Epithelial-to-Mesenchymal Transition and Ovarian Tumor Progression Induced by Tissue Transglutaminase

Tissue transglutaminase (TG2), an enzyme overexpressed in epithelial malignancies (1, 2), participates in Ca²⁺-dependent protein posttranslational modifications and cross-linking (3). We reported that TG2 is upregulated in ovarian tumors compared with normal ovarian surface epithelium (1, 4). The enzyme exerts pleiotropic functions in tumors. It strengthens integrin-dependent cell adhesion, remodels the extracellular matrix (ECM), and modulates intracellular signaling (5). In cancer cells, TG2 activates focal adhesion kinase, protein kinase B (Akt; ref. 6), cyclic AMP response element binding protein (7), and the nuclear factor-κB (NF-κB) complex (2). Activation of oncogenic signaling affects the invasive behavior of tumor cells. We recently showed that TG2 plays a role in i.p. dissemination of ovarian cancer cells (1) and regulates at the transcription level the secretion of matrix metalloproteinase-2 (MMP-2; ref. 7). To better understand its function in metastasis, we investigated the involvement of TG2 in epithelial-to-mesenchymal transition (EMT), a critical process in cancer progression.

EMT is characterized by breakdown of cell junctions and loss of cell polarity, rendering epithelial cells motile and invasive (8). EMT plays an important role in development, particularly in gastrulation and neural crest migration (9). In cancer, EMT causes cells to lose epithelial characteristics and acquire a mesenchymal phenotype, initiating invasion and metastasis. A critical component is the loss of type I cadherins that maintain stable cell-cell contacts through adherens junctions and desmosomes (10, 11). The extracellular immunoglobulin-like domains of E-cadherin bond neighboring cells through junctional complexes (12, 13). To preserve cellular shape and polarity, the intracellular domains of cadherins connect to the actin cytoskeleton through α-catenin and β-catenin (14). Downregulation of E-cadherin in cancer cells is a critical disruptor of epithelial homeostasis, augmenting cell invasiveness (8, 12, 15). A “cadherin switch” has been described, whereby E-cadherin loss is accompanied by gain of N-cadherin (16), a mesenchymal marker (17). The molecular mechanisms of this switch are not known (18).

Loss of E-cadherin during cancer progression is regulated genetically and epigenetically. Promoter hypermethylation in breast cancer cells downregulates E-cadherin expression (19). Genetic mechanisms involve transcriptional repression at Ets sites or palindromic E-boxes (20) by repressors, such as the zinc-finger domain–containing factors Snail, Slug, Zeb1, and Zeb2 and the basic helix-loop-helix factors E47 and Twist (21, 22). Their function is regulated by oncogenic pathways, particularly by ras (23), Src (24), Akt (25), GSK-3β (26), and NF-κB (27). Integrin signaling triggered by anchoring of cancer cells in the ECM activates integrin-linked kinase (ILK; refs. 28, 29), which regulates E-cadherin function and the assembly of actin filaments by modulating the balance between Rho and Rac (28).

Given the role of TG2 in cell adhesion modulated by β-integrins (1, 30) and its role in activating oncogenic signaling (2), we speculated that it may be involved in EMT. Here we show that ovarian cancer cells expressing TG2 adopt a mesenchymal phenotype, characterized by cadherin switch and invasive behavior in a Matrigel matrix. TG2 modulates E-cadherin loss at the transcriptional level by altering the expression of several repressors including Zeb1. This function is mediated through activation of the NF-κB complex by TG2. We show by using an orthotopic xenograft ovarian model that TG2-dependent induction of EMT leads to increased tumor formation, peritoneal metastases, and malignant ascites.

Human SKOV3 and OV90 ovarian cancer cell lines from the American Type Culture Collection were cultured in growth medium containing 1:1 MCDB 105 (Sigma) and M199 (Cellgro) supplemented with 10% fetal bovine serum (Cellgro) and 1% antibiotics.

