of the bioactive metabolite desmosterol, boosting the bio- to comprehensively understand and modify lipid metabolism as a synthesis of polyunsaturated fatty acids (PUFA) and the down- key process for the production and homoeostasis of a plethora of stream production of antiinflammatory mediators. Our report lipid-derived mediators. In this context, we aimed at modifying in- integrates distal cholesterol biosynthesis and its intermediate flammation through global lipid metabolism changes. We hypoth- desmosterol with endogenous PUFA biosynthesis and resolution esize that global changes in lipid metabolism can be controlled in of inflammation, rendering this pathway attractive for further such a way that an intrinsically antiinflammatory/proresolving phe- developments in the context of proresolving therapies. notype is observed. We argue that this would require a shift of lipid Author contributions: P.C.N.R., Y.W., V.M., and M.G. designed research; A.K., E.Z., C.M., metabolism toward the increased production of antiinflammatory and M.G. performed research; S.H., F.B., and M.G. contributed new reagents/analytic and proresolving mediators. As n3-polyunsaturated fatty acids tools; A.K., E.Z., C.M., Y.M., P.C.N.R., Y.W., V.M., and M.G. analyzed data; A.K., E.Z., (PUFAs) have particularly been described as important precursors C.M., Y.M., F.B., P.C.N.R., Y.W., V.M., and M.G. wrote the paper; V.M. contributed to study of several antiinflammatory and proresolving mediators (7), we design; and M.G. guided overall study design. therefore questioned how their endogenous production and me- The authors declare no conflict of interest. tabolism could be boosted, especially during an inflammatory event. This article is a PNAS Direct Submission. Mainly through the recent work of Glass and coworkers (8, 9), Published under the PNAS license. 24 we identified Δ -dehydrocholesterol reductase (DHCR24; also 1To whom correspondence may be addressed. Email: [email protected]. called Seladin-1) as a promising target for controlling lipid metab- This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. olism, possibly allowing its reprogramming. DHCR24 is the terminal 1073/pnas.1911992116/-/DCSupplemental. enzyme in cholesterol biosynthesis, catalyzing the transformation of First published September 23, 2019.

www.pnas.org/cgi/doi/10.1073/pnas.1911992116 PNAS | October 8, 2019 | vol. 116 | no. 41 | 20623–20634 Downloaded by guest on September 30, 2021 well as the antiinflammatory/proresolving potential, of controlled desmosterol accumulation through selective inhibition of DHCR24 using a recently published, highly in vivo active chemical probe (SH42) (half-maximal inhibitory concentration = 5nM)(21).Ini- tially, we evaluated the cross-reactivity of SH42 with LXRs and a set of additional transcription factors; next, we investigated the probes’ effects on the course of zymosan A (zyA)-induced peritoneal in- flammation. Subsequently, lipidomic analysis was carried out in order to understand the observed phenotypic changes on a molec- ular level. We quantitatively assessed more than 900 lipid species in blood, liver, and peritoneal lavage samples. We monitored serum and peritoneal lavage chemokine and gene expression profiles and conducted protein array analysis. Using tandem mass spectrometry (MS) analysis, we investigated PUFA levels and the production of specialized proresolving mediators (SPMs) and tissue-regenerative mediators. Ultimately, we investigated the therapeutic use of SH42 in animal experiments and investigated functional and phenotypic immune cell responses. Results The Chemical Probe SH42 Does Not Interact Directly with LXRs. To exclude direct LXR activation caused by the employed chemical probe SH42 (Fig. 1), we investigated the regulation of well-known LXR target genes under treatment with this compound. Since cholesterol biosynthesis in cell culture is only active when extracel- lular cholesterol is low (22), we treated RAW264.7 MΦ with high concentrations of SH42 in the presence and absence of fetal calf serum (FCS) as an exogenous source of cholesterol. As can be seen in Fig. 1, the LXR target genes (23) Abca1 and Abcg1 (Srebp1c only at 5 μM) are only activated when no extracellular cholesterol is available. In turn, SH42 blocks cholesterol biosynthesis (21) and causes LXR activation via desmosterol accumulation, but not in a direct fashion; otherwise, target gene regulation would also take place when extracellular cholesterol is available. In order to further strengthen this line of argumentation, we repeated the incubation of Fig. 1. Distal cholesterol biosynthesis, structure of chemical probe SH42, Φ and selectivity assessment. (A) Structure and action of SH42; the red circle RAW264.7 M with and without FCS under cotreatment with the emphasizes desmosterol, accumulating due to blockage of DHCR24. 7-DHCR; cholesterol-lowering drug atorvastatin, which interferes with proxi- 7-dehydrocholesterol reductase; FF-MAS, follicular fluid meiosis-activating sterol. mal cholesterol biosynthesis. By doing so, the accumulation of ste- (B) LXR target gene activation (qPCR) with (inactive cholesterol biosynthesis) and roidal precursors, such as desmosterol, was prevented. As can be without (active cholesterol biosynthesis) FCS. ctrl, control. (C) LXR target gene seen in Fig. 1, atorvastatin did block SH42-regulated target gene regulation under cotreatment with atorvastatin. For all experiments, results are activation when no extracellular cholesterol (FCS) was present. expressed as mean ± SEM (n = 5). *P < 0.05, **P < 0.01; nonparametric Student’s Again, from these results, we conclude that SH42 causes LXR target t test (Mann–Whitney) comparing each group separately with the control group gene regulation only in conjunctionwithactivatedcholesterolbio- (no treatment). synthesis, hence not directly interacting with LXRs but via a desmosterol-related loop. As an additional line of evidence, we re- peated the aforementioned experiments in combination with the When focusing on LXRs as master regulators of the aforemen- LXR inverse agonist/antagonist GSK2033 (24). As can be seen in SI tioned effects, there is no doubt about the possible value of LXR Appendix,Fig.S1, treatment with GSK2033 inhibits the expression agonists in the context of several diseases (18). Nevertheless, no of LXR target genes (e.g., Abca1, Abcg1) mediated by SH42, without LXR agonist has yet reached the clinic. The main drawback of affecting Srebp1c gene expression. In addition, we monitored pos- unspecific LXR activators can be attributed to SREBP1c activation, sible cross-reactivities with other nuclear receptors, namely, farnesoid resulting in severely increased hepatic lipogenesis (19). This has led X receptor (FXR), pregnane X receptor (PXR), and retinoid X researchers to develop specific LXR-β agonists, as well as tissue- receptor (RXR). We monitored the target gene expression for Shp and cell-specific drug formulations (18, 20). In contrast, desmosterol (25) (FXR), Cyp7a1 (26) (FXR and PXR), Cyp3a11 (27) (PXR), has been described as an LXR agonist and selective interaction and Cyp27a1 (28, 29) (PXR and RXR). As can be seen in SI Ap- partner of SREBP in that it can likely activate a subset of SREBP1 pendix,Figs.S1andS2, SH42 did not induce target gene expression target genes, despite inhibiting SREBP1 binding to response ele- of any of these transcription factors. As previously reported, SH42 ments. In turn, DCHR24 integrates cholesterol biosynthesis, LXR does pose selective actions on DHCR24 within distal cholesterol activation, and inflammation. This raises the question of whether biosynthesis (21). Taken together, these results provide evidence that the endogenous substrate of DHCR24, desmosterol, might be an the actions of SH42 are mediated by a desmosterol-related loop attractive candidate, fulfilling important requirements for successful without a direct interaction with the transcription factors tested here. LXR activation: (1) Desmosterol has been shown to bind to LXRs in the low micromolar range (16), (2) it possesses a selective Inhibition of DCHR24 Leads to Increased Inflammation Resolution and SREBP interaction potential, and (3) it inhibits inflammatory gene Selectively Decreases Proinflammatory Cell Influx. Next, we aimed expression (8). Recently these processes have been demonstrated in at verifying that inhibition of DHCR24 indeed has a profound MΦ using desmosterol and desmosterol mimetics (9). effect on acute inflammation. One of the most frequently used an- Along these lines, we reasoned that inhibition of DHCR24, imal models for such investigations is the murine zyA-induced leading to controlled desmosterol accumulation, is ideally suited peritonitis model (30). Mice were treated once daily (intraperito- as a master regulator of lipid metabolism, allowing initiation of neally [i.p.]) with 0.5 mg of SH42 for 3 d, followed by i.p. injection of an antiinflammatory and proresolving phenotype. We therefore zyA. This was an empirically found dose of SH42 at which we did not investigated the metabolic and immunological consequences, as observe any acute or long-term toxicity (up to 8 wk). Desmosterol

