Stearate‐derived very long‐chain fatty acids are indispensable to tumor growth

Abstract

Reprogramming of lipid metabolism is emerging as a hallmark of cancer, yet involvement of specific fatty acids (FA) species and related enzymes in tumorigenesis remains unclear. While previous studies have focused on involvement of long‐chain fatty acids (LCFAs) including palmitate in cancer, little attention has been paid to the role of very long‐chain fatty acids (VLCFAs). Here, we show that depletion of acetyl‐CoA carboxylase (ACC1), a critical enzyme involved in the biosynthesis of fatty acids, inhibits both de novo synthesis and elongation of VLCFAs in human cancer cells. ACC1 depletion markedly reduces cellular VLCFA but only marginally influences LCFA levels, including palmitate that can be nutritionally available. Therefore, tumor growth is specifically susceptible to regulation of VLCFAs. We further demonstrate that VLCFA deficiency results in a significant decrease in ceramides as well as downstream glucosylceramides and sphingomyelins, which impairs mitochondrial morphology and renders cancer cells sensitive to oxidative stress and cell death. Taken together, our study highlights that VLCFAs are selectively required for cancer cell survival and reveals a potential strategy to suppress tumor growth.

ACC1‐dependent de novo synthesis of very long chain fatty acids promotes tumor growth by stabilizing mitochondrial morphology and reducing cellular stress.

Introduction

Fatty acids are a diverse group of aliphatic hydrocarbons with a polar carboxylic head group. These molecules vary in length, number of carbon atoms, and saturation, including the number and position of double bonds. Based on the saturation, fatty acids can contain no double bonds, one double bond, or more than one double bond, termed saturated (SFAs), monounsaturated (MUFAs), and polyunsaturated fatty acids (PUFAs), respectively. Mammalian cells can synthesize fatty acids de novo from acetyl‐CoA that is mainly derived from glucose, glutamine, and acetate (Pietrocola et al, 2015). Acetyl‐CoA is then converted to malonyl‐CoA by acetyl‐CoA carboxylases. There are two isozymes, acetyl‐CoA carboxylase 1 (ACC1) and acetyl‐CoA carboxylase 2 (ACC2), that mediate distinctive physiological functions within the cell (Wang et al, 2022). ACC1 is localized primarily to the cytosol, while ACC2 is associated with the mitochondrial outer membrane. Therefore, ACC1 produces malonyl‐CoA in the cytosol, which is the major carbon donor for fatty acid synthesis. Seven malonyl‐CoA molecules are used as the elongation unit, and together with acetyl‐CoA as the initiator, they are catalyzed by fatty acid synthase (FASN) to form one 16‐carbon saturated fatty acid, palmitate (Jones & Infante, 2015; Menendez & Lupu, 2022). By contrast, malonyl‐CoA synthesized by ACC2 is located near the mitochondrial surface and is thought to work as an inhibitor of carnitine palmitoyl transferase 1 (CPT1), thus regulating transport of long‐chain fatty acids into mitochondria for subsequent β‐oxidation (Wang et al, 2022).

Mammalian cells can produce double bonds at the Δ9 position of the hydrocarbon chain with stearoyl‐CoA desaturase (SCD) to form n‐9 MUFAs, such as palmitoleate and oleate. However, these cells lose the ability to make unsaturated fatty acids with double bonds at the Δ3 or Δ6 positions. Therefore, mammalian cells must take up n‐3 and n‐6 unsaturated fatty acids, mainly linoleate and α‐linolenate, as essential nutrients (Nakamura & Nara, 2004). These fatty acids can be further elongated to very long‐chain fatty acids (VLCFAs, with C ≥ 22). Elongation of VLCFAs predominantly occurs in the endoplasmic reticulum (ER) through repetition of the elongation cycle; this consists of four sequential transformations catalyzed by elongases (ELOngation of Very Long fatty acids, ELOVL1‐7), 3‐ketoacyl‐CoA reductase (KAR), 3‐hydroxyacyl‐CoA dehydratases (HACD1‐4), and trans‐2‐enoyl‐CoA reductase (TECR), respectively (Jakobsson et al, 2006; Kihara, 2016). Although PUFAs can be generated by fatty acid desaturases (FADS1‐3) and are elongated primarily from the dietary essential linoleate and α‐linolenate, many PUFAs are essential and must instead be taken up (Nakamura & Nara, 2004).

Reprogramming of lipid metabolism is now emerging as one of the hallmarks of cancer (Boroughs & DeBerardinis, 2015; Grunt, 2018; Corn et al, 2020; Snaebjornsson et al, 2020). In the past decades, the roles of ectopic de novo fatty acid synthesis were extensively studied in cancers (Currie et al, 2013; Rohrig & Schulze, 2016), and the involved enzymes, ACC1 and FASN, were regarded as potential anti‐tumor targets (Rohrig & Schulze, 2016; Snaebjornsson et al, 2020). However, it remains to be determined how these enzymes that are involved in fatty acid synthesis affect cellular lipid metabolism. In the current study, we carry out lipid metabolomics analyses on FASN‐knockout and ACC1/2‐knockout cells and systematically compare these data. We ultimately reveal that tumor growth is susceptible to regulation of VLCFAs but not LCFAs, including palmitate, and further dissect the underlying mechanism.

Results

Tumor growth is more vulnerable to ACC1‐KO, rather than FASN‐KO

Cancer cells actively synthesize fatty acids, primarily palmitate, de novo through ACC1 and FASN (Fig 1A), and this is thought to play an essential role in cell growth and tumor progress (Menendez & Lupu, 2007). However, whether tumor growth can be effectively repressed by blocking ACC1 or FASN remains controversial (Jeon et al, 2012). Here, we used a CRISPR/Cas9 system to knock out FASN, ACC1, and/or ACC2 in MDA‐MB‐231 cells (Fig 1B). We found that knockout of ACC1 significantly suppressed cell proliferation, colony formation, invasion, and xenograft tumor growth in nude mice, which was exacerbated by the additional deletion of ACC2 (ACC1/2‐double knockout, DKO; Fig 1C–F). In contrast, knockout of FASN or ACC2 had no effect on these characteristics (Fig 1C–F). Furthermore, we observed similar effects of knocking out these genes in HeLa and MCF‐7 cells. ACC1‐KO or ACC1/2‐DKO significantly reduced cell proliferation, invasion, colony formation ability, and tumor growth, all of which were less extensive or not observed with FASN‐KO (Figs 1C right and F right, and Fig EV1A–F). These results suggest that tumor growth is more vulnerable to ACC1‐KO than to FASN‐KO.

Figure 1. Tumor growth is more vulnerable to ACC1‐KO, rather than FASN‐KO.

1. A scheme to show de novo synthesis of palmitate and fatty acid elongation achieved by enzymes, ELOLVs, KAR, HACDs, and TECR.
2. Western blot verification of knockout of FASN, ACC1, and/or ACC2 in MDA‐MB‐231 cells.
3. Cell proliferation rate of MDA‐MB‐231 and HeLa cells with knockout of FASN, ACC1, and/or ACC2.
4. Colony formation ability of MDA‐MB‐231 cells with knockout of FASN, ACC1, and/or ACC2.
5. Invasion ability of MDA‐MB‐231 cells with knockout of FASN, ACC1, and/or ACC2. The right panel shows the pictures of invasive cells, and the cell numbers are quantified in the left panel. Scale bar: 20 μm.
6. Tumor formation ability in nude mice of MDA‐MB‐231 and HeLa cells with gene knockout as indicated.

Data information: In (C–E), data are from triplicate experiments, and all experimental data are verified in at least two independent experiments. In (F), n = 4 for MDA‐MB‐231‐derived tumors, n = 6 for HeLa‐derived tumors. Error bars represent mean ± SD. *P < 0.01, **P < 0.01 (t‐test); black asterisk shows comparison with wild‐type group.

Source data are available online for this figure.

Figure EV1. Effects of knockouts of fatty acid synthesis enzymes on HeLa and MCF7 cells.

1. Western blot verification of knockout of FASN, ACC1, and/or ACC2 in HeLa cells.
2. Colony formation ability of HeLa cells with knockout of FASN, ACC1 and/or ACC2.
3. Western blot verification of knockout of FASN, ACC1, and/or ACC2 in MCF7 cells.
4. Cell proliferation rate of MCF7 cells with knockout of FASN, ACC1, and/or ACC2.
5. Colony formation ability of MCF7 cells with knockout of FASN, ACC1, and/or ACC2. The right panel shows the pictures of colonies, and the colony numbers are quantified in the left panel.
6. Invasion ability of MCF7 cells with knockout of FASN, ACC1, and/or ACC2. The right panel shows the pictures of invasive cells, and the cell numbers are quantified in the left panel. Scale bar: 50 μm.

