A Novel TGFβ Trap Blocks Chemotherapeutics-Induced TGFβ1 Signaling and Enhances Their Anticancer Activity in Gynecologic Cancers Free

Gynecologic cancers, including cervical cancer and ovarian cancer, are leading causes of cancer-related death in women worldwide. Ovarian cancer is the most lethal gynecologic malignancy with an estimated 238,700 new cases and 151,900 deaths in 2012 worldwide (1). Cervical cancer is the fourth most commonly diagnosed cancer and the fourth leading cause of cancer-related death among women worldwide, which accounts for nearly 3.7% of the total newly diagnosed cancer cases and 3.3% of the total cancer deaths in 2012 (1). Chemotherapy remains the main treatment option for ovarian cancer and advanced/relapsed cervical cancer. However, the response to chemotherapeutic treatment is often inadequate, with a subpopulation of tumor cells surviving chemotherapy. Consequently, relapse is a common event and these recurrent tumors are associated with increased aggressiveness, resistance to various chemotherapeutics, and a high mortality rate. Thus, chemotherapy resistance remains a major therapeutic hurdle in the management of gynecologic malignancies and novel strategies to enhance antitumor activity of chemotherapeutics agents are urgently needed.

TGFβ belongs to a family of homodimeric peptide growth factors that regulate a wide variety of cellular processes, including proliferation, differentiation, invasion, immunosurveillance, and stem cell maintenance (2). There are three mammalian isoforms of TGFβ ligand, named TGFβ1, -β2, and -β3 with significant homology and similarities in function. All three isoforms are secreted in a latent form and are activated via various mechanisms (3). Active TGFβ binds to three different cell surface receptors called type I (TβRI), type II (TβRII), and type III (TβRIII) receptors (4). TβRI and TβRII are serine/threonine kinase receptors, whereas TβRIII, also known as betaglycan (BG), serves as an accessory ligand-binding receptor. TGFβ ligand signals through TβRII, which recruits and activates TβRI kinase through transphosphorylation. The activated TβRI phosphorylates intracellular Smad2 and Smad3, which then interact with Smad4 protein to regulate gene expression in the nucleus (5). Overexpression of TGFβ1 has been reported in various gynecologic malignancies including ovarian cancer and cervical cancer (6–9). In patients with ovarian cancer and cervical cancer, elevated levels of TGFβ1 were associated with tumor progression, treatment resistance, and poor outcome (6, 9, 10). These pro-malignant functions of TGFβ can be collectively attributed to its unique abilities in modulating tumor microenvironment and stimulating stem cell-like features in tumor cells (8, 11).

Given its multifaceted role in driving malignant progression, it is important to identify modalities that can activate TGFβ signaling and to develop effective means for the blockade of TGFβ signaling. We and others have previously shown that some chemotherapeutic agents, such as doxorubicin, paclitaxel, and irradiation can activate TGFβ signaling, which in turn causes therapy resistance (12–15). As a result, administration of TβRI kinase inhibitor or TGFβ neutralizing antibody significantly enhanced antitumor activity of chemo- and radiation therapy in mouse models of breast cancer (12, 13, 15). However, the underlying mechanisms of how TGFβ pathway is activated by chemotherapeutics causing chemo-resistance are not well understood. Furthermore, whether the stimulation of TGFβ signaling during chemotherapy is unique to certain drugs such as doxorubicin and paclitaxel or more universal to various classes of chemotherapeutic drugs remains to be determined.

In this study, we investigated the effects of four chemotherapeutic agents commonly used in gynecologic cancers, including cisplatin, paclitaxel, doxorubicin, and camptothecin, on TGFβ signaling in cervical and ovarian cancer cells because the TGFβ signaling pathway was found to be strongly activated pathways by chemotherapy in ovarian cancer and cervical cancer. Using a novel TGFβ trap protein, we demonstrate that these drugs activate TGFβ signaling by stimulating TGFβ1 production via transcriptional and/or posttranscriptional mechanisms and the sequestration of TGFβ by the novel TGFβ trap protein was highly effective in blocking the TGFβ-like activities of the drugs in vitro and in vivo.

This study was approved by the ethical committee of the Second Affiliated Hospital of Wenzhou Medical University and conducted according to the Helsinki declaration. Informed consent was obtained from all subjects prior to participation in the study. Cervical tissue sections used for this study were cut from leftover tissue blocks from consented patients, who were treated with neoadjuvant chemotherapy. All animal experiments were conducted following appropriate guidelines. They were approved by the Institutional Animal Care and Use Committee and monitored by the Department of Laboratory Animal Resources at the University of Texas Health Science Center at San Antonio.