20624 | www.pnas.org/cgi/doi/10.1073/pnas.1911992116 Körner et al. Downloaded by guest on September 30, 2021 levels in whole blood, liver, and peritoneal exudate were monitored turn, the effects described here on cellular influx and population using gas chromatography (GC)-MS analysis (22). Blood desmosterol observed in the murine peritonitis model can likely be attributed levels proved stable at ∼5 μg/mL (13 μM) (Fig. 2A), while desmosterol to selective inhibition of DHCR24. This is followed by increased was basically absent in the control group. Cellular influx into the levels of desmosterol, which modulates inflammatory outcome peritoneal cavity was monitored using fluorescence-activated cell by its known actions on LXRs (13, 16). sorting (FACS) analysis. As can be seen in Fig. 2 B–H, treatment with SH42 did cause a decrease in leukocyte influx (Fig. 2B), with a Lipidomic Analysis of Serum, Liver, and Residential Peritoneal Cells. specific decrease in polymorphonuclear cells (PMNs) (Ly6Ghi)inthe Next, we aimed to obtain a better understanding of the under- early phase of acute inflammation (Fig. 2C). To quantify the kinetics lying molecular factors possibly explaining our findings. In first of leukocytes, we determined the resolution index (Ri) (31) dis- instance, we targeted lipid biosynthesis modulation under treat- playing a 58% reduction in Ri from 19 h to 8 h in mice treated with ment with SH42. We furthermore aimed to exclude classical side SH42. This indicated a strong acceleration of inflammation resolu- reactions observed for LXR agonists, such as hypertriglyceridemia tion (Fig. 2D). When focusing on the resolution phase, in which and increased lipogenesis (19). Initially, we monitored liver des- and MΦ dominate, we found a marked reduction in mosterol levels, averaging 6.2 ng/mg in the treatment group and classical Ly6Chi monocytes (Fig. 2E) and a significant increase in the 1.9 ng/mg in the control group (Fig. 3A). Next, we investigated nonclassical Ly6Clo monocytes (Fig. 2F)andMΦ (Fig. 2G), which desmosterol levels in the resident peritoneal cell fraction, aver- were accompanied by a significant enhancement in MΦ clearance of aging 17.3 versus 1.8 ng/1 × 106 cells in the control group (Fig. 3B). apoptotic PMNs (Fig. 2H). Next, we analyzed pro- and antiin- Using our established GC-MS assay in the scan mode (22), we also flammatory levels. As can be seen in Fig. 2 I–K,proin- screened for other steroidal precursors of cholesterol (33) as well as flammatory were down-regulated, while antiinflammatory/ oxysterols, neither of which could be detected under our experimental proresolving cytokines (Fig. 2 L and M) were up-regulated. This conditions. Additionally, we found increased blood and serum des- observation correlated well with the fact that exudate levels of mosterol levels under treatment with SH42, as described above and interleukin-10 (IL-10) and transforming growth factor-β (TGF-β), elsewhere (21). To gain further insights into molecular changes linked known to be crucial in the resolution of acute inflammation and to SH42 treatment, we carried out comprehensive quantitative tissue regeneration, were strongly enhanced (Fig. 2 L and M). To- lipidomic analysis of serum and liver samples after a 3-d treatment gether, these data indicate that SH42 displays an antiinflammatory, with SH42 (without zyA) (34). The serum lipid composition showed a proresolving, and proregenerative impact during acute inflammation. significant reduction in the cholesteryl ester (CE) fraction, accom- While it has been shown before that activation of LXRs in the panied by a strong trend toward higher free fatty acid (FFA) levels

zyA-induced murine peritonitis model alleviates inflammation (SI Appendix,Fig.S3). This finding is in line with SH42 blocking INFLAMMATION (32), it is important to note that the chemical probe SH42 does cholesterol biosynthesis, leading to lower CE levels. Importantly, no IMMUNOLOGY AND not possess direct LXR activation capacity as detailed above. In significant increase in the serum triacylglyceride (TAG) fraction was

Fig. 2. Peritoneal cell analysis. Red lines and circles represent SH42 treatment, and black lines and squares represent vehicle injection. C57BL/6 mice were exposed to zyA-induced peritonitis after 3 d of treatment with SH42, and peritoneal lavages were collected at 4, 12, 24, and 48 h. (A) Blood desmosterol levels hi after the indicated time points (x axis). (B) Total leukocytes were enumerated by light microscopy. (C) Cell numbers of PMN Ly6G cells. (D) Ri values. Ψmax, maximum PMN numbers; Tmax, the time point of maximum PMN numbers; T50, the time point where PMN are reduced to 50%. (E) Cell numbers of classical + Ly6Chi monocytes. (F) Cell numbers of nonclassical Ly6Clow monocytes. (G) Cell numbers of F4/80 peritoneal MΦ.(H) Cell numbers of -derived MΦ as analyzed by FACS through intracellular staining of PMNs. (I)TNF-α analysis. (J) IL-6 analysis. (K)IL-8analysis.(L and M) IL-10 and TGF-β levels all measured in the peritoneal fluids (4 h after zyA). The results represent at least 2 independent experiments and are expressed as mean ± SEM (n = 9to11per group). All experiments were analyzed by means of an unpaired Student’s t test, 1 per time point. *P < 0.05; **P < 0.01. Further details are provided in the main text.

Körner et al. PNAS | October 8, 2019 | vol. 116 | no. 41 | 20625 Downloaded by guest on September 30, 2021 under treatment with SH42. As can be seen in Fig. 3 C–E,wedid indeed observe significant changes in the FA composition, mainly in residential peritoneal cells. Our main finding was an increased content of PUFA in the FFA lipid fraction, including important precursor FAs for the downstream biosynthesis of SPMs such as, for example, (FA20:5) and (FA22:6). Further details are shown in Fig. 3.

Gene Expression Analysis in Liver and Peritoneal Cells. To sub- stantiate our findings, we next performed gene regulation analysis (qPCR) in liver and peritoneal exudate cells. It has recently been reported that desmosterol and desmosterol mimetics have a cell type-specific effect on LXR target gene regulation. Muse et al. (9) showed that desmosterol selectively activates the LXR target gene ABCA1 (Abca1)inMΦ but not in hepatocytes. Along these lines, we tested the effects of our chemical probe SH42 in vivo in peritoneal lavage cells and liver homogenates. As indicated in SI Appendix,Fig.S5, the chemical inhibition of DHCR24 decreased expression of Abca1 alongside Fasn without affecting expression of Scd in liver homogenates. Gene expression analysis in peritoneal cells focused on Scd-1, FA desaturase 6 (Fads6), Elovl1, Elovl5, Elovl6,andAlox-12/15. We chose Fads6 because it is one of the rate-limiting steps in endogenous PUFA biosynthesis (35). Fur- ther, we monitored Elovl5 as it is particularly involved in PUFA elongation and Elovl6, which is an important gene (enzyme) for the elongation of medium- to long-chain FA (36). Alox-12/15 was monitored as an important gene related to PUFA metabolism and, in particular, the production of SPMs. As can be seen in Fig. 4, when comparing 3-d SH42-treated mice with vehicle control, we observed a trend for increased expression of Scd-1(P = 0.19) as well a significant increase for Fads6 expression at time point 0. This finding, together with our lipidomic data (Fig. 3), pinpointed to an involvement of Fads6 in increased endogenous PUFA pro- duction predominantly in residential peritoneal cells. Our findings are vastly in line with a previous report on the activation of PUFA biosynthesis mediated by the nonselective LXR agonist T0901317 Fig. 3. Analysis of the fatty acid composition represented in the different (15). Next, we aimed at comparing the changes found in resi- lipidomes and desmosterol after a 3-d treatment with SH42. (A)Liver dential peritoneal cells with infiltrating cells during zyA-induced desmosterol levels (n = 3 to 5). Ctrl, control. (B) Desmosterol amounts found in the residential peritoneal cell fraction (n = 4 to 5). (C–E) Volcano plot peritonitis. We carried out qPCR analysis of the same genes 4 h analysis of the FA composition of the different lipidomes; each black dot after injecting zyA into the peritoneum. As depicted in Fig. 4, represents a lipid species linked to a specific FA. Lipids are marked in red (red Fads6 was now down-regulated in the treatment group, while dot). When a fold change >1.5 and P value <0.05 were observed, vehicle Elovl-1 and, in particular, Alox-12/15 became up-regulated at the groups were compared with SH42 treatment. The FFA-free FA lipid class is 4-h time point. Taken together, our data, in combination with an denoted as FA, followed by the carbon number and double-bond count earlier report on the activation of PUFA biosynthesis by T0901317 (e.g., FFAFA204 is free with 20 carbons and 4 double bonds). (15), point to increased PUFA biosynthesis mediated by SH42 Liver lipidomic (C, Upper), residential peritoneal cell lipidomic (D, Middle) and serum lipidomic (E) analyses are shown. The blue dotted lines show the fold change and P value thresholds. The control group was considered as the reference in all comparisons. Results are expressed as mean ± SEM. In A and B,**P < 0.01; nonparametric Student’s t test (Mann–Whitney).

