Triglycerides Promote Lipid Homeostasis during Hypoxic Stress by Balancing Fatty Acid Saturation

Abstract

Lipid droplets, which store triglycerides and cholesterol esters, are a prominent feature of clear cell renal cell carcinoma (ccRCC). Although their presence in ccRCC is critical for sustained tumorigenesis, their contribution to lipid homeostasis and tumor cell viability is incompletely understood. Here we show that disrupting triglyceride synthesis compromises the growth of both ccRCC tumors and ccRCC cells exposed to tumor-like conditions. Functionally, hypoxia leads to increased fatty acid saturation through inhibition of the oxygen-dependent stearoyl-CoA desaturase (SCD) enzyme. Triglycerides counter a toxic buildup of saturated lipids, primarily by releasing the unsaturated fatty acid oleate (the principal product of SCD activity) from lipid droplets into phospholipid pools. Disrupting this process derails lipid homeostasis, causing overproduction of toxic saturated ceramides and acyl-carnitines as well as activation of the NF-κB transcription factor. Our work demonstrates that triglycerides promote homeostasis by "buffering" specific fatty acids.

Keywords: cancer metabolism; clear cell renal cell carcinoma; diglyceride acyltransferase; fatty acid saturation; hypoxia; lipid droplets; lipid homeostasis; lipidomics; stable isotope tracing; triglycerides.

Graphical abstract

Figure 1

DGAT Loss Reduces Tumor Growth and Alters Lipid Composition In Vivo (A) Diagram of fatty acid and lipid synthesis and the influence of O₂ and exogenous lipid. (B) Growth curves for A498 xenograft tumors with induced (doxycycline chow) and un-induced (control chow) DGAT1 and DGAT2 shRNAs (hereafter called DGAT shRNA). (C) Tumor weights after necropsy. (D) Immunohistochemistry for cleaved caspase-3 and Ki67 in xenograft tumors collected on day 5 of treatment, with accompanying quantification. (E) Total TG abundance derived from summing individual TG species abundance after liquid chromatography-mass spectrometry (LC-MS) quantification. (F) TG species binned according to the number of fully saturated FA chains present and the abundance of each category summed and displayed as a ratio of doxycycline-treated versus control groups. All results are means of n = 10 tumors (2 tumors per mouse) per arm; error bars represent ± SD (B, D, and F) or ± SEM (C). Statistical significance by t test or ANOVA, as appropriate; ^∗p < 0.05, ^∗∗p < 0.01, and ^∗∗∗p < 0.001. ACC, acetyl-CoA carboxylase; CE, cholesterol ester; DG, diglyceride; DGAT, diglyceride acyltransferase; FASN, fatty acid synthase; ns, non-significant; PC, phosphatidylcholine; PE, phosphatidylethanolamine; PS, phosphatidylserine; SCD(i), stearoyl-CoA desaturase (inhibitor); SFA, saturated FA; TG, triglyceride; UFA, unsaturated FA. See also Figure S1.

Figure 2

TGs Promote Cell Viability in Low O₂ and Serum by Absorbing FA Saturation (A) Viability of A498 cells expressing inducible shRNA against DGAT1 and DGAT2 mRNAs (DGAT shRNA), assessed after 72 hr under the indicated conditions (hypoxia = 0.5% O₂; serum deprivation = low serum, 0.5% fetal bovine serum [FBS]) by Annexin-propidium iodide (PI) flow cytometry assay. (B) Viability of cells expressing inducible DGAT shRNAs after 72 hr under the indicated conditions (SCDi, 1 μM CAY10566) by Annexin-PI assay using flow cytometry. (C) Volcano plot showing fold change and significance of alterations in the lipidome of A498 cells cultured in low (0.5%) versus high (5%) serum. Lipids with ≥ 1.5 fold change and p ≤ 0.05 are displayed in color to denote lipid class. (D) Changes in FA composition or saturation of TGs, calculated by aggregating TG abundances for species containing 0, 1, or 2+ SFA chains separately. Values are normalized to control conditions (5% serum). (E) Lipid class-specific saturation indices (defined by (palmitate + stearate) / oleate) for A498 cells cultured under hypoxic (0.5% O₂) versus normoxic conditions (both in low serum). (F) As (E) but with pharmacological SCD inhibition (1 μM CAY10566) instead of hypoxia. (G) Effect of serum deprivation and DGAT shRNA on total TG abundances. (H) Changes in FA makeup of TGs following DGAT knockdown; values were calculated by aggregating TG abundances for species containing 0, 1, or 2+ SFA chains separately. Values were normalized to the control condition (vehicle [Veh] treatment). (I) TG saturation indices for the indicated conditions. Values are relative to normoxic untreated cells. (J) As (G) but with pharmacological SCD inhibition (1 μM CAY10566). Values are relative to the untreated vehicle control. Data are means of 3 (A, B, and D–J) or 5 (C) replicate wells and were confirmed in independent experiments; error bars represent SD. Statistical significance by t test or ANOVA, as appropriate. ^∗∗p < 0.05, and ^∗∗∗p < 0.005. See also Figure S2.

