Serine-Phosphorylated STAT3 Promotes Tumorigenesis via Modulation of RNA Polymerase Transcriptional Activity

Elevated tyrosine phosphorylation at residue 705 (pY705) of STAT3 is an indirect indicator of its transcriptional activity on diverse gene networks, which promote many cancer-associated cellular processes, such as proliferation and survival, inflammation, angiogenesis, metastasis, and immunosuppression (1, 2). The IL6 cytokine family (IL6, IL11, IL27, among others) is a prominent activator of STAT3 via dimerization of the gp130 signal-transducing coreceptor. Here, activation of gp130-associated Janus kinases (JAK) tyrosine phosphorylate STAT3, enabling formation of STAT3 homodimers, or heterodimers with additional transcription factors (e.g., STAT1, NANOG, c-Jun/c-Fos, OCT-1) and coactivators (e.g., p300/CBP), which then translocate to the nucleus (3, 4). Also, pY-STAT3 can indirectly influence gene transcriptional programming by inducing the expression of other transcription factors (e.g., MYC; ref. 4).

In a nonphosphorylated state, STAT3 can also associate with NF-κB to drive a distinct transcriptional signature comprising genes implicated in oncogenic and immune responses (5). It has also emerged that posttranslational modifications of STAT3, namely acetylation, methylation, S-glutathionylation, ubiquitination, and SUMOylation, can collectively modulate its dimerization, nuclear retention, and DNA binding capacity, and thus influence transcriptional outputs (6, 7). Notably, the tumor promoting potential of STAT3 extends to its role in the mitochondria, where it acts as a central regulator of cellular metabolism. Although nuclear STAT3 can reprogram metabolism in cancer cells by directing the transcription of MYC and HIF1A, both of which are master regulators of the “Warburg effect,” STAT3 can also enter the mitochondria and directly regulate the activity of the electron transport chain that is dependent on serine phosphorylation at residue 727 (pS727; refs. 8–10). Mitochondrial pS–STAT3 appears to be a requisite for mutant Ras-driven tumors, whereas the extent of its role in the pathogenesis of other cancer types remains unknown (6–8, 11).

There are contrasting reports on the requirement of pS727 for STAT3 DNA binding and transcriptional activities, and modulation of Y705 phosphorylation (12–17). Among these, embryonic fibroblasts derived from Stat3^SA/SA mice harboring an S727A knock-in substitution in the endogenous Stat3 locus (SA allele) exhibited a 50% reduction in the IL6-dependent transcriptional response compared with wild-type cells (13). Furthermore, Stat3^SA/SA mice used in a model of angiotensin II–induced hypertension demonstrate reduced expression of STAT3-dependent cardiac remodeling genes (17). These observations suggest a role for pS–STAT3 in transcriptional regulation, along with its potential contribution to physiologic and pathologic states, albeit ill defined.

We have previously generated a STAT3-driven genetic mouse model (gp130^F/F) of spontaneous intestinal-type gastric cancer (18–20). These mice are homozygous for a phenylalanine (F) knock-in substitution of the cytoplasmic Y757 residue in endogenous gp130, which blocks binding of the negative regulator suppressor of cytokine signaling (SOCS)3, leading to exaggerated IL11-driven pY-STAT3 levels in the gastric compartment (18, 19). The causal role for dysregulated IL11/gp130-dependent STAT3 activation in the development of gastric tumors in gp130^F/F mice was shown by the suppressed tumorigenesis observed upon either heterozygous or homozygous ablation of Stat3 or Il11rα1, respectively (18, 19). However, whether pS–STAT3 plays any role in gastric cancer is unknown. Here, we use a genetic strategy to demonstrate that constitutive serine phosphorylation is essential for the oncogenic activity of STAT3 in the gastric epithelium. Moreover, by combining a proteomics-based screen with CRISPR/Cas9-mediated functional validation, we reveal RuvB-like AAA ATPase 1 (RUVBL1/Pontin) and enhancer of rudimentary homolog (ERH) proteins as interacting partners of pS–STAT3 essential for maximal transcriptional efficiency of RNA polymerase on STAT3-regulated genes in gastric cancer.

Antral gastric biopsies (tumor/matched nontumor) were collected by surgical resection from patients with gastric cancer at Monash Medical Centre (MMC, Melbourne, Australia) and Xin Hua Hospital (Shanghai, China; Supplementary Table S1). Noncancerous (normal control) antral gastric tissue was collected from cancer-free individuals undergoing endoscopy (MMC). Biopsies were snap-frozen in liquid nitrogen or stored in 10% formalin, the latter for histopathology and Helicobacter pylori status (21). Written informed patient consent was obtained, and collections were approved by Monash Health Human Research Ethics Committee and Xin Hua Hospital Ethics Committee. Patient studies were conducted in accordance with the World Medical Association Declaration of Helsinki statement on the ethical principles for medical research involving human subjects.

The gp130^F/F, gp130^F/F:Stat3^+/−, gp130:Il11ra^−/−, and Stat3^SA/SA mice (13, 18, 19, 22), along with genetically matched (129Sv × C57BL/6) wild-type mice as littermate controls, were housed under specific pathogen-free conditions. All experiments were approved by the MMC “B” Animal Ethics Committee.

Recombinant human IL11 (5 μg; PeproTech) or PBS was intraperitoneal injected into 6- to 8-week-old mice. For xenografts, MKN-28 STAT3-WT, STAT3-SD, or STAT3-SA cells (1.75 × 10⁶) resuspended in 50% (v/v) PBS and Matrigel (Cultrex) were subcutaneously injected into NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ/Arc (NSG) mice (Animal Resources Centre, Australia), and tumor size was measured weekly using electronic callipers to calculate tumor volume (2 × width × length)/2 = V mm³).

Six- to 8-week-old gp130^F/F mice were lethally irradiated with a single 9.5 Gy dose, following which, mice were reconstituted via intravenous injection with 10⁷ cells from unfractionated bone marrow of donor mice. Recipient mice were culled at 10 to 12 weeks posttransplant.