Data information: In (B, D, E and F), data are from triplicate experiments, and all experimental data were verified in at least two independent experiments. Error bars represent mean ± SD. **P < 0.01 (t‐test); asterisk shows comparison with wild‐type group.

Source data are available online for this figure.

The elongation of VLCFAs, but not de novo synthesis of palmitate, is associated with tumor progression

As expected, FASN‐KO led to the accumulation of its substrate malonyl‐CoA, while ACC1‐KO dramatically reduced malonyl‐CoA (Fig 2A). In particular, ACC1/2‐DKO almost completely depleted cellular malonyl‐CoA (Fig 2A), indicating that the two isoforms predominantly account for its production. By contrast, none of the knockouts affected the cellular level of acetyl‐CoA (Fig 2A). Malonyl‐CoA is used as the two‐carbon unit for de novo biosynthesis of palmitate and also for the elongation of VLCFAs synthesized by ER membrane‐embedded enzymes (Jakobsson et al, 2006). In addition, acetyl‐CoA can also be utilized as the elongation unit for VLCFAs in the mitochondrial pathway (Jump, 2009). Therefore, to further determine how these knockouts affect the biosynthesis of fatty acids, we used ¹³C₆‐glucose to trace fatty acid biosynthesis in MDA‐MB‐231 and HeLa cells. Using this approach, acetyl‐CoA or derived malonyl‐CoA would contain ¹³C from ¹³C₆‐glucose or no labeled carbon, thus we expected the labeled fractions to fit a binomial distribution model (Fig 2B). VLCFAs can be produced from de novo synthesized palmitate or elongated on the already existing palmitate or other precursors; these two processes can be distinguished based on the labeling model of fatty acids (Fig 2B). Our tracing results showed that de novo synthesis mainly occurred in saturated fatty acids and to some degree in n‐9 unsaturated fatty acids, but not in n‐6 or n‐3 polyunsaturated fatty acids (Figs 2C and EV2A). This finding is consistent with the typical notion that human cells are unable to de novo synthesize n‐3 and n‐6 fatty acids (Nakamura & Nara, 2004). However, active elongation was observed in most of the fatty acids with > 16 carbons including all VLCFAs (Figs 2C and EV2A). ACC1‐KO suppressed both de novo synthesis and elongation of fatty acids, whereas ACC2‐KO did not affect either process. Notably, ACC1/2‐DKO almost totally blocked biosynthesis including elongation of all fatty acids tested in both cell lines. This suggests that cancer cells dominantly use malonyl‐CoA, not acetyl‐CoA, as the two‐carbon units for fatty acid elongation. In contrast, FASN‐KO completely inhibited de novo synthesis of fatty acids, but the elongation of fatty acids was unaffected (Figs 2C and EV2A). Therefore, de novo synthesis of palmitate is less important for tumor growth in comparison to elongation of VLCFAs.

Figure 2. ACC1‐KO inhibits both de novo synthesis and elongation of fatty acids, while FASN‐KO only inhibits de novo synthesis.

1. Relative cellular levels of acetyl‐CoA and malonyl‐CoA in MDA‐MB‐231 cells with knockout of FASN, ACC1, and/or ACC2.
2. Binomial distribution model of ¹³C₆‐glucose‐labeled fatty acids indicates de novo synthesis of fatty acids and elongation of existing precursors.
3. Mass isotopomer analysis of fatty acids in MDA‐MB‐231 cells with knockout of FASN, ACC1, and/or ACC2 cultured with 10 mM of ¹³C₆‐glucose for 48 h. Solid arrows show the reported pathways; dotted arrows indicate the possible pathways based on the labeling results.
4. Relative cellular levels of fatty acids in MDA‐MB‐231 cells with knockout of FASN, ACC1 and/or ACC2. Each colored datum shows the mean from three independent cultures.

Data information: In (A and C), data are from three independent cultures. Error bars represent mean ± SD. *P < 0.01, **P < 0.01 (t‐test).

Figure EV2. Fatty acids synthesis in HeLa knockout cells.

1. Mass isotopomer analysis of fatty acids in HeLa cells with knockout of FASN, ACC1, and/or ACC2 cultured with 10 mM of ¹³C₆‐glucose for 48 h. Results are shown as mean ± SD from three independent cultures.
2. Relative cellular levels of fatty acids in HeLa cells with knockout of FASN, ACC1, and/or ACC2. Each colored datum shows the mean from three independent cultures.

Decreased cellular VLCFAs are associated with slow growth of tumors

Next, we saponified and measured the total cellular fatty acids in MDA‐MB‐231 and HeLa cells. Surprisingly, our results showed that FASN‐KO, as well as ACC2‐KO, enhanced almost all detected fatty acids (Figs 2D and EV2B). Even in the cells with ACC1‐KO or ACC1/2‐DKO, only the cellular contents of some VLCFAs were dramatically decreased (Figs 2D and EV2B). Interestingly, among the detected PUFAs, several [FA(16:2), FA(18:2), FA(22:1), FA(16:3), FA(18:3), FA(18:4), FA(20:4), FA(20:5), FA(22:5), and FA(22:6)] were not labeled or were ineffectively labeled (Figs 2C and EV2A). This suggests that their cellular levels were not reliant on biosynthesis and were most likely supplied through nutritional uptake. Indeed, the cellular levels of these fatty acids were not impaired by knockout of the genes involved in the fatty acid synthesis (Figs 2D and EV2B). Therefore, it appears that the impaired elongation of VLCFAs is highly associated with inhibition of tumor growth.

We then carried out lipid metabolomics on MDA‐MB‐231 cells with different deleted genes. ACC1/2‐DKO, ACC1‐KO, FASN‐KO, and ACC2‐KO primarily showed differential levels of cellular triglycerides, ceramides, and glucosylceramides, including Hex1Cer, Hex2Cer, Hex3Cer, GM3, and CerG3GNAC1 (Fig 3A). Moreover, most VLCFAs in these lipids were reduced to a greater extent by ACC1/2‐DKO or ACC1‐KO, compared to FASN‐KO (Fig 3B–D). Among the measured lipid fatty acids, ACC1/2‐DKO produced the largest changes and showed high correlation with ACC1‐KO, low correlation with FASN‐KO, and no correlation with ACC2‐KO (Fig 3E). Moreover, the degree of reduction in lipid fatty acids from ACC1/2‐DKO cells was positively correlated with the number of carbons (Fig 3F). Taken together, these results suggest that tumor growth is susceptible to decreases in cellular VLCFAs.

Figure 3. ACC1‐KO reduces the cellular levels of VLCFAs.

• A
  Relative cellular levels of lipids in MDA‐MB‐231 cells with knockout of FASN, ACC1, and/or ACC2. Each colored datum shows the mean from three independent cultures. TG, triglyceride; DHCeramide, dihydroceramide; Hex1Cer, hexosylceramide; Hex2Cer, dihexosylceramide; Hex3Cer, trihexosylceramide; GM3, monosialodihexosylganglioside; SM, sphingomyelin; PC, phosphatidylcholine; PE, phosphatidylethanolamine; PG, phosphatidylglycerol; PI, phosphatidylinositol; PS, phosphatidylserine; PC(o) and PE(o), ether lipids.
• B–D
  Relative cellular levels of lipid‐containing fatty acids in MDA‐MB‐231 cells with knockout of FASN, ACC1, and/or ACC2. The top curve plot indicates the fraction of the corresponding fatty acid. Each colored datum shows the mean from three independent cultures.
• E
  Correlation of lipid‐containing fatty acids in MDA‐MB‐231 cells with knockout of FASN, ACC1 or ACC2 with those of ACC1/2‐DKO cells. Each datum is mean from three independent cultures.
• F
  Relationship between cellular level and carbon number of lipid‐contained fatty acids in MDA‐MB‐231 cells with ACC1/2‐DKO.

Elongation of VLCFAs is required for tumor growth

The above results suggest that the elongation of VLCFAs may be necessary for cancer cell proliferation. The elongation of VLCFAs consists of four steps that are catalyzed by the corresponding enzymes ELOVL, KAR, HACD, and TECR. ELOVL and HACD have seven and four isoforms, respectively, whereas no additional isoforms have been reported for KAR or TECR. Therefore, we aimed to block elongation by knocking out KAR and TECR in MDA‐MB‐231 cells (Fig 4A). Their knockouts did not affect the levels of acetyl‐CoA or malonyl‐CoA (Appendix Fig S1A), but did significantly suppress cell proliferation, colony formation, and in vivo tumor growth (Fig 4B–D). In particular, TECR‐KO completely blocked tumor formation, which was restored by the re‐expression of TECR (Fig 4E). Next, we also saponified and measured the total fatty acids in MDA‐MB‐231 cells and found that KAR‐ or TECR‐KO significantly decreased the cellular levels of VLCFAs (Fig 4F). It was further confirmed by lipid metabolomic analysis that VLCFAs involved in lipid components were widely decreased in KAR‐ or TECR‐KO cells (Fig 4G). Furthermore, the ¹³C₆‐glucose tracing results showed that knockout of the two genes did not affect de novo synthesis of palmitate (Fig 4H). Interestingly, knockout of KAR or TECR only partially repressed each cycle of elongation of fatty acids, such that the knockout preferentially blocked the synthesis of VLCFAs with longer chains (Fig 4H and Appendix Fig S1B). In addition, these results suggest that cells may express alternative enzymes that compensate for the loss of KAR or TECR and act to elongate fatty acids.

