Пространственное картирование оси AA-PGE2-EP при множественных повреждениях склероза.

Introduction

Multiple sclerosis (MS) is a chronic inflammatory disease of the central nervous system (CNS) that manifests itself by focal demyelinating lesions within the brain and spinal cord of people with MS (PwMS) [ 62 ]. MS lesion development is temporally diverse, but typically spans weeks to months [ 20 , 67 ] and is best described by a spectrum across different lesion stages [ 35 ]. In general, white matter lesions can be classified by their degree of neuroinflammation and demyelination. For example, microglia activation and clustering (termed microglia nodules) without clear signs of demyelination is suggested to predispose normal-appearing white matter (NAWM) to form pre-active lesions [ 9 , 76 , 77 ]. Progressing from there, active, mixed active/inactive (A/I) and chronic inactive (CIA) MS lesions all display myelin degradation but can be differentiated by their inflammatory profile (presence and activation status of microglia/macrophages). Within this profile, active MS lesions are characterized by the infiltration and accumulation of peripheral immune cells including lymphocytes and the presence of myelin phagocytosing microglia/macrophages throughout the lesion, whereas these cells are confined to the lesion border in mixed A/I lesions (hypocellular center) and almost absent in CIA [ 35 ]. To date, the local factors that drive the regional differences during MS lesion evolution remain largely elusive, and ways to further reveal the local metabolic environment that drives lesion inflammation and progression are, therefore, of high therapeutic value [ 27 ].

(Neuro)inflammation is partially orchestrated by essential polyunsaturated fatty acids (PUFAs), which are metabolized into bioactive lipid mediators (LMs) following, e.g., cytokine (e.g., TNFα or Il-1β) or reactive oxygen species stimulation [ 59 , 61 ]. PUFAs include omega-3 (e.g., docosahexaenoic acid) and omega-6 FAs [e.g., arachidonic acid (AA)], which are hydrolyzed upon stimulation by cytoplasmic phospholipase A2 and converted into LMs including prostaglandins and leukotrienes [ 53 ]. Selected products of the AA pathways are elevated in the plasma of PwMS and correlate with the disease disability score [ 5 ]. Cyclooxygenases (COX1 and 2) and prostaglandin synthases convert AA to LM including prostaglandin E 2 (PGE 2 ), which is abundantly present in cerebrospinal fluid of PwMS compared to controls [ 19 , 46 ], and lymphocytes increase their PGE 2 production and secretion just before MS disease onset/symptoms occur [ 11 ]. While current disease-modifying therapies effectively modulate peripheral inflammatory episodes, the (low-grade) inflammatory state of the CNS persists. This highlights the crucial need to obtain more information on both PUFA and LM distribution in the brain and their effect on MS lesion inflammation and progression to provide insight into molecular mechanisms and more effective treatment strategies.

PGE 2 signals via its four G-protein coupled receptors (GPCRs), termed Prostaglandin E Receptor 1-4 (EP1-4), and can evoke beneficial and/or detrimental effects [ 4 , 31 , 41 , 52 ]. Specifically, prostaglandin E2 receptor 2 (EP2) and prostaglandin E2 receptor 4 (EP4) activation have been reported in the context of inflammatory processes such as T-cell expansion and microglia activation [ 15 , 30 , 63 ]. Inhibition of both receptors synergistically suppressed the development of experimental autoimmune encephalomyelitis (EAE), an animal model of MS [ 14 ]. Of note, EP2 and EP4 are both associated with G αs proteins and adenylyl cyclase-cAMP signaling; however, the potential differences in their signal transduction pathways within the MS brain remain unclear.

In this study, we applied mass spectrometry imaging (MSI) to study the spatial distribution of AA and one of its key metabolites, PGE 2 , in human post-mortem brain tissue of non-neurological controls (NNCs) and PwMS. We observed lower levels of AA in MS compared to NNCs and specifically in MS lesions compared to peri-lesional tissue. Furthermore, COX2 ( PTGS2 ) and prostaglandin E synthase ( PTGES ) mRNA expression were higher in MS lesion tissue blocks compared to controls, and concomitantly PGE 2 /AA is increased in demyelinated WM lesions in MS. Of note, the expression of PGE 2 receptor EP4 decreases in MS tissue lysates, while EP2 levels increase, specifically in activated microglia. PGE 2 signaling in pro-inflammatory human-induced pluripotent stem cell (iPSC)-derived microglia evoked extensive cytokine-receptor pathway stimulation, strengthening of PGE 2 synthesis and initiation of homeostatic/resolving signaling, the latter mainly directed through EP2. Overall, we demonstrate a lipid distribution map in WM MS lesions that has enhanced PGE 2 -EP2 signaling, which potentially acts as a resolution mechanism to dampen neuroinflammation and thereby counteracts the initiating pro-inflammatory effects of PGE 2 in MS.

Materials and methods

Human tissue samples

Post-mortem fresh-frozen (FF) and formalin-fixed paraffin-embedded (FFPE) human white matter (WM) tissue blocks were obtained from people with clinically diagnosed MS (PwMS) (total n  = 31) and non-neurological controls (NNC) (total n  = 19) and provided by the MS Center Amsterdam, the Normal Aging Brain Collection Amsterdam, and the Netherlands Brain Bank. Relevant clinical information and the experimental application of each donor sample are listed in Table  1 (extended version in Table S4 ). All donors or their next of kin had given fully informed consent for autopsy and use of material for research from the Netherlands Brain Bank under ethical approval by the Medical Ethics Committee of the Free University Medical Center in Amsterdam (2009/148), project number 1127 and project number 412 in the University of Groningen. Table 1 Clinical and demographic data of MS and NNC subjects Case ID Age at death Sex PMD (h) Cause of death Type of MS Tissue preservation Lesion type Application NNC1 41 F 13:30 Pulmonary hemorrhage n/a FF n/a WB NNC2 98 M 8:40 Aortic dissection n/a FF n/a WB NNC3 82 M 13:35 Heart failure n/a FF n/a WB NNC4 53 M 14:25 Cardiac complications n/a FF n/a WB NNC5 78 M 5:55 Cardiac failure n/a FF n/a WB NNC6 77 F 9:15 Sudden death n/a FF n/a WB NNC7 88 F 5:40 Cardiac failure n/a FF n/a WB NNC8 77 F 19:45 Malignant lymphoma n/a FF n/a WB NNC9 91 F 5:45 Sudden death n/a FF n/a WB NNC019 68 M 8:40 Euthanasia n/a FF, FFPE n/a MSI, IHC (COX2, EP2), mRNA NNC095 68 M 7:35 Esophageal carcinoma, narrowing aa. coronaries n/a FFPE n/a IHC (EP2) NNC880 77 M 11:25 Pneumonia n/a FFPE n/a IHC (EP2) NNC945 59 F 8:10 Euthanasia, terminal COPD n/a FF n/a MSI, IHC(COX2), mRNA NNC989 71 F 7:50 Lung carcinoma n/a FF n/a MSI, IHC (COX2), mRNA NNC999 74 M 10:20 Euthanasia n/a FF n/a IHC (COX2), mRNA NNC227 93 M 6:00 Cachexia, cerebrovascular accident/kidney failure n/a FF n/a mRNA NNC254 75 F 5:40 Cachexia n/a FF n/a mRNA NNC275 82 F 5:10 Pneumonia by hemothorax n/a FF n/a mRNA NNC215 79 M 9:00 Pneumonia and metastasized kidney carcinoma n/a FF n/a mRNA MS1 70 F 6:30 Probably urosepsis SPMS FF unk WB MS2 53 F 10:45 Euthanasia SPMS FF unk WB MS3 48 F 4:50 Euthanasia SPMS FF unk WB MS4 48 F 8:10 Euthanasia SPMS FF unk WB MS5 75 F 8:00 Pneumonia SPMS FF unk WB MS6 43 M 8:30 Pneumonia SPMS FF unk WB MS7 53 M 5:53 Pneumonia PPMS FF unk WB MS8 52 F 8:25 Pneumonia PMS FF unk WB MS9 48 F 5:50 Congestive cardiac failure SPMS FF unk WB MS10 66 F 6:20 Cancer SPMS FF unk WB MS11 71 F 10:15 Post-surgery respiration problems PMS FF unk WB MS12 64 M 7:30 End stage prog. MS PPMS FF unk WB MS13 65 M 10:35 Urosepsis SPMS FF unk WB MS14 77 M 4:15 Cerebral vascular accident PPMS FF unk WB MS68 48 F 9:20 Pneumonia unk FF AI/I MSI, IHC (COX2), mRNA MS46 51 F 9:10 Euthanasia SPMS FF, FFPE A MSI, IHC (COX2, EP2), mRNA MS71 67 M 7:55 Euthanasia unk FFPE A/I IHC (EP2) MS85 67 F 11:25 Pneumonia unk FFPE A/I IHC (EP2) MS16 77 F 9:45 Aspiration pneumonia SPMS FF, FFPE A/I MSI, IHC (COX2, EP2), mRNA MS31 53 F 5:50 Euthanasia PMS FFPE A/I IHC (EP2) MS60 60 F 5:05 Euthanasia SPMS FF A/I MSI, IHC (COX2), mRNA MS42 82 F 7:30 Cardiac arrest, sudden death PPMS FF A/I MSI, IHC (COX2), mRNA MS32 54 M unk Progressive dyspnea unk FF A MSI, IHC (COX2), mRNA MS116 66 F 9:30 Euthanasia SPMS FF CIA MSI, IHC (COX2), mRNA MS120 73 M 8:45 Urosepsis SPMS FF A/I mRNA MS100 71 F 7:05 Cachexia with PMS and metastatic breast cancer PMS FF A/I mRNA MS298 53 F 10:45 Euthanasia PMS? FF unk mRNA MS51 47 M 7:15 Urosepsis with organ failure SPMS FF unk mRNA MS139 56 M 8:00 Pneumonia SPMS FF unk mRNA PMD post−mortem delay, unk unknown, n/a not applicable, mRNA messenger RNA, PMS progressive MS, PPMS primary progressive MS, SPMS secondary progressive MS, FF fresh frozen, FFPE formalin−fixed paraffin−embedded, CIA chronic inactive, A/I mixed active/inactive, A active, MSI Mass Spectrometry Imaging

