STAT3 Expression, Activity and Functional Consequences of STAT3 Inhibition in Esophageal Squamous Cell Carcinomas and Barrett’S Adenocarcinomas

ORIGINAL ARTICLE STAT3 expression, activity and functional consequences of STAT3 inhibition in esophageal squamous cell carcinomas and Barrett’s adenocarcinomas

S Timme1,11, S Ihde1,11, CD Fichter1,2,11, V Waehle1, L Bogatyreva3, K Atanasov1, I Kohler1, A Scho¨ pﬂin1, H Geddert4, G Faller4, D Klimstra5, L Tang5, T Reinheckel6,7,8, D Hauschke3, H Busch6,9,10, M Boerries6,9,10, M Werner1,8,9,10 and S Lassmann1,7,8,9,10

Signal transducer and activator of transcription 3 (STAT3) is altered in several epithelial cancers and represents a potential therapeutic target. Here, STAT3 expression, activity and cellular functions were examined in two main histotypes of esophageal carcinomas. In situ, immunohistochemistry for STAT3 and STAT3-Tyr705 phosphorylation (P-STAT3) in esophageal squamous cell carcinomas (ESCC, n ¼ 49) and Barrett’s adenocarcinomas (BAC, n ¼ 61) revealed similar STAT3 expression in ESCCs and BACs (P ¼ 0.109), but preferentially activated P-STAT3 in ESCCs (P ¼ 0.013). In vitro, strong STAT3 activation was seen by epidermal growth factor (EGF) stimulation in OE21 (ESCC) cells, whereas OE33 (BAC) cells showed constitutive weak STAT3 activation. STAT3 knockdown significantly reduced cell proliferation of OE21 (P ¼ 0.0148) and OE33 (P ¼ 0.0243) cells. Importantly, STAT3 knockdown reduced cell migration of OE33 cells by 2.5-fold in two types of migration assays (P ¼ 0.073, P ¼ 0.015), but not in OE21 cells (P ¼ 0.1079, P ¼ 0.386). Investigation of transcriptome analysis of STAT3 knockdown revealed a reduced STAT3 level associated with significant downregulation of cell cycle genes in both OE21 (Po0.0001) and OE33 (P ¼ 0.01) cells. In contrast, genes promoting cell migration (CTHRC1) were markedly upregulated in OE21 cells, whereas a gene linked to tight-junction stabilization and restricted cell motility (SHROOM2) was downregulated in OE21 but upregulated in OE33 cells. This study shows frequent, but distinct, patterns of STAT3 expression and activation in ESCCs and BACs. STAT3 knockdown reduces cell proliferation in ESCC and BAC cells, inhibits migration of BAC cells and may support cell migration of ESCC cells. Thereby, novel STAT3-regulated genes involved in ESCC and BAC cell proliferation and cell migration were identified. Thus, STAT3 may be further exploited as a potential novel therapeutic target, however, by careful distinction between the two histotypes of esophageal cancers.

Oncogene (2014) 33, 3256–3266; doi:10.1038/onc.2013.298; published online 5 August 2013 Keywords: STAT3; esophageal cancer; inhibition; cell proliferation; cell migration

INTRODUCTION Irrespective of its cause, deregulation of STAT3 was reported 7,8 9 Signal transducer and activator of transcription 3 (STAT3) is a 92- for several epithelial cancers, such as lung, head and neck, 10,11 12,13 kDa large transcription factor, whose gene is located on stomach and colorectal carcinomas. In fact, inhibition of chromosome 17q21. STAT3 becomes activated upon phosphor- STAT(3) signaling is currently discussed for therapeutic ylation on critical tyrosine residues upon cytokine and/or growth intervention, alone or in combination with inhibitors to factor receptor stimulation. Phosphorylation of STAT3 tyrosine 705 upstream signaling molecules, such as receptor tyrosine 6 results in STAT(3) hetero- or homodimerization, nuclear transloca- kinases or janus kinases Jak1 or Jak2. Indeed, direct STAT3 tion of STAT(3) dimers and binding of STAT(3) dimers to their inhibition was successful in vitro by interference with STAT3 respective target gene promoters.1–3 phosphorylation and/or dimerization using peptidomimetics or 14 In normal cells, STAT3 expression and activity is under tight small molecules targeting the SH2 domain, or by interference control to ensure physiological cell proliferation, survival, differ- with nuclear translocation and/or DNA binding using 15,16 entiation and motility. In contrast, malignant transformed cells, decoy oligonucleotides. In addition, synergistic effects of respective human cancers, frequently exhibit STAT3 ‘over’expres- direct STAT3 inhibitors with the epidermal growth factor sion and constitutive activation, which contribute to increased receptor (EGFR) drug cetuximab were demonstrated in cell lines 17,18 cancer cell proliferation and cell migration.4–6 This may also be derived from squamous cell carcinomas of the head and neck triggered by alterations in upstream cytokines and/or (receptor) or in cell lines and xenograft models of pancreatic 19,20 tyrosine kinases. adenocarcinomas.