Patients with stage IB2 or IIA2 (bulky, primary tumor >4 cm in diameter) cervical cancer seen at the Second Affiliated Hospital of Wenzhou Medical University, Wenzhou, China, between January 2007 and August 2014, were recruited for a pilot study aimed to identify predictive biomarkers for responses to neoadjuvant chemotherapy. A total of 30 subjects were enrolled with a median age of 44 years (range, 25–59 years). None of the patients had received any antitumor therapy before the specimen collection. Paired tumor samples from each patient were obtained during cervical biopsy (pre-chemotherapy) or surgery (post-chemotherapy).

All eligible patients received one or two courses of cisplatin-based neoadjuvant chemotherapy, as previous described (16): cisplatin, 60 mg/m² on day 1; 5-fluorouracil, 750 mg/m² on day 1; mitomycin (8 mg/m²) on day 1 for one or two courses, every 28 days. All of these chemotherapeutics were administrated via uterine artery injection. All patients were treated with radial hysterectomy and bilateral pelvic lymphadenectomy two to 3 weeks after completion of the neoadjuvant chemotherapy regimen as described previously (17).

Gene expression profiles of malignant carcinoma samples from ovarian cancer patients were obtained from GEO (GSE7463) (18). Only the carcinoma samples were included in this study. Probes were normalized by quantile normalization with preprocess Core in R (19). Differential gene expression analysis was performed by comparing samples from ovarian cancer patients treated with neo-adjuvant chemotherapy (source_name_ch1 as Cancer) to samples from patients without chemotherapy treatment (source_name_ch1 as Carcinoma), controlling tumor stage and histology, by limma in R (20). Significantly differentially expressed genes were defined as those with FDR below 0.05. Upstream regulator prediction was analyzed through the use of Qiagen's ingenuity pathway analysis (IPA; Qiagen) with default setting. Gene Ontology of Biological Processes was analyzed in DAVID Bioinformatics Resources 6.7 (21, 22). Heatmap of significantly differentially expressed genes was plotted with heatmap.2 (gplotsin R; ref. 23).

Human cervical cancer cell lines (HeLa and C-4I) and ovarian cancer cell lines (OVCAR-3 and TOV-21G) were originally obtained from the ATCC. All four cell lines were authenticated with the DNA markers used by ATCC. Our cell lines were all stocked mycoplasma free in liquid nitrogen tanks and used for experiments for 4 to 5 months with at least one additional test for mycoplasma. The mycoplasma-free cultures were maintained in RPMI1640 medium (Invitrogen) supplemented with 10% FBS and 0.1% penicillin in 5% CO₂ at 37°C. For drug treatment, cisplatin, doxorubicin, paclitaxel, and camptothecin were purchased from Sigma and dissolved in water (cisplatin) or DMSO (doxorubicin, paclitaxel, and camptothecin). Their aliquots were stored at −80°C.

RER was designed and synthesized in our laboratory (24). Briefly, RER was produced by transient transfection of HEK293F cells grown in suspension in Freestyle 293 medium at 8% CO₂, 80% humidity, and rotating at 80 rpm (Infors HT). The proteins were purified from the conditioned medium seven days posttransfection using a combination of metal affinity and size exclusion chromatography as previously described (24). HTS466284 (HTS) used in our study was reported previously to be an ATP-competitive inhibitor of the TGFβRI kinase domain (25, 26). The chemical name of the compound is [3-(pyridine-2yl)-4-(4-quinonyl)]-1Hpyrazole, which was synthesized according to the procedure described by Sawyer and colleagues (25).

Both cell and tissue samples were homogenized and lysed in Laemmli buffer with a cocktail of protease inhibitors. The total protein concentrations were quantified by the BCA protein assay (Thermo Scientific). Equal amounts of total protein were resolved by SDS PAGE, transferred to a nitrocellulose membrane under constant voltage and blocked with TBST containing 5% nonfat dried milk. Primary antibodies and secondary antibodies were diluted in TBST or 3% nonfat dried milk and applied with a washing step in between. Proteins were detected using the Amersham ECL Western blotting Detection Kit (GE Healthcare). Primary antibodies used including: phosphorylated Smad2 (P-Smad2) (Cell Signaling Technology), P-Smad3 (Abcam), T-Smad2 (Cell Signaling Technology), T-Smad3 (BD Pharmingen), Snail (Santa Cruz Biotechnology), and E-cadherin (Santa Cruz Biotechnology).