Figure 4. Knockout of elongation enzymes suppreses tumor growth.

• A
  Western blot verification of knockout of KAR or TECR in MDA‐MB‐231 cells.
• B, C
  Cell proliferation rate and colony formation ability of MDA‐MB‐231 cells with knockout of KAR or TECR. Error bars represent mean ± SD. Data are from triplicate experiments, and all experimental data were verified in at least two independent experiments. **P < 0.01 (t‐test); asterisk shows comparison with wild‐type group.
• D, E
  Tumor formation ability in nude mice of MDA‐MB‐231 cells with gene knockout, as indicated. Error bars represent mean ± SD (n = 6 for D, n = 8 for E).
• F
  Relative cellular levels of fatty acids in MDA‐MB‐231 cells with knockout of KAR or TECR. Each colored datum shows the mean from three independent cultures.
• G
  Relative cellular levels of lipid‐containing fatty acids in MDA‐MB‐231 cells with knockout of KAR or TECR. The top curve plot indicates the fraction of the corresponding fatty acid. Each colored datum shows the mean from three independent cultures.
• H
  Mass isotopomer analysis of fatty acids in MDA‐MB‐231 cells with knockout of KAR or TECR cultured with 10 mM of ¹³C₆‐glucose for 48 h. Error bars represent mean ± SD. Data are from three independent cultures.

Source data are available online for this figure.

We also deleted KAR and TECR in HeLa cells (Fig EV3A) and found that TECR‐KO largely suppressed colony formation and tumor growth in a xenograft model of nude mice; KAR‐KO only slightly suppressed colony formation and tumor growth (Fig EV3A and B). Our results further demonstrated that TECR‐KO decreased the levels of VLCFAs more significantly, compared with KAR‐KO (Fig EV3C). This most likely resulted from more effective compensation pathways for KAR in HeLa cells, as demonstrated by our ¹³C₆‐glucose tracing analysis where KAR‐KO indeed blocked biosynthesis of saturated and monounsaturated VLCFAs much less effectively than TECR‐KO (Fig EV3D). Similarly, we also observed that TECR‐KO exerted more potently suppressive effects on proliferation and colony formation of MCF‐7 cells than KAR‐KO (Fig EV3E–G). Furthermore, we analyzed the expression of elongation enzymes in 15 tumor and para‐carcinoma tissues based on the TCGA database. We found that TECR was highly expressed in most (10 of 15) tumor tissues compared to the corresponding para‐carcinoma tissues, similar to ACC1 (Appendix Fig S2A). By contrast, KAR was only highly expressed in colorectal and stomach cancers (Appendix Fig S2A), also suggesting the presence of other important KAR isoform(s) for VLCFA elongation in cancers. In addition, the most abundant isoforms, ELOVL1 and HACD3, were also expressed to a greater degree in most tumors compared to para‐carcinoma tissues (Appendix Fig S2A). We also observed low levels of TECR in epithelial MCF‐10A cells, compared with cancer cells (Appendix Fig S2B). Thus, we aimed to knock out TECR in MCF‐10A cells, and our results showed that TECR‐knockout also suppressed cell proliferation of MCF10A cells, although this effect was much weaker than that observed in MDA‐MB‐231 cells (Appendix Fig S2C). Taken together, these data demonstrate that elongation of VLCFAs is critical to tumor growth.

Figure EV3. Effects of KAR‐ or TECR‐KO on fatty acid synthesis in HeLa and MCF7 cells.

1. Western blot verification of knockout of KAR or TECR in HeLa cells and colony formation ability of HeLa cells with knockout of KAR or TECR.
2. Tumor formation ability in nude mice of HeLa cells with KAR or TECR knockout as indicated. The tumor picture was shown in Fig 1F.
3. Relative cellular levels of fatty acids in HeLa cells with knockout of KAR or TECR. Each colored datum shows the mean from three independent cultures.
4. Mass isotopomer analysis of fatty acids in HeLa cells with knockout of KAR or TECR cultured with 10 mM of ¹³C₆‐glucose for 48 h. Results are shown as mean ± SD from three independent cultures.
5. Western blot verification of knockout of KAR or TECR in MCF7 cells.
6. Cell proliferation rate of MCF7 cells with knockout of KAR or TECR.
7. Colony formation of MCF7 cells with knockout of KAR or TECR. The right panel shows the pictures of colonies, and the colony numbers are quantified in the left panel.

Data information: In (A, B, F and G), data are from triplicate experiments, and all experimental data were verified in at least two independent experiments. Error bars represent mean ± SD. **P < 0.01 (t‐test); asterisk shows comparison with wild‐type group.

Source data are available online for this figure.

However, we analyzed the correlation between survival time of patients of different cancers and protein levels of ACC1, FASN, ELOVL1‐7, KAR, HACD1‐4, or TECR (Appendix Table S1) and found that these correlations were very variant between cancers. As for breast cancers, the correlations of KAR or TECR with survival time were not significant. This was most likely attributed to the differential regulation of metabolic enzymatic activities that were much more important than their protein levels. Alternatively, they probably were not the rate‐limiting enzymes, thus their protein levels could not determine the rate of VLCFA biosynthesis.

Stearate‐derived VLCFAs support cancer cell growth

To further verify the requirement of VLCFA elongation for cancer cell growth, we cultured MDA‐MB‐231 and HeLa cells in medium with delipidized serum (DFBS; Appendix Fig S3A). Cell lines with deleted FASN, ACC1, or ACC1/2 did not survive, whereas the ACC2‐KO cells did survive (Fig 5A and Appendix Fig S3B). Palmitate supplementation robustly rescued FASN‐KO cells and slightly protected ACC1‐KO cells from cell death induced by DFBS, but did not rescue the ACC1/2‐DKO or TECR‐KO cells that lacked elongation ability (Fig 5A and Appendix Fig S3B). Supplementation of LCFAs (16:0–20:0) also did not protect ACC1/2‐DKO and TECR‐KO cells from death induced by lipid deprivation (Fig 5B, and Appendix Fig S3C and D). However, VLCFAs (22:0–26:0) significantly rescued ACC1/2‐DKO and TECR‐KO cells cultured in DFBS medium (Fig 5B, and Appendix Fig S3C and D). These results confirm that VLCFA elongation plays a critical role in cell survival. Considering the metabolic plasticity and complexity of fatty acid biosynthesis in cells (Fig 5C), we also tested the rescuing effects of palmitate (16:0), stearate (18:0), oleate (18:1), and arachidate (20:0), along with the essential precursors of n‐6 and n‐3 polyunsaturated fatty acids, linoleate (18:2) and α‐linolenate (18:3), on MDA‐MB‐231/FASN‐KO and HeLa/FASN‐KO cells cultured in DFBS medium. Oleate (18:1) slightly supported the growth of FASN‐KO cells in DFBS medium, while linoleate (18:2) and/or α‐linolenate (18:3) did not exert any preventive effects (Fig 5D and Appendix Fig S3E). By contrast, stearate (18:0) restored cell proliferation to a better extent compared with palmitate (16:0) and arachidate (20:0; Fig 5D). As the proximal downstream substrates of stearate (18:0; Fig 5C), arachidate (20:0), and oleate (18:1) in combination rescued cell viability to a similar extent as stearate (Fig 5D).

Figure 5. Stearate‐derived VLCFAs support cancer cell growth.

1. Proliferation rate of MDA‐MB‐231 cells with gene knockout, as indicated, in FBS medium or delipidized FBS (DFBS) medium supplemented with 0, 10, or 20 μM of palmitate.
2. Proliferation rate of MDA‐MB‐231/ACC‐DKO in FBS medium or DFBS medium supplemented with or without exogenous LCFAs (2 μM FA 16:0, 2 μM 18:0, 2 μM FA 20:0) or VLCFAs (2 μM FA 22:0, 2 μM FA 24:0, 2 μM FA 26:0).
3. A scheme to show the elongation of fatty acids from precursors.
4. Proliferation rate of MDA‐MB‐231/FASN‐KO cells in FBS medium or DFBS medium supplemented with or without 10 μM of fatty acids as indicated.
5. Mass isotopomer analysis of fatty acids in MDA‐MB‐231/FASN‐KO cells in FBS medium or DFBS medium supplemented with or without 20 μM of stearate. Cells were cultured with 25 mM of ¹³C₆‐glucose for 36 h.
6. Relative cellular levels of fatty acids in MDA‐MB‐231/FASN‐KO cells in FBS medium or DFBS medium supplemented with or without 20 μM of stearate.

