Abstract

   Cancer cells in acidic tumor regions are aggressive and a key
   therapeutic target, but distinguishing between acid-exposed and hypoxic
   cells is challenging. Here, we use carbonic anhydrase 9 (CA9)
   antibodies to mark acidic areas in both hypoxic and respiring tumor
   areas, along with an HRE-GFP reporter for hypoxia, to isolate distinct
   cell populations from 3D tumor spheroids. Transcriptomic analysis of
   CA9-positive, hypoxia-negative cells highlights enriched fatty acid
   desaturase activity. Inhibiting or silencing stearoyl-CoA desaturase-1
   (SCD1) induces ferroptosis in CA9-positive acidic cancer cells and
   delays mouse tumor growth, an effect enhanced by omega-3 fatty acid
   supplementation. Using acid-exposed cancer cells and patient-derived
   tumor organoids, we show that SCD1 inhibition increases acidic cancer
   cell reliance on external mono-unsaturated fatty acids, depriving
   hypoxic cells of essential resources. This bystander effect provides
   unbiased evidence for a lack of full overlap between hypoxic and acidic
   tumor compartments, highlighting a rationale for targeting desaturase
   activity in cancer.

   Subject terms: Cancer microenvironment, Cancer metabolism
     __________________________________________________________________

   The interplay between hypoxic and acidic conditions in the tumor
   microenvironment is unclear. Here, the authors show that hypoxic and
   acidic tumor cells do not fully overlap and display a symbiotic
   interaction, relying on unsaturated fatty acids, which can be
   therapeutically targeted by inhibiting stearoylCoA desaturase 1 (SCD1).

Introduction

   Tumor acidosis primarily results from a compromised capillary perfusion
   incapable of carrying away H^+ released (together with lactate) by
   hypoxic, glycolytic cancer cells^[60]1,[61]2. While this statement
   might suggest a necessary association between acidosis and hypoxia, the
   interplay between these two cancer hallmarks is more complex. First,
   the preferred glucose fermentation into lactate together with two H^+
   equivalents can occur in non-hypoxic tumor areas, a phenomenon
   initially described by Otto Warburg^[62]3. Second, glycolysis-impaired
   or lactate dehydrogenase-deficient cancer cells retain the ability to
   acidify the extracellular environment in vivo^[63]4–[64]6. Third, a
   high respiratory rate of cancer cells may directly contribute to
   acidosis upon CO[2] generation and hydration in the extracellular
   medium (CO[2] + H[2]O → HCO[3]^− + H^+)^[65]7. Acidosis may thus
   develop in tumor areas where O[2] is available but with a limited
   capacity to rapidly get rid of extracellular H^+. This claim is further
   supported by various sets of experimental data documenting a
   non-overlapping relationship between extracellular pH (pH[e]) and
   pO[2]^[66]8,[67]9. A model may thus be proposed where, besides hypoxic
   (acidic) cancer cells located at a distance from blood vessels, and
   well-perfused cancer cells bathing in a well-buffered extracellular
   medium, a third intermediary tumor compartment is made of non-hypoxic,
   acidic cancer cells^[68]1; non-hypoxic being understood in its large
   acceptance, i.e. moderately oxygenated.

   We previously reported a shift from glycolysis to fatty acid metabolism
   in cancer cells of distinct origins chronically exposed to an acidic
   pH[e] (vs. corresponding cancer cells maintained at physiological
   pH[e])^[69]10. Our lab further documented that acidic cancer cells
   could be concomitantly engaged in FA synthesis and oxidation^[70]11.
   More recently, a genome-wide CRISPR-Cas9 approach aiming to identify
   critical survival pathways for acidic cancer cells strengthened these
   findings with the identification of genes the expression of which
   supports mitochondrial respiration^[71]12. These data are also in line
   with the well-known inhibition of glycolytic enzymes in response to
   pH[i] reduction and fit the recently reported enhanced PPAR pathway in
   acid-resistant cancer cells^[72]13. The search for such distinct
   pathways to therapeutically target acidic, non-hypoxic cancer cells
   gains particular relevance when contemplating the phenotype associated
   with this tumor compartment. Indeed, while acidosis is often associated
   with a reduction in cell proliferation^[73]14, acidic cancer cells
   undergo an epithelial-to-mesenchymal phenotype under the control of
   different actors including TGFβ2^[74]15 and RhoA^[75]16 signaling,
   making them more prone to invasiveness and metastatic spreading. Rohani
   and colleagues also documented that tumor acidosis areas (not
   restricted to hypoxic regions) overlie highly proliferative and
   invasive zones at the tumor–stroma interface^[76]17. Targeting this
   tumor's acidic compartment thus represents an attractive strategy to
   kill cancer cells among the most aggressive tumors.

   In this study, we, therefore, aimed to follow an unbiased strategy to
   isolate acidic, non-hypoxic cancer cells and profile them to formally
   prove the existence of this tumor compartment and identify potential
   vulnerabilities to target it. The challenge was to discriminate this
   cell population from the hypoxic cancer cells. To dissociate acidic
   from non-acidic cancer cells regardless of the hypoxic environment, we
   considered the double role of carbonic anhydrase CA9 with its
   extracellular-facing catalytic sites. Indeed, in a hypoxic area where
   reduction in intracellular pH (pH[i]) requires dynamic export of H^+
   (together with lactate) in the extracellular medium, CA9 facilitates
   consecutive extracellular H^+ diffusion upon enzymatic CO[2] formation
   (accelerating the otherwise slow spontaneous reaction kinetics:
   H^+ + HCO[3]^− → H[2]O + CO[2]). Interestingly, Swietach and colleagues
   documented that in the presence of environmental CO[2], CA9 activity
   could instead reduce pH[e] by converting respiration-derived CO[2] to
   extracellular H^+ ^[77]18,[78]19. These data aligned with the
   documented lack of correlation between CA9 expression and various
   hypoxia markers^[79]20–[80]22. CA9 expression has thus the potential to
   act as a marker of acidosis not only in deeply hypoxic tumor regions
   but also in tumor areas where respiration supports cancer cell
   oxidative metabolism (and the associated reduction in pO[2] reaching a
   threshold compatible with CA9 expression). Combination with bona fide
   hypoxia markers could then offer discriminating tools to further
   isolate acidic cancer cells engaged in oxidative metabolism from
   hypoxic, glycolytic cancer cells. To maximize the tractability of this
   approach, we used 3D tumor spheroids where all the cells located at the
   same distance of the center share the same phenotype, making the
   different compartments mentioned above more defined. FACSorting of
   CA9-positive but deep hypoxia-negative cancer cells led us to identify
   desaturase SCD1 as a druggable vulnerability of acidic cancer cells, as
   further validated in mouse tumor models and patient-derived tumor
   organoids. This work also uncovered a symbiotic interplay between
   hypoxic and acidic cancer cells, the former being addicted to exogenous
   mono-unsaturated FA spared by SCD1-expressing cancer cells under
   acidosis, further reinforcing the rationale of targeting desaturase
   activity in cancer.

Results

CA9 expression does not strictly overlap with hypoxia markers

   We first compared the extent of CA9-positive staining with the hypoxic
   area within 3D tumor spheroids. To detect hypoxia, we used either
   pimonidazole staining of spheroid slices (Fig. [81]1a) or the detection
   of GFP under the regulation of hypoxia-responsive elements (HREs) in
   sections of cleared spheroids using light sheet microscopy
   (Fig. [82]1b). The CA9-positive area was significantly larger than the
   hypoxic zone in 3D tumor spheroids generated from head and neck FaDu
   and colorectal HCT116 cancer cells (see bar graphs in Fig. [83]1a, b),
   indicating the presence of a CA9-positive (acidic) compartment that is
   not hypoxic or more exactly with a pO[2] above the threshold of
   pimonidazole and HRE-GFP detection; similar results were obtained with
   colorectal HT-29 spheroids (Fig. [84]S1a). 3D imaging reconstruction of
   the whole spheroids (imaged by light sheet microscopy) confirmed that
   hypoxic and non- (or less) hypoxic CA9-positive compartments could be
   discriminated (Fig. [85]1c, see also Supplementary Movie [86]1). Mouse
   tumors generated from FaDu and HCT116 cancer cells also revealed that
   some CA9-positive areas did not overlap with pimonidazole-positive
   (hypoxic) areas (Fig. [87]1d); a differential CA9 and pimonidazole
   staining was also observed in HT-29 colorectal tumors (Fig. [88]S1b).
   Finally, we examined by flow cytometry the level of CA9 expression in
   cancer cells adapted to pH 6.5 and corresponding cancer cells
   maintained at pH 7.4 (named 6.5/ and 7.4/cancer cells here below,
   respectively), both cell populations being cultured under oxygenated
   conditions. We found a net increase in CA9 expression in acid-exposed
   cancer cells (Figs. [89]1e and [90]S1c, d), further validating CA9 as a
   marker to help track the acidic, oxidative tumor compartment.

Fig. 1. CA9 expression does not overlap with hypoxia.

