doi: 10.1038/s41586-019-0904-1.
Epub 2019 Feb 6.
Evidence for an alternative fatty acid desaturation pathway increasing cancer plasticity
Affiliations
- PMID: 30728499
- PMCID: PMC6390935
- DOI: 10.1038/s41586-019-0904-1
Item in Clipboard
Evidence for an alternative fatty acid desaturation pathway increasing cancer plasticity
Nature.
2019 Feb.
Abstract
Most tumours have an aberrantly activated lipid metabolism1,2 that enables them to synthesize, elongate and desaturate fatty acids to support proliferation. However, only particular subsets of cancer cells are sensitive to approaches that target fatty acid metabolism and, in particular, fatty acid desaturation3. This suggests that many cancer cells contain an unexplored plasticity in their fatty acid metabolism. Here we show that some cancer cells can exploit an alternative fatty acid desaturation pathway. We identify various cancer cell lines, mouse hepatocellular carcinomas, and primary human liver and lung carcinomas that desaturate palmitate to the unusual fatty acid sapienate to support membrane biosynthesis during proliferation. Accordingly, we found that sapienate biosynthesis enables cancer cells to bypass the known fatty acid desaturation pathway that is dependent on stearoyl-CoA desaturase. Thus, only by targeting both desaturation pathways is the in vitro and in vivo proliferation of cancer cells that synthesize sapienate impaired. Our discovery explains metabolic plasticity in fatty acid desaturation and constitutes an unexplored metabolic rewiring in cancers.
Conflict of interest statement
AH, CCK, AS, PS, SvC and SG have competing interests as employees of Bayer AG. KKW is a founder and equity holder of G1 Therapeutics and he has Consulting/Sponsored Research Agreements with AstraZeneca, Janssen, Pfizer, Array, Novartis, Merck, Takeda, Ono, Targimmune and BMS. SMF has received funding from Bayer AG and Merck.
Figures
(a) Schematic overview of fatty acid metabolism. AcCoA:
Acetyl-coenzyme A; SCD1/5: Stearoyl-CoA desaturase 1 and 5; Elovl5/6:
elongation of very long chain fatty acids protein 5 and 6. (b-e) SCD desaturation activity based on the
palmitoleate to palmitate ratio, oleate to stearate ratio, palmitoleate and
palmitate synthesis upon Merck Frosst Cpd 3j treatment (HUH7, A549: 2 nM;
H460, DU145: 1 nM; MDA-MB-468, T47D: 0.5 nM; panel b-d: n=3;
panel e: HUH7 n=3, A459 n=3, H460 n=6 (control) n=4 (SCD
inhibitor), DU145 n=3, MDA-MB-468 n=5, T47D n=5 (control) n=6 (SCD
inhibitor)). Unpaired two-sided Student’s T-tests with Holm-Sidak
multiple comparisons. (f-h) Correlation between SCD independence
and palmitate synthesis, growth rate or total fatty acid abundance (n=3).
SCD independence was defined as area under the cell number curve of Figure 1a. Palmitate synthesis was
derived from (e). Total fatty acid abundance was derived from Extended Data Figure 2a. Trend line
(dashed line) and 95% confidence intervals (dotted lines) are depicted. Cancer cell experiments were performed in low FBS DMEM (1%: HUH7;
0.5%: others) with treatment of 72 h. Error bars represent mean ± SD
from biological independent samples.