obtained, which is a known side effect of nonselective LXR agonists (19). When evaluating serum lipid concentrations, we found a sig- nificant increase in the FFA concentration with a median of 441 nmol/g in the SH42 group versus 363 nmol/g in the control group (SI Appendix,Fig.S4). Additionally, we investigated the liver lipidome under treatment with SH42. As can be seen in SI Appendix,Fig.S3,a reduced diacylglyceride (DAG) and TAG fraction was observed. This observation was also true for the absolute concentrations of DAG and TAG lipids found per gram of liver tissue, being 956 and 3,718 nmol/g (DAG and TAG) for SH42-treated mice and 1,351 and 6,205 nmol/g for control mice, respectively (complete lipidomic data are provided in Datasets S1–S4). To gain more mechanistic insights and shed light on the effects of SH42 on the course of zyA- induced peritonitis, we carried out fatty acid (FA) compositional Fig. 4. mRNA expression under treatment with SH42. mRNA expression was analysis of serum, liver, and residential peritoneal cells. As LXRs determined in peritoneal lavage cells from 3-d treated mice versus vehicle at have been described to control FA elongation and desaturation time point 0 h of zyA-induced peritonitis and at time point 4 h of zyA- (15), we were particularly interested in the FA composition of these induced peritonitis. Results are expressed as mean ± SEM (n = 10 to 15). lipidomes as we expected to find specific metabolic adaptations *P < 0.05, **P < 0.01; nonparametric Student’s t test (Mann–Whitney).

20626 | www.pnas.org/cgi/doi/10.1073/pnas.1911992116 Körner et al. Downloaded by guest on September 30, 2021 treatment. Mainly, the finding that Alox-12/15 became up-regulated RICTOR signaling, which plays an important role in regulating MΦ in the treatment group 4 h after zyA administration made us in- metabolism to promote M2 polarization (42). Notably, the latter terested in a possible link between our initial observations in the observation is in line with our results observed by lipidomic analysis. zyA-induced self-limiting murine peritonitis model (Fig. 2) and This finding was reflected in increased levels of Alox-12/15 and, increased endogenous PUFA production, as well as the possibly finally, enhanced generation of SPMs, such as MaR1, suggesting altered downstream production of important PUFA metabolites that SH42 induces polarization toward the M2 prohealing and such as eicosanoids and . proresolving phenotype. The results obtained for the entire protein array are presented in Dataset S5. Inhibition of DHCR24 Results in Increased Biosynthesis of Specialized Proresolving Mediators. Based on findings that PUFAs are well- Therapeutic Treatment with SH42. Next, we questioned whether known precursors for the biosynthesis of SPMs and other antiin- SH42 would also exert acute treatment effects in addition to flammatory lipids (7) and on our own lipidomic and qPCR data prophylactic treatment effects. For this, we chose the 24-h time obtained from residential peritoneal cells, we questioned whether point resembling the maximum cellular influx after zyA injection inhibition of DHCR24 and the effects described here are linked by (Fig. 2). Using in vivo mouse experiments, we initially monitored an increased production of proresolving and antiinflammatory blood desmosterol levels after a single (1 d) or repeated (2 d) lipids. To investigate this hypothesis, we used a standardized liquid injections of SH42 in order to investigate whether desmosterol chromatography (LC)-tandem MS (MS/MS) approach (37) and accumulates after a single dose only. This experiment was crucial profiled peritoneal LM concentrations. As can be seen in Fig. 5, to correctly interpret the results of the treatment study. If no we observed higher PUFA levels particularly at the early 4-h time desmosterol would have accumulated after a single SH42 injection, point, while other FAs such as linoleic acid and α-linolenic acid no causal link could be made between desmosterol and the observed remained basically unchanged between the treatment and con- biological phenotype. If so, treatment effects would likely have to be trol groups. In view of several recent reports about the immu- attributed to SH42 itself. As can be seen in Fig. 7, we did indeed nological function of cytochrome P450-derived epoxy metabolites, observe increased blood desmosterol levels after a single dose of we monitored not only SPMs () but also 19,20-epoxy SH42, which further increased after the second dose (without (19,20-EpDPA) and its inactivation prod- treatment, desmosterol is undetectable in blood samples as out- uct 19,20-dihydroxydocosapentaenoic acid (19,20-DiHDPA) (38, lined in Fig. 2A). Subsequently, we monitored changes in LM 39). As the mediators identified here are known to actively control biosynthesis and cellular influx at the 48-h time point. As can be inflammation resolution and cellular influx (39), we postulated seen in Fig. 7 C and D, significantly increased concentrations of

that their increased biosynthesis under treatment with SH42 might several PUFAs and LMs were observed. Additionally, cell numbers INFLAMMATION

pose a link between LXR activation, increased production of and populations were significantly altered in the treatment group IMMUNOLOGY AND antiinflammatory and proresolving LMs, and alleviated inflammatory (Fig. 7B), vastly resembling the results of the prophylactic treat- processes. In this context, we also monitored prostaglandin E2 ment. Taken together, this points to the fact that SH42 can be used (PGE2) and (LTB4) as established mediators of in an antiinflammatory therapeutic setting. inflammation (resolution) (40). As can be seen in Fig. 5L and SI Appendix,Fig.S6,PGE2 was significantly up-regulated in the Proresolving Effects and the Role of LXRs and 12/15-. treatment group, while LTB4 showed a strong trend (P = 0.06) Ultimately, we aimed at sketching a more functional link between toward higher concentrations. For the identification of 1 increased desmosterol levels, the increased production of antiin- (MaR1), please refer to the MS/MS analysis and comparison with flammatory lipids, and inflammation resolution, as well giving some genuine synthetic standard as provided in SI Appendix,Fig.S7.The insight into effects observed in human-derived MΦ.Toprovide Φ relative retention times, relative to the internal standard d4-LTB4, these relations, we focused our efforts on M phenotype, polari- for MaR1 observed in our samples were 0.997 and 0.998 for an zation, , and the roles of LXRs and 12/15-lipoxygenase authentic standard. In order to substantiate our findings, we carried (12/15-LOX) (43), as well as activation of the cytochrome P450 out a validation experiment at the 4- and 12-h time points for (CYP), 15-LOX, and (COX) pathways in human docosahexaenoic acid and its bioactive mediators MaR1 and 19,20- MΦ. Phagocytosis (efferocytosis) is a crucial process during in- EpDPA. These validation experiments confirmed our findings, as flammation resolution. We initially investigated the effects of SH42 can be seen in SI Appendix,Fig.S8. and desmosterol on the phagocytic capacity of human monocyte- derived MΦ. As can be seen in Fig. 8, SH42 was capable of in- Protein Array Analysis of the Phosphatidylinositol 3-Kinase/Protein ducing increased phagocytosis starting from ∼125 μg/mL, while Kinase B and Mammalian Target of Rapamycin Pathways. In order desmosterol showed a positive effect already at doses of no more to expand on these findings and shed more light on the effects of than 0.01 μg/mL This points to the fact that desmosterol is the accumulating desmosterol (not cholesterol) under SH42 treat- actual mediator of the observed effects (Fig. 8 A and B). When ment on the proteomic level, we next carried out protein array coincubating with the LXR antagonist GSK2033, phagocytosis was analysis of residential peritoneal monocytes/MΦ comparing SH42 blocked, and this effect could not be rescued by either SH42 or treatment and vehicle. We focused on the phosphatidylinositol desmosterol (Fig. 8C). In turn, this points to a possible role of 3-kinase (PI3K)/protein kinase B (AKT) and mammalian target of LXRs during this process. To provide additional proof that rapamycin (mTOR) pathways, as they are crucial in restricting desmosterol is the actual mediator of the positive effects of SH42, proinflammatory responses, thereby promoting antiinflammatory/ we quantified desmosterol after a 2-h incubation of human monocyte- proresolving responses and activating monocyte/MΦ differentia- derived MΦ with high concentrations of SH42. As can be seen in tion and polarization toward a proresolving phenotype (41). To Fig. 8D, we could indeed observe a strong trend toward increased understand if and how SH42 affects these pathways, we carried desmosterol concentrations. The other suspected mediators for the out protein array analysis of peritoneal monocytes after a 3-d described effects of SH42 are MaR1 and 19,20-EpDPA; hence, we treatment with SH42 (Fig. 6). In the treatment group, marked subsequently investigated the effects of these lipids on MΦ changes in the expression and phosphorylation of key proteins in phagocytosis (Fig. 8 E and F). This showed that both lipids posi- both the AKT/PI3K and mTOR pathways were observed. As can tively influenced phagocytosis at low concentrations (10 ng/mL). be seen in Fig. 6 A and B, most signaling pathways downstream of While 19,20-EpDPA is a CYP product, MaR1 is biosynthesized by AKT did point to antiinflammatory effects and to changes espe- the actions of 12/15-LOX. To this end, we investigated the effects cially in MΦ survival, cell proliferation, cell growth, and gluco- of SH42 on MΦ isolated from either wild-type or 12/15-knockout neogenesis. In order to substantiate our protein array data, we also (KO) mice, as we had no KO model related to the production of carried out Western blot analysis of mTOR and AKT as well as 19,20-EpDPA available. As can be seen in Fig. 8G, treatment of their phosphorylated variants (Fig. 6C). Besides these aspects, 12/15-LOX KO MΦ did result in diminished phagocytosis. As a SH42 regulated the activation of the mTOR pathway and particularly next step in integrating our results with inflammation resolution,

