Delivery of Therapeutics Targeting the mRNA-Binding Protein HuR Using 3DNA Nanocarriers Suppresses Ovarian Tumor Growth Free

Ovarian cancer is the second most common gynecologic cancer in the United States. It has been perceived as a silent killer due to its asymptomatic nature during the early stages of disease progression. Ovarian cancers are classified into a number of different subtypes depending on cellular origin and histologic characteristics, epithelial ovarian cancer being by far the most common type (1). The current clinical regimen for treating ovarian carcinoma is typically surgical debulking followed by chemotherapy with paclitaxel and a platinum-based therapy (cisplatin or carboplatin). Although 80% of patients receiving this treatment have an initial favorable response, most tumors will eventually relapse accompanied by the development of chemoresistance. A more effective therapeutic strategy is clearly needed to treat this deadly cancer.

Gene dysregulation is a hallmark of cancer. Cancer cells acquire a unique gene expression profile to continue to proliferate and survive the harsh tumor microenvironment. Unlike genetic modification that usually takes years to achieve, cancer cells adopt a more rapid and efficient way to alter gene expression through regulation of transcribed mRNAs (2–4). Posttranscriptional regulation (i.e., modification of mRNA stability and/or translational efficiency) is achieved through a tightly controlled network of interactions between specific target mRNAs with transacting miRNAs and/or RNA-binding proteins (5–7). By regulating these interactions, cancer cells rapidly alter the stability and/or translational efficiency of a large subset of mRNAs that encode proto-oncogenes, cytokines, cell-cycle regulators, and other regulatory proteins that promote cancer cell survival and tumor progression (8–11).

HuR (human antigen R, aka ELAVL1) is a ubiquitously expressed RNA-binding protein that is highly expressed in many ovarian tumors (12, 13). HuR functions in normal, healthy cells as a critical molecule involved in posttranscriptional gene regulation. When cells are stressed (e.g., by low oxygen levels), HuR binds to AREs, AU-rich RNA elements typically located in 3′-untranslated regions (UTR) of mRNAs, and potently influences translation of key survival and growth-related mRNAs in the cytoplasm by several mechanisms including active transport of mRNAs out of the nucleus, mRNA stabilization, and direct facilitation of translation. The role of HuR in the stress response is most likely part of a survival mechanism used by the most aggressive cancer cells in a tumor (14–16). High cytoplasmic HuR expression correlates with high histologic grade of ovarian cancer and poor prognosis in serous ovarian cancer (12, 17).

Identification and targeted inhibition of a single protein involved in tumorigenesis and tumor progression is one strategy to treat ovarian cancer (18). The role of HuR in posttranscriptional regulation of multiple genes that promote survival of ovarian tumor cells provides an opportunity to develop an alternative strategy that targets multiple core signaling pathways at once. Indeed, proof-of-concept studies using HuR knockout mice and intracranial injection of mice with genetically silenced HuR primary glioblastoma cells have shown a reduction in colon tumor and glioblastoma growth, respectively (19, 20). One approach to inhibit HuR is through targeted delivery of functional siRNA. We have recently reported suppression of pancreatic xenograft growth in mice following direct injection of tumors with a nanoparticle-delivered siHuR (21). Targeting systemically administered siRNA-based therapies to tumor cells, however, has been challenging. To address this challenge, in this study, we use a novel DNA dendrimer nanocarrier, 3DNA, derivatized with folic acid (FA) for systemic administration to target siHuR delivery to tumors in mice bearing ovarian tumors that have high cell surface expression of folate receptor-α.

A2780 cells (T. Hamilton, Fox Chase Cancer Center, Philadelphia, PA, obtained in 2004) and OVCAR-5 cells (A. Klein-Szanto, Fox Chase Cancer Center, obtained in 2013) were authenticated and maintained in RPMI1640 medium (Cellgro) supplemented with 10% FBS (Gemini) at 37°C in 5% CO₂. OVCAR3 cells, purchased from ATCC, were grown in media as recommended by ATCC. ID8 cells (K. Roby, University of Kansas Medical Center, Kansas City, KS, obtained in 2006), stably transfected with CAG/firefly luciferase (22), were authenticated and maintained as described previously (23). Frozen aliquots of cells were prepared upon receipt. All cells were used within 6 months of testing negative for mycoplasma using a PCR kit (Sigma). Cells were transiently transfected with siHuR (Ambion #4407268) using Lipofectamine 2000 (Invitrogen). Viable cell number was determined using EZ Count Kit (Lankenau Development, Inc.).

