Identification of essential sites of lipid peroxidation in ferroptosis
铁死亡中脂质过氧化重要位点的鉴定

The iron-dependent form of lipid-peroxidation-mediated cell death known as ferroptosis, although remaining somewhat enigmatic, represents an emerging modality for treatment of several illnesses, including cancers and degenerative diseases^1,2,3,4,5. There is a suite of compounds that induce and inhibit this form of regulated cell death, highlighting the therapeutic potential of controlling this mechanism of cell death. Induction of ferroptosis, pharmacologically or genetically, has been shown to slow and regress tumor growth in vivo^6,7. In addition, some approved medications may kill cancer cells by ferroptosis, albeit not with high selectivity^8,9,10,11. Compounds inhibiting ferroptosis have shown therapeutic potential in models of degenerative diseases and ischemia-reperfusion injuries^{12,13,14,15,16,17,18}. The development of potent and selective ferroptosis-inducing or ferroptosis-inhibiting drugs will benefit from a deeper understanding of the mechanisms of this form of cell death.
铁依赖形式的脂质过氧化介导的细胞死亡称为铁死亡，尽管仍然有些神秘，但却代表了治疗多种疾病的新兴方式，包括癌症和退行性疾病^{1 , 2 , 3 , 4 , 5} 。有一系列化合物可以诱导和抑制这种形式的受调节细胞死亡，凸显了控制这种细胞死亡机制的治疗潜力。药理学或遗传学上的铁死亡诱导已被证明可以减缓和逆转体内肿瘤的生长^{6 , 7} 。此外，一些批准的药物可能通过铁死亡杀死癌细胞，尽管选择性不高^{8 , 9 , 10 , 11} 。抑制铁死亡的化合物已在退行性疾病和缺血再灌注损伤模型中显示出治疗潜力^{12 , 13 , 14 , 15 , 16 , 17 , 18} 。更深入地了解这种细胞死亡形式的机制，将有助于开发有效和选择性的铁死亡诱导或铁死亡抑制药物。

Ferroptosis occurs when lipid hydroperoxides accumulate in cellular membranes^19,20,21. The increased accumulation of such hydroperoxides can be due to inhibition of lipid reactive oxygen species (ROS) detoxification systems or by direct lipid ROS generation^12,22,23,24. Polyunsaturated fatty acyl (PUFA) moieties, incorporated into phospholipids, are specific substrates of iron-dependent peroxidation during ferroptosis due to their easily abstractable bis-allylic hydrogens^21,25,26. Because PUFA moieties could be incorporated into membranes throughout the cell, the question of which cellular membranes are particularly susceptible to, or necessary for, ferroptosis is still enigmatic. Previous work exploring the distribution of ferrostatin-1 (fer-1), an anti-ferroptotic agent, revealed that, although fer-1 localizes to the endoplasmic reticulum (ER) membrane, mitochondrial membranes and lysosomal membranes, accumulation in neither the mitochondria nor the lysosomes was necessary for activity, suggesting that the ER membrane may serve as a primary site of action for this compound²⁷. The role of mitochondria in ferroptosis was further explored by Gao et al.²⁸, who found that mitochondria play a role in cysteine-deprivation-induced ferroptosis, but they are dispensable for GPX4-inhibition-induced ferroptosis. Additionally, work by Mao et al.²⁹ identified that inhibition of the mitochondrial reductase dihydroorotate dehydrogenase (DHODH) sensitizes cells to ferroptosis, indicating another role for mitochondria in the proliferation of lipid peroxides in ferroptotic death.
当脂质氢过氧化物在细胞膜中积聚时，就会发生^{铁死亡19、20、21} 。这种氢过氧化物积累的增加可能是由于脂质活性氧（ROS）解毒系统的抑制或直接脂质^{ROS的产生12,22,23,24} 。掺入磷脂中的多不饱和脂肪酰基 (PUFA) 部分由于其易于提取的双烯丙基氢而成为铁死亡过程中铁依赖性过氧化的特定底物^{21 , 25 , 26} 。由于 PUFA 部分可以掺入整个细胞的膜中，因此哪些细胞膜特别容易受到铁死亡的影响，或者是铁死亡所必需的问题仍然是个谜。先前的研究探索了抗铁死亡剂铁他汀-1 (fer-1) 的分布，结果表明，虽然 fer-1 定位于内质网 (ER) 膜、线粒体膜和溶酶体膜，但它既不积累于线粒体，也不积累于线粒体中。溶酶体对于活性是必需的，这表明内质网膜可能作为该化合物的主要作用位点²⁷ 。高等人进一步探讨了线粒体在铁死亡中的作用。 ²⁸ ，他们发现线粒体在半胱氨酸剥夺诱导的铁死亡中发挥作用，但它们对于GPX4抑制诱导的铁死亡来说是可有可无的。此外，Mao 等人的工作。²⁹发现抑制线粒体还原酶二氢乳清酸脱氢酶（DHODH）会使细胞对铁死亡敏感，这表明线粒体在铁死亡中脂质过氧化物增殖中的另一个作用。

The plasma membrane (PM) has also been implicated in ferroptosis. The anti-ferroptotic effect of the monounsaturated fatty acid (MUFA) oleic acid was observed to be exerted in conjunction with accumulation in the PM³⁰. Furthermore, the protein FSP1/AIFM2 was discovered to act as a CoQ₁₀-dependent suppressor of ferroptosis, possibly working at the PM^31,32. These prior studies raised the question of which cellular membranes are involved in the ferroptotic death process and how they are interrelated. In the present study, we aimed to elucidate the roles of cellular membranes in ferroptotic death.
质膜（PM）也与铁死亡有关。观察到单不饱和脂肪酸(MUFA)油酸的抗铁死亡作用与PM ³⁰中的积累一起发挥。此外，发现蛋白质 FSP1/AIFM2 可以作为 CoQ ₁₀依赖性铁死亡抑制剂，可能在 PM ^{31 、 32}上发挥作用。这些先前的研究提出了哪些细胞膜参与铁死亡过程以及它们如何相互关联的问题。在本研究中，我们旨在阐明细胞膜在铁死亡中的作用。

To identify essential membranes for ferroptosis, we evaluated the structure–activity–distribution profile of ferroptosis inhibitors and inducers. This new approach involves imaging the subcellular distribution of ferroptosis-modulating compounds and then generating analogs with altered distributions or genetically and pharmacologically modulating the compounds’ sites of accumulation. Two imaging techniques were used in this work: confocal fluorescence microscopy of compounds with fluorescent tags and stimulated Raman scattering (SRS) imaging of Raman-active compounds with high sensitivity and resolution^33,34. Although fluorescence imaging is more commonly employed and highly sensitive, SRS imaging offers the advantage of requiring only minor or even no modifications to the structures of the molecules of interest. Compared to conventional confocal Raman microscopy, SRS offers orders of magnitude higher imaging speed and sensitivity. Furthermore, SRS represents an orthogonal technique to fluorescence imaging, thereby eliminating interference during fluorescence measurement of cellular stains.
为了确定铁死亡的必需膜，我们评估了铁死亡抑制剂和诱导剂的结构-活性-分布概况。这种新方法涉及对铁死亡调节化合物的亚细胞分布进行成像，然后生成分布改变的类似物或通过遗传和药理学调节化合物的积累位点。这项工作使用了两种成像技术：带有荧光标签的化合物的共焦荧光显微镜和具有高灵敏度和分辨率的拉曼活性化合物的受激拉曼散射（SRS）成像^{33 、 34} 。尽管荧光成像更常用且高度敏感，但 SRS 成像具有只需要对感兴趣分子的结构进行少量甚至不需要修改的优点。与传统的共焦拉曼显微镜相比，SRS 的成像速度和灵敏度提高了几个数量级。此外，SRS 代表了一种荧光成像正交技术，从而消除了细胞染色荧光测量过程中的干扰。

The primary ferroptosis-modulating compounds explored in this study were ferroptosis-inhibiting polyunsaturated fatty acids deuterated at their bis-allylic sites (D-PUFAs) and ferroptosis-inducing class IV inducers, including FINO₂ and related compounds.
本研究探索的主要铁死亡调节化合物是双烯丙基位点氘化的铁死亡抑制多不饱和脂肪酸 (D-PUFA) 和诱导铁死亡的 IV 类诱导剂，包括 FINO ₂和相关化合物。

D-PUFAs potently inhibit ferroptosis due to the primary kinetic isotope effect. Bis-allylic hydrogens, situated between two double bonds, are particularly prone to hydrogen atom abstraction, which results in carbon oxygenation and subsequent lipid peroxide formation. In contrast, heavier deuterium atoms at the same sites slow this abstraction sufficiently to block peroxide propagation^35,36. The potent anti-peroxidation activity of these compounds has been demonstrated in vitro as well as in vivo^{17,24,35,36,37,38}. Here, we aimed to use these D-PUFAs as a chemical tool, hypothesizing that the observed sites of accumulation may point to essential protection points against ferroptosis.
由于主要动力学同位素效应，D-PUFA 可有效抑制铁死亡。位于两个双键之间的双烯丙基氢特别容易被氢原子夺取，从而导致碳氧化和随后的脂质过氧化物形成。相反，相同位点的较重氘原子会减慢这种提取速度，足以阻止过氧化物的^{传播35、36} 。这些化合物的有效抗过氧化活性已在体外^{和体内得到证明17,24,35,36,37,38 。}在这里，我们的目标是使用这些 D-PUFA 作为化学工具，假设观察到的积累位点可能指向防止铁死亡的重要保护点。

We also explored the class IV ferroptosis inducer FINO₂, which is an endoperoxide-containing 1,2-dioxolane that induces ferroptosis through oxidation of iron and indirect inactivation of GPX4 (ref. ²⁴). Because FINO₂ is a lipophilic compound, we hypothesized that FINO₂ accumulates in membranes and directly induces PUFA peroxidation in PUFA-containing phospholipids (PUFA-PLs) in those locations. Other ferroptosis inducers target specific proteins and disrupt anti-ferroptotic pathways, meaning that their localization is not directly connected to sites of lipid peroxide formation. However, because FINO₂ directly delivers a peroxide to cells, its subcellular localization is of interest. We, therefore, aimed to determine if FINO₂ targets specific membranes that allow it to exert its pro-ferroptotic action and whether redirection to other subcellular sites would alter the ability of FINO₂-type compounds to induce ferroptosis.
我们还探索了 IV 类铁死亡诱导剂 FINO ₂ ，​​它是一种含有内过氧化物的 1,2-二氧戊环，可通过铁氧化和间接灭活 GPX4 来诱导铁死亡（参考文献²⁴ ）。由于 FINO ₂是一种亲脂性化合物，我们假设 FINO ₂在膜中积累，并直接诱导这些位置的含 PUFA 磷脂 (PUFA-PL) 中的 PUFA 过氧化。其他铁死亡诱导剂靶向特定蛋白质并破坏抗铁死亡途径，这意味着它们的定位与脂质过氧化物形成位点不直接相关。然而，由于 FINO ₂直接将过氧化物传递至细胞，因此其亚细胞定位令人感兴趣。因此，我们的目的是确定 FINO ₂是否靶向特定膜，使其发挥其促铁死亡作用，以及重定向至其他亚细胞位点是否会改变 FINO ₂型化合物诱导铁死亡的能力。

We report here the analysis of subcellular distributions of numerous fatty acids and class IV ferroptosis inducers, finding that lipid peroxidation in the ER membrane and PM is essential to ferroptosis, with a contribution from the mitochondria. We determined that modulating the lipid composition of these membranes alters cell sensitivity to ferroptosis. We found that ferroptosis can be induced through lipid peroxidation in various organelles. Finally, we observed a consistent pattern of lipid ROS spread in cells treated with all four classes of ferroptosis inducers: lipid peroxidation initially propagated through the ER membrane and later accumulated in the PM. Together, these results identify the ER membrane as a key site of ferroptotic lipid peroxidation and an important target of ferroptosis-inducing and ferroptosis-inhibiting compounds.
我们在这里报告了对多种脂肪酸和 IV 类铁死亡诱导剂的亚细胞分布的分析，发现 ER 膜和 PM 中的脂质过氧化对于铁死亡至关重要，线粒体也有贡献。我们确定调节这些膜的脂质组成会改变细胞对铁死亡的敏感性。我们发现多种细胞器中的脂质过氧化可以诱导铁死亡。最后，我们观察到用所有四类铁死亡诱导剂处理的细胞中脂质 ROS 扩散的一致模式：脂质过氧化最初通过 ER 膜传播，后来在 PM 中积累。总之，这些结果表明内质网膜是铁死亡脂质过氧化的关键位点，也是铁死亡诱​​导和抑制化合物的重要靶点。

Deuterated PUFAs accumulate perinuclearly and in puncta

D-PUFAs will potently inhibit ferroptosis due to the primary kinetic isotope effect—heavier deuterium atoms at these sites slow the abstraction sufficiently to block peroxide propagation^35,36. We reasoned that identifying the subcellular sites of accumulation of D-PUFAs would reveal essential sites of lipid peroxidation in ferroptotic death.
由于主要动力学同位素效应，D-PUFA 将有效抑制铁死亡 - 这些位点较重的氘原子会减慢提取速度，足以阻止过氧化物的传播^{35 、 36} 。我们推断，识别 D-PUFA 积累的亚细胞位点将揭示铁死亡中脂质过氧化的重要位点。

We evaluated the potency and distribution of three different D-PUFAs: arachidonic acid-d₆ (ARA-d₆), eicosapentaenoic acid-d₈ (EPA-d₈) and docosahexaenoic acid-d₁₀ (DHA-d₁₀) (Fig. 1a). Treatment with 20 µM of ARA-d₆, EPA-d₈ or DHA-d₁₀ potently prevents death by all four classes of ferroptosis inhibitors (Fig. 1b), and concentrations as low as 1 µM were sufficient to cause some suppression of cell death (Extended Data Fig. 1a).
我们评估了三种不同 D-PUFA 的效力和分布：花生四烯酸- d ₆ (ARA- d ₆ )、二十碳五烯酸- d ₈ (EPA- d ₈ ) 和二十二碳六烯酸- d ₁₀ (DHA- d ₁₀ )（图1a ）。用 20 µM ARA- d ₆ 、EPA -d ₈或 DHA -d ₁₀治疗可有效防止所有四类铁死亡抑制剂导致的死亡（图1b ），低至 1 µM 的浓度足以对细胞产生一定程度的抑制死亡（扩展数据图1a ）。

a, Structures of D-PUFAs. b, Rescue of HT-1080 cells treated with each of four classes of ferroptosis inducers after a 24-hour pre-treatment with D-PUFAs. Data are represented as mean ± s.e.m., n = 3. c, SRS images of HT-1080 cells treated for 24 hours with 20 µM DHA-d₁₀. The protein (CH3) and lipid (CH2) cell-intrinsic vibrational frequencies are shown as well for comparison. Red arrow points to lipid droplets; white arrow points to perinuclear accumulation. d, SRS imaging of HT-1080 cells treated for 24 hours with the three different D-PUFAs used in this study: ARA-d₆ (80 µM), EPA-d₈ (20 µM) and DHA-d₁₀ (20 µM). Red arrow points to lipid droplets; white arrow points to perinuclear accumulation. e, SRS imaging of Panc-1, N27 and HT-22 cells treated for 24 hours with 20 µM DHA-d₁₀.
a ，D-PUFA 的结构。 b ，用 D-PUFA 预处理 24 小时后，用四类铁死亡诱导剂处理的 HT-1080 细胞的拯救。数据表示为平均值±sem， n = 3。c ，用20 µM DHA- d ₁₀处理24 小时的HT-1080 细胞的SRS 图像。还显示了蛋白质 (CH3) 和脂质 (CH2) 细胞固有振动频率以进行比较。红色箭头指向脂滴；白色箭头指向核周积累。 d ，用本研究中使用的三种不同的 D-PUFA 处理 24 小时的 HT-1080 细胞的 SRS 成像：ARA -d ₆ (80 µM)、EPA -d ₈ (20 µM) 和 DHA -d ₁₀ (20 µM) ）。红色箭头指向脂滴；白色箭头指向核周积累。 e ，用 20 µM DHA- d ₁₀处理 24 小时的 Panc-1、N27 和 HT-22 细胞的 SRS 成像。