Table 1 Clinical and demographic data of MS and NNC subjects Case ID Age at death Sex PMD (h) Cause of death Type of MS Tissue preservation Lesion type Application NNC1 41 F 13:30 Pulmonary hemorrhage n/a FF n/a WB NNC2 98 M 8:40 Aortic dissection n/a FF n/a WB NNC3 82 M 13:35 Heart failure n/a FF n/a WB NNC4 53 M 14:25 Cardiac complications n/a FF n/a WB NNC5 78 M 5:55 Cardiac failure n/a FF n/a WB NNC6 77 F 9:15 Sudden death n/a FF n/a WB NNC7 88 F 5:40 Cardiac failure n/a FF n/a WB NNC8 77 F 19:45 Malignant lymphoma n/a FF n/a WB NNC9 91 F 5:45 Sudden death n/a FF n/a WB NNC019 68 M 8:40 Euthanasia n/a FF, FFPE n/a MSI, IHC (COX2, EP2), mRNA NNC095 68 M 7:35 Esophageal carcinoma, narrowing aa. coronaries n/a FFPE n/a IHC (EP2) NNC880 77 M 11:25 Pneumonia n/a FFPE n/a IHC (EP2) NNC945 59 F 8:10 Euthanasia, terminal COPD n/a FF n/a MSI, IHC(COX2), mRNA NNC989 71 F 7:50 Lung carcinoma n/a FF n/a MSI, IHC (COX2), mRNA NNC999 74 M 10:20 Euthanasia n/a FF n/a IHC (COX2), mRNA NNC227 93 M 6:00 Cachexia, cerebrovascular accident/kidney failure n/a FF n/a mRNA NNC254 75 F 5:40 Cachexia n/a FF n/a mRNA NNC275 82 F 5:10 Pneumonia by hemothorax n/a FF n/a mRNA NNC215 79 M 9:00 Pneumonia and metastasized kidney carcinoma n/a FF n/a mRNA MS1 70 F 6:30 Probably urosepsis SPMS FF unk WB MS2 53 F 10:45 Euthanasia SPMS FF unk WB MS3 48 F 4:50 Euthanasia SPMS FF unk WB MS4 48 F 8:10 Euthanasia SPMS FF unk WB MS5 75 F 8:00 Pneumonia SPMS FF unk WB MS6 43 M 8:30 Pneumonia SPMS FF unk WB MS7 53 M 5:53 Pneumonia PPMS FF unk WB MS8 52 F 8:25 Pneumonia PMS FF unk WB MS9 48 F 5:50 Congestive cardiac failure SPMS FF unk WB MS10 66 F 6:20 Cancer SPMS FF unk WB MS11 71 F 10:15 Post-surgery respiration problems PMS FF unk WB MS12 64 M 7:30 End stage prog. MS PPMS FF unk WB MS13 65 M 10:35 Urosepsis SPMS FF unk WB MS14 77 M 4:15 Cerebral vascular accident PPMS FF unk WB MS68 48 F 9:20 Pneumonia unk FF AI/I MSI, IHC (COX2), mRNA MS46 51 F 9:10 Euthanasia SPMS FF, FFPE A MSI, IHC (COX2, EP2), mRNA MS71 67 M 7:55 Euthanasia unk FFPE A/I IHC (EP2) MS85 67 F 11:25 Pneumonia unk FFPE A/I IHC (EP2) MS16 77 F 9:45 Aspiration pneumonia SPMS FF, FFPE A/I MSI, IHC (COX2, EP2), mRNA MS31 53 F 5:50 Euthanasia PMS FFPE A/I IHC (EP2) MS60 60 F 5:05 Euthanasia SPMS FF A/I MSI, IHC (COX2), mRNA MS42 82 F 7:30 Cardiac arrest, sudden death PPMS FF A/I MSI, IHC (COX2), mRNA MS32 54 M unk Progressive dyspnea unk FF A MSI, IHC (COX2), mRNA MS116 66 F 9:30 Euthanasia SPMS FF CIA MSI, IHC (COX2), mRNA MS120 73 M 8:45 Urosepsis SPMS FF A/I mRNA MS100 71 F 7:05 Cachexia with PMS and metastatic breast cancer PMS FF A/I mRNA MS298 53 F 10:45 Euthanasia PMS? FF unk mRNA MS51 47 M 7:15 Urosepsis with organ failure SPMS FF unk mRNA MS139 56 M 8:00 Pneumonia SPMS FF unk mRNA PMD post−mortem delay, unk unknown, n/a not applicable, mRNA messenger RNA, PMS progressive MS, PPMS primary progressive MS, SPMS secondary progressive MS, FF fresh frozen, FFPE formalin−fixed paraffin−embedded, CIA chronic inactive, A/I mixed active/inactive, A active, MSI Mass Spectrometry Imaging

Clinical and demographic data of MS and NNC subjects

PMD post−mortem delay, unk unknown, n/a not applicable, mRNA messenger RNA, PMS progressive MS, PPMS primary progressive MS, SPMS secondary progressive MS, FF fresh frozen, FFPE formalin−fixed paraffin−embedded, CIA chronic inactive, A/I mixed active/inactive, A active, MSI Mass Spectrometry Imaging

Mass spectrometry imaging

FF human tissues were cryo-sectioned (CryoStar NX70, Thermo Fisher Scientific, Waltham, USA) into 10 µm, thaw-mounted onto super-frost microscopy glass slides, and stored at − 80 °C until further use. Before mass spectrometry imaging (MSI), excess salt was removed from the gently thawed tissue sections by immersing the slides for 1 min six times in Milli-Q water (18.2 MΩ) and drying the slides at room temperature under N2 flow. Silver-doped pneumatically assisted nanospray desorption electrospray ionization (PA nano-DESI) MSI was performed as previously described [ 12 ]. Briefly, the PA nano-DESI probe comprised two fused silica capillaries (50 µm inner diameter, 150 µm outer diameter) placed at ca. 90° angle to each other. A liquid bridge was formed at the intersection of the primary and the secondary capillary by delivering the extraction solvent through the primary capillary with a syringe pump (KD Scientific, Holliston, USA) at a flow rate of 0.5 µL/min. The extraction solvent was composed of acetonitrile:methanol v/v 9:1 (0.1% formic acid) spiked with 10 ppm 107 Ag + and internal standards (1 µM Prostaglandin E2 (PGE 2 )-d9, 1 µM arachidonic acid (AA)-d8) for normalization and quantitation [ 13 ]. Pneumatic assistance was provided by co-axial nitrogen flow to the tip with a gas pressure of 3.5 bar.

The tissue-containing slides were moved under the probe with an XYZ stage (Zaber Technologies, Canada) controlled by an in-house developed LabView program [ 37 ]. Lines were acquired by scanning at 50 µm/s in the X direction and stepping 150 µm in the Y direction, thus generating pixels of ~ 25 × 150 µm. The experiments were performed on an IQ-X mass spectrometer (Thermo Fischer Scientific) operated at a capillary voltage of 3.5 kV, and a heated capillary temperature of 275 °C. Data were acquired by alternating full scan (m/z 200–1100) and targeted selected ion monitoring (tSIM) (m/z 425–475) scan events with a resolution of 120 000 (at m/z 200), and the orbitrap automatic gain control target was set to 50% and 40%, respectively. Ion images were created with the in-house developed ion-to-image (i2i) application [ 42 ] using 5 parts per million (ppm) mass tolerance and an automatic contrast adjustment was applied to the 99th percentile of intensities. Quantitative ion images were generated in i2i by normalizing endogenous intensities pixel-by-pixel to the respective internal standard and ratio images by normalizing two endogenous signals.

Region selection and quantification

Consecutive brain sections of the tissues visualized by PA nano-DESI MSI were stained for myelin proteolipid protein (PLP) and Human Leukocyte Antigen–DR isotype (HLA-DR) to classify the amount of (de)myelination and inflammation in NNC and MS tissue. To this end, full tissue optical scans at 20 × magnification were obtained with the Olympus VS200 slide scanner (Evident, Tokyo, Japan). Brain tissue areas were created and stratified based on PLP and HLA-DR stainings using QuPath software (version 0.4.4) along with visually prominent intensity differences in ion images generated with PA nano-DESI MSI. The details of lesion classification are outlined in the Results section [ 21 , 35 ]. For analysis of the in-depth classification, ROIs with the same PLP or HLA-DR classification were averaged per donor. ROIs were transferred into the i2i application [ 42 ] and quantities were extracted for each lipid of interest using one-point calibration. Absolute values (µM/pixel) and relative values (PGE 2 /AA ratio) were used. Statistical analysis was performed with GraphPad Prism version 10.2.