   [91]Fig. 1
   [92]Open in a new tab

   a Representative pictures and quantification of Carbonic Anhydrase 9
   (CA9, red) and pimonidazole (green) immunostaining in FaDu and HCT116
   spheroid sections (N = 3, n = 2); DAPI (blue) nuclear staining was used
   to normalize measurements. The white delimitation represents the rim of
   the spheroid. b Representative pictures and quantification of CA9 (red)
   and Hypoxia-Responsive Element-dependent GFP reporter (HRE, green)
   wholemount fluorescence in FaDu and HCT116 spheroids (N = 5 and 7,
   respectively). c 3-dimensional modeling of CA9 (red), HRE-GFP (green),
   and DAPI (blue) staining in Fadu spheroids; this experiment was
   repeated twice with similar results. d Representative CA9 (purple) and
   pimonidazole (yellow) immunostaining in FaDu and HCT116 tumor sections;
   this experiment was repeated twice with similar results. e
   Representative flow cytometry analysis of CA9 staining in FaDu and
   HCT116 cancer cells maintained under normoxia at physiological pH 7.4
   or at acidic pH 6.5; this experiment was repeated twice with similar
   results. The scale bar represents 100 µm for 3D tumor spheroids (a–c)
   and 200 µm for mouse tumor sections (d). Data are plotted as the
   means ± SD (**P = 0.0017, ****P < 0.0001); N indicates the number of
   independent experiments and n indicates the number of biological
   replicates (when >1). Significance was determined by two-sided
   Student’s t-test (a, b). Source data are provided as a Source Data
   file.

CA9-positive HRE-negative cancer cells exhibit enrichment in FA desaturase
activity

   Next, we aimed to characterize the transcriptome of acidic, non-hypoxic
   cancer cells (as defined above) sorted from 3D spheroids using a
   combination of CA9 and HRE-GFP labeling (Fig. [93]2a). Four distinct
   cell populations isolated by FACS (Figs. [94]2b and [95]S2a) were
   submitted to RNA-seq analysis. The principal component analysis
   confirmed significant differences between the 4 quadrants
   (Fig. [96]2c). As a validation, Gene Set Enrichment Analysis (GSEA)
   revealed that hypoxia and glycolysis pathways were enriched in double
   positive CA9+/HRE+ cell population, i.e. the bona fide hypoxic,
   glycolytic cell compartment (Fig. [97]S2b). When compared with double
   positive CA9+/HRE+ cell population, CA9+/HRE- (i.e., acidic,
   non-hypoxic) cancer cells exhibited 1303 DE genes (FDR < 0.05)
   (Fig. [98]2d). KEGG analysis identified ferroptosis among the pathways
   enriched in CA9+/HRE− cells (Fig. [99]2e) further validating our
   sorting strategy. We have indeed documented that acidic cancer cells
   exhibit an increased susceptibility to lipid peroxidation and
   associated ferroptosis^[100]23. Interestingly, the top enriched
   pathway, i.e., the biosynthesis of mono-unsaturated fatty acid (MUFA)
   (Fig. [101]2e), could also be related to ferroptosis since MUFA are
   known to protect cells from ferroptosis^[102]24,[103]25. In this MUFA
   synthesis-related pathway, we found, among the upregulated genes,
   desaturases SCD1 and FADS1, two enzymes that catalyze rate-limiting
   steps of fatty acid unsaturation through the introduction of double
   bonds at the delta-9 and -5 positions, respectively. Enrichment in SCD1
   and FADS1 desaturases was confirmed by RT-qPCR performed on isolated
   CA9+/HRE− cells isolated from 3D FaDu spheroids (Fig. [104]2f). In
   6.5/FaDu cancer cells, we similarly found a significant increase in
   transcript abundance of SCD1 and FADS1 (Fig. [105]2g); the expression
   of FADS2 and SCD5 was also increased in 6.5/FaDu cancer cells
   (Fig. [106]2g). A similar increase in desaturase expression was found
   in 6.5/HCT116 and 6.5/Siha cancer cells (Fig. [107]S2c).

Fig. 2. CA9-positive HRE-GFP-negative cancer cells exhibit enrichment in FA
desaturase activity.

   [108]Fig. 2
   [109]Open in a new tab

   a, b Schematic protocol of cell sorting from 3D FaDu spheroids based on
   CA9 immunostaining and HRE-GFP expression, created in BioRender. Feron,
   O. (2024) BioRender.com/i40c288 (a) and representative FACS plots
   showing gating strategy for each of the four quadrants (b). c Principal
   component analysis (PCA) discriminating the four quadrants based on
   RNA-seq analysis performed on 3 independent sorting experiments. d
   Volcano plot of differentially expressed genes between CA9+/HRE− and
   CA9+/HRE+ FaDu cell populations. e. KEGG pathway enrichment analysis of
   the differentially expressed genes between CA9+/HRE− and CA9+/HRE+ FaDu
   cell populations. f, g Changes in the mRNA expression of the indicated
   desaturases in the four distinct FaDu spheroid compartments (f) (N = 3,
   n = 2), and in FaDu cancer cells maintained at physiological pH 7.4 or
   at acidic pH 6.5 (g) (N = 4); results are expressed as fold-change vs.
   mRNA levels in CA9-/HRE- double negative cell populations and in cancer
   cells at pH 7.4, respectively. Data are plotted as the means ± SD
   (P-values as indicated or ***P < 0.001, ****P < 0.0001); N indicates
   the number of independent experiments and n indicates the number of
   biological replicates (when >1). Significance was determined by a
   two-sided Student’s t-test with FDR adjustment (d), one-way ANOVA with
   Tukey’s multiple comparison test (f), or two-sided Student’s t-test
   (g). Source data are provided as a Source Data file.

SCD1 genetic silencing and pharmacological inhibition reduce the survival of
CA9-positive cancer cells

   Since SCD1 was the most consistently upregulated desaturase in
   different acid-exposed cancer cells, we next tracked the effects of a
   reduction in SCD1 activity in 3D spheroid models wherein acidic and
   non-acidic cancer cells co-exist. We first examined the effect of SCD1
   silencing using the CRISPR-Cas9 system. Three distinct clones, derived
   from different sgRNA combinations, were selected based on their ability
   to induce varying levels of gene knockout (KO). (Fig. [110]S3a). We
   found that SCD1 silencing led to significant growth inhibitory effects
   as determined based on a reduction in FaDu spheroid size and/or
   appearance of a halo of dead cells (Fig. [111]3a, b). Strikingly, when
   looking at an early time (i.e., before significant differences in
   spheroid size), CA9 expression was reduced in proportion to the
   reduction in SCD1 expression (Figs. [112]3c–e and [113]S3b). Similar
   results were obtained using spheroids generated from FaDu cancer cells
   transduced with a plasmid encoding for SCD1 shRNA (Fig. [114]S3c–e).

Fig. 3. SCD1 genetic silencing and pharmacological inhibition reduce the
survival of CA9-positive cancer cells.

   [115]Fig. 3
   [116]Open in a new tab

   a, b Representative contrast phase pictures (a) and quantification (b)
   of the effects of CRISPR-Cas9-based SCD1 gene invalidation on FaDu
   spheroid growth at day 7 post-formation (vs. control sgRNA); N = 3,
   n = 2. c–e Quantification of SCD1 (violet) (c) (N = 5) and CA9 (red)
   immunostaining (d) (N = 5), and representative pictures (e) of
   spheroids made of CRISPR-Cas9-based SCD1-silenced FaDu cells at day 10
   post-formation. f, g Spheroid growth (f) (N = 3, n = 3–4) and
   cytotoxicity (Incucyte Cytotox Green reagent) follow up (g) (N = 3,
   n = 4) after exposure to SCD1 inhibitor (or vehicle) for 72 h in FaDu
   spheroids. h, i Representative pictures (h) and quantification of CA9
   (red) (i) in sections of FaDu spheroids collected after 72 h exposure
   to SCD1 inhibitor (N = 3, n = 2). j Flow cytometry analysis of CA9
   labeling from FaDu cells isolated from spheroids after 72 h exposure to
   SCD1 inhibitor (or vehicle) (N = 3). k Quantification of pimonidazole
   (green) in sections of FaDu spheroids collected after 72 h exposure to
   SCD1 inhibitor (N = 3, n = 2). l, m Mitochondrial oxygen consumption
   rate (OCR) of l FaDu (N = 3, n = 7) and HCT116 (N = 3, n = 7–8)
   spheroids after 72 h SCD1 inhibition and (m) FaDu cancer cells
   transduced with the indicated SCD1 sgRNA (N = 3, n = 3) or control
   sgRNA (N = 3, n = 3–4). n Non-mitochondrial OCR of FaDu cells
   transduced with an SCD1-expressing vector or control plasmid and
   exposed to palmitate (N = 3, n = 3) or vehicle (FA-free BSA) (N = 3,
   n = 4). o–r Representative pictures (o) and quantification of SCD1
   (violet) (p), CA9 (red) (q), and pimonidazole (green) (r) staining in
   FaDu spheroids exposed to PPAR-γ inhibitor (or vehicle) for 72 h
   (N = 3); the effects of a PPAR-α inhibitor are also shown in graphs
   (p–r). All treatments (f–l) with SCD1 inhibitor (A939572, 32 µM) were
   initiated at day 7 after spheroid formation (i.e., timing 0 on graphs).
   All immunostaining quantifications (c, d, i, k, p–r) were normalized to
   the DAPI nuclear staining area. Data are plotted as the means ± SD
   (P-values as indicated or ***P < 0.001, ****P < 0.0001); N indicates
   the number of independent experiments and n indicates the number of
   biological replicates (when >1). Significance was determined by one-way
   ANOVA with Tukey’s multiple comparison tests (b–d), Dunnet’s multiple
   comparison tests (p–r), Sidaks’s multiple comparison tests (f, g),
   two-way ANOVA with Tukey’s multiple comparison tests (l–n) or two-sided
   Student’s t-tests (i–k). Source data are provided as a Source Data
   file.