To knock down TG2, an antisense construct (AS-TG2; ref. 31) cloned into pcDNA3.1 was transfected in SKOV3 cells, and stable clones were selected, as previously described (1). To overexpress TG2, TG2 subcloned in pQCXIP was transduced into OV90 cells, and stably expressing pooled colonies were selected, as described (7). To overexpress Zeb1 and reconstitute NF-κB and Akt activities in SKOV3 cells stably transfected with AS-TG2 or vector, full-length Zeb1 cDNA, constitutively active p65 subunit of NF-κB, and a constitutively active form of Akt that lacks the pleckstrin homology domain were transferred by retroviral infection and pooled colonies were selected. These vectors were previously described (27). Transfection of short interfering RNA (siRNA) was done using Dreamfect (Oz Biosciences) and siRNA targeting Zeb1 (Dharmacon) or scrambled siRNA (control).

Cells were lysed in radioimmunoprecipitation assay buffer containing protease inhibitors. Equal amounts of protein were separated by SDS-PAGE and electroblotted onto polyvinylidene difluoride membranes (Millipore). After blocking, membranes were probed with primary antibodies against TG2 (Neomarkers), glyceraldehyde-3-phosphate dehydrogenase (GAPDH; Biodesign International), Zeb1 (Santa Cruz), E-cadherin (Cell Signaling), N-cadherin (Cell Signaling), and vimentin (Sigma). After incubation with horseradish peroxidase–conjugated secondary antibodies, antigen-antibody complexes were visualized using enhanced chemiluminescence (Amersham Biosciences). Images were captured with an image analyzer (LAS 3000, Fuji Film) and densitometry analysis was done with MultiGauge 3.0 software (Fujifilm USA, Inc.). Western blots were repeated in independent conditions at least twice; representative blots are shown.

SKOV3 cells stably transfected with AS-TG2 or vector (pcDNA3.1) were injected orthotopically under the ovarian bursa of 7- to 8-wk-old female nude mice (nu/nu BALB/c) from Harlan. For the orthotopic injection, mice were anesthetized with a cocktail of acepromazine and Torbugesic, and a dorsal incision exposed the left ovary enshrouded in its oviductal fimbriae. Using a 27-gauge needle attached to a piece of polyethylene tubing (PE-20) and held in place with a micromanipulator, 1.0 × 10⁶ cells diluted in 5 μL of growth medium were injected into the ovarian bursa. Eight weeks after the injection, mice were euthanized and necropsy was done. In this first experiment, the SKOV3-pcDNA3.1 and SKOV3/AS-TG2 cells induced ovarian tumors but did not produce metastases. To produce a more aggressive phenotype, tumors were minced, treated with hyaluronidase, and cultured in medium supplemented with G418. Resultant xenograft-derived cell lines, SKOV3-pcDNA3.1× and SKOV3-AS-TG2×, were used in subsequent experiments.

Ovarian tumors were measured bidimensionally with calipers, and tumor volume was calculated as L × W²/2, where L is length and W is width. Peritoneal milliary seeding was estimated based on the presence of >10 disseminated nodules (yes/no), and peritoneal metastatic implants were counted. Because the primary goal was assessment of metastasis, animals that did not develop tumors within the ovary or that had discernable leakage at the time of injection were excluded from analysis. Four independent rounds of injection were done, with 10 to 11 mice being used each time. Results of these four rounds were analyzed and are presented together. Animal experiments were approved by the Indiana University Animal Care and Use Committee as being in compliance with federal regulations.

Invasion assay was done by using two-dimensional and three-dimensional cell cultures in a Matrigel matrix (BD-Biosciences). Briefly, 2.5 × 10⁴ cells suspended in 50 μL of cell culture medium were seeded onto solidified Matrigel in 24-well plates as a monolayer (two-dimensional model) or mixed and embedded within the matrix (three-dimensional model). After incubation at 37°C, invasive cells adhered to the surface of the gel and spread to form networks (two-dimensional model) or degraded and invaded into the gel (three-dimensional). These were observed under an inverted microscope and photographed live. The experiments were done in duplicate and repeated in independent conditions.

