Targeting ferroptosis protects against multiorgan dysfunction and death.

main of an iron-dependent type of cell death conceptualized as ferroptosis 8,9 . Here we show that pharmacological targeting of ferroptosis with our most potent ferrostatin-analogue 10 rescues from death in acute single and multiorgan dysfunction in mice, but not sepsis. Daily monitoring of critically ill ICU patients revealed that the peak level of reflecting excessive lipid peroxidation, correlates with multiorgan dysfunction and death. Our results demonstrate that ferroptosis targeting is life-saving in experimental models of critical illness and that monitoring of malondialdehyde can allow patient stratification. Therefore, controlling the extent of ferroptosis in non-septic patients with multiorgan dysfunction could become a novel treatment for one the major of global

Patients who suffer from critical illness after an inciting event, for instance major trauma, 53 surgery, or infection 4 , frequently require intensive care unit (ICU) support. Critical illness is 54 characterized by multiple organ dysfunction syndrome (MODS), often referred to as multiorgan 55 dysfunction. The extent of organ dysfunction in critically ill patients is correlated to an increase 56 in plasma catalytic iron 6,11,12 also known as labile iron or non-transferrin bound iron, which is 57 a transitional pool of both extra-and intracellular iron. An excess of iron can be sufficient to 58 induce ferroptosis 13,14 , a necrotic cell death type caused by iron-dependent peroxidation of 59 polyunsaturated phospholipids in cell membranes 15,16 , resulting in cell rupture 17,18 . Therefore, 60 we hypothesized that ferroptosis might be a detrimental factor in multiorgan dysfunction. ill adult patients enrolled in a prospective cohort study 21 . In this cohort, the median age was 60 70 (51-70) years. At enrolment, the median sequential organ failure assessment (SOFA) score was 71 9 (7-11), with 57% of patients suffering from sepsis and 25% of patients having septic shock.

72
The 30-day mortality rate was 23%. To monitor the dynamic fluctuations in these patients, 73 blood was sampled daily for up to 7 days. We found that the maximum value of Fec (Fec max ) 74 per patient showed a significant positive correlation with the SOFA score, reflecting the extent 75 of organ dysfunction (Fig. 1a). The Fec max values of patients who succumbed to their illness 76 were significantly higher than those of surviving patients (Fig. 1b), and higher Fec max values 77 were found for septic shock patients compared to sepsis patients (Fig. 1c, Extended data Fig.   78 1a-g). Similarly to Fec max values, the maximum value of MDA (MDA max ) per patient also 79 showed a significant positive correlation with the SOFA score (Fig. 1d) and was significantly 80 higher in the deceased group than in patients who survived (Fig. 1e). In contrast to Fec max , we 81 found no association of MDA max values with either sepsis or septic shock ( Fig. 1f and Extended 82 data Fig. 1h-n). It is well-known that an acute phase response during infection upregulates host 83 proteins to control free iron 22 , which might explain the dampened ferroptosis signature during 84 sepsis. Consistent with a stronger association of MDA max than Fec max with death, only MDA 85 values were significantly higher in the deceased group when analyzed per day (Extended data 86 Fig. 2a-n). A positive correlation between Fec and MDA within patients is evident from the Fec 87 levels being significantly higher on the day a patient reached MDA max compared to the day of 88 the minimum MDA value (MDA min ) (Fig. 1g). Interestingly, these MDA max values revealed a 89 bimodal distribution for the deceased patients (Fig. 1h). Stratification of all patients based on 90 the local minimum showed that patients with an MDA max >2.85 µM, representing 24.4% of all 91 patients, had a significantly lower survival probability (Fig. 1i). In fact, within this subgroup,  biomarker for necrosis, we also monitored aspartate aminotransferase (AST) and alanine 116 aminotransferase (ALT) to reflect liver injury, creatinine (Cr) and urea to monitor kidney 117 function, myoglobin (Mb) and creatine kinase (CK) to assess muscle injury, troponin T to 118 5 quantify myocardial injury and ferritin to investigate iron dysbiosis. Except for Cr and Mb, 119 which peaked at 2h post-iron overload, all other injury biomarkers peaked at 12h. The 120 exceptionally high levels of CK mainly originated from skeletal muscle tissue, as opposed to 121 heart or smooth muscle tissue (Extended data Fig. 4a-b). MDA levels were determined to 122 monitor excessive lipid peroxidation in multiple organs. A rapid increase, peaking at 30 min to 123 1h after FeSO4 injection, was observed in kidney, liver, ileum and skeletal muscle tissue, as 124 well as in plasma ( Fig. 2c and Extended data Fig. 3g). In addition, an increased number of dead 125 cells as a function of time was detected in kidney, liver and ileum tissue, reflected by a terminal 126 deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) staining ( Fig. 2d-g).