Nuclear and cytoplasmic fractions were isolated from fresh antral tissues using the NE-PER Nuclear and Cytoplasmic Extraction Reagents Kit (ThermoScientific). For epithelial gland isolation, gastric antral tissue was collected in ice-cold Hank's Balanced Salt Solution (HBSS; Gibco), and tissue was incubated in HBSS containing 30 mmol/L EDTA for 15 minutes at 37°C. Tissue was repeatedly flushed using a 1 mL pipette to liberate glands from the basement membrane. Glands were collected by centrifugation and lysed in RIPA buffer.

Following formalin fixation and paraffin embedding (FFPE), histologic assessment of mouse stomachs and MKN-28 xenografts was performed on 4 to 6 μm hematoxylin and eosin (H&E)-stained tissue sections. For IHC, sections were deparaffinized in xylene and rehydrated by sequential submersions in 100%, 70%, and 0% (v/v) ethanol/dH₂O. Antigen retrieval for pY-STAT3 IHC was performed in heated 1 mmol/L EDTA, pH 8.0 solution, and for other IHC in heated 10 mmol/L sodium citrate acid, pH 6.0 solution. Endogenous peroxidase was blocked in 3% (v/v) hydrogen peroxidase/methanol, and sections were blocked against nonspecific binding by incubating in 10% (v/v) serum/Tris-buffered saline, matched to the species of secondary antibody. IHC was performed with primary antibodies against pY-STAT3 (1:1,000 dilution), total STAT3 (1:1,000 dilution), and PCNA (1:5,000 dilution) (Cell Signaling Technology), pS–STAT3 (1:1,000 dilution; Santa Cruz Biotechnology), and CD45 (1:100 dilution; BD Biosciences), along with concentration matched rabbit-IgG control (Vector Labs). Following overnight 4°C incubations, sections were incubated at room temperature with biotinylated anti-rabbit IgG antibodies (Vector Labs), biotin labeled with HRP (Vectastain ABC HRP Kit; Vector Labs), and developed using a liquid diaminobenzidine chromogen substrate system (Dako). PCNA staining was performed using the Mouse on Mouse (M.O.M) Detection Kit (Vector Labs). Sections were counterstained with hematoxylin, images acquired using CellSens software (Olympus). PCNA and STAT3 (total, pY- and pS-) positivity was determined using the positive pixel count algorithm v9.0 (Leica Biosystems) on whole slide scans of stained sections; positivity was calculated as the total number of positive pixels/total number of pixels.

For mucosal thickness measurements, whole H&E-stained slides were imaged using an Aperio Scanscope AT Turbo digital pathology scanner (Leica Biosystems). Mucosal thickness was measured at 10 random points using the linear measurement tool of Aperio Image Scope software (Leica Biosystems), and the average taken. For PAS/Alcian blue staining, sections were incubated in 1% (w/v) Alcian blue [in 3% (v/v) acetic acid/dH₂O, pH 2.5] followed by oxidation in 1% (v/v) periodic acid/dH₂O. Slides were immersed in Schiff's reagent and counterstained with hemotoxylin prior to mounting.

FFPE human and mouse gastric sections were deparaffinized and subjected to antigen retrieval, along with endogenous peroxidase blockade, as with IHC. Following blocking of nonspecific binding with CAS-Block (Invitrogen), immunofluorescence was performed on sections with antibodies against pS–STAT3 (Santa Cruz Biotechnology), ERH (Abcam), and Pontin (Cell Signaling Technology), each at 1:100 dilution. Alexa Fluor conjugates (Invitrogen) were used as secondary antibodies. Nuclear staining was achieved using 4′,6-diamidino-2-phenylindole (DAPI). Sections that underwent the above staining protocol, but in the absence of primary antibodies, served as negative controls. Images were acquired using a Nikon C1 confocal microscope. To quantify cellular staining, digital images of photomicrographs (60× high power fields) were viewed using Image J software. Positive-staining cells were counted manually (n = 8 fields).

Human gastric cancer cell lines MKN-28 and MKN-1 (Japanese Collection of Research Bioresources Cell Bank) were maintained in Roswell Park Memorial Institute 1640 media (Gibco) containing 10% FCS (Bovogen). Cell line identification was authenticated by short-tandem repeat profiling (PowerPlex HS16 System Kit; Promega) in our laboratory after receipt in 2013, and cells were passaged during experiments for under 12 weeks at a time between freeze/thaw cycles. Lenti-X 293T cells (Clontech) were cultured in DMEM supplemented with 10% FCS. Cells were routinely tested for mycoplasma contamination (MycoAlert PLUS Mycoplasma Detection Kit; Lonza). Cells were incubated at 37°C supplemented with 5% CO₂ in a humidified chamber.

Murine primary gastric epithelial cells were isolated from 4- to 5-week-old mouse gastric antral tissues as before (23).

MKN-28 STAT3-WT, STAT3-SD, or STAT3-SA cells were generated by reconstituting STAT3-null cells (24) with STAT3-WT, STAT3-SD, or STAT3-SA (18, 25, 26) lentiviral expression constructs (pLVX-IRES-mCherry; Clontech). After viral transduction, STAT3-WT, STAT3-SD, or STAT3-SA cells were sorted for mCherry by flow cytometry to select for cell populations with equivalent STAT3 expression levels that approximated endogenous STAT3 expression in parental MKN-28 cells.

For the CRISPR RNA (crRNA) screen, Alt-R CRISPR/Cas9 crRNA and trans-activating crRNA (tracrRNA; Integrated DNA Technologies) were duplexed (single guide RNA) and transfected into MKN-28 cells stably expressing Cas9 (generated using lentiviral-encoded pLentiCas9-Blast vector). For transfection control, tracrRNA alone was used. The effect of crRNA target gene knockdown on SOCS3 mRNA induction was determined by qRT-PCR. All vectors were sequenced for validation, and sequences for crRNAs and tracrRNAs, along with cloning primers, are available upon request.

