Transient middle cerebral occlusion (MCAO) was performed as described previously [33]. In brief, mice were anesthetized with 1.5% isoflurane (Forene, Abbott, Wiesbaden, Germany) and received 0.1 mg kg ${}^{-1}$ buprenorphine (Temgesic, Essex Pharma, Munich, Germany) under spontaneous respiration. The focal ischemia was induced by inserting a standardized silicon-coated monofilament with a tip diameter of 0.23 mm (Doccol, Redlands, CA, USA). A midline cervical incision was followed by the introduction of the monofilament along the internal carotid artery until it occluded the ostium of the right middle cerebral artery. The reperfusion was initiated by withdrawing the monofilament after 3 hours of focal cerebral ischemia. After the surgery, mice were monitored and received food and regular drinking water ad libitum. Animals were assessed either 6 hours or 24 hours after ischemia onset, in order to describe their global neurological functions before being euthanized.

114 mice were subjected to 3 hours of MCAO followed by immediate imaging and reperfusion. Reperfusion was performed either 6 hours (n = 60) or 24 hours (n = 54) after ischemia onset. Within the 6-hour cohort, 22 mice were treated with warfarin and 38 without, whilst 15 mice received warfarin within the 24-hour cohort (Supplementary Fig. 1). 10 mice were used as sham-operated control. The operations were performed in an unblinded fashion since the operator did not apply any modifications such as drug treatment, however, sample provision for the mass spectrometry or 2,3,5-triphenyltetrazolium chloride (TTC) imaging were done in a blinded fashion.

To increase the chance of spontaneous HT following ischemic stroke (IS) induced by MCAO, we evaluated mice under warfarin treatment at the onset of ischemia versus non-anticoagulated mice. Accordingly, we administered warfarin through drinking water dissolving a 5 mg Coumadin tablet (warfarin sodium crystalline; Bristol Myers Squibb, Munich, Germany) in 375 mL tap water in accordance with previous reports [34]. Assuming a water consumption in rodents of 15 mL/100 g and a body weight of 20 g per 24 hours, a warfarin uptake of 0.033 mg (0.83 mg kg ${}^{-1}$ ) per mouse would be achieved within a 20-hour feeding period. Warfarin administration through bottled water was started at the same time. Following 20-hours of feeding ( $\pm$ Warfarin), MCAO was induced.

The neurological examination was performed 6 hours and 24 hours after ischemia onset, respectively. The modified Neurological Severity Score (mNSS) was applied to assess neurological deficits [35]. The mNSS is one of the most frequently referred to neurological scores for the functional assessment of mice after an induced stroke. A scoring system is used to quantify neurological deficits. Hemiparesis, walking behavior, coordination, and sensory function are the main areas assessed. The maximum score for mice is 18 points. The Bederson score was originally developed as an evaluation tool for the success of MCAO in the rat [36]. The neurological deficit is assessed using a 5-point scale and captures behavioral changes in the mouse based on its spontaneous movements. The grip test is used to evaluate motor function as well as coordination. The mice were placed on a wooden bar 30 cm above the ground and the time period to fall off was assessed.

The quantification of sphingoid bases and ceramides in tissue from the peri-infarct cortex collected 6 or 24 hours after ischemia onset as described in detail elsewhere [33, 37]. Briefly, samples of the peri-infarct cortex were snap-frozen with liquid nitrogen and stored at –80 °C until required for further analyses. The tissue samples were first mixed with water:ethanol (75:25, v/v) and homogenized to a suspension of 0.05 mg/ $\mu$ L tissue, using zirconium oxide grinding balls and a Mixer Mill MM400 (Retsch, Haan, Germany). For lipid extraction a volume of 20 $\mu$ L of the homogenate was used and mixed with 200 $\mu$ L extraction buffer (citric acid 30 mM, disodium hydrogen phosphate 40 mM, pH 4.2) and 20 $\mu$ L of the internal standard solution. The mixture was extracted once using 600 $\mu$ L methanol:chloroform:hydrochloric acid (15:83:2, v/v/v). Afterwards, the organic phase was divided into two aliquots (one for the determination of sphingoid bases and one for the determination of ceramides), evaporated and reconstituted using 50 $\mu$ L of tetrahydrofuran:water (9:1, v/v) containing 0.2% formic acid and 10 mM ammonium formate (ceramides) and 50 $\mu$ L methanol containing 5% formic acid (sphingolipids). The LC-MS/MS system consisted of a triple quadrupole mass spectrometer QTRAP 5500 (Sciex, Darmstadt, Germany) equipped with a Turbo Ion Spray source operated in positive electrospray ionization mode and an Agilent 1290 Infinity LC-system with binary HPLC pump, column oven and autosampler (Agilent, Waldbronn, Germany). Chromatographic separation of ceramides was achieved using a Zorbax RRHD Eclipse Plus C18 column (1.8 $\mu$ m 50 $\times$ 2.1 mm ID, Agilent, Waldbronn, Germany) and for the sphingoid bases using a Zorbax Eclipse Plus C8 UHPLC column (1.8 $\mu$ m 30 $\times$ 2.1 mm ID, Agilent, Waldbronn, Germany). For all analytes, the concentrations of the calibration standards, quality controls and samples were evaluated by Analyst software 1.7.1 and MultiQuant Software 3.0.3 (both Sciex, Darmstadt, Germany) using the internal standard method (isotope-dilution mass spectrometry).