Dual-Luciferase Assay (Promega) was done to quantify E-cadherin promoter activity in pcDNA3.1 and AS-TG2 SKOV3 cells. In brief, cells were transiently cotransfected with the experimental reporter plasmid, E-cadherin promoter firefly luciferase, containing the first 500 nucleotides and four E-box elements (32), and control reporter plasmid, cytomegalovirus promoter Renilla luciferase, at a ratio of 10:1 by using DreamFect Gold reagent (OZ Biosciences). Luminescence was measured by using a TD-20/20 luminometer (Turner Biosystems). Experiments were done in triplicate and repeated. To control for transfection efficiency, values for luminescence were normalized to Renilla activity.

Tumors harvested 8 wk after orthotopic injection were examined by immunohistochemistry for TG2, as described (1). For E-cadherin, a monoclonal antibody at 1:100 concentration (DAKO) and the avidin-biotin peroxidase system (DAKO) were used. Negative controls were run in parallel. Slides were scored from 0 to 3+, noting the percentage of staining cells, by a pathologist blinded to the identity of the samples.

RNA extracted by using the RNA STAT-60 reagent (Tel-Test, Inc.) was reverse-transcribed using the iScript cDNA synthesis kit (Bio-Rad). The PCR reactions used Taq polymerase (Promega) and the primers are described in Supplementary Table S1. For real-time PCR, we used the FastStart TaqMan Probe Master (Rox, Roche) on an ABI Prism 7900 platform (Applied Biosystems) according to the manufacturer's procedures. The primers and probes used are given in Supplementary Table S2. At the end of the PCR reaction, a melting curve was generated and the cycle threshold (C_t) was recorded for the reference gene and for GAPDH. Relative expression of gene of interest was calculated as ΔC_t, measured by subtracting the C_t of the reference gene from that of the control. Results are presented as mean ± SD of replicates. Each experiment was done in duplicate and repeated in independent conditions.

SKOV3 cells plated on fibronectin-coated chamber slides (BD Biosciences) were fixed in 4% paraformaldehyde and permeabilized with Triton X-100. After blocking, cells were incubated with primary antibodies, anti–E-cadherin (1:200), Cy5-labeled anti-vimentin (Sigma, 1:200), and isotype-specific IgG (control), followed by incubation with Alexa Fluor 488–labeled secondary antibody (1:1,000; Molecular Probes). Images were captured using a Zeiss LSM 510 confocal microscope system under UV excitation at 488 nm (Alexa Fluor 488), 630 nm (Cy5), and 340 nm [4′,6-diamidino-2-phenylindole (DAPI)].

Fisher's exact test was used to compare metastasis and ascites formation and intensity of E-cadherin expression between experimental groups. Student's t test compared mean tumor volumes, relative luciferase activity, and mRNA expression levels between groups. A P value of <0.05 was deemed significant. We estimated the Pearson correlation coefficient between mRNA expression levels for TG2 and E-cadherin repressors, Zeb1 and Zeb2, Twist1 and Twist2, and Slug and Snail1, by using the data downloaded from GEO⁷

Having previously shown that TG2 enhances i.p. tumor dissemination (1, 34), we sought to determine whether the enzyme plays a role in EMT. We used two ovarian cancer cell lines, one expressing endogenous TG2 (SKOV3) and one lacking TG2 expression (OV90). SKOV3 cells were stably transfected with an antisense TG2 construct (AS-TG2) or control vector (pcDNA3.1), and OV90 cells were stably transduced with full-length TG2 or control vector (pQCXIP). Decreased expression level of TG2 in AS-TG2 cells compared with vector-transfected cells was shown by Western blotting (Fig. 1A,, top left). Control cells expressing endogenous TG2 displayed poorly organized adhesive junctions and loss of cell polarity and exhibited a fibroblast-like morphology. In contrast, cells stably transfected with AS-TG2 appeared compact, flat, and cohesive (Fig. 1A,, top right). A similar difference in phenotype was noted in OV90 cells with or without TG2 overexpression. Increased expression level of TG2 in TG2-transduced compared with vector-transduced OV90 cells was shown by Western blotting (Fig. 1B,, bottom left). pQCXIP-transduced cells are compact and organized in tight structures, whereas TG2-transduced cells are elongated and dispersed (Fig. 1B , bottom right). These observations suggest that increased TG2 expression was associated with a mesenchymal morphology.