RNA extracted from the cultured cells or xenograft tumors was treated with DNase1 (Invitrogen) to remove genomic DNA contamination. Total RNA (2 μg) was reverse-transcribed into cDNA using random primers and M-MLV reverse transcriptase from Invitrogen Life Technology. Quantitative real-time PCR (qRT-PCR) was performed using Power SYBR Green PCR Mix from Life Technologies. All primers used in this study were designed by Primer Blast of NCBI and synthesized by Integrated DNA Technologies. Primer pair specificity was determined by generation of a single peak for dissociation curve (melting curve) at the end of RT-PCR cycling program. PRL27 was used as the internal control.

OVCAR-3 cells were grown to confluence in the complete medium, washed twice with PBS, and then incubated with a serum-free medium containing 10 nmol/L paclitaxel (PTX), 250 nmol/L camptothecin (CPT), 10 μmol/L cispaltin (DDP), 100 nmol/L doxorubicin (Doxo), or no drug for 24 hours. The conditioned medium was collected, filtered using a 0.45-μm syringe filter, and frozen at −80°C until ready for use. Total and active TGFβ1 present in the conditioned medium was quantified using a sandwich ELISA Kit from R&D Systems following the manufacturer's protocol.

IHC staining was performed on paraffin-embedded 4 μm tissue sections and mounted on poly-l-lysine-coated slides. Briefly, after deparaffinization in xylenes and rehydration through graded ethanol solutions, antigen retrieval was performed by submerging the sections into a sodium citrate solution (10 mmol/L, pH 6.0) or a EDTA solution (1 mmol/L, pH 8.0) at 95°C for 15 or 30 minutes, in a microwave oven. The tissue sections were then treated with 3% hydrogen peroxide in methanol to suppress the endogenous peroxidase activity. Tissue sections were then incubated with antibodies to P-Smad2 (Cell Signaling Technology), T-Smad2 (Cell Signaling Technology), slug (Cell Signaling Technology), CD133 (Miltenyi Biotec), and CD49f (LifeSpan Biosciences) at 4°C overnight. After washing, the sections were incubated with pre-diluted secondary antibody (BD Pharmingen), followed by further incubation with 3,3-diaminobenzidine tetrahydrochloride (DAB). Finally, the slides were counterstained with hematoxylin and mounted in an aqueous mounting medium. Appropriate positive and negative controls were stained in parallel. For negative controls, primary antibodies were replaced with PBS.

IHC staining of total Smad2 and P-Smad2 was assessed by two independent observers, who were blinded to the study. Expression of the two markers was determined by an individual labeling score combining the percent stained cells and the staining intensity of positive cells (27). Intensity of stained cells was graded semiquantitatively into four levels as following: 0 (no staining); 1 point (weak staining: light yellow); 2 points (moderate staining: yellow brown), and 3 points (strong staining: brown). The percentage was scored as following: 0 (0–5%), 1 point (6–24%), 2 points (25–49%), 3 points (50–74%), and 4 points (75–100%). The intensity score and the fraction of positive cell scores were multiplied for each marker to derive the immunoreactive score.

OVCAR-3 cells were treated with DDP (1 μmol/L) in the presence or absence of RER (40 nmol/L) for 4 days. The medium was changed every alternate day with the addition of DDP and/or RER. The harvested control and treated cells, after trypsinization, were stained for flow cytometry at a concentration of 100,000 cells per 100 μL of buffer (PBS pH 7.4, 2% PBS, 2 mmol/L EDTA) containing conjugated antibodies (2 mg/100 cells) against CD133 (MiltenyiBiotec) and CD44 (BD Biosciences) at room temperature for 1 hour. The analysis of stained cells was carried out using FACSAria flow cytometer (Becton Dickinson) at the core imaging facility of the UT Health Science Center at San Antonio, Texas.

Cell migration assays were performed in 24-well transwells with 8-μm pore polycarbonate membranes (BD Biosciences). Cells at a density of 20,000 to 40,000 cells/well in serum-free medium with or without treatment were seeded in the upper insert in triplicates. Complete medium with or without treatment was added in the lower chamber. After 18 hours for OVCAR-3 cell and 6 hours for HeLa cell, the cells that did not migrate across the membrane were removed with a cotton swab and the migrated cells were stained with the Hema 3 Stain 18 Kit (Fisher Scientific) according to the manufacturer's protocol. Migrated cells were counted under a microscope with 100× magnification.