Human iPSC-derived microglia

Human-induced pluripotent stem cells (hiPSC) were generated from NNC, with the approval of the LUMC scientific ethical committee, and obtained informed consent (NL45478.058.13/P13.080) [ 6 ]. The generation of iPSC-derived microglia (hiPSC microglia) was performed according to the protocol of Kenkhuis et al. and previously reported [ 21 , 32 ]. In short, iPSCs (cell line 114-1, female, 49 years) were used to develop mesodermal embryoid bodies, which were plated and supplemented with medium containing macrophage colony-stimulating factor (rhM-CSF, 100 µg/ml; 300-25, Peprotech) and Interleukin 3 (25 µg/ml; #200-03-B, Peprotech). Half of the media was refreshed weekly for 3–4 weeks and once mEBs started producing myeloid precursor cells, they were harvested weekly. To obtain mature microglia, the harvested myeloid precursor cells were plated in plates coated with 0.01 g/ml poly-D-lysine (Sigma, Saint Louis, MI, USA) and 0.01% gelatin type B (Sigma, Saint Louis, MI, USA) and cultured for 2 weeks in Advanced DMEM/F12 (Invitrogen, Waltham, MA, USA) supplemented with 100 U/mL penicillin–streptomycin (Invitrogen, Waltham, MA, USA), 2 mM Glutamax (ThermoFisher, Walthman, MA, USA), 50 µM 2-β-mercaptoethanol (ThermoFisher, Walthman, MA, USA), N2 supplement (ThermoFisher, Walthman, MA, USA), 2 µg/ml recombinant human GM-CSF (Peprotech, Rocky Hill, NJ, USA), and 20 µg/ml recombinant human IL-34 (Peprotech, Rocky Hill, NJ, USA). The phenotypic identity of the differentiated cells was confirmed by the immunocytochemistry analysis of microglia markers (Iba1, P2RY12, TMEM119). The acquired matured iPSC microglia were either used untreated (resting) or were subjected to treatment with lipopolysaccharide (LPS) (100 ng/ml; L2630, Sigma-Aldrich) and IFNγ (20 ng/ml; #300–02, Peprotech) (pro-inflammatory) for 24 h. For every batch, cell phenotypes were assessed by qPCR analysis. Subsequently, cells were washed twice with PBS and treated with PGE 2 (1 μM; P0409, Sigma-Aldrich) or vehicle (0.1% ethanol) for 24 h in serum-free medium. In addition, selective EP2 (10 μM; TG6-10-1, Calbiochem) and EP4 receptor inhibitors (1 μM; ONO AE3 208, Tocris) or vehicle (DMSO) were applied, together with the PGE 2 treatment. For RNA sequencing applications, hiPSC microglia (1 biological, 5 technical replicates per condition) were washed with PBS and lysed in 275 µl Buffer RLT Plus (#1053393; Qiagen), after which they were scraped, collected, and stored at − 70 °C until further use.

Immunostainings

For immunostainings, FF or FFPE human WM tissue blocks were sliced in 5 μm sections, dried or deparaffinized, and washed with Milli-Q water (Millipore). FFPE sections were subjected to heat-mediated antigen retrieval for 10 min (10 mM sodium citrate buffer, pH 6) and then brought to room temperature on ice for 30 min. FF sections were fixed with 4% PFA for 10 min at room temperature. All sections were washed with PBS and blocked with PBS containing 10% bovine serum albumin (BSA, Fraction V, Roche Diagnostics) or normal species serum (NSS) and 0.05% Tween20 (Sigma-Aldrich) for 30 min. Primary antibodies for EP2 (1:200; Abcam, ab124419), Iba1 (1:500; Abcam, ab5076), TMEM119 (1:50; R&D systems, MAB10313), COX2 (1:100; Cayman Chemical. 20,198), CD45 (1:200; Abcam ab10559,), Collagen IV (1:300, Invitrogen 51-9871-82), HLA-DR (1:500; Hybridoma), PLP (1:300; Serotec, MCA839G), and UEA-I (1:1000, Vector Labs B-1065) were diluted in PBS containing 1% BSA/NSS and 0.05% Tween20, and incubated in the dark overnight at 4 °C. Sections were washed with PBS and incubated with the Alexa fluorophore-conjugated secondary antibody diluted in PBS containing 0.05% Tween20, for 1 h at room temperature in the dark. For nuclear staining, Hoechst (33,258, Thermo Fisher Scientific) was diluted in PBS to a final concentration of 10 μg/mL and applied for 1 or 7 min in the dark. Partially, sections were submerged in Sudan Black B (0.03% in 70% EtOH) for 10 min to reduce autofluorescence. Lastly, sections were washed with PBS, mounted with Mowiol and a coverslip (Menzel-Glaser, thickness #1), and stored in the dark at 4 °C until image acquisition.

Microscopy and image acquisition

EP2-Iba1-TMEM119 immunoreactivity was assessed with a 10 × overview scan, after which ROIs were imaged with a 40× air objective using the Akoya Vectra Polaris Automated Quantitative Pathology Imaging System. Images were spectrally unmixed using inForm (version 3.0, Akoya Biosciences, Menlo Park, CA) and batch-analyzed using NIS elements (version 5.30.03, Nikon Europe B.V., Amsterdam, The Netherlands). Cells were defined by the overlap of nuclear and marker staining, and cell counts plus mean fluorescent intensity were extracted. Images of PLP, HLA-DR, COX2, CD45, UEA-I, and CollagenIV immunoreactivity were taken at the Olympus VS200 slide scanner using a 20 × overview scan and 60 × images (Olympus X line, 1.42 NA, oil) within the respective ROIs used previously for lipid quantification (Fig.  1 c). Images were deconvolved using Huygens Professional 21.10 software (scientific volume imaging B.V.) and (if possible) batch-analyzed using NIS elements. The mean fluorescent intensity of COX2 was measured in an enlarged nuclear mask in CD45 + cells. Fig. 1 Broad and in-depth tissue classification of NNC and MS human brain tissue. a Study sampling of three NNCs and seven MS human post-mortem brain tissues. b ROIs were defined by differences in neuropathology, assessed by PLP and HLA-DR reactivity. Within the MS tissues, broad tissue classes encompass a representative ROI for a lesion, adjacent lesion rim and distant peri-lesional tissue; scale bar: 5 mm. For in-depth quantification, ROIs were stratified by PLP reactivity ( A full, B partial and C absent/demyelinated), and HLA-DR density and morphology (1: ramified, low density and inactive lesion; 2: ramified, high density, mixed active/inactive lesion; 3: rounded, ameboid, high density and active lesion); scale bar: 50 µm c Immunohistochemical depiction of PLP and HLA-DR in MS tissues (color-coded) and annotated ROIs for broad and in-depth tissue classification; scale bar: 5 mm.

Broad and in-depth tissue classification of NNC and MS human brain tissue. a Study sampling of three NNCs and seven MS human post-mortem brain tissues. b ROIs were defined by differences in neuropathology, assessed by PLP and HLA-DR reactivity. Within the MS tissues, broad tissue classes encompass a representative ROI for a lesion, adjacent lesion rim and distant peri-lesional tissue; scale bar: 5 mm. For in-depth quantification, ROIs were stratified by PLP reactivity ( A full, B partial and C absent/demyelinated), and HLA-DR density and morphology (1: ramified, low density and inactive lesion; 2: ramified, high density, mixed active/inactive lesion; 3: rounded, ameboid, high density and active lesion); scale bar: 50 µm c Immunohistochemical depiction of PLP and HLA-DR in MS tissues (color-coded) and annotated ROIs for broad and in-depth tissue classification; scale bar: 5 mm.

Western blot

FF human brain tissue was cryo-sectioned and lesions were roughly excised by needle excision (yielding 50 µg tissue) [ 71 ]. Tissue was homogenized with a douncer on ice in 10 mM Tris (pH 7.4), 2 mM EDTA, 0.25 M sucrose, and a mix of protease inhibitors and stored at − 80 ºC until use. hiPSC microglia were washed, lysed on ice in RIPA buffer for 15 min, scraped from the plate, and lysates were stored at − 20 ºC until further use. Western blots were performed similarly to previous work [ 29 ]. In short, samples were further diluted in Laemmli Buffer (#161-0747; Bio-rad 4x), after which they were heated to 95 °C for 5 or 10 min, centrifuged, and applied on a 10% or 12% SDS-PAGE. After transfer to a nitrocellulose (cell lysate samples) or PVDF membrane (tissue homogenate samples), the blots were incubated for 1 h at room temperature in blocking solution intercept (TBS) blocking buffer (927-60001, LI-COR) diluted 1:1 in TBS containing 0.1% Tween20. Blots were incubated with EP2 (1:250; Abcam; ab124419), and EP4 (1:100, Santa Cruz Biotechnology; sc-55596) antibodies diluted in blocking solution overnight at 4 °C. After washing with TBS containing 0.1% Tween20, IRDye secondary antibodies were incubated for 1 h at room temperature. Blots were washed with TBS containing 0.1% and visualized by Azure Sapphire Biomolecular Imager for cell lysates and Odyssey Clx Imaging System for tissue homogenates. β-Actin (1:5000; Sigma-Aldrich; A5441) and α-tubulin (1:5000; Cedarlane; CLT9002) were used as loading control for sample normalization. Densitometric analysis was performed with ImageJ (version 1.49v) for cell lysates and Image Studio Lite software (version 5.2) for tissue homogenates. Raw blots are added in Additional File 1.