For flow cytometric analysis of peripheral blood immune cells, 1 mL of venous blood after peripheral venipuncture was sampled from a Heparin-lithium monovette. Flow cytometry was performed as described elsewhere [38]. In brief, S1P ${}_1$ expression was fixed for 30 minutes at 4 °C in a total volume of 10 mL in calcium- and magnesium-depleted PBS/2mM EDTA/2% fatty-acid free BSA/0.1% paraformaldehyde (PFA). Subsequently, the sample was centrifuged (10 minutes, 4 °C, 400 g), the supernatant removed, and the red cells lysed at room temperature in the dark for 15 minutes according to the manufacturer’s instructions (420301, Biolegend). The solution was centrifuged (5 minutes 4 °C, 400 g) and immune cells resuspended in calcium- and magnesium-depleted PBS/2mM EDTA/2% fatty-acid free BSA (staining buffer). 5 $\times$ 10 ${}^5$ cells were dispensed into polystyrene tubes, stained for 60 minutes at 4 °C in the dark, washed with 1 mL of staining buffer and assessed on a FACS Canto II (BD Biosciences, Heidelberg, Germany). The following antibodies were used: anti-human CD45-FITC (clone 30F11, Miltenyi Biotec, Bergisch Gladbach, Germany), anti-human CD3-PerCP (clone UCHT1, Biolegend, San Diego, CA, USA), anti-human S1P1-eFluor660 (clone SW4GYPP, ThermoFisher, Waltham, MA, USA), Mouse IgG ${}_1$ kappa isotype control (clone P3.6.2.8.1, ThermoFisher).

The ischemic lesions were visualized by staining murine coronal slices with a 2% solution of 2,3,5-triphenyltetrazoliumchloride (TTC) 6 hours or 24 hours after ischemia onset as described elsewhere [39]. In addition, MCAO mice were assessed by magnet resonance imaging (MRI) 3 hours after reperfusion. MRI measurements were acquired on a 3T Magnetom TRIO (Siemens Medical Solutions, Erlangen, Germany). During MRI, mice were spontaneously breathing after receiving intraperitoneal anesthesia comprising of Ketamine (Ketavet®, 100 mg kg ${}^{-1}$ bw, Parke-Davis, Berlin, Germany), Xylazine (Rompun®, 20 mg kg ${}^{-1}$ bw, Bayer, Leverkusen, Germany), and Acepromazine (Vetranquil®, 3 mg kg ${}^{-1}$ bw, CEVA Tiergesundheit, Düsseldorf, Germany). 75–150 $\mu$ L of each drug was injected intraperitoneally per mouse in accordance with its body weight. Due to the cardiodepressive effect of the anesthesia, a low volume of application was started and injected as needed. The intertoe reflex was used to assess the depth of anesthesia in mice. Mice were continuously protected from hypothermia with an infrared lamp during anesthesia and the subsequent MRI recording.

GraphPad Prism 8 (GraphPad Software, LLC, La Jolla, CA, USA) was used for statistical analyses. Data are illustrated as median $\pm$ interquartile range (IQR) except where stated otherwise. Statistical significance was assessed using the Mann-Whitney test for unpaired samples unless stated otherwise. A p value of 0.05 was considered statistically significant.