Four-week-old female nude mice were used for in vivo animal experiments. The animals were housed under specific pathogen free condition. Exponentially growing OVCAR-3 cells (5 × 10⁶ cells/120 μL/mouse) suspended in 50% Matrigel (Corning Life Sciences) in cold PBS were injected subcutaneously into the back of the mice. After tumor cell inoculation for 1 week, growing tumors were observed and their size was recorded twice a week. The length and width of each tumor were measured using a caliper, and the volumes were calculated by the following formula: volume (mm³) = length × width × width/2. After another 2 to 3 weeks, mice with tumor burden ≥ 100 mm³ in volume were ranked and divided into six groups (five mice for each group) with matched mean tumor volumes and treated as follows: control (normal saline), RER (5 mg/kg), low dose of DDP (2.5 mg/kg), high dose of DDP (5 mg/kg), low dose of DDP (2.5 mg/kg) and RER (5 mg/kg), high dose of DDP (5 mg/kg) and RER (5 mg/kg). RER was administered daily and DDP was given once a week by intraperitoneal injection. After treatment for 29 days, xenograft tumors were isolated from mice. A portion of the tumors tissue was fixed in 4% paraformaldehyde for histological study, and the rest were frozen for other experiments.

Two-tailed Student t test was used to compare the means of two groups. One-way analysis of variance with Tukey–Kramer post hoc test was used for analyzing data when means from more than two groups were compared. Results are expressed as mean ± SEM. P < 0.05 was considered to be statistically significant.

Gene expression profiles of malignant carcinoma samples from ovarian cancer patients were obtained from GEO (GSE7463) (18). Comparison of samples from ovarian cancer patients with chemotherapy treatment to samples without chemotherapy treatment identified a total of 790 upregulated and 929 downregulated probeSets (Supplementary Table S1). These differential expression probes correctly cluster patients based on whether they have undergone chemotherapy or not, except for two patients treated with chemotherapy being clustered into patients without chemotherapy (Fig. 1A). These genes are significantly enriched in Gene Ontology (GO) terms associated with cell cycle regulation (Fig. 1B; Supplementary Table S2), which is the expected effect of chemotherapy. Next we examined the potential upstream regulators of these differentially expressed genes to identify potential master regulators mediating the effects of chemotherapy, using Upstream Regulator Prediction from Qiagen's ingenuity pathway analysis (IPA; Qiagen). As indicated by the results in Fig. 1C and Supplementary Table S3, it is not a surprise that TP53 was the top activated upstream regulator in response to chemotherapy with a significant positive z-score and lowest P value, which is consistent with the observation made by Moreno and colleagues (18). Regulators associated with the estrogen pathway, including beta-estradiol and ESR1, in the Supplementary Table S3, were predicted to be most significantly inhibited with negative z-scores, suggesting a unique and interesting response to chemotherapy in ovarian cancer. Relevant to this study, TGFβ1 was the second most significantly activated regulator upon chemotherapy (Fig. 1C; Supplementary Table S3). Out of the genes corresponding to the 1,719 probeSets, which are significantly altered by chemotherapy, 98.57% of them are regulated by TGFβ1 in various cellular compartments as shown in Supplementary Fig. S1, suggesting TGFβ signaling pathway is an important master regulator in chemotherapy response.

Because of the lack of a similar gene expression profiling study in cervical cancer, we explored the effect of chemotherapy on TGFβ signaling in cervical cancer by examining phosphorylated Smad2 (P-Smad2) and total Smad2 (T-Smad2) levels in 30 matched primary cervical cancer specimens before and after cisplatin-based neoadjuvant chemotherapy using immunohistochemistry. As shown in Fig. 2, P-Smad2 and T-Smad2 protein expression in the tumor tissues were detected mainly in the nucleus with relatively weaker staining in cytoplasm. Prechemotherapy cervical cancer tissues consistently showed weak positive staining of P-Smad2, whereas postchemotherapy tissue consistently showed moderate or intense positive staining. Using a Wilcoxon test, the immunoreactive score, which reflects total staining extent and intensity in both nuclei and cytoplasm, for P-Smad2 expression was significantly increased in postchemotherapy samples compared with prechemotherapy samples (P < 0.05). With regard to T-Smad2, no significant difference was observed between prechemotherapy and postchemotherapy cervical samples. These data suggest that TGFβ signaling is activated after chemotherapy in cervical cancer.