Source data  源数据

Full size image  全尺寸图像

To determine the subcellular sites of accumulation of D-PUFAs, we made use of SRS imaging^33,34. Because the vibration of C–D bonds is Raman active in a cell-silent region, it is possible to directly image these C–D-containing D-PUFAs with high specificity, sensitivity and spatial resolution inside living cells. Imaging of DHA-d₁₀ in HT-1080 fibrosarcoma cells revealed perinuclear accumulation (Fig. 1c, white arrow) decorated with bright puncta (Fig. 1c, red arrow). This distribution was also observed for ARA-d₆ and EPA-d₈ (Fig. 1d) as well as in other cell types (Fig. 1e), including Panc-1 (a human pancreatic cancer line), N27 (a rat dopaminergic neuronal line) and HT-22 (a mouse hippocampal neuronal line) cells. Thus, D-PUFAs have a consistent and predominant perinuclear and punctal localization. This concentration of each of these PUFAs was sufficient to inhibit ferroptosis by all four classes of inducers, indicating that the areas of accumulation were candidates for primary activity sites.
为了确定 D-PUFA 积累的亚细胞位点，我们利用了 SRS 成像^{33 、 34} 。由于 C-D 键的振动在细胞沉默区域具有拉曼活性，因此可以在活细胞内以高特异性、灵敏度和空间分辨率直接对这些含有 C-D 的 D-PUFA 进行成像。 HT-1080 纤维肉瘤细胞中的 DHA -d ₁₀成像显示核周积聚（图1c ，白色箭头），并饰有明亮的斑点（图1c ，红色箭头）。这种分布在 ARA- d ₆和 EPA -d ₈中也观察到（图1d ）以及其他细胞类型（图1e ），包括 Panc-1（人类胰腺癌细胞系）、N27（大鼠多巴胺能细胞）神经元系）和 HT-22（小鼠海马神经元系）细胞。因此，D-PUFA 具有一致且占主导地位的核周和泪点定位。这些PUFA中每一种的浓度足以抑制所有四类诱导剂的铁死亡，表明积累区域是主要活性位点的候选区域。

D-PUFAs in lipid droplets do not inhibit ferroptosis
脂滴中的 D-PUFA 不会抑制铁死亡

We hypothesized that the puncta observed by SRS imaging of D-PUFAs were either lipid droplets or lysosomes. Cells treated with DHA-d₁₀ were stained with LysoTracker to label lysosomes and Nile Red to label lipid droplets. These experiments revealed that DHA-d₁₀ accumulates in lipid droplets but not in lysosomes (Fig. 2a). We did not observe D-PUFA accumulation in peroxisomes (Extended Data Fig. 1b). Lipidomic analysis of cells treated with DHA-d₁₀ supported localization in lipid droplets, as the PUFA was incorporated into triacylglycerides (TAGs) and cholesterol esters (CEs), which are stored in lipid droplets (Fig. 2b).
我们假设 D-PUFA 的 SRS 成像观察到的斑点是脂滴或溶酶体。用DHA -d ₁₀处理的细胞用LysoTracker染色以标记溶酶体，用尼罗红染色以标记脂滴。这些实验表明，DHA -d ₁₀在脂滴中积累，但不在溶酶体中积累（图2a ）。我们没有观察到 D-PUFA 在过氧化物酶体中积累（扩展数据图1b ）。用 DHA -d ₁₀处理的细胞的脂质组学分析支持脂滴中的定位，因为 PUFA 被掺入到储存在脂滴中的三酰基甘油酯 (TAG) 和胆固醇酯 (CE) 中（图2b ）。

a, SRS and fluorescence imaging of HT-1080 cells treated with 20 µM DHA-d₁₀ and then stained with Nile Red (shown in green for consistency with SRS image) and LysoTracker Green (in red). b, Heat map showing relative incorporation of deuterated DHA into HT-1080 triglycerides, as measured by LC–MS. Vehicle showed no deuterated incorporation, whereas DHA-d₁₀ had varying levels of incorporation into TAGs and CEs, with different sum total numbers of carbons and double bonds. Data shown are an average of absolute signal intensity of three biological replicates. c, SRS and fluorescence imaging of HT-1080 cells treated with DHA-d₁₀ (20 µM) ± co-treatment with DGAT inhibitors PF-06424439 (1 µM) and A922500 (1 µM) and then stained with Nile Red. d, Rescue of HT-1080 cells from erastin or RSL3 lethality with pre-treatment of DHA-d₁₀ ± co-treatment of DGAT inhibitors. Data are represented as mean ± s.e.m., along with individual data points, n = 3. Statistics were performed using two-sided unpaired t-test. LC–MS, liquid chromatography–mass spectrometry.
a ，用 20 µM DHA- d ₁₀处理然后用尼罗红（以绿色显示，与 SRS 图像一致）和 LysoTracker Green（红色）染色的 HT-1080 细胞的 SRS 和荧光成像。 b ，热图显示了通过 LC-MS 测量的氘代 DHA 相对掺入 HT-1080 甘油三酯的情况。媒介物没有表现出氘代掺入，而 DHA -d ₁₀具有不同水平的 TAG 和 CE 掺入，具有不同的碳和双键总数。显示的数据是三个生物重复的绝对信号强度的平均值。 c ，用DHA -d ₁₀ (20 µM)±与DGAT抑制剂PF-06424439 (1 µM)和A922500 (1 µM)共同处理然后用尼罗红染色的HT-1080细胞的SRS和荧光成像。 d ，通过DHA -d ₁₀ ± DGAT抑制剂的预处理，从erastin或RSL3致死中拯救HT-1080细胞。数据表示为平均值±sem，以及各个数据点， n = 3。使用两侧不配对t检验进行统计。 LC-MS，液相色谱-质谱法。

Source data  源数据

Full size image  全尺寸图像

We next aimed to determine whether the accumulation in lipid droplets was relevant to the anti-ferroptotic activity of D-PUFAs. To inhibit lipid droplet formation, synthesis of TAGs was blocked by pharmacologically inhibiting diglyceride acyltransferase (DGAT) enzymes. A co-treatment with 1 µM of inhibitors of each DGAT isozyme (PF-06424439 and A922500) completely abated lipid droplet accumulation in cells treated with exogenous PUFAs, as demonstrated by SRS imaging (Fig. 2c). Inhibition of lipid droplet formation by DGAT inhibitors had no impact on the anti-ferroptotic potency of DHA-d₁₀ (Fig. 2d), demonstrating that accumulation in lipid droplets does not play a role in the ability of D-PUFAs to prevent ferroptosis and, moreover, that lipid droplets are not needed for ferroptosis.
接下来我们的目的是确定脂滴中的积累是否与 D-PUFA 的抗铁死亡活性相关。为了抑制脂滴形成，通过药理学抑制甘油二酯酰基转移酶 (DGAT) 来阻断 TAG 的合成。如 SRS 成像所示，与每种 DGAT 同工酶（PF-06424439 和 A922500）的 1 µM 抑制剂共同治疗完全减少了用外源 PUFA 处理的细胞中的脂滴积累（图2c ）。 DGAT 抑制剂对脂滴形成的抑制对 DHA -d ₁₀的抗铁死亡效力没有影响（图2d ），表明脂滴中的积累对 D-PUFA 预防铁死亡的能力没有作用。此外，铁死亡不需要脂滴。

D-PUFAs incorporate primarily into ER phospholipids
D-PUFA 主要融入 ER 磷脂中

We next aimed to determine the identity of the perinuclear site of accumulation of D-PUFAs. Upon inhibition of lipid droplet formation, it became clear that D-PUFAs were likely accumulating in the ER membrane. This hypothesis was confirmed by SRS imaging of cells treated with DHA-d₁₀ (with and without co-treatment of DGAT inhibitors) and stained with ER-Tracker Green, a fluorescent ER label (Fig. 3a). To determine whether D-PUFAs were additionally accumulating in the Golgi body membrane, cells treated with DHA-d₁₀ were co-stained with ER-Tracker and BODIPY TR Ceramide, a fluorescent Golgi stain. The resulting correlative fluorescence and SRS images enhanced by x–z and y–z axis images revealed that accumulation of D-PUFAs occurs primarily in the ER membrane, with little co-localization of DHA-d₁₀ with the Golgi stain (Fig. 3b, white arrows). Finally, lipidomic analysis showed incorporation of DHA-d₁₀ into phospholipids (Fig. 3c). There was marked incorporation into phosphatidylethanolamine (PE) phospholipids and ether phospholipids. Ferroptosis has been reported to involve PE phospholipids²⁰; thus, incorporation of D-PUFAs into PEs with an observed ER distribution potentially explains the potent protective effect of these fatty acids. With regard to ether phospholipids, plasmalogens have been implicated in ferroptosis, with observations that ether phospholipids containing PUFAs are substrates of lipid peroxidation during ferroptosis³⁹ and that plasmalogen synthesis proteins are necessary for some fatty-acid-induced ferroptosis⁴⁰.
我们接下来的目标是确定 D-PUFA 积累的核周位点的身份。在抑制脂滴形成后，很明显 D-PUFA 可能会在内质网膜上积聚。该假设通过对用 DHA -d ₁₀处理的细胞（有或没有 DGAT 抑制剂的共同处理）进行 SRS 成像得到证实，并用 ER-Tracker Green（一种荧光 ER 标记）染色（图3a ）。为了确定 D-PUFA 是否另外在高尔基体膜中积累，用 ER-Tracker 和 BODIPY TR Ceramide（一种荧光高尔基体染色剂）对用 DHA -d ₁₀处理的细胞进行共染色。通过x – z和y – z轴图像增强的相关荧光和 SRS 图像显示，D-PUFA 的积累主要发生在 ER 膜中，DHA- d ₁₀与高尔基染色剂几乎没有共定位（图3b） ，白色箭头）。最后，脂质组学分析显示DHA -d ₁₀掺入磷脂中（图3c ）。明显掺入磷脂酰乙醇胺（PE）磷脂和醚磷脂。据报道，铁死亡与 PE 磷脂有关²⁰ ；因此，将 D-PUFA 掺入具有观察到的 ER 分布的 PE 中可能解释了这些脂​​肪酸的有效保护作用。 关于醚磷脂，缩醛磷脂与铁死亡有关，观察发现含有 PUFA 的醚磷脂是铁死亡过程中脂质过氧化的底物³⁹ ，并且缩醛磷脂合成蛋白是某些脂肪酸诱导的铁死亡所必需的⁴⁰ 。

a, SRS and fluorescence imaging of HT-1080 cells treated with DHA-d₁₀ (20 µM) ± co-treatment with DGAT inhibitors PF-06424439 (1 µM) and A922500 (1 µM) and then stained with ER-Tracker Green. b, SRS and fluorescence imaging of HT-1080 cells treated with DHA-d₁₀ (20 µM) and then stained with ER-Tracker Green and BODIPY TR Ceramide (a Golgi stain). White arrows indicate the Golgi region. c, Heat map showing incorporation of deuterated DHA into HT-1080 phospholipids as measured by LC–MS. Vehicle shows no deuterated incorporation, whereas DHA-d₁₀ has varying incorporation depending on the lipid species and fatty acid compositions. PC, phosphatidylcholine; PI, phosphatidylinositol; PS, phospatidyserine. Data shown are an average of absolute signal intensity of three biological replicates. LC–MS, liquid chromatography–mass spectrometry.
a ，用 DHA -d ₁₀ (20 µM) ± 与 DGAT 抑制剂 PF-06424439 (1 µM) 和 A922500 (1 µM) 共同处理，然后用 ER-Tracker Green 染色的 HT-1080 细胞的 SRS 和荧光成像。 b ，HT-1080 细胞经 DHA- d ₁₀ (20 µM) 处理，然后用 ER-Tracker Green 和 BODIPY TR Ceramide（高尔基染色剂）染色的 SRS 和荧光成像。白色箭头表示高尔基体区域。 c ，热图显示通过 LC-MS 测量，氘化 DHA 掺入 HT-1080 磷脂中。媒介物未显示出氘代掺入，而 DHA -d ₁₀具有不同的掺入，具体取决于脂质种类和脂肪酸组成。 PC、磷脂酰胆碱； PI，磷脂酰肌醇； PS，磷脂酰丝氨酸。显示的数据是三个生物重复的绝对信号强度的平均值。 LC-MS，液相色谱-质谱法。

Source data  源数据

Full size image  全尺寸图像

Given the nuanced role of mitochondria in ferroptosis and the proximity of the ER and mitochondria, we sought to determine incorporation of D-PUFAs into mitochondria. Initial images did not show overlap with mitochondria, and higher-resolution SRS imaging showed a distribution pattern more consistent with ER localization than mitochondrial localization (Extended Data Fig. 1c,d). Due to the proximity of the ER and mitochondria in these images, however, it was difficult to accurately quantify ER versus mitochondrial incorporation definitively. We, therefore, attempted an orthogonal approach, by treating cells with D-PUFA and purifying ER and mitochondrial membranes. We quantified relative incorporation of D-PUFA using high-resolution mass spectrometry (Extended Data Fig. 1d,e) and found that there was a significantly greater normalized incorporation into the ER as compared to the mitochondria. Notably, we observed contamination of the mitochondrial fraction with ER as measured by the ER-resident protein PDI on western blot, suggesting that the true D-PUFA level in the mitochondria is even lower. We conclude that the perinuclear localization of D-PUFAs is primarily in the ER, but we cannot rule out minor incorporation into the mitochondria.
鉴于线粒体在铁死亡中的微妙作用以及 ER 和线粒体的邻近性，我们试图确定 D-PUFA 是否掺入线粒体中。初始图像没有显示与线粒体重叠，并且更高分辨率的SRS成像显示与ER定位比线粒体定位更一致的分布模式（扩展数据图1c，d ）。然而，由于这些图像中 ER 和线粒体的距离很近，因此很难准确量化 ER 与线粒体掺入的关系。因此，我们尝试采用正交方法，用 D-PUFA 处理细胞并纯化 ER 和线粒体膜。我们使用高分辨率质谱法量化了 D-PUFA 的相对掺入量（扩展数据图1d,e ），发现与线粒体相比，内质网中的标准化掺入量明显更高。值得注意的是，我们通过蛋白质印迹中 ER 驻留蛋白 PDI 测量，观察到线粒体部分被 ER 污染，这表明线粒体中真正的 D-PUFA 水平甚至更低。我们得出的结论是，D-PUFA 的核周定位主要位于内质网中，但我们不能排除少量掺入线粒体中的可能性。