Whole-cell extracts were prepared in RIPA buffer containing proteinase inhibitors. Extract protein concentrations were determined using the BCA Protein Assay Kit (Pierce). Soluble proteins were separated on 10% SDS-PAGE gels and analyzed by Western blotting using HuR mAb clone 3A2 (1:1,000; Santa Cruz Biotechnology), or GAPDH mAb clone 5C6 (1:8,000; Ambion), and a secondary antibody, horseradish peroxidase–labeled goat anti-mouse (1:8,000, Thermo Scientific). Membranes were developed with ECL (Pierce).

Four-micron sections of fixed tumors were deparaffinized, antigen retrieval was performed using citrate buffer, and endogenous peroxidase was quenched. Sections were incubated with primary antibody overnight at 4°C [HuR mAb 19F12 (1:5,000; Clonegene); rabbit anti-Ki-67 (1:150; Zymed)] and then with biotinylated secondary antibody for 30 minutes at room temperature. Signal was amplified and visualized using the TSA-Plus Fluorescence System (Perkin Elmer) or avidin/biotin complex system (VECTASTAIN Elite ABC kit, Vector Laboratories) followed by DAB visualization and hematoxylin counterstaining. Sections were viewed with a Zeiss Axiovert 200M microscope.

To generate shHuR- or shControl (shCtrl)-expressing lentiviruses, Lipofectamine 2000 (Invitrogen) was used to transfect 293T cells for 24 hours with plasmids including a HuR shRNA lentiviral construct (called shHuRc257; “c” for constitutive) or a control shRNA construct (A. Ristimäki, University of Helsinki, Helsinki, Finland; ref. 24), packaging plasmids (pRSV-Rev and pMDLg/pRRE), and a VSVG-coding envelope vector (J. Azizkhan-Clifford, Drexel University, Philadelphia, PA). Transfection mixture was replaced with fresh cell growth medium (DMEM + 10% FBS). Forty-eight hours after transfection, supernatant was collected, sterile-filtered, and used to infect A2780 and OVCAR5 cells in medium containing 8 μg/mL polybrene (Santa Cruz Biotechnology). Following viral infection, shHuRc257- or shCtrl-expressing cells were selected in medium containing hygromycin B (Gemini).

To generate shRNA-expressing plasmids, two DNA oligonucleotides encoding shRNA targeting different human HuR sites (referred to as shHuRi289 and shHuRi699; “i” for inducible; Integrated DNA Technologies) were cloned into Tet-pLKO-puro lentiviral plasmid (J. Azizkhan-Clifford, Drexel University, Philadelphia, PA) as described by Wiederschain and colleagues (25). Final shHuR plasmid constructs were confirmed by DNA sequencing. Targeted sense sequences were: shHuRi289 = 5′-GCAGCAUUGGUGAAGUUGAAUCU-3′; shHuRi699 = 5′-GCCCAUCACAGUGAAGUUUGCA-3′. To generate OVCAR3 doxycycline-inducible cells, OVCAR3 cells were infected with shHuRi289- or shHuRi699-expressing lentivirus in medium containing 8 μg/mL polybrene (Santa Cruz Biotechnology). After viral infection, shHuR- or shCtrl-expressing cells were selected in medium containing 1 μg/mL puromycin dihydrochloride (Gemini).

A total of 7.5 × 10⁵ cells were seeded in 6-well culture plates to ensure 100% confluent growth the next day. The cell monolayer was scratched with a p200 pipet tip, then rinsed with DPBS to remove debris, followed by addition of fresh culture medium containing 5% FBS. Sixteen hours later, photographs of the scratches were taken. The mean width of each scratch was measured.

The soft agar assay was performed as described previously (26). Matrigel invasion was performed using Corning's Cell Invasion Assay protocol as described previously (21).

To generate xenografts, 2 × 10⁶ OVCAR5-shHuR or OVCAR5-shCtrl cells suspended in 100-μL PBS containing 20% cold Matrigel (BD Biosciences) were injected subcutaneously in the flank of 8-week-old female athymic nude mice (Harlan). Once a week, calipers were used to measure tumor length (L) and width (W), and tumor volumes were estimated by the formula V = (L × W²) × 0.52.