127
Hematologic analysis revealed leukocytosis, in particular neutrophilia and lymphopenia, which 128 is also typically observed in patients with acute iron overload 23 (Extended data Fig. 4c,d). hydroperoxides, these mice are sensitized to ferroptosis 30,31 . When subjected to acute iron 150 overload, they showed a strong sensitization with significantly higher levels of plasma injury 151 biomarkers compared to their littermate controls (Extended data Fig. 6a). Finally, we used a half-life (t1/2) of around 3-4h was determined for plasma, kidney, lung, and intestine, with t1/2 167 of muscle being slightly shorter (2h) (Extended data Fig. 7b,c). The median blood to plasma   Interestingly, treatment with UAMC-3203 had no effect on the plasma injury biomarkers in  Similarly, mice overexpressing GPX4 showed no or even slightly decreased survival after 186 respectively CLP-or lipopolysaccharide (LPS)-induced lethal shock (Extended data Fig. 8c,d). 187 This could imply that ferroptosis inhibition is a promising strategy to control non-septic patients 188 with multiorgan dysfunction (e.g. trauma), while for septic shock patients with multiorgan 189 dysfunction a combination treatment might be needed to control systemic inflammation as well, 190 as we previously reported viz. simultaneous neutralization of IL-1 and -18 35 . Noteworthy, 191 reduced levels of plasma iron were detected after TNF or CLP challenge (Extended data Fig.   192 8e,f), presumably as a protective strategy to limit microbial growth in an attempt to reduce their bowman´s spaces (Fig. 3i). Daily injection of UAMC-3203 following TAM treatment in 208 Gpx4 RTEKO mice could significantly delay death (Fig. 3k). In renal ischemia reperfusion injury, 209 UAMC-3203 also protected by attenuating tubular damage in the kidney (Extended data Fig.   210 9d,e). In the case of ferroptosis-driven acute liver injury, Gpx4 HEPKO mice showed very high 211 ALT, AST and LDH levels concomitant with severe cell death and morphological liver tissue 212 changes (Fig. 3l,m and Extended data Fig. 9f) when sacrificing the mice upon a drop in body 213 temperature. Tissue damage was characterized by enlarged nuclei, chromatin aberrations and 214 paling of both the hepatocellular nuclei and cytoplasm, likely reflecting death cell corpses (Fig.   215 3l). In the centrilobular region, mild inflammatory infiltrates were detected (Arrow heads Fig.   216 3l). For Gpx4 HEPKO mice, UAMC-3203 treatment showed a strong protection against TAM-217 induced acute liver dysfunction and subsequent death (Fig. 3n), with almost normalized liver 218 plasma injury biomarkers by day 21 when the mice were sacrificed (Extended data Fig. 9g).

8
This outcome strongly contrasted with the inability of Fer1 to rescue the mice or prolong 220 survival (Fig. 3n). The superior life-saving activity of UAMC-3203 in liver compared to kidney 221 might be due to the conversion of UAMC-3203 in the liver to a more active metabolite, which 222 is still under investigation.

223
In conclusion, we found that the severity of multiorgan dysfunction and the probability of death  We thank the VIB Flow Core and the VIB Bioimaging Core for training, support, and access 253 to the instrument park and are grateful for the statistical support provided by M. Vuylsteke. We