Total RNA was isolated from human gastric cancer cell lines and human/mouse gastric tissues using TRI Reagent Solution (Sigma), followed by on-column RNeasy Mini Kit RNA clean-up and DNase treatment (Qiagen). RNA was transcribed using the Superscript III First-Strand Synthesis System (Invitrogen) and Transcriptor High Fidelity cDNA Synthesis Kit (Roche), respectively. qRT-PCR was performed on samples in technical triplicates using the 7900HT Fast RT-PCR System (Applied Biosystems), and data acquired and analyzed (27). Mouse and human primer sequences are in Supplementary Table S2.

Genomic DNA was isolated from snap-frozen tissue using the ISOLATE II Genomic DNA Kit (Bioline), and subjected to RNAse-A treatment (Qiagen). The ratio of mitochondrial DNA copy number to the amount of nuclear DNA was calculated by qRT-PCR (7900HT Fast RT-PCR System using SensiMix SYBR Hi-ROX (Bioline) as described previously (28).

Gene expression profiling of Cyanine-3–labeled cRNA from 4-week-old mouse gastric antral tissues (n = 6/genotype) was performed on Agilent SurePrint G3 Mouse Gene Expression 8 × 60K chips. Scanned images of array slides were analyzed with Feature Extraction Software 11.0.1.1 (Agilent) using default parameters (GE1-1100_Jul11 and Grid: 028005_D_F_20120201) to obtain processed signal intensities. Data were imported and integrated using Genespring 13.0 (Agilent).

JavaGSEA Desktop Application v2.2.2 was used on microarray datasets comprising 45,578 native features (Agilent probes), which were collapsed to 9,300 genes, and h.all.v5.2.symbols.gmt, c2.all.v5.symbols.gmt, and c5.all.v5.2.symbols.gmt gene sets (https://software.broadinstitute.org/gsea/msigdb/) were used. A total of 1,000 “gene_set” permutations were used to test statistical analysis. Significant, positive enrichments were set at normalized enrichment score >0 and FDR q value of <0.05. All basic and advanced fields were set to default.

Overall survival analyses were performed on the “Gastric Cancer Project ‘08 Singapore’” (GSE15459) intestinal-type gastric cancer patient cohort (29) using low (bottom third) versus high (top third) median expression values for ERH, RUVBL1, or SOCS3. Treatment among the patients (n = 99) comprised surgery alone (n = 68) or adjuvant 5-fluorouracil therapy (n = 12), with the remainder unknown (n = 19).

Antibodies against pY-STAT3, pS–STAT3, STAT3, vimentin, and E-cadherin (Cell Signaling Technology), tubulin and OXPHOS antibody cocktail (Abcam), and Actin and FLAG-M2 (Sigma-Aldrich), were used. Membranes were analyzed using the Odyssey CLx Imaging System (Li-Cor). Protein expression was quantified using ImageJ (NIH).

Detailed methodology for in vitro chromatin immunoprecipitation (ChIP) on MKN-28 cells and in vivo ChIP on mouse gastric tissue is provided in the “Supplementary Materials and Methods.” For quantitative in vitro and in vivo ChIP data analyses, % input and fold enrichment were calculated using the 2-ΔΔC_t method (30).

In vitro rapid immunoprecipitation mass spectrometry of endogenous (RIME) on MKN-28 cells and in vivo RIME on mouse gastric tissue was performed using modified published protocols (31, 32), which are extensively detailed in the “Supplementary Materials and Methods.”

Cells were seeded in μ-Slide chambers (Ibidi), cultured and IL11-treated as per ChIP and RIME experiments. Cells were fixed in 4% formaldehyde/PBS and permeabilized with 100% ice-cold methanol. Proximity ligation assay (PLA) was performed using the Duolink In Situ Red Starter Kit (Mouse/Rabbit; Sigma). Antibody combinations included STAT3 (Cell Signaling Technology) with either ERH (Abcam) or Pontin (Cell Signaling Technology). STAT3-deficient cells were used to determine assay background. Images were acquired on a Nikon C1 confocal microscope, and analysis and quantification were performed using CellProfiler software version 3.0.0 (Broad Institute Imaging Platform).

Statistical analyses were performed using GraphPad Prism V7.0 software or R package. Statistical significance (P < 0.05) between the means of 2 groups was determined using unpaired t tests or Mann–Whitney U tests, and for matched datasets involved Wilcoxon signed-rank tests. Statistical significance between the means of multiple groups were determined using ordinary 1-way ANOVA or Kruskal–Wallis tests. Data are presented as the mean ± SEM. The log-rank test was used to calculate the statistical significance of the difference in survival between 2 groups.

In gp130^F/F mice, gastric adenomatous hyperplasia occurs by 6 weeks of age with 100% penetrance, followed by formation of adenomas (tumors) by 12 weeks of age, which continue to grow until a maximal tumor size is reached by 52 weeks (18, 19). In wild-type gp130^+/+ gastric tissues, pY-STAT3 levels are low, whereas pY-STAT3 levels are elevated in 4-week-old pretumor bearing and 12-week-old tumor-bearing gp130^F/F gastric tissues (Fig. 1A; Supplementary Fig. S1A and S1B). By contrast, comparable pS–STAT3 levels were detected in gastric tissues from 4-week-old and 12-week-old gp130^+/+ and gp130^F/F mice (Fig. 1A; Supplementary Fig. S1A and S1B).

