An arylthiazyne derivative is a potent inhibitor of lipid peroxidation and ferroptosis providing neuroprotection in vitro and in vivo

Lipid peroxidation-initiated ferroptosis is an iron-dependent mechanism of programmed cell death taking place in neurological diseases. Here we show that a condensed benzo[b]thiazine derivative small molecule with an arylthiazine backbone (ADA-409-052) inhibits tert-Butyl hydroperoxide (TBHP)-induced lipid peroxidation (LP) and protects against ferroptotic cell death triggered by glutathione (GSH) depletion or glutathione peroxidase 4 (GPx4) inhibition in neuronal cell lines. In addition, ADA-409-052 suppresses pro-inflammatory activation of BV2 microglia and protects N2a neuronal cells from cell death induced by pro-inflammatory RAW 264.7 macrophages. Moreover, ADA-409-052 efficiently reduces infarct volume, edema and expression of pro-inflammatory genes in a mouse model of thromboembolic stroke. Targeting ferroptosis may be a promising therapeutic strategy in neurological diseases involving severe neuronal death and neuroinflammation.

Orally administered ADA-409-052 shows fast absorption and good brain penetration. Using a pharmacokinetic study in mice, we established an in vivo treatment protocol. Orally administered ADA-409-052 at the dose of 10 mg/kg showed individual (n = 3) maximal plasma concentrations within 0.25-1 h time window, followed by a fast distribution and plasma clearance in a time window of 2-8 h (data not shown). At the 0.75 h time point after orally dosing 10 mg/kg ADA-409-052, plasma and brain concentrations (n = 3) were in the ranges 1080-1324 ng/ml and 1100-1526 ng/g tissue, respectively. Oral administration of 30 mg/kg ADA-409-052 resulted in a mean plasma concentration of 3771 ng/ml and a mean brain concentration of 3237 ng/g at 0.75 h. The concentrations dropped to 23.7 ng/ml and 30.9 ng/g at 4 h, respectively, and below detection limit at 24 h (n = 3 per time point). These results indicated fast and linear dose-dependent oral exposures with excellent brain penetration with a relatively short half-life in the plasma and brain.
ADA-409-052 decreased the expression of pro-inflammatory genes after ischemic stroke. Stroke is well known to trigger inflammation that contributes to ischemic damage. Since ADA-409-052 provided a mild protection against inflammation-induced neuronal death in vitro, we evaluated whether ADA-409-052 ameliorates ischemia-induced inflammation by analyzing the expression of astrocytic Gfap, and allograft inflammatory Aif1 encoding ionized calcium-binding adapter molecule 1 (Iba1), a commonly used microglia marker. Already during the phase of early acute inflammation at 24 h post-stroke, ADA-409-052 suppressed the stroke-induced expression of Gfap in the ipsilateral peri-ischemic area (from 1.059 ± 0.397 to 0.793 ± 0.197; p < 0.05; Fig. 6a), while the expression of Aif1 remained unchanged at the given time point (Fig. 6b). However, in comparison to vehicle-treated mice, ADA-409-052 decreased the mRNA expression of Ccl2, in the ipsilateral perilesional tissue (ADA-409-052: 6.6 ± 0.294 to vehicle: 10.05 ± 0.584; p = 0.0278; Fig. 6c). Furthermore, the expression of Arg1 (arginase-1) was reduced in the ipsilateral peri-ischemic area of ADA-409-052-treated mice when compared to contralateral peri-ischemic area (ipsilateral: 0.639 ± 0.334 vs contralateral: 1.288 ± 0.501; p < 0.001; Fig. 6d Fig. 6g). Although the levels of MCP-1 and TNF-α appeared visibly higher in minocycline-treated mice when compared with ADA-409-052-treated mice, the averages remained statistically insignificant between all treatment groups (Fig. 6h,i). The gene expression data of Il10, however, are in line with the peripheral secretion into the blood plasma of the anti-inflammatory interleukin (Fig. 6j), as both remained unaltered.

