SMURF1 Amplification Promotes Invasiveness in Pancreatic Cancer

Pancreatic cancer is a deadly disease, and new therapeutic targets are urgently needed. We previously identified DNA amplification at 7q21-q22 in pancreatic cancer cell lines. Now, by high-resolution genomic profiling of human pancreatic cancer cell lines and human tumors (engrafted in immunodeficient mice to enrich the cancer epithelial fraction), we define a 325 Kb minimal amplicon spanning SMURF1, an E3 ubiquitin ligase and known negative regulator of transforming growth factor β (TGFβ) growth inhibitory signaling. SMURF1 amplification was confirmed in primary human pancreatic cancers by fluorescence in situ hybridization (FISH), where 4 of 95 cases (4.2%) exhibited amplification. By RNA interference (RNAi), knockdown of SMURF1 in a human pancreatic cancer line with focal amplification (AsPC-1) did not alter cell growth, but led to reduced cell invasion and anchorage-independent growth. Interestingly, this effect was not mediated through altered TGFβ signaling, assayed by transcriptional reporter. Finally, overexpression of SMURF1 (but not a catalytic mutant) led to loss of contact inhibition in NIH-3T3 mouse embryo fibroblast cells. Together, these findings identify SMURF1 as an amplified oncogene driving multiple tumorigenic phenotypes in pancreatic cancer, and provide a new druggable target for molecularly directed therapy.


Introduction
Pancreatic ductal adenocarcinoma (hereafter, pancreatic cancer) is nearly always fatal, with a five year survival rate less than 5% [1]. It is often disseminated at diagnosis, and can metastasize widely. Early detection can improve survival, but surgical resection is rarely curative [2]. Pancreatic cancer is also largely resistant to conventional chemotherapy. Therefore, new therapies are urgently needed. In particularly, it will be important to discover and validate new targets for molecularly-directed therapy.
The molecular genetics of pancreatic cancer are in part known [3,4]. Somatic activating mutations of KRAS (sometimes occurring with gene amplification) are found in .90% of pancreatic cancers. Also common are inactivating mutations/deletions of tumor suppressors CDKN2A (.95% of cancers), TP53 (50-75%), and SMAD4 (also known as DPC4) (55%), an effector of TGFbmediated growth inhibition. Other gene mutations, each occurring in less than 5% of cancers, impact these and other core cancer signaling pathways [5].
Four specimens (the AsPC-1 cell line, and three xenografts) had genomic profiles that were particularly informative in delimiting the amplicon boundaries within 7q21-q22 (Fig. 1A). The smallest common region of gain spanned just 325 Kb within cytoband 7q22.1, and contained just two RefSeq [14] genes, SMURF1 (SMAD specific E3 ubiquitin protein ligase) and KPNA7 (karyopherin alpha 7). SMURF1 is a known inhibitor of TGFb signaling (by promoting degradation of its receptor TGFbRI, and signaling mediator SMAD4 [15,16]), a pathway frequently disrupted in pancreatic cancer. Given an obvious connection to pancreatic carcinogenesis, we therefore focused subsequent efforts on SMURF1.
Consistent with an oncogenic role, SMURF1 transcript (measured by microarray) was significantly elevated in cell lines/ xenografts with 7q22.1 gain/amplification (as were several coamplified neighboring genes) (Fig. 1B). To evaluate SMURF1 amplification in primary pancreatic tumors, we also carried out FISH on a tissue microarray containing 105 pancreatic cancer cases. Four of 95 (4.2%) evaluable cases exhibited SMURF1 amplification (locus/centromere ratio .2.5) (Fig. 1C), comparable to our CGH findings for early-passage xenografts. We were unable to identify a suitable antibody and staining conditions to evaluate SMURF1 expression by immunohistochemistry.