1Institute of Clinical Pathology, University Medical Center, Albert-Ludwigs-University, Freiburg, Germany; 2Faculty of Biology, Albert-Ludwigs-University, Freiburg, Germany; 3Institute of Medical Biometry and Medical Informatics, University Medical Center, Albert-Ludwigs-University, Freiburg, Germany; 4Department of Pathology, St Vincentius Kliniken, Karlsruhe, Germany; 5Department of Pathology, Memorial Sloan-Kettering Cancer Center, New York, NY, USA; 6Institute of Molecular Medicine and Cell Research, University Medical Center, Albert-Ludwigs-University, Freiburg, Germany; 7BIOSS Centre for Biological Signalling Studies, Albert-Ludwigs-University, Freiburg, Germany; 8Comprehensive Cancer Center Freiburg, University Medical Center, Albert-Ludwigs-University, Freiburg, Germany; 9German Cancer Consortium (DKTK), Heidelberg, Germany and 10German Cancer Research Center (DKFZ), Heidelberg, Germany. Correspondence: Professor Dr S Lassmann, Institute of Clinical Pathology, University Medical Center, Breisacherstrasse 115A, 79106 Freiburg, Germany. E-mail: [email protected] 11These authors contributed equally to this work. Received 18 October 2012; revised 13 June 2013; accepted 13 June 2013; published online 5 August 2013 STAT3 expression and function in esophageal cancer S Timme et al 3257 The two main histotypes of esophageal cancers—esophageal As ESCCs and BACs display distinct patterns of carcinogenesis squamous cell carcinoma (ESCC) and Barrett’s adenocarcinoma and behavior, STAT3 expression and activation were assessed in (BAC)—are highly aggressive tumors with a very poor prognosis ESCCs versus BACs. These analyses showed that STAT3 protein and low 5-year survival rate, especially if diagnosed at later expression was statistically similar in ESCCs and BACs (P ¼ 0.109; stages.21–23 ESCCs and BACs evolve via distinct pathways of Table 1, Figure 1a, middle column), although ESCCs more carcinogenesis, with so far few alterations for targeted therapeutic frequently appeared to be STAT3 positive (91.2%, 41/45 of cases) intervention having been identified. To date, there is little than BACs (78.3%, 47/60 of cases). Importantly, nuclear P-STAT3- information about STAT3 expression and activation and its positive tumor cells were significantly more frequent in ESCCs potential therapeutic relevance in esophageal cancers. (71%, 35/49 of cases) as compared with BACs (47%, 29/61 of cases) In the experimental setting, Andl et al.24 reported an EGFR- (P ¼ 0.013; Table 1, Figure 1a, right column). In addition, nuclear mediated Jak-dependent activation of STAT1/STAT3 dimer activity P-STAT3 expression was significantly associated with STAT3 and cell migration in esophageal keratinocytes. In addition, several expression in ESCCs (P ¼ 0.003) or BACs (Po0.001). studies implicated bile acid or, more generally, inflammation- Finally, to exclude that potential heterogeneous STAT3 or associated STAT3 activation in esophageal cancer cell models.25,26 P-STAT3 protein expression was missed in the small biopsies, Recently, STAT3 was shown to cooperate with polo-like kinase 1 primary resected ESCCs and BACs were examined. Indeed, STAT3 (PLK1) in ESCC cell proliferation,27 an observation also of interest and P-STAT3 protein expression was homogeneously distributed in view of the availability of novel PLK1 inhibitors. In a more throughout the entire tissue sections (Figure 1b). clinicopathological situation, the data on STAT3 and phospho- Thus, STAT3 protein expression frequently occurs in esophageal STAT3 expression patterns in human tissue specimens of ESCCs cancers, and functionally active STAT3 (nuclear phospho-STAT3 at and BACs are limited. One study reported positive STAT3 protein Tyr750) is a prominent feature of ESCCs rather than BACs. Because expression in 90% of ESCC cases,28 another study detected of its homogeneous distribution within the tumors, STAT3 status phospho-STAT3 (Tyr705) positivity in 37% of ESCC cases27 and yet may equally be determined in small pre-therapeutic biopsies or another study revealed upregulation of phospho-STAT3 (Tyr705) primary resection specimens. expression along the dysplasia to carcinoma sequence in surveillance biopsies of patients diagnosed with Barrett’s esophagus.25 STAT3 expression and activity in ESCC and BAC cell lines From these studies, it remains open whether or not STAT3 To evaluate the expression, regulation and activity of STAT3 expression and activation are correlated within single cases of in vitro, two established and validated cell lines31,32 were selected esophageal carcinomas and whether or not this is different for further in vitro analyses. between ESCCs and BACs. Moreover, in view of current discussions Immunofluorescence analyses (Figure 2a) showed strong about therapeutic targeting STAT3 signaling, the functional cytoplasmic STAT3 (OE21, OE33) and negligible (OE21) or weak consequences of interfering with STAT3 in ESCC and BAC cells (OE33) cytoplasmic P-STAT3 expression in unstimulated cells. so far remain elusive. Thus, a better knowledge of STAT3 Upon EGF stimulation, unphosphorylated STAT3 translocated to alterations and associated cellular behavior in ESCCs and BACs is the nucleus in OE21 (as in tissue specimens of some ESCC cases, highly valuable for potential future treatment options in these two for example, Figure 1, case #1), but not OE33, cells. Importantly, histotypes of esophageal cancer. EGF stimulation markedly induced a strong nuclear P-STAT3 In the present study, we therefore investigated STAT3 and expression in OE21 cells, whereas P-STAT3 expression remained phosphorylated STAT3 at tyrosine 705 (hereafter termed P-STAT3) unchanged in OE33 cells (Figure 2a). protein expression and intracellular localization in serial sections To further validate and quantify these findings, time-course of human tissue specimens of the two major histotypes of analyses for STAT3 and P-STAT3 protein levels post EGF esophageal cancer. Furthermore, in established ESCC and BAC cell stimulation were performed (Figure 2b): STAT3 protein levels lines, we investigate the functional consequences of STAT3 remained stable in unstimulated and EGF-stimulated OE21 or inhibition on proliferation and migration of esophageal cancer OE33 cells. In contrast, EGF stimulation caused prominent cells, as well as direct transcriptional regulation of several STAT(3)- induction of P-STAT3 in OE21 cells (P ¼ 0.0016), whereas the low associated target genes, including also PLK126 or Cathepsins, CTSB constitutive P-STAT3 levels remained unchanged in OE33 cells and CTSL.29,30 This was supplemented by high-throughput (P ¼ 0.1721). Thereby, quantitative real-time PCR (qRT–PCR) transcriptome profiling that revealed gene targets and pathways analyses revealed no significant effect of EGF stimulation on regulated and/or affected by the STAT3 knockdown. STAT1, STAT3 or STAT5 and selected target genes (cyclin-D1, PLK1, Bxl-xl, Survivin, Mcl-1 and MMP9) messenger RNA (mRNA) expression (Supplementary Figure S1). RESULTS Thus, irrespective of other cellular mechanisms that may lead to STAT3 and P-STAT3 are frequently, but differentially, expressed in STAT3 activation, the cell lines reflect the preferential activation of ESCCs and BACs STAT3 in tissue specimens of ESCCs and represent appropriate Serial sections of the same tumor areas or tumor cell groups, in vitro models for further functional analyses. partially containing even the same tumor cells of 110 esophageal carcinomas, were stained for STAT3 and P-STAT3, followed by evaluation for STAT3 in 105/110 cases (owing to the loss of small STAT3 knockdown and effect on STAT1, STAT3, STAT5 and STAT3 carcinoma lesions on serial sections of the biopsies in five cases) target gene expression and all 110 cases for P-STAT3. Representative IHC stainings are To study the functional consequences of STAT3 knockdown in given in Figure 1 and data are summarized in Table 1. esophageal cancer cells, a protocol for long-term, transient small The majority of cases showed clearly detectable cytoplasmic interfering RNA (siRNA) transfection was established to implement STAT3 expression (88/105, 83.8%, score 1), whereas 17/105 (16.2%) consistent STAT3 knockdown for cell proliferation and cell cases were STAT3 negative (score 0). Moreover, 64/110 (58.2%) migration assays (Figure 3a). cases also exhibited clear nuclear P-STAT3 protein expression This was achieved by optimizing our previous short-term siRNA (score 1), with 46/110 (41.8%) cases being P-STAT3 negative (score protocols33 and resulted in STAT3 knockdown in both cell lines 0). Neither STAT3 (P ¼ 0.844) nor P-STAT3 (P ¼ 0.323) expression starting from 72 to 144 h (60–80% to 485% reduction of STAT3 was associated with histological grading of the esophageal mRNA expression and complete absence of protein expression carcinomas. (Supplementary Figure S2). In contrast, a control siRNA did not

& 2014 Macmillan Publishers Limited Oncogene (2014) 3256 – 3266 STAT3 expression and function in esophageal cancer S Timme et al 3258 H&E STAT3 P-STAT3

ESCC #1

Score 1 Score 1

ESCC #2

Score 1 Score 1

BAC #3

Score 1 Score 0

BAC #4

Score 1 Score 0

STAT3 P-STAT3

ESCC

Score 1 Score 1

BAC

Score 1 Score 0

Figure 1. STAT3 and P-STAT3 protein expression in human tissue specimens of esophageal carcinomas. (a) The panels provide representative H&E stainings as well as STAT3 and P-STAT3 immunohistochemical stainings of two cases of ESCCs (#1, #2; top two rows) and BACs (#3, #4; bottom two rows) each. Note that ESCCs and BACs display similar patterns of cytoplasmic STAT3 expression, whereas nuclear P-STAT3 is more frequent in ESCCs compared with BACs. Case #1 (ESCC) also shows nuclear localization of STAT3 in tumor cells (some exemplarily indicated by arrows). Scores of immunohistochemistry are given in panels. Refer to Table 1 for detailed quantification and statistical evaluation of all investigated cases. Images of H&E stainings were taken at Â 10 magnification, and images of STAT3 and P-STAT3 were taken at Â 40 fold magnification. Boxes in H&E stainings indicate the location of depicted STAT3 and P-STAT3 IHC stainings on serial sections. (b) The panels show low magnification ( Â 5; enlarged inserts to visualize cytoplasmic/nuclear staining) of representative STAT3 and P-STAT3 stainings in entire tissue sections of primary resected esophageal cancers. Note that ESCCs and BACs show homogeneous patterns of STAT3 and P-STAT3 expression.

significantly affect STAT3 mRNA expression or protein levels at involved in cell proliferation and cell survival (cyclin-D1, PLK1, these time points. Bcl-xl, Survivin, Mcl-1) also remained unaltered at the protein level In parallel to functional assays (120 h, Figure 3a and see below), upon STAT3–siRNA treatment, as shown by western blot analyses STAT1, STAT5 and STAT3 mRNA expression levels were analyzed. (Supplementary Figure S3). This clearly showed that, although there was a significant Together, these data demonstrate the significant, consistent reduction of STAT3 mRNA expression, STAT1 and STAT5 mRNA and specific inhibition of STAT3 for up to 6 days by siRNA expression levels were unaffected by control- or STAT3–siRNA treatment, without directly affecting transcriptional regulation of treatment in OE21 and OE33 cells (Figure 3b). Similarly, selected the investigated genes involved in cell proliferation, cell survival target genes involved in cell proliferation, cell survival and cell and cell migration. migration (cyclin-D1, PLK1, Bcl-xl, Survivin, Mcl-1, MMP9 and the Cathepsins B/L/Z) were not consistently and significantly altered by STAT3–siRNA treatment at this time point (Figure 3b) or at STAT3 knockdown reduces ESCC and BAC cell proliferation and earlier time points (72 and 96 h, except for MCL-1 (96 h; P ¼ 0.036, BAC cell migration OE21 cells) and cyclin-D1 (96 h; P ¼ 0.001, OE33 cells) To test the functional consequences of STAT3 knockdown, (Supplementary Figure S3). Moreover, the selected target genes analyses of esophageal cancer cell proliferation and/or migration