To define whether constitutive pS–STAT3 expression is required for gastric tumorigenesis, we crossed gp130^F/F mice with pS–STAT3–deficient Stat3^SA/SA mice (13). At 4 weeks of age, the increased gastric mucosal thickness, stomach mass, and IHC staining for the PCNA cell proliferation marker, each indicative of the hyper-proliferative response within the gastric epithelium of gp130^F/F mice, were markedly reduced in gp130^F/F:Stat3^SA/SA mice (Supplementary Fig. S1C–S1G). Also, stomachs of 12-week-old gp130^F/F:Stat3^SA/SA mice were free of tumors, and characterized by significant reductions (compared with age-matched gp130^F/F mice) in stomach mass, gastric mucosal thickness, and PCNA-positive cell numbers, comparable to those in gp130^+/+ controls (Fig. 1B–F). From 26 weeks of age onwards, one-third of gp130^F/F:Stat3^SA/SA mice developed small gastric adenomas (Supplementary Fig. S2A and S2B). Furthermore, gp130^F/F:Stat3^+/SA mice containing only 1 Stat3^SA allele displayed a marked reduction in tumorigenesis at 12 and 26 weeks of age, indicating a gastric gene dosage effect of pS–STAT3 (Fig. 1B, D, and E; Supplementary Fig. S2A and S2B). Indeed, the magnitude of gastric tumor suppression in gp130^F/F:Stat3^SA/SA mice was comparable to that in gp130^F/F:Stat3^+/− mice, where the total signaling capacity of STAT3 has been genetically reduced to wild-type levels by heterozygous ablation of 1 endogenous Stat3 allele (Supplementary Fig. S2A–S2H; refs. 18, 19). The comparable rescue in gastric tumorigenesis of gp130^F/F:Stat3^SA/SA and gp130^F/F:Stat3^+/− mice aligned with a significant survival benefit, with both genotypes displaying 71% and 76% survival at 1-year, respectively, compared with 13% of gp130^F/F mice (Supplementary Fig. S2H and S2I). Therefore, these data reveal an obligate requirement for constitutive pS727 by STAT3 for its protumorigenic actions in the stomach.

In support of our in vivo findings, IHC revealed strong and comparable pS–STAT3 levels in gastric cancer patient biopsies (62% and 50% positive staining in nontumor and tumor tissue sections, respectively) and cancer-free controls (57% positive staining in normal tissue sections; Supplementary Fig. S3A–S3C; Supplementary Table S1). By contrast, pY-STAT3 was barely detectable in cancer-free controls (4% positive staining), and increased pY-STAT3 staining intensity was observed in gastric cancer patient nontumor (11% positive staining) and tumor (18% positive staining) tissues, albeit at levels markedly below those for pS–STAT3 (Supplementary Fig. S3A–S3C). Furthermore, immunoblotting of a panel of human gastric cancer cell lines indicated that pS–STAT3 expression was detected in all cell lines, whereas pY-STAT3 levels were only detected in 4 of 7 cell lines (Supplementary Fig. S3D). These observations demonstrate that pS–STAT3, similar to the gp130^F/F mouse model, is constitutively expressed in both malignant and nonmalignant human gastric tissues.

We next evaluated whether constitutive pS–STAT3 was required for the growth of established human gastric cancer cell line–derived xenografts. Upon CRISPR/Cas9-mediated knockout of STAT3 expression in human MKN-28 GC cells (24), cells were reconstituted with either wild-type STAT3 (STAT3-WT), a STAT3 mutant (serine to aspartic acid substitution at site 727; STAT3-SD) that mimics constitutive S727 phosphorylation (26), or S727 phosphorylation-defective STAT3 (STAT3-SA), via lentiviral transduction. The growth of STAT3-SA–expressing human MKN-28 xenografts in NSG mice was significantly impaired compared with STAT3-WT and STAT3-SD MKN-28 xenografts (displaying a comparable growth rate), and was associated with reduced numbers of PCNA-positive proliferating cells (Supplementary Fig. S4A–S4D). These data support a key role for constitutive pS–STAT3 in promoting the proliferative potential of the gastric epithelium during gastric cancer.

We next evaluated whether the suppression of gastric tumorigenesis in gp130^F/F:Stat3^SA/SA mice coincided with changes to inflammation. Histopathologic scoring of H&E-stained gastric antrum sections from 12-week-old gp130^F/F and gp130^F/F:Stat3^SA/SA mice, along with wild-type littermates, demonstrated that the marked gastric inflammation observed in gp130^F/F mice was significantly attenuated in gp130^F/F:Stat3^SA/SA mice (Fig. 2A). Similarly, IHC revealed that CD45-positive infiltrating immune cells present throughout the gastric compartment of gp130^F/F mice were largely absent in gp130^F/F:Stat3^SA/SA stomachs, the latter of which were comparable to wild-type controls (Fig. 2B).

To address whether pS–STAT3 expression in immune cells contributed to gastric tumorigenesis, we initially performed IHC for pS–STAT3 on gastric tumor sections from 12-week-old gp130^F/F mice. This indicated that pS–STAT3 was widely expressed in inflammatory cell aggregates distributed within the submucosal, muscularis, and serosal layers, as well as throughout the tumor mucosal epithelium (Fig. 2C). Immunoblotting also demonstrated that pS–STAT3 expression levels were comparable among protein extracts isolated from the glandular epithelium and stroma (containing immune cells) of 12-week-old gp130^F/F mouse tumors (Fig. 2D). Next, we assessed whether immune cells expressing pS–STAT3 contributed to gastric tumorigenesis by reconstituting bone marrow of irradiated gp130^F/F recipient mice with donor bone marrow from gp130^F/F:Stat3^SA/SA mice. However, the appearance of stomachs of gp130^F/F recipients reconstituted with either gp130^F/F:Stat3^SA/SA or control gp130^F/F bone marrow, along with their stomach mass and tumor burden, were comparable (Fig. 2E–G).

