Fatty acids and cancer-amplified ZDHHC19 promote STAT3 activation through S-palmitoylation.

Signal transducer and activator of transcription 3 (STAT3) has a critical role in regulating cell fate, inflammation and immunity1,2. Cytokines and growth factors activate STAT3 through kinase-mediated tyrosine phosphorylation and dimerization3,4. It remains unknown whether other factors promote STAT3 activation through different mechanisms. Here we show that STAT3 is post-translationally S-palmitoylated at the SRC homology 2 (SH2) domain, which promotes the dimerization and transcriptional activation of STAT3. Fatty acids can directly activate STAT3 by enhancing its palmitoylation, in synergy with cytokine stimulation. We further identified ZDHHC19 as a palmitoyl acyltransferase that regulates STAT3. Cytokine stimulation increases STAT3 palmitoylation by promoting the association between ZDHHC19 and STAT3, which is mediated by the SH3 domain of GRB2. Silencing ZDHHC19 blocks STAT3 palmitoylation and dimerization, and impairs the cytokine- and fatty-acid-induced activation of STAT3. ZDHHC19 is frequently amplified in multiple human cancers, including in 39% of lung squamous cell carcinomas. High levels of ZDHHC19 correlate with high levels of nuclear STAT3 in patient samples. In addition, knockout of ZDHHC19 in lung squamous cell carcinoma cells significantly blocks STAT3 activity, and inhibits the fatty-acid-induced formation of tumour spheres as well as tumorigenesis induced by high-fat diets in an in vivo mouse model. Our studies reveal that fatty-acid- and ZDHHC19-mediated palmitoylation are signals that regulate STAT3, which provides evidence linking the deregulation of palmitoylation to inflammation and cancer.

STAT3 is an important regulator of immunity, inflammation, and tumorigenesis[1,2]. Previously, JAK or Src-mediated tyrosine phosphorylation is considered as a major regulatory mechanism of STAT3[3,4]. Posttranslational S-palmitoylation or S-fatty acylation attaches fatty acids, such as palmitate, to the cysteine residues of a protein, and regulates diverse functions of proteins under physiological and pathological conditions[5,6]. Using chemical reporters of protein palmitoylation[7,8], we identified that STAT3 is palmitoylated (Extended Data Fig. 1a), consistent with other proteomic studies using acyl exchange methods in adipocytes[9]. Further studies using chemical reporters in multiple cell lines confirmed that ectopically expressed and endogenous STAT3 is indeed S-palmitoylated through thioester bonds, which are cleavable by hydroxylamine (NH2OH) treatment (Fig. 1a and Extended Data Fig. 1b, c. Original scans of all blots in Supplementary Fig. 1). In addition, we detected S-palmitoylation of STAT1 (α and β), STAT5B and STAT6, but not of STAT2 and 4 (Extended Data Fig. 1d–h), indicating that palmitoylation among STAT family members is different. Moreover, STAT3 can be efficiently labeled by palmitoylation (Alk-C16) or stearoylation (Alk-C18) probes, but much less efficiently labeled by probes with C14 or C20 chain length (Fig. 1b), suggesting that palmitoyl (C16) and stearoyl (C18) are the major acyl groups for STAT3 modification. Interestingly, nuclear STAT3 was highly palmitoylated (~75%) compared with cytoplasmic STAT3 (~20%) as showed by acyl-PEG exchange (APE) assay, which enables the separation and semi-quantification of acylated protein (Fig. 1c, d). In pulse-chase experiments, we found that the half-life of STAT3 palmitoylation turnover was ~1.8 hours, while the total protein remained stable during the assay period (Fig. 1e, and Extended Data Fig. 1i), suggesting that STAT3 palmitoylation is a dynamic process. Furthermore, the SH2 domain deletion (ΔSH2) and the C-terminal truncation (STAT31−585) mutants cannot be palmitoylated, suggesting that the palmitoylation sites might be located in SH2 domain or C-terminus (Extended Data Fig. 1j, k). Mutating the evolutionarily conserved C687 and C712 located in this region to serine, singly (C687S, C712S) or in combination (C687/712S, or 2CS) could markedly block palmitoylation (Fig. 1f, and Extended Data Fig. 1l–n), suggesting that C687 and C712 are required for STAT3 palmitoylation. We did not observe significant changes between STAT3 wild type (WT) and 2CS mutant in thermal stability and the pattern of disulfide bond formation, suggesting that these mutations might not alter overall folding of STAT3 significantly (Extended Data Fig. 1o–r).

Extended Data Figure 1.

STAT3 is S-palmitoylated at evolutionarily conserved cysteine residues and mutation of STAT3 palmitoylation sites does not affect disulfide formation and protein stability.

(a) HEK293A cells were incubated for 4 hours with DMSO or 50 μM clickable probes. Cell lysates were reacted with azide-biotin for enrichment of labeled proteins with streptavidin beads and identified by mass spectrometry. The peptide spectral counts were shown in the table. (b) HEK293A cells were labeled with 50 μM chemical probe (Alk-C16) for 4 hours. Cell lysates were reacted with biotin-azide and precipitated with streptavidin beads and subjected to western blot using anti-STAT3 antibody. Endogenous STAT3 palmitoylation was analyzed by western blot. (c) HEK293A cells were transfected with Myc-STAT3 WT and labeled with 50 μM chemical probe Alk-C16 for 4 hours, cell lysates were subjected to Streptavidin blot following anti-Myc IP and subsequent Click reaction. (d) HEK293A cells were transfected with empty vector or GFP-STAT1 and labeled with 50 μM chemical probe Alk-C16 for 4 hours. After Click reaction, cell lysates were subjected to streptavidin blot showing detection of STAT1. (e) Similar analysis as panel (c) was performed in HEK293A cells transfected with Flag-STAT1α and Flag-STAT1β. (f) HEK293A cells were transfected with Flag-STAT3 or Flag-STAT5 and labeled with 50 μM chemical probe Alk-C16 for 4 hours, Cell lysates were reacted with biotin-azide and precipitated with streptavidin beads and subjected to western blot using anti-Flag antibody. (g-h) Similar analysis as panel (e) was performed in HEK293A cells transfected with Flag-STAT3, Flag-STAT2, Flag-STAT4 and Flag-STAT6. (i) HEK293 cells were pulse labeled with 50 μM chemical probe (Alk-C16) for 4 hours and subsequently chased by the addition of an excess of 50 μ M BSA-conjugated palmitic acid. Cells were harvested at the indicated time point and subjected to analysis of STAT3 palmitoylation. (j) Schematic representation of Flag-STAT3 and Flag-STAT3 truncation mutant constructs. (k) HEK293A cells were transfected with vector control, Flag-tagged wild type (WT), SH2 deletion (ΔSH2), STAT3 1–585 or STAT3 586–770 mutant, and labeled with 50 μM chemical probe (Alk-C16) for 4 hours. STAT3 palmitoylation levels were analyzed by Click reaction and streptavidin bead pull-down, followed by western blotting using anti-Flag antibody. (l) Alignment of STAT3 protein sequence among different species. (m) HEK293A cells were transfected with Flag-tagged wild type (WT), C687S, C712S or C687/712S (2CS) mutant of STAT3, and total cell lysates were subjected to Acyl-PEG exchange (APE) assay. Samples were analyzed by western blot using anti-Flag antibody. The upper band indicated the palmitoylated STAT3. (n) The palmitoylated STAT3 in panel (m) were quantified using Image J. The data are presented as mean ± s.e.m., n = 3 biologically independent experiments. P value is determined by two-tailed t-test. (o) HEK293A cells were transfected with Flag-tagged wild type (WT), C687S, C712S or C687/712S (2CS) mutant of STAT3, and total cell lysates were reduced with DTT. Samples were analyzed by western blot using anti-Flag antibody. (p) HEK293A cells were transfected with Flag-tagged wild type (WT) or C687/712S (2CS) mutant of STAT3, and free thiols of protein were blocked by NEM. Disulfide bonds were reduced by DTT. Maleimide-PEG was applied to attach at the reduced thiols. Samples were analyzed by western blot using anti-Flag antibody. (q) HEK293A cells were transfected with Flag-tagged wild type (WT), C687S, C712S or C687/712S (2CS) mutant of STAT3, and total cell lysates were heated at indicated temperature. Samples were analyzed by western blot using anti-Flag antibody. Soluble fractions were plotted versus temperatures. (r) HEK293A cells were transfected with Flag-tagged wild type (WT), C687S, C712S or C687/712S (2CS) mutant of STAT3, and cells were heated at indicated temperature. Total cell lysates were analyzed by western blot using anti-Flag antibody. Soluble fractions were plotted versus temperatures. In (b-i, k, o-r), the experiments were independently repeated at least 3 times with similar results. For gel source data, see Supplementary Figure 1.

Figure 1.

STAT3 is S-palmitoylated at evolutionarily conserved cysteine residues.

(a) Myc-STAT3 is palmitoylated in HEK293 and MDA-MB-231 cells. Streptavidin blots showed detection of STAT3 S-palmitoylation band (Palm-STAT3) after metabolic labeling with 50 μM chemical reporter of palmitoylation (Alk-C16) with or without hydroxylamine treatment. (b) Analysis of STAT3 fatty acylation using different chemical reporters of fatty acylation (Alk-C14 to Alk-C20). STAT3 fatty acylation levels (Acylated-STAT3) were analyzed by streptavidin bead pull--down, and followed by western blotting. (c) Analysis of STAT3 palmitoylation in cytoplasm and nucleus by APE assay, following fractionation. HDAC1 and α-Tublin are nuclear and cytoplasmic fraction controls, respectively. STAT3-PEG bands indicated palmitoylated STAT3. (d) Quantification of STAT3 palmitoylation percentage in cytoplasm and nucleus in APE assays using Image J. All The data are presented as mean ± s.e.m., n = 3 biologically independent samples. P value is determined by two-tailed student’s t-test. (e) Determination of the half-life of STAT3 palmitoylation turnover from pulse-chase results. Plots were fitted with one-phase exponential decay curve in GraphPad. The data are presented as mean ± s.e.m., n = 4 biologically independent samples. (f) Palmitoylation levels of Flag-STAT3 wild type (WT), C687S, C712S and C687/712S (2CS) mutant analyzed by metabolic labeling with Alk-C16, Click reaction and streptavidin bead pull-down, and followed by western blotting. Palm-STAT3 band indicated palmitoylated STAT3. In a, b, f, the experiments were independently repeated at least 3 times with similar results. For gel source data, see Supplementary Figure 1.

As JAK-kinase phosphorylation site Y705 is located near C687 and C712, we tested whether phosphorylation and palmitoylation could influence each other. We observed that IL-6 or interferon-γ (IFN-γ) stimulation markedly enhanced, and the selective JAK1/2 inhibitor ruxolitinib decreased STAT3 palmitoylation (Fig. 2a–c, Extended Data Fig. 2a). Moreover, the enhanced palmitoylation following IL-6 stimulation was attenuated by C687S mutation (Extended Data Fig. 2b). Interestingly, the phosphorylation-deficient, dominant-negative STAT3 mutant (DN-STAT3, Y705F) showed decreased palmitoylation levels compared to the WT, but the mutation did not completely abolish its palmitoylation (Fig. 2d). Taken together, these results suggest that cytokine-induced STAT3 phosphorylation can enhance, but is not required for its palmitoylation.