Oncogene (2014) 3256 – 3266 & 2014 Macmillan Publishers Limited STAT3 expression and function in esophageal cancer S Timme et al 3259 were performed (starting 96 h onward, Figure 3a) using the XCelligence and ORIS systems. Table 1. STAT3 and P-STAT3 protein expression in esophageal First, in four independent experiments for cell proliferation in carcinomas the XCelligence system (measurements in time period from 96 to 144 h, see Figure 3a) of untreated, control- or STAT3–siRNA- All ESCC BAC P-value treated OE21 and OE33 cells, cell proliferation was significantly reduced in both OE21 (P ¼ 0.0148) and OE33 (P ¼ 0.0243) upon n/N % n/N % n/N % STAT3–siRNA treatment, with cell-doubling times increasing by up STAT3 to twofold relative to untreated or control–siRNA-treated cells a (Figure 4a, left panels, black bars). In control–siRNA-treated OE21 Score 0 17/105 16.2 4/45 8.9 13/60 21.7 0.109 or OE33 cells, cell proliferation was unaffected, with doubling time Score 1 88/105 83.8 41/45 91.1 47/60 78.3 similar to untreated cells (Figure 4a, left panels, gray bars). P-STAT3 Second, cell migration was examined in three independent Score 0 46/110 41.8 14/49 28.6 32/61 52.5 0.013b experiments in untreated, control- and STAT3–siRNA-treated OE21 Score 1 64/110 58.2 35/49 71.4 29/61 47.5 and OE33 cells using the XCelligence system (Figure 4a, right panels): compared with untreated or control–siRNA-treated OE21 Abbreviations: BAC, Barrett’s adenocarcinomas; ESCC, esophageal squa- cells, there was no significant and specific effect of STAT3–siRNA mous cell carcinomas; STAT3, signal transducer and activator of transcrip- on cell migration of OE21 cells (P ¼ 0.1079). In contrast, STAT3– tion 3. The table summarizes the number (n) and frequency (%) of siRNA treatment markedly reduced migration of OE33 cells by observed STAT3 and P-STAT3 patterns for all analyzed (N) cases and the more than twofold relative to untreated or control–siRNA-treated statistical correlation (P-value) between the histological subtypes ESCC and BAC. The bold value indicate statistical significance. aMann–Whitney test. OE33 cells (P ¼ 0.073). bFisher’s exact test. To investigate this further and to exclude that the transwell 8-mm-pore XCelligence assays introduced a bias into the migratory behavior of the large, squamous-differentiated OE21 cells, an

Unstimulated EGF-stimulated STAT3P-STAT3 STAT3 P-STAT3

OE21

OE33

10 p=0.0016 15min60min 4hrs 8hrs 8 EGF -+-+ - + - + 6 PSTAT3 4 OE21 STAT3 2 β-Actin 0 Relative protein levels STAT3 P-STAT3 10 p=0.1721 15min60min 4hrs 8hrs 8 EGF -+-+ - + - + 6 PSTAT3 4 OE33 STAT3 2 β-Actin

Relative protein levels 0 STAT3 P-STAT3 Figure 2. Effect of EGF stimulation on STAT3 and P-STAT3 protein expression in ESCC and BAC cell lines. (a) The panels show representative 3-dimensional immunofluorescence images of unstimulated and EGF-stimulated ESCC (OE21) and BAC (OE33) cells stained for STAT3 and P-STAT3 (red; cell nuclei counterstained with DAPI in blue). Note the marked nuclear STAT3 and P-STAT3 protein expression in EGF-stimulated ESCC (OE21), but not BAC (OE33), cells. Note that BAC (OE33) cells have constitutive high levels of (cytoplasmic) P-STAT3. (b) Representative western blots (left panels) and their quantification (right graphs) of unstimulated (EGF-) and EGF-stimulated (EGF þ ) ESCC (OE21) and BAC (OE33) cells at the indicated time points. Graphs show mean±standard deviation of three independent experiments; white bars ¼ 15 min, light gray bars ¼ 1 h, dark gray bars ¼ 4 h and black bars ¼ 8 h. Note similar STAT3 levels in untreated OE21 and OE33 cell lines. Note weak P-STAT3 levels in unstimulated OE21 cells, which are significantly induced upon EGF stimulation (P ¼ 0.0016). Weak P-STAT3 levels are also seen in unstimulated BAC (OE33) cells, but these are not increased upon EGF stimulation (P ¼ 0.1721).

& 2014 Macmillan Publishers Limited Oncogene (2014) 3256 – 3266 STAT3 expression and function in esophageal cancer S Timme et al 3260 Cell 1st 2nd plating siRNA siRNA

hours -24 0487224 96 120 144

Validation of STAT3 knock-down >Fig. 3B, Suppl. Fig. S2

STAT3 target evaluation >Fig. 3B, Fig. 5, Suppl. Fig. S3

Cell proliferation assays >Fig. 4 Cell migration assays

3 p=0.011

OE21 2

Rel. mRNA expression 0

PLK1 Mcl-1 CTSL STAT1 STAT5 STAT3 Bcl-xl MMP9 CTSB CTSZ Survivin Cyclin-D1 5

3 p=0.003

OE33 2

Rel. mRNA expression 0

PLK1 Mcl-1 CTSL STAT1 STAT5 STAT3 Bcl-xl MMP9 CTSB CTSZ Survivin Cyclin-D1 Figure 3. Long-term STAT3 inhibition and effect on STAT1, STAT5 and STAT3, as well as target gene expression. (a) The diagram provides an overview of the modified siRNA protocol, associated validation and functional assays performed. See indicated figures for detailed data presentation. (b) The graphs show results of qRT–PCR analyses at 120 h, giving relative mRNA expression of control- (gray bars) and STAT3 (black bars)-siRNA-treated as compared with non-treated (expression set to 1) OE21 and OE33 cells. Graphs show mean±standard deviation of three (OE33) and four (OE21) independent experiments. Note significant knockdown of STAT3, but not STAT1 or STAT5, mRNA expression (arrows, P-values). No significant and specific STAT3–siRNA effect on selected genes involved in cell proliferation, cell survival and cell migration was seen. OE33 cells had undetectable MMP9 mRNA expression. Refer to Supplementary Figure S3 for western blot analyses of cyclin-D1, PLK1, Bcl-xl, Mcl-1 and Survivin protein levels.

additional adapted scratch assay using the ORIS system was OE21 and OE33 cells at an early time point after siRNA induction performed with three independent experiments (Figure 4b). (72 h) by using Illumina HT-12 Expression Bead Chips. For details of Again, treatment of OE21 cells with STAT3–siRNA did not have data analysis, see Materials and methods. any observable effect on their migration (P ¼ 0.386). In contrast, A network representation of a gene-set enrichment34 between STAT3–siRNA-treated OE33 cells showed a significant 2.5-fold STAT3- and control–siRNA-treated OE21 and OE33 cells is depicted reduction of cell migration compared with control–siRNA-treated in Figure 5a, clearly showing a significant downregulation of cell OE33 cells (P ¼ 0.015), in line with the results observed from the cycle-related gene sets in OE21 (Po0.0001) and OE33 (P ¼ 0.01) XCelligence system. cells. Within the ‘cell cycle’ gene group, several cyclins and cyclin- Taken together, STAT3 knockdown significantly reduces ESCC dependent kinases were downregulated in both cell lines and BAC cell proliferation as well as cell migration of BAC cells in (Figure 5b). In addition, cell migration genes, such as Collagen two types of cell migration assays. triple helix repeat containing-1 gene (CTHRC1) associated with increased cancer cell motility,35 were significantly upregulated in OE21 (Po0.001), but not in the OE33 cells at this time point High-throughput transcriptome profiling identifies cell (Figure 5b). Vice versa, while the gene Shroom family member 2 proliferation and cell migration signaling pathways deregulated (SHROOM2, former APXL), a gene involved in tight junction by STAT3 knockdown in esophageal cancer cells stabilization36 and restricted cell migration,37 was upregulated in To identify STAT3-associated mediators responsible for the OE33 cells, it was downregulated in OE21 cells (Figure 5b). observed phenotypic alterations by STAT3 knockdown, we next Hence, the transcriptome gene set enrichment analysis of measured the transcriptome of STAT3- and control–siRNA-treated control- and STAT3–siRNA-treated OE21 and OE33 cells clearly