We next determined whether additional features of the intestinal-type gastric tumor phenotype of gp130^F/F mice, namely augmented angiogenesis and production of acidic mucins, were also dependent on pS–STAT3. Expression profiling by qRT-PCR revealed a significant reduction in mRNA levels of angiogenesis-related genes in gastric tissues of 12-week-old gp130^F/F:Stat3^SA/SA versus gp130^F/F mice (Fig. 2H). In addition, Alcian blue-Periodic Acid Schiff staining demonstrated that stomachs of gp130^F/F:Stat3^SA/SA (and wild-type) mice were devoid of acidic mucins that are indicative of the intestinal-type gastric tumor pathology of gp130^F/F mice (Fig. 2I). Collectively, these data support the notion that pS727 plays a critical role in facilitating the oncogenic potential of STAT3—via cell proliferation, inflammation, and angiogenesis—during the initiation and establishment of gastric cancer in a cell intrinsic manner.

We next examined the molecular basis by which pS–STAT3 promotes the onset of gastric cancer by DNA microarray profiling the transcriptome of pretumorigenic (and noninflamed) gastric antrum tissue from 4-week-old gp130^F/F mice, and age-matched wild-type and gp130^F/F:Stat3^SA/SA mice. A total of 987 genes were significantly differentially expressed (>2-fold, P < 0.05) in gastric antrum of gp130^F/F compared with wild-type mice, with 477 and 510 genes upregulated and downregulated, respectively (Fig. 3A; Supplementary Table S3). Strikingly, the gastric antrum transcriptome of gp130^F/F:Stat3^SA/SA mice resembled that of wild-type mice, with a 3-fold reduction in the number of significantly differentially-expressed genes (322; 185 upregulated, 137 downregulated) compared with gp130^F/F gastric antrum (Fig. 3A).

Because pS–STAT3 is implicated in mitochondrial function and associated metabolic reprogramming, a hallmark of cancer (33–35), we investigated whether gene networks associated with these processes were differentially modulated by pS–STAT3 in gp130^F/F gastric antrum. Heat map analyses of DNA microarrays indicated that mitochondrial-encoded genes exhibited a comparable expression profile among the gastric antrum genotypes (Fig. 3B). In addition, qRT-PCR and immunoblotting demonstrated that mRNA and protein levels of mitochondrial-encoded and nuclear-encoded mitochondrial genes were similar among gastric antrum from wild-type, gp130^F/F and gp130^F/F:Stat3^SA/SA mice aged 4 and 12 weeks, the latter including tumor and matched nontumor tissues from gp130^F/F mice (Fig. 3C–E; Supplementary Fig. S5A). Similarly, the mitochondrial DNA copy number was not affected by altered gp130-dependent signaling in gp130^F/F and gp130^F/F:Stat3^SA/SA mice (Supplementary Fig. S5B and S5C). Notably, the expression profile of metabolic genes for glycolysis and the citric acid (TCA) cycle were also unchanged (Fig. 3F and G). By contrast, gene set enrichment analysis (GSEA) illustrated an enrichment of gene signatures for inflammatory response, regulation of innate immunity, epithelial wound healing (migration) and JAK-STAT3 pathways in gp130^F/F (versus wild-type) compared with gp130^F/F:Stat3^SA/SA gastric antrum (Fig. 3H). Further interrogation of a signature comprising 30 STAT3 target genes (e.g., Socs3, Il11, Myc, Mmp9, Vegfa; refs. 18, 19) indicated that pS–STAT3 deficiency restored their expression to wild-type levels in gp130^F/F:Stat3^SA/SA gastric antrum (Fig. 3I; Supplementary Fig. S5D and S5E). Collectively, these data show that pS727 plays a central role in modulating the transcriptional activity of STAT3 in gastric tumorigenesis.

Because the IL11-gp130-JAK-STAT3 signaling axis promotes gastric tumorigenesis in gp130^F/F mice (18), we reasoned that reduced gastric IL11 expression levels in gp130^F/F:Stat3^SA/SA mice may dampen the overall STAT3 oncogenic signal output, leading to suppressed tumorigenesis (Fig. 3I; Supplementary Fig. S5D and S5E). Indeed, immunoblotting and/or IHC showed that pY-STAT3 and IL11 protein levels were reduced in the gastric antrum of gp130^F/F:Stat3^SA/SA versus gp130^F/F mice (Figs. 1A and 4A–E; Supplementary Fig. S1A), and gastric pY-STAT3 levels were similarly lower in both nuclear and cytoplasmic subcellular fractions from gp130^F/F:Stat3^SA/SA mice (Fig. 4F). Furthermore, we demonstrated a rescue effect on gastric pY-STAT3 levels in gp130^F/F:Stat3^SA/SA mice upon the administration of IL11 over 6 hours, coincident with augmented protein levels of c-Myc (a representative protumorigenic STAT3-regulated target; Fig. 4G).

Considering these observations, we investigated whether IL11 upregulated pS–STAT3 levels in addition to pY-STAT3 during gastric tumorigenesis. However, in 4-week-old gp130^F/F:Il11r^−/− mice deficient in IL11 signaling (18), gastric pS–STAT3 levels remained unaltered, whereas pY-STAT3 levels were substantially reduced compared with gp130^F/F mice (Fig. 4H). Similarly, pY-STAT3 levels were robustly increased in gastric antrum tissues of wild-type mice or human MKN-1 gastric cancer cell lines treated with IL11, whereas pS–STAT3 levels remained unchanged (Fig. 4I and J). Furthermore, treatment of these human gastric cancer cell lines with JAK inhibitors abrogated pY-STAT3 while having no effect on pS–STAT3 (Fig. 4K). Collectively, these data suggest that the upstream pathways leading to pY-STAT3 and pS–STAT3 signaling are not interdependent, but rather cooperate to drive gastric cancer. Moreover, the reduction in pY-STAT3 levels in gp130^F/F:Stat3^SA/SA mice are likely caused by reduced IL11 expression, which is the major driver of pY-STAT3 in this gastric cancer model.