   We also observed a net growth inhibitory effect on established
   spheroids upon exposure to SCD1 inhibitor (SCD1i) A939572, as
   determined based on reduction in FaDu spheroid size (Fig. [117]3f) and
   increased cytotoxicity as determined using IncuCyte detection of
   Cytotox Green Reagent accumulation (Fig. [118]3g). Similar results were
   obtained with colorectal HCT116 and HT-29 spheroids (Fig. [119]S3f, g).
   Also, propidium iodide labeling confirmed that cell death was more
   prominent inside the spheroid wherein CA9 was expressed before exposure
   to SDCi (Fig. [120]S3h). As observed with SCD1-silenced cancer cells,
   we found that as soon as 72 h after SCD1i exposure (i.e., before
   significant alterations in spheroid size), a dramatic reduction in the
   pool of CA9-positive cells could be detected by immunofluorescence
   (Figs. [121]3h, i and [122]S3i) or flow cytometry from treated 3D
   spheroids (Fig. [123]3j). Interestingly, this reduction in SCD1
   expression was associated with a decrease in the abundance of lipid
   droplets (LD) as evidenced using PLIN2 staining (Fig. [124]S3j).

   While SCD1i-induced cell death could account for the loss of CA9 signal
   from killed acidic cancer cells, we also examined whether the
   reoxygenation process occurring after SCD1i exposure could also
   contribute to the reduction of CA9 expression. Pimonidazole staining
   confirmed the reduction in the hypoxic core of FaDu and HCT116
   spheroids (see green staining in Figs. [125]3h and [126]S3i,
   quantification in Figs. [127]3k and [128]S3i) in response to SCD1i
   treatment. Seahorse experiments further validated that SCD1i exposure
   reduces OCR in 3D spheroids (Fig. [129]3l) and in acid-adapted cancer
   cells (Fig. [130]S3k). We found a similar reduction in OCR using
   SCD1-silenced cancer cells (Figs. [131]3m and [132]S3l), also supported
   by a reduction in pimonidazole staining in spheroids generated from
   these cells (Fig. [133]S3d, e). Conversely, supplementation of
   spheroids with stearate as a source of SFA led to an increased extent
   of CA9 and pimonidazole staining (Fig. [134]S3m). We also showed that
   the O[2]-dependent enzymatic activity of SCD1 could contribute to OCR
   and thus lead to spheroid reoxygenation upon SCD1 inhibition. For this
   purpose, we used SCD1-overexpressing cancer cells (Fig. [135]S3n) and
   showed that non-mitochondrial OCR was significantly increased in
   response to palmitate as an exogenous source of SFA (Fig. [136]3n).
   Altogether these data indicate that the disappearance of CA9 signal in
   spheroids upon SCD1 blockade or silencing results from the preferential
   killing of cancer cells residing in the acidic compartment, along with
   a global reoxygenation process.

   The preferred uptake of fatty acids that we previously reported to
   occur in acid-adapted cancer cells^[137]11,[138]15,[139]23 led us to
   examine the role of the PPAR transcription factor. Interestingly,
   PPAR-γ inhibitor (but not PPAR-α inhibitor) led to SCD1 downregulation
   and recapitulated the phenotype observed upon alteration in SCD1
   expression or activity, namely the extinction of CA9 and pimonidazole
   staining (Figs. [140]3o–r and [141]S3o).

The obligatory SCD1 expression for the survival of acid cancer cells is void
upon DHA exposure

   While the above data identified the requirement to desaturate FA as a
   vulnerability associated with the enhanced FA metabolism in
   acid-exposed cancer cells, we then reasoned that supplementation with
   PUFA could further increase the stress level in these cells. This
   strategy is further supported by the observed SCD1i-induced reduction
   in LD (Fig. [142]S3j) while PUFA excess is known to stimulate LD
   formation^[143]23. We found that n-3 PUFA docosahexaenoic acid (DHA)
   further enhanced the preferential cytotoxic effects of SCD1 inhibitor
   (SCD1i) A939572 in 3 out of 4 6.5/cancer cells (vs. corresponding
   7.4/cancer cells) (Figs. [144]4a and [145]S4a). In 6.5/Fadu cancer
   cells, DHA-enhancing effects were revealed upon the combination of
   SCD1i with an inhibitor of FADS1 (FADS1i) (but not FADS2i)
   (Figs. [146]4a and [147]S4a). In parallel, we found that DHA
   additionally heightened the elevated levels of triacylglycerol (TAG)
   induced by acidic conditions (Fig. [148]4b). However, while the
   abundance of the most prominent MUFA (i.e., C18:1 oleate) was directly
   proportional to the TAG increase in acid-exposed cancer cells (compare
   black and white bars in Fig. [149]4b, c), DHA exposure did not further
   increase the oleate contribution in TAG (compare black and red bars in
   Fig. [150]4c). The increase in total TAG levels in 6.5/cancer cells
   mostly resulted from the incorporation of C22:6 DHA (Fig. [151]4c). A
   more detailed analysis of the composition of TAG revealed that DHA
   emerges as the predominant constituent of TAG, with 45% of them
   incorporating at least one DHA, while the TAG content in C18:1 oleate
   decreased from 46% to 22% in acid-exposed cancer cells (Fig. [152]S4b).
   In the same experimental conditions, Bodipy staining revealed that the
   presence of lipid droplets was increased in 6.5/cancer cells and
   further enhanced in the presence of DHA (Fig. [153]S4c, d). More
   importantly, while acidic pH and DHA exposure independently increased
   the amounts of TAG to a similar extent (see black and orange bars in
   Fig. [154]4b), the PUFA/MUFA ratio in the neutral lipid (NL) fraction
   was increased by 5-fold in the DHA condition (see black and orange bars
   in Fig. [155]4d). The PUFA/MUFA ratio eventually peaked in the NL but
   also phospholipid (PL) fractions of acid-exposed cancer cells when
   subjected to DHA treatment (see red bars in Fig. [156]4d, e).

Fig. 4. The obligatory SCD1 expression for the survival of acid cancer cells
is void upon PUFA exposure.

   [157]Fig. 4
   [158]Open in a new tab

   a Effects of inhibitors of SCD1 (A939572), FADS2 (SC-26196), and FADS1
   (T3364366) on the viability of 7.4/ and 6.5/cancer cells (48 h, 32 µM
   for single-inhibitor treatment or 16 µM of each inhibitor for
   combination) in the presence of 50 µM DHA or vehicle (FA-free BSA)
   (N = 3, n = 2). b–e Lipidomic analysis of 6.5/ and 7.4/HCT116 cancer
   cells after 24 h exposure to 25 µM DHA or vehicle (N = 2). b
   Triacylglycerol (TAG) content expressed as % of the value measured in
   7.4/cells, c TAG composition in esterified FA, d, e PUFA/MUFA ratio in
   neutral lipid (d) and phospholipid (e) fractions. f, g Representative
   pictures (f) and quantification (g) of SCD1 (green) immunofluorescence
   signal in the indicated 7.4/ and 6.5/cancer cells exposed or not to
   20 µM DHA for 72 h (N = 4). h SCD1 immunoblotting from the indicated
   6.5/ and 7.4/cancer cells exposed or not to 20 µM DHA for 72 h (N = 1).
   Data are plotted as the means ± SD (ns: non-significant, P-values as
   indicated or ***P < 0.001, ****P < 0.0001); N indicates the number of
   independent experiments and n indicates the number of biological
   experiments (when >1). Significance was determined by one-way ANOVA
   with Tukey’s multiple comparison test (b–e, g) or two-way ANOVA with
   Tukey’s multiple comparison test (a). Source data are provided as a
   Source Data file.

   Collectively, the aforementioned data suggested that despite SCD1
   upregulation in response to an acidic environment, there is not enough
   MUFA to counterbalance the increased uptake of PUFA and the associated
   risk of membrane lipid peroxidation^[159]23. We next examined the
   reason for this paradox and found that remarkably, although SCD1 was
   consistently increased in acidic 6.5/cancer cells vs. 7.4/cancer cells
   (Fig. [160]4f), DHA reduced the expression of SCD1 in control 7.4/cells
   and also limited its upregulation in response to acidosis
   (Fig. [161]4f, g). Upregulation of SCD1 under acidosis as well as the
   blunting effect of DHA on desaturase expression were both confirmed by
   Western blotting (Fig. [162]4h). Collectively, these data confirm that
   cancer cells are particularly susceptible to SCD1 inhibition under
   conditions of both acidosis and DHA exposure, as acidosis stimulates
   desaturase expression while DHA suppresses it.