Körner et al. PNAS | October 8, 2019 | vol. 116 | no. 41 | 20627 Downloaded by guest on September 30, 2021 Fig. 5. Changes of free FA and the downstream metabolites 19,20-EpDPA and 19,20-DiHDPA in peritoneal lavages in the zyA-induced peritonitis model, with and without SH42 treatment. For all experiments, n = 6 to 10; data were obtained from 2 independent experiments with an individual n = 3to5(singleex-

periment for PGE2 with n = 4to5).*P < 0.05, **P < 0.01; nonparametric Student’s t test. (A–H) Longitudinal behavior of various FAs during zyA-induced peritonitis with and without SH42 treatment (nonparametric Student’s t test [Mann–Whitney]). Ctrl, control. (I) Chromatographic trace showing the transition m/z 343→281 was used to detect 19,20-EpDPA. AU, arbitrary units. (J) Enzymatic formation of 19,20-EpDPA and 19,20-DiHDPA catalyzed by CYP450 and soluble epoxide hy-

drolase. (K and L) Levels of MaR1 and PGE2 observed at the 4-h and 12-h time points. Results are expressed as mean ± SEM. *P < 0.05, **P < 0.01; nonparametric Students t test (Mann–Whitney). (M) MS/MS spectra identifying 19,20-EpDPA compared with an authentic standard (1 ng/mL, blue boxed inlay).

we focused on monocyte differentiation and MΦ phenotypes as inflammation resolution (46), we quantified the corresponding both processes are crucial for inflammation resolution. As cell mRNA expression and found a significant up-regulation of several shape and MΦ phenotype are correlated (44), we carried out cell important receptors (mRNA) (Fig. 9C). In order to further sub- shape analysis assessing monocyte differentiation toward the M1 or stantiate our functional phagocytosis data, we also monitored the M2 phenotype (45). As can be seen in Fig. 8 H–P, a 7-d treatment expression of well-described phagocytic markers (47). As we had of human-derived monocytes (peripheral blood mononuclear cells observed a number of MΦ-related functional changes associated [PBMCs]) did give a phenotype closely resembling the with SH42 and desmosterol treatment, we wondered if we could colony-stimulating factor (M-CSF) M2 phenotype. As expected, metabolically integrate these findings with the 15-LOX, CYP, and the effects of desmosterol were significantly lower than observed COX pathways. Indeed, when MΦ were treated with zyA with and for SH42, as desmosterol will be further converted into cholesterol without SH42, dramatic metabolic changes could be observed. if no DCHR24 inhibitor is present (Fig. 8 K–P). In order to gain While zyA-treated MΦ relied mainly on COX-driven PUFA me- more proof that desmosterol and SH42 influence MΦ polarization, tabolism producing high amounts of PGE2, cotreatment with SH42 we assessed messenger RNA (mRNA) expression markers for both not only increased the intracellular amount of free PUFA but also the M1 and M2 phenotypes. As can be seen in Fig. 9 A and B,the caused a marked shift toward the CYP and 15-LOX pathways. The M2 markers Arg-1 and IL-10 were especially increased after latter shows that treatment with SH42 stimulates a metabolic treatment with both SH42 and desmosterol. Next, as several specific shift toward an antiinflammatory phenotype as shown by surface receptors are known to play an important role in mediating increased levels of the 15-LOX pathway markers 15-HETE

20628 | www.pnas.org/cgi/doi/10.1073/pnas.1911992116 Körner et al. Downloaded by guest on September 30, 2021 INFLAMMATION IMMUNOLOGY AND

Fig. 6. Protein array analysis of the mTOR (A) and PI3K/AKT (B) pathways. (Left) Main proteins involved in the mTOR and PI3K/AKT pathways are shown alongside their known physiological functions. Increased protein levels for the SH42 group are shown in red, and decreased proteins are shown in green. (Right) Corresponding protein fold-changes, including phosphoprotein. (C) Western blot analysis.

(15-hydroxyeicosatetraenoic acid) and 17-HDHA (17-hydroxydo- Discussion cosahexaenoic acid), as well as the CYP-derived LM 19,20-EpDPA A summary of the observed effects of SH42 and the mechanism (Fig. 9F). For validation, we repeated the experiment with the of its proresolving effect described here can be found in Fig. 10. LOX inhibitor baicalein and could block the actions of SH42 (SI The fact that DHCR24 has been proposed as a potential drug Appendix, Fig. S10). target for several diseases made us interested in the metabolic and

Körner et al. PNAS | October 8, 2019 | vol. 116 | no. 41 | 20629 Downloaded by guest on September 30, 2021 Fig. 7. Therapeutic application of SH42. (A) Experimental setup of SH42 administration: 24 h after zyA-induced inflammation, 0.5 mg of SH42 was ad- ministered i.v. via the tail vein. Twenty-four hours later at the 48-h time point, mice were killed and peritoneal cells were analyzed by FACS analysis.(B, Left to Right) FACS results: total number of peritoneal lavage cells, PMNs, classical monocytes, nonclassical monocytes, peritoneal MΦ, percentage of phagocyted PMNs, and MΦ efferocytosis. (C and D) Levels of blood desmosterol after 1-d and 2-d treatments, followed by PUFA and LM levels in the peritoneal exudates at the 48-h time point. Results are expressed as mean ± SEM (n = 4 to 5 for desmosterol analyses and n = 9 to 10 for FACS and LM analysis, combined from 2 individual experiments). *P < 0.05, **P < 0.01, ****P < 0.001; nonparametric Student’s t test (Mann–Whitney) for LM and desmosterol analysis and unpaired Student’s t test for cellular analysis. ns, not significant.

