Somatic mutations activating STAT3 in human inflammatory hepatocellular adenomas

Among 114 screened hepatocellular adenomas, we identified 7 STAT3 mutations in 6 tumors (Table I and Fig. S1, A and B). All of these adenomas were inflammatory (6 out of 75 IHCA), and no additional somatic mutations were identified in hepatocyte nuclear factor 1α (HNF1A), catenin (cadherin-associated protein), β1 (CTNNB1), or IL6ST, genes that are recurrently altered in IHCA and HCA (Bluteau et al., 2002; Zucman-Rossi et al., 2006; Rebouissou et al., 2009). Thus, STAT3 IHCA mutations are exclusive from all other known genetic alterations in hepatocellular tumors. STAT3-mutated IHCA were associated with obesity (n = 4) and/or alcohol abuse (n = 3), and multiple (≥4) nodules were present in half of these cases. All STAT3 mutations were somatic, as they were not observed in adjacent nontumor liver tissues that were steatotic in 5/6 cases. All mutations were also monoallelic and led to amino acid substitutions in 5 cases or to in-frame insertions of 1–4 aa in the remaining 2 cases. In one case (#379), we identified 3 nucleotide mutations that led to amino acid substitutions at codons 502 and 658. Sequencing cloned STAT3 cDNA from this tumor showed that the three different nucleotide mutations were carried by the same allele. We also showed in three different cases (#379, #966, and #1351) that the mutations were harbored by all of the isoforms of STAT3, including splicing of nucleotides 2099–2101 and 2145–2194 (STAT3β) that lead to p.Ser701del and p.Thr716PhefsX8, respectively (Fig. S1 C; Schaefer et al., 1995). Sequencing STAT3 RT-PCR products in mutated tumors showed that all cases expressed the normal and mutated STAT3 alleles at comparable levels (Fig. S1, A and B). Thus, the expression of one mutated STAT3 allele is a rare but recurrent genetic event in IHCA that is selected for during benign hepatocellular tumorigenesis.

Four of the seven identified mutations were located in the domain of STAT3 (residues 585–688) that shares homology with Src homology 2 (SH2) domains (Fig. 1 A). The STAT3 SH2 domain mediates STAT3 dimerization via binding of phosphotyrosine residue Y705 (Shuai et al., 1994; Wen et al., 1995). Moreover, three of these mutations clustered in a hotspot in the SH2 domain at codons 657 and 658. Interestingly, introduction of cysteine at residues 662 and 664 promotes STAT3 dimerization and creates a constitutively active transcription factor (Stat3-C; Bromberg et al., 1999). The fourth mutation, Y640F, was in the PYTK motif conserved in STAT1, STAT2, and STAT3. Interestingly, substitution of the corresponding Y631 to phenylalanine in STAT2 promotes type I IFN signaling (Scarzello et al., 2007; Constantinescu et al., 2008). The three other mutations found in IHCA were distributed throughout the protein: the altered leucine-78, which contributes to dimerization (Chen et al., 2003); glutamate-166, which is part of helix α1 involved in the interaction with gp130 (Zhang et al., 2000a); and aspartate-502, which is located in the α-helical “connector” domain. Remarkably, none of the mutations identified in IHCA were similar to the 150 germline-inactivating STAT3 heterozygous mutations described in patients with Job’s syndrome, which causes hyper-IgE syndrome (Holland et al., 2007; Minegishi et al., 2007).

To investigate possible functional consequences of these STAT3 IHCA mutations, we introduced the seven different IHCA mutations into a STAT3α (full-length) cDNA and analyzed their effects in Hep3B cells, a hepatocellular carcinoma line that activates the acute inflammatory phase after treatment with IL-6 (Coulouarn et al., 2005). The results were also validated in two additional hepatocellular cell lines (HepG2 and Huh7). As previously described (Bromberg et al., 1999), in the absence of IL-6 ligand and serum, overexpression of WT STAT3 (WT) alone was not sufficient to activate a STAT3 reporter (Fig. 1 B) or the downstream acute-phase inflammatory genes (Fig. 1 C). In contrast, the six IHCA mutants and the mouse Stat3-C mutant (Bromberg et al., 1999) activated the STAT3 reporter in the absence of IL-6 and induced targets typical of the acute inflammatory response, including CRP and SOCS3 (Fig. 1, B and C). Moreover, activation of STAT3 was proportional to the amount of the transfected mutant STAT3 (Fig. 1 D). The last IHCA mutant D502Y/K658Y, identified in tumor #379, has two mutations on the same allele. The D502Y mutation did not impart constitutive activity, whereas K658Y activated the inflammatory response. Additionally, the dual mutant (D502Y + K658Y) produced a hyperactive STAT3 mutant compared with K658Y alone (Fig. 1 E). Thus, mutation at D502 alone is not sufficient to activate STAT3, but cooperates with the mutation at K658 present in the SH2 domain.