Oncogene (2014) 3256 – 3266 & 2014 Macmillan Publishers Limited STAT3 expression and function in esophageal cancer S Timme et al 3261 p=0.0148 3 3 p=0.1079

2 2

OE21 1 1 Relative migration 0 Relative doubling time 0

3 3 p=0.0243 2 2 p=0.073

OE33 1 1 Relative migration

Relative doubling time 0 0

120 hrs 144 hrs 1.5 p=0.386 Control- siRNA 1

OE21 0.5 STAT3- siRNA Relative migration 0

120 hrs 144 hrs 1.5

Control- p=0.015 siRNA 1 OE33 0.5 STAT3- siRNA Relative migration 0

Figure 4. STAT3 knockdown inhibits ESCC and BAC cell proliferation and reduces cell migration in BAC cells. (a) The graphs show results of cell proliferation and cell migration assays using the XCelligence system of control- (gray bars) or STAT3–siRNA (black bars)-treated OE21 and OE33 cells from time point 96 to 144 h (see Figure 3a). Y-axes represent relative cell-doubling time (left panels) or relative cell migration (right panels) with untreated cells set to 1 (mean±s.e.m. of independent experiments). Note the significant increase of doubling time of OE21 (P ¼ 0.0148) and OE33 (P ¼ 0.0243) cells upon STAT3–siRNA treatment (left panels). Note also that cell migration was markedly reduced by more than twofold in STAT3–siRNA-treated OE33 (P ¼ 0.073), but not OE21 (P ¼ 0.1079), cells (right panels). (b) The panels show phase-contrast pictures of the ORIS migration assay of control– and STAT3–siRNA-treated OE21 and OE33 cells from 120 h to 144 h of the siRNA protocol (cells plated at 96 h, cell stopper removed at 120 h and migration captured at 144 h, see Figure 3a). Dotted lines indicate the border of cell migration into the initially central free area of the fibronectin-coated wells. At 144 h, cells that migrated to the central area were counted. Quantitative evaluation is given in the right graphs, showing cell migration of STAT3–siRNA (black bars) relative to control–siRNA (gray bars)-treated cells (mean±s.e.m. of three independent experiments). Note the prominent 2.5-fold reduction of cell migration in STAT3–siRNA-treated OE33 (P ¼ 0.015) cells. corroborates the phenotype of reduced cell proliferation in both Esophageal cancers still represent a major challenge of cell lines as well as their different regulation of cell migration. interdisciplinary oncology, although several (therapeutically targetable) molecular alterations were identified in the two main histotypes: esophageal squamous cell carcinoma (ESCC) and Barrett’s adenocarcinoma (BAC).21–23 So far, few studies DISCUSSION addressed STAT3 expression, activation and function in Signal transducer and activator of transcription (STAT) proteins esophageal cancers: STAT3 was implicated in bile acid and have an essential role in oncogenic signaling1–6 in the inflammation-associated BAC,25,26 in b-Catenin/TCF4 signaling28 carcinogenesis and progression of several epithelial cancers,7–13 and in PLK1-guided cell proliferation of ESCC cells.27 Hence, the and may hence represent a valuable therapeutic target—either present study comprehensively addressed STAT3 expression, alone14–16 or in synergy with receptor tyrosine kinases and/or src activation and function in both ESCC and BAC tissue specimens kinase family inhibitors.17–20 and cell lines.

& 2014 Macmillan Publishers Limited Oncogene (2014) 3256 – 3266 STAT3 expression and function in esophageal cancer S Timme et al 3262 Cell Cycle OE21 OE33 DNA Replication Pre-Initiation M/G1 Transition Prostaglandin Leukotriene metabolism Activation of ATR in response to replication stress Post NMDA receptor activation events Arachidonic Acid Metabolism E2F mediated regulation of DNA replication Mitotic G1-G1/S phases RNA transport colanic acid building blocks biosynthesis G1/S Transition DNA replication ATR signaling pathway CREB phosphorylation through the activation of Ras G2/M CheckpointsDNA Replication Transport of Mature mRNA derived from an Intron-Containing Transcript CREB phosphorylation through the activation of CaMKII DNA strand elongation Mitochondrial Gene Expression RNA Polymerase III Transcription Initiation From Type 3 Promoter Cell cycle S Phase Metabolism of non-coding RNA Mitotic M-M-G1 phases Interactions of Rev with cellular proteins snRNP Assembly Mitochondrial Fatty Acid Beta-Oxidation Ras activation uopn Ca2+ infux through NMDA receptor Traf cking and processing of endosomal TLR Cell Cycle, MitoticMitotic M-M/G1 phases G2/M Transition In uenza Life Cycle Rev-mediated nuclear export of HIV-1 RNA Doxorubicin Pathway, Pharmacokinetics Cell Cycle Processing of Capped Intron-Containing Pre-mRNA Cell Cycle Cyclin A/B1 associated events during G2/M transition Late Phase of HIV Life Cycle Extension of Telomeres M/G1 Transition APC/C-mediated degradation of cell cycle proteins Transcription-coupled NER (TC-NER) DNA Repair mRNA Splicing - Major Pathway Assembly of the pre-replicative complex Mitotic G2-G2/M phases DNA Repair Resolution of Abasic Sites (AP sites) HIV InfectionmRNA Processing DNA replication M Phase PLK1 signaling events Lagging Strand Synthesis Activation of the pre-replicative complex mRNA processing Ribosome biogenesis in eukaryotes HIV Life Cycle DNA strand elongation Regulation of mitotic cell cycle Polymerase switching Mitotic Prometaphase Mitotic M-M/G1 phases APC-C-mediated degradation of cell cycle proteins mRNA Splicing DNA Repair Transcription of the HIV genome DNA Replication Unwinding of DNA Telomere C-strand (Lagging Strand) Synthesis Aurora B signaling mRNA Splicing - Minor Pathway DNA Replication Pre-Initiation Gap- lling DNA repair synthesis and ligation in TC-NER Cell Cycle, Mitotic Spliceosome Repair synthesis for gap- lling by DNA polymerase in TC-NER Mitotic M-M-G1 phases RNA Polymerase II Pre-transcription Events Homologous recombination Polymerase switching on the C-strand of the telomere Cell Cycle Activation of ATR in response to replication stress BARD1 signaling events Ribosome biogenesis in eukaryotes Gap- lling DNA repair synthesis and ligation in GG-NER RNA Processing G2/M Checkpoints Repair synthesis of patch ~27-30 bases long by DNA polymerase Mitochondrial Protein Import Proteasome Electron Transport Chain Huntington,s disease Fanconi anemia pathway Degradation of beta-catenin by the destruction complex Fanconi anemia pathway Aurora A signaling The citric acid (TCA) cycle and respiratory electron transport ATR signaling pathway Parkinson,s disease

Cell Cycle Cell Migration OE21 OE33 OE21 OE33 Fold Change [siRNA−CTL] Fold Change [siRNA−CTL] Fold Change [siRNA−CTL] Fold Change [siRNA−CTL]

Figure 5. RNA profiling reveals STAT3-regulated genes and pathways in OE21 and OE33 cells. (a) Significantly downregulated signaling pathways due to STAT3 knockdown in OE21 and OE33 cells. The panels depict a network representation of the gene set enrichment analyses using gene sets from the Consensus Path DB. All nodes are significantly downregulated in STAT3– compared to control–siRNA-treated cells (using significance cut-offs with OE21: Q-valueo0.01; OE33: P-valueo0.05). The node size and color are proportional to gene set size and significance (dark red ¼ least significance; yellow–white ¼ highest significance). Gene set sizes range from n ¼ 10 genes (‘APC-C-mediated degradation of cell cycle proteins’) to n ¼ 350 genes (‘Cell Cycle’). Nodes are connected if they share at least 30% of their genes. The gray oval backdrops are added to cluster functionally overlapping gene sets. (b) Fold-change difference for individual genes of the gene sets ‘Cell Cycle’ and ‘Cell Migration’. The bar plots depict the average fold change of STAT3– compared with control–siRNA-treated OE21 and OE33 cells. Only genes that have a significant fold change between the two conditions (Po0.05, moderated t-test) are shown. In the ‘Cell Cycle’ gene set, all cyclin-dependent kinases, cyclins and cell division cycle genes are annotated by their gene symbol. The inserted legends show the significance for the directed regulation of these pathways according to the gene set enrichment.