immunological role of DHCR24 during inflammation. Using a re- found in the peritoneal cavity might reflect an altered reaction to cently developed highly selective and potent DHCR24 inhibitor zyA stimulation, mediated through Toll-like receptor (TLR) (SH42) as a chemical probe (21), we aimed at investigating the stimulation. Ito et al. (48) showed that LXRs control a plethora consequences of controlled DHCR24 inhibition. In order to sepa- of genes involved in lipid metabolism; this function correlates rate direct and indirect effects on DHCR24/LXRs, it was manda- with their ability to modulate organization and tory to establish the probes’ selectivity, allowing for streamlined thereby influence TLR signaling (18, 23). A direct link between data interpretation without observing possibly mixed activities. LXRs and Alox-12/15 has been established, particularly in hu- Moreover, we wanted to exclude known side effects from nonselective man M2 MΦ (49). However, to our knowledge, no report has yet LXR activation (19), aiming to exploit the selective functions of pointed out an LXR-dependent increase in the expression of desmosterol. We particularly aimed at boosting endogenous PUFA peritoneal Alox-12/15. Nevertheless, it seems conceivable that biosynthesis and identified desmosterol as a suitable endogenous changes in lipid metabolism, caused by desmosterol-driven LXR metabolite mediating such effects. In order to test this hypothesis, activation, might ultimately translate into a differential response we initially investigated the probe’s influence on the course of to zyA stimulation. This might explain altered gene expression zyA-induced peritonitis (30). SH42 did have a marked influence profiles at the early 4-h time point. on relevant numbers and types of inflammation-related immune Having observed these marked changes in lipid metabolism cells. Particularly antiinflammatory/tissue-regenerative cell types and gene expression of related enzymes, we next investigated were found up-regulated upon treatment with SH42. In line with PUFA metabolism. Using a standardized LC-MS/MS platform, these findings, increased levels of IL-10 and TGF-β and in- we analyzed peritoneal LM profiles under treatment with SH42 creased efferocytotic capacity were found. In order to shed some during murine peritonitis. As can be seen in Fig. 5, we observed light on the underlying molecular mechanisms, we carried out increased free PUFA levels in mice under treatment with SH42 quantitative lipidomic analysis. at the 4-h time point, and this finding correlated well with our Monitoring almost 1,000 individual lipid species, we observed lipidomic results. It is likely that the increased levels of PUFA increased amounts of longer chain FA as well as increased PUFA stem from the altered lipid metabolism/composition of the res- content. This change in lipid metabolism, which was mainly ob- idential MΦ, as described by lipidomic and gene expression served in residential peritoneal cells, proved that our hypothesis of analysis. Additionally, increased levels of the well-established boosting PUFA biosynthesis through the controlled accumulation eicosanoids LTB4 and PGE2, as well as the proresolving of desmosterol is possible. As SPMs are well-known and charac- MaR1 and the epoxy metabolite 19,20-EpDPA terized downstream products of PUFA (7), we rationalized that and its inactive metabolite 19,20-DiHDPA, were observed. LTB4 altered lipid metabolism, leading to increased endogenous PUFA levels were enhanced at the 4-h time point in the treatment group, production, might ultimately lead to changed LM profiles during with a rapid decrease toward the 12-h time point. The levels of onset and offset of inflammation. Indeed, we found a significant PGE2 remained elevated at 12 h in the treatment group. Although increase in the gene expression of the desaturase Fads6 in resi- this finding might initially seem counterintuitive, it is important to dential peritoneal cells (Fig. 4) from mice treated with SH42. realize that PGE2 has been described as an important mediator of After inducing inflammation, a significant difference between LM class switching in the early phase of inflammatory processes vehicle and SH42 treatment in the expression of the key pro- (40). Moreover, several recent reports have underlined possible resolving enzyme Alox-12/15 was observed. This result clearly antiinflammatory effects exerted by PGE2 (50). In turn, the in- pointed to a molecular modulation of the course of murine creased levels of PGE2 found at 4 and 12 h are likely contribut- peritoneal inflammation. The increased expression of Alox-12/15 ing to the observed effects of SH42. Additionally, the increased

20630 | www.pnas.org/cgi/doi/10.1073/pnas.1911992116 Körner et al. Downloaded by guest on September 30, 2021 INFLAMMATION IMMUNOLOGY AND

Fig. 8. Functional studies of SH42, 19,20-EpDPA, and MaR1. (A) Phagocytosis under treatment with SH42. Ctrl, control. (B) Phagocytosis with desmosterol treatment. (C) Phagocytosis under cotreatment with GSK2033. Results in A–C represent at least 2 independent experiments and are expressed as mean ± SEM (n = 9 to 11 per group); 1-way ANOVA with Dunnett’s multiple comparison test. (D) Desmosterol analysis in SH42-treated MΦ (2 h); nonparametric Student’s t test. (E and F) Phagocytosis with MaR1 (E) and phagocytosis with 19,20-EpDPA (F) treatment. (G) Phagocytosis comparing wild-type and 12/15-LOX KO MΦ under treatment (both) with SH42. (E–G) Unpaired 2-tailed Student’s t test. Cell shape analysis of human PBMCs stimulated for 7 d with vehicle (H), GM-CSF (I), M-CSF (J), desmosterol (K), or SH42 (L) is shown. (Scale bar: 100 μm.) (M–P) Statistical analysis of cell shape, cell area, cell perimeter, and cell length (n = 3to11 independent experiments expressed as mean ± SEM). ns, not significant. (A–F and M–P)*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001; 1-way ANOVA with Bonferroni correction. Detailed information about the experimental procedures used for cell shape analysis in H–L can be found in SI Appendix,S9.

production of MaR1 was in line with the altered gene expression of possibly lipid rafts, altering cellular signaling. Integration of LXR Alox-12/15, a key enzyme necessary for the biosynthesis of antiin- and AKT signaling mediated by changes in lipid rafts has been flammatory/proresolving mediators. With respect to the involved described in cancer cells (54). However, to our knowledge, the receptors, to our knowledge, no definite receptor has yet been control of AKT/PI3K and mTOR in a DHCR24/LXR-dependent identified for 19,20-EpDPA (51). However, MaR1 has been shown fashion has not yet been shown. to be a partial agonist of recombinant human BLT1 receptor, likely Next, we tested if SH42 would also prove beneficial in a ther- antagonizing the actions of the proinflammatory mediator LTB4 apeutic setting and carried out additional mouse experiments in (52). The dissociation constant (Kd) values for several proresolving order to gain more mechanistic and functional insights into the LMs are described to be in the pico- to nanomolar range (53), a relation between desmosterol, LXRs, and the observed antiin- concentration range that can possibly be reached in our model. flammatory effects. As can be seen in Fig. 7, SH42 did also give Along this line, we argued that Alox-12/15–derived LMs, as well as rise to antiinflammatory effects when administered at the peak of the increased concentrations of epoxy metabolites (i.e., 19,20- inflammation (24-h time point). It was important to see that after EpDPA), are the final downstream products possibly explaining the a single dose of SH42, we could already observe the accumulation actions of SH42 (38). of desmosterol, as, otherwise, possible antiinflammatory effects In order to gain more insight on the proteomic level, we car- would have had to be attributed to SH42 itself. Nevertheless, it ried out protein array analysis of the AKT/PI3K and mTOR is important to stress that we cannot entirely exclude other bio- pathways (Fig. 6) in residential peritoneal cells. Also, on the logical effects related to SH42, beyond inhibiting DHCR24. Fi- proteomic level, we found a significant skewing toward a more nally, we investigated the mechanistic relationship between SH42, antiinflammatory phenotype. Interestingly, these results point to desmosterol, MaR1, 19,20-EpDPA, LXRs, and ALOX-12/15. For an intertwined change of lipidome and proteome, jointly changing this, we chose monocyte differentiation and polarization, as well as the metabolism and actions of peritoneal monocytes, thereby MΦ phagocytosis, as functional readouts, 2 important hallmarks of mediating the actions of SH42. It seems plausible that these ob- inflammation resolution. As can be seen in Fig. 8, MaR1, 19,20- servations are mediated by changes in lipid metabolism and EpDPA, SH42, and desmosterol were capable of inducing the

Körner et al. PNAS | October 8, 2019 | vol. 116 | no. 41 | 20631 Downloaded by guest on September 30, 2021 Fig. 9. Analysis of MΦ phenotype and PUFA, 15-LOX, CYP, and COX pathway markers. M1 MΦ were stimulated with TNF-α and, in combination with SH42 or desmosterol gene expression, were analyzed for M1 markers (A), M2 markers (B), proresolving receptors (C), and phagocytic markers (D). (E) Groups are described; all mRNA levels were quantified by RT-PCR (n = 9 to 11). The results are representative of 3 to 11 independent experiments and are expressed as the mean ± SEM. *P < 0.05, **P < 0.01; unpaired 2-tailed Student’s t test. (F) LC-MS/MS analysis of PUFA, 15-LOX, CYP, and COX pathway markers (n = 6 for all experiments, with 3 biological replicates and 2 technical repeats; mean ± SEM). **P < 0.01, nonparametric Student’s t test (Mann–Whitney). ns, not significant.