(a) Heat map representing fatty acid abundances
with(out) Merck Frosst Cpd 3j treatment (HUH7, A549: 2 nM; H460, DU145: 1
nM; MDA-MB-468, T47D: 0.5 nM) normalized to highest abundance of each fatty
acid across all cell lines/conditions (Figure
1b, Supplementary Table
1a). Over 90% reduction: white, no reduction: dark
green. (b,c) Desaturation activity to sapienate upon Merck
Frosst Cpd 3j treatment (HUH7, A549: 2 nM; H460, DU145: 1 nM; MDA-MB-468,
T47D: 0.5 nM; n=3). Unpaired two-sided Student’s T-tests with
Holm-Sidak multiple comparisons. (d) Sapienate to palmitate ratio in HUH7 (n=6)
versus freshly isolated primary human hepatocytes (PHH;
n=3), DU145 (n=6) versus RWPE-1 (n=6) prostate cells, and
MDA-MB-468 (n=6) and T47D (n=6) versus MCF10A (n=6) breast
cells. Unpaired Student’s T-tests and Welch’s correction (HUH7
versus PHH; DU145 versus RWEP-1);
one-way ANOVA with Dunnett’s multiple comparisons (MDA-MB-468, T47D
versus MCF10A). (e) Tumor weight of HUH7 subcutaneous xenografts
treated with(out) Merck Frosst Cpd 3j (n=8 one experiment; 1.5 mg per kg
twice daily p.o.). Unpaired Student’s T-test with Welch’s
correction. (f,g) FADS2 gene
expression in cells with(out) Merck Frosst Cpd 3j as described in (b,c)
normalized to T47D cells (n=3). One-way ANOVA with Tukey’s multiple
comparisons (f); unpaired Student’s T-tests with Holm-Sidak multiple
comparisons (g). (h) FADS2 protein expression in the same conditions as
in (d). Statistics as described in (d). n=3. (i) FADS2 gene/protein expression in HUH7 and A549
cells upon FADS2 silencing normalized to control (Gene:
HUH7 n=3, A549 n=6; protein n=3 except for A549 shFADS2-2 n=2). One-way
ANOVA with Dunnett’s multiple comparisons. Cancer cell experiments were performed in low FBS DMEM (1%: HUH7;
0.5%: others) with treatment of 72 h. Error bars represent SD (in
vitro) or SEM (in vivo) from mean of
biological independent samples (in vitro) or animals
(in vivo).
(a) Relative FADS2 gene/protein expression and
desaturation activity to sapienate in MDA-MB-468 control and FADS2
overexpression cells with DMSO or 0.5 nM Merck Frosst Cpd 3j normalized to
control (n=3). Unpaired two-sided Student’s T-test. (b) Relative FADS2 gene expression in
tumor nodules from HUH7 control or FADS2 knockdown
orthotopic xenografts with vehicle or Merck Frosst Cpd 3j (1.5 mg per kg
twice daily per oral; p.o.; n=4; one experiment) normalized to control.
One-way ANOVA with Tukey’s multiple comparisons. (c,d) Relative desaturation activity from
palmitate to sapienate or palmitoleate in normal adjacent liver (L) and
tumor nodules (T) in the same model as described in (f) normalized to normal
control livers. Control+vehicle-L n=18 (c) n=20
(d); control+vehicle-T n=18 (c) n=20
(d); control+SCD inhibition-L n=14 (c,
d); control+SCD inhibition-T n=13 (c) n=14
(d); shFADS2-2+vehicle-L n=19 (c,
d); shFADS2-2+vehicle-T n=18 (c,
d); shFADS2-2+SCD inhibition-L n=15 (c) n=16
(d); shFADS2-2+SCD inhibition-T n=15 (c,
d); two experiments. Two-way ANOVA with Sidak’s
multiple comparisons. (e,f) Desaturation activity from linoleate
to γ-linolenate based on the γ-linolenate to linoleate ratio
and arachidonate abundance in HUH7 and A549 control (non-targeting shRNA)
and FADS2 knockdown (shFADS2) cells (n=3). One-way ANOVA
with Dunnett’s multiple comparisons. (g,h) Linoleate and arachidonate abundance
in normal adjacent murine liver and tumor nodules from HUH7 control
(non-targeting shRNA) or FADS2 knockdown (shFADS2)
orthotopic xenografts treated with vehicle or Merck Frosst Cpd 3j (1.5 mg
per kg twice daily per oral; p.o.). Control+vehicle-L n=12 (g)
n=14 (h); control+vehicle-T n=13 (g) n=14
(h); control+SCD inhibition-L n=14 (g) n=15
(h); control+SCD inhibition-T n=14 (g) n=16
(h); shFADS2-2+vehicle-L n=14 (g) n=16
(h); shFADS2-2+vehicle-T n=13 (g) n=15
(h); shFADS2-2+SCD inhibition-L n=15 (g) n=18
(h); shFADS2-2+SCD inhibition-T n=15 (g) n=16
(h); two experiments. Two-way ANOVA with Tukey’s
multiple comparisons. Cancer cell experiments were performed in low FBS DMEM (1% : HUH7;
0.5% : others) with treatment of 72 h. Error bars represent SD (in
vitro) or SEM (in vivo) from mean of biolocial
independent samples (in vitro) or animals (in
vivo).