To confirm this, we examined the expression of E-cadherin and other EMT markers in cells engineered to express variable levels of TG2. A significant increase in E-cadherin expression was observed in AS-TG2 compared with control cells (Fig. 1B,, left). Correspondingly, decreased E-cadherin expression was noted in OV90 cells transduced with TG2 compared with control cells (Fig. 1B,, middle). Knockdown of TG2 in SKOV3 cells was associated with a cadherin switch characterized by increased E-cadherin expression and decreased N-cadherin expression compared with control cells. Expression levels of other mesenchymal markers (fibronectin and vimentin) were decreased in cells transfected with AS-TG2 compared with control cells (Fig. 1B,, right). This phenotype was confirmed by immunofluorescent staining. SKOV3 control cells dimly expressed E-cadherin and strongly stained with a vimentin antibody. In contrast, E-cadherin expression was increased and vimentin staining was diminished in AS-TG2 cells (Fig. 1C).

EMT renders cells motile and invasive. We next examined whether the mesenchymal morphology observed in TG2-expressing cells is associated with invasiveness. For this, cells were plated in two-dimensional and three-dimensional cultures in Matrigel matrix. SKOV3 cells transfected with vector formed networks in Matrigel, whereas AS-TG2 cells remained clumped together (Fig. 1D). Likewise, OV90 cells transduced with TG2 invaded through the matrix as trabecullae, whereas vector-transduced OV90 cells grew in cohesive, noninvading clumps (Fig. 1D). These results show that TG2 promotes EMT and invasiveness of ovarian cancer cells.

The process of EMT is linked to initiation of metastasis. We previously showed that TG2 enhances i.p. tumor dissemination (1). However, the i.p. model artificially dispersed ovarian cancer cells in the peritoneal cavity and did not appreciate the first steps of metastasis, specifically invasion and shedding of cells from the ovary. To overcome these limitations, here we used an orthotopic ovarian xenograft model, where control and AS-TG2–transfected SKOV3 cells were injected into the ovarian bursa of nude mice.

In an initial experiment, necropsies were done 8 weeks after ovarian cancer cell implantation. Primary ovarian tumors were formed in mice injected with either AS-TG2 or control cells and their sizes were similar. However, no distant metastases or malignant ascites were generated (Supplementary Table S3). Knowing that the metastatic potential of tumor cells is enhanced through animal passaging, we used ovarian tumors harvested from this experiment to generate xenograft-derived cultures. Western blotting showed that the expression level of TG2 in xenograft-derived cells remained lower in AS-TG2 xenograft–derived cells (AS-TG2×) compared with pcDNA3.1 xenograft–derived cells (pCDNA3.1×; Fig. 2A,, left). These cells, passaged once as ovarian xenografts, were used for a new round of orthotopic ovarian implantation. After 8 weeks, a significant difference in tumor development was observed (Fig. 2A; Table 1). The mean volume of primary tumors was significantly smaller in mice injected with AS-TG2× cells compared with that in mice injected with pcDNA3.1× cells (600 ± 187 versus 1,468 ± 257 mm³; P = 0.01; Fig. 2B). Mice injected with pcDNA3.1× cells developed disseminated metastatic implants on the omentum, mesentery, and peritoneal surface of abdominal flanks compared with mice injected with AS-TG2× cells. Specifically, 13 of 17 mice injected with pcDNA3.1× cells developed disseminated milliary metastases [>10 implants; Fig. 2A, (middle) and C], compared with 7 of 19 mice injected with AS-TG2× cells [P = 0.02; Fig. 2A, (right) and C]. The average number of peritoneal implants in mice injected with control cells was 24 ± 5 and in mice injected with AS-TG2 cells was 9 ± 3 (P = 0.04; Table 1). Ascites formation was also different between the two groups, with 10 of 17 mice injected with pcDNA3.1× cells having developed ascites compared with 4 of 19 mice injected with AS-TG2× cells (P = 0.038; Table 1). TG2 expression was verified by immunohistochemistry. Suppressed TG2 expression was observed in tumors generated from AS-TG2× cells (Fig. 2D,, sections 3 and 4) compared with tumors generated from pcDNA3.1× cells (Fig. 2D , sections 1 and 2). Occasional small and isolated islands of TG2-positive cells were observed in AS-TG2× tumors, consistent with the emergence of TG2-positive subpopulations in the absence of G418 selection in vivo.