proresolving M2 MΦ type, as well as enhancing phagocytosis. have been linked to the development of asthma, arthritis, and Much higher doses of SH42 were needed when compared with atherosclerosis in humans. Products of this pathway (MaR1 and desmosterol, MaR1, and 19,20-EpDPA. This speaks for the fact LXA4) have been described in a human skin blister model (55, 56). that desmosterol, MaR1, and 19,20-EpDPA are the actual medi- In summary (Fig. 10), we describe here the antiinflammatory ef- ators of the effects observed under SH42 treatment. We sub- fects of SH42 as a highly potent and selective inhibitor of DHCR24. stantiated our findings by gene expression analysis, showing a shift SH42 leads to accumulation of desmosterol, subsequently activating toward the prohealing MΦ phenotype (Fig. 9). Additionally, our LXRs and affecting PUFA metabolism mainly in residential functional findings were accompanied by an increased expression of monocytes. These changes are downstream-translated into the several important proresolving receptors (GPR32, ChemR23, ALX, increased production of proresolving mediators, likely mediating and GPR18), strengthening our line of argumentation. These re- the observed effects. In turn, targeting lipid metabolism at a ceptors have just recently been shown to be involved in inflammation presumably distinct point (e.g., cholesterol biosynthesis) can have resolution in a human skin blister model (55). Ultimately, we profound effects on the general lipid composition, with funda- monitored important CYP, 15-LOX, and COX pathway markers mental consequences for cellular phenotypes and inflammatory under treatment of human MΦ with zyA, with and without SH42. outcome. Endogenous n3-PUFA production has long been con- We found a dramatic shift toward the 15-LOX and CYP pathways sidered negligible when compared with dietary sources. It has under cotreatment with SH42 (Fig. 9). MΦ only stimulated with zyA been established that humans almost entirely depend on the di- almost solely relied on COX metabolism, producing large amounts etary uptake of n3-PUFA, and the possibility of boosting endoge- of PGE2. The aforementioned findings might lead to novel treat- nous n3-PUFA biosynthesis to target crucial physiological processes ment opportunities, as polymorphisms in the 15-LOX pathway such as inflammation resolution is vastly unexplored. However, our

20632 | www.pnas.org/cgi/doi/10.1073/pnas.1911992116 Körner et al. Downloaded by guest on September 30, 2021 Inhibitor experiments with GSK 2033 were carried out as follows. Cells were treated as described above. In the GSK 2033-treated groups, cells were pretreated with GSK 2033 (5 μM)for2hbeforeSH42treatment. Cells were ultimately in- cubatedwithSH42(5μM) with or without pretreatment for 8 h with either 10% FCS or 0.5% BSA in DMEM (no FCS). Control groups were adjusted to 0.1% ethanol and/or DMSO if necessary. RNA was isolated using Tripure RNA Isolation Reagent (Roche) according to the manufacturer’s protocol. Total RNA (1 μg) was reverse-transcribed using M-MLV (Moloney murine leukemia virus) Reverse Transcriptase (Promega) for RT-qPCR analysis according to the manufacturer’s instructions to produce complementary DNA. mRNA expressions were normalized to β-actin and cyclophilin A mRNA expressions and expressed as fold change compared with the vehicle-treated group using the ΔΔCT method. Primer se- quences can be found in SI Appendix,S11.

GC-MS Analysis of Desmosterol Levels. Desmosterol levels were determined as described elsewhere, with some modifications (21). A detailed description of the method is provided in SI Appendix,S12.

zyA-Induced Peritonitis and FACS Analysis. All animal protocols were performed Fig. 10. Overview and summary of the observed actions of SH42 in a sche- in accordance with the regulations of the Regierungspräsidium Tübingen and the matic cell. The effects of SH42 are shown in orange, zyA activation is shown in local ethics committee. For all experiments, SH42 was prepared in a 50% sterile blue, and combined effects are shown in orange/blue. SH42 selectively blocks Cremophor solution (Merck) to a final concentration of 50 μg/μL. Six- to 8-wk-old DHCR24 [1]. Inhibition of DHCR24 leads to increased levels of desmosterol [2]. female C57BL/6 mice (Charles River Laboratories) were injected intravenously (i.v.) Desmosterol activates LXRs [3]. LXR activation leads to increased endogenous via the tail vein. Mice were injected daily i.v. with SH42 (0.5 mg) or vehicle control PUFA biosynthesis [4]. LXR activation leads to changes in lipid composition and in a total volume of 150 μL for a period of 3 d. The mice were then injected i.p. metabolism [5]. The altered lipidome results in increased production of anti- with 1 mL of zyA (1 mg/mL; Z4250; Sigma–Aldrich). Peritoneal lavages (obtained inflammatory and proresolving LMs after zyA stimulation [6]. LMs are known in 4 mL of phosphate-buffered saline [PBS] without calcium and magnesium − − effectors of immune cell phenotype and function; we also observed significant [PBS / ; Sigma]) and tissues were obtained at 4, 12, 24, and 48 h. Collected exu- changes in their corresponding receptor expression [7]. Taken together, these dates were counted using a hemocytometer. A 2-mL peritoneal lavage aliquot actions result in accelerated inflammation resolution and modulation [8]. was then spun down (300 × g, 5 min, 4 °C), and the pellet was resuspended in 300

− − INFLAMMATION μL of flow cytometry staining buffer (FCSB) (1× PBS / ,1%BSA,2mMNaN3).A IMMUNOLOGY AND 100-μL sample volume was then transferred to a 96-well plate (round bottom). data presented here strongly indicate that interfering with lipid Antibodies were diluted with FCSB before use. For differential cell counting, cells metabolism at an appropriate junction can trigger endogenous PUFA were blocked with mouse anti-CD16/CD32 (101320; BioLegend; 1:50 [vol/vol]) biosynthesis and the downstream production of proresolving lipids, antibodies for 10 min at room temperature and then stainedwithanti-mouse which, in turn, translates into important functional and phenotypic e450-F4/80 (48-4801-82; eBioscience; 1:100 [vol/vol]), Allophycocyanin-Ly6G (APC- changes. We are confident that our study can form the fundament for Ly6G) (127614; BioLegend; 1:250 [vol/vol]) and FITC-Ly6C (128006; BioLegend; a better understanding of the interplay between cholesterol bio- 1:250 [vol/vol]) antibodies for 30 min at 4 °C. For determination of MΦ synthesis, LXRs, and the production of proresolving lipids, pos- phagocytosis of apoptotic PMNs, cells were permeabilized using a fixation and sibly posing a novel means for targeting inflammatory diseases. permeabilization kit (554714; Becton Dickinson) according to the manufac- turer’s instructions, following staining with anti-mouse PerCP-Cy5.5–conju- Materials and Methods gated anti-Ly6G (127616; BioLegend; 15:1,000 [vol/vol]) for 30 min at 4 °C. Data acquisition was performed on a FACSCanto II flow cytometer (Becton Dickinson ) The chemical probe SH42 was prepared as described elsewhere (21). All using FACSDiva software (Becton Dickinson), and data were analyzed with animal experiments were approved by the Local Ethical Commission of the FlowJo (TreeStar). The gating strategy can be found in SI Appendix,S13(45). For Leiden University Medical Center or the Regierungspräsidium Tübingen. All therapeutic treatment with SH42, peritonitis was initiated as described above – chemicals were from Sigma Aldrich and were of pro analysi or LC-MS grade, using 1 mL of a zyA solution. A single dose of SH42 was then administered i.v. if not stated otherwise. 24 h later, and mice were killed at the 48-h time point.