SCD silencing or inhibition exerts tumor growth inhibitory effects in mice,
an effect further exacerbated by lipid metabolism rewiring in response to a
PUFA-rich diet

   To validate in vivo the above findings, we first compared tumor growth
   using SCD1-silenced clones and confirmed a significant tumor growth
   delay (Fig. [163]5a) and prolonged mouse survival (Fig. [164]5b). Since
   genetic silencing of SCD1 reduced the growth of tumors from early time,
   we next evaluated pharmacological inhibition of SCD1 in established
   tumors in mice fed a control diet or an isocaloric omega-3 PUFA-rich
   diet. The combination of the PUFA-rich diet and SCD1 inhibition
   significantly enhanced the extent of tumor growth inhibition
   (Fig. [165]5c) and prolonged mouse survival (Fig. [166]5d) whereas
   SCD1i treatment in mice fed the control diet did not reveal significant
   differences (Fig. [167]5c, d). Of note, in the absence of SCD1i
   treatment, tumor growth was slightly slowed down in mice fed PUFA-rich
   diet (vs. control diet) (Fig. [168]5c, d).

Fig. 5. SCD silencing or inhibition exerts tumor growth inhibitory effects in
mice, an effect further exacerbated by lipid metabolism rewiring in response
to a PUFA-rich diet.

   [169]Fig. 5
   [170]Open in a new tab

   a, b Tumor growth (a) and survival (b) of mice injected with SCD1 knock
   out FaDu cancer cells (top), clone #C1 (n = 5 mice per group,
   *P = 0.0110, ***P < 0.001) and bottom, clone #C2 (n = 6 mice per group,
   *P = 0.0151, **P = 0.0022). c, d Effects of daily SCD1 inhibitor
   administration (A939572, 30 mg/kg) on tumor growth (*P = 0.0352) (c)
   and survival (**P = 0.0033) (d) of mice injected with FaDu cancer cells
   and fed with control (n = 5 mice per group) or PUFA-rich diet (n = 6
   mice per group). e–g Total FA (e), PUFA (f), and MUFA (g) in the
   neutral lipid fraction of tumors from mice treated with SCD1 inhibitor
   and fed with control (n = 4) or PUFA-rich diet (n = 5). h–j n-3
   PUFA/MUFA ratios in NL (h) (n = 5), PL (i) (n = 6), and FFA (j) (n = 6)
   fractions of tumors from mice treated with SCD1 inhibitor and fed with
   control or PUFA-rich diet. Data plotted as the means ± SEM (P-values as
   indicated or ***P < 0.001, ****P < 0.0001). Significance was determined
   by one-way ANOVA with Tukey’s multiple comparison tests (h–j), two-way
   ANOVA (a) with Tukey’s multiple comparison tests (c), or Fisher’s Least
   Significative difference (e–g) and Long-rank Mantel–Cox test (b, d).
   Source data are provided as a Source Data file.

   To get more insights into the influence of the PUFA-rich diet on the
   extent of SCD1 inhibition, we next examined the profiles of neutral
   lipids (NL) in mouse tumors (vs. mice fed the control diet). In the
   absence of treatment, the amounts of FA in the neutral lipid (NL)
   fraction were increased in mice on the PUFA-rich diet, regardless of
   their nature (i.e., SFA, MUFA and PUFA) (Fig. [171]5e). Strikingly,
   while SCD1 inhibition led to a reduction in NL content in mice fed the
   PUFA-rich diet, a significant accumulation of NL was observed in mice
   fed the control diet (Fig. [172]5e). Moreover, while the PUFA-rich diet
   globally increased the amount of PUFA in NL regardless of SCD1i
   treatment (Fig. [173]5f), changes in PUFA concentrations marginally
   influenced the total amounts of NL (see Fig. [174]5e). The opposed
   effects of SCD1 inhibition on the NL profile could actually be
   attributed to alterations in MUFA (Fig. [175]5g, compare bars 1 vs. 2,
   and 3 vs. 4). SCD1 inhibition led to a 3-fold reduction in the amount
   of MUFA in the NL fraction of tumors from mice fed the PUFA-rich diet
   whereas a 30% increase in MUFA was observed in the same NL fraction of
   tumors fed the control diet (Fig. [176]5g). A significant increase in
   SFA in the NL fraction was also observed in tumors of SCD1i-treated
   mice fed the control diet (Fig. [177]S5a). The differential
   contribution of FA to NL accounted for a net increase in the n-3
   PUFA/MUFA ratio in mice fed the PUFA-rich diet, an effect further
   exacerbated in the presence of SCDi (Fig. [178]5h). This n-3 PUFA/MUFA
   ratio was not significantly altered in tumors of SCDi-treated mice
   under control diet conditions (Fig. [179]5h). Remarkably, this
   difference was also very pronounced in the phospholipid (PL) and free
   FA (FFA) fractions (Fig. [180]5i-j). In these fractions, the n-3
   PUFA/MUFA ratios were more largely increased in the presence of SCD1i
   reaching values up to 2-fold higher than in tumors of untreated
   PUFA-fed mice (Fig. [181]5i, j). Overall, these data suggest that the
   increased uptake of fatty acids (including MUFA) enables cancer cells
   to survive when MUFA synthesis is inhibited and that this rescue
   process is overwhelmed in the presence of PUFA (along with further
   reinforcement of SCD1 downregulation).

SCD inhibition is partly compensated by exogenous MUFA but exerts bystander
cytotoxic effects on hypoxic cancer cells

   The absence of MUFA reduction in the NL fraction of tumors of mice fed
   the control diet and treated with SCD1i led us to examine the possible
   compensatory contribution of exogenous MUFA and whether it could partly
   mitigate the therapeutic effects of SCD1 inhibition in mice fed the
   DHA-rich diet. To explore this hypothesis, we first used
   patient-derived tumor organoids (PDTO) to evaluate (in a more tractable
   model than mouse tumors) the capacity of oleate supplementation to
   reduce DHA-induced cytotoxicity. We found that DHA-exacerbated
   cytotoxic effects of SCD1 inhibition were indeed reversed by oleate
   (Fig. [182]6a, b). Similarly, using pH-adapted cancer cells, we found
   that oleate abrogated cytotoxic effects of the combo SCD1i and DHA at
   pH 7.4 (Fig. [183]S6a) and partially reversed them at pH 6.5
   (Fig. [184]6c). We further proved that the combination of SCD1i and DHA
   led to enhanced ferroptosis in acid-adapted FaDu and HCT116 cancer
   cells as revealed using Bodipy-C11 as a reporter of peroxidation
   (Figs. [185]6d and Fig. [186]S6b for quantification). Interestingly, in
   the same set-up, the addition of exogenous oleate partially prevented
   ferroptosis, comparable to the effects of the ferroptosis inhibitor
   α-tocopherol (Figs. [187]6d and [188]S6b).

Fig. 6. SCD inhibition is partly compensated by exogenous MUFA but exerts
bystander cytotoxic effects on hypoxic cancer cells.

   [189]Fig. 6
   [190]Open in a new tab

   a, b Representative pictures (a) and quantification (b) of colorectal
   cancer patient-derived organoids after treatment with SCD1 inhibitor
   (A939572, 20 µM) in the presence of OA (100 µM) and/or DHA (50 µM) for
   96 h (N = 2, n = 2). c, d Effect of SCD1 inhibitor (12 µM) on the
   viability (96 h) (c) and lipid peroxidation BODIPY-C11 staining (72 h)
   (d) of 6.5/Fadu (N = 3, n = 2) and 6.5/HCT116 (N = 3, n = 4) cancer
   cells in the presence of OA (100 µM) and/or DHA (50 µM) and/or
   α-Tocopherol (10 µM) or vehicle(s). e Representative pictures of CA9
   immunofluorescence signal (red) in equatorial sections of FaDu and
   HCT116 spheroids exposed for 72 h to SCD1 inhibitor (A939572, 32 µM) in
   the presence of OA (100 µM) or not; this experiment was repeated twice
   with similar results. f Representative flow chart depicting the effects
   of SCD1 inhibitor on the uptake of TopFluor oleate by 6.5/Fadu and
   6.5/HCT116 cancer cells maintained under normoxia or hypoxia (1% O[2])
   (N = 4, n = 4). g Schematic representation of the symbiotic
   relationship between hypoxic cancer cells (unable to synthesize MUFA
   and thus dependent on exogenous MUFA) and acidic, non-hypoxic cancer
   cells that may use both MUFA sources. Inhibition of SCD1 however forces
   the latter cell compartment to capture exogenous MUFA thereby depriving
   hypoxic cells from a vital source of MUFA. h Viability of FaDu and
   HCT116 cancer cells maintained under hypoxia (1% O[2]) and exposed for
   48 h to the conditioned medium (CM) from normoxic 6.5/FaDu and
   6.5/HCT116 exposed or not to SCDi (24 h, 15 and 25 µM A939572,
   respectively) (N = 3, n = 7). In some experiments, CM was supplemented
   by oleate (50 µM) (N = 3, n = 4). Control conditions consist of CM +
   fresh addition of SCDi (N = 3, n = 7), and non-conditioned medium (NCM)
   (N = 3, n = 3). i, j MUFA amounts (i) and SFA/MUFA ratio (j) determined
   in the conditioned medium (CM) of normoxic 6.5/FaDu and 6.5/HCT116
   cancer cells exposed or not to SCDi as above (N = 2). Data are plotted
   as the means ± SD (ns: non-significant, P-values as indicated or
   ***P < 0.001, ****P < 0.0001); N indicates the number of independent
   experiments and n indicates the number of biological replicates (when
   >1). Significance was determined by two-way ANOVA with Tukey’s multiple
   comparison test (b–h). Source data are provided as a Source Data file.