Figure2.

A signaling relay involving STAT3 phosphorylation and palmitoylation promotes STAT3 dimerization in response to cytokine and fatty acids.

(a) Flag-STAT3 palmitoylation levels were analyzed by APE assay and western blotting upon IL-6 stimulation with or without hydroxylamine treatment. STAT3-PEG bands indicated palmitoylated STAT3. (b) Quantification of STAT3 palmitoylation percentage from APE assays in (a), n = 3 biologically independent samples. (c) Palmitoylation and Y705 phosphorylation of endogenous STAT3 in HEK293 cells, treated with IL-6 and/or JAK inhibitor ruxolitinib. Palmitoylation of STAT3 (Palm-STAT3) is detected by chemical reporter (Alk-C16) labeling, Click reaction, followed by Streptavidin pull-down and western blotting. (d) HEK293A cells were transfected with Flag-tagged wild type (WT) or Y705F mutant. The Palmitoylation levels (Palm-STAT3) of STAT3 WT or Y705F mutant were analyzed same as in (c). (e) Co-immunoprecipitation (Co-IP) assay to detect homodimerization of STAT3 WT or palmitoylation-deficient C687/712S (2CS) mutant in HEK293A cells treated with IL-6. Whole cell lysates were analyzed by anti-Flag immunoprecipitation followed by immunoblotting using the indicated antibodies (f) Percentage of STAT3 palmitoylation in mouse lung and liver tissues fed with normal-fat diet (NFD) or high-fat diet (HFD) were analyzed by APE assay, n = 5 animals. . (g) HEK293A cells were transfected with Flag-STAT3 and treated with BSA-conjugated palmitic acid (PA) at the indicated doses. STAT3 palmitoylation levels (indicated by STAT3-PEG bands) were analyzed by the APE assay. (h) Quantification of STAT3 palmitoylation percentage in (g). n = 3 biologically independent samples. . (i) Detection of endogenous STAT3 dimerization using disuccinimidyl glutarate (DSG) crosslinking assay in HEK293A cells, treated with IFN-γ, IL-6 or BSA-conjugated palmitic acid (PA, 100μM). (j) Co-IP assay to detect homodimerization of STAT3 WT or palmitoylation-deficient C687S mutant in HEK293A cells, treated with BSA-conjugated palmitic acid (PA, 100μM). Whole cell lysates were analyzed by anti-Flag IP followed by immunoblotting using the indicated antibodies. In c-e, i, j, the experiments were independently repeated at least 3 times with similar results. For gel source data, see Supplementary Figure 1. All the data in bar graph (b, f, h) are presented as mean ± s.e.m. P value is determined by two-tailed student’s t test.

Extended Data Figure 2.

STAT3 palmitoylation promotes STAT3 nuclear translocation and synergically enhances STAT3 signaling activity with phosphorylation.

(a) HEK293A cells were pretreated with ruxolitinib (1 μM) for 0.5 hour and then labeled with 50 μM chemical probe (Alk-C16) for 2 hours with the incubation of IFNγ (1 ng/ml). Endogenous STAT3 palmitoylation was analyzed by Click reaction and streptavidin beads pull-down and subjected to western blotting. (b) HEK293A cells were transfected with Flag-STAT3 WT or Flag-STAT3 WT C687S and then labeled with 50 μM chemical probe Alk-C16 for 2 hours with the incubation of IL-6 (20 ng/ml). STAT3 palmitoylation was analyzed by western blot following Click reaction and streptavidin beads pull down. (c) HEK293A cells were transfected with Myc-STAT3-WT, Myc-STAT3-C687S or Myc-STAT3-Y705F and treated with IL-6 (20ng/ml, 1 hour). Whole cell lysates were immunoprecipitated with anti-Myc antibody, and followed with immunoblotting using indicated antibodies. (d) Confocal microscopy showing changes in subcellular localization of STAT3 WT and C687/712 mutant in HEK293A stable cell lines. Cells were stained with anti-Flag antibody (green), anti-Lamin B1 antibody (red) and DAPI (blue). Lamin B1 displays the nuclear membrane. The red line indicates the position of the Z stack. Western blot showing STAT3 WT and C687/712 mutant were expressed at comparable levels in HEK293A stable cell lines. The levels of nuclear-localized STAT3 were quantified by measuring the Flag (STAT3) fluorescence intensity in the nucleus using the confocal software to define the selected ROI (Region of Interest) based on nuclear DAPI signal. The data are presented as mean ± s.e.m., n = 225 or 300 cells in each group, respectively. P value is determined by two-tailed student’s t-test. Scale bars represent 20 μm in all images. (e) HEK293A cells were transfected with Flag-STAT3-WT, Flag-STAT3-C687/712S or Flag-STAT3-K685S with or without CBP. Whole cell lysates were analyzed by western blot using indicated antibodies. (f) HEK293A cells were co-transfected with Flag-STAT3-WT or Flag-STAT3–2CS along with GFP-STAT1. Whole cell lysates were analyzed by anti-Flag IP followed by immunoblotting using the indicated antibodies. (g) Similar analysis as panel (f) was performed in HEK293A cells transfected with Flag-STAT3 and HA-JAK1 followed by anti-HA IP. (h) HEK293A cells were co-transfected with Flag-STAT3 and Myc-STAT3 and pretreated with 5, 15-DPP (10 μM) and then treated with IL-6 (20 ng/ml, 1hour). Whole cell lysates were analyzed by anti-Flag IP followed by immunoblotting using the indicated antibodies. (i) HEK293A cells were transfected with Flag-STAT3-WT treated with 5, 15-DPP alone or along with IL-6 (20 ng/ml, 1 hour) and labeled with 50 μM chemical probe (Alk-C16). STAT3 palmitoylation was analyzed by Click reaction and streptavidin beads pull-down and subjected to western blotting. (j) The acyl chain binding pocket according to the STAT3 crystal structure (PDB 4E68), hydrophobic amino acids are shown in yellow. (k) Mutational analysis of selected residues involved in STAT3 palmitoylaiton. (l) HEK293A cells were transfected with Myc tagged and Flag tagged STAT3 as indicated. Whole cell lysates were immunoprecipitated with anti-Myc antibody, and followed with immunoblotting using indicated antibodies. (m) STAT3 null (STAT3–/–) mouse embryonic fibroblast (MEF) cells were transfected with vector control, Flag-tagged wild type (WT) or C687/712S (2CS) mutant of STAT3. Cells were labeled with 50 μM Alk-C16 probe for 2 hours with the incubation of IL-6 (20 ng/ml). Western blotting using anti-Flag antibody in the streptavidin bead pull-down samples indicates the palmitoylation levels of STAT3. (n) STAT3−/− MEF cells were co-transfected with STAT3 reporter construct (m67-luciferase reporter) and Renilla luciferase control construct, and a vector control, a Flag-tagged STAT3 wild type (WT) or C687/712S mutant. Cells were then treated with IL-6 (20 ng/ml) for 8 hours. Luciferase activity was obtained from triplicated experiments and normalized to the Renilla luciferase. Relative fold induction was plotted as shown. (o) Similar experiment as (n) was performed in HEK293A cells. (p) STAT3−/− MEF cells were transfected with vector control, Flag-tagged STAT3 wild type (WT) or C687/712S mutant. Cells were treated with IL-6 (20 ng/ml) for 8 hours. The mRNA levels of STAT3 target genes (BCL2, BCL2L1 and MMP9) were analyzed by qRT-PCR. (q) Similar experiment as (p) was performed in HEK293A cells. In (n-q), the data are presented as mean ± s.e.m., n = 3 biologically independent samples. P value is determined by two-tailed student’s t-test. In (a-c, e-i, k-m), the experiments were independently repeated at least 3 times with similar results. For gel source data, see Supplementary Figure 1.

Palmitoylation has been known to regulate protein membrane localization and trafficking[10,11]. However, we did not observe membrane localized STAT3. Interestingly, palmitoylation-deficient STAT3 2CS mutant can still be phosphorylated at Y705 at similar levels as WT upon IL-6 stimulation, but showed significantly decreased nuclear localization (Extended Data Fig. 2c, d). In addition, STAT3 2CS showed no significant change on K685 acetylation compared to WT, with or without the expression of acetyltransferase p300/CBP[12] (Extended Data Fig. 2e). Surprisingly, STAT3 2CS mutant showed markedly lower homodimerization (Fig. 2e) or STAT1–STAT3 heterodimerization, but not JAK1–STAT3 heterodimerization (Extended Data Fig. 2f, g). Blocking STAT3-SH2 dimerization using 5, 15-diphenylporphyrin (5, 15-DPP) has no effect on STAT3 palmitoylation (Extended Data Fig. 2h, i). Taken together, palmitoylation might regulate STAT3 dimerization, which is not due to alternation of its phosphorylation and acetylation. It has been shown that all SH2 domains harbor putative lipid-binding pockets[13], which could potentially accommodate fatty acyl chain binding and enhance dimerization. We generated mutants with reduced size of the hydrophobic pocket, which might limit the potential lipid binding in STAT3 SH2 domain (Extended Data Fig. 2j). Indeed, STAT3 A651V mutant showed significantly reduced palmitoylation levels and homodimerization (Extended Data Fig. 2k, l). Therefore, it is possible that palmitoylation could enhance STAT3 dimerization through engaging the lipid-binding pocket in SH2 domain.

To test the function of STAT3 palmitoylation, we re-expressed STAT3 WT and 2CS mutant in STAT3–/– mouse embryonic fibroblast (MEF) cells. Consistently, the STAT3 2CS mutant lost palmitoylation in basal and IL-6-induced conditions (Extended Data Fig. 2m), and blocked IL-6-induced transcriptional activation and expression of STAT3 target genes (BCL2, BCL2L1 and MMP9) (Extended Data Fig. 2n–q) in both MEF and HEK293A cells, functioning as a dominant negative mutant. As cytokine-stimulation can promote STAT3 palmitoylation, palmitoylation may serve as a positive feedback response for cytokine stimulation, and promote STAT3 dimerization and activation.

HFD has been shown to increase free fatty acid levels, and linked to STAT3 activation and inflammation in animals[14,15]. We found that HFD notably increased STAT3 palmitoylation and phosphorylation levels compared to normal-fat diet (NFD) in mouse lung and liver tissues (Fig. 2f, Extended Data Fig. 3a, b), indicating that HFD may enhance STAT3 activation by promoting its phosphorylation and palmitoylation. In addition, palmitic acid and stearic acid could directly activate STAT3 reporter, consistent with the finding that STAT3 can be palmitoylated and stearoylated (Extended Data Fig. 3c). Indeed, exogenous palmitic acid could increase STAT3 palmitoylation levels in a dose-dependent manner (Fig. 2g, h). Compared to IL-6, palmitic acid induced relatively moderate STAT3 nuclear localization (Extended Data Fig. 3d), and delayed transcriptional activation in STAT3 reporter and target gene expression analyses (Extended Data Fig. 3e, f). Similar to IFN-γ and IL6, palmitic acid alone could induce STAT3 dimerization in disuccinimidyl glutarate (DSG) crosslinking assay (Fig. 2i). Depalmitoylation by hydroxylamine treatment or C687S mutation could almost abolish STAT3 homodimerization in response to palmitic acid (Extended Data Fig. 3g,–h, and Fig. 2j), suggesting that palmitoylation contributed to palmitic acid-induced STAT3 dimerization. Interestingly, combining IL-6 and palmitic acid further increased STAT3 transcriptional activity compared to IL-6 or palmitic acid treatment alone in STAT3–/– MEF cells expressing STAT3 WT. However, such effects were lost in cells expressing the STAT3 2CS mutant (Extended Data Fig. 3i, j). Taken together, palmitic acid could activate STAT3 and synergize with cytokines to enhance STAT3 transcriptional activity through palmitoylation. HFD could also induce cytokine secretion[16], therefore, it is possible that a positive feedback loop composed of STAT3 phosphorylation and palmitoylation may lead to strong activation of STAT3 in response to fatty acid and cytokine stimulation.