Discussion
The brain is the most susceptible mammalian tissue to the oxidative stress caused by an imbalance of redox reactions. This is due to the brain being extremely rich in lipids with unsaturated fatty acids, which are major substrates for ROS production, while simultaneously consuming about 20% of the body's oxygen 29,30 . In addition, most brain areas have a high iron concentration, needed for iron-catalyzed processes, such as oxygen transportation, oxidative phosphorylation, myelin synthesis, and neurotransmitter metabolism 31 , leaving the brain sensitive to abnormal iron homeostasis and iron-dependent LP 32 . The risk of unbalanced iron homeostasis and iron-dependent LP is high, especially in the elderly, as iron accumulates into the brain during aging 3,33 . Moreover, abnormally high iron concentrations are found in affected brain areas in neurodegenerative diseases, such as AD, PD, amyotrophic lateral sclerosis and stroke 5,29,31,33,34 . Hence, ferroptosis is considered a major form of cell death in neurodegenerative diseases and stroke 5 , making it a potential target for therapies of age-related brain diseases.
Our study demonstrates that in neuronal cell lines ADA-409-052, a condensed benzo[b]thiazine derivative with an arylthiazine backbone, strongly inhibits LP and the subsequent cell death induced by GSH depletion or blockade of GPx4, the key-triggers of ferroptosis. Both GSH depletion and GPx4 deficiency are highly relevant conditions, as GPx4 catalyzes the reduction of lipid peroxides at the expense of GSH, and as levels of both GSH and GPx4 are reduced in stroke and neurodegenerative diseases 35 . Attempts to boost GPx4 directly or indirectly in animal models of neurodegenerative diseases and stroke have been successful [36][37][38] , but remained unfeasible and inefficient in clinics [39][40][41] . An advantage of ADA-409-052 and molecules alike is that their ability to inhibit ferroptosis is independent on the availability of GSH, suggesting that ADA-409-052 could protect the brain even when GSH levels decrease. Furthermore, ADA-409-052 inhibited glutamate-induced changes of mitochondrial morphology, facilitating better cell survival.
Neuroinflammation is a well-known immune response to protect brain tissue and neurons from toxic molecules or noxious structures in all neurodegenerative diseases and acute brain insults. Still, when overwhelming or prolonged, it may turn harmful and promote neuronal damage and loss [42][43][44][45] . An important inflammatory pathway is the metabolism of arachidonic acid (AA), a major component of cell membrane lipids, into precursors of bioactive pro-inflammatory mediators, such as the eicosanoids, prostaglandins, interleukins, TNF and leukotrienes, that promote inflammatory cascades 46,47 . There is a complex relationship between ferroptosis, AA metabolism, and pro-inflammatory mediators. Ferroptotic cell death has been reported to be enhanced by proinflammatory polarization of macrophages and microglia 48 . In addition, previous studies indicate that, at least in cancer cells, ferroptosis can have pro-inflammatory effects and initiate inflammation by increasing the expression of PTGS2 (which encodes cyclooxygenase-2 (COX-2)), accelerating AA metabolism and promoting the secretion of pro-inflammatory molecules [49][50][51] . On the other hand, inflammatory cytokines (such as TNF, PGE2, IL-1, IL-1β, and IL-6) have been shown to directly alter the level and activity of GPx4 in cancer cells 52 and TNF treatment of cells leads to the sustained downregulation of GPx4 that may trigger ferroptosis 50 . The reduced secretion of MCP-1 and TNF-α, measured from ADA-409-052-stimulated BV2 cells using the BD CBA mouse inflammation kit, further indicates that the compound improves the activity of GPx4, countervailing pro-inflammatory and ferroptotic conditions. To further prove the compound's efficacy, we tested ADA-409-052 on neuronal death in a neuron (N2a)-macrophage (RAW 264.7) co-culture model of LPS/IFN-γ-induced neuroinflammation. Our Figure 6. ADA-409-052 administration in vivo reduced the expression levels of pro-inflammatory markers (a-f) and altered the peripheral cytokine secretion (g-j) at 1 day post-ischemia. The mRNA expression of tissue samples from the ipsi-and contralateral peri-ischemic area at 24 h was analyzed using qPCR. Expression levels of (a) Gfap and (c) Ccl2 were markedly reduced in the ipsilateral peri-ischemic area after ADA-409-052 treatment when compared to vehicle. (b) While, Aif1 remained unchanged throughout all groups, ADA-409-052 lowered (d) Arg1 and (e) Il6 mRNA expression in the ipsilateral samples when compared to contralateral. (f) Ischemic mice treated with ADA-409-052 showed reduced