Data information: In (A, B and D), data are from triplicate experiments, and all experimental data are verified in at least two independent experiments. In (E and F), data are from three independent cultures. Error bars represent mean ± SD. *P < 0.05, **P < 0.01 (t‐test); black asterisk shows comparison with wild‐type group, and gray asterisk shows comparison with DFBS group.

Since FASN‐KO in MDA‐MB‐231 cells only blocked de novo fatty acid synthesis while keeping fatty acid elongation ability mostly intact, we next traced the ¹³C₆‐glucose‐labeled fatty acids in FASN‐KO cells cultured in FBS or DFBS medium. Our results showed that the labeled fractions of stearate‐derived saturated fatty acids, FA(20:0) to FA(28:0) detected here, were obviously declined in cells cultured with DFBS medium (Fig 5E and Appendix Fig S4). These results indicate that the synthesis rate of stearate‐derived saturated fatty acids may depend on cellular concentrations of the precursors. This speculation was confirmed by our finding that stearate supplementation largely restored the ¹³C₆‐glucose‐derived fractions of saturated VLCFAs (Fig 5E) and increased the cellular levels of stearate and the downstream saturated VLCFAs (Fig 5F). In addition, we also observed that the labeled fractions of FA(20:1), FA(24:1), FA(26:1), FA(20:2), FA(24:2), and FA(26:2) were slightly dependent upon stearate supplementation in the FASN‐KO cells with DFBS medium (Appendix Fig S4). Therefore, these data suggest that stearate‐derived VLCFAs play a significant role in supporting cancer cell proliferation.

Fatty acid deficiency sensitizes cells to oxidative stress

To further determine the mechanism underlying cell death induced by fatty acid deficiency, we cultured ACC‐DKO cells in DFBS medium in the presence or absence of z‐VAD‐FMK (an apoptosis inhibitor; Fearnhead et al, 1995), necrostatin‐1 (a necroptosis inhibitor; Wang et al, 2007), disulfiram (a pyroptosis inhibitor; Hu et al, 2020), or ferrostatin‐1 and deferoxamine mesylate (DFOM; two ferroptosis inhibitors; Bogacz & Krauth‐Siegel, 2018). Our results showed that z‐VAD‐FMK and ferrostatin‐1 significantly enhanced survival of HeLa/DKO and MDA‐MB‐231/DKO cells in DFBS medium (Figs 6A and EV4A). However, another ferroptosis inhibitor, DFOM, did not rescue viability of DKO cells (Figs 6A and EV4A). Ferrostatin‐1 scavenges lipid reactive oxygen species (ROS), while DFOM is an iron chelator. This means that ROS may be the true cause of fatty acid deficiency–induced cell death, while “ferroptosis” may simply be a side effect associated with ROS. Next, we treated cells with two other antioxidants, trolox and N‐acetylcysteine (NAC), and found that both obviously promoted survival of HeLa/DKO and MDA‐MB‐231/DKO cells in DFBS medium (Figs 6A and EV4A). Interestingly, ferrostatin‐1, trolox, and NAC completely restored proliferation of HeLa/Wt and MDA‐MB‐231/Wt cells in DFBS medium (Figs 6A and EV4A). These data suggest that ROS are involved in growth suppression and cell death induced by lipid‐deprivation. Therefore, we measured the levels of cellular ROS in wild‐type and DKO cells and found that in complete medium, the basic levels of ROS in wild‐type and DKO cells were similar (Figs 6B and EV4B). However, lipid deprivation induced much higher levels of ROS in HeLa/DKO and MDA‐MB‐231/DKO cells than in wild‐type cells, and this was significantly reduced by supplementation with exogenous VLCFAs (FA(22:0), FA(24:0), and FA(26:0); Figs 6B and EV4B). This finding suggests that DKO cells lose the ability to maintain redox homeostasis under stress, at least partially resulting from VLCFA deficiency. This speculation was further confirmed by the results that MDA‐MB‐231/DKO and HeLa/DKO cells, as well as MDA‐MB‐231/TECR‐KO cells, were much more sensitive to hydrogen peroxide (H₂O₂) and superoxide induced by phenazine methosulfate (PMS; Chan & Weiss, 1987), in comparison to wild‐type cells (Figs 6C, and EV4C and D).

Figure 6. ACC1/2‐DKO cells sensitize to oxidative stress and mitochondria‐dependent apoptosis.

1. MDA‐MB‐231 Wt cells or ACC‐DKO cells were pretreated for 24 h and then cultured in FBS medium or DFBS medium with the indicated treatment for 5 days. Treatments were as follows: z‐VAD‐FMK (z‐VAD, 10 μM), necrostain‐1 (Nec‐1, 2 μM), Disulfiram (0.4 μM), ferrostain‐1 (Ferr‐1, 10 μM), Trolox (100 μM), N‐Acetyl‐L‐cysteine (NAC, 3 mM), and deferoxamine mesylate (DFOM, 10 μM). Data are presented relative to the values obtained for the control cells cultured in FBS medium. Dotted line shows the survival level of cells in DFBS medium without treatment (control).
2. Relative cellular levels of ROS in MDA‐MB‐231/ACC‐DKO and Wt cells cultured in FBS medium or DFBS medium for 5 days with or without exogenous VLCFA supplement.
3. The inhibitory effects of H₂O₂ and PMS on MDA‐MB‐231 ACC‐DKO or Wt cells cultured in FBS medium for 24 h with indicated concentrations.
4. The protein levels of PARP1 and caspase‐3 in MDA‐MB‐231/ACC‐DKO and Wt cultured in FBS or DFBS medium.
5. Annexin‐V & Dead cell (7‐AAD) flow cytometry analysis of MDA‐MB‐231/ACC‐DKO and Wt cultured in DFBS medium. The right panel shows the quantification of annexin V‐positive cells in the left panels.
6. The relative cell survival of MDA‐MB‐231/ACC‐DKO or Wt with control vector or Bcl‐2 or Bcl‐xL overexpression cultured in FBS medium or DFBS medium for 5 days. The left panels show the cell survival, and the right panels verify the overexpression of Bcl‐2 and Bcl‐xL.
7. The transmembrane potential in mitochondria of MDA‐MB‐231/ACC‐DKO and Wt cells cultured in FBS medium or DFBS medium was detected by FACS analysis. The bar graphs show the percentage of JC‐1 monomer.
8. MDA‐MB‐231/ACC‐DKO and Wt cells were stained for Mito Tracker to visualize mitochondria. Scale bar, 10 μm. Mitochondrial morphology was scored in the cell types indicated. Data represent three independent experiments, and at least 100 cells were counted per cell type.

Data information: In (A, B, C, E, F and G), data are from triplicate experiments, and all experimental data were verified in at least two independent experiments. Error bars represent mean ± SD. *P < 0.05, **P < 0.01 (t‐test). In (H), asterisk shows comparison with wild‐type group.

Source data are available online for this figure.

Figure EV4. VLFAs are necessary to maintain antioxidant ability.

• A
  HeLa/ACC‐DKO cells or Wt cells were pretreated for 24 h and then cultured in FBS medium or DFBS medium with indicated treatment for 5 days. Treatments were as follows: z‐VAD (10 μM), Ferr‐1 (10 μM), Trolox (100 μM), and NAC (3 mM). Data are presented relative to the values obtained for the control cells cultured in FBS medium. Dotted line shows the survival level of cells in DFBS medium without treatment (control).
• B
  Relative cellular levels of ROS in HeLa/ACC‐DKO and Wt cells cultured in FBS medium or DFBS medium for 6 days.
• C
  The inhibitory effects of H₂O₂ and PMS on HeLa/Wt and ACC‐DKO cells cultured in FBS medium for 48 h with indicated concentrations.
• D
  The inhibitory effects of H₂O₂ and PMS on MDA‐MB‐231/Wt and TECR‐KO cells cultured in FBS medium for 24 h with indicated concentrations.
• E
  The protein levels of PARP1 and caspase‐3 in HeLa/ACC‐DKO and Wt cultured in FBS medium or DFBS medium. The left panels show the cell survival, and the right panels verify the overexpression of Bcl‐2 and Bcl‐xL.
• F
  The relative survival of HeLa/Wt and ACC‐DKO cells with control vector or Bcl‐2 or Bcl‐xL overexpression cultured in FBS medium or DFBS medium for 5 days.
• G, H
  The transmembrane potential in mitochondria of HeLa /ACC‐DKO (G) and MDA‐MB‐231/TECR‐KO (H) cells cultured in FBS medium or DFBS medium was detected by FACS analysis.