SMURF1 amplification promotes cell invasion and anchorage-independent growth
To evaluate possible oncogenic functions of SMURF1, we first used RNAi to knockdown SMURF1 expression in the relevant context of gene amplification, using AsPC-1 cells. Transfection of four different small interfering RNAs (siRNAs), or a pool of the four together, each led to reduced SMURF1 protein levels (by Western blot), compared to a non-targeting siRNA pool ( Fig. 2A). Knockdown of SMURF1 did not alter cell proliferation, measured by WST-1 assay (Fig. 2B,C), and by BrdU incorporation (Fig. 2D), but led to significantly decreased cell invasion through Matrigel, measured by Boyden chamber assay (Fig. 2E). Decreased invasion was seen with each of the four siRNAs targeting distinct SMURF1 sequences, while the growth rate of the cells remained unchanged within the same time period (Fig. 2C), strongly supporting the specific role of SMURF1 in the invasiveness phenotype.
Given its known, antagonistic function in the TGFb pathway, we also sought to evaluate the effect of SMURF1 knockdown on TGFb signaling, using a TGFb responsive transcriptional reporter (p3TP-Lux) [17]. Knockdown of SMURF1 did not enhance TGFb pathway-mediated transcription in AsPC-1 cells (Fig. 2F). Of note, however, AsPC-1 cells (like most pancreatic cancers) harbor a mutated SMAD4, here SMAD4 (R100T) [18], characterized to be inactivating [19,20]. Therefore, AsPC-1 cells are likely incapable of a TGFb pathway transcriptional response. More generally, these findings suggest that the main effect(s) of SMURF1 amplification/overexpression are likely mediated through pathways distinct from TGFb signaling.
We also sought to evaluate the effect of siRNA knockdown in other pancreatic cancer cell lines. We chose two cell lines, BxPC-3 cells which (like AsPC-1 cells and most pancreatic tumors) have mutated SMAD4 (here by homozygous deletion) [21], and Hs700T cells which are wildtype for SMAD4 and have an intact TGFb growth-inhibitory pathway (Fig. S1). Notably, neither of these lines harbors focal amplification of SMURF1 (AsPC-1 cells are the only established line with focal amplification), nor elevated SMURF1 protein levels (Fig. 4A). Knockdown of SMURF1 (validated by Western blot; Fig. 4B) led to modestly reduced cell proliferation in BxPC-3 cells (Fig. 4C), and more so in TGFb-growth inhibitory pathway-intact Hs700T cells (Fig. 4D). SMURF1 knockdown also resulted in reduced cell invasion in BxPC-3 cells (Fig. 4E), though not significantly so. However, given that SMURF1 is neither focally amplified nor overexpressed in these lines, a simple explanation for the discordant phenotypes (compared to AsPC-1) is that SMURF1 may not function as an oncogenic driver in these cell contexts.
Finally, in complementary, overexpression studies, we transfected SMURF1 cDNA (expressed from a CMV promoter) into NIH-3T3 mouse fibroblasts. Overexpression of SMURF1, confirmed by Western blot (Fig. 5A), led to a significant loss of contact inhibition (i.e. increased foci), compared to a vector control (Fig. 5B). Notably, transfection of a catalytically-inactive mutant of SMURF1 (C699A) [22] did not reduce contact inhibition (Fig. 4B), indicating that this oncogenic activity is dependent on the E3 ubiquitin protein ligase activity of SMURF1.