In our study, the majority of esophageal carcinomas expressed surgical treatment.23 Our data show that representative STAT3 protein (83.8%) irrespective of their specific histotype. In evaluation of STAT3 and P-STAT3 protein is indeed possible in contrast, nuclear P-STAT3 protein expression was significantly such biopsies, as expression patterns were mostly homogeneous associated with ESCCs, which agrees with other studies of P-STAT3 in the entire tissue sections of primary resection specimens. expression in ESCCs.25,27 This may relate to a preferential Importantly, these STAT3 and P-STAT3 patterns were also ‘over’expression of EGFR in ESCCs.38,39 Indeed, P-STAT3 and mirrored by the ESCC and BAC cell lines, allowing further EGFR expression levels were associated in ESCCs (P ¼ 0.048), but translational clinicopathologic investigations. not BACs (P ¼ 0.336), in our cases (C Fichter, personal To this end, we examined the functional effects of STAT3 communication). Our present data of significant induction and knockdown using a transient, but long-term (6 days), siRNA nuclear translocation of P-STAT3 in ESCC (OE21) cells further treatment protocol. STAT3 knockdown resulted in a significant, underlines this and is consistent with data obtained from other twofold reduction of cell proliferation in ESCC (OE21) and BAC ESCC cell lines.27 (OE33) cells, as well as in a significant 2.5-fold reduction of cell Interestingly, STAT3 and P-STAT3 were both similarly expressed migration in BAC (OE33) cells. in BAC tissue specimens, and this was reflected in vitro by BAC Surprisingly, the observed inhibition of cell proliferation and/or (OE33) cells exhibiting equal, constitutive levels of STAT3 and cell migration by STAT3 knockdown in ESCC (OE21) and BAC P-STAT3. This may result from a potential STAT3 gene amplifica- (OE33) cells was not reflected by direct and strong transcriptional tion on chromosome 17q21, a region frequently amplified in regulation of cyclin-D1, PLK1, Bcl-xl, Survivin, Mcl-1, MMP9 and the BACs.40,41 Besides, inflammation-associated (constitutive) STAT3 Cathepsins B/L/Z, as measured by qRT–PCR at early and later time activity in BAC is conceivable,25,26 at least in vivo. points (72, 96 and 120 h) post STAT3 knockdown. This was Esophageal cancer specimens received for primary diagnosis contrary to the recent findings of, for example, a STAT3/PLK1 are frequently small endoscopic biopsies, that is, those obtained positive feedback loop in controlling cell proliferation in ESCC.27 before potential neoadjuvant chemo-/radiotherapeutic and/or In contrast to our experiments, Zhang et al.27, however, used

Oncogene (2014) 3256 – 3266 & 2014 Macmillan Publishers Limited STAT3 expression and function in esophageal cancer S Timme et al 3263 Table 2. Primary antibodies and newly established qRT–PCR primer and probes used in this study

Protein Antibody; manufacturer Dilution (method)

b-actin clone AAC-15; Sigma-Aldrich, Taufkirchen, GER 1:1000 (WB) Bcl-xl Polyclonal; Cell Signalling Technology, Danvers, USA 1:1000 (WB) Cyclin-D1 clone SP-4; DCS Innovative Diagnostik, Hamburg, GER 1:1000 (WB) Mcl-1, clone D35A5; Cell Signalling Technology, Danvers, USA 1:1000 (WB) PLK1 Polyclonal; AbCam, Cambridge, UK 1:1000 (WB) STAT3 clone 7D1; Cell Signalling Technology, Danvers, USA 1:100 (IHC, IF); 1:1000 (WB) P-STAT3 (T705) clone D3A7; Cell Signalling Technology, Danvers, USA 1:50 (IHC), 1:100 (IF); 1:1000 (WB) Survivin Polyclonal; NovusBiologicals, Cambridge, UK 1:1000 (WB)

Gene Primers and Probe Amplicon Size

STAT1 forward primer 50-gagcaggttcaccagctttatgat-30 124 bp reverse primer 50-aacggatggtggcaaatga-30 probe 50-Fam-aaagcaagactgggagcacgctgc-30-Tamra STAT5 forward primer 50-gaccgccccaaggatga-30 102 bp reverse primer 50-actcagggaccacttgcttga-30 probe 50-Fam-tcttctccaagtactacactcctgtgctggctaaag-30-Tamra MMP9 forward primer 50-gacgcagacatcgtcatcca-30 140 bp reverse primer 50-aactcgtcatcgtcgaaatgg-30 probe 50-Fam-ctttcctcctggccccggcatt-30-Tamra

Abbreviations: IHC, immunohistochemistry; IF, immunoﬂuorescence; qRT–PCR, quantitative real-time PCR; WB, western blot. See also Materials and Methods for additional published qRT–PCR assays.

Kyse-series cell lines with different origin and p53 mutation status either in the setting of treatment for non-metastasized BAC, that (that is, influencing cell proliferation). is, for preventing metastasis, or of metastasized ESCCs or BACs in Indeed, our RNA transcriptome analyses then readily revealed combination with other targeted inhibitors as seen for other the entire PLK1 signaling events as one of the pathway nodes epithelial cancers.17–20 Finally, we describe novel STAT3-regulated affected by STAT3–siRNA in OE21 cells. In addition, cell cycle genes involved in ESCC and BAC cell proliferation or cell genes, such as cyclin-dependent kinases, cyclins other than cyclin- migration. D1 and/or cell division cycle genes, were clearly downregulated in OE21 and OE33 cells upon STAT3 knockdown, thereby leading to the loss of cell proliferation. MATERIALS AND METHODS Similarly, our single-gene qRT–PCR analyses did not reveal any Tissue specimens STAT3-regulated mediators of the distinct pattern of ESCC (OE21) The study included formalin-fixed and paraffin-embedded (FFPE) tissue and BAC (OE33) cell migration upon STAT3 knockdown. Again, specimens of 110 patients with esophageal carcinomas from three centers with RNA transcriptome analyses, we were able to identify novel (University Medical Center/Freiburg and St. Vincentius Hospital/Karlsruhe, STAT3-regulated genes affecting cell morphology and/or cell Germany; Memorial Sloan-Kettering Cancer Center, NY, USA), and was conducted in agreement with institutional ethics regulations (#65/07). migration: The gene Shroom2 (SHROOM2, APXL) was down- Tissue specimens comprised 93/110 (84.5%) pretreatment biopsies and regulated in OE21 cells with unimpaired cell migration, but it was 17/110 (15.5%) surgical resection specimens. None of the patients had upregulated in OE33 cells with loss of migration. This relates well received neoadjuvant radio-/chemotherapy. The histology was esophageal 36 to the known function of Shroom2 in tight-junction stabilization squamous cell carcinomas (ESCC) in 49/110 cases (44.5%) and Barrett’s and in restricting endothelial cell migration via Rho kinase adenocarcinoma (BAC) in 61/110 (55.5%) cases.21,22 Of the carcinomas, 6/ (Rock),37 which may further relate to known interaction of Rho- 110 (5.5%), 81/110 (73.6%) and 23/110 (20.9%) were well, moderate and 44 GTPases with STAT3.42 In contrast, RNA transcriptome analyses poorly differentiated, respectively. revealed upregulation of the gene Collagen triple helix repeat containing-1 (CTHRC1) in still migrating OE21 upon STAT3 Immunohistochemistry knockdown. This is in line with a recently shown stimulation of Sections of the 110 FFPE tissue specimens were processed for STAT3 and motility of pancreatic cancer cells upon CTHRC1 overexpression.35 P-STAT3 staining according to previously established protocols.12 Briefly, Surely, these novel data shall be confirmed by further analyses, for 2-mm serial sections were deparaffinized, subjected to antigen retrieval example, also including other means of STAT3 inhibition such as (STAT3: pH 6.0; P-STAT3: pH 6.0) and incubated for 1 h with primary decoy oligonucleotides15,16 and extensive confirmative functional antibodies to STAT3 and P-STAT3 at tyrosine 705 (Table 2). After washing, analyses in vitro. all reactions were continued using the Dako REAL Detection System As seen from the above, analyses of selected gene(s) by qRT– (alkaline phosphatase/RED, rabbit/mouse) on the Autostainer System (all DakoCytomation, Glostrup, Denmark). PCR may miss important coordinated signaling events involved in Evaluation of STAT3 (cytoplasmic and/or nuclear) and P-STAT3 (nuclear) phenotypic alterations, which may also include influences of expression in esophageal carcinoma cells was in a two-tiered approach as 43 suppressors of cytokine signaling, protein inhibitors of STATs follows: score 0 ¼ negative or marginal positive in p20% tumor cells, and (PIAS) and/or protein phosphatases.5 In addition, the potential score 1 ¼ moderate and strong positive in 420% of the tumor cells.45,46 association of STAT3 and Cathepsins (CTSB,CTSL)30 also in our For this, a minimum of 300 tumor cells were counted in 3–4 high-power model systems is an important aspect, which prompts further fields ( Â 40 magnification). analyses at the protein/proteomic level. In summary, our present study provides novel experimental Cell cultures and EGF stimulation evidence in vivo and in vitro that inhibition of STAT3 may be a Esophageal cancer cell lines (OE21, OE33)31,32 were from ATCC/LGC valuable novel target in the two histotypes of esophageal cancers, Standards (Wesel, Germany). Cells grew adherent in RPMI 1640 medium