Extended Data Figure 3.

Palmitic acid promotes STAT3 nuclear localization and activity through induction of dimerization in synergy with cytokine stimulation.

(a) C57BL/6 mice were fed with normal fat diet (10 kcal% fat, NFD) or high fat diet (60 kcal% fat, HFD) for 2 weeks. STAT3 palmitoylation and p-Y705 levels were analyzed by APE and western blot in mouse lung tissues. (b) Similar experiment as (a) was performed in liver tissue. (c) U3A cells were treated with various fatty acids at 100 μM or IL-6 (20 ng/ml) for 6 hours. Luciferase activity was measured and relative fold induction from triplicate experiments. (d) Confocal microscopy showing changes in subcellular localization of STAT3 in HEK293A cells treated with palmitic acid (PA, 50 μM, overnight) and/or IL-6 (20 ng/ml, 1 hour). Cells were stained with anti-STAT3 antibody (red) and DAPI (blue). The yellow line indicates the position of the Z stack. The levels of nuclear-localized STAT3 were quantified by measuring the STAT3 fluorescence intensity in the nucleus using the confocal software to define the selected ROI (Region of Interest) based on nuclear DAPI signal. Scale bars represent 20 μm in all images. The data are presented as mean ± s.e.m., n = 217, 246, 261, 288 cells in each group, respectively. P value is determined by two-tailed student’s t-test. (e) HEK293A cells were transfected with STAT3-luciferase reporter with Renilla control constructs and treated with 20 ng/ml IL-6 or 100 μM PA for indicated time points. Luciferase activity was measured and relative fold induction from triplicate experiments was plotted with mean ± s.e.m., n = 3 biologically independent samples. (f) HEK293A cells were treated with 20 ng/ml IL-6 or 100 μM palmitic acid, the expression of indicated genes was quantified with qPCR. (g) HEK293A cells were transfected with Flag-STAT3 along with or without Myc-STAT3 WT and treated with 100 μM palmitic acid for 2 hours. Whole cell lysates were analyzed by anti-Flag IP followed by immunoblotting using the indicated antibodies. (h) Similar experiments as in (g) were performed in HEK293A cells with or without NH2OH treatment. (i) HEK293A cells were co-transfected with m67-luciferase reporter and Renilla reporter constructs along with vector control, Flag-tagged STAT3 wild type (WT) or C687/712S mutant. After 48 hours, cells were treated with palmitic acid (100 μM, 8 hours) and/or IL-6 (20 ng/ml, 8 hours). Normalized luciferase activity from triplicate experiments was plotted as shown. (j) HEK293A cells were transfected with vector control, Flag-tagged STAT3 wild type (WT) or C687/712S mutant. Cells were treated with IL-6 (20 ng/ml) for 8 hours. The mRNA levels of STAT3 target gene BCL2 were analyzed by qRT-PCR. Relative fold change (normalized to GAPDH) was shown from triplicate experiments. In (c, f, i, j), the data are presented as mean ± s.e.m., n = 3 biologically independent samples. P value is determined by two-tailed student’s t-test. In (a, b, g, h), the experiments were independently repeated at least 3 times with similar results. For gel source data, see Supplementary Figure 1.

In screens for potential palmitoyl acyltransferases (PATs) regulating STAT3, we identified ZDHHC5, 18 and 19 as potential primary hits[17]. In reconfirmation assays, ZDHHC19 showed consistent STAT3 palmitoylation activities and binding to STAT3 (Extended Data Fig. 4a–f). We generated the catalytically inactive mutant of ZDHHC19 (C142S) by mutating active site cysteine residue to serine[18] (Extended Data Fig. 4g, h), which failed to promote STAT3 palmitoylation under basal conditions or with IL-6 stimulation (Fig. 3a–c). Consistently, ectopic expressing ZDHHC19 WT, but not C142S mutant, could significantly induce the activation of STAT3 reporter (Extended Data Fig. 4i). Furthermore, ZDHHC19 knockdown by shRNA blocked STAT3 palmitoylation and its target gene expression, which can be rescued by re-expressing shRNA-resistant mouse WT ZDHHC19, but not C142 mutant (Fig. 3d, Extended Data Fig. 4j). Loss of ZDHHC19 also abolished STAT3 target gene expression induced by IL-6, palmitic acid, or their combination (Extended Data Fig. 4k–m). Taken together, our results suggest that ZDHHC19 is the major palmitoylating enzyme regulating STAT3 palmitoylation, and is required for palmitic acid-induced STAT3 activation.

Extended Data Figure 4.

ZDHHC19 mediates STAT3 palmitoylaiton and Ras is not involved in ZDHHC19 mediated STAT3 signaling.

(a-b) HEK293A cells were co-transfected with construct encoding Flag-STAT3 and HA-ZDHHCs. After 48 hours, cells were labeled for 4 hours with 50 μM palmitoylation probe Alk-C16. Cells lysates were reacted with azide-biotin and subjected to Streptavidin beads pull-down. STAT3 palmitoylation was shown by western blot. (c) HEK293A cells were transfected with Flag-STAT3 alone or along with HA-ZDHHC19, HA-ZDHHC 5 and HA-ZDHHC 18 and labeled with 50 μM probe Alk-C16 for 4 hours following analysis of STAT3 palmitoylation. (d) HEK293A cells were transfected with Flag-STAT3 alone or along with HA-ZDHHC19, HA-ZDHHC 5 and HA-ZDHHC 18. Whole cell lysates were analyzed by anti-Flag IP followed by immunoblotting using the indicated antibodies. (e) HEK293A cell were transfected as in panel (d). Total cell extracts were analyzed by western blot after IP with anti-HA antibody. (f) HEK293A cells were transfected with Flag-STAT3-WT or Flag-STAT3-K685S with or without ZDHHC19. Palmitoylation of STAT3 was analyzed as in panel (c). (g-h) HEK293A cells were transfected with HA-ZDHHC19-WT or inactive mutant C142S (CS) and labeled with 50 μM Alk-C16 or 16C-BYA. Cell lysates were subjected to streptavidin beads pull-down. STAT3 palmitoylation was showed by western blot. (i) HEK293A cells were transfected with STAT3-luciferase reporter and Renilla control, co-transfected with ZDHHC19-WT and ZDHHC19-C142S as shown treated with IL-6 (20 ng/ml) and/or palmitic acid (PA, 100 μM) for 8 hours. Luciferase activity was measured and relative fold induction from triplicate experiments. (j) HEK293 cells were co-transfected with ZDHHC19-shRNA and ZDHHC19-WT or ZDHHC19- C142S, the expression of indicated genes was quantified with qRT-PCR. (k-m) KNS-62 control cells and ZDHHC19 KO stable cells were treated with IL-6 (20 ng/ml) and/or palmitic acid (100 μM) for 8 hours. The mRNA levels of STAT3 target genes (BCL2, BCL2L1 and MMP9) were analyzed by qRT-PCR. (n) Schematic representation of Grb2 wild type and SH3-domain truncation mutants and alignment of ZDHHC19 SH3 binding motif sequences. (o) HEK293A cells were co-transfected with Myc-tagged ZDHHC19 and GFP-tagged Grb2 WT, N terminal SH3 deletion (ΔN-SH3) or C terminal SH3 deletion (ΔC-SH3). Whole cell lysates were subjected to immunoprecipitation using anti-Myc antibody, followed by western blotting using the indicated antibodies. (p) HEK293A cells were co-transfected with Myc-tagged ZDHHC19 WT or P18A mutant, Flag-tagged STAT3 and GFP-tagged Grb2 following Co-IP assay. (q) HEK293A cells were treated with IL-6 (20 ng/ml) and analyzed by anti-STAT3 Co-IP assay followed by western blotting using the indicated antibodies. (r-s) HA tagged ZDHHC19 was co-transfected with HA tagged HRAS (r) or NRAS (s) into HEK293A cells. After 24 hours, cells were incubated with fresh media containing 10% dialyzed FBS for 2 hours and subsequently labeled with 50 μM Alk-16C probe for 4 hours. Cells lysates were reacted with biotin-azide and subjected to SDS-PAGE. Streptavidin blot was used to detect HRAS (r) or NRAS (s) palmitoylation as previously described. Comparable protein loading was confirmed by anti-HA and β-Actin western blotting. (t-u) Depalmitoylase inhibitor Palmostatin B (10 μM) treatment or overexpression of depalmitoylase ABHD17A has no effect on the expression of STAT3 target genes (BCL2, BCL2L1 and MMP9) in HEK293A cells expressing oncogenic HRAS G12V (t) or NRAS G12V (u). (v) Treatment with different concentrations of Palmostatin B (1 μM, 5 μM and 10 μM) has no effect on the expression of STAT3 target genes (BCL2, BCL2L1 and MMP9) in melanoma SK-MEL-2 cells containing NRAS Q61R mutation. In (i-m, t-v), the data are presented as mean ± s.e.m., n = 3 biologically independent samples. P value is determined by two-tailed student’s t-test. In (a-h, o-s), the experiments were independently repeated at least 3 times with similar results. For gel source data, see Supplementary Figure 1.

Figure 3.

ZDHHC19 mediates STAT3 palmitoylation.

(a) Palmitoylation of Flag-STAT3 was analyzed by APE assay and western blotting, with co-transfection of vector control, ZDHHC19 wild type (WT), or the catalytically inactive C142S mutant. STAT3-PEG bands indicated palmitoylated STAT3. (b) Quantification of STAT3 palmitoylation percentage in (a). The data are presented as mean ± s.e.m., n = 3 biologically independent samples. P value is determined by two-tailed student’s t-test. (c) Palmitoylation of Flag-STAT3 in cells co-transfected with vector control, ZDHHC19 WT or C142S mutant, then treated with or without IL-6. Palm-STAT3 bands indicated palmitoylated STAT3, detected by Alk-C16 labeling, Click reaction, followed streptavidin pulldown and western blotting. (d) Palmitoylation of endogenous STAT3 in HEK293 cells stably transfected with shRNA targeting human ZDHHC19, then re-expressed mouse HA-ZDHHC19 wild type (WT) or C142S mutant. STAT3 palmitoylation (Palm-STAT3) was analyzed the same as in (c). (e) Complex formation of GFP-Grb2, Myc-ZDHHC19 and Flag-STAT3 in HEK293 cells upon IL-6 stimulation and/or JAK inhibitor ruxolitinib treatment. Whole cell lysates were subjected to Co-IP using anti-Flag antibody, following western blotting using indicated antibodies. (f) Palmitoylation of endogenous STAT3 in HEK293A cells transfected with shRNA control or shRNA targeting Grb2 (sh2 or sh4), and cells were treated with or without IL-6. STAT3 palmitoylation (Palm-STAT3) was analyzed the same as in (c). In c-f, the experiments were independently repeated at least 3 times with similar results. For gel source data, see Supplementary Figure 1.