To generate tumors in the peritoneum, 6- to 8-week-old female C57BL/6J mice (Jackson Laboratory) were injected intraperitoneally with 2 × 10⁶ ID8-Fluc cells in 200-μL DMEM containing no supplements. To assess tumor growth, mice were optically imaged using an IVIS 100 series Bioluminescence Imaging System (Caliper Life Sciences), as described previously, with a 5-minute integration time for image acquisition (22). Luciferase activity, a measure of tumor load, was quantified as relative light units (RLU).

Two-layer 3DNA nanocarriers (Genisphere) were prepared having 2 unique single stranded sequences (arms) presented on the outer surface as described previously (27, 28). The diameter, determined by DLS, was 70 nm and the zeta potential was −28 ± 2 meV. FA-targeted 3DNA reagents were prepared by first conjugating FA to an oligonucleotide having complementarity to one of the 2 single-stranded sequences on the surface of the 3DNA and hybridizing this FA-modified oligonucleotide to 3DNA nanocarrier to achieve a final ratio of 6 FA molecules per 3DNA molecule. FA-targeted 3DNA or untargeted 3DNA was combined with either a Cy3 (GE Healthcare) labeled oligonucleotide complementary to the second outer surface 3DNA, or a modified siHuR [Integrated DNA Technologies (IDT)] or control siRNA (IDT), complementary to the second outer surface oligo to prepare fluorescent-tagged or silencing reagents, respectively. Both the siHuR and control silencing sequences had a 31 nucleotide extension added to the 5′-end of the sense strand as well as base modifications (2′ o-methyl C, 2′ fluoro G, 2′ fluoro A, and phosphorothiolate bonds) to enhance stability against nuclease digestion and minimize immunogenicity. The final molecular ratio of siRNA to 3DNA was 18:1. The siHuR(mouse) sense sequence is: 5′- GCCUGUUCAGCAGCAUUGGdTdT-3′; and the siHuR(human) sense sequence is: 5′-GCGUUUAUCCGGUUUGACAdTdT-3′.

For efficacy and life span studies, 3DNA formulations were administered in a blinded fashion to mice bearing ID8-Fluc tumors by intraperitoneal injection (bi-weekly × 4; 100 μL; 3 μg ds siHuR/injection). Mice were optically imaged once a week. Four criteria defined life span endpoints: (i) >20% body weight loss from the pretreatment weight; (ii) ascites development; (iii) physical signs of distress; and (iv) unknown death.

For biodistribution analysis, mice were administered FA-3DNA-Cy3 (intravenous injection, 100 μL) and sacrificed 24 hours later. Multiple tissues were collected, fixed, and processed for paraffin sectioning. Sections were examined for fluorescence and representative fields were photographed using a Nikon E800 Eclipse microscope equipped with Image Pro Plus software and Evolution camera.

All procedures performed on mice in this study were done in accordance with the protocols approved by the Lankenau Institutional Animal Care and Use Committee.

Apoptotic cells were identified by terminal deoxynucleotide transferase dUTP nick end labeling (TUNEL) assay using an In Situ Detection Kit (Roche Boehringer Mannheim) as described previously (29).

To identify cancer-related transcripts that were differentially expressed between A2780 cells transfected with siHuR or control siRNA, 200 ng of total RNA for each sample, isolated using a RNeasy Mini Kit (Qiagen), was analyzed with nCounter GX Human Cancer Reference Kit (NanoString Technologies). Negative control probes were used to determine genes reliably detected. Statistically significant differences in gene expression were determined using Student t test (P < 0.05). Principal component analysis (PCA) was performed using the top 4 components, which accounted for >99% of the variance. Genes found to contain at least one ARE site in previous PARalyzer analysis of four HEK293 cell PAR-CLIP libraries were used to identify HuR targets (30). All analyses were performed in R using FactoMineR for the PCA, pHeatmap for hierarchical clustering, and heatmap analysis and ggplot2 for the scatterplots.

Survival distributions were estimated using the Kaplan–Meier method, and treatment groups were compared using the log-rank test. Pairwise comparison adjustments were made using the Dunnett–Hsu method. Statistical significance of differences in tumor growth rates were determined by a multivariate repeated measures ANOVA and differences in tumor mass were determined using Student t test. Analysis was performed using SAS 9.2 or later.