To investigate whether pS727 plays a critical role in modulating the transcriptional output of STAT3 by influencing its recruitment to target gene promoters, we performed ChIP. The Socs3 gene promoter was used as a representative STAT3-regulated target because Socs3 gene expression levels strongly align with STAT3 activity in gastric cancer (18, 24, 27). To avoid any bias conferred by different gastric pY-STAT3 expression levels (observed between gp130^F/F and gp130^F/F:Stat3^SA/SA mice), we used gastric tissues from IL11-treated wild-type and Stat3^SA/SA mice, which results in comparable pY-STAT3 induction in the stomach (Fig. 5A). Although equivalent gastric pY-STAT3 protein levels bound to the Socs3 promoter region in IL11-treated wild-type and Stat3^SA/SA mice, there was a significant reduction in Socs3 mRNA levels in the Stat3^SA/SA gastric antrum (Fig. 5B and C). This finding was confirmed by ChIP in IL11-stimulated human MKN-28 gastric cancer cells expressing either STAT3-WT or STAT3-SA (Fig. 5D–F).

The above observations dissociate pY-STAT3 promoter recruitment from the transcriptional induction of STAT3-regulated genes, and suggest that the requirement of pS727 for maximal transcriptional capacity of STAT3 is independent of its loading onto target promoters. We therefore hypothesized that the impaired transcriptional response in the absence of pS–STAT3 could be due to inefficient promoter recruitment of RNA polymerase II (Pol II), the initiation of transcription, or elongation of nascent RNA transcripts. To investigate this, we performed ChIP on IL11-stimulated human MKN-28 GC cells expressing STAT3-WT and STAT3-SA for the binding of distinct serine phosphorylation sites within Pol II that are associated with transcription initiation and 5′-end mRNA processing (RNA Pol II pSer5 CTD) or transcription elongation (RNA Pol II pSer2 CTD; Fig. 5G) at the SOCS3 promoter. The accumulation of pSer5-RNA Pol II on the active SOCS3 promoter was comparable in IL11-stimulated cells expressing STAT3-WT (44% increase) and STAT3-SA (33% increase), which suggests that pS–STAT3 does not have a major influence on the recruitment of RNA Pol II during transcription (Fig. 5H). By contrast, in IL11-stimulated STAT3-SA cells, pSer2-RNA Pol II levels bound to the SOCS3 gene were significantly impaired (by 90%) compared with IL11-stimulated STAT3-WT cells (STAT3-WT, 270% increase vs. STAT3-SA, 25% increase; Fig. 5H). Notably, in STAT3-deficient cells compared with STAT3-WT and STAT3-SA cells, SOCS3 promoter loading by pSer5-RNA Pol II was equivalent in the absence of IL11 stimulation, yet was not increased following IL11 stimulation (Fig. 5H). Collectively, these data indicate that STAT3 is required for increased promoter recruitment of pSer5-RNA Pol II following stimulation, yet this is independent of pS727. Rather, pS–STAT3 primarily regulates transcription by augmenting RNA Pol II-mediated elongation, but not initiation.

The ability of pS–STAT3 to regulate RNA Pol II elongation but not promoter recruitment of either STAT3 or RNA Pol II suggests that additional transcriptional coregulators may interact with pS–STAT3. To identify interacting partners specific for pS–STAT3, we undertook a combinatorial in vitro and in vivo proteomics RIME approach using gastric tissues of 4-week-old gp130^F/F and gp130^F/F:Stat3^SA/SA mice together with IL11-treated human MNK-28 GC cells expressing STAT3-WT and STAT3-SA (Fig. 6A; Supplementary Fig. S6A and S6B). Using this unbiased strategy, 14 candidate serine phosphorylation-dependent STAT3-interacting proteins were identified (Fig. 6A).

To validate the functional requirement of the 14 candidate pS–STAT3-binding proteins for the transcriptional capacity of STAT3, we established a CRISPR/Cas9 (36) screen to knockout each of these 14 genes in human MKN-28 GC cells, along with that of STAT3 as a control, and assess the effect on SOCS3 gene transcription in response to IL11 stimulation. Although transfection of tracrRNA alone had no impact on SOCS3 induction, in contrast, transfection with the tracrRNA-crRNA duplex targeting STAT3 reduced SOCS3 transcript levels (by 56%) to that in unstimulated cells. Among these 14 genes, genetic ablation of either RUVBL1 (encoding Pontin) or ERH markedly impaired the capacity of IL11 to upregulate mRNA levels of SOCS3 and the additional STAT3 target gene, MMP9 (Fig. 6B–D; Supplementary Fig. S6C). The requirement of Pontin and ERH for IL11-dependent transcription was specific for STAT3-regulated genes (e.g., SOCS3), but not genes regulated by the related IL11/STAT1 axis (e.g., SOCS1; Fig. 6E; ref. 18).

To confirm the physical interaction of pS–STAT3 with Pontin and ERH, we used in situ PLA on IL11-treated human MKN-28 STAT3-WT and STAT3-SA cells using antibodies against STAT3 and either ERH or Pontin. The abundance of puncta denoting STAT3–Pontin or STAT3–ERH interactions was significantly diminished in STAT3-SA expressing cells, confirming the requirement for pS–STAT3 to form complexes with Pontin and ERH (Fig. 6F and G). STAT3–ERH interactions were predominantly localized to the nucleus of MKN-28 STAT3-WT cells (Fig. 6F), consistent with an exclusive nuclear functional role for ERH (37). By contrast, Pontin is localized to both the nucleus and cytoplasm (38), which likely accounts for the nuclear and cytoplasmic STAT3–Pontin interactions observed in MKN-28 STAT3-WT cells (Fig. 6F). These observations support ERH and Pontin as in vivo interacting partners of pS–STAT3 to augment the transcriptional induction of STAT3 target genes in gastric cancer.