We explored the mechanisms of how cytokine stimulation promotes STAT3 palmitoylation, and found that ZDHHC19 contains a highly conserved SH3 binding site around Pro18 and Pro20. ZDHHC19 could interact with the SH3 domain of Grb2, a key adaptor protein, which assembles signaling complexes in JAK-STAT signaling (Extended Data Fig. 4n, o)[19]. ZDHHC19 SH3 binding motif mutant (P18A) showed significantly reduced interaction with Grb2 and STAT3 (Extended Data Fig. 4p), suggesting that Grb2 may recruit ZDHHC19 to STAT3 through its SH3 domain. Furthermore, we found that IL-6 stimulation significantly increased (Extended Data Fig. 4q), while JAK inhibitor ruxolitinib blocked the complex formation of STAT3, Grb2 and ZDHHC19 (Fig. 3e). Consistently, Grb2 knockdown substantially blocked cytokine stimulation-induced STAT3 palmitoylation (Fig. 3f). Taken together, upon cytokine stimulation, Grb2 may serve as a critical adaptor protein, recruiting ZDHHC19 through its SH3 domain to the membrane complex, allowing efficient ZDHHC19–STAT3 association and STAT3 palmitoylation. Although Grb2 is known to be involved in Ras signaling[20], ZDHHC19 does not regulate HRAS and NRAS palmitoylation[21]. Oncogenic HRAS (G12V) and NRAS (G12V or Q61R), or their palmitoylation did not affect STAT3 transcriptional activities either (Extended Data Fig. 4r–v).

To explore the roles of ZDHHC19 and ZDHHC19-mediated STAT3 palmitoylation in pathogenesis, we found that ZDHHC19 is located in 3q26–29, a region frequently amplified in multiple human cancers, especially in lung squamous cell carcinoma (LSCC) in bioinformatics analyses (Extended Data Fig. 5a)[22]. Among the 56 genes in the focal 3q29 amplicon, ZDHHC19 expression has the strongest correlation with STAT3 target genes (BCL2 and BCL2L1) expression (Extended Data Fig. 5b–d). In addition, ZDHHC19 showed co-occurrence instead of mutual exclusivity with known oncogenes (PIK3CA, PRKCI and WWTR1 etc.) and tumor suppressors (XRN1, AIM2 and PYHIN1) in LSCC (Extended Data Fig. 5e–g). Consistently, ZDHHC19 expression is significantly higher in LSCCs, compared to adenocarcinoma (Extended Data Fig. 5h, i). In addition, high ZDHHC19 expression correlated significantly with poor patient survival (Extended Data Fig. 5j, k). The expression of STAT3 target genes (BCL2 and BCL2L1) is positively and significantly correlated with ZDHHC19 expression levels in two LSCC patient datasets (Extended Data Fig. 5l–o). We further carried out immunohistochemistry (IHC) studies of 131 tumor samples from a cohort of 85 LSCC patients, and confirmed that high ZDHHC19 protein levels showed significant correlation with nuclear STAT3 staining (Figure 4a, Extended Data Fig. 6, Supplementary Table 1).

Extended Data Figure 5.

ZDHHC19 gene is amplified and highly expressed in lung squamous cell carcinoma and correlated with poor clinical outcomes.

(a) Alteration frequency and oncoprint diagram of ZDHHC19 gene alteration summary in cancer patients (11,413) obtained from cBioPortal (TCGA). (b) The regions of the genome that are significantly amplified across a set of lung squamous cell carcinoma samples from TCGA were identified using GISTIC2 from Gene Pattern (Broad Institute). ZDHHC19 was confirmed as one of the genes in the amplified 3q29 region. (c) Genes in amplified 3q29 region are correlated with STAT3 target genes BCL2L1 and BCL2 cross two Geo DataSets (GSE28571 (n = 100 samples) and GSE73403 (n = 69 samples)). 28 genes are analyzed. P value is calculated by two-tailed Pearson correlation. (d) Venny diagram of genes in amplified 3q29 region significantly correlated with STAT3 target genes BCL2L1 and BCL2 cross two Geo DataSets (GSE28571 (n = 100 samples) and GSE73403(n = 69 samples)). 36 genes in GSE28571 and 30 genes in GSE73403 are analyzed. P value is determined by two-tailed Pearson correlation. (e) The plots of p/q value of oncogenes showing co-occurrence or mutual exclusivity with ZDHHC19 in lung squamous cell carcinoma from TCGA (n = 501 samples). 231 genes (162 co-occurrence and 69 mutual exclusivity) are analyzed. p and q value are determined by Fisher’s exact test and FDR-test. (f) The plots of p/q value of tumor suppressors showing co-occurrence or mutual exclusivity with ZDHHC19 in lung squamous cell carcinoma from TCGA (n = 501 samples). 180 genes (115 co-occurrence and 65 mutual exclusivity) are analyzed. p and q value are determined by Fisher’s exact test and FDR-test.. (g) Gene list of oncogenes or tumor suppressors showing significant co-occurrence or mutual exclusivity with ZDHHC19 in lung squamous cell carcinoma (n = 501 samples). p and q value are determined by Fisher’s exact test and FDR-test. (h) Comparison of ZDHHC19 expression in lung squamous cell carcinoma (Lung Squ) and lung adenocarcinoma (Lung Adeno) patient samples obtained from cBioPortal TCGA dataset (n = 1097 samples). P values were determined by two-tailed student’s t-test. (i) The expression level of ZDHHC19 in human lung cancer cell lines was grouped in 4 subtypes: Lung squamous cell carcinoma cell lines (Lung squ) (n = 29 samples); Lung large cell carcinoma cell lines (Large lung) (n = 14 samples); lung adenocarcinoma cell lines (Lung adeno) (n = 53 samples); and, Lung small cell lung carcinoma cell lines (Small lung) (n = 53 samples). Data were obtained from Cancer Cell Line Encyclopedia database (CCLE, www.broadinstitute.org/ccle) and graphed by GraphPad Prism. P value is determined by two-tailed student’s t-test. (j) Kaplan-Meier curves of 68 lung squamous cell carcinoma patients after stratification by the median level of ZDHHC19 were used for depicting survival time. Patient data were derived from GSE 73403 and analyzed by GraphPad Prism. P values is determined by Log-rank (Mantel-Cox) test. (k) Kaplan-Meier curves of lung squamous cell cancer patients after stratification by the median level of ZDHHC19 were used for depicting Survival time. Patient data were analyzed by http://kmplot.com. P values is determined by Log-rank (Mantel-Cox) test. (l) Pearson analysis of gene expression data from lung cancer patients (GSE 73403 dataset (n = 69 samples)) was used for depicting the correlation between BCL2 and ZDHHC19. Data were graphed and analyzed by GraphPad Prism software. (m) Similar analysis as panel (l) of gene expression data from lung cancer patients (GSE 73403 dataset (n = 69 samples)) was used for depicting the correlation between BCL2L1 and ZDHHC19. (n) Pearson analysis of gene expression data from lung cancer patients (GSE 28571 dataset (n = 100 samples)) was used for depicting the correlation between BCL2 and ZDHHC19. (o)Similar analysis of gene expression data from lung cancer patients (GSE 28571 dataset (n = 100 samples)) was used for depicting the correlation between BCL2L1 and ZDHHC19. 100 samples were graphed and analyzed by GraphPad Prism software. In (l-o), center line indicates a line of best fit through the data of two variables in Pearson correlation coefficient model, and the dotted lines indicates 95% confidence band.

Figure 4.

ZDHHC19 is amplified in lung squamous cell carcinoma (LSCC) and promotes tumorigenesis in vitro and in vivo.

(a) Representative images of IHC staining of STAT3 and ZDHHC19 in paraffin-embedded LSCC samples, n = 131 biologically independent samples from 85 patients (detailed information in Supplementary Table 1). Analysis was done once for each sample. (b) Representative images of tumorsphere formation of KNS62 cells stably infected with shRNA control or shRNA targeting ZDHHC19 (shZDHHC19) under limiting dilution conditions. Cells were treated with vehicle control or palmitic acid (PA, 100μM). The numbers of wells with tumorsphere/total wells were labeled in the panels in indicated conditions. (c) Quantification of KNS62 stem cell frequency in (b). The center line is the log-active cell fraction and the dotted lines give the 95% confidence interval. P values are determined by two-tailed student’s t-test. (d) Xenograft tumor growth of HCC95 wild type (WT) control or with CRISPR/Cas9 knockout of ZDHHC19. Animals were fed with normal-fat diet (NFD) or high-fat diet (HFD) for 4 weeks, n = 10 biologically independent tumors per group. (e) Quantification of the percentage of nuclear-localized STAT3 in tumor samples in (d) by IHC staining. n = 10 biologically independent tumor samples per group. (f) Tumor growth curve of LSCC patient-derived xenograft (PDX) model under NFD or HFD. n = 8 (in NFD) or 10 (in HFD) biologically independent tumors. (g) Quantification of tumorsphere numbers from PDX tumor cells infected with shRNA control (pLKO) or shRNA-targeting ZDHHC19 (shZDHHC19), and then treated with vehicle control or palmitic acid (PA, 100 μM), n = 6 biologically independent samples. Data in d-g, are presented as mean ± s.e.m., P value is determined by two-tailed students’s t-test in (e and g) or two-way ANOVA followed by Bonferroni’s test for time courses in (d and f).

Extended Data Figure 6.

ZDHHC19 expression correlates with STAT3 nuclear localization in lung squamous cell carcinoma.

(a) Immunohistochemistry staining showing correlation of ZDHHC19 expression with STAT3 nuclear localization. 131 biologically independent samples were analyzed. (b) Statistical analysis of Pearson correlation between immunohistochemistry staining scores of ZDHHC19 expression and STAT3 nuclear localization. (c) Summary of lung SCC patient tissue samples and the scores of ZDHHC19 expression and STAT3 nuclear localization.

Consistently, in LSCC cell line HCC95, IL-6 stimulation markedly increased endogenous STAT3 palmitoylation and ZDHHC19–STAT3 interaction, which could be blocked by JAK inhibitor ruxolitinib (Extended Data Fig. 7a, b). Knocking down ZDHHC19 in multiple LSCC cell lines (HCC95, KNS-62 and/or SK-MES) reduced STAT3 palmitoylation and its nuclear localization (Extended Data Fig. 7c–f). Furthermore, we found that knockdown or knockout of ZDHHC19 markedly inhibited cell proliferation, colony formation, and cell migration of LSCC cells (Extended Data Fig. 7g–o). Interestingly, ectopic expression of the constitutive active form of STAT3 (STAT3C)[23] could rescue the inhibitory effects of ZDHHC19 knockdown in colony formation of KNS62 cells, suggesting that ZDHHC19 regulates LSCC cell clonal growth through STAT3 (Extended Data Fig. 7p).