To date, studies in ovarian cancer have shown that cytoplasmic HuR accumulation correlates with poor disease progression, HuR expression is high across different grades of ovarian cancer, and strong nuclear HuR staining is often found in high-grade ovarian carcinoma (13, 17). Few comparisons of HuR expression in normal ovary and ovarian tumors have been made (12). Here, we analyzed and compared HuR expression levels between ovarian cancer and primary cell lines as well as between 5 normal ovaries and 31 serous ovarian tumors. In a Western blot analysis, we found that all ovarian cancer cells tested (A2780, OVCAR3, OVCAR5, and OVCA420) express higher levels of HuR (2–6 fold) compared with ovarian primary cells (HIO80 and HIO120; Fig. 1A). Immunostaining human ovarian tumors for HuR revealed cytoplasmic HuR expression in 29 of 31 tumors. Nuclear expression was high in the same 29 tumors (two tumors had no nuclear and cytoplasmic HuR expression). In contrast, normal ovary specimens had low nuclear HuR expression relative to tumors and no cytoplasmic HuR (Fig. 1B), in agreement with the observations by Denkert and colleagues of HuR expression and localization in three normal ovaries (12). These results suggest that HuR may be a good therapeutic target for the treatment of ovarian cancer.

To study the effect of HuR on ovarian cancer cell growth, we took three approaches to knock down HuR expression in vitro, each targeting a different location in the HuR coding sequence: transient transfection of siRNA (siHuR) or stable expression (inducible and constitutive) of shRNAs (shHuR; see Supplementary Fig. S1 for targeted HuR sequences). First, we directly transfected A2780 ovarian cancer cells with siHuR615. HuR suppression was confirmed by Western blot analysis and immunostaining (Fig. 2A). HuR mRNA was significantly reduced as well in siHuR-transfected cells (Supplementary Fig. S2A). We assessed cell viability at 24, 48, and 72 hours using a Cell Count EZ Kit to measure the oxidative pentose phosphate cycle (OPPC). HuR inhibition resulted in a 32% reduction in viable cells 72 hours after transfection as compared with cells transfected with control siRNA (P < 0.05; Fig. 2A).

To further investigate the function of HuR in cell proliferation, we utilized a lentiviral gene transduction system to generate ovarian cell lines that stably express shRNAs targeting HuR mRNA. First, we generated A2780 and OVCAR5 cell lines that stably and constitutively express either shHuRc or shCtrl. HuR protein expression was effectively silenced in a sustained way in shHuRc-expressing cells, whereas HuR expression remains at the parental cell level in shCtrl-expressing cells in both early and later passages (Fig. 2B and Supplementary Fig. S2B). HuR mRNA levels were also reduced 7-fold in shHuRc-expressing cells (P < 0.0005; Supplementary Fig. S2C). The number of A2780 and OVCAR5 cells was decreased 19% and 39% (compared with parental lines), respectively, in shHuRc-expressing cells when assayed at 72 hours after the cells were plated (P < 0.05; Fig. 2B).

We also generated two OVCAR3 cell lines that stably express two distinct doxycycline-inducible HuR-targeted shRNAs, shHuR_i289 and shHuR_i699. We detected a 25% and 7% decrease in cell number of OVCAR3-shHuR_i289 and OVCAR3-shHuR_i699, respectively, at 120 hours after doxycycline treatment (Fig. 2C). Significant suppression of HuR protein and mRNA was detected in both cell lines in response to doxycycline treatment (Fig. 2C and Supplementary Fig. S3A). mRNA knockdown was 53% and 67% in OVCAR3-shHuRi289 and OVCAR3-shHuRi699, respectively (P < 0.005). Protein knockdown reached a maximum of 36% to 64% in both cell lines at 5 days of doxycycline treatment (1 μg/mL; Fig. 2C). Importantly, doxycycline treatment by itself is not toxic and did not affect OVCAR3 cell growth (Supplementary Fig. S3B and S3C). Taken together, these results demonstrate that inhibition of endogenous HuR expression compromises the normal proliferation of ovarian cancer cells.

Being able to survive and grow in the absence of adherence to the extracellular matrix and neighboring cells is a characteristic of cancer cells. To examine whether HuR is required for anchorage-independent growth of ovarian cancer cells, we assayed OVCAR5-shHuRc257 and control cells for soft agar colony formation over a 7-week period (Fig. 3A). HuR suppression resulted in a 58% and 42% decrease in colony number compared to parental (P < 0.005) and shCtrl-expressing cells (P < 0.05), respectively.