& 2014 Macmillan Publishers Limited Oncogene (2014) 3256 – 3266 STAT3 expression and function in esophageal cancer S Timme et al 3264 with 10% FCS and 2 mM glutamine (all from PAA Laboratories, Pasching, several protein interaction and pathway databases, such as WikiPathways, 52 Austria) at 37 1C in a 5% CO2 atmosphere. Gene Ontology, KEGG or Pathway Interaction Database. For EGF stimulation, 3 Â 104 (for RNA analyses) and 8 Â 104 (for protein The functional relationships were visualized by constructing a gene set analyses) cells were seeded, washed, starved (no FCS) overnight and then network in which two nodes are connected if they share at least 30% of incubated with 100 ng/ml EGF for 15 min, 1 h, 4 h and 8 h. RNA and protein their genes.52 This thresholding results in disjoint network clusters that analyses were performed as described below. All experiments were group functionally related gene sets. performed three times and means±s.d. were calculated.

Protein extraction and western blotting Immunofluorescence 31,32 As published before , each 5–10 mg sample of total proteins of cultured 32,33 As reported earlier, cultured cells were fixed, permeabilized, blocked cells (Qproteome Kit; Qiagen) was blotted, membranes were blocked and and incubated with primary antibodies (STAT3, P-STAT3; Table 2) overnight then incubated at 4 1C overnight with primary antibodies (STAT3, P-STAT3, at 4 1C. Washed cells were incubated with secondary antibodies (1:200, Bcl-xl, cyclin-D1, Mcl-1, Survivin, PLK1, b-Actin; Table 2). Thereafter, Alexa488- or Alexa568-conjugated goat anti-rabbit or goat anti-mouse IgG, secondary antibodies (1:25000; HRP-conjugated goat anti-mouse or anti- Invitrogen/Life Technologies, Darmstadt, Germany) for 1 h at room rabbit IgG, Dianova, Hamburg, Germany) were incubated for 1 h at room temperature, washed and mounted with 4’,6-diamidino-2-phenylindole temperature. Visualization was with the ECL western blotting detection counterstain. Visualization was by the ‘Axioplan 2 imaging with ApoTome’ reagents or Lumigen TMA-6 (GE Healthcare, Freiburg, Germany). Images taking image stacks at 1 mm (Plan Apochromat Â 63/1.3 oil objective), and were captured (CN3000-WL Infinity System, Infinity Version 14.2; PeqLab subsequent conversion of image stacks to 3-dimensional view was Biotechnologies, Erlangen, Germany), and the protein pixel density was performed using the AxioVision 4.8 software (all Carl Zeiss MicroImaging measured using the ImageJ software, with data presented as mean±s.d. Inc., Go¨ttingen, Germany). for all experiments.

STAT3 knockdown by siRNA Cell proliferation and cell migration assays using XCelligence The design of the long-term siRNA protocol was adapted from our System previous published work33 and associated assays are provided in Figure 3a Cell proliferation and migration were measured using the XCelligence and Supplementary Figure S2. System, culture plates and the RTCA software 1.2 (Roche Diagnostics For each independent experiment, 3 Â 104 cells were seeded in 24-well GmbH, Basel, Switzerland) starting 96 h after initial seeding (that is, 24 h plates (RNA analyses, triplicates) or 8 Â 104 cells in 12-well plates (protein after the second siRNA treatment; Figure 3a). analyses, duplicates). These were transfected twice with 100 nM STAT3 For each independent experiment on cell proliferation (E-plates 16), siRNA (pool of 4 STAT3 sequences, siGENOMESMARTpool, Dharmacon 6 Â 104 OE21 and 4.5 Â 104 OE33 cells were seeded per well in triplicate for RNAi Technologies, Thermo Fisher Scientific, Lafayette, CO, USA) or Silencer medium, STAT3- and control–siRNA. negative siRNA control (Invitrogen/Life Technologies GmbH,) using 1 mlof For each independent experiment on cell migration (CIM-plate 16/8 mm DharmaFECT (Dharmacon RNAi Technologies, Thermo Fisher Scientific) pores), 12 Â 104 OE21 and 9.2 Â 104 OE33 cells were seeded per well into transfection reagent for OE33 cells or siPORTTM NeoFXTM (Invitrogen/Life upper chambers in triplicate with serum-free medium, STAT3- and control– Technologies GmbH) transfection reagent for OE21 cells. siRNA. Lower chambers all contained 10% FCS medium. Cells treated with the STAT3-specific siGENOMESMARTpool or the Cell proliferation and cell migration were evaluated by taking Silencer negative siRNA control are referred to as ‘STAT3–siRNA’ and impedance measurements once every 15 min for up to 50 h. Control- ‘control–siRNA’-treated cells. Effects of siRNA treatments were evaluated and STAT3–siRNA-treated cells are presented as ‘relative doubling time’ for both STAT3–siRNA and control–siRNA-treated cells relative to untreated and ‘relative cell migration’ in comparison with untreated cells (set to 1), cells receiving normal medium throughout the experiments. respectively. Data are given as mean±s.e.m. of independent experiments.