RNA isolation and real-time quantitative polymerase chain reaction (qRT-PCR)

After treatment, hiPSC microglia were washed with PBS, and RNA was extracted using TRIzol (#15596-018, Thermo Fisher Scientific), whereas RNA from human brain tissue blocks was extracted using the RNeasy Lipid Tissue Mini Kit (#174804, Qiagen). Total RNA quantity was evaluated using a Nanophotometer (Implen, Westlake Village, USA) and cDNA was synthesized using the High-Capacity cDNA Reverse Transcription Kit (#4368813, Thermo Fisher Scientific). Transcripts of interest were visualized and measured using SYBR Green (#4309155, Thermo Fisher Scientific) and the QuantStudio ™ 3 Real-Time PCR System (# A28567 , Thermo Fisher Scientific). Expression was measured in experimental and technical replicates and the 2 −  ΔΔ CT relative quantification method was applied to normalize the expression to the housekeeping genes Polrf2 (hiPSC microglia) and β-actin (human brain tissue block lysates). Primer sequences are summarized in Table 2 Table 2 Primer details Target Forward primer (5'–3') Reverse primer (5'–3') ACTB GGG AAA TCG TGC GTG ACA TTA AG TGT GTT GGC GTA AGG TCT TTG POLRF2 GAA CTC AAG GCC CGA AAG TGA TGA TGA GCT CGT CCA C PTGER2 TGC CTT TCA CGA TTT TTG CA ACG CAT TAG TCT CAG AAC AGG A PTGER4 TACTCATTGCCACCTCCCTGGT GACTTCTCGCTCCAAACTTGGC PTGS1 TCT TGC TGT TCC TGC TCC TG GTC ACA CTG GTA GCG GTC AA PTGS2 ACA GGC TTC CAT TGA CCA G TCA CCA TAG AGT GCT TCC AAC PTGES CAC CGG AAC GAC ATG GAG A TCC AGG CGA CAA AAG GGT TA PTGES2 AGC AAG CGA CTC AAG AGC AG GCC ATA CAG CGC CAA ATC AG PTGES3 TGC CCC GTT CAC AAT ACT CC CAG GCT GCT TTT CCA AAG ACA T

Table 2 Primer details Target Forward primer (5'–3') Reverse primer (5'–3') ACTB GGG AAA TCG TGC GTG ACA TTA AG TGT GTT GGC GTA AGG TCT TTG POLRF2 GAA CTC AAG GCC CGA AAG TGA TGA TGA GCT CGT CCA C PTGER2 TGC CTT TCA CGA TTT TTG CA ACG CAT TAG TCT CAG AAC AGG A PTGER4 TACTCATTGCCACCTCCCTGGT GACTTCTCGCTCCAAACTTGGC PTGS1 TCT TGC TGT TCC TGC TCC TG GTC ACA CTG GTA GCG GTC AA PTGS2 ACA GGC TTC CAT TGA CCA G TCA CCA TAG AGT GCT TCC AAC PTGES CAC CGG AAC GAC ATG GAG A TCC AGG CGA CAA AAG GGT TA PTGES2 AGC AAG CGA CTC AAG AGC AG GCC ATA CAG CGC CAA ATC AG PTGES3 TGC CCC GTT CAC AAT ACT CC CAG GCT GCT TTT CCA AAG ACA T

Primer details

RNA isolation and Quant Seq 3′mRNA-Seq

Lysates were thawed on ice, after which RNA was extracted using the RNeasy Plus Micro kit (#74034; Qiagen). Obtained RNA concentrations were measured using the Qubit® 3.0 Fluorometer (ThermoFisher), and the Quant Seq 3′ mRNA-Seq V2 Library Prep Kit FWD with Unique Dual Indices (#UDI12A_0001-0096; Lexogen) was used according to manufacturer’s instructions to prepare the sequencing libraries. Quality control was assessed using the 4200 TapeStation (Agilent Technologies). Sequencing libraries were pooled equimolarly, and sequenced was performed on a NextSeq 2000 (Illumina). Sequencing data were pre-processed by Lexogen Bioinformatics Services (Vienna, Austria) using their in-house pipeline, which includes quality control and adaptor trimming, alignment to the most recent human reference genome (homo_sapiens_GRCh38_ensembl_release_10 7_ERCC_SIR), and PCR deduplication trough unique molecule identifier (UMI) processing.

Differential expression analysis

Downstream bioinformatics was applied in R (v4.3.2) using RStudio (v2023.12.1) [ 73 , 74 ]. Data processing and normalization and subsequent differential expression (DE) analysis were executed using the edgeR (v4.0.16) and limma (v3.58.1) packages [ 7 , 65 ]. Principal component analysis (PCA) plots visualizing sample clustering across conditions were generated using ggplot2 (v3.5.1) [ 79 ]. Differentially expressed genes (DEGs) were based on an adjusted p-value of less than 0.05 and a Log|FC| greater than 1, which were depicted in Venn diagrams and volcano plots produced using the ggvenn (v0.1.10) and EnhancedVolcano (v1.20.0) packages [ 39 , 82 ]. Over-representation analysis (ORA) was used to identify Kyoto Encyclopedia of Genes and Genomes (KEGG) terms enriched based on the DEG lists, using the clusterProfiler (v.4.4.4) package [ 80 ].

Statistical analysis

Data are shown as boxplots displaying minimum to maximum values, with the middle line representing the median. All statistical tests (except RNA seq. data) were performed in GraphPad Prism v10.02 (GraphPad Software, La Jolla, USA) and based on at least three independent experiments. In the case of a sample size > 10, outliers were identified and removed using Grubb’s method ( α  = 0.05), which was only the case in MS lesion tissue block lysate RT-qPCR samples. Data normality was assessed using the Shapiro–Wilk test. In multiple (> 2) group comparisons, we used a two-tailed one-way analysis of variance (ANOVA) with Tukey’s/Dunnett’s (paired) multiple comparisons test. Friedman test (paired) or Kruskal–Wallis test was used for non-normally distributed data followed by Dunn’s post hoc analysis. For comparing two experimental groups, we used a two-tailed Student’s t-test or a Mann–Whitney test in the case of non-parametric data. To correlate lipid data with pathological hallmarks, we used Spearman’s rank correlation coefficient (non-parametric data). Analysis details are stated in figure legends, accompanied by corresponding p-values, of which p  < 0.05 (red) was considered statistically significant.

Results

Classifications of NNC and MS white matter brain tissue

To provide a spatial distribution map of AA and PGE 2 in MS lesions, FF human brain tissues from NNCs ( N  = 3) and MS lesions ( N  = 7) (Fig.  1 a) were used to stratify MS brain tissue, using PLP and HLA-DR immunoreactivity to reveal the level of myelination and neuroinflammation, respectively. Based on these histological markers, we created two separate classifications for the MS tissue to allow for a general (“ broad ”) and a more differentiated (“ in-depth ”) analysis of the LM profile in MS brain tissue. A paired analysis within these classifications was performed in the MS tissues to provide tissue comparisons. For the broad classification, three regions of interest (ROIs) per donor, corresponding to the lesion, lesion rim, and peri-lesion (distant from the lesion), were chosen based on one selected lesion within the tissue (Fig.  1 b,c). For the in-depth tissue classification, PLP reactivity was used to classify multiple areas into “A—fully myelinated (full)” (representing control brain tissue, normal-appearing white matter (NAWM), pre-lesion), “B—diffusely myelinated/partially demyelinated” (dirty-appearing WM (DAWM)), and “C—absent myelin/completely demyelinated” (MS lesions) (Fig.  1 b, c). In addition, HLA-DR reactivity was stratified by cell morphology and quantity (assessed visually) into “ramified, low density (1)”, “ramified, high density (2)” and “rounded, ameboid (3)”. For lesions specifically, the terms active (high density and ameboid), mixed active/inactive (medium density, rounded), and inactive (low density, ramified to rounded) were used based on previously used lesion classifications. The in-depth classification allows us to cluster multiple regions within the tissues and highlight potential differences based on both the level of myelination and inflammation. Selected ROIs per donor for broad and in-depth classification are depicted in Fig.  1 c.