Hmox1 expression in the ipsilateral periischemic area when compared to contralateral expression levels, to ipsilateral expression of vehicle-treated mice respectively. Minocycline decreased the expression of Hmox1 only ipsilateral when compared to contralateral. mRNA expression levels of peri-ischemic tissue samples from ipsi-and contralateral, data are expressed as mean ± s.e.m.; unpaired, two-tailed student's t-test; n = 9-13; *p < 0.05, **p < 0.005, ***p < 0.001. (g) The secretion of the inflammation-mediating IL-6 was significantly reduced in ADA-409-052-treated mice when compared to vehicle-treated mice, as detected from the blood plasma, collected from ischemic mice at 1 dpi, by using the BD CBA mouse inflammation kit. (h,i) Both MCP-1 and TNF-α were equally low in ADA-409-052-treated mice when compared with vehicle treatment and showed a clear trend towards reduction when compared with minocycline-treated mice (changes did not reach significance). (j) The peripheral secretion of IL-10 remained unaltered by any given treatment. Data are expressed as mean ± s.e.m.; unpaired, two-tailed student's t-test; n = 7; *p = 0.027. www.nature.com/scientificreports/ data show that ADA-409-052 suppresses pro-inflammatory activation of microglia-and macrophage-mediated neuronal death. Our findings are in line with previous studies showing that intracerebral injection of ferrostatin-1, a ferroptosis inhibitor, reduced the expression of COX-2 with concomitant neuroprotection 20 . This indicates that ADA-409-052 and established ferroptosis inhibitors reduce inflammation and inflammation-induced injury also in brain tissue.
To investigate whether the brain penetrating compound ADA-409-052 can provide neuroprotection in vivo, we chose thromboembolic stroke in mice, as this model results in distinct and relatively fast brain injury and as it mimics the type of stroke that is very common in clinics 53,54 . The model evolves a rather small, focal lesion ensuring a good survival rate and its mechanisms resemble those of vascular occlusion in patients 55 . We also measured the blood flow during and after thrombin injection to secure proper and consistent clot formation, which we found to be reflected as very homogenous infarct volumes measured by MRI analysis, indicating the model's reliability. We found that treatment with ADA-409-052 resulted in a marked neuroprotection as detected by significantly reduced lesion volumes post-ischemic stroke, which equaled to the reduction obtained by our positive control, minocycline. The protective effect of minocycline replicated in our study the results of previous studies [56][57][58] . Importantly, only ADA-409-052 restricted the expansion of stroke-induced edema. This effect of ADA-409-052 is in line with the previous finding that intracerebral injection of a ferroptosis inhibitor reduces brain inflammation in a rat model of intracerebral hemorrhage 20 .
Brain swelling is an important clinical measure as it is associated with deterioration and manifestation of neurological deficits and elevated mortality 59,60 . Our data thus confirm the previous findings that ferroptosis is an important mechanism of both brain infarction and brain edema. Furthermore, the results suggest that ADA-409-052, a small molecule inhibiting ferroptotic neuronal death, may be able to efficiently prevent ischemic neuronal death and harmful stroke-induced edema in thromboembolic stroke and prove beneficial in various brain diseases.
Minocycline was used as our reference compound in the stroke study as well as in key in vitro experiments. While we could have compared ADA-409-052 with established inhibitors of ferroptosis, we preferred the compound minocycline, which is extensively characterized and tested in a large number of studies, including our own experiments. It is a small molecule with excellent brain penetration and proven efficacy in vivo 27,58,[61][62][63][64][65][66][67] . Moreover, minocycline inhibits LP, attenuates iron-induced brain injury following experimental intracerebral hemorrhage 66,68 , and results in expression of molecules counteracting the development of ferroptosis 67 . It is also one of the most potent small molecules against experimental ischemic stroke supported by a large number of studies 27,69,70 . Our data show that minocycline protected against ferroptotic death only at relatively high concentrations, which is in line with previous studies suggesting that it is not a direct ferroptosis inhibitor. Assuming ferroptosis is a major mechanism of stroke-induced secondary edema, it is unsurprising that ADA-409-052 and not minocycline significantly mitigated the brain edema that developed 24 h after ischemic insult in this study's TE model of stroke.