To confirm the role of TGFβ signaling in response to chemotherapy, we initially investigated TGFβ signaling activity in cancer cell lines treated with various chemotherapeutic agents in vitro. Two cervical cancer cell lines, HeLa and C-4I, and two ovarian cancer cell lines, OVCAR-3 and TOV-21G, were used. Four chemotherapeutic agents, cisplatin (DDP), paclitaxel (PTX), doxorubicin (Doxo), and camptothecin (CPT), which are commonly used as therapeutics for gynecological malignancy, were investigated in this study. The four cell lines were treated with 10 μmol/L DDP, 10 nmol/L PTX, 100 nmol/L Doxo, or 250 nmol/L CPT for 1, 6, or 24 hours. Because TGFβ signal is mediated through the phosphorylation of intracellular Smad2 and Smad3 proteins thereby affecting gene expression in the nucleus, we measured the levels of phosphorylated Smad2 and Smad3 to evaluate the activation of TGFβ signaling pathway by the drugs. As showed in Fig. 3, all four drugs stimulated phosphorylation of Smad2 and Smad3 with varying efficacy and time kinetics in cervical cancer cells (Fig. 3A and B) and ovarian cancer cells (Fig. 3C and D). The density of each P-Smad2 or P-Smad3 was normalized to its corresponding T-Smad2 or T-Smad3, respectively, and presented in Supplementary Fig. S2.

Next, we set out to determine whether chemotherapeutics-activated Smad2 and Smad3 phosphorylation can be blocked by TGFβ inhibitors. Two TGFβ inhibitors, HTS466284 (HTS) and RER, were used. HTS466284 (HTS) is an ATP competitive inhibitor of TGFβ type I receptor kinase (25, 26). RER, a novel recombinant trivalent TGFβ trap protein comprised of the endoglin (E) domain of TβRIII flanked by the extracellular domain of TβRII (R), is designed and synthesized in our laboratory (24). RER showed similar or more potent activity than the kinase inhibitor in blocking TGFβ1-induced Smad2 and Smad3 phosphorylation in HeLa and C-4I cells at the concentrations used (Fig. 4A; Supplementary Fig. S3). All four cell lines shown in Fig. 4 were pretreated with or without HTS (100 nmol/L) or RER (40 nmol/L) for 1 hour followed by treatment with 10 μmol/L DDP, 10 nmol/L PTX, 100 nmol/L doxorubicin, or 25 0 nmol/L CPT for 6 or 24 hours. Consistent with the data in Fig. 3, the levels of phosphorylated Smad2 and/or Smad3 protein were in general increased after treatment with the indicated drugs. This action was blocked by both HTS and RER in most cases, suggesting that the drugs appeared to increase extracellular active TGFβ levels, which was neutralized by RER (Fig. 4).

Because TGFβ signaling has been widely reported to increase cancer stem cell (CSC) population, we hypothesized that chemotherapeutics-induced TGFβ signaling might also lead to increased CSCs. CD133, CD44, CD49f, and ABCG were previously reported as CSC markers in ovarian cancer and CD133⁺CD49f⁺ cells sorted from the OVCAR-3 ovarian cancer cell line were shown to have CSC features (28–30). Upon DDP treatment, OVCAR-3 cells displayed higher mRNA levels of stem cell markers including CD44, CD133, and ABCG, as compared with untreated cells, which were dampened by RER or HTS treatment (Fig. 5A). We confirmed the increase of stem cell population with DDP treatment by quantifying CD133⁺CD44⁺ OVCAR-3 cells with flow cytometry. Interestingly, the majority of our OVCAR-3 cells express CD133, but not CD44 (Fig. 5B). Treatment with 1 μmol/L DDP increased CD133⁺CD44⁺ cells by 20-fold in comparison with the untreated control. Addition of RER to the DDP treatment decreased CD133⁺CD44⁺ cells from 15.6% to 3.1%. A live/dead gate was presented in Supplementary Fig. S4. These data indicate that DDP treatment can increase stemness properties of OVCAR-3 cells and this induction is partly dependent on TGFβ signaling.

Epithelial–mesenchymal transition (EMT) is associated with acquisition of tumor stem-like properties (31–33). Furthermore, TGFβ is a potent inducer of EMT and plays an important role in tumor cell motility and migration (34). These raised the possibility that chemotherapeutics may act like TGFβ and have the potential to promote tumor metastasis by stimulating tumor cell motility and invasion via EMT. Therefore, we tested the effect of the drugs on the migration and EMT of the cervical cancer HeLa cell line and ovarian cancer OVCAR-3 cell line. DDP and doxorubicin treatment significantly promoted migration in both cell lines, which was blocked by TGFβ inhibitor RER and HTS (Fig. 5C). CPT significantly promoted migration in HeLa cells, but not in OVCAR-3 cells. On the contrary, PTX-inhibited migration in these two indicated cancer cell lines, likely due to its role in disrupting microtubule dynamics and consequently cell migration. With respect to EMT, we observed a decreased expression of the epithelial cell marker E-cadherin protein after TGFβ1 and DDP treatment, which was reversed by the addition of HTS or RER in HeLa and C-4I cell lines (Fig. 5D). However, the mesenchymal marker Snail was increased after the treatment of TGFβ1 or DDP in these cancer cell lines (Fig. 5D). Like TGFβ1, DDP appears to transcriptionally stimulate E-cadherin and repress Snail as their mRNA levels were increased or decreased, respectively, by DDP in OVCAR-3 cells, which were again reversed by RER or HTS (Supplementary Fig. S5). These results indicate that chemotherapeutics, especially DDP and doxorubicin, can stimulate EMT and tumor cell migration in a TGFβ-dependent manner.