   In 3D spheroids, in the absence of an exogenous source of PUFA, we also
   found that the reduction in the pool of CA9+ cancer cells observed in
   response to SCD1 inhibition was attenuated upon oleate supplementation
   (Fig. [191]6e). Although obtained by adding excess oleate, these data
   indicate that exogenous MUFA has the potential to, at least in part,
   compensate for the deficit in MUFA synthesis and thereby possibly
   support partial resistance of cancer cells to SCD1 inhibition. Still,
   the net tumor growth inhibition observed in mice led us to search for a
   potential bystander effect, i.e. a phenomenon where cells that are not
   directly targeted by SCD1 inhibition could be nonetheless affected as a
   result of interactions with nearby treated acid-exposed cells.

   We reasoned that an increased uptake of MUFA in response to desaturase
   inhibition should lead to a steeper gradient in MUFA availability in
   tumors. The consequence could be lipid starvation for cancer cells the
   most distant from the tumor blood vessels. To explore this hypothesis,
   we first tracked the uptake of TopFluor-OA into acid-exposed cancer
   cells in the presence of oxygen or under hypoxia. We found that the
   efficiency of OA uptake tends to be lower in hypoxic cancer cells than
   in non-hypoxic acid-exposed cancer cells and, importantly, that SCD
   inhibition led to the accumulation of significantly larger OA amounts
   in the latter than in hypoxic cancer cells (Figs. [192]6f and
   [193]S6c). Hypoxic cells could thus suffer in vivo from a reduction in
   MUFA availability resulting from an increased uptake by acidic
   (non-hypoxic) cancer cells (see the model in Fig. [194]6g). Since the
   verification of this hypothesis could be biased by the reduction in
   tumor size, we designed an in vitro experiment wherein acidic cancer
   cells were exposed to SCD1i in the presence of a 1:1 mixture of OA and
   PA and supernatants were collected after 24 h. We found that the
   resulting conditioned medium (CM) led to significant growth inhibitory
   effects in hypoxic cancer cells (vs. CM from untreated acidic cancer
   cells) (Fig. [195]6h) while CM supplementation with OA rescued the
   survival of hypoxic cancer cells (Fig. [196]6h). Strikingly, direct
   exposure of hypoxic cancer cells to SCD1i did not induce cytotoxicity
   (Fig. [197]6h), in agreement with the need of O[2] to support SCD1
   enzymatic activity^[198]26. Measurements of FA concentrations confirmed
   that CM of acidic cancer cells exposed to SCD1i were deprived in MUFA
   (Fig. [199]6i) and exhibited a net increase in the SFA/MUFA ratio
   (Fig. [200]6j) more prone to induce apoptosis in hypoxic cancer cells.
   This is further supported by an SCD1i-triggered increase in ATF3
   staining (previously identified as a very sensitive ER stress marker
   resulting from SFA toxicity^[201]27) which accompanies the reduction in
   CA9 staining (Fig. [202]S6d).

Discussion

   CA9 may exert a dual role by facilitating CO[2] hydration into
   bicarbonate and H^+, or alternatively, by managing excess (glycolytic)
   H^+ through CO[2] venting^[203]18,[204]28. On the premise that all
   tumor cells located at the same distance from the center of a spheroid
   share the same phenotype, we were able to isolate a population of cells
   harboring CA9 expression but not facing deep hypoxia as detected by
   using pimonidazole or HRE-GFP reporter. The enrichment of the MUFA
   biosynthesis pathway in these cells enabled us to reveal a major role
   for FA desaturases in the survival of non-hypoxic, acid-exposed cancer
   cells. Pharmacological inhibition and genetic silencing of SCD1
   desaturase, the rate-limiting enzyme in MUFA synthesis, both identified
   the loss of the CA9+ cell population in 3D spheroids as an early event
   in the observed growth inhibitory effects. This observation prompted us
   to unveil two major phenomena with major therapeutic consequences: (i)
   the absolute requirement of MUFA to detoxify PUFA accumulation when
   cells concomitantly face an acidic environment, and (ii) a symbiotic
   relationship between acid and hypoxic tumor compartments, the latter
   being dependent on the uptake of exogenous MUFA not consumed by the
   former. The combination of these two findings substantiates the
   anticancer synthetic lethality of combining PUFA-rich diet and SCD1
   inhibition.

   Our lipidomic study first revealed that the capacity of MUFA to support
   TAG formation to incorporate PUFA (and thereby reduce the risk of
   peroxidation)^[205]23 was altered under acidic conditions. Indeed,
   while acidic conditions and PUFA (i.e., DHA exposure) both increased
   the amounts of intracellular TAG, only acidic pHe was associated with
   an increase in MUFA (i.e. C18:1 oleate) in the NL fraction, mitigating
   the increase in PUFA/MUFA ratio otherwise observed upon DHA exposure
   (Fig. [206]4b–d). Moreover, when PUFA exposure was combined with acidic
   conditions, while the TAG level was significantly increased (to a
   larger extent than each treatment separately), the MUFA contribution
   did not increase further than in response to acidic pHe alone
   (Fig. [207]4b, c). As a consequence, the PUFA/MUFA ratio was yet more
   largely enhanced (than upon exposure to DHA alone) in the NL but also
   the PL fractions (Fig. [208]4d, e), supporting an increased sensitivity
   to ferroptosis (Fig. [209]6d). Importantly, we documented that this
   observation was largely dependent on the blunting of the
   acidosis-driven SCD1 upregulation in the presence of exogenous PUFA
   (Fig. [210]4f–h). This finding underscores the bidirectional nature of
   the relationship between PUFA and SCD1 since earlier studies indicated
   that a lack of PUFA leads to an upregulation of SCD1 expression and
   activity, subsequently elevating plasma triglyceride
   concentrations^[211]29. Strikingly, while a repressing PUFA-responsive
   element was previously identified in the SCD1 promoter region^[212]30
   and is likely to account for the SCD1 downregulation in acid-exposed
   cells, we found that PUFA-sensitive PPAR transcription factor largely
   drives SCD1 regulation in these cells, as previously reported in the
   liver^[213]31. It is likely that a biphasic PUFA dose effect could
   explain this seemingly paradoxical regulation. Interestingly, Kondo et
   al. identified SREBP2 as another transcription factor highly sensitive
   to acute exposure to acidic pHe^[214]32, suggesting that FA but also
   cholesterol metabolism are major hallmarks of tumor acidosis.

   The second major finding of this study that supports synthetic
   lethality of the PUFA and SCD1 inhibitor combination, was related to
   the MUFA deprivation of hypoxic cancer cells resulting from the
   triggered dependency of acid-exposed cancer cells on exogenous MUFA.
   SCD1 activity was previously shown to be critical under specific
   nutrient (lipid)-compromised environments including low serum culture
   conditions,^[215]33 poorly vascularized CNS niches (ie, cerebrospinal
   fluid)^[216]34, and calorie restriction diet^[217]35. Young et al. also
   documented that SCD1 inhibition in normoxic cancer cells phenocopied a
   deficit in MUFA under hypoxia in cells exhibiting oncogenic mTOR
   activation^[218]36. MUFA supplementation (or mobilization from lipid
   droplets)^[219]37 but also SCD1 upregulation have been associated with
   a reduction in SFA-driven ER stress^[220]38–[221]40, displacement (or
   replacement) of PUFA to reduce the risk of lipid ROS accumulation
   within membrane phospholipids^[222]41 or both^[223]24. Our work adds
   the spatial dimension to the above studies by documenting that within
   the same tumor, attempts to compensate SCD1 inhibition by enhanced
   uptake of exogenous MUFA may work for one cell compartment but at the
   detriment of another one. We showed indeed that PUFA supplementation by
   dampening SCD1 expression greatly reinforces the addiction of
   acid-exposed cancer cells to the external MUFA source. As a
   consequence, upon exposure to SCD1 inhibitor, acid-exposed cancer cells
   avidly consume MUFA, depriving the extracellular milieu from them. This
   effect combined with the incapacity of SCD1 to desaturate SFA in the
   absence of O[2] accounts for the killing of hypoxic cancer cells unable
   to mitigate ER stress. This collateral impact on surrounding hypoxic
   cells (bystander effect) bolstered by the paradoxical response of
   tumor-bearing mice (fed either a control or a PUFA-rich diet) to SCD1
   inhibition (i.e., increase and decrease in the tumor TAG levels,
   respectively) was confirmed in an in vitro model decomposing the
   hypoxic and acidic (non-hypoxic) cell compartments. Remarkably, our
   study also emphasizes the concomitance and the mutual reinforcement of
   two distinct modes of cell death caused by a deficit in MUFA
   availability in the same tumor, namely apoptosis resulting from
   SFA-induced ER stress^[224]27,[225]36,[226]40 and alterations in
   cardiolipin^[227]33 but also PUFA-driven ferroptosis in acid-exposed
   cancer cells (this study). In this context, it is worth noting that
   acid-exposed cancer cells (contrary to hypoxic cancer cells) exhibit
   very low basal ER stress, as indicated by ATF4 and Bip markers
   (Fig. [228]S6e), supporting the beneficial effects of enhanced MUFA
   synthesis in these cells. This finding also aligns with the growing
   interest in pro-ferroptotic approaches (rather than pro-apoptotic
   chemotherapies) to prevent the emergence of resistance from unaltered
   tumor compartments^[229]42. More generally our work identifies SCD1
   activity as a control point for the O[2] balance between different
   tumor regions, directly contributing to the O[2] gradient formation.
   Indeed, we previously reported that FAO is upregulated (concomitantly
   to FAS) in acid-exposed cancer cells in part through enhanced PPAR
   activity^[230]11,[231]15,[232]23, itself responding to increased
   availability of FA, including MUFA. This may explain why O[2]
   consumption is reduced when MUFA synthesis is inhibited together with
   the ETC modulation upon alterations in the inner mitochondrial
   membrane^[233]33 and the suppression of O[2] consumption by SCD1
   itself.