Extended Data Figure 7.

ZDHHC19 mediated STAT3 palmitoylaiton facilitates LSCC tumor cell growth, colony formation and migration in vitro through induction of STAT3 activity.

(a) HCC95 lung squamous cell carcinoma cells were pretreated with ruxolitinib (1 μM) for 30 minutes and then labeled with 50 μM probe Alk-C16 for 2 hours with the incubation of IL-6 (20 ng/ml). Endogenous STAT3 palmitoylation was analyzed by Click reaction and streptavidin beads pulldown, and followed by western blotting. (b) HCC95 cells were treated as in panel (a). Whole cell lysates were analyzed by anti-STAT3 IP followed by western blotting using the indicated antibodies. (c) Lung squamous cell carcinoma cell lines HCC95 and KNS-62 were transfected with shRNA control or shZDHHC19 (3 and 5) and labeled with 50 μM Alk-C16 probe for 4 hours, Cell lysates were reacted with biotin-azide and precipitated with streptavidin beads. Endogenous STAT3 palmitoylation was determined by western blot. (d) SK-MES-1 cells were transfected with ZDHHC19-shRNA and labeled with 50 μM palmitoylation probe Alk-C16 for 4hours. Cells lysates were reacted with azide-biotin and subjected to streptavidin beads pull-down. STAT3 palmitoylation was shown by western blot. (e) HCC95 shRNA control cells and Sh3 ZDHHC19 stable cells were pretreated with ruxolitinib (1 μM) for 30 minutes and then labeled with 50 μM probe Alk-C16 for 2 hours with the incubation of IL-6 (20 ng/ml). Endogenous STAT3 palmitoylation was determined by western blot. (f) Confocal immunofluorescence imaging showed STAT3 localization in HCC95 Sh3 ZDHHC19 stable cells and vector control cells. Cells were stained with anti-STAT3 antibody (green) and DAPI (blue). The levels of nuclear-localized STAT3 were quantified by measuring the STAT3 fluorescence intensity in the nucleus using the confocal software to define the selected ROI (Region of Interest) based on nuclear DAPI signal. The scale bar represents 20 μm. The data are presented as mean ± s.e.m., n = 275, 320 cells. P value is determined by two-tailed student’s t-test. (g) Cell proliferation was determined in HCC95 Sh3 ZDHHC19 stable cells or shRNA control cells. The data are presented as mean ± s.e.m., n = 3 biologically independent samples. P value is determined by two-way ANOVA followed by Bonferroni’s test. (h) Cell proliferation was determined in HCC95 cells with ZDHHC19 CRISPR knockcout or control. The data are presented as mean ± s.e.m., n = 6 biologically independent samples. P value is determined by two-way ANOVA followed by Bonferroni’s test. (i) Colony formation of HCC95 cells with ZDHHC19 CRISPR knockcout or control. The colony number was quantified. The data are presented as mean ± s.e.m., n = 5 biologically independent samples. P value is determined by two-tailed student’s t-test.. (j) Cell proliferation showing KNS62 cells with ZDHHC19 knockdown or control shRNA. The data are presented as mean ± s.e.m., n = 12 biologically independent samples. P value is determined by two-way ANOVA followed by Bonferroni’s test. (k) Cell proliferation was determined in KNS62 cells with ZDHHC19 CRISPR knockcout or control. The data are presented as mean ± s.e.m., n = 6 biologically independent samples. P value is determined by two-way ANOVA followed by Bonferroni’s test. (l) Colony formation of HCC95 cells with ZDHHC19 CRISPR knockcout or control. The colony number was quantified. The data are presented as mean ± s.e.m., n = 6 biologically independent samples. P value is determined by two-tailed student’s t-test. (m) Cell proliferation was determined in SK-MES-1 cells with ZDHHC19 CRISPR knockcout or control. The data are presented as mean ± s.e.m., n = 12 biologically independent samples. P value is determined by two-way ANOVA followed by Bonferroni’s test. (n) Colony formation of SK-MES-1 cells with ZDHHC19 CRISPR knockcout or control. The colony number was quantified. The data are presented as mean ± s.e.m., n = 6 biologically independent samples. P value is determined by two-tailed student’s t-test.. (o) The migration ability of HCC95 Sh3 ZDHHC19 stable cells or shRNA control cells was measured using the transwell migration assay. Numbers of invading cells were quantified. The data are presented as mean ± s.e.m., n = 5 biologically independent samples. P value is determined by two-tailed student’s t-test.(p) Crystal staining showing cell growth of KNS62 ZDHHC19 knockdown or control stable cells transfected with STAT3C or vehicle control. In (a-e, p), the experiments were independently repeated at least 3 times with similar results. For gel source data, see Supplementary Figure 1.

LSCC cells can form tumorspheres when cultured at three-dimensional conditions, which enrich cancer stem-like cells and recapitulate many aspects of tumorigenesis ex vivo[24]. Consistently, we observed tumorsphere formation in multiple LSCC cell lines, with high expression of “stemness” markers (SOX2, OCT4, NANOG, ALDHA1 and CD133) (Extended Data Fig. 8a–d). Interestingly, palmitic acid significantly enhanced tumorsphere formation of KNS62 cells (sphere number, size, and the stem cell frequency) in limiting dilution assays, suggesting that fatty acid could promote LSCC “stemness” (Fig. 4b, c). In addition, loss of STAT3 or ZDHHC19 significantly blocked tumorsphere formation under basal conditions or with palmitic acid stimulation (Extended Data Fig. 8e–h). Interestingly, ectopic expression of STAT3C could rescue the inhibitory effects of ZDHHC19 knockdown in tumorsphere formation (Extended Data Fig. 8i), suggesting that ZDHHC19–STAT3 signaling axis plays an important role in maintaining LSCC “stemness”.

Extended Data Figure 8.

STAT3 palmitoylaiton through ZDHHC19 involves in maintaining the cancer stem cell niche.

(a) Photomicrographs of HCC95, KNS62, and HCC827 parental adherent cells (left) and tumorspheres (right) in low-adherence culture. The experiment was independently repeated at least 3 times with similar results. (b) qRT-PCR analysis of the expression level of stem cell markers in tumorsphere of HCC95 cells compared to parental HCC95 cells. Fold change was normalized to 18S rRNA. The data are presented as mean ± s.e.m., n = 3 biologically independent samples. P value is determined by two-tailed student’s t-test. (c-d) Similar experiments were performed in KNS62 and HCC827 cells. The data are presented as mean ± s.e.m., n = 3 biologically independent samples. P value is determined by two-tailed student’s t-test. (e) HCC95 empty vector control cells and STAT3 or ZDHHC19 knock out stable cells were cultured in low-attachment plates with 25 μM palmitic acid for 7 days. Phase contrast photomicrographs showing tumorsphere formation. Numbers of spheres were counted from 5 randomly selected fields. The data are presented as mean ± s.e.m., n = 5 biologically independent samples. P value is determined by two-tailed student’s t-test. (f) KNS62 empty vector control cells and ZDHHC19 shRNA knockdown stable cells were cultured in low-attachment plates with 25 μM palmitic acid for 5 days. Fluorescent (calcein AM staining) and phase contrast photomicrographs showing tumorsphere formation. The data are presented as mean ± s.e.m., n = 12 biologically independent samples. P value is determined by two-tailed student’s t-test. (g) SK-MES-1 empty vector control cells and ZDHHC19 shRNA knockdown stable cells were cultured in low-attachment plates with 25 μM palmitic acid for 5 days. Fluorescent (calcein AM staining) and phase contrast photomicrographs showing tumorsphere formation. The data are presented as mean ± s.e.m., n = 12 biologically independent samples. P value is determined by two-tailed student’ t-test. (h) KNS62 empty vector control cells and ZDHHC19 shRNA knockdown stable cells were cultured in low-attachment plates with 25 μM palmitic acid for 5 days. The cells were trypsinized and seeded in plate again. Fluorescent (calcein AM staining) photomicrographs showing secondary tumorsphere formation after 5 days. The data are presented as mean ± s.e.m., n = 4 biologically independent samples. P value is determined by two-tailed student’s t-test. (i) KNS62 empty vector control cells and ZDHHC19 shRNA knockdown stable cells transfected with vehicle control or constitutive active STAT3C were cultured in low-attachment plates for 5 days. Fluorescent (calcein AM staining) photomicrographs showing tumorsphere formation. Tumorsphere numbers were quantified. The data are presented as mean ± s.e.m., n = 4 biologically independent samples. P value is determined by two-tailed student’s t-test.

To validate the roles of ZDHHC19 in LSCC in vivo, we established a xenograft model of HCC95 (control or CRISPR/Cas9-mediated ZDHHC19 knockout) in mice, fed with NFD or HFD. HFD promoted body weight gain, and accelerated tumor growth of control HCC95 cells. ZDHHC19 knockout significantly blocked HFD-induced tumor growth (Fig.4d, and Extended Data Fig. 9a–c), and decreased HFD-induced STAT3 palmitoylation, nuclear localization and target genes expression (BCL2, BCL2L1 and MMP9), and reduced Ki67-positive proliferative tumor cells (Extended Data Fig. 9d–g, Fig. 4e). Similar results were obtained in xenograft model with shRNA-knockdown of ZDHHC19 (Extended Data Fig. 9h–i). Furthermore, we tested a human LSCC patient-derived xenograft (PDX) model with high ZDHHC19 expression (PDX7 in Extended Data Fig. 10a, Supplementary Table 2). Consistently, HFD significantly promoted PDX tumor growth (Fig.4f, Extended Data Fig. 10b, c), correlated with the upregulation of STAT3 target gene expression, increased STAT3 palmitoylation and nuclear localization (Extended Data Fig. 10d–f). We isolated primary cells from the PDX tumor, and found that palmitic acid could markedly enhance tumorsphere formation in primary LSCC cells, which could be significantly inhibited by ZDHHC19 knockdown (Fig.4g, Extended Data Fig. 10g, h). Taken together, our data suggest that ZDHHC19-mediated STAT3 palmitoylation plays an important role in LSCC and HFD-related tumorigenesis in vivo.

Extended Data Figure 9.

High-fat diet induced tumor growth of lung squamouse cell carcinoma in vivo is dependent on ZDHHC19 mediated STAT3 palmitoylaiton.