We made extensive attempts to modulate the ER through genetic and pharmacological means. None of the approaches explored, however, resulted in significant or persistent changes in ER size or composition. Knockdown of lipid metabolism genes, including acyl-CoA synthase (ACSL) genes and 1-acylglycerol-3-phosphate O-acyltransferase 3 (AGPAT3), did cause decreases in D-PUFA potency (Extended Data Fig. 2a–c), but these effects were not consistent. A decrease in PUFA incorporation was not reliably detected under these conditions (Extended Data Fig. 2d). We attempted to alter incorporation of exogenous lipids into the ER with various compounds, including thimerosal, a lysophosphatidylcholine acyltransferase (LPCAT) inhibitor⁴¹; miltefosine, a phosphatidylcholine synthesis inhibitor⁴²; and N-ethylmaleimide, a phospholipase A2 (PLA2) activator⁴³. None of these modulators was effective, however (Extended Data Fig. 3a–c). We also tried to increase the ER size by inhibiting ER-phagy pathways and to decrease ER area by directing organelle-depleting proteins to the ER. Knockdown of ER-phagy proteins FAM134B, SEC62 and RTN3 (ref. ⁴⁴) did result in an increase sensitivity to FINs, but there was no detectable expansion of ER volume (Extended Data Fig. 3d–f). Considering that PLA2G16 and FUN14 domain-containing 1 (FUNDC1) both play roles in targeted organelle degradation^45,46, we overexpressed these proteins on the surface of the ER to induce ER-phagy. GFP, PLA2G16, FUNDC1 and the active cytosolic N-terminal region of FUNDC1 were directed to the ER using a cytochrome B5 tag (Extended Data Fig. 4a–c), resulting in no detectable change in ER size. We concluded that modulation of ER size and lipid composition is challenging, and perhaps future developments will allow more straightforward and applicable manipulation of ER surface area and content.
我们进行了广泛的尝试，通过遗传和药理学手段来调节 ER。然而，所探索的方法都没有导致 ER 大小或组成发生显着或持续的变化。脂质代谢基因的敲除，包括酰基辅酶A合酶（ ACSL ）基因和1-酰基甘油-3-磷酸O-酰基转移酶3（ AGPAT3 ），确实导致D-PUFA效力下降（扩展数据图2a-c ），但是这些效果并不一致。在这些条件下，没有可靠地检测到 PUFA 掺入量的减少（扩展数据图2d ）。我们尝试用各种化合物改变外源脂质掺入内质网，包括硫柳汞，一种溶血磷脂酰胆碱酰基转移酶 (LPCAT) 抑制剂⁴¹ ；米替福辛，一种磷脂酰胆碱合成抑制剂⁴² ；和N-乙基马来酰亚胺，一种磷脂酶 A2 (PLA2) 激活剂⁴³ 。然而，这些调制器均无效（扩展数据图3a-c ）。我们还尝试通过抑制内质网吞噬途径来增加内质网大小，并通过将细胞器消耗蛋白引导至内质网来减少内质网面积。 ER 吞噬蛋白 FAM134B、SEC62 和 RTN3（参考文献⁴⁴ ）的敲低确实导致对 FIN 的敏感性增加，但没有检测到 ER 体积的扩张（扩展数据图3d-f ）。考虑到 PLA2G16 和 FUN14 结构域包含 1 (FUNDC1) 都在靶向细胞器降解中发挥作用^{45 , 46} ，我们在内质网表面过表达这些蛋白以诱导内质网自噬。 使用细胞色素 B5 标签将 GFP、PLA2G16、FUNDC1 和 FUNDC1 的活性胞质 N 端区域定向到 ER（扩展数据图4a-c ），导致 ER 大小没有可检测到的变化。我们的结论是，内质网大小和脂质成分的调节具有挑战性，也许未来的发展将允许更直接和适用的内质网表面积和含量的操纵。

Modulating ER and PM lipids alters ferroptotic sensitivity
调节 ER 和 PM 脂质可改变铁死亡敏感性

We next sought to explore if the distribution observed for D-PUFAs remained consistent for other fatty acids relevant to ferroptosis. The fatty acids examined included pro-ferroptotic PUFAs and anti-ferroptotic MUFAs. We hypothesized that, if we enriched the ER membrane with these fatty acids, we would observe corresponding changes in sensitivity to ferroptosis.
接下来，我们试图探讨观察到的 D-PUFA 的分布是否与铁死亡相关的其他脂肪酸保持一致。检查的脂肪酸包括促铁死亡的PUFA和抗铁死亡的MUFA。我们假设，如果我们用这些脂肪酸丰富内质网膜，我们就会观察到铁死亡敏感性的相应变化。

To determine the distribution of PUFAs and MUFAs, SRS imaging was again employed. Fatty acids labeled with deuterium atoms were again used—PUFAs ARA-d₁₁ and DHA-d₅, deuterated at non-bis-allylic sites (and, therefore, pro-ferroptotic with abstractable bisallylic hydrogens), and MUFAs oleic-d₁₇ acid (OA-d₁₇) and palmitoleic-d₁₃ acid (PA-d₁₃) (which are anti-ferroptotic due to their lack of bisallylic sites as they harbor a monounsaturated fatty acyl moiety) (Fig. 4a). All four fatty acids displayed the same distribution pattern as the anti-ferroptotic D-PUFAs (Fig. 4b). As a control, the distributions of two other lipids were tested as well: myristic-d₂₇ acid (MA-d₂₇) and cholesterol-d₆. MA-d₂₇, a saturated fatty acid (SFA), accumulated into the ER and lipid droplets like other fatty acids, demonstrating that this distribution appears to be common to various fatty acid types, including SFAs that do not influence ferroptosis. Cholesterol-d₆ did not accumulate in the ER, demonstrating that the ER is not simply the default site of accumulation for any lipid (Extended Data Fig. 5a–c). Solution Raman spectra of all these compounds showed an observable C–D peak (Extended Data Fig. 5d).
为了确定 PUFA 和 MUFA 的分布，再次采用 SRS 成像。再次使用用氘原子标记的脂肪酸——PUFA ARA- d ₁₁和 DHA -d ₅ ，在非双烯丙基位点氘化（因此，具有可提取的双烯丙基氢的促铁焦性），以及 MUFA 油酸-d ₁₇酸(OA- d ₁₇ ) 和棕榈油酸- d ₁₃酸 (PA- d ₁₃ )（其中由于它们缺乏双烯丙​​基位点（因为它们含有单不饱和脂肪酰基部分），因此具有抗铁焦亡性（图4a ）。所有四种脂肪酸均显示出与抗铁死亡 D-PUFA 相同的分布模式（图4b ）。作为对照，还测试了其他两种脂质的分布：肉豆蔻酸-d ₂₇酸（MA- d ₂₇ ）和胆固醇-d ₆ 。 MA- d ₂₇是一种饱和脂肪酸 (SFA)，像其他脂肪酸一样积累到 ER 和脂滴中，表明这种分布似乎是各种脂肪酸类型所共有的，包括不影响铁死亡的 SFA。胆固醇-d ₆不会在 ER 中积聚，这表明 ER 不仅仅是任何脂质的默认积聚位点（扩展数据图5a-c ）。所有这些化合物的溶液拉曼光谱均显示出可观察到的 C-D 峰（扩展数据图5d ）。

a, Structures of deuterated fatty acids. Of note, the PUFAs are deuterated at non-bis-allylic positions and, therefore, do not inhibit ferroptosis. b, SRS images of HT-1080 cells treated with each of the deuterated fatty acids. c, Effect of pre-treatment (24 hours) with fatty acids on ferroptosis induced by RSL3 in HT-1080 cells, as compared to ethanol vehicle. Data are represented as mean ± s.e.m., n = 3. Two-sided unpaired t-test was performed with P values of 0.00001, 0.000053, 0.010389 and 0.012767. d, Increase in sensitivity to ferroptosis in cells overexpressing GFP-ACSL4 when treated with equivalent dose of PUFAs, as compared to parental HT-1080 cells. Data are represented as mean ± s.e.m., n = 3. Two-way ANOVA with Tukey’s multiple comparisons test was performed with P values of 0.5037, 0.0004, <0.0001, <0.0001, <0.0001, <0.0001, <0.0001, 0.0001 and 0.0033. e, Representative imaging of cells stained with ER-Tracker Green and FM 4-64 to label ER and PM, respectively, and the resulting masks generated in MATLAB. f, Quantification of relative fatty acid incorporation as C–D signal to general lipid CH₂ signal ratio within the PM and ER masks as shown in e. Data are represented as mean ± s.e.m., with each point representing an individual cell. Two-sided unpaired t-test resulted in ARA n = 37 P < 0.0001, OA n = 41 P < 0.0001 and MA n = 26 P = 0.016. For all panels, GraphPad Prism P value style of 0.1234 (NS), 0.0332 (*), 0.0021 (**), 0.0002 (***) and <0.0001 (****) was used. FA, fatty acid; NS, not significant; WT, wild-type.
a ，氘化脂肪酸的结构。值得注意的是，PUFA 在非双烯丙基位置被氘化，因此不会抑制铁死亡。 b ，用每种氘化脂肪酸处理的 HT-1080 细胞的 SRS 图像。 c ，与乙醇载体相比，脂肪酸预处理（24小时）对HT-1080细胞中RSL3诱导的铁死亡的影响。数据表示为平均值±sem， n = 3。进行两侧不配对t检验， P值为0.00001、0.000053、0.010389和0.012767。 d ，与亲代HT-1080细胞相比，当用等剂量的PUFA处理时，过度表达GFP-ACSL4的细胞对铁死亡的敏感性增加。数据表示为平均值±sem， n = 3。使用Tukey多重比较检验进行双向方差分析， P值为0.5037、0.0004、<0.0001、<0.0001、<0.0001、<0.0001、<0.0001、0.0001和0.0033。 e ，分别用 ER-Tracker Green 和 FM 4-64 染色以标记 ER 和 PM 的细胞的代表性成像，以及在 MATLAB 中生成的结果掩模。 f ，相对脂肪酸掺入的量化，作为 PM 和 ER 掩模内的 C-D 信号与一般脂质 CH ₂信号的比率，如e所示。数据表示为平均值±sem，每个点代表一个单独的细胞。两侧不配对t检验结果为 ARA n = 37 P < 0.0001，OA n = 41 P < 0.0001 和 MA n = 26 P = 0.016。对于所有面板，使用 GraphPad Prism P值样式为 0.1234 (NS)、0.0332 (*)、0.0021 (**)、0.0002 (***) 和 <0.0001 (****)。 FA，脂肪酸； NS，不显着； WT，野生型。

Source data  源数据

Full size image  全尺寸图像

Once the subcellular distributions of these lipids were assessed, their effects on ferroptotic death were evaluated. As expected, pre-treatment with the MUFAs OA-d₁₇ or PA-d₁₃ potently inhibited RSL3-induced ferroptosis, whereas PUFAs ARA-d₁₁ and DHA-d₅ significantly increased cell death (Fig. 4c). By contrast, myristic acid and cholesterol did not significantly alter sensitivity to ferroptosis (Extended Data Fig. 5e). To increase incorporation of PUFAs into membranes via increased intracellular PUFA-CoA concentration, an HT-1080 ACSL4 overexpression cell line was generated (Extended Data Fig. 5f). HT-1080 ACSL4 OE cells exhibited increased sensitivity to ferroptosis, as well as an increased pro-ferroptotic effect of pre-treatment with PUFAs (Fig. 4d), confirming that enriching PUFAs in the ER enhanced ferroptosis. We also explored the possibility of localization of ACSL4 playing a role in PUFA distribution. Immunofluorescence imaging of HT-1080 cells stained for ACSL4, however, showed a distribution diffusely throughout cells (Extended Data Fig. 5g), with overlap with the ER, labeled by calnexin staining. A western blot of ER and mitochondrial fractions from HT-1080 cells identified ACSL4 in both fractions (Extended Data Fig. 5h), suggesting that ACSL4 is spread throughout the cell.
一旦评估了这些脂质的亚细胞分布，就评估了它们对铁死亡的影响。正如所预期的，用MUFA OA- d ₁₇或PA -d ₁₃预处理有效抑制RSL3诱导的铁死亡，而PUFA ARA- d ₁₁和DHA -d ₅显着增加细胞死亡（图4c ）。相比之下，肉豆蔻酸和胆固醇并没有显着改变对铁死亡的敏感性（扩展数据图5e ）。为了通过增加细胞内PUFA-CoA浓度来增加PUFA掺入膜中，产生了HT-1080 ACSL4过表达细胞系（扩展数据图5f ）。 HT-1080 ACSL4 OE细胞表现出对铁死亡的敏感性增加，以及用PUFA预处理的促铁死亡效果增加（图4d ），证实了ER中丰富的PUFA增强了铁死亡。我们还探讨了 ACSL4 在 PUFA 分布中发挥作用的本地化可能性。然而，ACSL4 染色的 HT-1080 细胞的免疫荧光成像显示在整个细胞中广泛分布（扩展数据图5g ），与通过钙联蛋白染色标记的 ER 重叠。来自 HT-1080 细胞的 ER 和线粒体组分的蛋白质印迹鉴定出两个组分中的 ACSL4（扩展数据图5h ），表明 ACSL4 遍布整个细胞。

Because it has been observed that inhibition of OA incorporation into the PM correlated with diminished anti-ferroptotic effect of this MUFA³⁰, we sought to determine the degree of accumulation of MUFAs and PUFAs in the PM. Perhaps due to the narrowness of the PM in image cross-sections, as well as the lower magnification of the SRS images, no fatty acids were initially detected in the PM. Higher magnification, however, revealed PM incorporation. Staining with ER-Tracker Green and FM 4-64 (PM) was used to generate ER and PM masks in MATLAB (Fig. 4e), which were then applied as masks to the SRS images of the deuterated FAs and used to quantify relative incorporation as compared to the general SRS lipid signal. We found that there was significantly more PUFA, MUFA and SFA incorporation into the ER than the PM, although relative levels were closer in the case of myristic acid (Fig. 4f).
因为已经观察到抑制 OA 掺入 PM 与该 MUFA ³⁰的抗铁死亡作用减弱相关，所以我们试图确定 PM 中 MUFA 和 PUFA 的积累程度。也许是由于 PM 图像横截面较窄，以及 SRS 图像放大倍数较低，最初在 PM 中未检测到脂肪酸。然而，更高的放大倍数显示出颗粒物的掺入。使用 ER-Tracker Green 和 FM 4-64 (PM) 染色在 MATLAB 中生成 ER 和 PM 掩模（图4e ），然后将其用作氘化 FA 的 SRS 图像的掩模，并用于量化相对掺入与一般 SRS 脂质信号相比。我们发现，与PM相比，ER中掺入的PUFA、MUFA和SFA明显更多，尽管肉豆蔻酸的相对水平更接近（图4f ）。