   Regarding in vivo therapeutic options derived from this extended
   knowledge of metabolic addiction of acid-exposed cancer cells, we
   combined the administration of SCD1 inhibitors with a dietary PUFA
   enrichment in line with DHA supplementation of a normal diet^[234]23.
   This combined approach repositions SCD1 inhibition as a realistic
   anticancer target despite the physiological compensatory response
   resulting from exogenous MUFA uptake. This strategy may be related to
   the capacity of DGAT inhibitors to prevent the formation of TAG and
   thereby tip the balance toward an increase in plasma membrane PUFA that
   increases the risk of peroxidation^[235]23. Interestingly, a close
   vicinity of SCD1 and DGAT2 was recently reported, the former promoting
   the channeling of MUFA to the latter^[236]43 and indirectly supporting
   a preference for MUFAs in TAG synthesis.

   In summary, our study identified the acidic tumor compartment as
   exquisitely dependent on the availability of MUFA preferentially
   synthesized intracellularly using desaturases such as SCD1 but also
   captured from the extracellular medium in response to desaturase
   inhibition. Our work unraveled the interruption of a symbiotic dialog
   between cells able to use SCD1 to generate endogenous MUFA and those
   lacking O[2] and thus more dependent on exogenous MUFA. Upon SCD1i
   exposure, the forced uptake of exogenous MUFA by the former induced
   cytotoxicity in the latter. On top of this, in the presence of PUFA
   supplementation, acidic cancer cells will also directly suffer from
   SCD1i, the effect of which is further reinforced by the downregulation
   of the enzyme expression and the consecutive increase in the PUFA/MUFA
   ratio prone to support ferroptosis.

Methods

Ethical statement

   This research complies with all relevant ethical regulations.
   Experiments involving mouse xenograft received the approval of the
   ethics committee from UCLouvain (approval ID 2020/UCL/MD032) and were
   carried out according to national care regulations.

Cell culture

   Human head and neck hypopharyngeal (FaDu, HTB-43), colorectal (HCT116,
   CCL-247; HT-29, HTB-38), and cervix (SiHa, HTB-35) cancer cell lines
   were purchased from ATCC. Cell lines were stored according to the
   supplier’s instructions and used within 6 months after resuscitation of
   frozen aliquots. All cells were cultured in DMEM supplemented with 10%
   heat-inactivated FBS, 10 mM D-glucose, 2 mM L-glutamine (or glutamax),
   and if needed 25 mM of both PIPES and HEPES before adjusting pH to 7.4
   or 6.5. Acidosis pH-adapted tumor cells were established as previously
   described^[237]10. All cell lines were tested for mycoplasma
   contamination with the MycoAlert™ Mycoplasma Detection kit (LT07-318;
   Lonza) before being used. After collection, cell number was assessed by
   cell counting on a hematocytometer with Trypan Blue mortality exclusion
   dye.

Cell transfection and lentiviral infection

   Cell transfection of 5×HRE/GFP plasmid (#46926, Addgene) was carried
   out with Lipofectamine LTX with plus reagent (#15338100, Thermo Fisher
   Scientific), according to manufacturer’s instructions. Transfected cell
   lines were selected with G418 treatment (2 mg/mL) for 1 week.

   For CRISPRS-Cas9 gene invalidation, 15 µg of plasmids containing the
   constitutive Cas9 gene (VB221215-1151jyw, VectorBuilder) were
   transfected by Calcium Phosphate (#CAPHOS-1KT, Merck) technique in
   HEK293T packaging cells together with 15 µg of psPAX (#12260, Addgene)
   and pMD2G (#12259, Addgene) second-generation packaging plasmids by
   following the manufacturer’s indications. The produced virions were
   transferred to target cell lines after filtration. After 24 h of
   transduction, the virion-containing media was removed and the
   transduced cells were selected with 8 µg/mL blasticidin for 1 week. A
   subsequent packaging and transduction were performed on Cas9 expressing
   cell lines with single guide (sg) RNA necessary for SCD1 gene
   invalidation at the position 43 (CCTGCTTGGCAGCGGATAAA, VectorBuilder)
   and 109 (GCCTTTAAATTCCCGGCTCG, VectorBuilder) of the target gene or a
   CTL (sg) RNA (GTGTAGTTCGACCATTCGTG, VectorBuilder), Cancer cells were
   infected with sg190 (C2) or a combination of sg43 and sg109 (C1 and
   C3). Following infection, the cells were subjected to a 1-week long
   puromycin selection (1 µg/mL) and a clonal selection to isolate
   individual clones for further analysis.

   For the overexpression of SCD1 (VB900124-8400vbs, VectorBuilder) and
   shRNA-mediated downregulation of SCD1 (target sequence:
   CGTCCTTATGACAAGAACATT, VectorBuilder), second-generation packaging and
   transduction were performed, as previously described. Scramble controls
   were used for both conditions, with mRFP for overexpression and a
   scramble sequence (CCTAAGGTTAAGTCGCCCTCG) for downregulation.
   Post-transduction, cells were subjected to a 1-week selection using
   1 µg/mL puromycin.

3D spheroid models

   Spheroids were formed with FaDu, HCT116, and HT-29 cell lines and
   cultured in commercial DMEM media supplemented with 10%
   heat-inactivated FBS, 10 mM of D-glucose, and 2 mM of glutamax. They
   were seeded at a density of 1500 cells/well in 96-well ultra-low
   attachment plates (7007, Costar or 83.3925.400, BIOFLOAT) and cultured
   for 7 days before performing further experiments.

Fatty acid supplementation and treatments

   Fatty acid supplementation was performed in full media with
   lipid-depleted FBS (S181L-500, VWR) and BSA-conjugated fatty acids at a
   concentration of 50 µM of docosahexaenoic acid (DHA, 10-2206, Larodan),
   100 µM of palmitic acid (PA, 10-1600, Larodan) and/or oleic acid (OA,
   10-1801, Larodan) during 24, 48 or 72 h. 3D spheroids were supplemented
   with 100 µM of DHA and OA or 50 µM of PA and SA (stearic acid,
   10-1800-17, Larodan) during 24 or 72 h in full media. Fatty acids were
   conjugated with BSA in PBS to obtain a fatty acid/BSA ratio of 4:1
   (mol:mol). Aliquoted fatty acids were stored at −80 °C under light
   protection and directly used upon unfreezing by the addition of the
   requested concentrations into the wells.

   2D and 3D culture treatments were performed with full media, during
   72 h with 32 µM of SCD1 inhibitor (A939572, #HY-50709, Med Chem
   Express) or during 24 h with 25 µM of PPARα inhibitor (GW6471, S2798,
   Bioconnect) and PPARγ inhibitor (GW9662, HY-16578, MedChemExpress).

   Cell density was determined by using the Presto Blue reagent (Thermo
   Fisher Scientific) according to the manufacturer’s instructions.
   Spheroid growth was monitored using a live-cell phase contrast
   microscope (Axio Observer, Zeiss) or IncuCyte automated system (#SX5,
   Sartorius) and the spheroid area was measured by Fiji software or
   IncuCyte software.

Patient-derived tumor organoids

   Tumor organoids were generated from a local library of specimens
   collected from CRC patients undergoing surgery or tumor biopsy at
   Cliniques universitaires St Luc in Brussels (ethics committee approval
   n°ONCO-2015-02 updated on 13-05-2019 in accordance with the principles
   of the Declaration of Helsinki). The organoids used in this study were
   obtained from a male patient with CRC who did not receive any
   compensation. Signed informed consent was obtained before tumor
   collection and all personal/clinical data remained strictly
   confidential for the investigators. Organoids were cultured in Advanced
   DMEM/F12 media (Gibco) supplemented with glutamine,
   penicillin/streptomycin, primocin 100 µg/ml (Invivogen), B27 without
   vitamin A 1× (Thermo Fisher Scientific), N2 1× (Thermo Fisher
   Scientific), Nicotinamide 100 mM (Sigma), N-acetylcysteine 1.25 mM
   (Sigma) and hEGF 50 ng/ml (STEMCELL Technologies) and embedded in
   Cultrex BME type 2 (Biotechne), using Corning Costar 24 well plates
   (Ref:3524). The organoids were allowed to grow in their media at
   physiological pH but over the duration of the SCDi test, the organoids
   media was set at pH 6.9. The organoids have been treated for 96 h with
   vehicle or SCDi 20 µM, and the corresponding fatty acids (OA 100 µM
   and/or DHA 50 µM) or BSA (used in the same amounts as the fatty acids
   accordingly).