Because RER blocks TGFβ signaling by neutralizing extracellular TGFβs, the blockade of TGFβ-like activities of chemotherapeutics by RER indicated that the drugs likely increased extracellular TGFβ levels. To explore the mechanism of activation of TGFβ pathway by chemotherapeutics, two assays were performed. Using quantitative real-time PCR, we initially investigated the effect of the drugs on TGFβ1 and TGFβ2 mRNA expression in the OVCAR-3 cell line treated with the four drugs for 24 hours. The result showed DDP, doxorubicin, and CPT increased TGFβ1 transcript with no or very moderate effect on TGFβ2 transcript (Fig. 6A). In contrast, PTX treatment showed no effect on either TGFβ1 or TGFβ2 transcript level. To determine whether chemotherapeutics stimulated TGFβ1 production and secretion or induced extracellular TGFβ1 activation, we initially treated confluent cultures of OVCAR-3 cell lines with or without the indicated drugs for 24 hours in a serum-free medium. The medium was then collected for the measurements of total and active forms of TGFβ1 with a sandwich ELISA Kit from R&D Systems. As shown in Fig. 6B, the secreted active and total TGFβ1 were significantly increased by all four chemotherapeutics. Thus, PTX increases TGFβ1 expression in a posttranscriptional manner while the other three drugs appear to increase extracellular TGFβ1 via both transcription and posttranscriptional mechanisms.

Our in vitro studies demonstrated that chemotherapeutics increased extracellular TGFβ1 level and consequently activated TGFβ signaling and CSC enrichment in human cervical cancer, as well as ovarian cancer cells. To confirm these observations in vivo and further investigate whether blockade of TGFβ singling with the novel TGFβ trap, RER, can enhance cisplatin antitumor activity in vivo, a nude mouse xenograft model was established using OVCAR-3 cells. Mice bearing growing OVCAR-3 xenografts were divided into six groups with similar mean tumor volumes of greater than 100 mm³ and were then treated with a low (2.5 mg/kg weekly) or a high dose (5 mg/kg weekly) of DDP, or RER (5 mg/kg daily), as single agent or in combination for 29 days. As showed in Fig. 7A and B, whereas the treatment with RER alone showed no effect on tumor growth, the relative tumor volumes of the other treatment groups were significantly lower than that of the control at the end of the experiment. Treatment with RER and low-dose DDP was more effective than the treatment with low-dose DDP alone in inhibiting tumor growth (Fig. 7A and B) and the relative tumor volumes at the last two measurements were statistically different (Fig. 7A). Similar to the >50% reduction of the terminal tumor volume, we also observed >50% reduction of the terminal tumor weight (Fig. 7B). However, the reduction in tumor weight was not statistically significant at P < 0.05, which was apparently due to their larger coefficient of variation, possibly related to the difficulty in our ability to accurately separate tumors from their surrounding tissues resulting in the large variation in tumor weights. Similarly, RER appeared to also enhance the tumor inhibitory activity of the high-dose DDP resulting in complete regression of two tumors, although the tumor volumes between the two treatment groups were not statistically different (Fig. 7A and B). These data suggested that RER appeared to enhance the antitumor activity of DDP by neutralizing DDP-induced protumor activity of TGFβ. Indeed, low-dose DDP treatment significantly increased the mRNA and active forms of TGFβ1 levels in the xenograft tumors, which were reduced by the combination treatment with RER (Fig. 7C). Consistently, RER treatment also reduced DDP-induced TGFβ signaling in tumors as shown by the reduction of phosphorylated Smad2 levels of tumor specimens from DDP-treated mice with Western immunoblotting analysis (Fig. 7D). There were not enough tumor tissues from the high-dose DDP group for us to perform these assays.

EMT has been associated with increased malignancy, chemotherapy resistance, and poor prognosis in ovarian cancer (35). Recent reports indicated that EMT is involved in the maintenance and formation of stem-like cancer cells in ovarian cancer (31, 36). We next examined whether DDP treatment-induced EMT and CSC markers can be inhibited by RER in the OVCAR-3 xenograft tumors. Immunohistochemical staining showed that DDP increased the expression of mesenchymal marker protein slug and stem cell markers, including CD49f and CD133 (Fig. 7E). qRT-PCR data showed that the transcript levels of snail, CD44, CD133, and ABCG were increased by DDP treatment, and this could be attenuated by RER (Fig. 7F). Thus, DDP increased the expression of EMT and stem cell markers, which was blocked by RER. These results indicate a beneficial combination effect on reducing aggressive cancer cell population in the xenograft tumors.