The translational relevance of our findings was validated in human gastric cancer patients, whereby ERH, RUVBL1, and SOCS3 expression was significantly upregulated in 85% (17/20), 70% (14/20), and 70% (14/20) of tumor versus matched nontumor tissues, respectively, and increased expression correlated with worse overall patient survival, albeit not significant for ERH (Fig. 7A–C; Supplementary Fig. S7A). Furthermore, mRNA levels for ERH and RUVBL1 demonstrated significant positive correlations with those for SOCS3, and with pS–STAT3 protein levels (Fig. 7D and E). Interestingly, the increased expression of these genes was independent of disease stage (Supplementary Fig. S7B). Importantly, immunofluorescence staining revealed a pronounced colocalization of pS–STAT3 with ERH and Pontin in human gastric cancer tumor and matched nontumor tissues, with dual pS–STAT3/ERH or pS–STAT3/Pontin staining higher in tumors (Fig. 7F–H). The colocalization of pS–STAT3 with ERH and Pontin was also confirmed in tumors of the gp130^F/F gastric cancer model (Supplementary Fig. S7C and S7D), thus supporting ERH and Pontin as in vivo interacting partners of pS–STAT3 in gastric cancer.

In cancer, emerging studies have suggested an essential requirement for pS–STAT3 by oncogenic Ras-addicted malignancies that is independent of its canonical nuclear (i.e., gene transcription) role, but rather is dependent on mitochondrial function via augmenting the mitochondrial electron transport chain to support aerobic glycolysis over oxidative phosphorylation for the production of cellular ATP (the “Warburg effect”; refs. 8, 11, 27). However, in contrast to Ras-driven oncogenesis, the relative role of pS–STAT3 nuclear and mitochondrial activities to the molecular pathogenesis of many epithelial malignancies driven by excessive cytokine secretion or tyrosine kinase activity, such as gastric cancer, is unknown.

Here, we define an obligatory and unprecedented role for pS–STAT3 in the initiation and maintenance of gastric cancer. Specifically, in the STAT3-driven gp130^F/F gastric cancer model, we demonstrate that pS–STAT3 deficiency suppresses gastric tumorigenesis and is associated with the attenuation of proliferative, inflammatory, angiogenic, and metaplastic processes. Furthermore, our bone marrow reconstitution studies, in concert with human gastric cancer (epithelial) cell line xenografts, imply that the requirement for pS–STAT3 is intrinsic to the nonhematopoietic (i.e., epithelial) compartment, consistent with our previous data for STAT3-dependency by the gastric tumor epithelium (18, 27). Notably, the marked tumor suppressive effect, along with improved survival rate, of ablating pS–STAT3 in gp130^F/F:Stat3^SA/SA mice was comparable to that achieved by monoallelic deletion of Stat3 (gp130^F/F:Stat3^+/−), which also phenocopies gp130^F/F mice lacking the capacity to transduce IL11 signals (18). In this regard, microarray-based transcriptomic analyses assigned the protumorigenic actions of pS–STAT3 to the transcriptional modulation of nuclear encoded STAT3 target genes—implicated in numerous oncogenic cellular processes, independent of mitochondrial function and metabolism—among which includes IL11, the primary upstream activator of STAT3 in gastric cancer (18, 27).

SOCS3 is a negative regulator of STAT3 transcriptional function, and epigenetic silencing of SOCS3 is observed in some cancer types (e.g., endometrial). However, in gastric cancer, the transcriptional upregulation of SOCS3 is likely reflecting elevated STAT3 activation (18, 19, 27), and high SOCS3 mRNA levels significantly correlate with poor survival outcomes (Fig. 7C). Therefore, SOCS3 was used in our current study as a robust molecular read-out of STAT3 transcriptional activity in gastric cancer. In this regard, we demonstrate that the magnitude of oncogenic STAT3 transcriptional activity is enhanced by the capacity of pS727 to augment RNA Pol II-mediated transcript elongation of STAT3-target genes (e.g., SOCS3), rather than facilitate RNA Pol II (or STAT3) promoter recruitment and transcription initiation. This is supported by the report that pS727 did not influence the capacity of STAT3 to accumulate on the SOCS3 promoter (14). Moreover, our demonstration that RNA Pol II occupies the SOCS3 promoter in the absence IL11-stimulated STAT3 signaling is consistent with the rapid transcriptional induction of many primary response genes associated with oncogenesis, including those upregulated by STAT3 (e.g., Myc), whose promoters are preloaded with a proximally paused RNA Pol II that is primed for the transition from RNA Pol II transcript initiation to elongation (39).

Another key finding of our study was the identification of Pontin and ERH as pS–STAT3–dependent interacting partners that augment the transcriptional activation of STAT3 target genes in gastric cancer. Pontin is a highly conserved AAA+ ATPase family member, and whereas STAT3 can interact with other ATPases, in particular BRG1, which is critical for Ser2 phosphorylation of RNA Pol II and transcriptional elongation, unlike our current study, the role of phosphorylated STAT3 in the interaction with BRG was not investigated (40). Pontin has been shown to directly modulate gene transcription, as well as participate in protein super complexes to regulate diverse cellular processes including chromatin remodeling (INO80 complex), histone modification (NuA4/TIP60 complex), snoRNP biogenesis, and RNA Pol II assembly (R2TP complex; refs. 38, 41–44). Interestingly, Pontin can interact with the transcription activation domain of STAT2 to facilitate maximal RNA Pol II recruitment and initiation at the promoters of IFN-stimulated genes (45, 46). This contrasts our current finding supporting a role for Pontin (via its interaction with pS–STAT3) in the RNA Pol II-driven transcription elongation phase through STAT3 target genes, suggesting that the role for Pontin in augmenting STAT-driven gene transcription may be dependent on cellular context, the upstream STAT-activating stimuli and/or the specific STAT family member. Indeed, we show here that the genetic ablation of Pontin (and ERH) in human gastric cancer cells had no effect on the transcription of the STAT1-regulated gene SOCS1. With respect to ERH, its diverse molecular functions include regulation of transcription, cell cycle, RNA splicing, and pyrimidine metabolism (47). Notably, ERH can interact with SPT5 and FCP1, essential factors that modulate the processing capacity of RNA Pol II, and thus transcriptional elongation rates (48–50). Therefore, the actions of Pontin and ERH in promoting transcription elongation of RNA Pol II are consistent with our data assigning such a transcriptional role for pS–STAT3. Furthermore, because STAT3 target genes affect numerous oncogenic cellular processes, such as proliferation, apoptosis, invasion, and angiogenesis (2, 4), the increased expression of RUVBL1 and ERH that we observed in human gastric cancer would likely further augment the oncogenic transcriptional capacity of STAT3, thus favoring a more aggressive tumor phenotype in patients.