STAT3 activation requires phosphorylation of tyrosine 705, dimerization, and translocation to the nucleus. Accordingly, in STAT3-mutated IHCA, immunohistochemical analyses demonstrated marked overexpression of phospho-Tyr705-STAT3 in the nucleus of these tumor hepatocytes (Fig. 2 A). Similarly, in Hep3B-transfected cells, we showed that the STAT3 mutants L78R, E166Q, Y640F, and G656_Y657insF were constitutively phosphorylated on Tyr705 and translocated to the nucleus, whereas overexpressed WT was unphosphorylated and cytoplasmic in the absence of cytokine (Fig. 2, B–D). STAT3 IHCA mutants did not exhibit increased phosphorylation of serine 727, a second residue involved in STAT3 activation (Wen et al., 1995), when compared with IL-6–activated WT (Fig. S2 A). Moreover, in STAT3 Y640 and G656 mutants, substitution of tyrosine 705 with phenylalanine totally abrogated activation of the inflammatory response. Thus, tyrosine phosphorylation is essential for activation of IHCA mutant STAT3 (Fig. 2 E).

A critical step in STAT3 activation is the formation of homodimers through reciprocal phospho-Tyr705–SH2 domain interactions. To assess dimerization by IHCA STAT3 mutants, we performed coimmunoprecipitation assays with epitope-tagged STAT3 proteins. The Y640F mutant homodimerized or heterodimerized with WT independent of IL-6 (Fig. 3 A). Moreover, mutant homodimers were highly phosphorylated at tyrosine 705. In contrast, homodimers of WT were only faintly detected and were not phosphorylated on tyrosine 705. As expected, IL-6 treatment does not enhance dimer formation, but slightly increases tyrosine 705 phosphorylation (Haan et al., 2000). In addition, overexpression of WT impaired the activity of the STAT3 mutant Y640F in a dose-dependent manner (Fig. 3 B). Thus, WT can compete with the mutant STAT3 and impair the formation of active IHCA mutant homodimers. Interestingly, analysis of the STAT3 homodimer structure (Becker et al., 1998) showed that 4 out of 6 STAT3 mutants targeted amino acids (Y640, K658, Y657, and Y657-M660) that reside near the transcription activation domain (TAD; residues 701–722; Fig. 3 C and Fig. S2 B). For example, Tyr-640 engages in hydrophobic interactions with twofold related Phe-710 and residues 656–659 pack against TAD residues 711–716 (Fig. 3 C and Fig. S2 B). These modifications could thus enhance mutant STAT3 function, and they must also somehow facilitate phosphorylation of tyrosine 705 by kinases in hepatocytes, which would then drive dimerization and nuclear localization of mutant STAT3. After cytokine exposure, phosphorylated STAT3 homodimers translocate to the nucleus where they bind to specific IL-6–responsive elements to activate the transcription of target genes (Hattori et al., 1990). In Hep3B cells, nuclear Y640 and G656 STAT3 mutants bind to recognition elements found in endogenous STAT3 target genes in the absence of IL-6, whereas, as expected, WT was defective in binding in the absence of IL-6 (Fig. 3 D).

To investigate the role of the IL-6–JAK pathway in the activation status of IHCA STAT3 mutants, we tested the effects of increasing doses of IL-6 on their activity. Indeed, even low doses of IL-6 (1 ng/ml) dramatically augmented the constitutive activity of the Y640 STAT3 mutant compared with the WT (Fig. 4 A). This hypersensitive response to IL-6 signaling was confirmed for L78, E166, G656, and Y657 STAT3 mutants (Fig. 4 B and Fig. S2 C). Together with gp130, JAK1 is known as the major kinase that phosphorylates and activates STAT3 in IL-6 signaling. Accordingly, silencing JAK1 using small interfering RNA (siRNA) abolished the hyperstimulation of the STAT3 mutants by IL-6 (Fig. 4 B).