Figure 3

¹³C-Oleate Tracing Reveals a Critical Buffering Role for TG-Resident Unsaturated FAs (A) Effect of SCDi on total TG abundances as measured by LC-MS. (B) Effect of oleate pre-loading with or without DGAT shRNA on subsequent A498 cell survival (by Annexin-PI) during serum limitation and SCD inhibition. (C) Schematic of the experimental workflow. DGAT2 knockout cells were serum-starved for 24 hr and then loaded for 24 hr with 10 μM [U¹³C]-oleate (C18:1) ± DGAT1 inhibitor (T863, 2 μM). The medium was then replaced and the tracer removed, and cells were subjected to a 48-hr washout. (D) TG labeling patterns after 24-hr loading with [U¹³C]-oleate with or without DGATi, where numbers of mono-unsaturated FA (MUFA) and FA carbons are indicated. 1×, 2×, and 3× indicate whether TGs have one, two, or three oleates (includes [¹³C₁₈]-20:1) conjugated to their glycerol backbones. (E) BODIPY and DAPI staining directly after [U¹³C]-oleate loading with or without DGATi. (F) Labeling patterns as assessed by incorporation of the ¹³C label in 18:1 and 20:1 FAs in TG, DG, PC, and PE species. (G) Model of the metabolic mechanism by which TGs alleviate the saturation of certain lipid classes (e.g., PCs) under conditions of unsaturated lipid deprivation by releasing stored oleate. Data are means of triplicate wells confirmed in independent experiments (A, B, and D) or means of three independent experiments each conducted in triplicate (F); error bars represent SD. Statistical significance by t test or ANOVA, as appropriate. ^∗p < 0.05, ^∗∗p < 0.05, and ^∗∗∗p < 0.005. See also Figure S3.

Figure 4

DGAT Loss Modifies Lipid Homeostasis, Elevates Ceramide, Acyl-ceramide, and Acyl-carnitine Levels, and Activates NF-kB Target Gene Expression (A) Effect of SCD and DGAT inhibition on ceramide levels in serum-deprived A498 cells in vitro. (B) Effect of DGAT loss on ceramides in vivo (i.e., A498 xenografts). (C) Effect of DGAT loss on acyl-ceramides in vivo (i.e., A498 xenografts). (D) Effect of hypoxia on the FA composition of acyl-carnitines (CARs) on serum-deprived A498 cells in vitro. (E) Effect of DGAT loss on the FA composition of acyl-carnitines (CARs) on serum-deprived A498 cells in vitro. (F) Effect of DGAT loss on the FA composition of acyl-CARs in A498 xenograft tumors. (G) Gene set enrichment analysis (GSEA) on RNA as assessed by microarray comparisons performed on A498 DGAT shRNA xenograft tumors after 5 days of control or doxycycline chow. Normalized enrichment score (NES) allow comparison of enrichment between different gene sets. (H) Effect of DGAT shRNA, NF-κB inhibition, and proteasome inhibition on NF-κB luciferase reporter activity. (I) Effect of serum deprivation on NF-κB luciferase reporter activity. (J) Schematic of the consequences of DGAT inhibition preceding a period of unsaturated lipid deprivation. These conditions lead to increased acyl-CARs and ceramides as well as increased incorporation of saturated FAs into the PL pool. (K) Published data comparing the TG composition of ccRCC and normal tissue (Saito et al., 2016) were reanalyzed to investigate shifts in TG saturation. TG species were binned according to the number of fully saturated FA chains present, and the abundance of each category was aggregated and is displayed as a ratio of the abundance in normal tissue. For (A), (D), (E), data are means of 5 and for (H) and (I) of 3 replicate wells, and results were confirmed in independent experiments. For (B), (C), (F), and (G), data are means of tumors from 4 tumors. Error bars represent SD. Statistical significance by t test or ANOVA, as appropriate. ^∗p < 0.05, ^∗∗p < 0.05, and ^∗∗∗p < 0.005. See also Figure S4.