Data information: In (A–D, F–H), data are from triplicate experiments, and all experimental data were verified in at least two independent experiments. Error bars represent mean ± SD. *P < 0.05, **P < 0.01 (t‐test). In (A), asterisk shows comparison with wild‐type group.

Source data are available online for this figure.

VLCFAs are necessary to maintain normal mitochondrial morphology

The pan‐caspase inhibitor, z‐VAD‐FMK, rescued ACC‐DKO cells from lipid deprivation, suggesting the involvement of caspases‐mediated apoptosis. We also detected cleaved PARP1 and caspase‐3, both apoptotic indicators, in DKO cells treated with lipid deprivation (Figs 6D and EV4E). We further used Annexin‐V staining to confirm that lipid deprivation significantly induced apoptosis in DKO cells (Fig 6E). Next, to determine whether mitochondria were playing a critical role in DKO cells, we overexpressed Bcl‐2 or Bcl‐xL, which are anti‐apoptotic proteins that protect mitochondrial membrane potential (Shimizu et al, 1998). Our results showed that overexpression of Bcl‐2 or Bcl‐xL significantly increased survival of DKO cells in FBS or DFBS medium; survival of wild‐type cells was also increased in DFBS medium (Figs 6F and EV4F). Furthermore, we found that in normal conditions, wild‐type and DKO cells along with MDA‐MB‐231/TECR‐KO cells had similar mitochondrial potentials (Figs 6G, and EV4G and H). This was detected by JC‐1 as formation of red fluorescent aggregates upon high mitochondrial potential or green fluorescent monomers at low membrane potential (Smiley et al, 1991). However, in conditions of lipid deprivation, a much larger fractions of DKO and TECR‐KO cells lost mitochondrial potential compared to wild‐type cells (Figs 6G, and EV4G and H).

We then used MitoTracker Red to detect mitochondrial morphology and observed that in most MDA‐MB‐231/Wt and HeLa/Wt cells, more than ~90% of mitochondria in a cell displayed elongated tubules (Figs 6H and EV5A). By contrast, more than 90% of ACC‐DKO cells had fragmented or intermediate mitochondria (Figs 6H and EV5A). MDA‐MB‐231/TECR‐KO cells also displayed similar changes in mitochondrial morphology (Fig EV5B). To investigate how fatty acid deficiency affected mitochondrial lipid components, we separated mitochondria from MDA‐MB‐231/WT and DKO cells to perform lipid metabolomics. Our results showed that ACC‐DKO significantly reduced mitochondrial dihydroceramides, ceramides, and glucosylceramides, including Hex1Cer, Hex2Cer, Hex3Cer, GM3, and CerG3GNAC1 as well as sphingomyelins (Fig 7A). These findings were similar to the changes in total cellular lipids, with the exception of sphingomyelins (Fig 3A). Glucosylceramides and sphingomyelins can be synthesized from ceramides and dihydroceramides (Fig 7B). SPTLC1 is a subunit of serine palmitoyltransferase (SPT), which catalyzes the first step of de novo sphingolipid synthesis, and was reported to cause mitochondrial structural abnormalities (Myers et al, 2014). Here, we knocked out SPTLC1 in MDA‐MB‐231 cells and observed an increased in fragmented mitochondria, along with a significant decrease in cell growth and colony conformation ability (Fig EV5C–F). Therefore, these results suggest that the low levels of cellular VLCFAs most likely influence their incorporation into dihydroceramides and downstream metabolites. Next, we analyzed the composition of fatty acids involved in these lipids and confirmed that the contents of VLCFAs were significantly reduced (Fig 7C), consistent with the changes in total lipids (Fig 3B and C). We also observed that supplementation with VLCFAs significantly restored mitochondrial shape in DKO cells (Fig 7D and E). Taken together, these results suggest that VLCFAs play a critical role in maintaining mitochondrial morphology and function.

Figure EV5. VLCFAs are necessary to maintain mitochondrial morphology.

• A, B
  HeLa/ACC‐DKO (A) and MDA‐MB‐231/TECR‐KO cells were stained with MitoTracker to visualize mitochondria. Scale bar, 10 μm. Mitochondrial morphology was scored in the cell types indicated. Data represent three independent experiments, and at least 100 cells were counted per cell type. **P < 0.01 (t‐test). Asterisk shows comparison with wild‐type group.
• C
  Western blot verification of knockout of SPTLC1 in MDA‐MB‐231 cells.
• D–F
  The cell proliferation rate (D), colony formation (E) and mitochondrial morphology (F) of SPTLC1‐knockout MDA‐MB‐231 cells. Scale bar, 10 μm. Mitochondrial morphology was scored in the cell types indicated. Data represent three independent experiments, and at least 100 cells were counted per cell type. *P < 0.05, **P < 0.01 (t‐test). Asterisk shows comparison with wild‐type group.

Source data are available online for this figure.

Figure 7. VLCFAs are necessary to maintain normal mitochondrial morphology.

1. Relative mitochondrial lipids in MDA‐MB‐231/ACC‐DKO cells. Lipids are abbreviated as in Fig 3A. Error bars represent mean ± SD, and data are from three independent cultures.
2. A scheme to show the synthesis of glucosylceramides and sphingomyelins.
3. Relative cellular levels of lipid‐containing fatty acids in MDA‐MB‐231/ACC‐DKO and Wt cells. The top curve plot indicates the fraction of the corresponding fatty acid. Each colored datum shows the mean from three independent cultures.
4. MDA‐MB‐231/ACC‐DKO cells were in FBS medium with or without exogenous LCFA or VLCFA supplementation and were stained for MitoTracker to visualize mitochondria. Scale bar, 10 μm.
5. Mitochondrial morphology in (D) was scored in the cell types indicated. Data represent three independent experiments, and at least 100 cells were counted per cell type. Error bars represent mean ± SD, and data are from triplicate experiments.

Data information: *P < 0.05, **P < 0.01 (t‐test).

Discussion

Actively dividing cells, such as cancer cells, must duplicate cellular membranes, and this requires a large amount of fatty acids, which are synthesized or absorbed from the microenvironment (Snaebjornsson et al, 2020). The dietary fatty acids primarily consist of long‐chain fatty acids, including palmitate, oleate, stearate, linoleate, and linolenate (Pepino et al, 2014). This makes cellular VLCFAs more reliant on de novo synthesis or elongation from shorter dietary fatty acids, which requires malonyl‐CoA as the elongation unit (Jakobsson et al, 2006); malonyl‐CoA is product of ACC1/2. Therefore, blocking ACC1/2, rather than FASN, effectively reduces the cellular levels of VLCFAs (Fig 2D). Herein, we demonstrate that VLCFAs are reduced in ceramides and downstream glucosylceramides or sphingomyelins with ACC1‐KO or ACC‐DKO, to a much greater extent than with FASN‐KO (Fig 3). Thus, VLCFAs may be more important for cancer growth in certain lipid species. Alternatively, there could exist a threshold of lipids required for cancer growth. Furthermore, VLCFA deficiency seems to impair mitochondrial morphology and renders cells sensitive to oxidative stress. These observations are consistent with previous reports showing that inactivation of SPTLC1 or ceramide synthase 2 (CerS2) involved in sphingolipid synthesis cause mitochondrial structural abnormality or functional disruption (Pewzner‐Jung et al, 2010a, 2010b; Zigdon et al, 2013; Myers et al, 2014). In addition, sphingolipids are also required for tumor drug resistance and maintenance of cancer stem cell (Ogretmen, 2018). This may, at least in part, account for the importance of VLCFAs in tumor progression.

Due to the differential abundance in dietary nutrients, our results now show that it is feasible to control the cellular levels of VLCFAs, but not LCFAs. Among the four enzymes involved in VLCFA elongation, several isoforms of the seven ELOVL family members are reported to be important in various types of cancer (Tamura et al, 2009; Yamashita et al, 2017). Notably, knocking out KAR and TECR does not completely inhibit the elongation of fatty acids (Fig 4H), suggesting the presence of compensatory counterparts, which remains to be further explored. Nonetheless, our results still suggest that TECR could be a potential target to control cellular VLCFAs and suppress tumor growth. In addition, targeting ACC1/2 could also be an alternative approach for controlling cellular VLCFAs because of its ability to provide malonyl‐CoA as the main two‐carbon elongation unit. However, malonyl‐CoA is also used to posttranslationally modify proteins for their normal functions (Bruning et al, 2018). Therefore, the serious inhibition of ACC1/2 could be potentially toxic to cells. To our surprise, FASN deletion increases the cellular levels of almost all fatty acids detected in our study, and this observation is also supported by a recent report that FASN‐mutant cells strongly depend on lipid uptake (Aregger et al, 2020). The increased uptake of fatty acids could largely attenuate the tumor suppressive effect of FASN inhibition and may create a challenge to the development of FASN inhibitors for treating related diseases, such as cancer and non‐alcoholic fatty liver disease.