In this study, our results showed, for the first time, that chemotherapeutics agents stimulate TGFβ1 production and secretion, and consequently activate the TGFβ pathway, EMT, and CSC features resulting in reduced chemo-sensitivity in gynecologic cancer cells and xenograft tumors. In addition, we also provide evidence that TGFβ signaling blocked by a novel trivalent TGFβ receptor trap, RER, is an effective treatment approach and can enhance the efficacy of chemotherapeutics in vivo.

Since its discovery in the early 1980s, TGFβ signaling pathway has been extensively investigated as a key regulator in carcinogenesis (37, 38). Abnormal TGFβ signaling activation has been frequently observed in a variety of human malignancies, including cervical cancer and ovarian cancer, and shown to promote tumor progression and regulate chemo-sensitivity (6, 7, 9). Our results indicated that ovarian cancer displayed an increased RNA transcript of genes associated with TGFβ signaling after chemotherapy. In addition, the expression of P-Smad2 was significantly upregulated in post-chemotherapy cervical cancer tissues compared with pre-chemotherapy samples. These results suggest TGFβ signaling pathway plays an important role in response of chemotherapy. Our result is consistent with the study of Marchini and colleagues (35). They reported that TGFβ signaling pathway was significantly upregulated in the chemo-resistant/relapsing ovarian cancers compared to chemo-sensitive ovarian tumors. Along similar lines, using reverse phase protein array, Carey and colleagues identified TGFβ signaling as an indicator for primary chemotherapy response in patients with advanced serous ovarian cancer (39). Collectively, these results support the notion that aberrant activation of TGFβ pathway is likely a potent mediator of chemo-resistance in ovarian cancer.

TGFβ1 is secreted in a latent form and activated via various mechanisms (3). The activation of TGFβ pathway often leads to tumor progression and drug resistance. Our previous studies showed that doxorubicin, an anthracycline drug widely used for breast cancer, activated TGFβ signaling in human breast cancer MDA-MB-231 and murine mammary cancer 4T1 cell (12). Similarly, Bhola and colleagues reported chemotherapeutic drug paclitaxel increased TGFβ signaling in breast cancer cells (13). Nevertheless, whether these observations are universal in various types of cancer remains to be determined. Thus, four drugs commonly used for gynecological cancers, including paclitaxel, camptothecin, cisplatin, and doxorubicin, were included in this study. We observed all of these four chemotherapeutics activated TGFβ pathway by stimulating phosphorylation of Smad2 and Smad3, which is consistent with the published data in breast cancer, indicating chemotherapy-associated activation of TGFβ pathway is a common event.

More recently, several lines of evidence suggest that TGFβ-mediated drug resistance may be largely due to its induction of cancer stem-like properties in carcinoma cells (13, 14). CSCs, also known as “tumor-initiating cells,” represent a small proportion of cancer cells, with the properties involved in drug resistance, metastasis and relapse of cancers (40). Ovarian cancer has been described as a stem cell disease recently (41). An increasing body of data has demonstrated a subpopulation of CSCs in ovarian cancer, which contributes to chemo-resistance and tumor relapse (29, 30, 41–44). We show here that chemotherapeutics treatment stimulated CSC-like properties as evidenced by increased expression of stem cell markers, such as CD44, CD133, and ABCG, when compared to untreated control ovarian cancer cells both in vitro and in vivo. In addition, chemotherapeutics increased the expression of EMT markers and this transformation has also been associated with acquisition of tumor stem-like properties (31–33). Because metastasis is an important characteristic feature of CSCs, chemotherapeutics-promoted migration provided additional evidence that chemotherapeutics increased CSC population. These results are consistent with the finding by Wiechert and colleagues that cisplatin was able to induce the CSC state as indicated by a GFP reporter driven by a NANOG-promoter in ovarian cancer (29).