We studied the sphingolipid profile in the acute phase after focal cerebral ischemia, in a model of middle cerebral artery occlusion for 3 hours, followed by reperfusion after 6 hours and 24 hours after ischemia onset. To establish hemorrhagic transformation (HT) more frequently, mice were additionally selected to receive anticoagulation with warfarin (Fig. 1A), which exacerbates the risk of HT [34]. HT was evaluated either using TTC staining or using MRI (Fig. 1B). Mice with effective oral anticoagulation with the vitamin-K-antagonist (VKA) warfarin (0.83 mg kg ${}^{-1}$ ) at the onset of cerebral ischemia displayed a higher frequency of HT both at a follow up time of 6 hours and 24 hours alike (VKA vs. no VKA: 6 h: 71.4% vs. 34.4%; 24 h: 77.8% vs. 42.9%, Fig. 1C). However, HT did not further worsen the neurological deficit of the mice in this experimental paradigm with large ischemic lesions as there were no significant differences in all three neurological scores tested 6 hours and 2 hours after ischemia onset (Fig. 1C).

Fig. 1.

Peri-infarct cortex tissue sampling to study and the impact of hemorrhagic transformation on sphingolipid mediator concentrations in an unbiased sphingolipid profiling approach. (A) Study design: C57/BL6 mice were either sham-operated (n = 10) or underwent ischemic occlusion of the middle cerebral artery for 3 hours (n = 69) either in the absence (n = 46) or presence (n = 23) of the vitamin-K-antagonist (VKA) warfarin. Mice were observed for 6 hours (n = 46) or 24 hours (n = 23) after ischemia onset. (B) Visualization of large hemispheric ischemic stroke (IS) and hemorrhagic transformation (HT) on TTC staining of native postmortem brain sections and MRI imaging. The peri-infarct cortex (PIC) from which analyte material was taken, is visualized. (C) C57/BL6 mice anticoagulated with VKA displayed hemorrhagic transformation after reperfusion with a higher frequency. In this model of large hemispheric infarctions, HT had no further significant impact on the functional status of the mice. The Mann-Whitney U-test was applied to calculate statistical differences. The data are presented as median $\pm$ IQR. (D) An unbiased temporal sphingolipid profiling approach by LC-MS/MS was utilized to assess the impact of HT occurring within the IS area on sphingolipid mediator tissue levels in the PIC (dashed rectangular box).

To determine subclinical changes, we next assessed changes on the molecular level in terms of the peri-infarct sphingolipidome by liquid chromatography mass spectrometry (LC-MS) (Fig. 1D). Tissue samples were isolated 6- or 24-hour after the MCAO intervention, homogenized, and analyzed for their sphingolipid metabolome. Results were correlated to the presence of HT that occurred in VKA-anticoagulated but also in non-anticoagulated mice of this study and were recorded for each individual animal.

We could show that in the acute phase after ischemic stroke S1P levels were significantly reduced and ceramides were almost unanimously more abundantly expressed. Indeed, S1P levels were significantly decreased in the periinfarct cortex 6 h after the onset of focal cerebral ischemia in comparison to sham-operated animals (IS ${}_{\text{6h}}$ vs. sham: 350.3 $\pm$ 151.1 pg mg ${}^{-1}$ vs. 1298.0 $\pm$ 442.9 pg mg ${}^{-1}$ , p 0.0001), and restored after 24 hours of observation as compared to 6 hours of observation (p = 0.06) in accordance with our previous observations [33].

Ceramides, however, displayed a reversed kinetic. For example, C ${}_{\text{16}}$ ceramide was nearly two-fold enhanced at 6 hours of follow up (IS ${}_{\text{6h}}$ vs. sham: 5.8 $\pm$ 1.3 ng mg ${}^{-1}$ vs. 2.7 $\pm$ 1.0 ng mg ${}^{-1}$ , p 0.001) and levels declined back to baseline at 24 hours), and significantly lower than at 6 hours (p 0.0001). Next, C ${}_{\text{18}}$ ceramides mirrored the previous kinetics (IS ${}_{\text{6h}}$ vs. sham: 253.5 $\pm$ 101.8 ng mg ${}^{-1}$ vs. 113.2 $\pm$ 24.8 ng mg ${}^{-1}$ , p = 0.0003; IS ${}_{\text{24h}}$ vs. sham: 98.9 $\pm$ 41.3 ng mg ${}^{-1}$ vs. 113.2 $\pm$ 24.8 ng mg ${}^{-1}$ , p $>$ 0.99; IS ${}_{\text{24h}}$ vs. IS ${}_{\text{6h}}$ : p 0.0001). Likewise, this behavior was also observed for C ${}_{\text{18:1}}$ ceramide, C ${}_{\text{20}}$ ceramide, and C ${}_{\text{24:1}}$ ceramide. C ${}_{\text{24}}$ ceramide displayed endogenously higher variability amongst sham-operated mice and such 24 hours after ischemia onset