Overall, our findings further increase the understanding of the involvement of fatty acid biosynthesis in cancer, including elongation of VLCFAs and uptake in cancer cells. Our results suggest potential mechanisms underlying the role of VLCFAs in cancer growth and survival and question the limited efficacy of FASN inhibition on tumor growth. Taken together, our findings have the potential to guide treatment strategies for cancer through controlling cellular fatty acids.

Materials and Methods

Reagents and Tools table

Reagent or Resource	Source	Identifier
Antibodies
Anti‐FASN (clone G‐11)	Santa Cruz	CAT# sc‐48357
Anti‐ACC1	ProteinTech	CAT# 21923‐1‐AP
Anti‐ACC2 (clone D5B9)	Cell Signaling	CAT# 8578
Anti‐KAR	Abcam	CAT# ab236990
Anti‐TECR	Sigma	CAT# HPA056488
Anti‐SPTLC1	ProteinTech	CAT#15376‐1‐AP
Anti‐β‐Actin	Abclonal	CAT# AC026
Anti‐β‐Tublin	Abclonal	CAT# AC008
Anti‐Bcl2	ProteinTech	CAT# 12789‐1‐AP
Anti‐Bcl‐xL	ProteinTech	CAT# 26967‐1‐AP
Goat Anti‐Mouse	Jackson	CAT# 115‐035‐003
Goat Anti‐Rabbit	Jackson	CAT# 111‐035‐003
Chemicals
FBS	Gemini	CAT# 900‐108
Delipidated FBS	PAN	CAT# P30‐3402
Dialyzed FBS	This paper	N/A
Dialyzed delipidated FBS	This paper	N/A
Matrigel	BD Biosciences	CAT# 356234
Penicillin and streptomycin	Gibco	CAT# 15140122
D‐GLUCOSE (U‐13C6)	CIL	CAT# CLM‐1396‐1
DMEM power	Caisson	CAT# DMP52‐10LT
Bovine Serum Albumin	Sigma	CAT# A8806
Lipofectamine™ 3000 Reagent	Invitrogen	CAT# L3000‐015
Acetyl‐Coenzyme A	Sigma	CAT# 10101893001
Coenzyme A hydrate	Sigma	CAT# C4282
Malonyl coenzyme A lithium salt	Sigma	CAT# M4263
Palmitate	Sigma	CAT# P0500
Stearate	Sigma	CAT# S4751
Arachidate	Sigma	CAT# A3631
Oleate	Sigma	CAT# O1008
Linoleate	Sigma	CAT# L1012
α‐Linolenate	Sigma	CAT# L2376
z‐VAD‐FMK	MCE	CAT# HY‐16658B
Necrostain‐1	Selleck	CAT# S8037
Disulfiram	MCE	CAT# HY‐B0240
Ferrostain‐1	MCE	CAT# S7243
Trolox	MCE	CAT# HY‐101445
N‐Acetyl‐l‐cysteine	Selleck	CAT# S1623
Deferoxamine mesylate	Selleck	CAT# S5742
Tween 80	MCE	CAT# HY‐Y1891
Methyl cellulose	MCE	CAT# HY‐125861
JC‐1	Abcam	CAT# ab141387
H2DCFDA	MCE	CAT# HY‐D0940
Mito Tracker	Invitrogen	CAT# M22425
Hoechst33258	Solarbio	Cat# C0021
Critical Commercial Assays
CellTiter‐Glo® Luminescent Cell Viability Assay	Promega	Cat# G7571
Experimental models: cell lines
Human: MDA‐MB‐231 (female, 51 years old)	ATCC	Cat# HTB‐26
Human: HeLa (female, 31 years old)	ATCC	Cat# CCL‐2
Human: HEK293T (female, fetus)	ATCC	Cat# CRL‐3216
Human: MCF7 (female, 69 years old)	ATCC	Cat# HTB‐22
Human: MCF‐10A (female, 36 years old)	ATCC	Cat# CRL‐10317
Experimental Models: Organisms/Strains
Mouse: Nude mice (female, 6–8 weeks old)	Experimental Animal Center of Nanjing Biomedical Research Institute	N/A
Recombinant DNA
pCDH‐TECR	This paper	N/A
pLentiGuide‐sgACC1‐1	This paper	N/A
pLentiGuide‐sgACC1‐2	This paper	N/A
pLentiGuide‐sgACC2‐1	This paper	N/A
pLentiGuide‐sgACC2‐2	This paper	N/A
pLentiGuide‐sgFASN‐1	This paper	N/A
pLentiGuide‐sgFASN‐2	This paper	N/A
pLentiGuide‐sgKAR‐1	This paper	N/A
pLentiGuide‐sgKAR‐2	This paper	N/A
pLentiGuide‐sgTECR‐1	This paper	N/A
pLentiGuide‐sgTECR‐2	This paper	N/A
lentiCRISPRv2‐sgTECR‐1	This paper	N/A
lentiCRISPRv2‐sgTECR‐2	This paper	N/A
lentiCRISPRv2‐sgSPTLC1‐1	This paper	N/A
lentiCRISPRv2‐sgSPTLC1‐2	This paper	N/A
Sequences of Oligonucleotides for sgRNA Constructs
ACC1‐sg1: GAAGACCTTAAAGCCAATGC	This paper	N/A
ACC1‐sg2: CAACAACAACTATGCAAATG	This paper	N/A
ACC2‐sg1: TACCTGCACGGGGATTCTCT	This paper	N/A
ACC2‐sg2: CGTTGTCGCCCAGACGCTAC	This paper	N/A
FASN‐sg1: CGACCCACCTCCGTCCACGA	This paper	N/A
FASN‐sg2: TACGCCACCATCCTGAACGC	This paper	N/A
KAR‐sg1: TGATGCAAAGTCAACAGCAA	This paper	N/A
KAR‐sg2: CCTAGAAGAAAAATTCAAAG	This paper	N/A
TECR‐sg1: AATATGACTTTACGTCCAGT	This paper	N/A
TECR‐sg2: TGTACTCACTGGGGTCCAGG	This paper	N/A
SPTLC1‐sg1: GTTGGATAACCCTAGGGTTA	This paper	N/A
SPTLC1‐sg2: ACGATGTTGTAGTTGAGAGC	This paper	N/A

Methods and Protocols

Cell lines

MDA‐MB‐231, HeLa, MCF7, MCF10A, and 293T were obtained from the American Tissue Culture Collection (ATCC). The MCF10A cells were cultured in F12/DMEM supplemented with 5% horse serum, 50 ng/ml epidermal growth factor (EGF), 10 μg/ml insulin, 0.5 μg/ml hydrocortisone, 100 IU/ml penicillin, 100 μg/ml streptomycin, and 0.1 μg/ml cholera toxin. MDA‐MB‐231, MCF7, HeLa, and 293T cells were cultured in DMEM (Corning) supplemented with 10% fetal bovine serum (FBS, Gemini) and 100 U/ml penicillin and streptomycin (Gibco). All cells were cultured in a humidified atmosphere of 5% CO₂ at 37°C.

For delipidized medium, cells were first seeded in the medium containing regular 10% FBS and were switched into the medium containing 10% delipidized FBS (PAN) upon treatment on the following day.

Animal

All procedures using animals were approved by the Capital Medical University Institutional Animal Care and Use Committee (IACUC). All experiments were initiated with mice aged 6–8 weeks. The mice were housed in a pathogen‐free animal barrier facility.

Generation of knockout cell lines

ACC1, ACC2, FASN, KAR, TECR, and both of ACC1 and ACC2 were deleted from MDA‐MB‐231, HeLa, and MCF7 cells using the CRISPR‐Cas9 system. pCDH‐Cas9‐2A‐GFP‐BSD was used to express Cas9. Single‐guide RNAs (sgRNAs) were cloned into pLentiGuide‐Vector linearized with BsmBI. sgRNA sequences are listed in the Reagents and Tools table. MDA‐MB‐231, HeLa, and MCF‐7 cells were co‐transfected with Cas9‐EGFP and pLentiGuide‐sgRNA plasmids using lipofectamine 3000 (Invitrogen). In order to generate the ACC1/2‐DKO cell lines, the Cas9‐GFP, sgACC1‐1, and sgACC2‐1 plasmids were co‐transfected into cells. Single cells were sorted into 96‐well plate based on green fluorescence and maintained until colonies formed. Clones were validated for knockout of related genes by Western blotting. TECR restoration was performed by infecting MDA‐MB231/TECR‐KO clone#1 with lentivirus expressing full‐length human TECR cDNA. Single‐guide RNAs (sgRNAs) were cloned into lentiCRISPRv2 (one vector system) linearized with BsmBI. Then, the lentivirus was produced in 293T cells and MDA‐MB‐231 or MCF10A cells were infected with indicated lentivirus. Stable integrates were selected with puromycin.