Class IV ferroptosis inducers localize to ER and Golgi
IV 类铁死亡诱导剂定位于内质网和高尔基体

To complement our work evaluating the key sites of ferroptosis-relevant fatty acid accumulation, we also explored the class IV ferroptosis inducer FINO₂ (1), which is an endoperoxide-containing 1,2-dioxolane that induces ferroptosis through oxidation of iron and indirect inactivation of GPX4 (ref. ²⁴). Because FINO₂ is a lipophilic compound, we hypothesized that FINO₂ accumulates in membranes and directly induces PUFA peroxidation in PUFA-PLs in those locations. We, therefore, aimed to determine if FINO₂ targets specific membranes that allow it to exert its pro-ferroptotic action and whether redirection to other subcellular sites would alter the ability of FINO₂-type compounds to induce ferroptosis.
为了补充我们评估铁死亡相关脂肪酸积累的关键位点的工作，我们还探索了 IV 类铁死亡诱导剂 FINO ₂ ( 1 )，它是一种含有内过氧化物的 1,2-二氧戊环，可通过铁的氧化和间接作用来诱导铁死亡。 GPX4 失活（参考文献²⁴ ）。由于 FINO ₂是一种亲脂性化合物，我们假设 FINO ₂在膜中积累，并直接诱导这些位置的 PUFA-PL 中的 PUFA 过氧化。因此，我们的目的是确定 FINO ₂是否靶向特定膜，使其发挥其促铁死亡作用，以及重定向至其他亚细胞位点是否会改变 FINO ₂型化合物诱导铁死亡的能力。

To identify the subcellular distribution of FINO₂, labeled analogs were prepared. Peroxide FINO₂-1 (2) resembles FINO₂ but contains a fluorescent naphthalimide moiety (Fig. 5a) that can be visualized by fluorescence microscopy. Analog FINO₂-0 (5) does not contain the endoperoxide moiety essential for biological activity²⁴, but it does contain the naphthalimide moiety (Extended Data Fig. 6a). FINO₂-1 was able to induce ferroptosis with similar potency as FINO₂ (Fig. 5b), whereas FINO₂-0 did not induce ferroptosis (Extended Data Fig. 6b), confirming that the endoperoxide moiety is necessary and that the naphthalene group does not exhibit toxicity.
为了鉴定FINO ₂的亚细胞分布，制备了标记的类似物。过氧化物FINO ₂ -1 ( 2 ) 类似于FINO ₂但含有荧光萘酰亚胺部分（图5a ），可以通过荧光显微镜观察。类似物FINO ₂ -0 ( 5 )不包含生物活性所必需的内过氧化物部分²⁴ ，但它确实包含萘酰亚胺部分（扩展数据图6a ）。 FINO ₂ -1 能够诱导铁死亡，其效力与 FINO ₂相似（图5b ），而 FINO ₂ -0 则不诱导铁死亡（扩展数据图6b ），证实内过氧化物部分是必要的，并且萘基团不表现出毒性。

a, Structures of FINO₂ and analog FINO₂-1 with an added fluorescent naphthalimide label. b, Dose–response curves of FINO₂ and FINO₂-1 ± fer-1 (2 µM). Data are represented as mean ± s.e.m., n = 3. c, Confocal fluorescence imaging of HT-1080 cells treated with FINO₂-1 (3 µM) and fer-1 (3 µM) for 3 hours, co-stained with ER-Tracker Red. d, Confocal fluorescence imaging of HT-1080 cells treated with FINO₂-1 (3 µM) and fer-1 (3 µM) for 3 hours, co-stained with BODIPY TR Ceramide. e, Structures of FINO₂ analogs FINO₂-3 and FINO₂-4. f, Confocal fluorescence imaging of HT-1080 cells treated with FINO₂-3 (3 µM) and fer-1 (3 µM) for 3 hours, co-stained with MitoTracker Red CMXRos. g. Confocal fluorescence imaging of HT-1080 cells treated with FINO₂-4 (3 µM) and fer-1 (3 µM) for 3 hours, co-stained with LysoTracker Red. h, Rescue of HT-1080 cells treated for 24 hours with FINO₂-1 (2.5 µM), FINO₂-3 (5 µM) or FINO₂-4 (5 µM) by DFO (left, 10 µM co-treatment), fer-1 (center, 2 µM co-treatment) or DHA-d₁₀ (right, 20 µM 24-hour pre-treatment). Data are represented as mean ± s.e.m., n = 3 biologically independent samples.
a ，FINO ₂和添加荧光萘酰亚胺标记的类似物 FINO ₂ -1 的结构。 b ，FINO ₂和 FINO ₂ -1 ± fer-1 (2 µM) 的剂量反应曲线。数据表示为平均值±sem， n = 3。 c ，用 FINO ₂ -1 (3 µM) 和 fer-1 (3 µM) 处理 3 小时的 HT-1080 细胞的共聚焦荧光成像，与 ER- 共染色追踪者红色。 d ，用 FINO ₂ -1 (3 µM) 和 fer-1 (3 µM) 处理 3 小时的 HT-1080 细胞的共聚焦荧光成像，与 BODIPY TR 神经酰胺共染色。 e ，FINO ₂类似物FINO ₂ -3和FINO ₂ -4的结构。 f ，用 FINO ₂ -3 (3 µM) 和 fer-1 (3 µM) 处理 3 小时的 HT-1080 细胞的共聚焦荧光成像，与 MitoTracker Red CMXRos 共染色。克。用 FINO ₂ -4 (3 µM) 和 fer-1 (3 µM) 处理 3 小时的 HT-1080 细胞的共焦荧光成像，并与 LysoTracker Red 共染色。 h ，用 DFO FINO ₂ -1 (2.5 µM)、FINO ₂ -3 (5 µM) 或 FINO ₂ -4 (5 µM) 处理 24 小时的 HT-1080 细胞的拯救（左图，10 µM 共处理） 、fer-1（中心，2 µM 共同处理）或 DHA -d ₁₀ （右侧，20 µM 24 小时预处理）。数据表示为平均值±sem， n = 3 个生物学独立样本。

Source data  源数据

Full size image  全尺寸图像

Next, confocal fluorescence imaging was performed to determine the distributions of these compounds. Because FINO₂-1 kills cells within a few hours of treatment, co-treatment with 3 µM of fer-1 was used to prevent cell death. As this co-treatment did not alter the distribution of the compound, co-treatment with fer-1 was used for the remainder of these analog imaging experiments (Extended Data Fig. 6c). FINO₂-1 displayed a perinuclear distribution that was confirmed to be the ER membrane (Fig. 5c). FINO₂-0 had the same distribution (Extended Data Fig. 6d). Accumulation in the Golgi membrane was identified as well (Fig. 5d).
接下来，进行共焦荧光成像以确定这些化合物的分布。由于 FINO ₂ -1 会在处理后数小时内杀死细胞，因此与 3 µM fer-1 共同处理可防止细胞死亡。由于这种共同处理没有改变化合物的分布，因此与 fer-1 的共同处理用于这些模拟成像实验的其余部分（扩展数据图6c ）。 FINO ₂ -1 显示核周分布，被确认为ER膜（图5c ）。 FINO ₂ -0 具有相同的分布（扩展数据图6d ）。还鉴定出在高尔基体膜中的积累（图5d ）。

To ensure that ER/Golgi membrane localization of FINO₂-1 was not due strictly to the addition of the fluorescent moiety, another analog, FINO₂-2 (6), was synthesized, which contained an SRS active diyne group instead of the naphthalimide fragment (Extended Data Fig. 6e). FINO₂-2 induced ferroptosis with similar potency as FINO₂-1 and FINO₂, and SRS imaging revealed that FINO₂-2 exhibited the same distribution as FINO₂-1 with some additional puncta (Extended Data Fig. 6f,g). Co-staining with LysoTracker and Nile Red was performed, showing that the puncta were neutral lipid bodies, likely induced by the higher concentrations necessary to detect the compound with SRS imaging (Extended Data Fig. 6g).
为了确保 FINO ₂ -1 的 ER/高尔基膜定位并非严格归因于荧光部分的添加，合成了另一种类似物 FINO ₂ -2 ( 6 )，其包含 SRS 活性二炔基团而不是萘酰亚胺片段（扩展数据图6e ）。 FINO ₂ -2 诱导的铁死亡具有与 FINO ₂ -1 和 FINO ₂类似的效力，并且 SRS 成像显示 FINO ₂ -2 表现出与 FINO ₂ -1 相同的分布，但有一些额外的斑点（扩展数据图6f,g ）。与 LysoTracker 和尼罗红进行共染色，显示斑点是中性脂质体，可能是由 SRS 成像检测化合物所需的较高浓度诱导的（扩展数据图6g ）。

Finally, given the localization of fatty acids to the PM, we sought to determine if class IV ferroptosis inducers, such as FINO₂, accumulate to any extent in the PM. Cells treated with FINO₂-1 were co-stained with CellMask Deep Red to stain the PM and imaged (Extended Data Fig. 6h). FINO₂-1 did not show any overlap with CellMask, demonstrating that accumulation in the PM is not essential for initiation of ferroptosis by class IV inducers.
最后，考虑到脂肪酸在 PM 中的定位，我们试图确定 IV 类铁死亡诱导剂（例如 FINO ₂ ）是否在 PM 中积累到任何程度。用FINO ₂ -1 处理的细胞与CellMask Deep Red 共染色以对PM 进行染色并成像（扩展数据图6h ）。 FINO ₂ -1 没有显示出与 CellMask 的任何重叠，这表明 PM 中的积累对于 IV 类诱导剂引发铁死亡并不是必需的。

Ferroptosis can be directly induced in various organelles
多种细胞器可直接诱导铁死亡

Imaging of FINO₂-1 and FINO₂-0 revealed that delivery of an endoperoxide moiety to the ER and Golgi results in ferroptotic cell death. We wondered if redistributing an endoperoxide to other organelles would induce ferroptosis.
FINO ₂ -1 和 FINO ₂ -0 的成像显示，将内过氧化物部分递送至内质网和高尔基体会导致铁死亡细胞。我们想知道将内过氧化物重新分配到其他细胞器是否会引起铁死亡。

Experiments with a series of analogs of FINO₂ demonstrated that protection of the ER is key to inhibiting ferroptosis, regardless of where in the cell it was initiated. These analogs were synthesized to redistribute the FINO₂ endoperoxide to other subcellular sites (Extended Data Fig. 7a). Although compounds FINO₂-5 (7) and FINO₂-6 (8) were successfully redirected to lysosomal membranes, they induced a non-ferroptotic cell death (Extended Data Fig. 7b,c). Compound FINO₂-7 (9) did not change distribution as compared to FINO₂-1 (Extended Data Fig. 7d). In contrast, FINO₂-3 (3) and FINO₂-4 (4) (Fig. 5e) successfully induced ferroptosis while being redistributed to the mitochondrial membranes and the lysosomal membranes, respectively (Fig. 5f,g). FINO₂-3 and FINO₂-4 did not accumulate in the PM when examined on higher magnification (Extended Data Fig. 6i) and retained similar lethal potency as FINO₂-1 (Extended Data Fig. 7e). Both FINO₂-3 and FINO₂-4 induced ferroptotic cell death that could be rescued by ferroptosis inhibitors with varying mechanisms and distributions: fer-1, deferoxamine and DHA-d₁₀ (Fig. 5h). Rescue by both DHA-d₁₀ and fer-1 points to the ER and mitochondria as essential sites of protection. As it has been demonstrated that mitochondria are not essential to ferroptosis^27,28, this finding implicates the ER as the key essential site of protection.
使用一系列 FINO ₂类似物进行的实验表明，保护 ER 是抑制铁死亡的关键，无论铁死亡是在细胞的哪个部位开始的。合成这些类似物是为了将 FINO ₂内过氧化物重新分配到其他亚细胞位点（扩展数据图7a ）。尽管化合物FINO ₂ -5 ( 7 ) 和FINO ₂ -6 ( 8 ) 成功地重定向至溶酶体膜，但它们诱导非铁死亡细胞死亡（扩展数据图7b，c ）。与FINO ₂ -1 相比，化合物FINO ₂ -7 ( 9 ) 没有改变分布（扩展数据图7d ）。相反，FINO ₂ -3 ( 3 ) 和 FINO ₂ -4 ( 4 ) (图5e ) 成功诱导铁死亡，同时分别重新分布到线粒体膜和溶酶体膜 (图5f,g )。当在更高的放大倍数下检查时（扩展数据图6i ），FINO ₂ -3 和FINO ₂ -4 没有在PM中积聚，并且保留了与FINO ₂ -1相似的致死效力（扩展数据图7e ）。 FINO ₂ -3 和FINO ₂ -4 均诱导铁死亡，可以通过具有不同机制和分布的铁死亡抑制剂来挽救：fer-1、去铁胺和DHA -d ₁₀ （图5h ）。 DHA -d ₁₀和 fer-1 的救援表明 ER 和线粒体是重要的保护位点。 由于已经证明线粒体对于铁死亡不是必需的^{27 , 28} ，这一发现表明 ER 是关键的重要保护位点。

To further characterize the role of mitochondria, we sought to determine how treatment with these different FINO₂ analogs would impact sensitivity of cells to inhibition of the mitochondrial reductase DHODH. Inhibition of DHODH with the compound brequinar (BQR) induced ferroptosis in HT-1080 cells and caused an increase in sensitivity to RSL3, as previously demonstrated by Mao et al.²⁹ (Extended Data Fig. 8a). We then co-treated cells with a fixed amount of FINO₂-1, FINO₂-3 or FINO₂-4 and increasing doses of BQR (Extended Data Fig. 8b). FINO₂-1 demonstrated a greater degree of sensitization, suggesting that the combination of simultaneously oxidizing the ER and mitochondria more effectively induces ferroptosis.
为了进一步表征线粒体的作用，我们试图确定这些不同的 FINO ₂类似物处理如何影响细胞对线粒体还原酶 DHODH 抑制的敏感性。使用化合物 brequinar (BQR) 抑制 DHODH 可诱导 HT-1080 细胞铁死亡，并导致对 RSL3 的敏感性增加，正如 Mao 等人先前所证明的那样。 ²⁹ （扩展数据图8a ）。然后，我们用固定量的FINO ₂ -1、FINO ₂ -3 或FINO ₂ -4 以及增加剂量的BQR共同处理细胞（扩展数据图8b ）。 FINO ₂ -1 表现出更大程度的致敏作用，表明同时氧化 ER 和线粒体的组合更有效地诱导铁死亡。

Ferroptosis results in peroxidation of ER followed by PM
铁死亡导致 ER 过氧化，随后导致 PM

To further define the roles that different membranes play in the ferroptotic death cascade, HT-1080 cells were treated with inducers from each of the four ferroptosis classes and imaged at various timepoints to capture cell death progression. Cells were stained with C11 BODIPY to detect lipid peroxidation and organelle stains for subcellular localization.
为了进一步确定不同膜在铁死亡级联中发挥的作用，用四种铁死亡类别的诱导剂处理 HT-1080 细胞，并在不同时间点成像以捕获细胞死亡进程。细胞用 C11 BODIPY 染色以检测脂质过氧化，并用细胞器染色进行亚细胞定位。