Our finding that chemotherapeutics-induced CSC properties was associated with enhanced TGFβ signaling and could be abrogated by TGFβ inhibitors indicates that TGFβ signaling mediates this unwanted side effect of chemotherapeutics. In this study, we report, for the first time, that chemotherapeutics agents stimulated TGFβ1 production and secretion in ovarian cancer cells. These data support our hypothesis that common chemotherapeutics may active TGFβ pathway by stimulating TGFβ1 production and secretion resulting in EMT and CSC expansion to cause drug resistance. Thus, combination therapy with a TGFβ inhibitor and chemotherapy should block drugs/TGFβ-induced EMT and CSC formation leading to enhanced inhibition of tumor growth and metastasis. Although our study has shown that the chemotherapeutics can stimulate TGFβ1 expression at transcriptional and/or posttranscriptional levels, detailed mechanisms by which the different types of chemotherapeutics stimulate active TGFβ1 production remains to be elucidated. Given that TGFβ1 mRNA and/or protein levels were significantly increased with just 24 hours of drug treatment, we believe the drugs directly regulated TGFβ1 transcript and/or protein production instead of selecting cancer stem-like cells with higher levels of TGFβ1 transcript and/or protein, which was a mechanism suggested by Bhola and colleagues (13).

Over the past decade, various components of TGFβ signaling pathway have been explored for the inhibition of this pathway, including both intracellular and extracellular targets. TGFβ receptor kinase inhibitors are the most commonly used TGFβ inhibitor for preclinical and clinical studies. Several published studies have shown efficacy of TβRI kinase inhibitors in attenuating malignant properties of cancer cells in vitro and in vivo (45–48). However, these inhibitors have the possibility of inhibiting other kinases, which may result in undesirable “off-target” side effects. In this study, we report a novel trivalent TGFβ trap, RER, which sequesters extracellular active TGFβ. This trap was shown to potently block TGFβ binding to TβRII and antagonize TGFβ signaling in cultured epithelial cells at picomolar concentrations, and it showed better anti-TGFβ activities than a pan-TGFβ neutralizing antibody and the TβRI kinase inhibitor HTS in prostate cancer cells (24). The data in this study showed that RER effectively blocked TGFβ/chemotherapeutics-induced Smad2/3 phosphorylation and inhibited TGFβ1/chemotherapeutics-stimulated EMT and CSCs expansion. Moreover, RER was in general more effective at blocking TGFβ/chemotherapeutics-induced Smad2/3 phosphorylation when compared with the TβRI kinase inhibitor. In addition, our in vivo data showed that the combination of cisplatin and RER enhanced the efficacy of cisplatin in inhibiting tumor growth in the OVCAR-3 xenograft model in comparison to single cisplatin treatment.

Collectively, the data presented in this study demonstrate that chemotherapeutics can activate TGFβ signaling by stimulating TGFβ1 production and secretion, resulting in EMT and CSC enrichment and decreased chemo-sensitivity in human ovarian cancer and cervical cancer cells. Combination therapy with the novel TGFβ trap RER and cisplatin neutralized cisplatin-stimulated TGFβ1, leading to more efficacious inhibition of ovarian cancer growth. Our studies shed light on an underlying mechanism of chemoresistance and potential utility of TGFβ traps for the treatment of gynecologic malignancies.

A.P. Hink, L.-Z. Sun, and C. Zwieb are listed as co-inventors on an issued patent on TGFβ type II–type III receptor fusions that is owned by the University of Texas Health Science Center at San Antonio. No potential conflicts of interest were disclosed by the other authors.

Conception and design: H. Zhu, X. Gu, C. Zwieb, A.P. Hinck, L.-Z. Sun, X. Zhu

Development of methodology: H. Zhu, X. Gu, C. Zwieb, J. Zhang, L.-Z. Sun

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H. Zhu, H. Bouamar, J. Yang, X. Ding, C. Zwieb, A.P. Hinck, L.-Z. Sun

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H. Zhu, X. Gu, Y. Zhou, J. Yang, X. Ding, C. Zwieb, L.-Z. Sun

Writing, review, and/or revision of the manuscript: H. Zhu, X. Gu, L. Xia, L.-Z. Sun, X. Zhu

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H. Zhu, L. Xia, J. Zhang, L.-Z. Sun

Study supervision: L.-Z. Sun, X. Zhu

This study was in part supported by NIH R01CA172886 awarded to L-Z. Sun and A.P. Hinck, and National Natural Science Foundation of China (81602305) and Medical and Technology Project of Zhejiang Province (grant no. 2016KYA141) to H. Zhu. Additional support was provided by the Flow Cytometry Shared Resource of the Cancer Therapy and Research Center, which is supported by the NIH NCI Cancer Center Support Grant P30 CA054174-17. H. Zhu was supported by a fellowship from the Second Affiliated Hospital of Wenzhou Medical University. L. Xia was in part supported by a fellowship from Xiangya School of Medicine, Central South University, Hunan, China. X. Gu was in part supported by Cancer Research Training Program grants RP140105 and RP170345 from Cancer Prevention and Research Institute of Texas (CPRIT).

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.