Φ Human M Differentiation, Polarization, and Lipid Analysis. Human PBMCs Determination of IL-10, IL-6, IL-8, TNF-α, and TGF-β. Analysis was carried out were isolated from human leukapheresis collars from the Blood Bank of Eberhard- according to the manufacturer’s instructions using a standard enzyme-linked Karls University of Tübingen. For differentiation experiments, cells were cultured immunosorbent assay (R&D Systems). in RPMI medium 1640 alone, with 10 ng/mL human recombinant granulocyte/ macrophage (GM)-CSF (Macs Milteny), 100 ng/mL M-CSF (Macs Milteny), Transcriptional Analysis of Peritoneal Cells and MΦ. Human PBMCs were dif- 0.01 μg/mL desmosterol, or 250 μg/mL SH42 at 37 °C for 7 d. For polarization ferentiated for 7 d in RPMI 1640 supplemented with GM-CSF (10 ng/mL) and experiments, M1 MΦ (monocytes cultured with GM-CSF for 7 d) were stim- then treated with TNF-α (100 ng/mL), SH42 (250 μg/mL), and desmosterol ulated with tumor necrosis factor-α (TNF-α; 100 ng/mL; Promokine) with and (0.01 μg/mL). Primer sequences can be found in SI Appendix,S11 without SH42 (250 μg/mL) or desmosterol (0.01 μg/mL) for 24 h before transcriptional analysis. For LM analysis, 5 × 106 M1 MΦ were stimulated Peritoneal lavages from SH42- and vehicle-treated mice were used fol- with SH42 (250 μg/mL) and/or in combination with zyA (50 μg/mL) for the lowing 0 and 4 h of zyA-induced peritonitis prior to transcriptional analysis. indicated time points (4 and 24 h), quenched with 1.5 mL of ice-cold Murine 18S expression as a housekeeping gene was evaluated. Primer se- methanol, and kept at −80 °C until analysis, as described below. In an ad- quences can be found in SI Appendix,S11. ditional set of experiments, baicalein (1 μg/mL) was added together with Φ SH42 and zyA, followed by incubation for 24 h. Protein Array Analysis. Peritoneal monocytes/M from 3-d SH42- and vehicle-treated mice were used following 4 h of zyA-induced peritonitis. Protein and phosphory- β LXR Selectivity Assessment. RAW 264.7 cells were maintained in high-glucose lation (TGF- Phospho Antibody Array, PTG176; FullMoonBioscience) profiling of Dulbecco’s modified Eagle’s medium (DMEM) with 10% FCS and 1% penicillin/ peritoneal monocytes (pooled lavages from 5 mice per condition) was performed according to the manufacturer’s instructions. Image acquisition was performed by streptomycin under a 5% CO2-humidified atmosphere at 37 °C. Before treat- ment, cells were seeded into 24-well plates and incubated for at least 12 h. the manufacturer. A detailed description is provided in SI Appendix,S14. Different concentrations of SH42 (dissolved in ethanol; 0, 0.5, 1, 5, and 10 μM) were tested, and treated cells were incubated for 8 h with 10% FCS or 0.5% Human and Murine MΦ Phagocytosis. Human MΦ were prepared from PBMCs bovine serum albumin (BSA) in DMEM. In the FCS-starved groups, medium using the Histopaque (Sigma–Aldrich) method and differentiated with GM-CSF; contained 0.5% BSA and 1% penicillin/streptomycin DMEM for 8 h before subsequently, 0.1 × 106 cells per well were incubated with test substances as SH42 treatment. In the atorvastatin-treated groups, cells were pretreated with stated. Murine MΦ were obtained by peritoneal lavage and plated on 48-well atorvastatin (dissolved in DMSO; 10/25 μM) for 8 h before SH42 treatment. plates for 1 h in PBS with calcium and magnesium to allow adherence. Next, MΦ

Körner et al. PNAS | October 8, 2019 | vol. 116 | no. 41 | 20633 Downloaded by guest on September 30, 2021 were stimulated with SH42 (250 μg/mL), MaR1 (10 ng/mL), or 19,20 EpDPA Quantitative Lipidomic Analysis. Quantitative lipidomic analysis was carried (10 ng/mL) as indicated. GSK 2033 was used at a concentration of 200 nM. out as described elsewhere (34), with some modification. A detailed de- Fluorescent zyA particles (Z2841; Thermo Fisher Scientific) were added at a 1:30 scription of the method is provided in SI Appendix, S15. ratio (MΦ/zyA) and incubated at 37 °C for 60 min to allow phagocytosis. Fluo- rescence was determined using a fluorescent plate reader (Tecan). Targeted LM Analysis. Targeted LM analysis was carried out as described elsewhere (45). Further details are available in SI Appendix, S16 and Table S1, Φ Western Blot Analysis. Peritoneal monocytes/M from 3-d SH42- and vehicle- and raw data are available in Dataset S6. treated mice were used following 4 h of zyA-induced peritonitis. Cell lysates were obtained using radioimmunoprecipitation buffer (89900; Thermo Fisher ACKNOWLEDGMENTS. We thank Shane Soogea, Evelyne Steenvoorden, and Scientific), determination of protein quantity was performed using the stan- Marieke Heijink for their help with lipid analysis. We thank Marije Kuipers dard protocol of the Pierce BCA Protein Assay Kit (23225; Thermo Fisher Sci- for help with the artwork of Fig. 10. This work was supported by 2 grants entific), and protein levels were normalized and loaded in NuPAGE Bis-Tris gels from the Deutsche Forschungsgemeinschaft (DFG; German Research Foun- (NP0335BOX; Thermo Fisher Scientific) and blotted on polyvinylidene difluoride dation): DFG-MI 1506/5-1 and Projektnummer 374031971–TRR 240 (to V.M.); membranes. The following antibodies were used: anti-mTOR (2972; Cell and IZKF-Fortüne Grant 2377-0-0 (to A.K.). This research was also supported Signaling Technology), anti–phospho-mTOR (T2446) (ab63552; Abcam), anti– by the Enabling Technologies Hotels Programme-ZonMW Grant 4350004007 pan-AKT (ab8805; Abcam), and anti–phospho-pan AKT (T308) (ab38449; (to Y.W. and M.G.). Y.W. is supported by VENI Grant 91617027 from the Abcam); for a loading control, an anti-GAPDH antibody (G9545; Sigma–Aldrich) Netherlands Organization for Scientific Research-ZonMW. M.G. is supported was used. A goat anti-rabbit antibody (375699; ABIN) conjugated with alkaline by and coordinator of the Horizon2020 Innovative Training Network (ITN) phosphatase was used as a secondary antibody. project, ArthritisHeal 812890.