(a-f) 13C enrichment of palmitate or
stearate from 13C6 glucose in HUH7 or A549 cells in
control condition (ethanol, black) or upon 12C sapienate
supplementation (blue). Cells were grown in 10% dialyzed FBS DMEM containing
4.5 g per L 13C6 glucose for 1 week, after which cells
were grown for 72 h in 0.5% FBS DMEM containing 4.5 g per L
13C6 glucose supplemented with ethanol or 20
µM 12C sapienate. The purpose of this experiment was to trace the incorporation of
carbons from sapienate into cis-8-octadecenoate. Palmitate and stearate were
measured as controls. Since 13C-labeled sapienate is not
commercially available, we performed a reverse labeling in which we
pre-labeled HUH7 and A549 cells with 13C6-glucose to
enrich cis-8-octadecenoate with 13C. Then, we supplemented these
cells with unlabeled sapienate in the presence of
13C6-glucose and determined the 13C
enrichment of octadecenoate. If sapienate is elongated to
cis-8-octadecenoate, we expect a shift in the 13C enrichment from
higher to lower octadecenoate isotopologues. Indeed, we found that
supplementation of unlabeled sapienate shifted the 13C enrichment
accordingly (a, d). Moreover, the largest 13C
enrichment increase was found in the M+2 isotopologue, indicating the
elongation of unlabeled sapienate to octadecenoate with 13C
labeled acetyl-CoA. As expected, sapienate supplementation did not or only
marginally change the 13C enrichment of palmitate and stearate
(b,c,e,f). Unpaired two-sided Student’s T-tests; n=3. Error bars
represent mean ± SD from biological independent samples.
(a) Relative cis-8-octadecenoate abundances in cancer
cells normalized to T47D cells. HUH7 n=3; A549 n=3, H460 n=5, DU145 n=3,
MDA-MB-468 n=5, T47D n=5. One-way ANOVA with Tukey’s multiple
comparisons. (b,c) Relative cis-8-octadecenoate
abundances in HUH7 and A549 control (non-targeting shRNA) and
FADS2 knockdown (shFADS2) cells in control condition
(ethanol) or upon 20 µM sapienate supplementation normalized to
control. HUH7: control n=6 (ethanol) n=3 (sapienate); shFADS2-1 n=3;
shFADS2-2 n=6 (ethanol) n=3 (sapienate); A549: control n=6; shFADS2-1 n=3;
shFADS2-2 n=3. Data values are shown in Supplementary Table 1c,
d. Two-way ANOVA with Tukey’s multiple
comparisons. (d) Relative proliferation of MDA-MB-468 cells with
ethanol (n=9) or 20 µM cis-8-octadecenoate (n=3) upon treatment with
DMSO or 0.5 nM Merck Frosst Cpd 3j. Data were normalized to control with
error bars representing SEM. Two-way ANOVA with Tukey multiple
comparisons. (e,f) Relative proliferation of HUH7 and
A549 control (non-targeting shRNA) and knockdown (shFADS2) cells with
ethanol or 20 µM cis-8-octadecenoate upon treatment with DMSO or 2 nM
Merck Frosst Cpd 3j. HUH7: control n=9; shFADS2-1 n=6; shFADS2-2 n=9; A549:
EtOH n=6; cis-8-C18:1 n=3. Data were normalized to control. Two-way ANOVA
with Tukey multiple comparisons. Only statistics for pair-wise comparisons
are depicted. Cancer cell experiments were performed in low FBS DMEM (1%: HUH7;
0.5%: all other cancer cells) with treatment of 72 h. Error bars represent
mean ± SD from biological independent samples, unless otherwise
noted.