(a) Representative image of the xenograft tumors isolated from HCC95 empty vector control and ZDHHC19 knock out xenografts as indicated. (b) Weight of the tumors was measured and plotted. The data are presented as mean ± s.e.m., n = 10 biologically independent samples. P value is determined by two-tailed student’s t-test. (c) Body weight change of mouse fed with normal or high fat diet. Data were normalized to original weight. The data are presented as mean ± s.e.m., n = 5 mice. (d) STAT3 palmitoylation were analyzed by APE and western blotting in human-derived cell tumor tissues. The palmitoylated STAT3 were quantified using Image J. The data are presented as mean ± s.e.m., n = 3 biologically independent samples. P value is determined by two-tailed student’s t-test. (e) Representative images of immunohistochemical (IHC) staining of Ki-67 in paraffin-embedded xenograft tumor tissues harvested from the indicated groups. The scale bar represents 200μm. Numbers of Ki-67 positive cells were counted from 5 randomly selected fields. The data are presented as mean ± s.e.m., n = 10 biologically independent samples. P value is determined by two-tailed student’s t-test. (f) Representative images of immunohistochemical (IHC) staining of STAT3. (g) qRT-PCR analysis of the expression level of STAT3 target genes (BCL2, BCL2L1 and MMP9) was performed in tumor tissues harvested at the end of the experiment. Fold change was normalized to 18S rRNA. The data are presented as mean ± s.e.m., n = 10 biologically independent samples. P value is determined by two-tailed student’s t-test. (h) HCC95 shRNA control and Sh3 ZDHHC19-stable cell lines were injected into mouse. Representative image of the xenograft tumors isolated from the indicated groups. (i) Tumor growth was monitored and readout by the tumor volume. The data are presented as mean ± s.e.m., Control n = 10 biologically independent samples, shZDHHC19 n = 9 biologically independent samples. P value is determined by two-way ANOVA followed by Bonferroni’s test. (j) Weight of the tumors was measured, and the data are presented as mean ± s.e.m., Control n = 10 biologically independent samples, shZDHHC19 n = 9 biologically independent samples. P value is determined by two-tailed student’s t-test. (k) qRT-PCR analysis of the expression level of STAT3 target genes in HCC95 shRNA control or Sh3 ZDHHC19 stable xenografts. The data are presented as mean ± s.e.m., Control n = 10 biologically independent samples, shZDHHC19 n = 9 biologically independent samples. P value is determined by two-tailed student’s t-test. (l) Representative images of immunohistochemical (IHC) staining of STAT3 and Ki-67 in paraffin-embedded xenograft tumor tissues harvested from the indicated groups. Ki-67 positive cells were quantified as in panel (e). The data are presented as mean ± s.e.m., n = 5 biologically independent samples. P value is determined by two-tailed student’s t-test.

Extended Data Figure 10.

High fat diet facilitates tumor growth in lung squamous cell carcinoma PDX model.

(a) qRT-PCR analysis of the expression level of ZDHHC19 was performed in a series of patient-derived xenograft (PDX) tumor tissues (PDX1-PDX7). Fold change was normalized to 18S rRNA. The data are presented as mean ± s.e.m., n = 3 technical replicates. (b) Representative image of the xenograft tumors isolated from PDX mouse with normal fat diet (NFD) or high fat diet (HFD). (c) Weight of the tumors was measured and the data are presented as mean ± s.e.m., NFD n = 8 biologically independent samples, HFD n = 10 independenlly samples. P value is determined by two-tailed student’s t-test. (d) STAT3 palmitoylation was analyzed by APE and western blotting in PDX tumor tissues. The palmitoylated STAT3 was quantified using Image J. The data are presented as mean ± s.e.m., n = 5 biologically independent samples. P value is determined by two-tailed student’s t-test. (e) qRT-PCR analysis of the expression level of STAT3 target genes (BCL2, BCL2L1 and MMP9) was performed in tumor tissues harvested at the end of the experiment. Fold change was normalized to 18S rRNA. The data are presented as mean ± s.e.m., NFD n = 8 biologically independent samples. HFD n = 10 biologically independent samples. P value is determined by two-tailed student’s t-test. (f) Representative images of immunohistochemical (IHC) staining of STAT3 and ZDHHC19 in paraffin-embedded PDX tumor tissues harvested from the indicated groups. (g) Transcriptional level of ZDHHC19 was shown the knockdown efficiency in primary PDX cells by qRT-PCR. The data are presented as mean ± s.e.m., n = 3 biologically independent samples. P value is determined by two-tailed student’s t-test. (h) Representative images of tumor sphere formation using primary tumor cells isolated from lung squamous cell carcinoma PDX tumor tissues. (i) Graphic scheme showing fatty acids and amplified ZDHHC19 promote STAT3 activation through S-Palmitoylation. In (f, h), the experiments were independently repeated at least 3 times with similar results.

In summary, we demonstrate that S-palmitoylation of STAT3 is critical for its homodimerization, nuclear localization, and transcriptional activity. STAT3 palmitoylation and phosphorylation regulate STAT3 through a positive feedback signaling network, contributing to fatty acid-induced activation of STAT3 and inflammation. HFD and obesity have been linked to multiple cancers[25]. Indeed, smoking and high intakes of saturated fat have been shown as significant risk factors for human LSCC[26]. HFD enhances stemness and tumorigenicity of intestinal progenitors[27], and the fatty acid translocase CD36 promotes tumor metastasis by regulating fatty acid uptake into cells[28]. Our findings of the fatty acid–ZDHHC19–STAT3 signaling axis might provide new mechanistic insights linking palmitoylation, inflammation and cancer (Extended Data Fig. 10i). Although ZDHHC19 is the major PAT regulating STAT3 in LSCC, other PATs, such as ZDHHC5, have also been implicated in lung cancer[29]. Taken together, these studies highlight the importance of protein palmitoylation in pathogenesis.

METHODS

Reagents

Tris(2-carboxyethyl) phosphine hydrochloride (TCEP) (C4706), Copper(II) sulfate (496130), Tris[(1-benzyl-1H-1,2,3-triazol-4-yl)methyl]amide (TBTA) (678937), Palmitic acid (P5585), Myristic acid (M3128), Palmitoleic acid (P9417), stearic acid (S4751), oleic acid (O1008), arachidic acid (A3631), PEG-maleimide 10 kDa (712469), N-Ethylmaleimide (NEM) (E3876), DL-Dithiothreitol (D9779) and Formamide (F9037), Anti-FLAG® M2 Magnetic Beads (M8823), Phosphatase inhibitor cocktail (P0044) and Crystal Violet (C0775) were purchased from Sigma-Aldrich. Biotin Picolyl Azide (1167), Alkynyl Palmitic acid (Alk-C16) (1165), Alkynyl Myristic Acid (Alk-C14) (1164) and Alkynyl Stearic Acid (Alk-C18) (1166) were purchase from Click Chemistry Tools. Disuccinimidyl glutarate (DSG) (20593), Streptavidin Agarose beads (SA10004), Trizol reagent (15596026) and 3-(4,5-Dimethylthiazol-2-yl)-2,5-Diphenyltetrazolium Bromide (MTT) (M6494) were purchased from Thermo Fisher. PEG-maleimide 5 kDa (146–109) was purchased from Layson Bio. 2-bromohexadec-15-ynoic acid (16-BYA) and Alkynyl arachidic acid (Alk-C20) were synthesized in-house (>98% of purity). Human Recombinant IL-6 (200–06) was purchased from Pepro tech. Human Recombinant IFN-γ (01–172) was purchased from Millipore. Protein A/G XPure Agarose Resin (P5030–1) was purchased from UBPBio. Laemmli (SDS-Sample Buffer, Reducing, 6X) (BP-111R) was purchased from Boston BioProducts. cOmplete EDTA-free protease inhibitors cocktail (05892791001) and Liberase DL (05401160001) were purchased from Roche. Polyethylenimine (PEI) (23966–1) was purchased from Polyscience. Corning® Matrigel® Basement Membrane Matrix (354234) was purchased from Corning.

Antibody

Myc-tag (sc-40), GFP-tag (sc-9996) and GAPDH (sc-47724) were purchased from Santa Cruz Biotechnology. Flag-tag (F1804) and α-Tubulin (T9026) were purchased from Sigma-Aldrich. Flag-tag (14793S), HA-tag (3724S), phospho-STAT3 (Tyr705) (9145S), STAT3 (9139S), Grb2 (3972), Lamin B1(68591S), HDAC1 (5356S), Ki67 (9027S), Anti-Rabbit HRP (7074S), Anti-Mouse HRP (7076S) were purchased from Cell Signaling. STAT3 (MA1–13042), Streptavidin-HRP (N100), Alexa Fluor 488 Goat anti-Rabbit (A11008) and Alexa Fluor 594 Goat anti-Mouse (A11032) were purchased from Thermo Fisher Scientific. ZDHHC19 (26–021) was purchased from Prosci. ZDHHC19-HRP (orb473136) was purchased from Biorbyt. β-Actin (ab6276) was purchased from ABCAM. Mouse TrueBlot® ULTRA Anti-Mouse Ig HRP (18-8817-33) and Rabbit TrueBlot® ULTRA Anti-Rabbit Ig HRP (18-8816-33) were purchased from Rockland.

Plasmids

Full-length human STAT3 cDNA was obtained from DF/HCC DNA Resource Core at Harvard Medical School, and was subcloned into pCMV-3Tag-6 vector (Agilent). The Myc -tagged STAT3 constructs (Myc-STAT3-WT and Myc-STAT3-Y705F) used in this study are the same as previously described[30]. All the mutants (STAT3 C687S, C712S, C687/712S; ZDHHC19 C142S, P18A) were generated by site-directed mutagenesis using the QuickChange II Site-Directed Mutagenesis kit (Agilent) following manufacturer’s instructions. The STAT3 SH2 deletion, N- and C-terminally truncated variants (STAT31−585 and STAT3586−770) were cloned into pCMV-3Tag-6 vector according to the previous report[31]. The constructs of STAT1α Flag pRc/CMV and STAT1β Flag pRc/CMV were gifts from Jim Darnell (Addgene plasmids #8691 and #8704). The EGFP STAT1 plasmid was a gift from Alan Perantoni (Addgene plasmid #12301). The HA-JAK1 plasmid was generously provided by Lixin Rui at University of Wisconsin School of Medicine and Public Health, Madison, WI). The Flag-tagged STAT2 plasmid was generously provided by Dr. Tom Maniatis (Mount Sinai School of Medicine, New York, NY). The other constructs (STAT3, STAT4, STAT5B, STAT3C and STAT3-specific m67-luciferase reporter) used in this study are the same as previously described[32,33]. The HA-ZDHHC constructs were gifts from Dr. Masaki Fukata (National Institute for Physiological Sciences, Japan). The Myc-tagged ZDHHC19 plasmid was purchased from GenScript (Piscataway, NJ).

Cell culture

Human MDA-MB-231, HCC95, SK-MES-1, KNS-62 HCC827, , Phoenix and HEK293A cell lines were obtained from ATCC (Manassas, VA) or Cell Line Depository at Massachusetts General Hospital (MGH) Cancer Center. U3A and STAT3 null (STAT3–/–) mouse embryonic fibroblasts (MEF) cells were provided by Dr. David A. Frank. All the cells were cultured in high-glucose Dulbecco’s Modified Eagle Media (DMEM) (Life technologies) with 10% (v/v) fetal bovine serum (FBS) (Thermo/Hyclone, Waltham, MA), 100 units/mL penicillin and 100 μg/mL streptomycin (Life technologies) at 37°C with 5% CO2. STAT3–/–MEF cells were grown in DMEM containing 10% FBS in the presence of 100 units/mL penicillin and100 μg/mL streptomycin at 37°C with 5% CO2. ATCC cell lines were characterized by Short Tandem Repeat (STR) profiling. None of cell lines used in this paper are listed in the database of commonly misidentified cell lines maintained by ICLAC.