Cancer cells regulate the expression of a discreet set of genes to migrate and invade distant sites. HuR has been shown to affect cell migration through posttranscriptional regulation of invasion- and metastasis-related genes such as Snail and MMP-9. High-grade serous carcinomas (type II) metastasize rapidly and attach to the abdominal peritoneum or omentum (31). Here, we investigated whether altering HuR expression affects ovarian cancer cell migration and invasion. First, we performed in vitro scratch assays with OVCAR5-shHuRc257 cells. We observed a significant decrease in the rate at which OVCAR5-shHuRc257 cells migrated to “repair” the scratched surface as compared with OVCAR5 and OVCAR5-shCtrl cells. After 16 hours, closure of the scratch was complete in parental cells, while closure in siCtrl cells was nearly complete at 91% and closure in shHuR-expressing cells was only 38% (P = 0.0001; Fig. 3B). Next, we performed Matrigel invasion assays in which cells were seeded in serum-free medium on transwell inserts coated with Matrigel and incubated for 24 hours with serum-rich medium in the bottom chambers serving as a chemoattractant to promote invasion. We observed a significant decrease in the number of OVCAR5-shHuRc257 cells that had invaded the Matrigel as compared with the number of invading parental OVCAR5 and OVCAR5-shCtrl cells (58% and 48% decrease, respectively, P < 0.05; Fig. 3C). Taken together, these results demonstrate that inhibition of endogenous HuR expression compromises the migration and invasion capabilities of ovarian cancer cells.

To test the effect of HuR inhibition on the growth of human ovarian tumor cells in vivo, we injected the hind flank of female nude mice subcutaneously with equal numbers of either OVCAR5-shHuRc257 or OVCAR5-shCtrl cells. Three weeks postinjection, tumors were resected and photographed, and their volumes measured (Fig. 4). The mean tumor volumes generated from OVCAR5-shHuRc257 cells were significantly smaller than those from OVCAR5-shCtrl cells (118 ± 21 mm³ vs. 357 ± 63 mm³; P = 0.0004; Fig. 4). Western blot analysis of total cell lysates prepared from tumors showed a mean reduction of 73% in HuR protein in OVCAR5-shHuRc tumors as compared with OVCAR5-shCtrl tumors (Fig. 4). We observed very few apoptotic cells in both OVCAR5-shHuRc and OVCAR5-shCtrl upon TUNEL staining of tumor sections (Supplementary Fig. S4A). We also evaluated whether HuR inhibition affected tumor cell proliferation. Human Ki67 staining revealed large areas of Ki67⁻ cells in OVCAR5-shHuRc tumors that were absent in OVCAR5-shCtrl tumors (Supplementary Fig. S4A), in agreement with our observation that HuR inhibition resulted in a decrease in tumor cell proliferation in vitro (Fig. 2). To complement the Ki67 staining of xenograft sections, we stained parental, shHuRc, and shCtrl cells with Ki67 in culture. The percent Ki67⁺ parental cells was 1.4 times higher than in shHuRc cells (P = 0.003) and 1.2 times higher than in shCtrl cells (Supplementary Fig. S4B). No apoptotic cells were observed upon TUNEL staining of all cell populations. Taken together, we conclude from these results that tumors in which HuR is suppressed are smaller than control tumors largely due to a reduction in cell proliferation rather than tumor cell death.

To test the therapeutic efficacy of siHuR in a mouse ovarian tumor model, we conjugated FA to 3DNA nanocarrier for in vivo–targeted siHuR delivery to tumor cells following systemic delivery (Supplementary Fig. S5A). The folate receptor has been shown to be overexpressed on the surface of many tumor cells, including ovarian cancer (32). We confirmed that ovarian cancer cells express higher amounts of folate receptor than do ovarian primary cells (Supplementary Fig. S5B). To improve siHuR stability in serum, we made several modifications to nucleotide bases (Supplementary Fig. S5C). Importantly, siHuR conjugated to 3DNA retains its ability to suppress HuR expression; inhibition of HuR expression following in vitro transfection of A2780 cells with FA-3DNA-siHuR is as effective as Lipofectamine-delivered siHuR (Supplementary Fig. S5D).