1. N. N. Pavlova, C. B. Thompson, The emerging hallmarks of cancer metabolism. Cell 29. C. M. Quinn, W. Jessup, J. Wong, L. Kritharides, A. J. Brown, Expression and regulation Metab. 23,27–47 (2016). of sterol 27-hydroxylase (CYP27A1) in human macrophages: A role for RXR and 2. M. M. Rinschen, J. Ivanisevic, M. Giera, G. Siuzdak, Identification of bioactive me- PPARgamma ligands. Biochem. J. 385, 823–830 (2005). tabolites using activity metabolomics. Nat. Rev. Mol. Cell Biol. 20, 353–367 (2019). 30.J.L.Cash,G.E.White,D.R.Greaves,“Zymosan‐induced peritonitis as a simple 3. A. Luengo, D. Y. Gui, M. G. Vander Heiden, Targeting metabolism for cancer therapy. experimental system for the study of inflammation” in Methods in Enzymology, Cell Chem. Biol. 24, 1161–1180 (2017). T. M. Handel, D. J. Hamel, Eds. (Academic Press, 2009), vol. 461, chap. 17 pp. 379–396. 4. P. Hiebert, S. Werner, Targeting metabolism to treat psoriasis. Nat. Med. 24, 537–539 31. C. N. Serhan, B. D. Levy, Resolvins in inflammation: Emergence of the pro-resolving – (2018). superfamily of mediators. J. Clin. Invest. 128, 2657 2669 (2018). 5. C. N. Serhan, Discovery of specialized pro-resolving mediators marks the dawn of 32. J.-Y. Choi et al., Mer signaling increases the abundance of the transcription factor LXR resolution physiology and pharmacology. Mol. Aspects Med. 58,1–11 (2017). to promote the resolution of acute sterile inflammation. Sci. Signal. 8, ra21 (2015). 6. J. A. van Diepen, J. F. P. Berbée, L. M. Havekes, P. C. N. Rensen, Interactions between 33. D. Mailänder-Sánchez et al., Antifungal defense of probiotic Lactobacillus rhamnosus GG is mediated by blocking adhesion and nutrient depletion. PLoS One 12, e0184438 (2017). inflammation and lipid metabolism: Relevance for efficacy of anti-inflammatory 34. B. K. Ubhi, “Direct infusion-tandem mass spectrometry (DI-MS/MS) analysis of complex drugs in the treatment of atherosclerosis. Atherosclerosis 228, 306–315 (2013). lipids in human plasma and serum using the lipidyzer platform” in Clinical Metabolomics: 7. C. N. Serhan, N. A. Petasis, Resolvins and protectins in inflammation resolution. Chem. Methods and Protocols, M. Giera, Ed. (Springer, New York, NY, 2018), pp. 227–236. Rev. 111, 5922–5943 (2011). 35. C. He et al., Inhibiting delta-6 desaturase activity suppresses tumor growth in mice. 8. N. J. Spann, C. K. Glass, Sterols and oxysterols in immune cell function. Nat. Immunol. PLoS One 7, e47567 (2012). 14, 893–900 (2013). 36. T. Matsuzaka, H. Shimano, Elovl6: A new player in fatty acid metabolism and insulin 9. E. D. Muse et al., Cell-specific discrimination of desmosterol and desmosterol mimetics sensitivity. J. Mol. Med. (Berl.) 87, 379–384 (2009). confers selective regulation of LXR and SREBP in macrophages. Proc. Natl. Acad. Sci. 37. M. Schlegel et al., The neuroimmune guidance cue netrin-1 controls resolution pro- – U.S.A. 115, E4680 E4689 (2018). grams and promotes liver regeneration. Hepatology 63, 1689–1705 (2016). 10. C. Wu, I. Miloslavskaya, S. Demontis, R. Maestro, K. Galaktionov, Regulation of cellular 38. Y. Shimanaka et al., Omega-3 fatty acid epoxides are autocrine mediators that control – response to oncogenic and oxidative stress by Seladin-1. Nature 432, 640 645 (2004). the magnitude of IgE-mediated mast cell activation. Nat. Med. 23, 1287–1297 (2017). 11. T. Takano et al., Augmentation of DHCR24 expression by hepatitis C virus infection 39. D. W. Gilroy et al., CYP450-derived oxylipins mediate inflammatory resolution. Proc. facilitates viral replication in hepatocytes. J. Hepatol. 55, 512–521 (2011). Natl. Acad. Sci. U.S.A. 113, E3240–E3249 (2016). 12. A. I. Cederbaum, Molecular mechanisms of the microsomal mixed function oxidases 40. B. D. Levy, C. B. Clish, B. Schmidt, K. Gronert, C. N. Serhan, Lipid mediator class switching and biological and pathological implications. Redox Biol. 4,60–73 (2015). during acute inflammation: Signals in resolution. Nat. Immunol. 2,612–619 (2001). 13. N. J. Spann et al., Regulated accumulation of desmosterol integrates macrophage 41. B. D. Manning, A. Toker, AKT/PKB signaling: Navigating the network. Cell 169, 381–405 lipid metabolism and inflammatory responses. Cell 151, 138–152 (2012). (2017). 14. H. Guo et al., Modulation of long noncoding RNAs by risk SNPs underlying genetic 42. A. J. Covarrubias et al., Akt-mTORC1 signaling regulates Acly to integrate metabolic predispositions to prostate cancer. Nat. Genet. 48, 1142–1150 (2016). input to control of macrophage activation. eLife 5, e11612 (2016). 15. A. Varin et al., Liver X receptor activation promotes polyunsaturated fatty acid syn- 43. M. C. Greenlee-Wacker, Clearance of apoptotic and resolution of in- thesis in macrophages: Relevance in the context of atherosclerosis. Arterioscler. flammation. Immunol. Rev. 273, 357–370 (2016). Thromb. Vasc. Biol. 35, 1357–1365 (2015). 44. F. Y. McWhorter, T. Wang, P. Nguyen, T. Chung, W. F. Liu, Modulation of macrophage 16. C. Yang et al., Sterol intermediates from cholesterol biosynthetic pathway as liver X phenotype by cell shape. Proc. Natl. Acad. Sci. U.S.A. 110, 17253–17258 (2013). receptor ligands. J. Biol. Chem. 281, 27816–27826 (2006). 45. A. Körner et al., Sympathetic nervous system controls resolution of inflammation via 17. Y. Oishi et al., SREBP1 contributes to resolution of pro-inflammatory TLR4 signaling regulation of repulsive guidance molecule A. Nat. Commun. 10, 633 (2019). by reprogramming fatty acid metabolism. Cell Metab. 25, 412–427 (2017). 46. C. D. Buckley, D. W. Gilroy, C. N. Serhan, Proresolving lipid mediators and mechanisms – 18. C. Hong, P. Tontonoz, Liver X receptors in lipid metabolism: Opportunities for drug in the resolution of acute inflammation. Immunity 40, 315 327 (2014). – discovery. Nat. Rev. Drug Discov. 13, 433–444 (2014). 47. S. Gordon, Phagocytosis: An immunobiologic process. Immunity 44, 463 475 (2016). 19. J. R. Schultz et al., Role of LXRs in control of lipogenesis. Genes Dev. 14, 2831–2838 48. A. Ito et al., LXRs link metabolism to inflammation through Abca1-dependent reg- ulation of membrane composition and TLR signaling. eLife 4, e08009 (2015). (2000). 49. J. T. Huang et al., Interleukin-4-dependent production of PPAR-γ ligands in macro- 20. S. Gadde et al., Development of therapeutic polymeric nanoparticles for the resolu- phages by 12/15-lipoxygenase. Nature 400, 378–382 (1999). tion of inflammation. Adv. Healthc. Mater. 3, 1448–1456 (2014). 50. G. A. FitzGerald, BIOMEDICINE. Bringing PGE in from the cold. Science 348, 1208– 21. C. Müller et al., New chemotype of selective and potent inhibitors of human delta 24- 2 1209 (2015). dehydrocholesterol reductase. Eur. J. Med. Chem. 140, 305–320 (2017). 51. D. Bishop-Bailey, S. Thomson, A. Askari, A. Faulkner, C. Wheeler-Jones, Lipid-metabolizing 22. M. Giera, F. Plössl, F. Bracher, Fast and easy in vitro screening assay for cholesterol CYPs in the regulation and dysregulation of metabolism. Annu. Rev. Nutr. 34,261–279 biosynthesis inhibitors in the post-squalene pathway. Steroids 72, 633–642 (2007). (2014). 23. A. C. Calkin, P. Tontonoz, Transcriptional integration of metabolism by the nuclear 52. R. A. Colas et al., Identification and actions of the maresin 1 metabolome in infectious – sterol-activated receptors LXR and FXR. Nat. Rev. Mol. Cell Biol. 13, 213 224 (2012). inflammation. J. Immunol. 197, 4444–4452 (2016). 24. W. J. Zuercher et al., Discovery of tertiary sulfonamides as potent liver X receptor 53. M. Spite, J. Clària, C. N. Serhan, Resolvins, specialized proresolving lipid mediators, – antagonists. J. Med. Chem. 53, 3412 3416 (2010). and their potential roles in metabolic diseases. Cell Metab. 19,21–36 (2014). 25. Y. Jiao, Y. Lu, X. Y. Li, Farnesoid X receptor: A master regulator of hepatic triglyceride 54. A. J. C. Pommier et al., Liver X Receptor activation downregulates AKT survival sig- and glucose homeostasis. Acta Pharmacol. Sin. 36,44–50 (2015). naling in lipid rafts and induces of prostate cancer cells. Oncogene 29, 26. J. Y. L. Chiang, Regulation of bile acid synthesis: Pathways, nuclear receptors, and 2712–2723 (2010). mechanisms. J. Hepatol. 40, 539–551 (2004). 55. M. P. Motwani et al., Pro-resolving mediators promote resolution in a human skin model 27. J. Zhou et al., A novel pregnane X receptor-mediated and sterol regulatory element- of UV-killed Escherichia coli-driven acute inflammation. JCI Insight 3, 94463 (2018). binding protein-independent lipogenic pathway. J. Biol. Chem. 281, 15013–15020 (2006). 56. J. Wittwer, J. Marti-Jaun, M. Hersberger, Functional polymorphism in ALOX15 results 28. T. Li, W. Chen, J. Y. L. Chiang, PXR induces CYP27A1 and regulates cholesterol me- in increased allele-specific transcription in macrophages through binding of the tabolism in the intestine. J. Lipid Res. 48, 373–384 (2007). transcription factor SPI1. Hum. Mutat. 27,78–87 (2006).

20634 | www.pnas.org/cgi/doi/10.1073/pnas.1911992116 Körner et al. Downloaded by guest on September 30, 2021