Discussion
Here, we set out to pinpoint and discover the oncogene driving 7q21-q22 amplification in pancreatic cancer. High-resolution genomic profiling of pancreatic cancer cell lines and early-passage xenografts defined a 325 Kb minimal amplicon spanning SMURF1. Transcript levels of SMURF1 were elevated in specimens with gain/amplification, and by FISH we confirmed SMURF1 amplification in primary pancreatic cancers. Using complementary approaches of knockdown (in focally-amplified AsPC-1 cells) and overexpression (in NIH-3T3 cells), we determined that SMURF1 amplification/overexpression does not alter cell proliferation, but promotes cell invasion, anchorageindependent growth, and loss of contact inhibition, of which at least the latter is dependent on its catalytic activity.
SMURF1 was initially an intriguing oncogene candidate because of its known connection to TGFb signaling. The TGFb pathway, at least early in tumor development, is growth suppressive [23]. Normally, TGFb binds to its receptors (TGFbRI, TGFbRII), leading to the phosphorylation of signal transducers SMAD2/SMAD3, which then shuttle to the nucleus and in complex with SMAD4 mediate transcription. Key transcriptional responses include induction of CDKN2B (p15Ink4b) and CDKN1A (p21Cip1), and repression of MYC, together leading to G 1 cellcycle arrest. The TGFb growth suppressive pathway is commonly disrupted in pancreatic cancer, most often through mutation/ deletion of SMAD4, but also through inactivation/loss of TGFbRI and TGFbRII [4].
SMURF1 is a HECT-domain E3 ubiquitin ligase (E3 ubiquitin ligases carry out the third and substrate-specific step in protein ubiquitination). SMURF1 promotes the nuclear export of TGFb pathway inhibitor SMAD7 (increasing its availability), and the destruction of TGFbRI and SMAD4 (through ubiquitinationmediated degradation) [15,16]. All these activities should serve to antagonize TGFb signaling, and together provide a strong rationale for SMURF1 amplification/overexpression in pancreatic cancer. It was notable then, that SMURF1 knockdown in AsPC-1 cells did not enhance TGFb-pathway transcription (though perhaps not surprising, given the inactivating mutation of SMAD4). Therefore, the oncogenic activities of SMURF1 must act at least in part independently of its functions in TGFb signaling (at least at the pathway level of SMAD4). To this end, SMURF1 has also  [34] called ratios (red line); Screen shot of the corresponding locus from the UCSC genome browser. The dashed lines bracket the 325 Kb minimal amplicon, which spans SMURF1. (B) SMURF1 is overexpressed when gained/amplified. Box plots show 25 th , 50 th (median) and 75 th percentile transcript levels (assayed by Agilent 44K array) for specimens with (red) or without (gray) 7q21-q22 gain, for SMURF1 and its nearest gene neighbors. Note, KPNA7 was not represented on the array. *, P,0.05; **, P,0.01; ***, P,0.001 (Mann-Whitney U-test). (C) FISH reveals SMURF1 amplification in primary pancreatic cancers. Shown are two pancreatic cancers with SMURF1 amplification (center, and right), along with a non-amplified control (BxPC3 cells, left). SMURF1 locus probe (red); control chr 7 centromere (CRP7) probe (green). doi:10.1371/journal.pone.0023924.g001 been shown to dissolve tight junctions (by degradation of RhoA) during epithelial-mesenchymal transition [24], and focal adhesions (by degradation of talin heads) to potentiate cell migration [25]. Additional studies should clarify the key SMURF1 substrates linked to invasiveness and anchorage-independent growth in pancreatic cancers with 7q22 amplification.
During the progress of this work, two other studies characterized the 7q21-q22 amplicon in pancreatic cancer. Suzuki et al. [26] by genomic profiling of cell lines identified the amplicon in AsPC-1 cells, with the amplicon peak spanning 11 genes. Further efforts focused on two genes, TRRAP and SMURF1, with significantly elevated expression when amplified. However, in contrast to our study, they reported that knockdown of SMURF1 inhibited AsPC-1 cell proliferation. Notably, though, they evaluated only one siRNA. Given our results that four independent siRNAs knocked down SMURF1 levels comparably and decreased invasion without affecting cell proliferation, we suggest that their finding might reflect a non-specific or specific off-target RNAi effect. Indeed, growth inhibition is a common non-specific effect, triggered by a type I interferon response to siRNA [27]. Suzuki et al. went on to show that SMURF1 overexpression in two pancreatic cancer cell lines enhanced colony growth on tissue culture plastic. Nonetheless, our findings based on knockdown in the physiologicallyrelevant context of focal SMURF1 amplification suggest that the main oncogenic function of SMURF1 relates to promoting cell invasiveness rather than proliferation.
In another recent study, Laurila et al. [28] by FISH analysis of cell lines delimited the 7q21-q22 amplicon to 0.77 Mb spanning 10 genes (including SMURF1), but focused their efforts on ARPC1A and ARPC1B, subunits of the Arp2/3 complex functioning in actin polymerization. Using RNAi, they found that knockdown of either reduced cell motility, and knockdown of ARPC1A also reduced cell invasion. Though our minimal amplicon excluded ARPC1A and ARPC1B, it is nonetheless possible that their amplification contributes to motility/invasion in tumors where they are amplified. It is not uncommon to find multiple driver oncogenes within tumor amplicons (e.g. ref. [29]). Indeed, our own studies do not resolve whether KPNA7, within our 325 Kb minimal amplicon, might also have an oncogenic role (along with SMURF1).
To summarize, by genomic profiling and functional analysis we identified SMURF1 as an amplified oncogene driving cell invasiveness in pancreatic cancer. Perhaps of most significance, as an enzyme SMURF1 represents a tractable drug target. Other E3 ubiquitin ligases have been linked to cancer, and because of their substrate specificity E3 ubiquitin ligases are thought to be attractive targets for therapy [30]. Indeed, several small molecule inhibitors (including against MDM2, a regulator of TP53) are presently being evaluated [31]. Our findings identify SMURF1 as a possible new target for molecularly-directed therapy against the devastating disease of pancreatic cancer.