Western blotting

After desired treatments as specified as indicated, cells were washed twice with PBS and lysed in buffer (20 mM Tris–HCl, pH7.5, 150 mM NaCl, 1 mM EDTA, 1% Triton X‐100, 2.5 mM sodium pyrophosphate, 1 mM b‐glycer‐ophosphate, 1 mM sodium vanadate, 1 mg/ml leupeptin, and 1 mM phenylmethylsulfonyl fluoride) on ice. Equal amounts of protein (30 μg) were loaded onto 8% SDS‐PAGE gels (for ACC1, ACC2, FASN, and pACC) or 12% gels (for KAR and TECR detection). Western detection was performed using a chemiluminescence Western Blot scanner (ChampGel 7000, SAGECREATION, Beijing, China). The goat anti‐rabbit IgG/HRP (Cat#ZDR‐5306, 1:10,000 dilution) and the goat anti‐mouse IgG/HRP (Cat#ZDR‐5307, 1:10,000 dilution) secondary antibodies were obtained from Jackson. For protein detection, the following antibodies were used: FASN (clone G‐11; Santa Cruz, 1:1,000 dilution), ACC1 (Polyclonal; Proteintech, 1:2,000 dilution), ACC2 (clone D5B9, Cell Signaling, 1:500 dilution), phosphorylation ACC (clone D5B9, Cell Signaling, 1:1,000 dilution), KAR (Abcam, 1:2,000 dilution), TECR (Sigma, 1:1,000 dilution), β‐Actin (Abclonal, 1:20,000 dilution), Bcl₂ (Proteintech, 1:2,000 dilution), and β‐tublin (Abclonal, 1:5,000 dilution).

Cell proliferation and survival assay

As for the cell proliferation, cells were plated in triplicate in 12‐well plates at 10⁴ cells per well. After the desired treatments as indicated in the figures, cells were washed twice with phosphate‐buffered saline (PBS) to remove dead cells and then were trypsinized. The cell number was determined using Muse Cell Analyzer (Luminex Cooperation), with cellular debris being excluded. For each well, the fold‐change in cell number relative to Day₀ was determined.

As for the cell survival assay, 2,000 cells per well were seeded in 96‐well plates and were treated on the following day either with vehicle or drugs with the indicated concentrations in the figures for 24 h. Then, cell viability was determined by using the CellTiter‐Glo Luminescent Cell viability Assay Kit (Promega).

Preparation of fatty acid/BAS complex

Fatty acid–rescuing experiments were performed in the full or delipidized FBS medium with fatty acids as indicated. The FA‐BSA was prepared according to a previously described protocol (Cousin et al, 2001). Briefly, palmitate, oleate, linoleate, and linolenate were solubilized in 0.1 M NaOH (heating at 70°C) at 100 mM separately and combined with 10% (wt/vol) FFA‐free Bovine Serum Albumin (BSA, Sigma; heating at 55°C) to make a fatty acid concentration of 5 mM. Stearate and other saturate FAs were solubilized in trichloromethane at 200 mM separately and combined with 10% (wt/vol) FFA‐free BSA (heating at 55°C) to make a concentration of 2 mM. Stored FFA/10% BSA stock solutions were heated for 15 min at 55°C and then cooled to room temperature before use.

Lentivirus production

Viral packing was done as previously described (Li et al, 2010). Briefly, pCDH‐CMV‐cDNA, pCMV‐Dr8.91, and pCMV‐VSV‐G were co‐transfected into 293T cells using polyethylenimie, and then media containing the virus was collected at 48 h after transfection. Cells were infected with medium containing viruses in the presence of polybrene (10 μg/ml) for 48 h, and then cells were selected with puromycin.

Colony formation assay

Five‐hundred cells were maintained in culture media in 12‐well plates for 10 days (MDA‐MB‐231) or 14 days (HeLa), fixed with 4% paraformaldehyde, stained with 0.1% crystal violet for colony observation, and counted using a light microscope.

Cell invasion assay

Cell invasion was assayed using 24‐well inserts with 8 μM pore inserts (Cat#3422, Costar) coated with Matrigel (BD Biosciences). 2 × 10⁴ indicated cells suspended in 0.5 ml medium without FBS were plated into the upper chamber. The chamber was then transferred to a well containing 0.5 ml of medium with 10% FBS. Non‐invading cells were removed after 16 h culture. The cells that migrated through the membranes were fixed with paraformaldehyde for 30 min, and stained with 0.1% crystal violet for 20 min. The number of cells was counted under an Olympus microscope over five random fields in each well. Each assay was performed in triplicate.

Dialyzed FBS preparation

The 80 ml FBS or delipidized FBS was put into 20 cm dialysis membrane (7 kDa cutoff, Cat#YA1076, Solarbio), and dialyzed against 2 l 0.15 M NaCl at 4°C for 2 h. Then the NaCl solution was refreshed for 2–3 times. The FBS was dialyzed until glucose < 5 mg/dl, and then proceed to be filtered with 0.22 μm membrane under sterile condition. This dialyzed FBS was used for isotope labeling experiments.

Acetyl‐CoA and Malonyl‐CoA levels analysis

LC–MS/MS analysis was conducted on a TSQ Quantiva triple quadrupole mass spectrometer networked to a Dionex Ultimate 3000 UPLC system (Thermo Fisher Scientific). Experiments were performed in phenol red‐free medium containing 10% FBS‐Dialyzed. Cells were grown in 60‐mm dishes until 80% confluent. The medium was aspirated completely and cells were washed with PBS three times quickly and gently. Cells were extracted in 1 ml 80% (vol/vol) methanol (pre‐chilled to −80°C) and incubated at −80°C for 2 h. Following extraction, samples were centrifuged for 20 min at 14,000 × g (cooled at 4°C). The supernatant was transferred to a vial and dried under nitrogen gas (room temperature). Dried samples were stored at −80°C and then resuspended in 50 μl 80% methanol for LC–MS/MS analysis. Each sample was loaded onto a Synergi Hydro‐RP column (2.1 × 100 mm, 100A, Phenomenex) for metabolite separation. Column temperature was set at 35°C. Mobile phases A and B were 10 mM tributylamine in aqueous with pH 5 and 100% methanol, respectively. The chromatographic gradient was set as follows: 0–3.5 min: 1%B; 3.5–22 min: from 1 to 70% B; 22–23 min: from 70 to 90% B; 23–25 min: 90% B; 25–30 min: 1% B. Data were acquired using positive/negative switching method. Spray voltages of 3.5 and 2.5 kV were applied for positive and negative modes, respectively. Q1 and Q3 resolution set at 0.7 and 1 s of cycle time was used in the method. For data analysis, we integrated the peak areas using the Thermo XCalibur Quan Brower Software (Thermo Fisher Scientific). The ion transitions were used as follows in the analysis: CoA: 768.3‐ > 261.1, acetyl‐CoA: 810.1‐ > 303.4 and malonyl‐CoA: 854.1‐ > 347.1. Relative abundance based on chromatographic peak area is used for quantitation.

Isotope tracing of fatty acids

Fatty acids analysis was performed on Q Exactive orbitrap mass spectrometer (Thermo Fisher Scientific). Experiments were performed in the medium containing 10% FBS‐Dialyzed or 10% delipidized FBS‐Dialyzed. DMEM without glucose, glutamine, pyruvate, and phenol red was prepared from power (Caisson) by adding 3.7 g NaHCO₃ per liter and adjusting the pH to 7.4, then supplemented with 10 or 25 mM of [U‐¹³C]‐glucose (as indicated in the legends), 2 mM glutamine and 1 mM pyruvate. For saponified fatty acid analysis, 6 × 10⁵ cells were seeded in 60‐mm dishes overnight, then rinsed with PBS and cultured with 5 ml [U‐¹³C]‐glucose‐containing medium for 48 h. Samples were prepared as previously described (Kamphorst et al, 2013). The medium was aspirated completely and cells were washed with PBS three times quickly and gently. Cells were extracted with 50% (vol/vol) methanol solution containing 0.1 M HCl (prechilled to −20°C) at 10⁶ cells/ml. The resulting liquid and cell debris were scraped into a glass vial. Chloroform (0.5 ml) was added, the mixture was vortexed for 1 min and then centrifuged for 15 min at 1,700 g, and the chloroform layer was transferred to a glass vial. Equal amounts of extracts from different treatments were dried under nitrogen gas (room temperature), reconstituted into 1 ml 90% methanol solution containing 0.3 mM KOH, incubated at 80°C for 1 h to saponify fatty acids, acidified with 0.1 ml of formic acid, extracted twice with 1 ml of hexane, and dried under nitrogen gas. Dried samples were stored at −80°C and then resuspended in 150 μl dichloromethane: methanol for LC–MS/MS analysis. Cortecs C18 column (2.1 × 100 mm; Waters) was applied for analysis. Mobile phase A was prepared by dissolving 0.77 g of ammonium acetate in 400 ml of HPLC‐grade water, followed by adding 600 ml of HPLC‐grade acetonitrile. Mobile phase B was prepared by mixing 100 ml of acetonitrile with 900 ml isopropanol. A 18‐min gradient with flow rate of 250 μl/min was used. Linear gradient was as follows: 0 min, 30% B; 2.5 min, 30% B; 8 min 50% B; 10 min, 98% B; 15 min 98% B; 15.1 min, 30% B; 18 min 30% B. Data with mass ranges of m/z 150–600 were acquired at negative ion mode. The full scan was collected with resolution of 70,000. The source parameters are as follows: spray voltage: 3 kv; capillary temperature: 320°C; heater temperature: 300°C; sheath gas flow rate: 35 Arb; auxiliary gas flow rate: 10 Arb. Isotope labeled fatty acids were assigned based on in‐house database containing ¹³C labeled medium to very long‐chain fatty acids according to accurate ion masses using Tracefinder 3.2 (Thermo Fisher Scientific). Mass tolerance of 10 ppm was applied for precursor mass matching. Chromatographic peak area was used for relative quantitation.