To investigate the effects of TG2 on markers of EMT in vivo, E-cadherin expression was assessed by immunohistochemistry (Fig. 3A). Twelve of 16 tumor specimens derived from AS-TG2× cells stained strongly (2+ and 3+) for E-cadherin, compared with 5 of 18 pCDNA3.1× derived xenografts (P = 0.015; Fig. 3B). For few E-cadherin–positive pcDNA3.1× xenografts, available metastatic implants were immunostained. Distant implants displayed negative E-cadherin staining, consistent with the concept that E-cadherin is downregulated at metastatic sited compared with primary tumor sites (Supplementary Fig. S1). These data show that TG2-induced E-cadherin repression is maintained in vivo, inversely correlating with the metastatic potential of the xenografts.

We next investigated the mechanisms of E-cadherin repression by TG2. Reverse transcription-PCR (RT-PCR) showed that E-cadherin mRNA level was upregulated in AS-TG2 cells compared with control cells (Fig. 4A). In contrast, E-cadherin mRNA level was downregulated in OV90 cells transduced with TG2 compared with vector-transduced cells (Fig. 4A). A gene reporter assay in which AS-TG2 and control cells were transfected with an E-cadherin-luciferase reporter construct (32) showed that its reporter activity was increased 2.6-fold by TG2 knockdown (Fig. 4A , bottom; P < 0.001). These data show that TG2 negatively regulates E-cadherin at the transcriptional level.

Several transcriptional repressors are recognized as critical modulators of EMT, including Snail, Slug, Zeb1, Zeb2, and Twist (21, 22, 35, 36). To find out whether TG2 expression is associated with altered mRNA expression of these repressors, semiquantitative RT-PCR and real-time PCR were used. Diminished mRNA expression levels for Slug, Zeb1, and Zeb2, but not for Twist1, Twist2, Snail1, and Snail 3, were observed in AS-TG2 compared with control cells (Fig. 4B,, left). Real-time PCR showed that Zeb1, Zeb2, and Slug mRNA expression levels were decreased by ∼80% in AS-TG2–transfected cells compared with controls, but that Snail1 expression level was not significantly different between the two cell types (Fig. 4C,, top). Similar trends were observed in OV90 cells. Slug, Zeb1, and Zeb2 mRNA levels were increased in OV90-TG2 compared with control cells. In addition, Twist1 and Snail 3 mRNA levels were slightly upregulated in TG2-overexpressing cells (Fig. 4B,, right). Real-time PCR confirmed a 75% increase in mRNA expression level for Zeb1 and Zeb2 in OV90-TG2 compared with control cells, but no significant difference for Slug and Snail1 (Fig. 4C , bottom). These data suggest that in ovarian cancer cells, TG2 expression correlates with expression of E-cadherin transcriptional repressors, particularly with that of Zeb1 and Zeb2.

Next, we investigated in human tumors the correlation between expression of TG2 and that of E-cadherin repressors by using an existing database derived from microarray gene expression analysis of 285 ovarian tumors (37). Statistically significant correlations were found between TG2 and Zeb1 (Pearson correlation coefficient = 0.24, P < 0.001; see Fig. 4D), between TG2 and Slug (Pearson correlation coefficient = 0.27, P <0.001; see Fig. 4D), and, to a lesser degree, between TG2 and Snail2 (Pearson correlation coefficient = 0.14, P = 0.02). TG2 mRNA expression level did not correlate with that of Twist1, Twist2, or Zeb2 in this database (Supplementary Table S4).