Immunofluorescence staining of 2D cell cultures

   To perform the immunostaining of SCD1, cancer cell monolayers were
   washed twice with PBS and fixed with PFA 4% for 20 min at room
   temperature (RT). After two washes, cells were permeabilized with 0.1%
   saponin for 20 min at RT and washed 3 times. Then, cells were blocked
   with a solution of 2% BSA/ 0.05% Saponin for 1 h at RT and incubated
   with an SCD antibody (#ab236868, Abcam, 1:100) diluted in 0.1% BSA/
   0.05% Saponin, overnight at 4 °C. After the incubation, the cells were
   washed three times with cold PBS and then incubated with an Alexa Fluor
   488-conjugated anti-rabbit secondary antibody (#A32790, Thermo Fisher
   Scientific, 1:600) and DAPI diluted in 0.1% BSA/ 0.05% Saponin, for 1 h
   at RT. Finally, cells were washed three times with PBS before being
   imaged with an LSM 800 confocal microscope (Zeiss).

   To stain lipid droplets, cancer cell monolayers were washed twice with
   warmed PBS and incubated with 2 µM BODIPY 493/503 (D3922, Thermo Fisher
   Scientific) in PBS for 20 min at 37 °C and 5% CO2. After three washes,
   cells were fixed with PFA 4% for 20 min at RT in the dark. Cells were
   washed three times more with PBS and imaged with LSM 800 confocal
   microscope (Zeiss). 2D immunofluorescence images were analyzed with
   ImageJ. Briefly, each cell was delimited to create a Mask and saved as
   a Text Image corresponding to pixels from cells. In parallel, the same
   threshold was applied to all images and saved as Text Image,
   corresponding to interest pixels (SCD or LD staining). Using the
   RStudio software, interest pixels per cell were quantified.

Immunofluorescence staining of 3D spheroids

   3D tumor spheroids were treated with 1 mg/mL of propidium iodide
   (P4170, Sigma) for 15 min or 1 µM of pimonidazole for 24 h, fixed with
   4% PFA during 24 h, incubated with 30% sucrose and embedded in OCT.
   Frozen spheroid sections (8 µm) were permeabilized in 0,5% of
   PBS-Triton X-100, blocked with 5% BSA during 1 h, incubated overnight
   at 4 °C with anti-CA9 (#AF2188, R&D, 1:500), anti-SCD1 (#ab236868,
   Abcam, 1:200), anti-pimonidazole (#Pab27, HPI, 1:200), anti-ATF3 (#
   MA5-31360, Thermo Fisher Scientific, 1:100) or anti-PLIN2 (#HPA016607,
   Merck, 1:100) primary antibodies. Spheroid sections were then incubated
   with Alexa Fluor 488-conjugated anti-rabbit and 568-conjugated
   anti-goat secondary antibodies (#A32790 and #A11057, respectively;
   Thermo Fisher Scientific) for 1 h and nuclei were counterstained with
   Hoechst. Slides were prepared with fluorescence mounting medium (Dako)
   and the staining was visualized by AxioImager.z1-ApoTome1 (Zeiss). All
   spheroid samples from the same experiment were imaged with the same
   gain and exposure settings. Immunofluorescence results were analyzed
   with Zen Blue (Zeiss) and quantified with QuPath software. The tissue
   total area and the area of the staining of interest were measured by
   creating a pixel classifier with a fluorescence thresholder. The area
   of a staining compartment was measured by creating an annotation of the
   zone where the staining is present.

Wholemount immunofluorescence and Light Sheet microscopy

   3D spheroids containing 5HRE/GFP plasmid were fixed with 4% PFA for
   24 h, blocked, and permeabilized at 4 °C overnight in 0.5% of
   PBS-Triton X-100, containing 5% of BSA. Whole spheroids were incubated
   with the anti-CA9 primary antibody (#AF2188, R&D, 1:500) at 4 °C for 5
   days under shaking and then incubated with 568-conjugated anti-goat
   secondary antibody (#A11057, Thermo Fisher Scientific) and Hoechst at
   4 °C for 4 days, under shaking. The antibody incubation was followed by
   2 h of 4% PFA fixation and spheroid clearing with CUBIC 1 for 4 days
   and CUBIC 2 for 3 days. CUBIC1 and 2 were prepared by following the
   instructions of Susaki et al.^[238]44. Spheroids were mounted into 1%
   agarose in capillaries and visualized with Lightsheet.z1 (Zeiss). All
   spheroid samples from the same experiment were imaged with the same
   gain and exposure settings. Immunofluorescence results were analyzed
   with Zen Blue (Zeiss), modelized with Imaris software, and quantified
   with QuPath software.

Multiplex immunofluorescence in tumor sections

   Tumors collected from mice were fixed with 4% PFA for 48 h and embedded
   in paraffin. Paraffin tumor sections (5 µm) were deparaffinized with 3
   consecutive baths of toluene and isopropanol. Endogenous peroxidases
   were inactivated with 3% H[2]O[2] diluted in 40% isopropanol. Before
   each primary antibody staining, antigen retrievals of tumor sections
   were performed with citrate buffer, followed by permeabilization with
   0.1% of TBS-Tween 80 and blocking with 5% BSA for 1 h. Consecutive
   primary antibody stainings were performed with anti-pimonidazole
   (#Pab27, HPI, 1:1000), anti-PLIN2 (#HPA016607, Merck, 1:100) and
   anti-CA9 (#AF2188, R&D, 1:500) antibodies during 1 h at RT. Slides were
   then incubated with anti-rabbit-HRP or anti-goat-HRP secondary
   antibodies (#K4003, Dako, ready to use; #705-035-147, Jackson Immuno
   Research, 1:100) during 40 min, followed by a 10 min incubation with
   Alexa Fluor 555-([239]B40955, Thermo Fisher Scientific, 1:200), Alexa
   Fluor 488- (#[240]B40953, Thermo Fisher Scientific, 1:200) and Alexa
   Fluor 750-coupled (#[241]B56131, Thermo Fisher Scientific, 1:1000)
   tyramide diluted in borate buffer with 0.003% H2O2, respectively.
   Nuclei were counterstained with Hoechst, slides were prepared with
   fluorescence mounting medium (Dako) and acquired with Axioscan.z1
   (Zeiss). All tumor samples from the same experiment were imaged with
   the same gain and exposure settings. Immunofluorescence results were
   analyzed with Zen Blue (Zeiss) and quantified with QuPath software.

Flow cytometry and FACSorting

   6.5 and 7.4 pH-adapted 2D cell lines were collected and stained with
   PE-coupled anti-CA9 antibody (130-123-340, Miltenyi, 1:200) for 20 min
   at 4 °C. To assess viability, cells were incubated with DAPI before
   acquisition. 3D spheroids were harvested and dissociated with Accumax
   (07921, STEMCELL) or Accutase (07922, STEMCELL) at RT under shaking for
   10 min. The staining was performed with PE-coupled anti-CA9 antibody
   (130-123-340, Miltenyi, 1:200) for 20 min at 4 °C. To assess viability,
   cells were incubated with DAPI before acquisition. After incubation,
   cells were acquired by flow cytometry on FACSCanto II (BD Biosciences)
   or sorted with FACSAria III (BD Biosciences) with a gating strategy
   excluding debris, double cells, and dead cells.

   For BODIPY-C11, 6.5 pH-adapted cell lines were treated with SCDi
   (12 µM) and/or DHA (50 µM) in the presence or not of OA (100 µM) or
   α-Tocopherol (20 µM) (#T3634, Sigma Aldrich). After 72 h, the cells
   were harvested and incubated with 2 µM of BODIPY-C11 (#D3861, Thermo
   Fischer Scientific) for 30 min at 37 °C. After incubation, cells were
   acquired by flow cytometry on FACSCanto II (BD Biosciences) with FITC
   channel corresponding to oxidized BODIPY-C11.

   For TopFluor-OA, 6.5 pH-adapted cell lines were plated at 40%
   confluence and maintained in Normoxia or Hypoxia (1% O[2]). After 24 h,
   cells were treated with SCDi (12 µM) for 24 h. Then, cells were
   incubated with 2 µM TopFluor-OA (810259, Avanti Research) in a
   serum-free medium for 30 min at 37 °C (still in normoxia or hypoxia).
   After incubation, cells were acquired by flow cytometry on FACSCanto II
   (BD Biosciences) with a FITC channel.

   Flow cytometry analysis was performed with the FlowJo software.

RNA Sequencing analysis

   1 million cells per condition were isolated from spheroids by
   FACSorting. The RNA quantity and quality evaluation, the RNA
   extraction, and the RNA Sequencing were performed at Marcogen. Briefly,
   samples (biological triplicates) were prepared with the TruSeq
   Standarded mRNA Library Prep Kit (#20020594; Illumina) from 1 μg RNA.
   The sequencing was performed using the NovaSeq 6000 Sequencing system
   with paired-end reads, all according to the manufacturer’s
   recommendations. All RNA-sequencing data were analyzed by using the
   Automated Reproducible MOdular workflow for preprocessing and
   differential analysis of RNA-seq data (ARMOR v1.5.4) pipeline. In this
   pipeline, reads underwent quality control with FastQC (Babraham
   Bioinformatics). Quantification and quality control results were
   summarized in a MultiQC report before being mapped to the human
   transcriptome using Salmon, which was built by applying all Ensembl
   cDNA sequences obtained in the Homo_sapiens.GRCh38.cdna.all.fa file.
   Next, estimated transcript abundances from Salmon were imported into R
   using the tximeta package and analyzed for differential gene expression
   with edgeR. Over Representation Analysis (ORA) and Gene Set Enrichment
   Analysis (GSEA) were performed with the WebGestaltR v.0.4.4
   Bioconductor package on the KEGG, GO, and HALLMARKS databases. Raw and
   processed RNA-seq data were deposited and made publicly available on
   the Gene Expression Omnibus.