Specimens
Pancreatic cancer cell lines, described previously [12], and NIH-3T3 cells were obtained from the American Type Culture Collection (Manassas, VA). Pancreatic cancer xenografts, which effectively enrich the tumor epithelial fraction for DNA analysis, were generated as described [32] at the Johns Hopkins Hospital, with approval from the Institutional Review Board (IRB) (protocol ID 05-04-14-02) and Animal Care and Use Committee (protocol ID MO05M466). Briefly, a 1 mm 3 piece of the primary tumor was soaked in Matrigel (Collaborative Biomedical Research), then implanted subcutaneously in a nu/nu mouse. Engrafted tumors were harvested when they reached 1-2 cm in diameter, and tumor cell enrichment confirmed by H&E-stained frozen section. DNA and RNA were isolated using the Qiagen (Valencia, CA) AllPrep kit. Eleven of the 48 xenografts were previously profiled by lowerresolution CGH on cDNA arrays [6].

Array CGH
CGH was done using Agilent (Santa Clara, CA) catalogue 244K CGH arrays. DNAs were labeled as described [33], then hybridized (vs. a pool of eight sex-matched normal leukocyte DNAs) following the manufacturer's instructions. Arrays were scanned using an Agilent G2505B scanner, and data extracted and normalized using Agilent Feature Extraction software (version 9.1) with default settings. Copy number alterations were called using cghFLasso [34], and low-level gains and higher-level amplifications defined by cghFLasso tumor/normal ratios .1.3 and .3.0, respectively. CGH data detailed herein are available at GEO (GSE19852); a complete description of the dataset is in preparation.

Expression profiling
Expression profiling was done using Agilent catalogue 44K Whole Human Genome arrays. RNAs were labeled using the Quick Amp Labeling kit (Agilent), then hybridized (vs. a universal RNA reference pool of 11 cancer cell lines [35]) following the manufacturer's instructions. After scanning and data extraction (as above), expression data were normalized (mean-centered) by array and by gene, and mean-centered log 2 ratios reported.

FISH
A tissue microarray containing 105 pancreatic ductal adenocarcinoma cases (archived at Stanford University, and used with IRB approval) was previously described [6]. Probe labeling and FISH were performed using Vysis (Downers Grove, Illinois) reagents according to the manufacturer's protocols. A locusspecific BAC mapping to SMURF1 at 7q22.1 (RP11-62N3; BACPAC Resources Centre, Oakland, CA) was labeled with SpectrumOrange, and co-hybridized with SpectrumGreen-la-beled chr 7 centromere probe (CEP7; Vysis). Slides were counterstained with DAPI, and imaged using an Olympus BX51 fluorescence microscope with Applied Imaging (San Jose, CA) Cytovision 3.0 software. Twenty-five tumor cells were scored per case, and amplification defined as an average SMURF1/ CEP7 ratio .2.5.  Table S1. AsPC-1 cells were grown at 37uC in complete media of RPMI-1640 (Invitrogen, Carlsbad, CA), 10% FBS, 50 U/ml penicillin, and 50 U/ml streptomycin. For transfection, 150,000 cells were seeded per 6well plate well, and transfected using Lipofectamine 2000 reagent (Invitrogen) according to the manufacturer's protocol. Cells were transfected with a final concentration of 50 nM siRNA for 6 hrs.