To determine how effectively FA targets delivery of 3DNA to tumor cells, we systemically administered Cy3-labeled 3DNA with or without FA-derivatization to C57BL/6J mice bearing tumors throughout the peritoneal cavity derived from murine ID8-Fluc cells. Twenty-four hours after administration, mice were sacrificed and multiple tissues and tumor nodules were collected for fluorescent microscopic analysis (Supplementary Fig. S6). We observed significant Cy3 fluorescence in ovarian tumors, lesser amount in normal ovaries of tumor-bearing mice, and very low amounts in brain, liver, spleen, and lung. No fluorescence was observed in heart and in the ovary of a nontumor bearing mouse. This result supports the use of FA-derivatized 3DNA to target siHuR to ovarian tumor cells.

To test the therapeutic efficacy of FA-derivatized 3DNA delivery of siHuR, we used the ID8-Fluc ovarian cancer mouse model. We intraperitoneally injected C57BL/6J mice with ID8-Fluc cells. Tumor load in these mice was assessed using optical imaging to detect and quantify luciferase bioluminescence as relative light units (RLU). Baseline optical images were obtained 4 to 6 weeks after mice were injected with cells. The total RLU per mouse for baseline images ranged from 5 × 10⁵ to 1 × 10⁶ RLU. Mice were distributed into 5 groups having equivalent tumor loads (n = 6 per group). The treatment schedule for these mice was 2 intraperitoneal injections of either FA-3DNA-siHuR, 3DNA-siHuR, FA-3DNA-siCtrl, or 0.9% saline per week for 4 weeks (Fig. 5A). Mice were optically imaged once a week. Representative optical images of mice in each group at baseline and at week 4 are shown in Fig. 5A. The mean tumor loads from each of the five treatment groups were determined at week 1, 2, 3, and 4. Suppression of tumor growth, although not quite significant, was observed only in mice treated with FA-3DNA-siHuR compared with those of the other groups (Fig. 5A). Ascites fluid accumulation in the abdomen attenuates the bioluminescence emitted by tumor cells. Thus, the observed bioluminescence may have resulted in low assessments of tumor load in the other groups and underestimated the effectiveness of the FA-targeted siHuR treatment on the suppression of solid tumor growth. Strikingly, only one of 6 mice treated with FA-3DNA-siHuR developed ascites compared with the rest of groups in which 67% to 100% of mice developed ascites (Fig. 5A). Histopathologic analysis of multiple tissues (liver, kidney, spleen, lung, brain, ovaries, and heart) from three mice treated with FA-3DNA-siHuR revealed minimal chronic inflammation (data not shown).

In a separate study, we examined whether FA-3DNA-siHuR treatment of mice bearing ID8-Fluc tumors prolongs life. Mice were injected twice weekly with the same 4 treatments as in Fig. 5A for 4 weeks. The median survival of FA-3DNA-siHuR–treated mice is approximately 1.5 times longer than either the FA-3DNA-siCtrl, or 3DNA only treated mice (43 vs. 29 days; Fig. 5B). The log-rank test of significance across the groups has a P value of <0.001, indicating a significant difference across the four treatment groups in survival. Pairwise comparisons adjusted using the Dunnett–Hsu method indicate that treatment with FA-3DNA-siCtrl is significantly different from FA-3DNA-siHuR and 3DNA (P = 0.0007 and P = 0.0093), respectively, but not significantly different from 3DNA-siHuR (P = 0.305). Nevertheless, it is noteworthy that treatment with the FA-derivatized 3DNA formulation improved both efficacy and life span outcomes as compared with the siHuR formulation lacking the FA conjugate.