In the absence of IL-6 signaling, silencing of gp130 or JAK2 did not affect the constitutive activity of IHCA STAT3 mutants (Fig. 4, C and D). In contrast, JAK1 knockdown by siRNA partially impaired constitutive IHCA STAT3 mutant activity (Fig. 4 E). Similarly, AG490, an inhibitor that disables the kinase activity of both JAK1 and JAK2; Ruxolitinib, a selective inhibitor of Jak1; and curcumin, a STAT3 inhibitor, partially decreased the activity of IHCA STAT3 mutants (Fig. S3, A and B). In contrast, treatment with SrcI-1 kinase inhibitor dramatically impaired both the constitutive and IL-6–induced activation of IHCA STAT3 mutants (Fig. 4 F and Fig. S3 C). Two other Src kinase inhibitors also impaired the activity of IHCA STAT3 mutants (Fig. S3, D and E). Thus, phosphorylation and activation of mutant STAT3 does not require gp130 or JAK2 but requires JAK1 and Src signaling. Depending on cell context, STAT3 can be activated directly by Src family kinases or via Jak1 (in NIH 3T3 cells) as an intermediate kinase between Src and STAT3 (Zhang et al., 2000b). Our findings suggest that in hepatocellular cell lines, the requirement of Jak1 is necessary for maximal activation of IHCA STAT3 mutants, whereas Src is essential for their constitutive activation.

In vivo, expression analyses of 16 IHCA revealed no significant differences in 28 target genes of IL-6 and IFN signaling pathways in STAT3-mutated adenomas (n = 5) versus gp130-mutated IHCA (n = 11; Fig. 4 G). Thus, the functional effects of mutating STAT3 and gp130 in human hepatocellular adenomas appear redundant.

In the liver, activation of the IL-6–STAT3 signaling is necessary for the acute-phase response during inflammation (Lütticken et al., 1994) and liver regeneration, particularly after transplantation (Cressman et al., 1996). The connection between inflammation and tumorigenesis has been extensively investigated over the past decade (Grivennikov and Karin, 2010). Depending of the model, inflammation can trigger tumor initiation or function to enhance tumor progression and/or metastasis. During these processes and in numerous tumor types, STAT3 activation has been recognized as a major oncogenic event. However, in our cohort, STAT3-activating mutations were found in 8% of the inflammatory hepatocellular adenoma, but were not present in a series of 218 malignant liver tumors, including hepatocellular carcinomas and cholangiocarcinomas (n = 11). Thus, STAT3-activating mutations appear to be largely or quasiexclusively associated with benign proliferation of hepatocytes in human.

Collectively, our findings show that most of the genetic alterations of HCA subtypes (HNF1A inactivation, β-catenin activation, or activation of STAT3 via mutation of STAT3 or gp130), are associated with benign hepatocyte tumorigenesis, where there remains a largely controlled rate of tumor cell proliferation. At present, only β-catenin–activating mutations are associated with malignant transformation to hepatocellular carcinoma (Zucman-Rossi et al., 2006), and here there are likely additional cooperating mutations that lead to frank malignancy. Regardless, to our knowledge this is the first identification of somatic mutation of STAT3 in human tumors, and these findings underscore the important role of STAT3 activation in hepatocellular benign tumorigenesis and reveal a new mechanism of recurrent STAT3 activation.

A series of 364 hepatocellular tumors, including 114 HCAs, 218 common hepatocellular carcinomas (HCCs), 11 cholangiocarcinomas and 20 hepatocholangiocarcinomas were previously collected (according to French ethical guidelines) and characterized (Bioulac-Sage et al., 2007; Zucman-Rossi et al., 2006). In brief, patients with HCA were predominantly female (80%) with a mean age of 39 yr; among them, 13% presented mutations of CTNNB1 gene, 19% presented mutations of HNF1α, 26% displayed IHCA-associated IL6ST mutations, 39.5% showed IHCA without IL6ST mutations, and 9.6% were nonclassified adenomas. Patients with HCC were predominantly male (79%), with a mean age of 64 yr, and 34.3% of these presented with cirrhosis. The underlying risk factors of patients with HCC were chronic hepatitis B (19%), chronic hepatitis C (25%) and heavy alcohol intake (38%). Patients with cholangiocarcinoma were predominantly male (72%) and were slightly younger (mean age, 61 yr) than patients with common HCC. Patients with hepatocholangiocarcinoma were also predominantly male (89%) and had a median age of 61 yr. All patients have given consent according to French law, and the ethical committee of Saint Louis Hospital approved the study.