Cell growth and invasion assays
Cell proliferation/viability was quantified by colorimetry based on the metabolic cleavage of the tetrazolium salt WST-1 in viable cells, according to the manufacturer's protocol (Roche, Indianapolis, IN). BrdU incorporation was determined by colorimetric ELISA using the BrdU Cell Proliferation Assay, according to the manufacturer's protocol (Cell Signaling Technology, Danvers, MA). Invasion was quantified by Boyden chamber assay (BD Biosciences, San Jose, CA). Briefly, 24 hrs after transfection, 50,000 cells were plated into 24-well Matrigel-coated inserts with a 0.5% to 10% FBS gradient. Seventy-two hrs later, cells were fixed, stained with crystal violet, and cells traversing the membrane counted. All assays were done as triplicate transfections, and all experiments were repeated at least once with similar results.

TGFb transcriptional reporter assay
Cells were co-transfected with 4 mg p3TP-Lux (Addgene, Cambridge, MA), a TGFb responsive firefly luciferase reporter containing three consecutive TPA response elements (TREs) and a portion of the plasminogen activator inhibitor 1 (PAI-1) promoter region [17], along with 0.4 mg pRL-TK (Promega, Madison, WI) expressing Renilla luciferase as an internal normalization control. Luciferase activity was assayed 48 hrs after transfection (by Lipofectamine 2000) using the dual luciferase reporter assay system (Promega, Madison, WI). Reporter activity is expressed as the ratio of firefly/Renilla. Assays were done as triplicate transfections, and repeated at least once with similar results.

Anchorage-independent growth
A pGIPZ shRNAmir construct targeting SMURF1 (V2LHS_229724), along with a non-targeting pGIPZ shRNAmir control, were obtained from Open Biosystems (Huntsville, AL). To create stably-transduced AsPC-1 cell pools, lentiviral constructs (along with Trans-lentiviral packaging mix plasmids) were transfected into 293TN producer cells (System Biosciences, Mountain View, CA), and supernatant packaged virus transduced into AsPC-1 cells following the manufacturer's instructions (Open Biosystems' Trans-lentiviral GIPZ packaging protocol). Two days post-infection, 1 mg/ml puromycin (Invitrogen) was added to the culture medium, and cells selected for 14 days. Anchorage independent growth was assayed by colony formation in soft agar. Briefly, 10,000 cells were embedded in a 6-well plate well within a top layer of 0.36% agarose in complete media, over a layer of 0.48% agarose in complete media. Cells were grown for 14-21 days, then visible colonies counted after staining with 0.015% Neutral Red solution. Assays were done as triplicate transductions, and repeated at least once with similar results.

NIH-3T3 focus formation assay
Full-length human SMURF1 cDNA expression vector, pcDNA3.1-SMURF1 was a kind gift from Di Chen (University of Rochester Medical Center, Rochester, NY), and the parent vector pcDNA3.1 was purchased from Invitrogen (Carlsbad, CA). A catalytic mutant SMURF1 (C699A) [22] was engineered using the QuickChange XL II Site-Directed Mutagenesis Kit from Stratagene (La Jolla, CA), with the following mutagenic primers: 59-CGTGGAGGAGACCGCCGGGTTTGCTGTGG -39 (de- generate, mutated bases denoted by bold text) and 59-CCACAG-CAAACCCGGCGGTCTCCTCCACG-39. Fifty thousand cells were seeded per 60 mm plate, and 8 mg of plasmid was transfected by Lipofectamine 2000 reagent (Invitrogen) according to the manufacturer's protocol. Two days after transfection, cells from each 60 mm plate were re-plated into two 10 cm plates and grown to confluence over 28 days. Visible foci were counted after methanol fixation and Giemsa staining. Assays were done as triplicate transfections, and repeated at least once with similar results.