In cells treated with ferroptosis inducers, we observed ER peroxidation initially, followed later by PM peroxidation (Fig. 6a,b). The peroxidation of each membrane was associated with morphological changes, including a decrease in ER membrane area and a narrowing and ballooning of the plasma membrane. ER peroxidation occurs within 2 hours for cells treated with RSL3, FIN56 and FINO₂. At this stage, there is not significant observable change in PM morphology (Fig. 6c). By 5 hours, most cells exhibited lipid peroxidation within the PM (Fig. 6d). Imidazole ketone erastin (IKE) induces a slower death than the other ferroptosis inducers; ER peroxidation was observed by 6 hours, with widespread PM peroxidation by 10 hours (Extended Data Fig. 8c,d).
在用铁死亡诱导剂处理的细胞中，我们首先观察到ER过氧化，随后观察到PM过氧化（图6a，b ）。每个膜的过氧化与形态变化相关，包括内质网膜面积的减少以及质膜的变窄和膨胀。对于用 RSL3、FIN56 和 FINO ₂处理的细胞，内质网过氧化在 2 小时内发生。在此阶段，PM形态没有明显可观察到的变化（图6c ）。 5小时后，大多数细胞在PM内表现出脂质过氧化（图6d ）。咪唑酮erastin (IKE) 比其他铁死亡诱导剂诱导的死亡速度更慢； 6 小时后观察到 ER 过氧化，10 小时后观察到广泛的 PM 过氧化（扩展数据图8c、d ）。

a, Quantification of C11 BODIPY oxidized:reduced ratio within the ER and PM of HT-1080 cells treated with DMSO, RSL3 (0.5 µM), FIN56 (10 µM) or FINO₂ (10 µM) at 2 hours and 5 hours. CellMask Deep Red (PM) and ER-Tracker Blue-White (ER) were used to select regions of interest in CellProfiler within which C11 BODIPY signal was quantified. Data are represented as mean ± s.e.m., with each point representing a single cell. Sample sizes are 2 hours ER (n = 18, 19, 25 and 18) and PM (n = 18, 19, 25 and 14) and 5 hours ER (n = 21, 24, 28 and 36) and PM (n = 21, 22, 28 and 36). Brown–Forsythe and Welch ANOVA with Dunnett T3 test for multiple comparisons was used with P values of: 2 hours ER (<0.0001, 0.0007 and <0.0001), 2 hours PM (0.6485, 0.9049 and 0.1531), 5 hours ER (<0.0001, <0.0001 and <0.0001) and 5 hours PM (<0.0001, <0.0001 and <0.0001). b, Quantification of C11 BODIPY oxidized:reduced ratio within the ER and PM of HT-1080 cells treated with DMSO or IKE (10 µM) at 6 hours and 10 hours. CellMask Deep Red (PM) and ER-Tracker Blue-White (ER) were used to select regions of interest in CellProfiler within which C11 BODIPY signal was quantified. Data are represented as mean ± s.e.m., with each point representing a single cell. Sample sizes are 6 hours ER (n = 20 and 32) and PM (n = 14 and 23) and 10 hours ER (n = 20 and 24) and PM (n = 14 and 19). Two-sided unpaired t-test was used with P values of: 6 hours ER (<0.0001), 6 hours PM (0.4385), 10 hours ER (<0.0001) and 10 hours PM (<0.0001). c, Representative images of HT-1080 cells treated with DMSO, RSL3 (0.5 µM), FIN56 (10 µM) or FINO₂ (10 µM) for 2 hours and stained with C11 BODIPY (oxidized and reduced overlay), CellMask Deep Red and ER-Tracker Blue-White. d, Representative images of HT-1080 cells treated with DMSO, RSL3 (0.5 µM), FIN56 (10 µM) or FINO₂ (10 µM) for 5 hours and stained with C11 BODIPY (oxidized and reduced overlay), CellMask Deep Red and ER-Tracker Blue-White. For all panels, GraphPad Prism P value style of 0.1234 (NS), 0.0332 (*), 0.0021 (**), 0.0002 (***) and <0.0001 (****) was used. NS, not significant.
a ，用 DMSO、RSL3 (0.5 µM)、FIN56 (10 µM) 或 FINO ₂ (10 µM) 处理 2 小时和 5 小时的 HT-1080 细胞的 ER 和 PM 内 C11 BODIPY 氧化：还原比率的定量。 CellMask Deep Red (PM) 和 ER-Tracker Blue-White (ER) 用于在 CellProfiler 中选择感兴趣的区域，在其中对 C11 BODIPY 信号进行定量。数据表示为平均值±sem，每个点代表一个细胞。样本量为 2 小时 ER（ n = 18、19、25 和 18）和 PM（ n = 18、19、25 和 14）以及 5 小时 ER（ n = 21、24、28 和 36）和 PM（ n = 21、22、28 和 36）。使用带有 Dunnett T3 检验的 Brown–Forsythe 和 Welch 方差分析进行多重比较， P值为：2 小时 ER（<0.0001、0.0007 和 <0.0001）、2 小时 PM（0.6485、0.9049 和 0.1531）、5 小时 ER（<0.0001） ，<0.0001 且<0.0001）和下午 5 小时（<0.0001、<0.0001 和 <0.0001）。 b ，用 DMSO 或 IKE (10 µM) 处理 6 小时和 10 小时的 HT-1080 细胞的 ER 和 PM 内 C11 BODIPY 氧化：还原比率的定量。 CellMask Deep Red (PM) 和 ER-Tracker Blue-White (ER) 用于在 CellProfiler 中选择感兴趣的区域，在其中对 C11 BODIPY 信号进行定量。数据表示为平均值±sem，每个点代表一个细胞。样本大小为 6 小时 ER（ n = 20 和 32）和 PM（ n = 14 和 23）以及 10 小时 ER（ n = 20 和 24）和 PM（ n = 14 和 19）。使用双边未配对t检验， P值为：6 小时 ER (<0.0001)、6 小时 PM (0.4385)、10 小时 ER (<0.0001) 和 10 小时 PM (<0.0001)。 c ，用 DMSO、RSL3 (0.5 µM)、FIN56 (10 µM) 或 FINO ₂ (10 µM) 2 小时，并用 C11 BODIPY（氧化和还原覆盖）、CellMask 深红和 ER-Tracker 蓝白染色。 d ，用 DMSO、RSL3 (0.5 µM)、FIN56 (10 µM) 或 FINO ₂ (10 µM) 处理 5 小时并用 C11 BODIPY（氧化和还原覆盖）、CellMask 深红和染色的 HT-1080 细胞的代表性图像ER-追踪器蓝白色。对于所有面板，使用 GraphPad Prism P值样式为 0.1234 (NS)、0.0332 (*)、0.0021 (**)、0.0002 (***) 和 <0.0001 (****)。 NS，不显着。

Source data

Full size image

To evaluate lipid peroxidation in the mitochondria, we compared correlation of oxidized C11 BODIPY signal to ER-Tracker versus MitoTracker. C11 BODIPY oxidized signal had a significantly higher correlation with the ER initially for all four FINs at the intial stages of death (Extended Data Fig. 8e). We performed the same quantification for cells treated with RSL3, BQR and RSL3 + BQR and observed more relative mitochondrial peroxidation with BQR as expected but, interestingly, still significantly greater peroxidation in the ER earlier in death (Extended Data Fig. 8f). Altogether, these data point to the ER as a primary essential target of lipid peroxidation in ferroptosis, with the peroxidation of the mitochondria and PM as a later stage of death.
为了评估线粒体中的脂质过氧化，我们比较了氧化 C11 BODIPY 信号与 ER-Tracker 和 MitoTracker 的相关性。对于死亡初始阶段的所有四个 FIN，C11 BODIPY 氧化信号与最初的 ER 具有显着更高的相关性（扩展数据图8e ）。我们对用 RSL3、BQR 和 RSL3 + BQR 处理的细胞进行了相同的定量，并观察到与预期相比，BQR 具有更多相对的线粒体过氧化作用，但有趣的是，死亡早期 ER 中的过氧化作用仍然显着更高（扩展数据图8f ）。总而言之，这些数据表明 ER 是铁死亡中脂质过氧化的主要靶标，线粒体和 PM 的过氧化是死亡的后期阶段。

Previous research pointed to the ER membrane as a possible primary site of protection by fer-1 against lipid ROS to prevent ferroptotic death²⁷. Here we demonstrated that directly introducing lipid ROS into the ER membrane is sufficient to initiate ferroptotic death, with our observation that the ER is the subcellular membrane target of FINO₂, a class IV ferroptosis inducer. It is not a requirement that lipid ROS be introduced exclusively to the ER membrane; however, we found that initiating lipid ROS in the mitochondrial or lysosomal membranes can still induce ferroptotic cell death, as demonstrated by the FINO₂ analogs FINO2-3 and FINO2-4, which localize to different subcellular compartments. Protecting lysosomes or mitochondria from lipid ROS with a radical trapping agent was not effective for blocking ferroptosis²⁷. At the same time, inhibition of DHODH is known to fuel ferroptotic death, indicating a relevant contribution of mitochondrial ROS. Interestingly, we found that inhibiting DHODH in combination with induction of lipid ROS in the ER membrane with FINO₂-1 resulted in augmented cell death. Altogether, these findings indicate that ROS in the lysosomes and mitochondria can initiate and drive ferroptosis.
先前的研究指出，ER 膜可能是 fer-1 对抗脂质 ROS 的主要保护位点，以防止铁死亡²⁷ 。在这里，我们证明直接将脂质 ROS 引入 ER 膜足以引发铁死亡，我们观察到 ER 是 FINO ₂ （一种 IV 类铁死亡诱导剂）的亚细胞膜靶标。不要求将脂质 ROS 专门引入 ER 膜；然而，我们发现在线粒体或溶酶体膜中启动脂质 ROS 仍然可以诱导铁死亡细胞死亡，正如 FINO ₂类似物 FINO2-3 和 FINO2-4 所证明的那样，它们定位于不同的亚细胞区室。使用自由基捕获剂保护溶酶体或线粒体免受脂质 ROS 的影响对于阻断铁死亡并不能有效²⁷ 。同时，已知抑制 DHODH 会促进铁死亡，表明线粒体 ROS 的相关贡献。有趣的是，我们发现抑制 DHODH 并结合 FINO ₂ -1 在内质网膜中诱导脂质 ROS 会导致细胞死亡增加。总而言之，这些发现表明溶酶体和线粒体中的 ROS 可以启动并驱动铁死亡。

Our findings showed that lipid droplets do not play a significant role in D-PUFA-mediated protection against ferroptosis. Recently published work by Dierge et al.⁴⁷, however, found that exogenous n-3 and n-6 PUFAs potentiated ferroptosis in acidic cancer cells and that this effect is magnified through inhibition of protective shunting of PUFAs into lipid droplets with DGAT inhibitors. Therefore, although inhibition of lipid droplets may not influence sensitivity to ferroptosis in all cases, or play a role in the activity of anti-ferroptotic FAs, perturbation of lipid droplet metabolism may have a role in modulating ferroptosis in certain contexts.
我们的研究结果表明，脂滴在 D-PUFA 介导的铁死亡保护中不起重要作用。 Dierge 等人最近发表的作品。然而， ⁴⁷发现外源性 n-3 和 n-6 PUFA 增强了酸性癌细胞中的铁死亡，并且通过用 DGAT 抑制剂抑制 PUFA 保护性分流到脂滴中，这种效应被放大。因此，尽管脂滴的抑制可能不会在所有情况下影响对铁死亡的敏感性，或者在抗铁死亡 FA 的活性中发挥作用，但在某些情况下，脂滴代谢的扰动可能在调节铁死亡中发挥作用。

Beyond lipid droplets, we identified the ER as the primary site of PUFA incorporation, with less incorporation into the PM or mitochondria. These results point to the ER, the largest membrane in the cell containing a high relative PUFA concentration, as a key target of lipid peroxidation in ferroptosis. We then aimed to deepen our understanding of the subcellular dynamics of lipid peroxidation to disentangle the roles of these membranes in influencing ferroptosis. For all ferroptosis inducers, PM lipid peroxidation was observed by C11-BODIPY 3–4 hours after ER lipid peroxidation was observed. This finding is consistent with previous observations by Magtanong et al.³⁰, who found that treatment with erastin2, an erastin analog, resulted in PM peroxidation 10 hours after treatment, as observed in our time-course for IKE. Our data indicate that ER peroxidation happens early in ferroptotic death, with mitochondrial and PM peroxidation as later events.
除了脂滴之外，我们还确定内质网是 PUFA 掺入的主要位点，很少掺入到 PM 或线粒体中。这些结果表明内质网（细胞中最大的膜，含有相对较高的多不饱和脂肪酸浓度）是铁死亡中脂质过氧化的关键目标。然后，我们的目标是加深对脂质过氧化亚细胞动力学的理解，以阐明这些膜在影响铁死亡中的作用。对于所有铁死亡诱导剂，在观察到 ER 脂质过氧化反应后 3-4 小时，通过 C11-BODIPY 观察到 PM 脂质过氧化反应。这一发现与 Magtanong 等人之前的观察结果一致。 ³⁰ ，他们发现用erastin2（一种erastin类似物）治疗会在治疗后10小时导致PM过氧化，正如我们在IKE时间过程中观察到的那样。我们的数据表明，ER 过氧化发生在铁死亡的早期，线粒体和 PM 过氧化是较晚的事件。

Our model proposes that the ER membrane is a key site of lipid peroxidation in ferroptosis induced by class I through class IV compounds. This finding couples with the previously observed activation of the ER stress-related unfolded protein response in ferroptosis and further underscores the role this organelle plays in ferroptosis^48,49. Lipid peroxidation may spread from the ER to other membranes, or peroxidation of membranes may occur independently at different stages and with different rates. If lipid peroxidation is spreading, how it spreads is not clear. Vesicular transport or membrane contact sites are both good hypotheses and warrant further exploration. It was recently shown by Riegman et al.⁵⁰ that ferroptotic cell lysis is osmotic, and perhaps it is the combination of lipid peroxidation in the cell membrane and the osmotic swelling that results in the final blow of cell death—a membrane that is both stretched and then weakened finally pops.
我们的模型提出，内质网膜是 I 类至 IV 类化合物诱导的铁死亡中脂质过氧化的关键位点。这一发现与之前观察到的铁死亡中内质网应激相关未折叠蛋白反应的激活相结合，进一步强调了该细胞器在铁死亡中所起的作用^{48 , 49} 。脂质过氧化可能从内质网扩散到其他膜，或者膜的过氧化可能在不同阶段以不同速率独立发生。如果脂质过氧化正在扩散，那么它是如何扩散的尚不清楚。囊泡运输或膜接触位点都是很好的假设，值得进一步探索。 Riegman 等人最近展示了这一点。 ⁵⁰铁死亡细胞裂解是渗透性的，也许细胞膜中的脂质过氧化和渗透性膨胀的结合导致了细胞死亡的最后一击——细胞膜先被拉伸，然后又变弱，最终破裂。

In summary, we found that endoperoxide delivery to different organelle membranes can induce ferroptosis and that the ER membrane is a primary target of lipid peroxidation in ferroptotic death. Although mitochondria are not required for ferroptosis, when present they play an important role in susceptibility. We also observe that oxidation of the PM occurs at a later, more final stage of ferroptotic death. These observations dovetail with the cell’s natural systems of protection against lipid peroxidation, with GPX4 serving as protector of the ER membrane and subcellular membranes in general, whereas DHODH protects the mitochondria, and FSP1 protects the PM^29,31,32. The use of SRS and fluorescence imaging to examine the structure–activity–distribution relationship of inducers and inhibitors of ferroptosis demonstrates the power of such a strategy in its ability to demystify the subcellular dynamics of this form of cell death and to provide important information for future development of ferroptosis-modulating drugs.
总之，我们发现内过氧化物递送到不同的细胞器膜可以诱导铁死亡，并且内质网膜是铁死亡中脂质过氧化的主要目标。尽管线粒体不是铁死亡所必需的，但当它们存在时，它们在易感性中发挥着重要作用。我们还观察到 PM 的氧化发生在铁死亡的后期、更最后的阶段。这些观察结果与细胞针对脂质过氧化的天然保护系统相吻合，GPX4 通常充当 ER 膜和亚细胞膜的保护者，而 DHODH 保护线粒体，FSP1 保护 PM ^{29 , 31 , 32} 。使用SRS和荧光成像来检查铁死亡诱导剂和抑制剂的结构-活性-分布关系，证明了这种策略的力量，能够揭开这种细胞死亡形式的亚细胞动力学的神秘面纱，并为未来提供重要信息。铁死亡调节药物的开发。