Arachidonic acid levels decrease in MS lesions

AA is the key substrate for a plethora of LMs including prostaglandins, leukotrienes, and lipoxins through enzymatic and non-enzymatic reactions [ 16 ]. We performed silver-doped PA nano-DESI MSI for AA, where ion images show regional heterogeneity of AA levels in NNC tissues, but profoundly sharp outlined differences within MS lesion tissues (Fig.  2 a). In particular, MS lesion areas characterized by demyelination (exemplified by white arrows) reveal lower AA intensities compared to myelinated regions within the same tissue section (exemplified by red arrows) (Fig.  2 a). The mean detected AA concentrations per donor are consistently lower in MS compared to NNCs ( p  = 0.017) (Fig.  2 b). Moreover, paired analysis within the MS tissue sections revealed significantly lower AA concentrations in the lesion area compared to the peri-lesion area ( p  = 0.01) ( broad classes; Fig.  2 c), and between lesion rim and lesions ( p  = 0.02), whereas no differences in AA were observed between the lesion rim and peri-lesional areas (Fig.  2 c). In-depth classification showed that AA levels were significantly lower in demyelinated areas (absent) compared to both full and partial myelination ( p  = 0.049; p  = 0.049) ( in-depth classes; Fig.  2 d). The different inflammatory states (low, high, active) did not show changes in AA ( in-depth classes; Fig.  2 d). Essential fatty acids, including AA, are transported from the blood circulation across brain endothelial cells into the CNS [ 60 ]. We observed that the vascularized area, assessed by endothelial marker Ulex Europaeus Agglutinin I (UEA-I), was similar between full, partially demyelinated and demyelinated areas (Fig.  S1 a, b). Interestingly, detected AA concentrations correlate positively with the vascularized area in fully myelinated tissue ( p  = 0.017, r s  = 1) and trend to correlate with demyelinated areas ( p  = 0.103, r s  = 0.771), but not in partially demyelinated regions ( p  = 0.241, r s  = 0.6) (Fig.  S1 c). Fig. 2 Decreased AA levels in WM MS lesions. a Ion images of AA [M + 107Ag] + (m/z 411.1448 ± 5 ppm) normalized to internal standard AA-d8 in human NNC ( n  = 3) and MS brain tissues ( n  = 7). The spatial distribution of AA is visualized by a min–max intensity scale within each ion image; hence, the images are not related to each other. White arrows point out exemplary demyelinated/lesion areas; red arrows indicate myelinated/peri-lesional tissue. Pixel size is 25 µm × 150 µm; scale bar: 5 mm. b AA signal intensities from the entire tissue section were extracted and normalized to the internal standard AA-d8 to yield the average detected concentration per pixel (µM/pixel). Each dot represents the tissue of one donor. c Paired analysis of AA within MS tissues based on broad classification (one representative ROI per tissue area). Each colored dot represents the tissue of one donor. d Paired analysis of AA levels based on in-depth classification based on PLP or HLA-DR (averages of multiple ROIs per class). Each colored dot represents the tissue of one donor. e
mRNA expression of various biosynthesizing enzymes in human brain tissue block lysates of NNC ( N  = 8) and MS tissues ( N  = 12). Tissues used both in MSI and qPCR are marked in red. f Representative images of COX2, CD45 (immune cell marker), and Collagen IV (Coll IV, vascular marker) immunoreactivity in MS WM tissue; scale bar: 50 µm, zoom in: 25 µm. g Quantification of COX2 mean fluorescent intensity (MI) measured within the nuclei and percentage of COX2 + cells in NNC ( N  = 4) and MS lesions ( N  = 6). h Paired tissue analysis of COX2 + cells in MS tissues (HLA-DR in-depth ) and Spearman correlation ( r s ) of detected AA concentration with COX2 + cells (HLA-DR in-depth ). Data is shown as box plots with median ± quartiles; whiskers extend to minimum and maximum. Data have been statistically tested for three groups by Friedman test (paired) for non-normally distributed data and Dunn’s post hoc analysis. For two groups, an unpaired student t-test with Welch’s correction was used when the variance of the groups was significantly different or the Mann–Whitney test for non-parametric datasets. Exact p-values are reported and statistical significance is set at p  < 0.05 (red)

Decreased AA levels in WM MS lesions. a Ion images of AA [M + 107Ag] + (m/z 411.1448 ± 5 ppm) normalized to internal standard AA-d8 in human NNC ( n  = 3) and MS brain tissues ( n  = 7). The spatial distribution of AA is visualized by a min–max intensity scale within each ion image; hence, the images are not related to each other. White arrows point out exemplary demyelinated/lesion areas; red arrows indicate myelinated/peri-lesional tissue. Pixel size is 25 µm × 150 µm; scale bar: 5 mm. b AA signal intensities from the entire tissue section were extracted and normalized to the internal standard AA-d8 to yield the average detected concentration per pixel (µM/pixel). Each dot represents the tissue of one donor. c Paired analysis of AA within MS tissues based on broad classification (one representative ROI per tissue area). Each colored dot represents the tissue of one donor. d Paired analysis of AA levels based on in-depth classification based on PLP or HLA-DR (averages of multiple ROIs per class). Each colored dot represents the tissue of one donor. e
mRNA expression of various biosynthesizing enzymes in human brain tissue block lysates of NNC ( N  = 8) and MS tissues ( N  = 12). Tissues used both in MSI and qPCR are marked in red. f Representative images of COX2, CD45 (immune cell marker), and Collagen IV (Coll IV, vascular marker) immunoreactivity in MS WM tissue; scale bar: 50 µm, zoom in: 25 µm. g Quantification of COX2 mean fluorescent intensity (MI) measured within the nuclei and percentage of COX2 + cells in NNC ( N  = 4) and MS lesions ( N  = 6). h Paired tissue analysis of COX2 + cells in MS tissues (HLA-DR in-depth ) and Spearman correlation ( r s ) of detected AA concentration with COX2 + cells (HLA-DR in-depth ). Data is shown as box plots with median ± quartiles; whiskers extend to minimum and maximum. Data have been statistically tested for three groups by Friedman test (paired) for non-normally distributed data and Dunn’s post hoc analysis. For two groups, an unpaired student t-test with Welch’s correction was used when the variance of the groups was significantly different or the Mann–Whitney test for non-parametric datasets. Exact p-values are reported and statistical significance is set at p  < 0.05 (red)

Decreasing AA concentrations in MS tissue and MS lesions could point towards increased metabolism of AA-derived LMs [ 78 ]. Hence, we next quantified the expression of biosynthesizing enzymes like cyclooxygenases and prostaglandin synthases involved in the AA to prostaglandin conversion in human tissue block lysates by RT-qPCR (Fig.  2 e) [ 64 ]. Transcripts for PTGS2 (encoding for COX2) and PTGES (encoding for mPGES-1) were significantly higher in MS compared to NNCs ( p  = 0.031; p  = 0.001, respectively), and PTGES2 (encoding for mPGES-2) showed a similar trend ( p  = 0.09), whereas PTGS1 (encoding for COX1) and PTGES3 (encoding for cytosolic PGES) were not differently expressed between MS and NNCs (Fig.  2 e). Next, we visualized the rate-limiting enzyme for PGE 2 biosynthesis, COX2, by immunohistochemistry (IHC) in consecutive brain slices to the MSI-imaged tissue sections [ 45 ]. We found COX2 immunoreactivity predominantly in larger vessels (panel 1, Fig.  2 f) but predominantly around numerous nuclei expressing CD45 (panel 2), which we confirmed using confocal imaging (Fig.  S1 e), both in NNC and MS cases (Fig.  2 f). Sporadically, we observed COX2 expression in CD45 + cells within the perivascular space in MS cases (Fig.  S1 d) [ 66 , 81 ]. However, quantification of COX2 mean intensity (MI) perinuclear and percentage of COX2 + cells (of total cells) did not differ between NNC and MS (Fig.  2 g). Moreover, within MS tissues, the number of COX2 + cells did not alter between the different inflammatory states. However, at both high and active states of HLA-DR + cells, COX2 + cells tend to positively correlate with the AA concentration in the same region ( r s  = 0.9, p  = 0.083; r s  = 0.8, p = 0.133 for the in-depth classes respectively) (Fig.  2 h). Collectively, these findings point towards an overall decrease of AA concentration in MS WM brain tissues compared to NNCs, including an MS lesion-specific decrease compared to surrounding tissue. Together with the (partial) increase in biosynthesizing enzymes, this might suggest a local activation of AA metabolism within MS brain tissue.