Owing to the scarcity of analyte material, several special sphingolipidome species were only analyzed in sham-operated mice and such after 24 hours of observation. Here, no differences could be ascertained comparing mice receiving the MCAO intervention with sham (p $>$ 0.05). Here, sphinganine (Spha), sphingosine (Spho), C ${}_{\text{18}}$ Spha, C ${}_{\text{24:1}}$ Spha, C ${}_{\text{16}}$ glucosylceramide (GlcCer), C ${}_{\text{18}}$ GlcCer, C ${}_{\text{24:1}}$ GlcCer, C ${}_{\text{16}}$ lactosylceramide (LacCer), C ${}_{\text{18}}$ LacCer, and C ${}_{\text{24:1}}$ LacCer have all been tested (Fig. 2).

Fig. 2.

Reversed kinetics of S1P and ceramide species in the peri-infarct tissue after ischemic stroke. Various sphingolipid species were tested by LC-MS from analyte material sampled from the peri-infarct tissue comparing mice receiving sham-intervention with such having undergone MCAO-intervention and followed up for 6 or 24 hours, respectively. S1P, C ${}_{\text{16}}$ ceramide, C ${}_{\text{18}}$ ceramide, C ${}_{\text{18:1}}$ ceramide, C ${}_{\text{20}}$ ceramide, C ${}_{\text{24}}$ ceramide, and C ${}_{\text{24:1}}$ ceramide were kinetically assessed. Due to scarcity of the analyte material, special sphingolipid species could only be characterized in sham mice and MCAO-operated mice with a follow up time of 24 hours (for abbreviations see text). A Kruskal-Wallis test with Dunn’s multiple comparisons test was applied to calculate statistical differences. Data are presented as median $\pm$ IQR; *p 0.05, **p 0.01, ***p 0.001, ****p 0.0001.

Next, we aimed to characterize the sphingolipidomic changes arising from the establishment of HT after ischemic stroke (IS). Initially, we had observed that HT occurred in a higher frequency if mice had been anticoagulated with VKA prior to the MCAO intervention [34]. Despite some scientific indication that warfarin might affect ceramide expression [40] via inhibition of the production of pro-inflammatory ceramides, we initially treated mice with VKA treatment (rectangles) to enrich for this complication of IS. Indeed, HT (red coloring) did occur more frequently (Fig. 3), but mice also passed away more often (6 h: VKA vs. no VKA: 13.6% vs. 0%, Supplementary Fig. 1). C ${}_{\text{16}}$ ceramide, a known pro-inflammatory ceramide for example, is a product of acid sphingomyelinase 1 [41], and was being produced less abundantly in mice comparing mice treated with VKA to such without in the absence of HT (VKA vs. no VKA: 4.9 $\pm$ 2.1 ng mg ${}^{-1}$ vs. 5.8 $\pm$ 1.3 ng mg ${}^{-1}$ ), although statistical differences could not be ascertained owing to the small sample size. Since we did not know the pharmacological consequences of warfarin onto other sphingomyelinase, and as a result of these two observations, we then discontinued the VKA model to induce HT in order to establish an unbiased sphingolipidomic landscape arising from HT because of IS in the absence of VKA anticoagulation (circles).

Fig. 3.

Gross assessment of the sphingolipidome within the peri-infarct tissue does not distinguish between IS and HT. Cerebral specimens from the peri-infarct cortex of mice that have developed hemorrhagic transformation (HT) of the ischemic core after MCAO were analysed regarding their differential sphingolipid expression profile as compared to mice without HT. A Kruskal-Wallis test with Dunn’s multiple comparisons test was applied to calculate statistical differences. Data are presented as median $\pm$ IQR; *p 0.05, **p 0.01 ***p 0.001. S1P, sphingosine 1-phosphate; Spho, sphingosine; Spha, sphinganine; Cer, ceramide.

Our data demonstrates that HT is not associated with any alterations of the sphingolipidome encompassing the assessment of S1P and various ceramide species neither at 6 hours nor 24 hours after follow up (FU).