(a-d) Heat map representing abundance changes of
phosphatidylcholine (a), phosphatidylethanolamine
(b), phosphatidylserine (c) and
phosphatidylinositol (d) species in control and
FADS2 knockdown HUH7 and A549 cells relative to
control. HUH7: control n=3; shFADS2-1 n=4; shFADS2-2 n=5; A549: n=5. Only
significant differences are depicted as log2 fold change compared
to control. X denotes blank or excluded values. Phospholipid species
carrying sapienate or palmitoleate are depicted in bold red and listed in
Supplementary Table
1e. Two-way ANOVA with Dunnett’s multiple
comparisons. (e) Relative distribution of phospholipid species in
HUH7 (n=2) and A549 (n=5) cell with non-targeting shRNA (control). PC:
phosphatidylcholine; PE: phosphatidylethanolamine; PS: phosphatidylserine;
PI: phosphatidylinositol; SM: sphingomyelin. (f) Membrane fluidity based on the ordered to
disordered ratio in HUH7 and A549 with a non-targeting shRNA (control;
black) or two different shRNA targeting FADS2 (brown and orange) normalized
to control (n=4). The higher the ordered to disordered ratio, the more
saturated lipids are present in the membrane. One-way ANOVA with
Dunnett’s multiple comparisons. (g) Lipid peroxidation sensitivity via MDA assay in
HUH7 with a non-targeting shRNA (control; black) or two different shRNA
targeting FADS2 (brown and orange) normalized to control (n=3). Cells were
treated with vehicle or 5 µM RSL3, the latter inhibiting glutathione
peroxidase 4 and inducing lipid peroxidation. Two-way ANOVA with
Sidak’s multiple comparisons. Cancer cell experiments were performed in low FBS DMEM (1%: HUH7;
0.5%: all other cancer cells) with treatment of 72 h. Data are presented as
mean ± SD from biological independent samples.
(a) Fatty acid concentrations of FBS (fetal bovine
serum; n=4). Low FBS condition (0.5-1% FBS) corresponds to a total fatty
acid concentration of 4.31-8.62 µM. (b) Sensitivity profile of cancer cells to Merck Frosst
Cpd 3j (white; HUH7, A549: 2 nM; H460, DU145: 1 nM; MDA-MB-468, T47D: 0.5
nM) in blood-like medium (BLM), normalized to
control. HUH7 n=3; A549 n=3; H460 n=6; DU145 n=6, MDA-MB-468 n=3; T47D n=9.
Two-way ANOVA with Dunnett’s multiple comparisons. (c, d) Sensitivity profile of HUH7 and
A549 control (non-targeting shRNA; black) and knockdown (shFADS2; brown and
orange) cells treated with DMSO (dark bars) or 2 nM Merck Frosst Cpd 3j
(light bars) in blood-like medium (BLM), normalized to
control (n=3). Two-way ANOVA with Holm-Sidak multiple comparisons. (e) Sensitivity profile of MDA-MB-468 control (black)
and FADS2 overexpression (green) cells DMSO (dark bars) or
0.5 nM Merck Frosst Cpd 3j (light bars) in blood-like medium (BLM), normalized to control (n=3). Two-way ANOVA with Holm-Sidak
multiple comparisons. (f) Desaturation activity to sapienate based on the
sapienate to palmitate ratio in cancer cells in conditions as described in
(b). n=3. Unpaired two-sided Student’s T-tests with Holm-Sidak
multiple comparisons. (g,h) Desaturation activity from palmitate to sapienate
based on the sapienate to palmitate ratio in the same conditions as
described in (c-f). n=3. One-way ANOVA with Dunnett’s multiple
comparisons (g); unpaired Student’s T-test (h). Cancer cell experiments were performed in low FBS BLM (1%: HUH7;
0.5%: all other cancer cells) with treatment of 72 h. Data are presented as
mean ± SD from biological independent samples.