RNA isolation, cDNA synthesis and qRT–PCR Cell migration assay using the ORIS System Total RNA isolation (RNeasy Mini kit; Qiagen, Hilden, Germany), reverse transcription and qRT–PCR were performed with primers and probes for To quantitatively assess the migration of single cells also in a non- STAT3, cyclin-D1, Survivin, Bxl-xl, Mcl-1, PLK1 and TBP.12,32,33,47–49 Primers transwell, adapted ‘scratch assay’, the ORIS System was used (Platypus Technologies, Madison, WI, USA). and probes for STAT1, STAT5 and MMP9 (intron-spanning) were 4 4 established in this study (Table 2). Cathepsin B, L and Z mRNA expression For this, 6 Â 10 OE21 and 8 Â 10 OE33 cells (duplicate wells of each was measured by Sybr-Green assays (primer sequences available upon STAT3- or control–siRNA-treated cells) were seeded, after 96 h (that is, 24 h request) using b-Actin50 as the reference gene. Relative mRNA expression after second siRNA transfection, Figure 3a), per well of fibronectin-coated was calculated by the comparative Ct-method [2 À ddCt]51 as plates, containing a central area covered by a cell stopper. Cell stoppers reported.12,32,33,47–50 Data reflect ‘relative mRNA expression’ (y-axes in were removed from confluent monolayers 24 h later (at 120 h, see graphs) of independent experiments (mean±s.e.m.). Figure 3a) and cell migration into the central well area was documented for 24 h (that is, until 144 h; see Figure 3a), after which the cells were fixed, stained with 4’,6-diamidino-2-phenylindole and evaluated by fluorescence High-throughput transcriptome profiling microscopy. Thereafter, quantification was carried out by automated image Isolated RNAs, 72 h after transfection, were processed with the Ambion WT analysis counting of the number of migrated cells (ImageJ) per total cells Expression kit (Ambion, Austin, TX, USA) as described by the manufacturer. plated and expressing this in STAT3- relative to control–siRNA-treated cells. ± Hybridization and scanning were performed according to the standard Data are given as mean s.e.m. of independent experiments. Illumina protocols. Illumina HT-12 Expression Bead Chips were normalized together using the quantile normalization algorithm without background subtraction. Probesets with known bad quality and those having no Entrez Statistical analyses ID annotation were discarded, resulting in 19584 EntrezID-annotated Statistical evaluation included STAT3 and P-STAT3 protein expression (IHC genes. In case of multiple probesets matching the same gene, the probeset scores) and pathological parameters (histotype, grading). Fisher’s exact having the respective highest interquartile range across all samples was test, two-sided for proportions, distribution-free Kruskal–Wallis or Mann– chosen for further analysis. Differential gene expression analysis between Whitney U tests were applied to compare parameters. A P-value less than treatment groups was performed by an empirical Bayes approach 0.05 was considered statistically significant. Statistical analyses were providing moderated t-statistics. Gene expression data are available for performed with the software package SPSS v 18 (SPSS Inc., Chicago, IL, the reviewers at GEO using the following link: http://www.ncbi.nlm.nih. USA). For in vitro experiments, the Student’s t-test was used. gov/geo/query/acc.cgi?token=vxczzsagiokywbk&acc=GSE47763. The gene set enrichment analyses were performed using the GAGE algorithm,34 which tests whether a gene set is highly ranked relative to other genes. For functional annotation, we used gene sets from the CONFLICT OF INTEREST Consensus Path DB, comprising a curated pathway compendium from The authors declare no conflict of interest.

Oncogene (2014) 3256 – 3266 & 2014 Macmillan Publishers Limited STAT3 expression and function in esophageal cancer S Timme et al 3265 ACKNOWLEDGEMENTS 21 Gabbert HE, Nakamura Y, Shimoda T, Field YK, Hainaut P, Inoue H. Esophageal We thank Martina Gansz for assistance with Cathepsin analyses. We acknowledge the squamous cell carcinomas. In: Bosman FT, Carneiro F, Hruban RH, Theise ND (eds.) support of the study by the Deutsche Forschungsgemeinschaft (SFB850 project C5 to WHO Classification of Tumours. Volume 3. (IARC Press, Lyon, France). SL, MW; SFB850 project B7 to TR; Gerok stipend of SFB850 project Z1 to ST). 22 Werner M, Lambert R, Flejou JF, Keller G, Hainaut P, Stein HJ, Ho¨ fler H. Adeno- Preliminary work for this study has been supported by the Mushett Family carcinoma of the esophagus. In: Bosman FT, Carneiro F, Hruban RH, Theise ND Foundation, Chester, NJ, USA (grant to SL, MW, LT, DK). TR and SL are member (eds.) WHO Classification of Tumours. Volume 3. (IARC Press, Lyon, France). and associate members of BIOSS Centre for Biological Signalling Studies, Albert- 23 Pennathur A, Gibson MK, Jobe BA, Luketich JD. Oesophageal carcinoma. Lancet Ludwigs-University, Freiburg, Germany. This study was supported by DFG SFB850 C5 2013; 381: 400–412. (MW, SL); DFG SFB850 B7 (TR); DFG SFB850 Z1 (Gerok position to ST); and Mushett 24 Andl CD, Mizushima T, Oyama K, Bowser M, Nakagawa H, Rustgi AK. EGFR-induced Family Foundation, Chester, NJ (DK, LT, MW, SL). cell migration is mediated predominantly by the JAK-STAT pathway in primary esophageal keratinocytes. Am J Physiol Gastrointest Liver Physiol 2004; 287: G1227–G1237. REFERENCES 25 Dvorak K, Chavarria M, Payne CM, Ramsey L, Crowley-Weber C, Dvorakova B et al. Activation of the interleukin-6/STAT3 antiapoptotic pathway in esophageal cells 1 Jove R. Preface: STAT signaling. Oncogene 2000; 19: 2466–2467. by bile acids and low pH: relevance to barrett’s esophagus. Clin Cancer Res 2007; 2 Bromberg J. Stat proteins and oncogenesis. J Clin Invest 2002; 109: 1139–1142. 13: 5305–5313. 3 Vinkemeier U. Getting the message across, STAT! design principles of a molecular 26 Zhang HY, Zhang Q, Zhang X, Yu C, Huo X, Cheng E et al. Cancer-related signaling circuit. J Cell Biol 2004; 167: 197–201. inflammation and Barrett’s carcinogenesis: interleukin-6 and STAT3 mediate 4 Yu H, Pardoll D, Jove R. STATs in cancer inflammation and immunity: a leading apoptotic resistance in transformed Barrett’s cells. Am J Physiol Gastrointest Liver role for STAT3. Nat Rev Cancer 2009; 9: 798–809. Physiol 2011; 300: G454–G460. 5 Aggarwal BB, Kunnumakkara AB, Harikumar KB, Gupta SR, Tharakan ST, Koca C 27 Zhang Y, Du XL, Wang CJ, Lin DC, Ruan X, Feng YB et al. Reciprocal activation et al. Signal transducer and activator of transcription-3, inflammation, and cancer: between plk1 and stat3 contributes to survival and proliferation of esophageal how intimate is the relationship? Ann NY Acad Sci 2009; 1171: 59–76. cancer cells. Gastroenterology 2011; 142: 521–530. 6 Sansone P, Bromberg J. Targeting the interleukin-6/Jak/stat pathway in human 28 Yan S, Zhou C, Zhang W, Zhang G, Zhao X, Yang S et al. beta-Catenin/TCF malignancies. J Clin Oncol 2012; 30: 1005–1014. pathway upregulates STAT3 expression in human esophageal squamous cell 7 Haura EB, Zheng Z, Song L, Cantor A, Bepler G. Activated epidermal growth factor carcinoma. Cancer Lett 2008; 271:85–97. receptor-Stat-3 signaling promotes tumor survival in vivo in non-small cell lung 29 Reiser J, Adair B, Reinheckel T. Specialized roles for cysteine cathepsins in health cancer. Clin Cancer Res 2005; 11: 8288–8294. and disease. J Clin Invest 2010; 120: 3421–3431. 8 Jiang R, Jin Z, Liu Z, Sun L, Wang L, Li K. Correlation of activated STAT3 expression 30 Kreuzaler PA, Staniszewska AD, Li W, Omidvar N, Kedjouar B, Turkson J et al. Stat3 with clinicopathologic features in lung adenocarcinoma and squamous cell controls lysosomal-mediated cell death in vivo. Nat Cell Biol 2011; 13: 303–309. carcinoma. Mol Diagn Ther 2011; 15: 347–352. 31 Boonstra JJ, van Marion R, Beer DG, Lin L, Chaves P, Ribeiro C et al. Verification 9 Seethala RR, Gooding WE, Handler PN, Collins B, Zhang Q, Siegfried JM, and unmasking of widely used human esophageal adenocarcinoma cell lines. Grandis JR. Immunohistochemical analysis of phosphotyrosine signal transducer J Natl Cancer Inst 2010; 102: 271–274. and activator of transcription 3 and epidermal growth factor receptor autocrine 32 Fichter CD, Herz C, Mu¨nch C, Opitz OG, Werner M, Lassmann S. Occurrence of signaling pathways in head and neck cancers and metastatic lymph nodes. Clin multipolar mitoses and association with Aurora-A/-B kinases and p53 mutations in Cancer Res 2008; 14: 1303–1309. aneuploid esophageal carcinoma cells. BMC Cell Biol 2011; 12: 13. 10 Gong W, Wang L, Yao JC, Ajani JA, Wei D, Aldape KD et al. Expression of activated 33 Herz C, Schlu¨ rmann F, Batarello D, Fichter CD, Scho¨pflin A, Mu¨nch C et al. signal transducer and activator of transcription 3 predicts expression of vascular Occurrence of Aurora A positive multipolar mitoses in distinct molecular classes endothelial growth factor in and angiogenic phenotype of human gastric cancer. of colorectal carcinomas and effect of Aurora A inhibition. Mol Carcinog 2012; 51: Clin Cancer Res 2005; 11: 1386–1393. 696–710. 11 Lee J, Kang WK, Park JO, Park SH, Park YS, Lim HY et al. Expression of activated 34 Luo W, Friedman MS, Shedden K, Hankenson KD, Woolf PJ. GAGE: generally signal transducer and activator of transcription 3 predicts poor clinical outcome in applicable gene set enrichment for pathway analysis. BMC Bioinformatics 2009; gastric adenocarcinoma. APMIS 2009; 117: 598–606. 27(10): 161. 12 Lassmann S, Schuster I, Walch A, Go¨bel H, Ju¨ tting U, Makowiec F et al. STAT3 35 Park EH, Kim S, Jo JY, Kim SJ, Hwang Y, Kim JM et al. Collagen triple helix repeat mRNA and protein expression in colorectal cancer: effects on STAT3-inducible containing-1 promotes pancreatic cancer progression by regulating migration targets linked to cell survival and proliferation. J Clin Pathol 2007; 60: 173–179. and adhesion of tumor cells. Carcinogenesis 2013; 34: 694–702. 13 Morikawa T, Baba Y, Yamauchi M, Kuchiba A, Nosho K, Shima K et al. STAT3 36 Etournay R, Zwaenepoel I, Perfettini I, Legrain P, Petit C, El-Amraoui A. Shroom2, a expression, molecular features, inflammation patterns, and prognosis in a myosin-VIIa- and actin-binding protein, directly interacts with ZO-1 at tight junctions. J Cell Sci 2007; 120: 2838–2850. database of 724 colorectal cancers. Clin Cancer Res 2011; 17: 1452–1462. 37 Farber MJ, Rizaldy R, Hildebrand JD. Shroom2 regulates contractility to control 14 Zhao W, Jaganathan S, Turkson J. A cell-permeable Stat3 SH2 domain mimetic endothelial morphogenesis. Mol Biol Cell 2011; 22: 795–805. inhibits Stat3 activation and induces antitumor cell effects in vitro. J Biol Chem 38 Hanawa M, Suzuki S, Dobashi Y, Yamane T, Kono K, Enomoto N, Ooi A. EGFR 2010; 285: 35855–35865. protein overexpression and gene amplification in squamous cell carcinomas of 15 Leong PL, Andrews GA, Johnson DE, Dyer KF, Xi S, Mai JC et al. Targeted inhibition the esophagus. Int J Cancer 2006; 118: 1173–1180. of Stat3 with a decoy oligonucleotide abrogates head and neck cancer cell 39 Yu WW, Guo YM, Zhu M, Cai XW, Zhu ZF, Zhao WX, Fu XL. Clinicopathological and growth. Proc Natl Acad Sci USA 2003; 100: 4138–4143. prognostic significance of EGFR over-expression in esophageal squamous cell 16 Souissi I, Najjar I, Ah-Koon L, Schischmanoff PO, Lesage D, Le Coquil S et al. carcinoma: a meta-analysis. Hepatogastroenterology 2011; 58: 426–431. A STAT3-decoy oligonucleotide induces cell death in a human colorectal carci- 40 Rauser S, Weis R, Braselmann H, Feith M, Stein HJ, Langer R et al. Significance of noma cell line by blocking nuclear transfer of STAT3 and STAT3-bound NF-kB. HER2 low-level copy gain in Barrett’s cancer: implications for fluorescence in situ BMC Cell Biol 2011; 12:14. hybridization testing in tissues. Clin Cancer Res 2007; 13: 5115–5123. 17 Boehm AL, Sen M, Seethala R, Gooding WE, Freilino M, Wong SM et al. Combined 41 Yoon HH, Shi Q, Sukov WR, Wiktor AE, Khan M, Sattler CA et al. Association of targeting of epidermal growth factor receptor, signal transducer and activator of HER2/ErbB2 expression and gene amplification with pathologic features and transcription-3, and Bcl-X(L) enhances antitumor effects in squamous cell carci- prognosis in esophageal adenocarcinomas. Clin Cancer Res 2012; 18: 546–554. noma of the head and neck. Mol Pharmacol 2008; 73: 1632–1642. 42 Reipschlaeger S, Kubatzky K, Taromi S, Burger M, Orth J, Aktories K, Schmidt G. 18 Bonner JA, Yang ES, Trummell HQ, Nowsheen S, Willey CD, Raisch KP. Inhibition of Toxin-induced RhoA activity mediates CCL1-triggered STAT signaling. J Biol Chem STAT-3 results in greater cetuximab sensitivity in head and neck squamous cell 2012; 287: 11183–11194. carcinoma. Radiother Oncol 2011; 99: 339–343. 43 Tischoff I, Hengge UR, Vieth M, Ell C, Stolte M, Weber A, Schmidt WE, Tannapfel A. 19 Jaganathan S, Yue P, Turkson J. Enhanced sensitivity of pancreatic cancer cells to Methylation of SOCS-3 and SOCS-1 in the carcinogenesis of Barrett’s concurrent inhibition of aberrant signal transducer and activator of transcription 3 adenocarcinoma. Gut 2007; 56: 1047–1053. and epidermal growth factor receptor or Src. J Pharmacol Exp Ther 2010; 333: 44 Sobin LH, Gospodarowicz MK, Wittekind C (eds) TNM Classification of Malignant 373–381. Tumours. 7th Edn (Wiley-Blackwell, UICC, Geneva, Switzerland, 2009). 20 Nagaraj NS, Washington MK, Merchant NB. Combined blockade of Src kinase and 45 Khoury JD, Medeiros LJ, Rassidakis GZ, Yared MA, Tsioli P, Leventaki V et al. epidermal growth factor receptor with gemcitabine overcomes STAT3-mediated Differential expression and clinical significance of tyrosine-phosphorylated resistance of inhibition of pancreatic tumor growth. Clin Cancer Res 2011; 17: STAT3 in ALK þ and ALK À anaplastic large cell lymphoma. Clin Cancer Res 2003; 483–493. 9: 3692–3699.