S1P can enhance immune cell migration as a chemotactic agent [33, 42] via S1P ${}_{\text{1}}$ -mediated up-regulation of ICAM-1 and E-selectin on endothelial cells [43, 44, 45, 46], which are important adhesion molecules for innate immune cells such as monocyte or granulocytes during diapedesis. Therefore, we sought to analyze the expression pattern of S1P ${}_{\text{1}}$ on peripheral immune cells to estimate their adhesiveness towards S1P-stimulated endothelium to obtain insights into putative pathophysiological consequences of this increase in peri-infarct cortext S1P levels. Indeed, CD45 ${}^{+}$ immune cells of the adaptive lineage, i.e., T cells (CD45 ${}^{+}$ CD3 ${}^{+}$ within the FSC/SSC properties for lymphocytes) or B cells (CD45 ${}^{+}$ CD3 ${}^{-}$ within the FSC/SSC properties for lymphocytes) had an abundant expression of S1P ${}_{\text{1}}$ of 37.8% or 80.2%, respectively, compared to isotype ( 5%, Fig. 4A). Interestingly, innate immune cells captured in the granulocyte or monocyte gate by FSC/SSC-properties, typically first responders in the context of acute inflammation, had a way superior expression pattern of S1P ${}_{\text{1}}$ (Fig. 4A). Here, monocytes had an abundant expression of 80.7%, whilst granulocytes were almost unanimously positive for S1P ${}_{\text{1}}$ . S1P ${}_{\text{1}}$ was shown to be a crucial factor regulating neutrophil recruitment [47].

Fig. 4.

Chemotactic recruitment of innate immune cells via sphingolipids and in-depth assessment of the sphingolipid subspecies as putative risk factors for HT. (A) As an example of chemotactic recruitment of innate immune cells, whole blood leukocytes were assessed for their expression pattern of the type 1 S1P receptor (S1P ${}_{\text{1}}$ ) by which cells are being recruited to effector organs. (B) Stratifying the assessed sphingolipid abundances by their topmost or bottommost 35% representative allowed the identification of putative species associated with HT in MCAO samples 24 hours after ischemia onset. Especially high levels of S1P and its immediate precursor sphingosine (Spho) or C ${}_{\text{18}}$ lactosylceramide (LacCer) appeared to predispose for HT. In contrast, C ${}_{\text{18}}$ sphinganine (Spha) or C ${}_{\text{16}}$ LacCer were conversely regulated, i.e., lower levels were more frequently linked to HT. Data are presented as median $\pm$ IQR, and no statistical testing was performed.

Seeing that S1P is consumed (via receptor binding and subsequent internalization by effector cells) in the acute phase of IS, we intended to risk stratify our established sphingolipidomic profiles by the topmost and bottommost 35% of sphingolipid species measured and qualitative check whether HT had occurred or not (Fig. 4B). In doing so, we identified that apart from C ${}_{\text{18}}$ sphinganine (Spha), the topmost expression levels of all other immediate S1P relatives and S1P itself were qualitatively linked to a more frequent conversion of HT. For example, the topmost 35% expression levels (Hi ${}_{\text{35\%}}$ ) for S1P showed HT in three out of four mice, whereas in the case of the bottommost 35% (Lo ${}_{\text{35\%}}$ ) only one mouse of four had converted to HT. Likewise, sphingosine, the closest relative of S1P had seen HT in four out of five mice in Hi ${}_{\text{35\%}}$ opposed to one out of five mice in Lo ${}_{\text{35\%}}$ . Conversely, for C ${}_{\text{18}}$ Spha, Lo ${}_{\text{35\%}}$ had seen HT in 60%, whilst in Hi ${}_{\text{35\%}}$ only one out of five mice had developed HT (20%). Apart from Spho, one ceramide subspecies also appeared to be closely linked to the establishement of HT. Here, C ${}_{\text{18}}$ lactosylceramide had seen HT in 80% in Hi ${}_{\text{35\%}}$ as opposed to only 20% in Lo ${}_{\text{35\%}}$ (Fig. 4B). We report an association of multiple sphingolipid species as putative risk factors for the conversion of HT, however, would be keen to assess their relevance as causative agents to drive innate immune cell recruitment and subsequent loss of endothelial integrity predisposing for HT in the future. Yet, we are only reporting associations.