Tumor and nontumor liver samples were frozen immediately after surgery or biopsy and stored at –80°C. Tissues samples were also included in 10% formalin, paraffin-embedded and stained with hematin-eosin and Masson’s trichrome. The diagnosis of HCA, HCC, cholangiocarcinoma, and hepatocholangiocarcinoma was performed using established histological criteria (International Working Party, 1995).

qRT-PCR was performed as previously described (Rebouissou et al., 2009) using predesigned primers and probe sets from Applied Biosystems (Table S1). Ribosomal 18S RNA (R18S) was used to normalize expression data and the 2-ΔΔCT method was applied.

DNA sequencing was performed as previously described (Rebouissou et al., 2009) using primers described in Table S2. All HCA samples were previously sequenced for CTNNB1 (exons 2–4), HNF1A (exons 1–10), and IL6ST (exon 6). All HCC samples were previously sequenced for CTNNB1 (exons 2–4) and P53 (exons 2–10) as previously described (Zucman-Rossi et al., 2006; Rebouissou et al., 2009). All mutations were validated by sequencing a second independent PCR product on both strands. In all cases, the somatic origin of the mutation found in tumor was verified by sequencing the corresponding adjacent, normal liver sample.

A full-length STAT3 open reading frame cloned in pCMV6-XL4 vector was purchased from OriGene (SC124165; available from GenBank/EMBL/DDBJ under accession no. NM_139276). Mutagenesis reactions were performed using the QuickChange XL site-directed mutagenesis kit (Stratagene) using primers described in Table S3. All constructs were verified by sequencing. The pCMV6-XL4 expression vector plasmid was used as a control. Stat3-C plasmid was purchased from Addgene, Inc. (plasmid 8722).

Hep3B, Huh7, and HepG2 cells (American Type Culture Collection) were grown in DME containing 10% FCS supplemented with penicillin (100 U/ml) and streptomycin (0.1 mg/ml). For transfections, cells were plated 16 h before transfection to produce monolayers that were 60% confluent, and these were transfected by using either Lipofectamine LTX (plasmid alone) or Lipofectamine 2000 (plasmid and siRNA) according to the manufacturer’s instructions (Invitrogen). Transfection efficiency was monitored by measuring the level of WT and mutated STAT3 mRNA using qRT-PCR or Western blot analysis. The same STAT3 expression level was observed in WT and IHCA mutant STAT3-transfected cells. For luciferase assays, Hep3B cells were co-transfected with a STAT3 luciferase reporter expressing a firefly luciferase reporter gene that contains three copies of a STAT3 consensus binding site linked to a minimal thymidine kinase promoter (provided by H. Gascan, Institut National de la Santé et de la Recherché Médicale UMR564, Angers, France; pSIEM-Luc 1 µg; Coqueret and Gascan, 2000) and the expression plasmid for WT or IHCA mutant STAT3 (1 µg). 1 d after transfection, cells were starved in a free-serum medium for 3 h, and then left untreated or treated with IL-6 (100 ng/ml) for 3 h. They were then lysed and the luciferase activity was determined according to the manufacturer’s recommendations (Promega). The activities were normalized to the protein amounts in each cell lysate. All analyses were performed in triplicate.

The efficiency of knockdown using siRNAs was monitored by measuring the level of JAK1, JAK2, or IL6ST mRNA using qRT-PCR (Fig. S3 F) or by Western blot analysis. Knockdown efficiency was >80%. JAK1 siRNA (S7646), JAK2 siRNA (S7651), and IL6ST (gp130) siRNA (S229006) were purchased from Applied Biosystems; siRNA control (Block-IT Alexa Fluor Red) was purchased from Invitrogen.