Cell lines  细胞系

HT-1080 (human (Homo sapiens) male fibrosarcoma), PANC-1 (human (H. sapiens) male pancreatic epithelial fibrosarcoma), HEK293T (human (H. sapiens) male fibrosarcoma), N27 (rat (Rattus norvegicus) dopaminergic neural cell line) and HT22 (mouse (Mus musculus) hippocampal neuronal cell line) were obtained from the American Type Culture Collection (https://www.atcc.org/). HT-1080, HEK293T and HT-22 cells were cultured in DMEM with 10% FBS, 1% non-essential amino acids and 1% penicillin–streptomycin (P–S). PANC-1 cells were cultured in DMEM with 10% FBS and 1% P–S. N27 cells were cultured in RPMI medium with 10% FBS, 2 mM L-glutamine and 1% P–S. All cells were cultured at 37 °C and 5% CO₂.
HT-1080（人（智人）男性纤维肉瘤）、PANC-1（人（智人）男性胰腺上皮纤维肉瘤）、HEK293T（人（智人）男性纤维肉瘤）、N27（大鼠（褐家鼠）多巴胺能神经细胞细胞系）和 HT22（小鼠（ Mus musculus ）海马神经元细胞系）获自美国典型培养物保藏中心（ https://www.atcc.org/ ）。 HT-1080、HEK293T 和 HT-22 细胞在含有 10% FBS、1% 非必需氨基酸和 1% 青霉素-链霉素 (P-S) 的 DMEM 中培养。 PANC-1 细胞在含有 10% FBS 和 1% P-S 的 DMEM 中培养。 N27 细胞在含有 10% FBS、2 mM L-谷氨酰胺和 1% P-S 的 RPMI 培养基中培养。所有细胞均在37℃和5%CO ₂下培养。

Chemicals and reagents

Reagents used in this study include: arachidonic acid-d₆ (Retrotope), eicosapentaenoic acid-d₈ (Retrotope), docosahexaenoic acid-d₁₀ (Retrotope), erastin (Cayman Chemical), IKE (Stockwell laboratory), RSL3 (Stockwell laboratory), FIN56 (gift of Rachid Skouta), FINO₂ (Woerpel laboratory), BQR (Cayman Chemical), PF-06424439 (Cayman Chemical), A922500 (Cayman Chemical), arachidonic acid-d₁₁ (Cayman Chemical), docosahexaenoic acid-d₅ (Cayman Chemical), oleic acid-d₁₇ (Cayman Chemical), palmitoleic acid-d₁₃ (Cayman Chemical), myristic acid-d₂₇ (Sigma-Aldrich), cholesterol-d₆ (Sigma-Aldrich), Thimerosal Ready Made Solution (Sigma-Aldrich), miltefisone (Cayman Chemical), N-ethylmaleimide (Sigma-Aldrich), LysoTracker Green DND-26 (Invitrogen), LysoTracker Red DND-99 (Invitrogen), Nile Red (Invitrogen), ER-Tracker Red (Invitrogen), ER-Tracker Green (Invitrogen), ER-Tracker Blue-White DPX (Invitrogen), BODIPY TR Ceramide complexed to BSA (Invitrogen), MitoTracker Red CMXRos (Invitrogen), Hoechst 33342 (Invitrogen), BODIPY 581/591 C11 (Invitrogen), CellMask Deep Red (Invitrogen) and FM 4-64 (Invitrogen).
本研究中使用的试剂包括：花生四烯酸- d ₆ (Retrotope)、二十碳五烯酸- d ₈ (Retrotope)、二十二碳六烯酸- d ₁₀ (Retrotope)、erastin (Cayman Chemical)、IKE (Stockwell 实验室)、RSL3 (Stockwell 实验室) ）、FIN56（Rachid Skouta 的礼物）、FINO ₂ (Woerpel 实验室)、BQR (Cayman Chemical)、PF-06424439 (Cayman Chemical)、A922500 (Cayman Chemical)、花生四烯酸- d ₁₁ (Cayman Chemical)、二十二碳六烯酸- d ₅ (Cayman Chemical)、油酸- d ₁₇ （开曼化学），棕榈油酸- d ₁₃ （开曼Chemical)、肉豆蔻酸- d ₂₇ (Sigma-Aldrich)、胆固醇- d ₆ (Sigma-Aldrich)、硫柳汞现成溶液 (Sigma-Aldrich)、米替松 (Cayman Chemical)、 N-乙基马来酰亚胺 (Sigma-Aldrich)、LysoTracker绿色 DND-26 (Invitrogen)、LysoTracker 红色 DND-99 (Invitrogen)、尼罗红(Invitrogen)、ER-Tracker Red (Invitrogen)、ER-Tracker Green (Invitrogen)、ER-Tracker Blue-White DPX (Invitrogen)、BODIPY TR 神经酰胺与 BSA 复合 (Invitrogen)、MitoTracker Red CMXRos (Invitrogen)、Hoechst 33342 (Invitrogen)，BODIPY 581/591 C11 (Invitrogen)、CellMask Deep Red (Invitrogen) 和 FM 4-64 (Invitrogen)。

Dose–response assays  剂量反应测定

Cells were seeded in a 384-well plate at 1,500 cells per well. In experiments using D-PUFAs, cells were treated at the time of seeding. After 24-hour incubation at 37 °C/5% CO₂, cells were treated with compounds of interest and incubated for a further 24 hours at 37 °C/5% CO₂. Cell viability was evaluated with CellTiter-Glo (Promega), and results were worked up in GraphPad Prism. All treatments were performed in triplicate.
将细胞以每孔 1,500 个细胞接种到 384 孔板中。在使用 D-PUFA 的实验中，细胞在接种时进行处理。在37℃/5%CO ₂下孵育24小时后，用感兴趣的化合物处理细胞并在37℃/5%CO ₂下再孵育24小时。使用 CellTiter-Glo (Promega) 评估细胞活力，并在 GraphPad Prism 中处理结果。所有处理均一式三份进行。

SRS imaging and analysis  SRS 成像和分析

Cells were seeded on 12-mm circular cover glasses (Thermo Fisher Scientific) in a 24-well plate at 50,000 cells per well. In experiments using deuterated fatty acids, cells were treated at the time of seeding and incubated for 24 hours at 37 °C/5% CO₂. In experiments with FINO₂-2, cells were treated after overnight incubation at 37 °C/5% CO₂. All treatments were performed in duplicate. SRS imaging was performed with a system coupling two spatially and temporally overlapped laser beams from an integrated laser source (picoEmerald, Applied Physics & Electronics) into a commercial confocal laser scanning microscope (FV1200MPE, Olympus). One of the laser beams for SRS is a tunable pump beam (720–990 nm, 5–6 ps); the other is a Stokes beam with fixed wavelength (1,064 nm, 6 ps, intensity modulated at 8 MHz). Both are at 80-MHz repetition rate. The two beams are focused onto the cell samples through a ×25 water objective (XLPlan N, 1.05 NA MP, Olympus), and the transmitted beams are then collected by a high-NA oil condenser lens (1.4 NA, Olympus) for detection. A high O.D. bandpass filter (890/220 CARS, Chroma Technology) is used to block the Stokes beam and leave only the pump beam to be collected by a silicon photodiode (FDS1010, Thorlabs) with a DC voltage of 64 V. The output current of the photodiode was terminated by 50 Ω and demodulated with a high-frequency lock-in amplifier (HF2LI, Zurich Instruments) at 8-MHz frequency. The stimulated Raman loss signal at each pixel is sent to the analog interface box (FV10-ANALOG, Olympus) of the microscope to generate the image. All images (512 × 512 pixels) are acquired with 30-μs time constant at the lock-in amplifier and 100-μs pixel dwell time. The on-sample power for SRS imaging was 80 mW for pump beam and 120 mW for Stokes beam.
将细胞以每孔 50,000 个细胞接种在 24 孔板中的 12 毫米圆形盖玻片 (Thermo Fisher Scientific) 上。在使用氘化脂肪酸的实验中，在接种时处理细胞并在37℃/5%CO ₂下孵育24小时。在使用 FINO ₂ -2 的实验中，在 37 °C/5% CO ₂下孵育过夜后处理细胞。所有处理均一式两份进行。 SRS 成像是通过将来自集成激光源（picoEmerald，应用物理与电子）的两束空间和时间重叠的激光束耦合到商用共焦激光扫描显微镜（FV1200MPE，奥林巴斯）中的系统进行的。 SRS 的激光束之一是可调谐泵浦光束（720–990 nm，5–6 ps）；另一种是固定波长的斯托克斯光束（1,064 nm，6 ps，8 MHz 强度调制）。两者的重复率均为 80 MHz。两束光束通过×25水物镜（XLPlan N，1.05 NA MP，Olympus）聚焦到细胞样品上，然后由高NA油聚光镜（1.4 NA，Olympus）收集透射光束进行检测。高 OD 带通滤波器（890/220 CARS，Chroma Technology）用于阻挡斯托克斯光束，仅留下泵浦光束，由具有 64 V 直流电压的硅光电二极管（FDS1010，Thorlabs）收集。输出电流光电二极管的端接电阻为 50 Ω，并使用频率为 8 MHz 的高频锁定放大器（HF2LI，Zurich Instruments）进行解调。每个像素处的受激拉曼损耗信号被发送到显微镜的模拟接口盒（FV10-ANALOG，奥林巴斯）以生成图像。所有图像（512 × 512 像素）均在锁定放大器处以 30 μs 时间常数和 100 μs 像素驻留时间采集。 SRS 成像的样品上功率对于泵浦光束为 80 mW，对于斯托克斯光束为 120 mW。

For evaluation of incorporation of D-FA into ER and PM, cells treated with 20 μM D-FA were incubated with 1 μM ER-Tracker Green (Invitrogen) in HBSS for 15 minutes at 37 °C and with 10 μg ml⁻¹ FM 4-64 (Invitrogen) in HBSS for 5 minutes at room temperature for ER and PM labeling. The CH₂ on-resonance channel was acquired at 2,850 cm⁻¹, and the CD on-resonance channel was acquired at 2,105 cm⁻¹. Off-resonance channels were acquired at 1,900 cm⁻¹ and were used for subtraction from the on-resonance channel to obtain the pure SRS signal. The PM and ER regions in cells were segmented on the CH₂ on-resonance channel with reference to the corresponding PM and ER fluorescence images using MATLAB R2020a. Concentrated lipid droplets gave saturated SRS signals and were removed from segmented regions. The pure SRS signals for CD and CH₂ were obtained by subtracting the on-resonance image with the off-resonance image at the same condition.
为了评估 D-FA 掺入 ER 和 PM，将用 20 μM D-FA 处理的细胞与 1 μM ER-Tracker Green (Invitrogen) 一起在 HBSS 中于 37°C 和 10 μg ml ^-1 FM 中孵育 15 分钟4-64 (Invitrogen) 在 HBSS 中室温放置 5 分钟以进行 ER 和 PM 标记。 CH ₂同共振通道在2,850 cm ^-1处获取，并且CD同共振通道在2,105 cm ^-1处获取。在1,900 cm ^-1处获取非共振通道，并将其用于从共振通道中减去以获得纯SRS信号。使用 MATLAB R2020a 参考相应的 PM 和 ER 荧光图像，在 CH ₂共振通道上对细胞中的 PM 和 ER 区域进行分割。浓缩的脂滴产生饱和的 SRS 信号，并从分段区域中去除。 CD和CH ₂的纯SRS信号是通过在相同条件下将共振图像与非共振图像相减而获得的。

Confocal fluorescence imaging
共焦荧光成像

Cells were seeded in four-well coverglass chambers (Nunc Lab-Tek) at 150,000 cells per well or eight-well cover glass chambers at 50,000 cells per well. In experiments using fatty acids, cells were treated at the time of seeding and incubated for 24 hours at 37 °C/5% CO₂. In experiments with FINO₂ analogs and FINs, cells were treated after overnight incubation at 37 °C/5% CO₂. After relevant staining as described below, cells were placed in HBSS, and images were taken using Zeiss LSM 700 and 800 confocal microscopes with Zeiss ZEN Blue 2.1 software and then worked up in Fiji/CellProfiler^51,52. For quantification of images, CellProfiler was used. For ER versus PM quantification, ER and PM of each cell were identified as objects using their respective stains. The ratio of the median signal for C11 BODIPY oxidized and reduced channels were then determined for every cell within the identified ER and PM regions, and these were plotted. For ER versus mitochondria quantification, co-localization of C11 BODIPY was used instead owing to the proximity of the ER and mitochondria stains and the small size of the mitochondria making object identification too difficult. The co-localization metrics were measured only for pixels above a certain threshold, which was set as 20% of the maximum intensity of the image to eliminate the background. The Manders coefficient was used to represent the correlation of pixel intensity between C11 BODIPY oxidized and either the ER stain or the mitochondrial stain. For quantification of ER size, objects were identified using adaptive Otsu two-class thresholding, and their size properties were calculated and exported with at least six images used for each sample, resulting in the sample sizes of number of cells defined as n.
将细胞以每孔 150,000 个细胞接种在四孔盖玻片室 (Nunc Lab-Tek) 中，或以每孔 50,000 个细胞接种在八孔盖玻片室中。在使用脂肪酸的实验中，在接种时处理细胞并在37℃/5%CO ₂下孵育24小时。在使用 FINO ₂类似物和 FIN 的实验中，在 37 °C/5% CO ₂下孵育过夜后处理细胞。如下所述进行相关染色后，将细胞置于 HBSS 中，使用 Zeiss LSM 700 和 800 共焦显微镜以及 Zeiss ZEN Blue 2.1 软件拍摄图像，然后在 Fiji/CellProfiler ^{51、52​​中}进行处理。为了量化图像，使用了 CellProfiler。对于 ER 与 PM 定量，使用各自的染色剂将每个细胞的 ER 和 PM 识别为对象。然后确定所识别的 ER 和 PM 区域内每个细胞的 C11 BODIPY 氧化通道和还原通道的中值信号比率，并将这些信号绘制出来。对于 ER 与线粒体定量，使用 C11 BODIPY 的共定位，因为 ER 和线粒体染色剂很接近，并且线粒体尺寸较小，使得对象识别过于困难。共定位指标仅针对高于特定阈值的像素进行测量，该阈值被设置为图像最大强度的 20% 以消除背景。 Manders 系数用于表示氧化 C11 BODIPY 与 ER 染色或线粒体染色之间的像素强度的相关性。 为了量化 ER 大小，使用自适应 Otsu 两类阈值来识别对象，并使用每个样本至少使用六张图像来计算和导出它们的大小属性，从而得到定义为n 的细胞数量的样本大小。