To identify genes whose expression is regulated by HuR in ovarian cancer cells, we used a NanoString nCounter GX Human Cancer Reference Kit to interrogate 230 genes (plus 6 housekeeping genes) in total RNA from A2780 cells 72 hours after they had been transfected with siHuR615 or with siCtrl (see ref. 21 for complete list of the interrogated genes). Four biologic replicates of each sample were analyzed using multivariate statistics. Both PCA and hierarchical clustering analysis showed that the transcript profile changed significantly after HuR suppression (Supplementary Fig. S7A and S7B). MO plots identified transcripts having significantly different expression after HuR knockdown (Supplementary Fig. S7C). 116 genes (50%) had significantly altered expression; 50 had upregulated expression and 66 had downregulated expression (Table 1). We used multiple existing PAR-CLIP datasets from HEK293 (30) and ribonucleoprotein immunoprecipitates from MiaPaca-2 cells (21) to identify those transcripts to which HuR binds directly, and indirect targets of HuR (i.e., downstream of direct targets). Thirty-one of 50 (62%) upregulated genes, and 37 of 66 (56%) downregulated genes are direct targets of HuR (Supplementary Fig. S7D; Table 1). Using the Database for Annotation, Visualization, and Integrated Discovery (DAVID) to perform a functional annotation enrichment analysis, we determined that HuR-regulated transcripts with altered expression were associated with multiple essential processes including cell proliferation, angiogenesis, cell-cycle regulation, apoptosis, and DNA repair (Supplementary Table S1).

This is the first study to demonstrate the therapeutic potential of targeting HuR for the treatment of ovarian cancer. Our previous studies in preclinical pancreatic cancer models (33–38), as well as studies by others (17, 19, 20, 24, 39–41), provided the rationale for investigating HuR inhibition in ovarian tumors We first established that suppression of HuR expression reduces proliferation, anchorage-independent growth, and invasion of ovarian cancer cells in vitro. We further demonstrate HuR inhibition–mediated reduction in tumor growth rate, delay in ascites development, and extension of lifespan by nearly 1.5-fold in two different ovarian cancer mouse models, a xenograft model and an orthotopic model in which mice bear tumors throughout the peritoneum.

We used a novel DNA dendrimer nanocarrier, 3DNA, to target delivery of siHuR directly to the peritoneum of mice bearing ID8-Fluc tumors. Derivatization of 3DNA with therapeutic payloads (e.g., siRNA, miRNA, DNA, small molecules), targeting moieties (e.g., antibodies, ligands) and tracking tags (e.g., fluorescent markers, radionuclides) allows for versatility in the design of 3DNAs having potential for a wide range of medical applications. The results of this study demonstrate that systemically administered siHuR delivered by 3DNA retains the ability to inhibit HuR expression and achieve desirable therapeutic endpoints.

While HuR acts as a stress-response protein in tumor cells by regulating the expression of multiple genes known to function in tumor cell survival, HuR is also expressed at low levels in normal cells and plays a vital role in these cells as well. HuR functions in normal cells to stabilize mRNA transcripts and regulate their translation. It also plays a role in polyadenylation and alternative spicing for selected pre-mRNA transcripts (42). In these capacities, HuR is an essential protein, as evidenced by the inability of HuR knockout mice to survive (43) and the lethal consequences of globally induced HuR silencing in adult mice (44). Making use of the versatility of the 3DNA nanocarrier, we conjugated the 3DNA to FA aiming to suppress HuR in tumor cells while avoiding toxicity in healthy cells. Folate receptor-α (FR) expression is frequently amplified on tumor cells, including ovarian tumor cells, allowing for preferential tumor uptake via receptor-mediated endocytosis (45). Our biodistribution study demonstrated efficient tumor targeting, as minimal fluorescence was observed in normal tissues following systemically administered Cy3-labeled FA-3DNA. In addition, mice showed no outward signs of distress and we did not observe any formulation-dependent toxicity in noncancerous tissues after multiple administrations of FA-3DNA-siHuR over a period of several weeks. However, while 3DNA-siHuR formulations demonstrated significant tumor growth inhibition compared with controls in efficacy studies, the difference of tumor growth suppression and lifespan extension between mice treated with FA-targeted and those treated with nontargeted siHuR formulations was not significant. It is noteworthy, however, that FA conjugation always yielded better outcomes than nontargeted formulations, suggesting that FA improved tumor-targeted delivery over and above that which would be achieved simply by the enhanced permeability and retention effect. Intratumoral heterogeneity of FR expression may be a contributing factor that attenuates FA-targeting effectiveness. It is important to keep in mind that not all ovarian tumors in patients are FR+. Screening tumors for FR expression can be used to identify those patients that would benefit from FA-targeted therapy (46).