ADA-409-052 down-regulated pro-inflammatory cytokines and chemokines induced by TE stroke as measured by qPCR, the BD CBA respectively. Among the genes with reduced expression after ADA-409-052 treatment was Ccl2 (MCP-1), a marker of activated pro-inflammatory microglia/macrophages and infiltrating neutrophils, associated with neuronal degeneration 71 . Even though, ADA-409-052 was unable to change the peripheral secretion of MCP-1, only minocycline caused a certain increase of MCP-1 secretion into the plasma. Importantly, Il6, a pro-inflammatory downstream product of toll-like receptor (TLR) 72 , was reduced in both gene expression of ischemic brain tissue and cytokine levels of blood plasma. In addition, ADA-409-052 reduced the expression of stroke-induced Hmox1 that encodes heme oxygenase-1, an essential enzyme for iron-dependent LP during ferroptotic cell death 73 . These data are in line with in vitro data indicating that ADA-409-052 efficiently inhibits ferroptosis and associated inflammation in brain tissue.
In conclusion, small molecule inhibitors of ferroptosis such as ADA-409-052 may be potent protective compounds against neurodegeneration and acute brain insults. As mechanisms of these brain diseases are complex and involve several detrimental signaling pathways, an advantage of ferroptosis inhibitors is their ability to reduce not only ferroptosis triggered by non-enzymatic LP, but also cell death mediated by enzymatic LP heavily contributed by inflammation. A combination of ferroptosis inhibitors with molecules targeting still other harmful pathways might offer attractive therapeutic approaches for brain disorders.

Materials and methods
In vitro lipid peroxidation analysis. C11-BODIPY 581/591 (ThermoFisher Scientific, MA, USA), was used as a fluorescent sensor to determine TBHP-induced LP in PC-12 cells. Analyses were done by cellular imaging or flow cytometry, as C11-BODIPY's fluorescence emission peak shifts from ~ 590 nm to ~ 510 nm upon oxidation 74 .
For the fluorescent cellular imaging, PC-12 cells (ATCC, VA, USA; for all experiments see suppl. information for cell culturing details) were pre-incubated 90 min with 5 µM C11-BODIPY. Next, cells were washed and exposed for 90 min to 1 mM TBHP (Sigma, MO, USA) in the presence or absence of ADA-409-052 at different concentrations (0.625, 1.25, 2.5, 5, and 10 µM); DMEM media served as control. Images were taken using the 10X magnification at a Zeiss Observer.Z1 fluorescent microscope (AxioCam MRm, Zeiss Zen Imaging Software, Zeiss, Germany). Using the ImageJ software (NIH, MD, USA) an appropriate threshold was set and the mean grey value for each channel (8-bit images) was calculated. Results are displayed as the red/green-ratio, normalized to control.

In vitro ferroptosis models.
To test the anti-ferroptotic properties of ADA-409-052, PC-12 cells were exposed for 24 h to either 0.25 µM RSL3 (Selleck Chemicals, TX, USA) or 20 mM glutamate (Glu, Sigma) with or without ADA-409-052 at following concentrations: 2.5, 5, 10, and 20 µM. Minocycline (Sigma) was used in parallel at the same concentrations. The compounds' efficacy in cell protection was measured as cell viability 24 h post-treatment using the resazurin assay. In brief, cells were incubated for 2 h at 37 °C with 10 µM resazurin in HBSS. The absorbance was measured at 485 nm. Additionally, ADA-409-052's effect on intracellular GSH levels was measured. We thus exposed PC-12 cells to 20 mM glutamate with or without 5 µM ADA-409-052. After 24 h of exposure, GSH was extracted from the collected cell pellet with 5% sulfosalicylic acid and the supernatant neutralized with 3 M Tris. The Amplite Rapid Fluorimetric Glutathione GSH/GSSG Ratio Assay Kit (AAT Bioquest, Sunnyvale, CA, USA) was used to measure the GSH concentration according manufacturer's instructions. All results were expressed as percentage of mean fluorescence from untreated control cells.