Hyper-activation of IL11/gp130/STAT3 signaling in gastric cancer recapitulates what is observed in many human tumors where oversupply of IL6 family cytokines converge, primarily via JAKs, to maintain pY-STAT3 levels (4, 18, 19, 51). However, we reveal here that pS–STAT3 is independent from IL11-driven upregulation of pY-STAT3, as evidenced by pS–STAT3 levels remaining unaltered in in vitro and in vivo gastric cancer models following either IL11 treatment, IL11R ablation, or JAK inhibition. Rather, constitutive pS–STAT3 is observed at comparable levels in mouse/human gastric nontumor/tumor tissues, and among human gastric cancer cell lines, which may be a consequence of at least one of the diverse upstream kinases that drive serine phosphorylation of STAT3, including ERK, p38, and JNK MAPKs, protein kinase C delta, glycogen synthase kinase 3α/β, and cyclin-dependent kinase-1, being engaged at all times (12, 52–56). We therefore speculate that STAT3 tyrosine and serine phosphorylation events may have coevolved to confer a broader spectrum of transcriptional potentials capable of graded regulation. Moreover, we hypothesise, at least in gastric cancer, that IL11 signaling within the gastric epithelium represents a response to tissue injury, infection, and/or inflammation, and engages JAK-bound gp130 receptor complexes to induce pY-STAT3. This tightly controlled molecular switch then facilitates STAT3 nuclear translocation and DNA binding affinity, thus licensing STAT3 with a broad transcriptional capacity. Upon entering the nucleus, basal pS–STAT3 then recruits and exploits various coactivators, such as Pontin and ERH, to augment the transcriptional elongation of STAT3 target genes by RNA Pol II. Accordingly, as evidenced by gp130^F/F and human gastric cancer cell xenograft models, in the absence of pS–STAT3, the induction of pY-STAT3 alone is insufficient to attain the threshold of STAT3 transcriptional output that is necessary for gastric tumorigenesis.

In summary, our current study demonstrates that STAT3-driven gastric cancer is dependent upon the capacity of pS727 to maximise RNA Pol II transcription rates of STAT3-target genes by facilitating interaction of STAT3 with Pontin and ERH, identified here as hitherto unknown transcriptional coactivators. Therefore, identifying key sites governing the interaction between STAT3 and ERH or Pontin, as well as the mechanism(s) by which ERH and Pontin regulate STAT3-driven gene transcription, provides opportunities for pharmacologic targeting of STAT3 in gastric cancer, and other STAT3-driven cancers (e.g., colorectal, pancreatic), which thus far has been problematic (1, 6). Although our data (Fig. 7F) suggest that the targeted disruption of pS–STAT3/Pontin and pS–STAT3/ERH interactions might occur in normal (i.e., nontumor) and tumor tissues, nonetheless, the specific blockade of S727 independently of Y705 in STAT3 represents an alternative approach to selectively suppress the oncogenic transcriptional activity of pS–STAT3 in gastric cancer, while simultaneously preserving the homeostatic functions (e.g., mucosal wound healing) of pY-STAT3. An advantage of such a therapeutic strategy is to maintain overall gastric function, and thus minimize nonspecific toxicities. In the absence of current pS–STAT3-specific inhibitors, the identification of the kinase(s) responsible for gastric STAT3 serine phosphorylation will foster the evaluation of existing drugs targeting candidate kinases (e.g., MEK/MAPK inhibitors, Trametinib and Binimetinib) as effective indirect pS–STAT3 inhibitors, and thus anticancer agents, in gastric cancer.

No potential conflicts of interest were disclosed.

Conception and design: J.J. Balic, D.J. Garama, D.J. Gough, B.J. Jenkins

Development of methodology: J.J. Balic, D.J. Garama, D.J. Gough, B.J. Jenkins

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J.J. Balic, D.J. Garama, M.I. Saad, L. Yu, A.C. West, A.J. West, P.S. Bhathal, D.J. Gough, B.J. Jenkins

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.J. Balic, D.J. Garama, M.I. Saad, L. Yu, A.C. West, A.J. West, T. Livis, D.J. Gough, B.J. Jenkins

Writing, review, and/or revision of the manuscript: J.J. Balic, D.J. Gough, B.J. Jenkins

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J.J. Balic, D.J. Garama, A.C. West T. Livis, P.S. Bhathal, D.J. Gough, B.J. Jenkins

Study supervision: D.J. Gough, B.J. Jenkins

Other (processing of tissue samples and histologic interpretation): P.S. Bhathal

This work was supported by a research grant awarded by the National Health and Medical Research Council (NHMRC) of Australia to B.J. Jenkins, as well as the Operational Infrastructure Support Program by the Victorian Government of Australia. J.J. Balic was supported by an Australian Postgraduate Awards scholarship from the Australian Government. A.C. West was supported by an NHMRC Early Career Fellowship and D.J. Gough by an NHMRC Career Development Fellowship and a grant from the United States Department of Defence. B.J. Jenkins was supported by an NHMRC Senior Medical Research Fellowship.

We thank N. Williamson and the Bio21 Mass Spectrometry and Proteomics Facility, the Monash Health Translational Precinct Medical Genomics Facility, and the Monash Histology Platform for providing core and technical services.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.