Prostaglandin E 2 /AA ratio increases in MS lesions

The observed increased expression of some prostaglandin synthesizing enzymes, particularly PTGES, in MS tissues suggests an increased prostaglandin biosynthesis. To verify this hypothesis, we next focused on the distribution of the bioactive LM prostaglandin E 2 (PGE 2 ) using PA nano-DESI MSI and report both the detected PGE 2 concentrations as well as the PGE 2 /AA ratio (Fig.  3 a). Similar to the spatial distribution of AA, PGE 2 delineated lesion areas within the MS tissues, while being quite homogenously distributed in the NNC tissues (Fig.  3 a). Interestingly, some MS lesions (white arrow) showed reduced PGE 2 levels compared to peri-lesional tissue (red arrow), while others showed a distinct increase of PGE 2 in the lesion (Fig.  3 a, MS46 and MS60 tissue, respectively). The PGE 2 /AA concentration ratio uniquely indicates the PGE 2 concentrations relative to the availability of AA substrate [ 8 , 40 ]. PGE 2 /AA ratio images reveal that increased relative PGE 2 was generally restricted to MS lesions and partially to the adjacent tissue, except for MS68, where lesions showed lower PGE 2 /AA ratio compared to the remaining tissue (Fig.  3 a, comparable scale between images). In general, whole-tissue PGE 2 levels were lower in MS compared to NNCs (Fig.  3 b). Fig.3 PGE 2 /AA levels are increased in demyelinated WM MS tissue. a Ion images of PGE 2 (first row) and PGE 2 /AA concentration ratio images (second row). PGE 2 ion images were constructed by visualizing m/z 459.1301 ± 5 ppm ([M +  107 Ag] + ) normalized pixel-by-pixel to internal standard PGE 2 -d9 in human NNC ( n  = 3) and MS lesions ( n  = 7). The spatial distribution of PGE 2 is visualized by a min–max intensity scale within each single ion image; hence, the images are not related to each other. In contrast, all the PGE 2 /AA concentration ratio images are fitted to the same scale and can be compared. White arrows point out exemplary lesion areas; red arrows indicate peri-lesional tissue with full myelination; scale bar: 5 mm. b PGE 2 signal intensities were extracted as average pixel intensity per tissue and normalized to the internal standard of PGE 2 to yield average detected concentration per pixel (µM/pixel). Each dot presents one donor. c Paired analysis of PGE 2 concentration and PGE 2 /AA concentration ratio based on broad classification (one representative ROI) within MS tissues. Each colored dot represents the tissue of one donor. d , e Paired analysis of PGE 2 and PGE 2 /AA ratio based on PLP or HLA-DR classification ( in-depth classes). Each colored dot represents the tissue of one donor. f Representative images of HLA-DR immunoreactivity in MS WM tissue; scale bar: 50 µm, zoom in: 10 µm. g Spearman correlation ( r s ) of PGE 2 /AA ratio values with HLA-DR MI in present, partial, and absent PLP tissue categories. Data is shown as box plots with median ± quartiles; whiskers extend to minimum and maximum. Data have been statistically tested for three groups by Friedman test (paired) for non-normally distributed data and Dunn’s post-hoc analysis. Exact p-values are reported and statistical significance is set at p  < 0.05 (red)

PGE 2 /AA levels are increased in demyelinated WM MS tissue. a Ion images of PGE 2 (first row) and PGE 2 /AA concentration ratio images (second row). PGE 2 ion images were constructed by visualizing m/z 459.1301 ± 5 ppm ([M +  107 Ag] + ) normalized pixel-by-pixel to internal standard PGE 2 -d9 in human NNC ( n  = 3) and MS lesions ( n  = 7). The spatial distribution of PGE 2 is visualized by a min–max intensity scale within each single ion image; hence, the images are not related to each other. In contrast, all the PGE 2 /AA concentration ratio images are fitted to the same scale and can be compared. White arrows point out exemplary lesion areas; red arrows indicate peri-lesional tissue with full myelination; scale bar: 5 mm. b PGE 2 signal intensities were extracted as average pixel intensity per tissue and normalized to the internal standard of PGE 2 to yield average detected concentration per pixel (µM/pixel). Each dot presents one donor. c Paired analysis of PGE 2 concentration and PGE 2 /AA concentration ratio based on broad classification (one representative ROI) within MS tissues. Each colored dot represents the tissue of one donor. d , e Paired analysis of PGE 2 and PGE 2 /AA ratio based on PLP or HLA-DR classification ( in-depth classes). Each colored dot represents the tissue of one donor. f Representative images of HLA-DR immunoreactivity in MS WM tissue; scale bar: 50 µm, zoom in: 10 µm. g Spearman correlation ( r s ) of PGE 2 /AA ratio values with HLA-DR MI in present, partial, and absent PLP tissue categories. Data is shown as box plots with median ± quartiles; whiskers extend to minimum and maximum. Data have been statistically tested for three groups by Friedman test (paired) for non-normally distributed data and Dunn’s post-hoc analysis. Exact p-values are reported and statistical significance is set at p  < 0.05 (red)

Paired analysis of PGE 2 concentrations within the MS tissue varied widely and did not follow a pattern within the immunohistochemistry-derived classifications of broad or in-depth (Fig.  3 c–e, first graph, respectively). However, the PGE 2 /AA ratio was increased in the lesion tissue compared to peri-lesional tissue ( p  = 0.01) in the broad tissue classification (Fig.  3 c) and also significantly higher ( p  = 0.023) in demyelinated regions compared to fully myelinated regions ( in-depth classes; Fig.  3 d). No difference was observed between partially demyelinated and fully demyelinated areas ( in-depth classes; Fig.  3 d). Categorized by HLA-DR, PGE 2 /AA concentration ratio did not significantly differ between the groups ( in-depth classes; Fig.  3 e).

We observed that the PGE 2 /AA ratio pattern partly co-localized with the HLA-DR reactivity in the same tissue (Fig.  S2 ). Hence, we quantified the MI of HLA-DR (Fig.  3 f) and correlated it with the PGE 2 /AA ratio within the PLP classes (Fig.  3 g). None of the correlations reached significance (partial r s  = − 0.143, p  = 0.783; absent: r s  = − 0.214, p  = 0.662) and PGE 2 /AA ratios in full myelinated regions showed a negative trend with immune cell activation ( r s  = − 0.714, p  = 0.136). Together, these findings suggest that the PGE 2 /AA ratio increases in MS lesions but no clear association between PGE 2 /AA levels and immune cell activation (MI HLA-DR) could be made.

Enhanced microglial EP2 expression in MS lesions

PGE 2 signaling following its local production occurs through four GPCR receptors (EP1-4), of which EP2 and EP4 have been previously related to neurological diseases and neuroinflammation (Fig.  4 a) [ 2 , 30 ]. To investigate the putative local PGE 2 -induced signaling pathway in MS tissues, we next analyzed protein levels of EP2 and EP4 in human brain tissue homogenates of NNCs and MS lesions. Receptor levels of EP4 were significantly lower in MS lesions ( p  = 0.0018) compared to NNCs, whereas EP2 levels were increased in MS lesions compared to NNCs ( p  < 0.0001) (Fig.  4 b, c). IHC analysis revealed a co-localization of EP2 with microglia (defined as Iba1 + TMEM119 + cells), and to a far lesser extent with the vasculature (Fig.  4 d). The percentage of EP2 + microglia of all cells as well as the percentage of EP2 + microglia from all Iba1 + TMEM119 + cells was significantly higher in the lesion (mixed A/I) compared to NNC ( p  = 0.005; p  = 0.015). Microglial EP2 expression (MI) was higher in the lesion ( p  = 0.029) and showed a positive trend in the peri-lesional tissue ( p  = 0.058) compared to NNC (Fig.  4 e). Together, MS WM lesions show a transition towards overall decreased EP4 receptor expression and an enhanced EP2 receptor expression, the latter being specifically expressed by microglia. Fig. 4 Increased microglial EP2 expression in MS lesions. a Schematic overview of PGE 2 receptors EP1-4 and their G-protein coupled signaling pathways. EP2 (blue) and EP4 (yellow) are highlighted. b Representative cropped immunoblots of EP2, EP4 (upper panel each), and β-actin (lower panel) from human brain tissue homogenates of NNC and MS lesions. c Densitometric quantification of EP2 and EP4, normalized to β-actin in NNC ( N  = 9) and MS lesions ( N  = 14). d Representative images of EP2 (cyan), Iba1 (magenta) and TMEM119 (yellow) immunoreactivity in WM of NNC and mixed A/I MS lesions (peri-lesion and lesion). Panels show outlined excerpts at higher magnification; scale bar: 50 µm. e Quantification of EP2 + cells and EP2 mean fluorescent intensity within microglia (Iba1 + TMEM119 + cells) in NNC and MS tissues ( N  = 5). f
mRNA expression of PTGER2 (encoding EP2) and EP2 protein were measured in human iPSC-derived microglia (hiPSC microglia) non-stimulated (resting) or stimulated with LPS + IFNγ for 24 h (pro-inflam). Data is shown as box plots with median ± quartiles; whiskers extend to minimum and maximum. Data have been statistically tested for three groups by ordinary one-way ANOVA with Dunnett’s correction or Kruskal–Wallis test for non-normally distributed data and Dunn’s post hoc analysis. For two groups, an unpaired student t-test with Welch’s correction was used when the variance of the groups was significantly different or the Mann–Whitney test for non-parametric datasets. Exact p-values are reported and statistical significance is set at p  < 0.05 (red)

Increased microglial EP2 expression in MS lesions. a Schematic overview of PGE 2 receptors EP1-4 and their G-protein coupled signaling pathways. EP2 (blue) and EP4 (yellow) are highlighted. b Representative cropped immunoblots of EP2, EP4 (upper panel each), and β-actin (lower panel) from human brain tissue homogenates of NNC and MS lesions. c Densitometric quantification of EP2 and EP4, normalized to β-actin in NNC ( N  = 9) and MS lesions ( N  = 14). d Representative images of EP2 (cyan), Iba1 (magenta) and TMEM119 (yellow) immunoreactivity in WM of NNC and mixed A/I MS lesions (peri-lesion and lesion). Panels show outlined excerpts at higher magnification; scale bar: 50 µm. e Quantification of EP2 + cells and EP2 mean fluorescent intensity within microglia (Iba1 + TMEM119 + cells) in NNC and MS tissues ( N  = 5). f
mRNA expression of PTGER2 (encoding EP2) and EP2 protein were measured in human iPSC-derived microglia (hiPSC microglia) non-stimulated (resting) or stimulated with LPS + IFNγ for 24 h (pro-inflam). Data is shown as box plots with median ± quartiles; whiskers extend to minimum and maximum. Data have been statistically tested for three groups by ordinary one-way ANOVA with Dunnett’s correction or Kruskal–Wallis test for non-normally distributed data and Dunn’s post hoc analysis. For two groups, an unpaired student t-test with Welch’s correction was used when the variance of the groups was significantly different or the Mann–Whitney test for non-parametric datasets. Exact p-values are reported and statistical significance is set at p  < 0.05 (red)

To mimic microglial PGE2 signaling in MS lesions in vitro, we cultured human iPSC-derived microglia (hiPSC microglia) under resting and pro-inflammatory conditions (LPS/IFNγ for 24 h). Pro-inflammatory microglia showed higher expression of PTGER2 /EP2 on RNA and protein levels ( p  = 0.054; p  = 0.001, respectively) when compared to the non-stimulated control (resting) (Fig.  4 f). PTGER4 /EP4 showed an upward trend in pro-inflammatory microglia compared to resting but no clear increase on the protein level (Fig.  S3 a, b), together suggesting that EP2 signaling might play a more dominant role under inflammatory conditions.