Fluorescent staining of live cells
活细胞的荧光染色

Fluorescent labels were used according to the manufacturer’s instructions. LysoTracker Green DND-26 (Invitrogen) and LysoTracker Red DND-99 (Invitrogen) were diluted to 50 nM in HBSS and incubated for 30 minutes at 37 °C. Nile Red (Invitrogen) was diluted to 1 µg ml⁻¹ in HBSS and incubated for 5 minutes at room temperature. ER-Tracker Green, ER-Tracker Red and ER-Tracker Blue-White (Invitrogen) were diluted to 1 µM in HBSS and incubated for 20 minutes at 37 °C. Golgi stain BODIPY TR Ceramide complexed to BSA (Invitrogen) was diluted to 5 µM and incubated at 37 °C for 30 minutes. MitoTracker Red CMXRos (Invitrogen) was diluted to 50 nM in HBSS and incubated at 30 minutes. Hoechst 33342 was diluted to 1 µg ml⁻¹ and incubated for 5 minutes at room temperature. BODIPY 581/591 C11 (Invitrogen) was diluted to 10 µM and incubated for 10 minutes at 37 °C. CellMask Deep Red Plasma Membrane Stain (Invitrogen) was diluted to 5 µg ml⁻¹ in HBSS and incubated for 10 minutes at 37 °C.
根据制造商的说明使用荧光标签。 LysoTracker Green DND-26 (Invitrogen) 和 LysoTracker Red DND-99 (Invitrogen) 在 HBSS 中稀释至 50 nM，并在 37 °C 下孵育 30 分钟。将尼罗红(Invitrogen)在HBSS中稀释至1μgml ^-1并在室温下孵育5分钟。将 ER-Tracker Green、ER-Tracker Red 和 ER-Tracker Blue-White (Invitrogen) 在 HBSS 中稀释至 1 µM，并在 37 °C 下孵育 20 分钟。将与 BSA (Invitrogen) 复合的高尔基体染色剂 BODIPY TR 神经酰胺稀释至 5 µM，并在 37 °C 下孵育 30 分钟。 MitoTracker Red CMXRos (Invitrogen) 在 HBSS 中稀释至 50 nM，并孵育 30 分钟。将Hoechst 33342稀释至1 µg ml ^-1并在室温下孵育5分钟。 BODIPY 581/591 C11 (Invitrogen) 稀释至 10 µM，并在 37 °C 下孵育 10 分钟。将 CellMask 深红色血浆膜染色剂 (Invitrogen) 在 HBSS 中稀释至 5 µg ml ^-1并在 37 °C 下孵育 10 分钟。

Immunofluorescence imaging
免疫荧光成像

HT-1080 cells were seeded on eight-well chambered slides (Thermo Fisher Scientific) at 0.05 million cells per well and incubated overnight. Cells were washed with PBS twice and fixed with 4% paraformaldehyde for 30 minutes at room temperature in the dark. Cells were than permeabilized by three washes with PBST (PBS with 0.1% Triton X-100). Cells were blocked with 5% normal goat serum in PBST for 1 hour at room temperature. Cells were stained with mouse anti-ACSL4 antibody (Invitrogen) 1:250 and rabbit anti-calnexin antibody (Invitrogen) 1:333 overnight at 4 °C. Cells were washed three times with PBST and stained with Alexa Fluor 594-conjugated anti-mouse (Invitrogen) 1:500 and Alexa Fluor 488-conjugated anti-rabbit antibody (Invitrogen) 1:500 at room temperature for 1 hour. Cells were washed three times with PBST. Slides were slightly dried, one drop of antifade mountant with DAPI (Invitrogen) added, and covered with cover glass. Slides were dried overnight and imaged using a Zeiss LSM 800 confocal microscope.
将 HT-1080 细胞以每孔 50 万个细胞接种在八孔室载玻片（Thermo Fisher Scientific）上，并孵育过夜。用PBS洗涤细胞两次，并用4%多聚甲醛在室温下避光固定30分钟。然后用 PBST（含 0.1% Triton X-100 的 PBS）洗涤 3 次来透化细胞。用 PBST 中的 5% 正常山羊血清在室温下封闭细胞 1 小时。将细胞用小鼠抗 ACSL4 抗体 (Invitrogen) 1:250 和兔抗钙联蛋白抗体 (Invitrogen) 1:333 在 4°C 下染色过夜。用PBST洗涤细胞3次，并用Alexa Fluor 594缀合的抗小鼠(Invitrogen) 1:500和Alexa Fluor 488缀合的抗兔抗体(Invitrogen) 1:500在室温下染色1小时。用PBST洗涤细胞3次。将载玻片稍微干燥，添加一滴含有 DAPI (Invitrogen) 的抗褪色封固剂，并用盖玻片覆盖。将载玻片干燥过夜并使用 Zeiss LSM 800 共焦显微镜成像。

Lipidomics with high-resolution mass spectrometry
高分辨率质谱脂质组学

HT-1080 cells were seeded at 5 million cells per 10-cm dish in triplicate, with either ethanol as vehicle or 20 µM PUFA, and incubated for 24 hours. Cells were harvested, and lipids were isolated and analyzed as previously described^6,53. In brief, cells were homogenized with a microtip sonicator in 250 µl of ice-cold methanol with 0.01% butylated hydroxyltoluene (BHT) and then transferred to glass tubes containing 850 µl of cold methyl-tert-butyl ether (MTBE) and vortexed for 30 seconds. Samples were then incubated on a shaker at 4 °C for 2 hours. Then, 200 µl of cold water was added, and they were incubated on ice for 20 minutes before centrifugation at 3,000 r.p.m. for 20 minutes at 4 °C. The organic layer was collected and dried under a stream of nitrogen gas on ice. Next, the samples were reconstituted in a solution of 2-propanol/acetonitrile/water (4:3:1, v/v/v) containing a splashlipidomix standard (Avanti Polar Lipids). Ultra-performance liquid chromatography was then performed at 55 °C on an Acquity UPLC HSS T3 Column, (1.8 µm, 2.1 mm × 100 mm) over a 20-minute gradient elution. Mobile phase A was acetonitrile/water (60:40, v/v); mobile phase B was 2-propanol/acetonitrile/water (85:10:5, v/v/v), both containing 10 mM ammonium acetate and 0.1% acetic acid. After injection, the gradient was held at 40% mobile phase B for 2 minutes. For the next 12 minutes, the gradient was ramped in a linear fashion to 100% B and held at this composition for 3 minutes. The eluent composition returned to the initial condition in 1 minute, and the column was re-equilibrated for an additional 1 minute before the next injection was conducted.The flow rate was 0.4 ml min⁻¹, and injection volumes were 6 µl. The SYNAPT G2 mass spectrometer (Waters) was operated in both positive and negative electrospray ionization (ESI) modes. All raw data files were converted to netCDF format using the DataBridge tool implemented in MassLynx software (Waters, version 4.1). They were then subjected to peak-picking, retention time alignment and grouping using the XCMS package (version 3.2.0) in the R (version 3.4.4) environment^54,55. After retention time alignment and filling missing peaks, an output data frame was generated containing the list of time-aligned detected features (m/z and retention time) and the relative signal intensity (area of the chromatographic peak) in each sample. All the extracted features were normalized to measured protein concentrations measured by BCA assay (Pierce). Structural assignment and structural characterization of significant lipid features were initially obtained by searching monoisotopic masses against the available online databases, such as METLIN, Lipid MAPS and HMDB, with a mass tolerance of 5 ppm and by confirming fragmentation patterns of HDMSE data in MS^E data viewer (version 1.3, Waters).
将 HT-1080 细胞以每 10 厘米培养皿 500 万个细胞一式三份接种，使用乙醇作为载体或 20 µM PUFA，并孵育 24 小时。如先前所述^{6 、 53}收获细胞并分离和分析脂质。简而言之，将细胞用微尖超声仪在 250 µl 含有 0.01% 丁基化羟基甲苯 (BHT) 的冰冷甲醇中匀浆，然后转移至含有 850 µl 冷甲基叔丁基醚 (MTBE) 的玻璃管中，并涡旋 30秒。然后将样品在 4°C 的摇床上孵育 2 小时。然后加入 200 µl 冷水，在冰上孵育 20 分钟，然后在 4 °C 下以 3,000 rpm 离心 20 分钟。收集有机层并在冰上氮气流下干燥。接下来，将样品在含有splashlipidomix 标准品（Avanti Polar Lipids）的 2-丙醇/乙腈/水（4:3:1，v/v/v）溶液中重构。然后在 55 °C 的 Acquity UPLC HSS T3 色谱柱（1.8 µm，2.1 mm × 100 mm）上进行 20 分钟梯度洗脱，进行超高效液相色谱分析。流动相A为乙腈/水（60:40，v/v）；流动相 B 为 2-丙醇/乙腈/水（85:10:5，v/v/v），均含有 10 mM 乙酸铵和 0.1% 乙酸。注射后，梯度保持在40%流动相B 2分钟。在接下来的 12 分钟内，梯度以线性方式升至 100% B，并在此组成下保持 3 分钟。洗脱液组合物在1分钟内恢复到初始条件，并且在进行下一次注射之前将柱再平衡1分钟。流速为0.4ml min ^-1 ，并且注射体积为6μl。 SYNAPT G2 质谱仪 (Waters) 在正电喷雾电离 (ESI) 模式和负电喷雾电离 (ESI) 模式下运行。使用 MassLynx 软件（Waters，版本 4.1）中实现的 DataBridge 工具将所有原始数据文件转换为 netCDF 格式。然后在 R（版本 3.4.4）环境中使用 XCMS 软件包（版本 3.2.0）对它们进行峰拾取、保留时间对齐和分组^{54 55} 。保留时间对齐并填充缺失峰后，生成一个输出数据帧，其中包含每个样品中时间对齐的检测特征（ m / z和保留时间）和相对信号强度（色谱峰面积）的列表。所有提取的特征均根据 BCA 测定 (Pierce) 测量的蛋白质浓度进行标准化。重要脂质特征的结构分配和结构表征最初是通过在可用的在线数据库（例如 METLIN、Lipid MAPS 和 HMDB）中搜索单同位素质量来获得的，质量公差为 5 ppm，并通过确认 MS ^E数据中 HDMSE 数据的碎片模式来获得查看器（版本 1.3，沃特世）。

Isolation of mitochondrial and ER fractions
线粒体和 ER 组分的分离

Samples were subject to fractionation using commercially available mitochondrial isolation (Thermo Fisher Scientific, 89874) and ER isolation (MilliporeSigma, ER0100-1KT) kits for each respective fraction.
使用市售的线粒体分离（Thermo Fisher Scientific，89874）和 ER 分离（MilliporeSigma，ER0100-1KT）试剂盒对样品进行分级分离。

In brief, for mitochondrial fractionation, cells were spun down at 850g for 2 minutes, resuspended in 1.6 ml of reagent A, vortexed for 5 seconds and incubated on ice for 2 minutes. Cells were homogenized in a 7-ml dounce homogenizer (Bellco Glass); 1.6 ml of reagent C was added; and suspension was centrifuged at 700g for 10 minutes at 4 °C. Supernatant was then transferred to new tubes and spun down at 3,500g for 15 minutes at 4 °C. After spin-down, supernatant was discarded and pellet resuspended in 500 µl of reagent C and centrifuged at 12,000g for 5 minutes at 4 °C. The resulting mitochondrial fraction pellet was isolated, flash frozen and maintained at −80 °C until use.
简而言之，对于线粒体分级分离，将细胞以 850 g离心 2 分钟，重悬于 1.6 ml 试剂 A 中，涡旋 5 秒，并在冰上孵育 2 分钟。将细胞在7ml杜恩斯匀浆器(Bellco Glass)中匀浆；加入1.6ml试剂C；将悬浮液在 4°C 下以 700 g离心 10 分钟。然后将上清液转移至新管中并在 4 °C 下以 3,500 g离心 15 分钟。离心后，弃去上清液，将沉淀重悬于 500 µl 试剂 C 中，并在 4 °C 下以 12,000 g离心 5 分钟。分离所得线粒体组分沉淀，快速冷冻并维持在-80°C直至使用。

For ER fractionation, cells were centrifuged at 600g for 5 minutes at 4 °C and supernatant discarded. Cells were washed with ten volumes of PBS and pelleted again under identical conditions. Cells were resuspended in 3 ml of hypotonic extraction buffer and incubated on ice for 20 minutes. Cells were then centrifuged at 600g for 5 minutes at 4 °C; the supernatant was discarded; and the pellet was resuspended in 2 ml of isotonic extraction buffer. Suspension was homogenized in 7 ml of dounce homogenizer, and homogenate was centrifuged at 1,000g for 10 minutes at 4 °C.
对于 ER 分级分离，将细胞在 4°C 下以 600 g离心 5 分钟，并弃去上清液。用十倍体积的PBS洗涤细胞并在相同条件下再次沉淀。将细胞重悬于3ml低渗提取缓冲液中并在冰上孵育20分钟。然后将细胞在 4°C 下以 600 g离心 5 分钟；弃去上清液；并将沉淀重悬于2ml等渗提取缓冲液中。将悬浮液在 7 ml 杜恩斯匀浆器中匀浆，并将匀浆在 4 °C 下以 1,000 g离心 10 分钟。

The central supernatant above the pellet but below the thin lipid layer on top was isolated by slowly puncturing the top lipid layer, and 1.5 ml of the central supernatant layer was transferred to a new tube and centrifuged again at 12,000g for 15 minutes at 4 °C. Again, 1 ml of the central layer was transferred to a new tube and then to a flat-bottom 250-ml glass beaker with a stir bar inside. Next, 8 mM calcium chloride solution was added dropwise while stirring to the beaker on ice, and, once added, the solution was stirred for an additional 15 minutes. Solution was then subject to centrifugation at 8,000g for 10 minutes at 4 °C; the supernatant was siphoned off; and the pellet was flash frozen and stored at −80 °C until use.
通过缓慢刺破顶部脂质层，分离沉淀上方但顶部薄脂质层下方的中央上清液，并将1.5ml中央上清液层转移至新管中，并在4°下再次以12,000g离心15分钟C.再次将 1 ml 中心层转移至新管中，然后转移至内部装有搅拌棒的平底 250 ml 玻璃烧杯中。接下来，在搅拌的同时将8mM氯化钙溶液滴加到冰上的烧杯中，并且一旦添加，将溶液再搅拌15分钟。然后将溶液在 4°C 下以 8,000 g离心 10 分钟；吸去上清液；将沉淀快速冷冻并储存在-80°C直至使用。