& 2014 Macmillan Publishers Limited Oncogene (2014) 3256 – 3266 STAT3 expression and function in esophageal cancer S Timme et al 3266 46 Nasr MR, Laver JH, Chang M, Hutchison RE. Expression of anaplastic lymphoma 49 Lassmann S, Kreutz C, Schoepflin A, Hopt U, Timmer J, Werner M. A novel kinase, tyrosine-phosphorylated STAT3, and associated factors in pediatric ana- approach for reliable microarray analysis of microdissected tumor cells from plastic large cell lymphoma: A report from the children’s oncology group. Am J formalin-fixed and paraffin-embedded colorectal cancer resection specimens. Clin Pathol 2007; 127: 770–778. J Mol Med 2009; 87: 211–224. 47 Gerlach U, Kayser G, Walch A, Hopt U, Schulte-Mo¨nting J, Werner M, Lassmann S. 50 Livak KJ, Schmittgen TD. Analysis of relative gene expression data using real-time Centrosome-, chromosomal-passenger- and cell-cycle-associated mRNAs are dif- quantitative PCR and the 2(-Delta Delta C(T)) method. Methods 2001; 25: 402–408. ferentially regulated in the development of sporadic colorectal cancer. J Pathol 51 Kamburov A, Stelzl U, Lehrach H, Herwig R. The ConsensusPathDB interaction 2006; 208: 462–472. database: 2013 update. Nucleic Acids Res 2013; 41: D793–D800. 48 Lassmann S, Danciu M, Mu¨ller M, Weis R, Makowiec F, Schulte-Mo¨ nting J et al. 52 Boerries M, Grahammer F, Eiselein S, Buck M, Meyer C, Goedel M et al. Molecular Aurora A is differentially expressed and regulated in chromosomal and micro- fingerprinting of the podocyte reveals novel gene and protein regulatory satellite instable sporadic colorectal cancers. Mod Pathol 2009; 22: 1385–1397. networks. Kidney Int 2013; 83: 1052–1064.

Supplementary Information accompanies this paper on the Oncogene website (http://www.nature.com/onc)

Oncogene (2014) 3256 – 3266 & 2014 Macmillan Publishers Limited