(a) Separation and detection of sapienate
(cis-6-hexadecenoate) and palmitoleate (cis-9-hexadecenoate) via gas
chromatography mass spectrometry. Separation was optimized using a standard
mix containing pentadecanoate, sapienate, palmitoleate, palmitate,
cis-8-octadecenoate, oleate, vaccinate, linoleate and stearate (upper
panel), and the method was subsequently validated by measurement of these
fatty acids in biological samples. A representative biological sample is
presented in the lower panel. (b) Separation and detection of cis-8-octadecenoate,
oleate (cis-9-octadecenoate) and vaccinate (cis-11-octadecenoate) via gas
chromatography flame ionization detector. Separation was optimized using a
standard mix containing cis-8-, cis-9- and cis-11-octadecenoate (upper
panel), and the method was subsequently validated by measurement of these
fatty acids in biological samples. A representative biological sample is
presented in the lower panel.
(a) Sensitivity profile of cancer cells treated Merck Frosst Cpd 3j
normalized to control (HUH7 n=6; A459 n=3, H460 n=3, DU145 n=6, MDA-MB-468 n=6,
T47D n=6). Two-way ANOVA with Dunnett’s multiple comparisons. (b) Heat map representing fatty acid abundances normalized to
highest abundance of each fatty acid across all cell lines. Over 90% reduction:
white, no reduction: dark green. (c) Correlation between SCD independence defined as area under the
cell number curve of (a) and desaturation activity to sapienate (Extended Data Figure 2b). Trend line (dashed
line); 95% confidence intervals (dotted lines). n=3. (d-f) Sapienate to palmitate ratio of HUH7 subcutaneous xenografts
treated with vehicle (n=8) or Merck Frosst Cpd 3j (n=8; 1.5 mg per kg twice
daily with oral (p.o.) gavage) and in diethylnitrosamine (DEN) or
genetically-induced HCC (normal n=15; DEN n=4); myrAKT-N-Ras (n=10),
Pten (n=8) and Stk (normal n=7; HCC n=6).
Unpaired two-sided Student’s T-test with Welch’s correction. (g,h) Correlation between FADS2 protein expression and SCD
independence or desaturation activity to sapienate (Extended Data Figure 2b). Trend line (dashed line); 95%
confidence intervals (dotted lines). n=3. (i,j) FADS2 gene expression in paired samples of
human HCC (n=4) and non-small cell lung adenocarcinoma (n=10)
versus normal adjacent tissue. (k) Desaturation activity to sapienate in HUH7 and A549 cells with a
non-targeting shRNA or shRNAs targeting FADS2 (n=3). One-way
ANOVA with Dunnett’s multiple comparisons. (l) Sapienate to palmitate ratio in normal adjacent liver and HUH7
orthotopic liver tumors with non-targeting shRNA or shRNA targeting
FADS2 (n=5). Two-way ANOVA with Sidak’s multiple
comparisons. Experiments were performed in low FBS (1%: HUH7; 0.5%: other) with treatment of
72 h. Error bars represent SD (in vitro) or SEM (in
vivo) from mean of biological independent samples (in
vitro) or animals (in vivo).
(a,b) Relative proliferation of MDA-MB-468 control (with or without
sapienate) and FADS2 overexpression cells upon treatment 0.5 nM Merck Frosst Cpd
3j normalized to control (a: n=9; b: control n=10,
overexpression n=12). Two-way ANOVA with Tukey’s multiple
comparisons. (c,d) Relative proliferation of HUH7 and A549 cells
(with or without sapienate) upon FADS2 knockdown with(out) 2 nM
Merck Frosst Cpd 3j normalized to control (c: control n=9;
shFADS2-1 n=6; shFADS2-2 n=6; d: n=6). Two-way ANOVA with Tukey
multiple comparisons (within different cell lines); one-way ANOVA with
Dunnett’s multiple comparisons (across different cell lines). Only
pair-wise comparisons are depicted. (e,f) Representative images of hematoxylin and eosin
stain and relative area of resected tumor nodules derived from HUH7 control
(non-targeting shRNA) or FADS2 knockdown (shFADS2-2) orthotopic
liver xenografts in mice treated with vehicle or Merck Frosst Cpd 3j (1.5 mg per
kg twice daily per oral; p.o.; control+vehicle n=13; control+SCD inhibition
n=12; shFADS2-2+vehicle n=12, shFADS2-2+SCD inhibition n=14 of one experiment).