Lipid analysis

   Lipid extraction from cultured cancer cells and collected tumor samples
   and data analysis methods were previously described. Briefly, gas
   chromatography-flame ionization detection (GC-FID) was used to quantify
   fatty acids as phospholipids, neutral lipids, and free fatty
   acids^[242]23 while lipid species were analyzed using a targeted
   multiple reaction monitoring (MRM) method^[243]45; MRM transitions are
   available in the Metabolomics Workbench repository.

RNA Extraction and RT-qPCR

   The total RNA of 2D cells was collected from an equal number of intact
   cells in TRI Reagent (#TR118; Molecular Research Center). RNA was
   recovered after separation in 1-bromo-3-chloropropane and precipitation
   with isopropanol, washed with ethanol (70%), and resuspended in
   RNAse-free water. The total RNA of cells isolated from 3D spheroids
   after FACSorting was extracted using the Maxwell RSC SimplyRNA tissue
   kit (#AS1340; Promega) according to manufacturer’s instructions
   including a DNAse treatment. RNA quantity and quality were evaluated
   with a Nanodrop 1000 spectrophotometer (Thermo Fisher Scientific).
   After reverse transcription on 500 ng of total RNA with the RevertAid
   Reverse-Transcriptase, oligo-dT, and random hexamers (Thermo Fisher
   Scientific), quantitative PCR amplification was performed on the CFX
   Connect Real-Time PCR detection system (#1855201, Biorad) using Takyon
   Low Rox SYBR® MasterMix dTTP Blue (#UFLSMT-B0701; Eurogentec) and
   sequence-specific primers for SCD (FW:TCATAATTCCCGACGTGGCT;
   REV:CCAGGTTTGTAGTACCTCCTCTG), SCD5 (FW:CATCTGGGGAGAGAGTCTGTG;
   REV:ATTATGGAAGCCTTCACCAATGGC), FADS1 (FW:TGCAATGTCCACAAGTCTGC;
   REV:AGCTGCCCTGACTCCTTTAG), and FADS2 (FW:GACCACGGCAAGAACTCAAAG;
   REV:GAGGGTAGGAATCCAGCCATT) genes. Relative gene expression was
   calculated using the ddCt method, with RPLP0 (FW: AGCCCAGAACACTGGTCTC;
   REV: ACTCAGGATTTCAATGGTGCC) as the reference gene.

Western blotting

   Cancer cells were washed with PBS and lysed in a RIPA buffer
   supplemented with a protease inhibitor. Protein concentration was
   determined using a bicinchoninic acid-based assay (Thermo Fisher
   Scientific). A sample denaturation was performed for 7 min at 95 °C
   with Laemmli sample buffer containing 100 mM dithiothreitol. Then,
   samples (25 mg per well) were separated by SDS-PAGE (15%
   acrylamide/bis-acrylamide gels) and transferred onto PVDF membranes.
   Membranes were then blocked with 5% skimmed milk diluted in TBS-0.1%
   Tween 20 (TTBS) and immunoblotted overnight at 4 °C with the specific
   SCD1 (#ab236868, Abcam, 1:1000), BiP (#3177, Cell Signaling Technology,
   1:1000), ATF4 (#11815 Cell Signaling Technology, 1:1000) primary
   antibody and HSP90 (#610419, BD Biosciences, 1:10,000) or β-actin
   (A5441, Sigma-Merck, 1:10,000) primary antibody as housekeeping. After
   several washes with TTBS, membranes were incubated for 1 h at room
   temperature with horseradish peroxidase (HRP)-conjugated secondary
   antibodies (Jackson Immunoresearch), and chemiluminescent signals were
   revealed by using ECL Western Blotting Detection Kit (GE Healthcare) on
   X-ray films in a dark chamber.

IncuCyte experiments

   The IncuCyte (#SX5, Sartorius) was used to assess spheroid growth and
   toxicity with the IncuCyte Cytotox Green Reagent (#4633, Sartorius,
   25 nM) according to the manufacturer’s recommendation. The Cytotox
   Green reagent was added to the culture together with inhibitors (7 days
   after spheroid formation). Spheroid images were acquired every 6 h with
   the 10× objective, the area, and the green fluorescence were measured
   with the Incucyte Software.

Seahorse experiments

   All monolayer cells were plated in on Seahorse Bioanalyzer XFe96
   culture plates overnight in DMEM media before treatment with SCD1i
   (15 µM), with or without palmitic acid or stearic acid (25 µM).
   Spheroids were treated in Ultra-Low attachment plates with SCD1i
   (15 µM) 24 h before being transferred to XFe96 Spheroid Microplates
   plates, pre-coated with poly-D lysine (50 µg/mL). For all experiments,
   cells were incubated for 45 min à 37 °C in Seahorse XF-base medium
   supplemented with 1 mM glucose and 1 mM glutamine, before the
   experiment. According to the test, pH was maintained at 7.4 or 6.5,
   with or without SCD1i (15 µM) and PA (25 µM). OCR was measured with
   Seahorse XFe96 Analyzer (Agilent) and data analysis was performed using
   Seahorse Wave 2.4 software.

Acidic-conditioned medium and hypoxic conditions

   When acidic adapted cancer cells reached 60% confluence of a 6-well
   plate, FADU, and HCT116 were treated respectively with 15 µM or 25 µM
   SCDi, in 6.5 medium complemented with 5% of delipidated FBS (S181L-500,
   VWR), 50 µM PA and 10 µM OA, for 24H. Half of the medium volume was
   used to quantify FA, the other was used as a conditioned medium to
   treat acidic cancer cells in a hypoxia chamber at 1% O2, 5% CO[2],
   37 °C (H35 Hypoxystation, Don Whitley Scientific). As a third, 50 µM OA
   was added in the control conditioned medium and as a fourth condition,
   SCDi was added in the control conditioned medium at the concentrations
   described above and a non-conditioned medium complemented as described
   above was used as control. After 48 h incubation, cell viability was
   determined by using the Presto blue reagent (A13261, Thermo Fisher
   Scientific) according to the manufacturer’s instructions.

In vivo mouse experiments

   All mice were housed in the same experimental room with the appropriate
   12 h light/12 h dark cycles and controlled for temperature (21-24 °C)
   and humidity (40-60%). Mice were fed with isocaloric control (Safe,
   U8978 P0177) or PUFA-rich anchovy/sardine oil-containing (Safe, U8978
   P0177) diet. Rj:NMRI-Foxn1nu/nu 5-week female mice (Janvier), were
   first accommodated to the specific diets for two weeks prior to cancer
   cell injection, the diets were maintained throughout the experiment.
   Before injection control or SCD1 knock out FaDu cancer cells were
   washed twice with PBS, resuspended in 0.9% NaCl and 2 × 10^6 cells were
   injected in the right flank of each mouse. When tumors reached
   50–100 mm^3, mice were treated with SCD1 inhibitor (30 mg/kg) diluted
   in DMSO by intraperitoneal injection for 5 days per week for three
   weeks. Control groups were treated in parallel with equivalent
   vehicles. Mice were weighed once per week and tumors were measured 3
   times a week. None of the tumor volumes of mice in this study exceeded
   the limits authorized by the local ethical committee. At the end of the
   experiment, when tumor size reached the ethical endpoint (500 mm^3 or
   visible necrosis), mice were sacrificed and blood, liver, and tumors
   were collected. Collected samples were separated in two for fatty acid
   quantification, and immunohistochemistry.

Statistics and reproducibility

   Statistical analyses were performed through GraphPad Prism 10 by using
   Student’s t-test, one-way ANOVA with Tukey’s, Sidak’s or Dunnet’s
   multiple comparison test, two-way ANOVA with Tukey’s multiple
   comparison tests or with Fisher’s LSD test and Long-rank Mantel–Cox
   test when appropriate. Statistical significance is indicated as
   follows: exact P-values are provided when *P ≤ 0.05 or **P < 0.01,
   while ***P < 0.001 and ****P < 0.0001 indicate threshold values. No
   statistical method was used to predetermine the sample size. The
   experimental sample size was chosen based on commonly accepted
   standards. No data were excluded from the analyses. Mice were randomly
   assigned to each group, other experiments were not randomized. The
   investigators were not blinded to allocation during experiments and
   outcome assessment.

Reporting summary

   Further information on research design is available in the [244]Nature
   Portfolio Reporting Summary linked to this article.

Supplementary information

   [245]Supplementary Information^ (8.5MB, pdf)
   [246]41467_2024_54435_MOESM2_ESM.docx^ (12.9KB, docx)

   Description of Additional Supplementary Information
   [247]Supplementary Movie 1^ (277.5MB, mp4)
   [248]Reporting Summary^ (1.6MB, pdf)
   [249]Transparent Peer Review file^ (1MB, pdf)

Source data

   [250]Source data^ (3.4MB, xlsx)

Acknowledgements