Treatment of ovarian tumor-bearing mice with siHuR significantly reduced the growth rate of tumors, but complete remission of tumors was not attained. While targeting HuR is an attractive therapeutic strategy because it leads to disruption of multiple genes in core signaling pathways, the overall effectiveness of a monotherapy based on HuR suppression depends on the amount of HuR present in tumor cells. Disparity in HuR expression and localization is likely to be heavily influenced by the tumor microenvironment, which can show significant inter- and intratumoral heterogeneity in terms of blood flow, oxygenation, and nutrient supply (47). We have shown that HuR is engaged under conditions of hypoglycemia and hypoxia (33). As such, HuR expression and cytoplasmic localization are expected to be high in regions within ovarian tumors with the least amount of vascularization (i.e., greatest hypoxia and hypoglycemia). Silencing HuR would have little effect on cancer cells that have low cytoplasmic HuR expression. In addition, serous ovarian tumors have a very high rate of genetic instability that leads to extensive heterogeneity within tumors (48). Mutation of non-HuR–regulated genes may promote tumor progression. It is also possible that the siHuR payload was not delivered to all tumor cells.

We are currently taking a realistic approach and investigating combination therapies in which tumor-targeted siHuR is combined with either another siRNA or with chemotherapeutic drugs. In particular, we are combining siHuR with a siRNA that silences CLDN3, a non-HuR–regulated gene, thus targeting a pathway outside the HuR binding'ome. We have previously shown that siRNA inhibition of the tight junction protein claudin 3 suppresses ovarian tumor growth in mice (22). We are also combining siHuR with carboplatin and paclitaxel, the standard-of-care drugs for first-line treatment for ovarian cancer. We have recently shown that silencing HuR sensitizes cells grown in culture against DNA damaging agents, including carboplatin and paclitaxel, due, in part, to HuR's acute upregulation of the mitotic inhibitor kinase Wee1 (36). Thus, in addition to enhancing efficacy directly as a combination therapy, HuR inhibition may also combat drug resistance. The versatility of the 3DNA nanocarrier may make it possible to target delivery of combination therapies to tumors using a single 3DNA formulation.

A similar siRNA/chemotherapy combination therapy in which a siRNA targeting the transmembrane protein EphA2 was combined with paclitaxel showed promising results in preclinical tests in ovarian tumor-bearing mice and has recently been approved by the FDA for use in a phase I clinical trial (18, 49). As in our study using siHuR as a monotherapy, despite significant reduction in tumor growth, complete tumor remission was not obtained with the siEphA2/paclitaxel combination therapy. Interestingly, EphA2 is among the many genes that are posttranscriptionally regulated by HuR (50). Given the numerous genes and miRNAs that have significantly altered expression in ovarian tumors and are either presumed or have been shown to be regulated by HuR, some of which are known to sensitize cells to chemotherapeutics (e.g., miR-200c, class III β-tubulin; see ref. 41), tumor-targeted HuR inhibition in combination with clinically relevant chemotherapeutics may yield better therapeutic outcomes than those we have observed with siHuR monotherapy.

K. Rhodes is a research scientist in Genisphere LLC. G.E. Gonye is a senior application scientist in NanoString Technologies Inc. R.C. Getts has ownership interest (including patents) in Genisphere LLC. J.A. Sawicki is a consultant/advisory board member for Genisphere. No potential conflicts of interest were disclosed by the other authors.

Conception and design: Y.-H. Huang, W. Peng, J.R. Brody, R.C. Getts, J.A. Sawicki

Development of methodology: Y.-H. Huang, W. Peng, R.C. Getts, J.A. Sawicki

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y.-H. Huang, W. Peng, N. Furuuchi, J.V. Gerhart, K. Rhodes, M. Jimbo

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y.-H. Huang, W. Peng, G.E. Gonye, J.R. Brody, R.C. Getts, J.A. Sawicki

Writing, review, and/or revision of the manuscript: Y.-H. Huang, W. Peng, N. Mukherjee, G.E. Gonye, J.R. Brody, R.C. Getts, J.A. Sawicki

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y.-H. Huang, W. Peng, R.C. Getts, J.A. Sawicki

Study supervision: R.C. Getts, J.A. Sawicki

The authors thank Charles Dunton and Gary Daum for procurement of archival ovarian tumor specimens from patients and Benjamin Leiby for assistance with statistical analysis.

This research was supported in part by a grant from the Marsha Rivkin Center for Ovarian Cancer Research (J.A. Sawicki), the Sharpe-Strumia Research Foundation of the Bryn Mawr Hospital (J.A. Sawicki), and the Sarah Parvin Foundation (J.R. Brody).

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.