Cells were exposed, in serum-free medium, for 6 h to Tyrphostin AG490 (Sigma-Aldrich), 9 h to Src inhibitor-1 (Sigma-Aldrich), Src inhibitor-5 (JS Res Chemical Trading), or PP2 (Sigma-Aldrich); for 18 h to Ruxolitinib (INCB-018424; JS Res Chemical Trading); and for 24 h to Curcumin (Sigma-Aldrich). In the last 3 h of treatment, cells were stimulated with 100 ng/ml IL-6, or not, as indicated.

Western blot analyses were performed as previously described (Rebouissou et al., 2009) using antibodies specific for STAT3 (Cell Signaling Technology; 1:500) and phospho-STAT3 Tyr705, (Cell Signaling Technology; 1:200). For dimerization assays, cell lysates were incubated with Immobilized Protein G agarose (Thermo Fisher Scientific) and anti-flag (Cell Signaling Technology; 1:50) or anti-myc (Cell Signaling Technology; 1:200) antibody, at 4°C overnight, before Western blot analysis.

Immunohistochemistry was performed using a Dako autostainer, on paraffin sections of 10% fixed tumor tissue using a monoclonal anti–phospho-STAT3 Tyr705 (Cell Signaling Technology; 1:50). For each immunohistochemical procedure, antigen retrieval was performed in citrate buffer; detection was amplified by the Dako Envision system.

1 d after transfection, cells were starved in a serum-free medium overnight, and the TransFactor Extraction kit (Takara Bio, Inc.) was used to prepare nuclear extracts following the manufacturer’s protocol. Nuclear extracts were analyzed for STAT3 DNA-binding activity using the TransFactor Universal STAT3-specific kits (Takara Bio, Inc.), which is an ELISA-based method.

24 h after transfection, cells were fixed in 4% formaldehyde for 15 min, washed with PBS, and permeabilized with 0.1% Triton X-100 for 15 min. Cells were saturated with 2% BSA for 30 min, washed with PBS, and then incubated with primary antibody overnight at 4°C. After 3 washes with PBS, cells were incubated with secondary antibodies for 1 h. The slides were washed, and then mounted with VECTASHIELD Mounting Medium with DAPI (Vector Laboratories). Immunofluorescence images were obtained using an HBO 100 optic microscope (Carl Zeiss, Inc.). All data from one experiment were collected at the same intensity and background level. The following monoclonal antibodies were used: rabbit anti–phospho-STAT3 Tyr705 (Cell Signaling Technology; 1:200), mouse anti-total STAT3 (Cell Signaling Technology; 1:100). The secondary antibodies were anti–mouse (Invitrogen; 1:200) and anti–rabbit (Invitrogen; 1:200).

Fig. S1 shows the somatic DNA alterations identified on STAT3 gene, and distribution of mutations according to the different isoforms. Fig. S2 shows serine 727 phosphorylation of STAT3 mutants. Fig. S3 shows the effect of inhibitors on STAT3 IHCA mutant activation. Table S1 lists the TaqMan predesigned gene expression assays used for qRT-PCR analyses. Table S2 lists the primers used in PCR and sequencing. Table S2 lists the primers used in in site-directed mutagenesis.

We warmly thank Gaelle Cubel for her participation to this work and Hugues Gascan who kindly provided the pSIEM-Luc plasmid. We also thank Jean Saric, Christophe Laurent, Brigitte Le Bail, Anne Rullier, Antonio Sa Cunha for contributing to the tissue collection (CHU Bordeaux) and all the participant to GENTHEP network (Génétique des tumeurs hépatiques).

This work was supported by the association pour la recherche Contre le Cancer (ARC, grant n°3194), Institut National de la Santé et de la Recherche Médicale (Réseaux de recherche clinique et réseaux de recherche en santé des populations), Collection Nationale des Carcinomes Hépatocellulaires, the Ligue Nationale Contre le Cancer (“Cartes d’identité des tumeurs” program) and the BioIntelligence collaborative program. C.P., M.A. and J-C.N. are supported by a fellowship from the MENRT, Inca and ARC, respectively. T.I. is supported by the National Institutes of Health grants GM071596, AI055894, and AI067949.

The authors have no conflicting financial interests