Visualization of mitochondria in an in vitro model of ferroptosis.
To demonstrate the effect of ADA-409-052 on mitochondria changes upon ferroptosis, PC-12 cells were exposed to glutamate (20 mM) in presence or absence of ADA-409-052 (10 µM) for 24 h, as described before. Live cells were stained with MitoTracker Red CMXRos (ThermoFischer Scientific) and imaged with a 63X Achroplan objective on a Zeiss Laser Scanning Microscope 710 with Zen Imaging Software (Zeiss) (see suppl. information for the full protocol).

In vitro inflammation model on BV2 microglial cells. BV2 microglial cells were stimulated with pro-
inflammatory LPS. Therefore, BV2 cells were incubated with 50 ng/ml LPS (serotype O111:B4, Sigma) for 24 h in the presence or absence of ADA-409-052 using 2.5, 5, 10, and 20 µM. NO release was assessed using the Griess reaction (Promega, WI, USA). A standard curve was prepared using 0-100 μM sodium nitrite. Nitrite concentrations were presented as percentages of NO release after normalization to LPS-exposed BV2 cells. Cytokines were measured from conditioned media of LPS-+ /− ADA-409-052-treated cells using the mouse inflammation kit of the BD CBA was performed according manufacturer's instructions (for details see suppl. information). Data were acquired on a BD Accuri C6 flow cytometer, and the FCAP Array Software v3.0 was used for analyses (all BD Bioscience). Data were normalized to control-treated cells, set as 100%.
Magnetic resonance imaging. Lesion and edema volumes were determined 24 h post-ischemia, using T2-weighted MRI imaging in the anaesthetized animal. MRI was performed on a vertical 9.4 T Oxford NMR 400 magnet as described elsewhere 76 . Multi-slice T2 weighted images (repetition time 3000 ms, echo time 40 ms, matrix 128*256, FOV 19.2 mm, 12 slices à 1 mm-thickness) were obtained. The MRI images were analyzed using an in-house aedes software under MATLAB environment (Math-Works, MA, USA). Lesion and edema volume were calculated blinded to study groups following Shuaib et al. 77 . Following MRI, various samples were collected from finally anaesthetized animals (details in suppl. information).
Quantitative real-time PCR analyses of mRNA levels. Total RNA was isolated from homogenized peri-ischemic brain tissue samples of the ipsi-and contralateral hemispheres, in TRIzol (Life technologies, CA, USA) according to the manufacturer's instructions. cDNA (2.5 ng/µl) was synthesized using 500 ng of total RNA, dNTP, Maxima reverse transcriptase, and random hexamer primers, combined with ribonuclease inhibitors (ThermoFisher Scientific). The relative expression levels of mRNAs encoding the selected genes were measured. According manufacturer's protocol by RT-qPCR (StepOnePlus, ThermoFisher Scientific) using specific TaqMan gene expression assays (ThermoFisher Scientific), the following genes were analyzed: Gfap, Arg1, Mpo, Il6, Il10, Tnfα, Ccl2, and Hmox1. The expression levels were normalized to b2m (beta-2 microglobulin), calculated as described previously 78 , presented as fold change.
Analyses of peripheral cytokines in blood plasma. The mouse inflammation kit of the BD CBA (BD Bioscience) was used to detect cytokines secreted into the plasma of 7 mice per group at 1 dpi. CBA was performed according manufacturer's instructions and data were acquired as described in suppl. information. Results are presented in pg/ml final concentration.
Statistical analysis. Statistical significant differences (p-values) between the means for two measurement groups were calculated using two-tailed distribution and two-sample equal variance of the student's t-test; Oneway ANOVA with Tukey's multiple comparison test was used whenever appropriate (GraphPad Prism software). The data are displayed as mean ± standard error of the mean (s.e.m.).
Altogether eight mice (13.34%) were excluded due to mortality, hemorrhagic transformation, or insufficient clot formation.

Data availability
The datasets generated and analyzed during the current study are available from the corresponding author on reasonable request.