Based on these results and on a recent report implicating Zeb1 in ovarian epithelial cell transformation (38), we focused on the effects of TG2 on Zeb1. Zeb1 expression was confirmed at the protein level. Consistent with RT-PCR data, lower Zeb1 expression was observed in AS-TG2 compared with control cells (Fig. 5A,, left). To test whether Zeb1 negatively regulates E-cadherin in ovarian cancer cells, we used siRNA targeting Zeb1. E-cadherin expression level was upregulated in SKOV3 cells transfected with Zeb1 siRNA compared with cells transfected with control siRNA (Fig. 5A , right).

To show the significance of Zeb1 to regulation of E-cadherin expression by TG2, we stably overexpressed Zeb1 in SKOV3 cells where TG2 was knocked down. The expression level of E-cadherin, which was markedly increased by TG2 knockdown compared with control cells (Fig. 5B,, left, lane 2 versus lane 1), was reduced significantly by overexpression of Zeb1 (Fig. 5B,, left, lane 3 versus lane 2). Overexpression of Zeb1 in cells with decreased expression of TG2 led to restoration of the invasive phenotype, comparable to that of cells expressing TG2, as assessed by a two- or three-dimensional Matrigel invasion assay (Fig. 5B , right). These data suggest that Zeb1 plays a critical role in TG2-mediated regulation of E-cadherin expression and the invasive phenotype of ovarian cancer cells.

Given the role of TG2 in activating the NF-κB complex and Akt (2, 34), and the known functions of these pathways in modulating EMT (25, 27), we investigated whether the effects of TG2 on Zeb1 and EMT are induced as a consequence of Akt or NF-κB activation. To test this, we overexpressed the p65 subunit of NF-κB or a constitutively active form of Akt (CA-Akt) in ovarian cancer cells where TG2 was knocked down and compared the expression levels of Zeb1 and E-cadherin to those in cells expressing endogenous TG2. Overexpression of p65 in AS-TG2 cells was confirmed by Western blotting for IκBα (Fig. 5C,, left) and overexpression of CA-Akt was verified by Western blotting for Akt (Fig. 5C,, middle). Zeb1 expression level was lower in AS-TG2 cells compared with control cells, and this corresponded to increased E-cadherin expression level (Fig. 5C,, right, lane 2 versus lane 1). Overexpression of p65 in AS-TG2 cells rescued Zeb1 expression level, repressing in turn E-cadherin level (Fig. 5C,, right, lane 3 versus lane 2). As assessed by a Matrigel invasion assay, overexpression of p65 in AS-TG2 cells led to an invasive phenotype comparable to that of control cells (Fig. 5D), suggesting that an important element in the process of TG2-induced EMT is the repressor Zeb1 induced by NF-κB. Overexpression of CA-Akt in AS-TG2 cells also repressed E-cadherin level (Fig. 5C,, right, lane 4 versus lane 2) and induced an invasive phenotype (Fig. 5D). However, CA-Akt overexpression did not affect Zeb1 (Fig. 5C , left, lane 4 versus lane 2), suggesting that other mechanisms are implicated downstream of Akt. Collectively, these data indicate that TG2 negatively regulates E-cadherin expression, leading to EMT, increased cancer cell invasiveness, and metastasis.

We previously showed that TG2 plays a role in peritoneal dissemination of ovarian tumors (1). In the current study, we provide strong evidence for its role in metastasis and present novel mechanistic insight into how TG2 facilitates tumor spread. We show that TG2 induces a mesenchymal phenotype and that this, in turn, translates into increased cell invasiveness and tumor metastasis. This is mediated by TG2 through effects on transcriptional repressors that regulate E-cadherin expression. To our knowledge, this is the first report implicating TG2 in EMT and establishes a critical link between the enzyme, cancer cell invasion, and metastasis.