Lipidomics analysis using LC–MS/MS

LC/MS analyses were conducted on a Q Exactive orbitrap mass spectrometer networked to a Dionex Ultimate 3000 UPLC system (Thermo Fisher Scientific). Experiments were performed in phenol red‐free medium containing 10% FBS‐Dialyzed. Cells were grown in 60‐mm dishes until 80% confluent. Media were aspirated, and cells were rinsed twice with room temperature PBS and then scraped with 1 ml room temperature PBS. The resulting liquid and cell debris were transferred into a tube. Four milliliter of CHCl₂: MeOH (2:1) was added, the mixture was vortexed for 1 min three times and then centrifuged at 1,700 g for 15 min. The chloroform layer was transferred to a glass vial. The extract was dried under nitrogen gas (room temperature). The gradient was as below: 0 min, 37% B; 1.5 min, 37% B; 4 min, 45% B; 5 min, 52% B; 8 min, 58% B; 11 min, 66% B; 14 min, 70% B; 18 min, 75% B; 20 min, 98% B; 22 min, 98% B; 22.1 min, 37% B; 25 min, 37% B. The detailed mass spectrometer parameters were as follows: spray voltage, 3.2 kV for positive and 2.8 kV for negative; capillary temperature, 320°C; sheath gas flow rate, 35 Arb; aux gas flow rate, 10 Arb; mass range (m/z), 240–2,000 for positive and 200–2,000 for negative. Full MS resolution, 60,000; MS/MS resolution, 15,000. Lipids were identified and quantified using LipidSearch 4.1.30 (Thermo, CA). Mass tolerance of 5 and 10 ppm was applied for precursor and product ions. Retention time shift of 0.25 min was performed in “alignment.” M‐score and chromatographic areas were used to reduce false positives.

Mitochondrial lipids targeted metabolomics

Mitochondria was isolated according to a previously described protocol (Frezza et al, 2007). Briefly, cells were washed and collected with PBS. Then, the cell pellet was resuspended in ice‐clod IBc buffer (10 μM Tris‐MOPS, 1 mM EGTA/Tris, 0.2 mM sucrose, pH 7.4) and homogenized using a Teflon pestle operated at 500 g. The homogenate was centrifuged at 600 g for 10 min, and the supernatant was transferred to a glass centrifuge tube and centrifuged at 7,000 g for 10 min. The mitochondrial pellet was washed with ice‐cold IBc and resuspended in 200 μl ice‐cold IBc. All of the above experiment operations were conducted at 4°C. Then the mitochondrial lipids were extracted, and LC–MS/MS analyses were conducted as described in the section of “Lipidomics Analysis using LC–MS/MS.”

Fatty acid levels in FBS and Delipidized FBS (LC/MS analysis)

In total, 800 μl CHCl₂: MeOH (2:1) was added to 200 μl FBS or delipidized FBS. The lipids in the serum were extracted as described in the section of “Lipidomics Analysis using LC–MS/MS” and proceeded to be saponified and extracted and detected as described in the section of “Isotope tracing of fatty acids.”

Mitochondrial morphology

Mitochondria were stained with 400 nM Mito Tracker (Cat#M22425, Invitrogen) for 40 min and nuclear were stained with 1 μg/ml Hoechst 33258 (Cat# C0021, Solarbio) for 5 min. Images of mitochondrial morphology of live cells were captured using a confocal laser‐scanning microscope (LSM880, Carl Zeiss) with a 63× oil‐immersion objective with 1.4 numerical aperture (Plan‐Apochromat 63×/1.4 Oil DIC M27, Carl Zeiss). A 594‐nm laser (excitation: 594 nm, emission: 667 nm) was used to visualize mitochondria. Mitochondrial morphology quantifications and comparisons were done as previously described (Karbowski et al, 2006). Cells were divided into three categories, “Elongated/tubular” with > 90% of mitochondria forming elongated interconnected networks, “Intermediate” with mixed tubular and short mitochondria, and “Fragmented” with > 90% short punctiform mitochondria.

Intracellular ROS measurement

Intracellular ROS levels were detected using H₂DCFDA (Cat#HY‐D0940, MCE) according to manufacturer's instructions. Cells were cultured with serum‐free DMEM medium containing 5 μM H₂DCFDA for 20 min. After removing the medium, cells were washed with PBS and trypsinized. Then, cells were collected and resuspended in PBS, and fluorescence intensity was examined using BD LSRFortessa cell analyzer (BD Biosciences).

Mitochondrial membrane potential

The cells were washed with PBS and trypsinized. Then cells were collected and incubated for 20 min with 1 μΜ 5,6‐Dichloro‐2‐[(E)‐3‐(5,6‐dichloro‐1,3‐diethyl‐1,3‐dihydro‐2H‐benzimidazol‐2‐ylidene)‐1‐prop‐1‐enyl]‐1,3‐diethyl‐1H‐benzimidazolium iodide (JC‐1, Cat#ab141387, Abcam) in culture medium at 37°C. After washing with PBS, fluorescence intensity of cells was examined using BD LSRFortessa cell analyzer (BD Biosciences).

Animal studies

All procedures using animals were approved by the Capital Medical University Institutional Animal Care and Use Committee (IACUC). All experiments were initiated with mice aged 6–8 weeks. The mice were housed in a pathogen‐free animal barrier facility.

ACC1‐KO (KO#1 of MAD‐MB‐231 or HeLa), ACC2‐KO (KO#1 of MDA‐MB‐231), ACC‐DKO (KO#1 of MDA‐MB‐231 or HeLa), FASN‐KO (KO#2 of MDA‐MB‐231 or HeLa), KAR‐KO (KO#1 of MDA‐MB‐231 or HeLa), and TECR‐KO (KO#1 of MDA‐MB‐231 or HeLa) cells were used. Tumor cells in a volume of 100 μl were injected subcutaneously into the hind flanks of 6‐week‐old athymic female nude mice. 5 × 10⁶ cells (Fig 1F left, MDA‐MB‐231 cells, N = 4; Fig 4D, MDA‐MB‐231 cells, N = 6), 2 × 10⁶ cells (Figs 1F right and EV3B, HeLa cells, N = 6), and 10⁶ cells (Fig 4E, MDA‐MB‐231 cells, N = 8) were injected for a period of 30 days (Figs 1F left and 4D) or 24 days (Figs 1F right and EV3B) or 18 days (Fig 4E), respectively. All of the mice were killed at the end and tumors were harvested and weighed.

Statistical analysis

Data are given as means ± SD. Statistical analyses were performed using unpaired two‐tailed Student's t‐test for comparison between two groups. Asterisks in the figures indicated statistical significances (*, P < 0.05; **, P < 0.01).

Author contributions

Qiaoyun Chu: Validation; investigation; methodology; writing – original draft. Ping Liu: Validation; investigation; methodology. Yihan Song: Validation; investigation; methodology. Ronghui Yang: Validation; investigation; methodology. Jing An: Investigation. Xuewei Zhai: Resources. Jing Niu: Resources; visualization. Chuanzhen Yang: Resources; visualization. Binghui Li: Conceptualization; formal analysis; supervision; visualization; writing – original draft; project administration; writing – review and editing.

Disclosure and competing interests statement

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Expanded View Figures PDF

Source Data for Expanded View and Appendix

PDF+

Source Data for Figure 1

Source Data for Figure 4

Source Data for Figure 6

The EMBO Journal (2023) 42: e111268

Data availability

The mass spectrometry lipidomics data were deposited to the National Omics Data Encyclopedia with the dataset identifier OEP003708 (http://www.biosino.org/node/project/detail/OEP003708).