Masks in (e) show the tissue contour and indicating tumor (black), necrotic
(grey) and liver (white) area. Scale bar represents 1,000 µm in all
cases. Box blots in (f) show box extending from the 25th to
75th percentiles, whiskers indicating the minimum and maximum,
and a line indicating the mean. One-way ANOVA with Tukey’s multiple
comparisons. Experiments were performed in low FBS (1%: HUH7; 0.5%: others) with treatment of
72 h. Error bars represent SD from mean of biological independent samples
(in vitro) or animals (in vivo), unless
stated otherwise.
(a,b) Relative fatty acid abundances in HUH7 and A549
cells treated 2 nM Merck Frosst Cpd 3j (n=3). Data were normalized to the
respective controls. (c) Palmitoleate and sapienate abundances in membrane phospholipids
in HUH7 cells with non-targeting shRNA and an shRNA targeting
FADS2 (n=4). (d,e) Relative phospholipid-bound palmitoleate, sapienate and
cis-8-octadecenoate abundances in HUH7 and A549 cells treated with control or 2
nM Merck Frosst Cpd 3j normalized to control (n=2 in e
sapienate+SCD inhibitor; n=3 in all other cases,). BDL denotes ‘below
detection limit’. When the respective control was BDL, the data were
normalized to the total abundance of all fatty acids measured and are
represented in arbitrary units. Experiments were performed in low FBS (1% : HUH7; 0.5% : others) with treatment
of 72 h. Error bars represent SD from mean of biological independent samples.
Unpaired two-sided Student’s T-tests.
(a) Sapienate to palmitate and palmitoleate to palmitate ratios in
HCC and normal liver tissue as well as in blood plasma from humans. Blood plasma
was from healthy volunteers or liver cancer patients, while normal liver was
adjacent non-cancerous tissue from liver cancer patients and non-transplanted
donor livers (normal: blood plasma n=23, tissue n=16 and cancer: blood plasma
n=33, tissue n=16). Grey indicates blood plasma and black indicates tissue.
Notably, blood plasma ratios from healthy volunteers are the same as in Figure
4b. (b) Sapienate to palmitate and palmitoleate to palmitate ratios in
lung cancers and normal lung tissue as well as in blood plasma from humans.
Blood plasma was from healthy volunteers or cancer patients, while normal lung
was adjacent non-cancerous tissue from cancer patients (normal: blood plasma
n=23, tissue n=15 and cancer: blood plasma: n=34, tissue n=15). Grey indicates
blood plasma and black indicates tissue. Notably, blood plasma ratios from
healthy volunteers are the same as in Figure 4a. (c) Sapienate metabolism is an alternative monodesaturation
pathway. Two-Way ANOVA with Tukey multiple comparisons. Error bars represent SEM of mean
from different individuals.
Comment in
-
Tumours use a metabolic twist to make lipids.Nature. 2019 Feb;566(7744):333-334. doi: 10.1038/d41586-019-00352-1. Nature. 2019. PMID: 30783267 No abstract available.
-
Uncovering Plasticity in Tumor Lipid Metabolism.Cancer Discov. 2019 Apr;9(4):456. doi: 10.1158/2159-8290.CD-NB2019-027. Epub 2019 Feb 27. Cancer Discov. 2019. PMID: 30814093
References
Publication types
MeSH terms
Substances
Grants and funding
LinkOut - more resources
Full Text Sources
Molecular Biology Databases