Generation of stable cell lines

Two different shRNAs targeting human ZDHHC19 (ZDHHC19 MISSION shRNA3: TRCN0000134807Sequence CCGGCCAGCAACTGGTATTTAACAACTCGAGTTGTTAAATACCAGTTGCTGGTTTTTTG and ZDHHC19 MISSION shRNA5: TRCN0000137220 CCGGGCCAGCAACTGGTATTTAACACTCGAGTGTTAAATACCAGTTGCTGGCTTTTTTG) were purchased from Sigma (St. Louis, MO). Each shRNA was cloned into the pLKO.1 lentiviral vector. Viruses were produced following the manufacturer’s instructions. Briefly, a 10-cm dish of 80% confluent HEK293T cells was transfected with 10 μg of the transfer plasmid, 5μg pVSVG, 7.5 μg psPAX2, 500 μL of jetPRIME buffer, and 25 μL of jetPRIME transfection reagent (Polyplus). Media was changed after overnight incubation. After 48 hours, viral supernatants were filtered through a 0.45 μm low protein binding membrane (Millipore) and used immediately. Transduction was performed in HCC95 cells, followed by selection with 1 μg/mL puromycin for 2 weeks. ZDHHC19 knockdown efficiency was analyzed by Western blot. For generation of ZDHHC19 knockout stable cells, sgRNAs targeting human ZDHHC19 were cloned into pLentiCRISPR v2 (a gift from Feng Zhang’s lab). Viruses were produced by following the published protocols from Zhang lab. The empty vector pLentiCRISPR v2 was taken as control. After infection, cells were selected with 1 μg/mL puromycin. ZDHHC19 knockout efficiency was analyzed by western blot. The mixed cell populations were used for the experiments. For generation of STAT3 expression stable cell lines, Flag-tagged STAT3 WT or C687/712S mutant was cloned into pBABE-hygro retroviral vector. Retroviruses were produced in the Pheonix packaging cell lines. Transduction of HEK293A stable cells was performed in 6-well plate and selected under 200 μg/mL hygromycin B for 3 days. Flag-STAT3 expression was analyzed by western blot using anti-FLAG M2 (F1804, Sigma Aldrich, 1:5000).

Immunoblotting and immunoprecipitation

Cells were lysed with RIPA buffer supplemented with protease inhibitors (Roche) and phosphatase inhibitors (Roche). Lysates were denatured by heating for 5 minutes at 95 °C and loaded onto 4–12% Bis-Tris polyacrylamide gel. NuPAGE MOPS or MES running buffer (Invitrogen) was used for the SDS-PAGE. The proteins were subsequently transferred to polyvinylidene fluoride (PVDF) membranes (Millipore). The membranes were blocked and incubated with primary antibodies and secondary HRP-conjugated antibodies, and developed by exposure to film. Antibody and dilutions used in the studies: anti-FLAG M2 (1:2000), anti-HA (1:1000), anti-GFP (1:1000), anti-STAT3 (1:1000), anti-pSTAT3 (1:1000), anti-ZDHHC19 (1:500), anti-GAPDH (1:2000), anti-c-Myc (1:1000), streptavidin-HRP antibody (1:5000). Anti-rabbit and anti-mouse IgG, HRP-linked antibodies were diluted 1:5000. For co-immunoprecipitation experiments, cells were lysed in IP buffer (20 mM Tris-HCl, pH 7.0, 250 mM NaCl, 3 mM EDTA, 3 mM EGTA, 0.5% NP-40) supplemented with protease inhibitors (Roche) and phosphatase inhibitors (Roche). 1 μg of indicated antibody was added into lysates and incubated with protein G-Sepharose (life technologies) at 4°C for overnight. Sepharose-enriched immunocomplexes were resolved on SDS-PAGE, transferred to PVDF membrane and analyzed with immunoblotting.

Mass spectrometry analysis

Proteins were precipitated and washed by methanol. The precipitated pellets were then dissolved in suspension buffer containing 4% (w/v) SDS, 50 mM Tris-HCl, pH 7.4, 150 mM NaCl and 10 mM EDTA and further diluted with. Brij97 buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 10 mM EDTA, 1% Brij97). Diluted proteins were incubated with streptavidin agarose beads for 1.5 hours at room temperature on a rotator. The captured proteins were incubated for 40 min in dark with 500 μL 8M urea, 50 μL 500 mM TCEP and 50 μL of 400 mM iodoacetamide and digested on beads. The supernatants were collected for LC-MS/MS analysis as described previously[34].

Labeling, Click reaction and streptavidin pull down

Cells were labeled with DMSO or probe for the indicated time. Cells were lysed in lysis buffer (50 mM TEA-HCl, pH 7.4, 150 mM NaCl, 1% Triton X-100, 0.2% SDS, cOmplete EDTA-free protease inhibitors) followed by Click reaction with biotin-azide. Proteins were precipitated with 9 volumes of 100% methanol for 2 hours at −80°C and recovered by centrifugation at 14,000 × g for 10 min. The pellets were dissolved in 100 ml suspension buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 5 mM EDTA, 2% SDS) and then diluted to 10 fold immunoprecipitation buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 5 mM EDTA, 0.5% NP40).Labeled cellular proteins were enriched using streptavidin agarose (Life technologies) at room temperature with rotation for 3 hours. Protein-bound streptavidin agarose beads were washed three times with immunoprecipitation buffer and bound proteins were eluted with elution buffer (10 mM EDTA pH 8.2, 95% formamide) for 10 min at 95°C. Samples were processed with western blot analysis.

Fractionation assay

Fractionation of cytoplasmic proteins was performed by using the Subcellular Protein Fractionation Kit (Cat# 78840, Thermofisher). The nucleus was lysed with lysis buffer (50 mM TEA, pH7.3, 150 mM NaCl, 4% SDS, 1× protease inhibitor cocktail (Roche, EDTA free), 1,500 units/ml benzonase nuclease, 2 mM PMSF). When the solution was clear, EDTA was added to the final concentration of 5 mM. The cytoplasmic proteins were diluted with the lysis buffer containing 5 mM EDTA before APE assay.

Acyl-PEG exchange (APE) assay

The assay was performed as reported[35]. Briefly, Cells were lysed in lysis buffer (50 mM TEA-HCl, pH 7.4, 150 mM NaCl, 1% Triton X-100, 2% SDS) with protease and phosphatase inhibitors (Roche) followed by incubation with 25 mM TCEP at 55°C for 1 hour. Free cysteine residues were alkylated with 50 mM NEM for 1 hour at room temperature. The mixture was subjected to Methanol/Chloroform protein precipitation and then incubated with NH2OH to cleave palmitoylation thioester bonds at 37°C for 1 hour. Proteins were precipitated by Methanol/Chloroform and incubated with 2 mM PEG at room temperature. After 1 hour, proteins were precipitated again by Methanol/Chloroform and then re-suspended with 1× SDS loading buffer and boiled at 95°C for 3 min and subjected to western blotting analysis.

DSG crosslinking assay

Cells were lysed with RIPA buffer supplemented with protease and phosphatase inhibitors (Roche). 20-fold molar excess of DSG was added into cell lysates. Allow the sample to react on ice for 2 hours. Unreacted DSG was quenched with 100 mM Tris-HCl, pH 7.4 for 10–15 minutes at room temperature. Samples were then boiled with 6× SDS loading buffer at 95°C for 3 min and processed with western blot analysis.

Disulfide detection assay (PEG-maleimide gel shift)

Disulfide detection assay was conducted according to the protocol as previously described[36]. Briefly, the cells were lysed in lysis buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 2 mM EDTA, 1% Triton X-100, 2% SDS) with protease and phosphatase inhibitors (Roche). Free cysteine residues were alkylated with 50 mM NEM for 1 hour at room temperature. The cell lysates were incubated with 50 mM DTT at 37°C for 1 hour. The proteins were precipitated by Methanol/Chloroform and incubated with 2 mM PEG-maleimide (10 kDa) at room temperature for 2 hours. The proteins were precipitated again by Methanol/Chloroform and then re-suspended with 1× SDS loading buffer and boiled at 95°C for 3 min and subjected to western blot analysis.

Transfection and infection

Expression constructs were transfected into cells using jetPRIME transfection reagent, Lipofectamine 2000 or Polyethylenimine (PEI) following previous report[37]. Retrovirus and lentivirus were prepared using Phoenix and 293T cells respectively.

RNA extraction and quantitative real-time PCR (qPCR)

Total RNA was extracted from cells using the TRIZOL (Invitrogen, Carlsbad, CA). Message RNA was converted to first-strand cDNA using high capacity cDNA reverse transcription kit (Thermo, Cat#4368814) and followed by real-time PCR reaction using LightCycler®480 SYBR green Master according to the manufacturer’s directions. Real-time PCR analyses were performed in triplicate by LightCycler®480 PCR machine (Roche). GAPDH was used for normalizing gene expression. Gene expression was calculated using comparative CT method (ΔΔCT method). Data from triplicates experiments were pooled and plotted as shown. The sequences of gene specific primers used for PCR are shown bellow:

Human BCL2 Forward: GCTACCGTCGTGACTTCGC, Reverse: CCCCACCGAACTCAAAGAAGG; Mouse BCL2 Forward: ATGCCTTTGTGGAACTATATGGC, Reverse: GGTATGCACCCAGAGTGATGC; Human BCL2L1 Forward: GGTCGCATTGTGGCCTTTTTC, Reverse: TGCTGCATTGTTCCCATAGAG; Mouse BCL2L1 Forward: GACAAGGAGATGCAGGTATTGG, Reverse: TCCCGTAGAGATCCACAAAAGT; Human MMP9 Forward: GGGACGCAGACATCGTCATC, Reverse: TCGTCATCGTCGAAATGGGC; Mouse MMP9 Forward: CTGGACAGCCAGACACTAAAG, Reverse: CTCGCGGCAAGTCTTCAGAG; Human GAPDH Forward: TGCACCACCAACTGCTTAGC, Reverse: GGCATGGACTGTGGTCATGAG; Mouse GAPDH, Forward: AGGTCGGTGTGAACGGATTTG, Reverse: TGTAGACCATGTAGTTGAGGTCA; Human 18S rRNA Forward: GTAACCCGTTGAACCCCATT, Reverse: CCATCCAATCGGTAGTAGCG; Human Nanog, Forward: CCCCAGCCTTTACTCTTCCTA, Reverse: CCAGGTTGAATTGTTCCAGGTC; Human OCT4, Forward: GGGAGATTGATAACTGGTGTGTT, Reverse: GTGTATATCCCAGGGTGATCCTC; Human Sox2 Forward: TACAGCATGTCCTACTCGCAG, Reverse: GAGGAAGAGGTAACCACAGGG; Human CD133 Forward: GAGAAAGTGGCATCGTGCAA, Reverse: CACGTCCTCCGAATCCATTC; Human ALDH1 Forward: GCCATAACAATCTCCTCTGCT, Reverse: CATGGAAACCGTACTCTCCC.

Transwell migration assay

The assay was performed as previously reported[38]. Briefly, cells were transfected with indicated plasmids and seeded into Trans-well inserts (Corning) in 24-well culture plates. After 16 hours, the cells were fixed and stained by crystal violet solution (1XPBS, 0.05% w/v crystal violet, 1% formaldehyde, 1% methanol). The migrated cells were photographed under Zeiss microscope and counted from 5 randomly selected fields. Data from triplicates experiments were pooled and plotted as shown.