EP2 signaling regulates both PGE 2 -related activation and resolution of microglia

To study the consequences of PGE 2 signaling in microglia in-depth, we next treated both resting and pro-inflammatory hiPSC microglia with PGE 2 (1 µM) or vehicle (0.1% ethanol) in combination with a selective EP2 or EP4 inhibitor (EP2i/EP4i) (Fig.  5 a). Using bulk RNA sequencing, we identified a total of 12,111 genes and PGE 2 treatment induced upregulation of 1201 genes in resting microglia and 1170 in pro-inflammatory microglia respectively, of which 628 genes were shared between the conditions (Fig.  5 b). 703 genes were downregulated in resting and 870 in pro-inflammatory microglia, with 446 genes shared between resting and pro-inflammatory microglia (Fig.  5 b) (Table  S1 ). PGE 2 addition to resting microglia induced differentially expressed genes (DEGs), including immunity-related genes ( CD86, MRC-1 (CD206) and IL7R ), ischemia-modulated genes ( HIF1A , SRGN ) and transcription-related genes ( RNAseI, ETS-2, CREM ) (Fig.  5 c) (Table  S1 ). Interestingly, EP2 but not EP4 inhibitors had a drastic effect on PGE 2 -treated resting microglia (Fig.  S4 a). Of note, PGE 2 addition to pro-inflammatory microglia also enhanced the expression of immunity-related genes, including CD86, CD40 , and members of the leukocyte immunoglobulin-like receptor (LIR) family (Fig.  5 d) (Table  S1 ). In addition, guanylate-binding proteins (GBP) family members 1,4,5, which play a central role in protective immunity against pathogens are upregulated by PGE 2 (Fig.  5 d) . In line with this, PGE 2 addition downregulated osteopontin ( SPP1 ), matrix metalloprotease 7 ( MMP7 ) and 24-dehydrocholesterol reductase ( DHCR24 ) expression, indicating initiation of a pro-resolving phagocytic phenotype (Fig.  5 d) (Table  S1 ). Fig. 5 PGE 2 signaling adds to immunological response in microglia. a Schematic overview of PGE 2 and EP inhibitor treatment strategy on resting and pro-inflammatory hiPSC microglia. b Venn diagram of significantly differentially expressed genes (DEG) presenting upregulated (top) and downregulated genes (bottom) in resting (green) vs. pro-inflammatory (blue) microglia treated with PGE 2 vs. vehicle; N TR  = 5. c Volcano plot representing significant DEGs comparing PGE 2 treatment versus vehicle in resting microglia with respect to -log 10 p adjusted in the y -axis and log 2 fold change in the x -axis. d Volcano plot representing significant DEGs comparing PGE 2 treatment vs. vehicle in pro-inflammatory microglia. e Volcano plot visualizing DEGs regulated by EP2i + PGE 2 vs EP4i + PGE 2 in pro-inflammatory microglia. f mRNA expression of CXCR4 , PTGES , TREM2, CCL13, and PAK1 in pro-inflammatory microglia treated with PGE 2  + EP inhibitors; N  = 3. g Over-representation analysis of enriched KEGG pathways in pro-inflammatory iPSC microglia treated with PGE 2  + EP2i vs. PGE 2  + EP4i. Data is shown as box plots with median ± quartiles; whiskers extend to minimum and maximum. Data have been statistically tested for four groups by Friedman test for non-normally distributed data and Dunn’s post hoc analysis. Exact p-values are reported and statistical significance is set at p  < 0.05 (red)

PGE 2 signaling adds to immunological response in microglia. a Schematic overview of PGE 2 and EP inhibitor treatment strategy on resting and pro-inflammatory hiPSC microglia. b Venn diagram of significantly differentially expressed genes (DEG) presenting upregulated (top) and downregulated genes (bottom) in resting (green) vs. pro-inflammatory (blue) microglia treated with PGE 2 vs. vehicle; N TR  = 5. c Volcano plot representing significant DEGs comparing PGE 2 treatment versus vehicle in resting microglia with respect to -log 10 p adjusted in the y -axis and log 2 fold change in the x -axis. d Volcano plot representing significant DEGs comparing PGE 2 treatment vs. vehicle in pro-inflammatory microglia. e Volcano plot visualizing DEGs regulated by EP2i + PGE 2 vs EP4i + PGE 2 in pro-inflammatory microglia. f mRNA expression of CXCR4 , PTGES , TREM2, CCL13, and PAK1 in pro-inflammatory microglia treated with PGE 2  + EP inhibitors; N  = 3. g Over-representation analysis of enriched KEGG pathways in pro-inflammatory iPSC microglia treated with PGE 2  + EP2i vs. PGE 2  + EP4i. Data is shown as box plots with median ± quartiles; whiskers extend to minimum and maximum. Data have been statistically tested for four groups by Friedman test for non-normally distributed data and Dunn’s post hoc analysis. Exact p-values are reported and statistical significance is set at p  < 0.05 (red)

Both EP2i and EP4i conditions separate clearly in pro-inflammatory microglia from the PGE 2 and vehicle conditions (Fig.  S4 b). DEGs between PGE 2  + EP2i treatment compared to PGE 2  + EP4i in pro-inflammatory microglia are visualized in the volcano plot (Fig.  5 e) (Table  S2 ). Compared to EP2i-treated cells, EP4i treatment induced expression of homeostatic/pro-resolving markers ( IGF1, TREM2 ), cytokine signaling ( CCL13 ), nuclear signaling and cytoskeleton rearrangement ( PAK1, TUBA1B, HCLS1, SRSF3 ) genes (Fig.  5 e) (Table  S2 ). EP2i promoted higher expression of genes related to chemokine-receptor signaling ( cxcr4 ) as well as PGE 2 synthase ( PTGES ) compared to EP4i (Fig.  5 e) (Table  S2 ). Validation experiments in pro-inflammatory microglia showed EP2i enhanced PGE 2 -induced CXCR4 ( p  = 0.049) and PTGES ( p  = 0.067) mRNA expression (Fig.  5 f). EP2i further reduced PAK1 ( p  = 0.049), CCL13 , and TREM2 expressions ; however, the latter two did not reach significance (Fig.  5 f). Of note, comparing the EP2 + versus EP2 − cells within the Iba1 + population in MS lesion IHC revealed an increased intensity of the homeostatic microglia marker TMEM119 in EP2 + cells ( p  = 0.0699) (Fig.  S5 ). In pro-inflammatory microglia, pathways over-representation analysis (KEGG pathways) showed a lower expression of genes associated with cell cycle pathways and some cytokine–receptor interactions when comparing EP2i to EP4i, but also a higher expression of genes in this pathway category (Fig.  5 g) (Table  S3 ). Taken together, these findings suggest that PGE 2 might stimulate an inflammatory response in homeostatic microglia, but that it also guides activated microglia towards inflammation resolution. A clear differentiation between EP2 and EP4 signaling is challenging but our findings suggest that PGE 2 -EP2 signaling could be responsible for a more resolving signature under inflammatory conditions.

Discussion

Chronic neuroinflammation in the CNS of PwMS presents a major challenge for therapeutic intervention, routed in the lack of understanding the underlying dysregulated inflammatory/resolving pathways. In our study, we depict a distribution map of the lipids AA and PGE 2 in WM MS brain lesions and highlight the role of PGE 2 -EP2 signaling in activated microglia at the lesion site. Specifically, we show a reduction of AA in MS brain tissues compared to NNCs and a distinct loss within MS lesions compared to surrounding tissue. PTGS2 and PTGES expression increased in MS tissue lysates and we report a higher PGE 2 /AA ratio in demyelinated tissue areas in MS. Within the same tissue location, PGE 2 receptor EP2 was upregulated in microglia, resulting in overall higher levels of EP2 in MS lesions compared to NNCs, whereas EP4 decreased in MS lesions. Finally, we focus on the genes modulated by PGE 2 signaling in resting and activated microglia and investigate specific pathways evoked through PGE 2 -EP2 and/or PGE 2 -EP4 signaling.