Cloning of overexpression retrovirus plasmids
过表达逆转录病毒质粒的克隆

NEBuilder HiFi DNA Assembly Cloning Kit (New England Biolabs) was used to clone inserts of interest into retroviral expression plasmid pMRX-IP-GFP (gift of Noboru Mizushima)⁴⁵. Gibson assembly primers for genes of interest were designed (Supplementary Table 1), and inserts were amplified by polymerase chain reaction (PCR) using Phusion High-Fidelity Polymerase (New England Biolabs) on a T100 Thermal Cycler (Bio-Rad). For ACSL4 and FUNDC1, insert amplification was performed with pCMV-Entry-ACSL4 (OriGene) and pcDNA4/TO/Myc-His-FUNDC1 (gift of Quan Chen), respectively, and pMRX-IP-GFP-Plaat3 (PLA2G16) was then digested using BamHI and XhoI in CutSmart Buffer (New England Biolabs). To introduce the cytochrome b5 sequence into pMRX-IP-GFP-PLA2G16, two inserts were made using long primers to introduce the new sequence in, resulting in a three-piece assembly that required digesting the vector with BamHI and BsiWI. For PLA2G16-cyb5, the pMRX-IP-GFP-PLA2G16 was used as template and vector for digestion. For FUNDC1-cyb5, the pMRX-IP-GFP-FUNDC1 was used as template, and the same primers as PLA2G16 were used with the exception of FUNDC1-cyb5 R1 substituted for PLA2G16-cyb5 R1. To make the pMRX-IP-GFP control plasmid, the three GFP primers in the table were used along with PLA2G16-cyb5 R2 to amplify the two inserts from pMRX-IP-GFP-PLA2G16, and vector was cut with HindIII and BsiWI. To make pMRX-IP-GFP-cyb5, pMRX-IP-GFP-PLA2G16-cyb5 was amplified with GFP-cyb5 F and PLA2G16-cyb5 R2, and pMRX-IP-GFP was digested with BamHI and BsiWI. To make the N-terminus cyb5-tagged FUNDC1 plasmid, the FUNDC1-cyb5 primers were used with GFP F1 and PLA2G16-cyb5 R2, with pMRX-IP-GFP-FUNDC1 and pMRX-IP-GFP-PLA2G16-cyb5 as templates and with pMRX-IP-GFP digested with HindIII and BsiWI. All PCR and digestion products were gel purified using the Zymoclean Gel Recovery Kit. The HiFi kit was then used according to the manufacturer’s instructions, and the reaction was incubated at 50 °C for 1 hour. The reactions were then transformed into Stable (New England Biolabs) or Stbl3 (Invitrogen) cells and plated on LB/agar plates. Resulting colonies were grown up with QIAprep Spin Miniprep Kit (Qiagen) and sent to GENEWIZ for Sanger sequencing. The primers used for these experiments are detailed in Supplementary Table 1.
NEBuilder HiFi DNA 组装克隆试剂盒（New England Biolabs）用于将感兴趣的插入片段克隆到逆转录病毒表达质粒 pMRX-IP-GFP（Noboru Mizushima 赠送）中⁴⁵ 。设计了感兴趣基因的吉布森组装引物（补充表1 ），并使用Phusion高保真聚合酶（New England Biolabs）在T100热循环仪（Bio-Rad）上通过聚合酶链式反应（PCR）扩增插入片段。对于 ACSL4 和 FUNDC1，分别用 pCMV-Entry-ACSL4 (OriGene) 和 pcDNA4/TO/Myc-His-FUNDC1 (Quan Chen 赠品) 进行插入扩增，然后将 pMRX-IP-GFP-Plaat3 (PLA2G16) 进行扩增。使用 CutSmart Buffer (New England Biolabs) 中的 BamHI 和 XhoI 进行消化。为了将细胞色素 b5 序列引入 pMRX-IP-GFP-PLA2G16，使用长引物进行了两个插入以引入新序列，从而形成需要用 BamHI 和 BsiWI 消化载体的三部分组装体。对于 PLA2G16-cyb5，使用 pMRX-IP-GFP-PLA2G16 作为模板和载体进行消化。对于FUNDC1-cyb5，使用pMRX-IP-GFP-FUNDC1作为模板，并且使用与PLA2G16相同的引物，除了FUNDC1-cyb5 R1替换PLA2G16-cyb5 R1之外。为了制作 pMRX-IP-GFP 对照质粒，将表中的三个 GFP 引物与 PLA2G16-cyb5 R2 一起使用，以扩增 pMRX-IP-GFP-PLA2G16 中的两个插入片段，并用 HindIII 和 BsiWI 切割载体。为了制备 pMRX-IP-GFP-cyb5，用 GFP-cyb5 F 和 PLA2G16-cyb5 R2 扩增 pMRX-IP-GFP-PLA2G16-cyb5，并用 BamHI 和 BsiWI 消化 pMRX-IP-GFP。 为了制作 N 末端 cyb5 标记的 FUNDC1 质粒，将 FUNDC1-cyb5 引物与 GFP F1 和 PLA2G16-cyb5 R2 一起使用，以 pMRX-IP-GFP-FUNDC1 和 pMRX-IP-GFP-PLA2G16-cyb5 作为模板，并使用pMRX-IP-GFP 用 HindIII 和 BsiWI 消化。所有 PCR 和消化产物均使用 Zymoclean 凝胶回收试剂盒进行凝胶纯化。然后根据制造商的说明使用 HiFi 试剂盒，并将反应在 50°C 下孵育 1 小时。然后将反应物转化至稳定(New England Biolabs)或Stbl3(Invitrogen)细胞中并铺于LB/琼脂平板上。所得菌落使用 QIAprep Spin Miniprep Kit (Qiagen) 培养，并发送至 GENEWIZ 进行 Sanger 测序。用于这些实验的引物详见补充表1 。

Generation of overexpression and stable knockdown cell lines by viral infection
通过病毒感染产生过表达和稳定敲低细胞系

For overexpression cell lines, genes were inserted into pMRX-IPU-GFP as described above. For this plasmid, the pUMVC gag-pol plasmid and pVSV-G envelope plasmid were used⁵⁶. For knockdown cell lines, glycerol stocks of Mission shRNA knockdown sequences in pLKO.1-puro were purchased from Sigma-Aldrich (Supplementary Table 3). For this plasmid, the psPAX2 gag-pol plasmid and pVSV-G envelope plasmid were used.
对于过表达细胞系，如上所述将基因插入pMRX-IPU-GFP中。对于该质粒，使用了pUMVC gag-pol 质粒和pVSV-G 包膜质粒⁵⁶ 。对于敲低细胞系，pLKO.1-puro 中 Mission shRNA 敲低序列的甘油储备购自 Sigma-Aldrich（补充表3 ）。对于该质粒，使用psPAX2 gag-pol 质粒和pVSV-G 包膜质粒。

To produce virus, 400,000 HEK293T cells per well were seeded in 2.5 ml in six-well plates and incubated overnight at 37 °C/5% CO₂. To 250 µl of OptiMEM (Gibco), 1.25 µg of knockdown or overexpression plasmid, 1.25 µg of gag-pol plasmid and 0.156 µg of VSV-G plasmid were added. Next, 8 µl of TransIT LT1 (Mirus) was added, and this solution was mixed and incubated at room temperature for 15–30 minutes. The solution was then added dropwise to HEK293T cells and incubated overnight at 37 °C/5% CO₂. Transfection medium was replaced with 5 µl of fresh medium, and cells were incubated at 32 °C/5% CO₂ for 24 hours. Viral supernatant was harvested; 250 µl of 1 M HEPES buffer (Gibco) was added; and the supernatant was filtered through a 0.45-µm syringe filter. Aliquots were then either frozen and stored at −80 °C or applied directly to cells of interest.
为了产生病毒，将每孔400,000个HEK293T细胞接种在2.5ml的六孔板中，并在37℃/5%CO ₂下孵育过夜。向250μl OptiMEM(Gibco)中添加1.25μg敲低或过表达质粒、1.25μg gag-pol质粒和0.156μg VSV-G质粒。接下来，添加 8 µl TransIT LT1 (Mirus)，混合该溶液并在室温下孵育 15-30 分钟。然后将溶液滴加至HEK293T细胞并在37℃/5%CO ₂下孵育过夜。将转染培养基替换为5μl新鲜培养基，并将细胞在32℃/5%CO ₂下孵育24小时。收获病毒上清液；添加 250 µl 1 M HEPES 缓冲液 (Gibco)；上清液通过 0.45 µm 注射器过滤器过滤。然后将等分试样冷冻并储存在-80°C或直接应用于感兴趣的细胞。

To infect cells, viral supernatant with a 1:2,000 dilution of 10 mg ml⁻¹ polybrene (Santa Cruz Biotechnology) was applied to cells of interest seeded in six-well plates and incubated overnight at 37 °C/5% CO₂. Fresh media was then applied, and cells were incubated for 24 hours. Selection media containing 1 µg ml⁻¹ puromycin was then added, and selection/expansion took place. Knockdowns/overexpressions were confirmed by qPCR or western blot as described below.
为了感染细胞，将用10mg ml ^-1聚凝胺(Santa Cruz Biotechnology)以1:2,000稀释的病毒上清液应用于接种在六孔板中的感兴趣的细胞，并在37℃/5%CO ₂下孵育过夜。然后应用新鲜培养基，并将细胞孵育24小时。然后添加含有1μgml ^-1嘌呤霉素的选择培养基，并进行选择/扩增。如下所述，通过 qPCR 或蛋白质印迹确认敲低/过表达。

qPCR  定量PCR

Cells were harvested, washed with PBS and lysed using QIAshredder (Qiagen). RNEasy kit (Qiagen) was used to isolate RNA, and RT–PCR was performed using MultiScribe Reverse Transcriptase (Invitrogen) on a T100 Thermal Cycler (Bio-Rad). qPCR reactions were then performed in triplicate using Power SYBR Green PCR Master Mix (Applied Biosystems) on a ViiA 7 Real-Time PCR instrument (Thermo Fisher Scientific). TBP was used as an internal reference. Differences in mRNA levels compared with TBP were computed between vehicle and experimental groups using the ΔΔCt method. The primers used in the study are detailed in Supplementary Table 2.
收获细胞，用PBS洗涤并使用QIAshredder (Qiagen)裂解。 RNEasy 试剂盒 (Qiagen) 用于分离 RNA，并使用 MultiScribe 逆转录酶 (Invitrogen) 在 T100 热循环仪 (Bio-Rad) 上进行 RT-PCR。然后使用 Power SYBR Green PCR Master Mix (Applied Biosystems) 在 ViiA 7 实时 PCR 仪器 (Thermo Fisher Scientific) 上一式三份进行 qPCR 反应。 TBP 用作内参。使用ΔΔCt方法计算媒介物组和实验组之间与TBP相比mRNA水平的差异。研究中使用的引物详见补充表2 。

Western blotting  蛋白质印迹法

Cells were harvested, washed with PBS and incubated in RIPA buffer containing cOmplete, Mini Protease Inhibitor Tablets (Sigma-Aldrich) on ice for 30 minutes. These were then centrifuged at 14,000g for 15 minutes at 4 °C. Protein concentrations of supernatants were determined using the BCA Protein Assay Kit (Pierce), and concentrations were normalized and diluted with 4× SDS running buffer (Invitrogen). NuPAGE Novex 4–12% Bis-Tris Midi Protein Gel (Invitrogen) was loaded at run in MES buffer (Invitrogen) for 30 minutes at 200 V. Proteins were then transferred to PVDF membrane (Invitrogen) using the iBlot2 machine. Membrane was washed in PBS and then blocked overnight at 4 °C in Intercept Blocking Buffer (LI-COR). Membrane was then incubated overnight at 4 °C with primary antibodies in a 1:1 solution of blocking buffer and PBST. Membrane was then washed three times with PBST, incubated for 1 hour at room temperature with secondary antibodies in 1:1 solution of blocking buffer and PBST and then washed three more times with PBST. It was then imaged with the ChemiDoc MP Imaging System (Bio-Rad). Antibodies used included mouse monoclonal anti-ACSL4 (F-4) (Santa Cruz Biotechnology) 1:500; rabbit monoclonal anti-pan-actin (D18C11) (Cell Signaling Technology) 1:1,000; mouse monoclonal anti-β-actin (8H10D10) (Cell Signaling Technology) 1:1,000; rabbit monoclonal anti-cytochrome C (D18C7) (Cell Signaling Technology) 1:1,000; rabbit monoclonal anti-PDI (C81H6) (Cell Signaling Technology) 1:1,000; IRDye goat anti-mouse 680 1:5,000; and IRDye goat anti-rabbit 800 (LI-COR) 1:5,000.
收获细胞，用 PBS 洗涤，并在含有 cOmplete、Mini 蛋白酶抑制剂片剂 (Sigma-Aldrich) 的 RIPA 缓冲液中冰上孵育 30 分钟。然后将它们在 4°C 下以 14,000 g离心 15 分钟。使用BCA蛋白质测定试剂盒(Pierce)测定上清液的蛋白质浓度，并将浓度标准化并用4×SDS运行缓冲液(Invitrogen)稀释。将 NuPAGE Novex 4–12% Bis-Tris Midi 蛋白凝胶 (Invitrogen) 加载到 MES 缓冲液 (Invitrogen) 中，在 200 V 下运行 30 分钟。然后使用 iBlot2 机器将蛋白质转移到 PVDF 膜 (Invitrogen)。将膜在 PBS 中洗涤，然后在拦截封闭缓冲液 (LI-COR) 中于 4 °C 封闭过夜。然后将膜与一抗在封闭缓冲液和 PBST 的 1:1 溶液中于 4 °C 孵育过夜。然后用PBST洗涤膜3次，在封闭缓冲液和PBST的1:1溶液中与二抗在室温下孵育1小时，然后用PBST再洗涤3次。然后使用 ChemiDoc MP 成像系统 (Bio-Rad) 对其进行成像。使用的抗体包括小鼠单克隆抗ACSL4（F-4）（Santa Cruz Biotechnology）1:500；兔单克隆抗泛肌动蛋白 (D18C11) (Cell Signaling Technology) 1:1,000；小鼠单克隆抗 β-肌动蛋白 (8H10D10) (Cell Signaling Technology) 1:1,000；兔单克隆抗细胞色素 C (D18C7) (Cell Signaling Technology) 1:1,000；兔单克隆抗 PDI (C81H6) (Cell Signaling Technology) 1:1,000； IRDye 山羊抗鼠 680 1:5,000；和 IRDye 山羊抗兔 800 (LI-COR) 1:5,000。

For synthetic methods, see Supplementary Note.
有关合成方法，请参阅补充说明。

Statistical analysis and reproducibility
统计分析和再现性

Statistical analysis was performed in Prism using ANOVA with multiple comparison control or two-sided unpaired t-test with 95% confidence. GraphPad Prism P value style of 0.1234 (NS), 0.0332 (*), 0.0021 (**), 0.0002 (***) and <0.0001 (****) was used. All SRS and confocal fluorescence images are representative. SRS imaging was performed with at least two independent experiments with at least three images. Confocal fluorescence imaging was performed at least twice for each FINO₂ analog with a minimum of three images per experiment. Immunofluorescence imaging was performed twice with a minimum of three images per experiment. C11 BODIPY imaging of ER and PM was repeated three times with at least five images per condition. C11 BODIPY imaging of ER and mitochondria was repeated twice with at least two images per condition.
使用带有多重比较对照的 ANOVA 或 95% 置信度的双边不配对t检验在 Prism 中进行统计分析。使用的 GraphPad Prism P值样式为 0.1234 (NS)、0.0332 (*)、0.0021 (**)、0.0002 (***) 和 <0.0001 (****)。所有 SRS 和共焦荧光图像均具有代表性。 SRS 成像是通过至少两次独立实验和至少三幅图像进行的。每个 FINO ₂类似物至少进行两次共焦荧光成像，每个实验至少三幅图像。免疫荧光成像进行两次，每个实验至少三幅图像。 ER 和 PM 的 C11 BODIPY 成像重复 3 次，每种情况至少 5 张图像。 ER 和线粒体的 C11 BODIPY 成像重复两次，每种情况至少两张图像。

Reporting summary  报告摘要

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
有关研究设计的更多信息，请参阅本文链接的《自然投资组合报告摘要》 。