Several oncogenic signal transduction pathways have been implicated in regulating E-cadherin expression in cancer cells (25, 27, 39). For instance, the NF-κB pathway, constitutively activated in many epithelial tumors, induces EMT by regulating the function of the Zeb family of transcription factors (27). Recent reports implicate Zeb1 and Zeb2 in mesenchymal transformation of ovarian cancer cells under the control of microRNA 200 (40). The results presented here point to Zeb1 as an important EMT regulator, downstream of TG2 and of the activated NF-κB complex. We recognize that other mechanisms may be recruited by TG2 to repress E-cadherin expression. A potential pathway may be activation of the serine/threonine kinase Akt by TG2 (6). It has been shown that Akt represses E-cadherin transcription by stabilizing the function of the repressors Snail and Slug (25), and it is possible that a similar mechanism may operate in ovarian cancer cells.

Aside from the general implications of EMT in tumor progression (8), there are specific nuances about EMT in ovarian cancer. It has been suggested that after shedding from the primary site, ovarian cancer cells float in the peritoneal cavity as cell aggregates or spheroids (33). Within spheroid structures, cells adopt a mesenchymal phenotype (33, 41) that is closely regulated by cytokines and the peritoneal hormonal milieu. Such influences may include regulation by estrogen (42) or transforming growth factor β (38), secreted in the peritoneal fluid. The mesenchymal phenotype allows cells within spheroids to become invasive when in contact with the mesothelium (41). We observed decreased spheroid formation by AS-TG2 cells compared with control cells in nonadherent culture conditions (not shown) supporting this function. Mesenchymal transformation of epithelial ovarian cells also increases secretion of MMPs (43), which remodel the ECM (44). Digestion of ECM by MMPs is critical to establishment of metastases and sprouting of new vessels. Recent data from our laboratory showed that TG2 regulates MMP-2 secretion by ovarian cancer cells (7). Those data in concert with the present findings support a critical role of the enzyme in regulating the invasiveness of ovarian cancer cells.

The use of orthotopic ovarian xenografts rather than i.p. tumor implantation (1) provides stronger evidence for the role of TG2 in metastasis. The orthotopic approach better mimics the phenomena of primary tumor formation and of tumor spreading initiated from the primary site. Critical steps of this process are invasion outside of the ovarian bursa followed by i.p. dissemination with invasion of the mesothelium, which cannot be appreciated with i.p. xenografts (45, 46). Previous studies using orthotopic ovarian models used either implantation of human (46) or murine (47) tumor fragments adjacent to the ovary or injection of cancer cell lines under the ovarian bursa (45). The approach using tumor fragments microsurgically sutured to the oviductal fimbriae cannot appreciate invasion of cells out of the bursa into the peritoneal space. Intrabursal injection of ovarian cancer cells causes tumor formation in the ovary; however, invasion and peritoneal spreading may not occur (45). In our first experiment, SKOV3 cells injected intrabursally formed large ovarian tumors but did not cause peritoneal dissemination. After only one passage as xenografts, cells injected intrabursally led to omental and mesenteric metastases. Passaging of cell lines as xenografts permits selection of more invasive cells with increased tumorigenicity (48). The orthotopic ovarian model using the animal-passaged cancer cells described here accurately reconstitutes primary tumor formation in the ovary followed by peritoneal dissemination after invasion through the bursa, recapitulating the steps of metastasis observed during the progression of ovarian cancer.

We show, using this model, that TG2 knockdown decreases tumor growth and peritoneal dissemination. The current findings are concordant with our previous results (1) and with a report showing a strong association between TG2 expression and adverse clinical outcome in ovarian cancer (49). The proinvasive and metastatic functions of TG2 may be specific to particular epithelial cancer types [e.g., pancreatic (50) and ovarian (1) cancers] that rely on peritoneal dissemination. The EMT phenotype has been also associated with chemotherapy resistance and may be implicated in TG2-mediated chemotherapy resistance reported by our group and by others (6, 34). Collectively, these data show that activation of oncogenic signaling by TG2 induces EMT and enhances tissue invasiveness and tumor dissemination.

No potential conflicts of interest were disclosed.

Grant support: U.S. Department of Veterans Affairs through a VA Merit Award (D. Matei).

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

We thank Drs. Tucholski (University of Alabama, Birmingham, AL) and Martignetti and Difeo (Mount Sinai Medical School, New York, NY) for the constructs, and Drs. Kenneth Nephew, Mircea Ivan, and David Donner for careful review of the manuscript.