Lipid brain constituents assessed by lipidomics on brain tissue homogenates is a useful tool for defining pathology-induced metabolic differences; however, it lacks the spatial element to pinpoint lipid changes relative to its environment [ 36 ]. Here, we used MSI on FF human WM brain tissue slides from NNC and PwMS to explore spatial lipid distribution within the CNS. With this technique, we created a precise map of AA and PGE 2 abundances in and around MS lesions and linked them to pathological hallmarks of MS such as demyelination and immune cell activation. We found that AA concentrations were significantly reduced in MS tissues compared to NNCs and for both AA and PGE 2 , we observed very distinct patterns of alterations in and around the MS lesions. In a recent mouse study, the lipid profile of lysolecithin-induced focal demyelinated lesion homogenates was tracked [ 57 ]. 14 days after injection, free AA and PGE 2 levels increased. As AA release from the membrane is stimulated by inflammation, free AA levels could initially increase (as seen after 14 days in the animal model) but later decrease due to an increased AA conversion into downstream bioactive LMs such as prostaglandins and leukotrienes [ 22 ]. Data on other AA-derived LMs such as hydroxy-eicosatetraenoic acids and leukotrienes in our data set further support this hypothesis (data not shown). Likely, reduced storage and enhanced metabolism synergistically create the overall and lesion-specific AA reduction observed in MS. Importantly, the cases used in our study do not equally represent both sexes, particularly in the MS cohort of the MSI experiments, which we acknowledge as a limitation of the study.

In MS tissues, we observed an increase of prostaglandin-converting enzymes PTGS2 (COX2) and PTGES (mPGES-1) mRNA expression which further supports the hypothesis of locally increased PGE 2 biosynthesis. Among the three prostaglandin isozymes, mPGES-1 is functionally coupled to COX2 and increases concomitantly with COX2 in response to various stimuli [ 68 ]. For example, COX2 and m-PGES-1 are induced in the MS mouse model experimental autoimmune encephalomyelitis (EAE) and in MS lesions [ 34 ]. Further, we report COX2 expression in the vasculature and perinuclear in CD45 + cells but observed it less in the cytoplasm. These observations are in line with previous reports in primary human cells [ 43 , 56 ]. The latter observation confirmed other studies on COX2 in macrophages and microglia of MS lesions, particularly in the process of demyelination [ 66 , 83 ]. Of note, despite increased mRNA expression of PTGS2 (COX2) in MS, we did not observe overall higher levels of COX2 immunoreactivity in our MS tissues compared to NNCs and also found COX2 + cells in our control tissues counter to a previous study [ 66 ]. Local heightened COX2 levels upon active lipid metabolism may stay undetected in the general comparison taken and are, therefore, still plausible.

The source of PGE 2 is proposed to be a production by inflamed endothelial cells, dying oligodendrocytes, activated microglia/macrophages as well as direct transport from the blood [ 25 , 38 , 50 ]. Our ion images show that PGE 2 /AA ratios are highest in demyelinated lesion areas where axonal damage and microglia activation take place [ 3 , 33 ]. Similarly, CNS PGE 2 detected concentration was found to be increased at the peak of the disease in EAE and cuprizone-fed mice, rodent models for MS representing the inflammatory and demyelinating aspects respectively [ 34 , 55 ]. Inhibition of PGE 2 production by COX2 inhibition in EAE and COX2 KO mice in the cuprizone model showed reduced clinical scores and less oligodendrocyte death [ 10 , 16 ]. Despite positive results in mouse models, COX2 inhibition in humans is connected to increased risk for cardiovascular events and (hemorrhagic) stroke [ 47 , 75 ]. Consequently, a more specific target, such as one/multiple of the four PGE 2 receptors, would be desirable to largely circumvent the impairment of crucial cellular functions for body homeostasis.

PGE 2 exerts beneficial and detrimental immunomodulatory functions by signaling through EP2 and EP4; however, the specific pathways initiated in the CNS of PwMS remain unclear. Our data show a decrease in EP4 expression in MS lesions, which contrasts with findings in EAE mouse models where EP4 is upregulated in spinal cord lysates and blood [ 34 , 69 ]. However, EP4 expression is also reduced in blood cells from MS patients [ 58 ]. Previous studies have indicated species-specific differences, with humans exhibiting higher EP4 expression in microglia compared to mice [ 24 ]. These discrepancies suggest that EP4 regulation may vary between species and is influenced by the pathological context (MS versus EAE). In our study, we report a significant increase of EP2 in the homogenates of MS tissues compared to NNCs. EP2 expression was predominantly observed in microglia at the MS lesion site, which together with the increased PGE 2 /AA ratio at this site, suggests enhanced PGE 2 -EP2 signaling in MS lesions. Induced EP2 and EP4 expression have been shown in vivo in the CNS of EAE and cuprizone mouse models, where EP2 increases prior to EP4 upregulation [ 15 , 34 , 55 ]. Previous studies revealed that HIF-1a, induced by high cell density, downregulates EP4 expression, but not EP2 [ 54 , 70 ]. As HIF-1a is also expressed in demyelinating WM MS lesions, it may regulate the reduced EP4 expression in MS compared to NNCs [ 1 , 18 ]. In addition to the shift of expression, EP2 receptors are to a lesser extent desensitized by PGE 2 binding than EP4, plus EP2 also respond (gradually less) to the downstream metabolite 15-keto-PGE 2 further strengthening the hypothesis of enhanced PGE 2 -EP2 signaling [ 51 ]. Importantly, we were not able to assess the cell-specific expression of EP4 in our brain tissue due to a lack of appropriate EP4 antibodies; hence, we can only make conclusions on its expression relative to NNC tissues but not cell-specific. In conclusion, our findings suggest enhanced PGE 2 –EP2 signaling in MS compared to NNC, mostly restricted to inflamed microglia in MS lesions.

Inhibiting EP2 or EP4 activity has been promising in diverse disease models for inflammation [ 14 ], demyelination [ 56 ], cognition [ 48 ], epileptic seizures [ 49 ], and Alzheimer’s disease pathology [ 23 ]. Specifically, during EAE, a T-cell-driven MS disease mouse model, EP4 inhibition during the immunization phase was effective in reducing the clinical score, while EP2-deficient mice did not show any beneficial effect unless used together with an EP4 antagonist [ 14 ]. On the contrary, EP2 antagonism was effective in reducing demyelination in a cuprizone model by preventing oligodendrocyte apoptosis and in LPS-injected (i.p.) mice by reducing microgliosis [ 28 , 56 ]. Hence, the development of a safe BBB-penetrating EP2 antagonist for clinical exploration is ongoing, and EP4 antagonists currently undergo clinical trials for cancer immunotherapy [ 17 , 26 , 44 ]. In light of our observed enhanced PGE 2 –EP2 signaling in MS, EP2 targeting might be the most interesting therapeutic candidate; however, effects are likely highly dependent on cell state and tissue microenvironment.

Focusing on the specific effect in the CNS, we lastly aimed to shed light on the bimodal nature of PGE 2 signaling in both resting and activated microglia using RNA sequencing. Primarily, we found that PGE 2 treatment polarizes resting microglia toward an inflammatory response and guides activated microglia toward resolution and a regulated immune response. Inhibiting EP2 signaling in resting microglia suppressed part of the PGE 2 -induced inflammation signature, whereas EP4 inhibition was less effective. Pro-inflammatory effects initiated by LPS/IFNγ treatment were dampened both through PGE 2 -EP2, but mainly PGE 2 -EP4 inhibition. Of note, EP2 blockage in activated microglia resulted in a trend towards downregulation of TREM2 , a key player in the maintenance of microglia metabolism and response to phagocytic triggers, and an upregulation of PTGES , potentially exacerbating further PGE 2 synthesis [ 72 ]. Contextualizing these findings to demyelinating lesion tissue of MS patients, in which we found increased microglial EP2 expression in the lesion and higher TMEM119 intensity in EP2 + cells, the enhanced EP2 signaling could be a response mechanism to counteract demyelination and induce a potential resolving microglial phenotype. Thus, although the underlying signaling remains to be validated, PGE 2 -EP2 signaling might also have crucial beneficial effects in MS lesions.

To conclude, in the current research, we provide a spatial distribution of AA and its derivative PGE 2 in WM MS lesions, which we coupled to enhanced PGE 2 -EP2 signaling. This could potentially be a chronic resolution response dampening neuroinflammation, counteracting the primary pro-inflammatory effects of PGE 2 in MS.

Supplementary Information

Below is the link to the electronic supplementary material. Supplementary file1 (XLSX 464 kb) Supplementary file2 (XLSX 154 kb) Supplementary file3 (XLSX 10 kb) Supplementary file4 (DOCX 37 kb) Supplementary file5 (TIF 6774 kb) Supplementary file6 (DOCX 386 kb) Supplementary file7 (DOCX 28 kb) Supplementary file8 (DOCX 25 kb) Supplementary file9 (TIF 13114 kb) Supplementary file10 (TIF 518 kb) Supplementary file11 (TIF 3477 kb) Supplementary file12 (TIF 1692 kb) Supplementary file13 (TIF 1745 kb)

Supplementary file1 (XLSX 464 kb)

Supplementary file2 (XLSX 154 kb)

Supplementary file3 (XLSX 10 kb)

Supplementary file4 (DOCX 37 kb)

Supplementary file5 (TIF 6774 kb)

Supplementary file6 (DOCX 386 kb)

Supplementary file7 (DOCX 28 kb)

Supplementary file8 (DOCX 25 kb)

Supplementary file9 (TIF 13114 kb)

Supplementary file10 (TIF 518 kb)

Supplementary file11 (TIF 3477 kb)

Supplementary file12 (TIF 1692 kb)

Supplementary file13 (TIF 1745 kb)