Luciferase assay

Cells were transfected with luciferase construct harboring m67 sequence and Renilla control construct. After 48 hours of transfection, cells were lysed with passive lysis buffer. The activities of Firefly and Renilla luciferase in the lysates were measured with the dual Luciferase Assay System (Promega, Madison, WI). The luminescence was measured using EnVision Multilabel Reader (Perkin Elmer, Waltham, MA). The relative luminescence was calculated by dividing Firefly to Renilla luminescence and normalized to control. Data from triplicates experiments were pooled and plotted as shown.

Cell proliferation assay

Cells were either mock-transfected or transfected with the indicated plasmid. After 24 hours of transfection, cells were seeded into 96-well plate. The cell viability was determined at the indicated time points. 100μL fresh media and 20μL MTT were added into each well. The plates were incubated at 37°C for 2 hours. After incubation, 100μL MTT solvent (methanol) was added into each well. Absorbance values (OD = 590nm) of stained cells were measured using Perkin Elmer EnVision Multilabel Reader. Data from triplicates experiments were pooled and plotted as shown.

Immunofluorescence

Cells were cultured for 48 hours on glass coverslips in 6-well plate and fixed with 4% paraformaldehyde in PBS for 10 min at room temperature, followed by permeabilization with 0.1% Triton X-100 for 5 min at room temperature, incubation with primary antibody for one hour, wash with 1% Triton X-100, and incubation with Alexa Fluor-conjugated secondary antibodies for one hour in the dark. The coverslips were stained with Hoechst 33342 (H3570, Invitrogen, 1:3000) before being mounted with VECTASHIELD mounting medium (H-1000, Vector laboratories). Images were captured by Leica TCS-NT 4D confocal microscope. Z stacks were collected with a spacing of 1 μm. Antibody and dilutions used in the studies: anti-Flag (14793S, Cell signaling, 1:100), anti-Lamin B1 (68591S, Cell signaling, 1:100), anti-STAT3 (MA1–13042, Thermo Fisher Scientific, 1:100). Alexa Fluor 488 Goat anti-Rabbit (A11008, Life technologies, 1:5000), Alexa Fluor 594 Goat anti-Mouse (A11032, Life technologies, 1:5000). The fluorescence-intensity profile along the Z-axis from confocal Z-stacks were shown. The fluorescence intensity of nuclear localized STAT3 was quantified using the confocal software to define the selected ROI (Region of Interest) area based on nuclear DAPI signal. More than 200 cells were quantified in at least three independent experiments.

Immunohistochemistry (IHC)

For paraffin embedded tumor tissue, slides were de-paraffinized and rehydrated using standard procedures. After antigen retrieval (10mM Sodium citrate, 0.05% Tween 20), the slides were processed using EXPOSE Mouse and Rabbit specific HRP/DAB Detection IHC kit (Abcam Cat#80436) according to the manufacturer’s instruction. In brief, the slides were quenched to block endogenous peroxidase and blocked with buffer. The anti-STAT3 (1:2000) or anti-Ki67 (1:500) antibody was incubated for overnight in a humidified chamber. The slides were developed with DAB Chromogen at optimized time. The images were captured using Zeiss microscope. Ki67 positive cells were counted from 5 randomly selected fields and plotted as shown.

IHC staining of STAT3 was quantified by nuclear localization. Score: 0 (negative), 1 (few nuclei), 2 (10%), 3 (10–50%), 4 (>50%); IHC staining of ZDHHC19 was quantified by Score: 0 (<10%), 1 (10–25%), 2 (25–50%), 3 (50–75%), 4 (>75%).

Tumorsphere formation assay

Stem like cells were enriched from HCC95, KNS62 and HCC827 by culturing in serum-free DMEM-F12 medium (Life Technologies, Grand Island, NY) containing 50 μg/ml insulin (Sigma-Aldrich St. Louis, MO), 0.4% Albumin Bovine Fraction V (Sigma-Aldrich St. Louis, MO), N-2 Plus Media Supplement (Life Technologies, Grand Island, NY), B-27 Supplement (Life Technologies, Grand Island, NY), 20 ng/mL EGF (STEMCELL technologies, Cambridge, MA) and 10 ng/ml basic FGF (STEMCELL technologies, Cambridge, MA) in ultra-low attachment plates (Corning, Corning, NY). The images were captured using Zeiss microscope. Spheres were counted from 5 randomly selected fields and plotted as shown.

in vitro limiting dilution assay

Tumor-initiating cell frequencies were calculated using the extreme limiting dilution analysis (ELDA) algorithm as previously described[39]. In brief, the cancer cells were serially diluted to obtain the following final cell concentrations: 1 (288 wells), 10 (192 wells), 100 (96 wells), 1000 (24 wells) cells/well in 96 well ultralow attachment plates (Corning, CLS3474–24EA). After 2 weeks of incubation, the plates were removed from the incubator and the tumorsphere formation was determined using microscope. The cancer cell initiating frequency and significance was determined using Extreme Limiting Dilution Analysis online software (http://bioinf.wehi.edu.au/software/elda/). Images were captured using Zeiss microscope.

Cellular thermal shift assay

Cellular thermal shift assay was conducted according to the protocol as previously described[40]. In brief, HEK293A cells were transfected with STAT3 plasmids and analyzed by intact cell-based and cell lysate-based thermal shift assay at 36 hours post-transfection. For intact cell-based assay, HEK293A cells were collected and washed with PBS buffer for three times before being distributed into different 0.2 mL PCR tubes (~1 million cells per tube). The cells were denatured at the indicated temperatures for 3 min on PCR instrument, following freeze-thaw twice using liquid nitrogen. The samples were centrifuged, and the supernatants were analyzed by Western blot. For cell lysate-based assay, HEK293A cells were collected and harvested with lysis buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 2 mM EDTA, 1% Triton X100). After denaturation at the indicated temperatures for 3 min, the lysates were centrifuged, and the supernatants were analyzed by Western blot.

Patient samples

PDX tumor (PDX7) bearing NSG mouse was purchased from the Jackson Laboratory (PDX LIVETM model TM00231 lung squamous cell carcinoma). The Jackson Laboratory received coded patient tumor samples, with all personal identifier removed, under a Jackson IRB approved PDX Resource Protocol (http://tumor.informatics.jax.org/mtbwi/pdxDetails.do?modelID=TM00231). PDX1 to PDX6 tumor samples were obtained from Dr. A. N. Hata’s lab at Massachusetts General Hospital Cancer Center, the PDX samples are collected in compliance with all relevant ethical regulations for human research participants with patient consent following Protocol 02–180 and 13–416 approved by Dana Farber/Harvard Cancer Center (DF/HCC) IRB. For tissue microarray, one sample set was purchase from US Biomax (OD-CT-RsLug02–001) with 32 cases of lung squamous cell carcinoma tissue. All human tissues are collected under HIPPA approved protocols with the donor being informed completely and with their consent. The US Biomax Inc. states to follow standard medical care and protect the donors’ privacy. Other three slides with 57 cases were obtained from Dr. Mari Mino-Kenudson of Department of Pathology, Massachusetts General Hospital (MGH). The tissues are collected from surgically removed tumors with patient consent following protocol (2009P001838) approved by MGH IRB. All information regarding the patient samples is listed in >Supplementary Table 1 and 2. All samples are de-identified by removing patient information and assigned a code, and the researcher who analyzed the samples could not trace back to patient personal information.

Animal studies

SCID mice were maintained at the Massachusetts General Hospital (MGH) animal facility. All animal studies were conducted in accordance with the Guide for the Care and Use of Laboratory Animals from NIH and a protocol (2013N000065) approved by MGH Institutional Animal Care and Use Committee (IACUC). In none of the experiments did xenograft tumor size surpass 1.5 cm in any two dimensions, and no animal had severe loss of body weight (>15%) or evidence of infections or wounds (endpoints that permitted by IACUC).

2×106 HCC95 WT or ZDHHC19 knockout stable cells, and shRNA control cells or shRNA targeting ZDHHC19 (sh3) stable cells were injected subcutaneously into mouse flanks. The mice were fed with autoclaved regular chow diet (normal fat diet, NFD, composed of 10% fat based on caloric content) or high-fat diet (HFD, composed of 60% fat based on caloric content) (Research Diet Cat#D12492, D12450J New Brunswick, NJ) for one week before cancer cell transplantation and their specific diet were maintained until the experiments were terminated. Group assignment and tumor monitoring were carried out in a double-blinded manner. Tumor volume was assessed by caliper measurement using the formula (width2× length/2) (mm3). After four weeks, the mice were euthanized, and the primary tumors were isolated for further experiments.

The animal tissues used for HFD-induced STAT3 palmitoylation studies were prepared by The Jackson Laboratory (Bar Harbor, ME). Briefly, 10 C57BL/6 mice were randomly divided equally into two groups and fed with NFD or HFD at The Jackson Laboratory. The mice were sacrificed at Day 14. All harvested samples were shipped on dry ice and stored at −80°C for further analysis.

Establishment of patient-derived xenografts in mice

PDX tumor (PDX7) bearing NSG mouse was purchased from the Jackson Laboratory (PDX LIVETM model TM00231 lung squamous cell carcinoma). Once the tumors were established, the mice were euthanized, and the tumors were dissected and dissociated. 2–3 mm PDX tumor fragments mixed with Matrigel were loaded into syringe with 14 gauge and subcutaneously implanted into both flanks of SCID mice. When the tumors were established at the week 3 post-inoculation, normal fat diet or high fat diet were administered. Tumor volume was determined by caliper measurement using the formula (width2× length/2) (mm3).

Isolation of primary cancer cells from PDX tumor

Tumor tissue collected from PDX mouse was minced into small pieces and enzymatically dissociated into single cells with 0.28 Unit/ml Liberase DL (Roche #05401160001) for 1–2 hours at 37°C. Cells were then washed with Hank’s balanced salt solution containing 2% FBS. The single cell suspension was then cultured in advanced DMEM-F12 medium supplemented with 10% FBS, 50 μg/ml insulin, 20 ng/mL EGF, 10 ng/mL basic FGF, 6 mmol/L glutamine, 1% of penicillin/streptomycin and 40 ng/mL dexamethasone. The isolated cells were cultured in tumorsphere media mentioned in Tumorsphere formation assay to determine stem like cell enrichment.

Statistical analysis

Results were presented as mean ± s.e.m of at least three independent experiments. Measurements at single time points were analyzed by ANOVA and if significant, further analyzed by a two-tailed Student’s t-test. Time courses were analyzed by repeated measurements (mixed model) ANOVA with Bonferroni post-tests. Survival analysis was estimated by the Kaplan-Meier method. Survival in different groups was compared using the log-rank test with GraphPad Prism, version 6.0 (GraphPad Software, San Diego, CA). The Pearson’s correlation coefficient (r) was used to measure the strength of the association between two variables. P < 0.05 was considered to indicate statistical significance.

Data availability

The data supporting the findings of this study are available within the paper. Uncropped raw images from western blots are shown in Supplementary Information. Source Data for all graphs are available online. Tumor sample information is shown in Supplementary table. All other data are available from the corresponding author upon reasonable request.