|Home | About | Journals | Submit | Contact Us | Français|
The comprehensive characterization of a large number of cancer genomes will eventually lead to a compendium of genetic alterations in specific cancers. Unfortunately, the number and complexity of identified alterations complicate endeavors to identify biologically relevant mutations critical for tumor maintenance, because many of these targets are not amenable to manipulation by small molecules or antibodies. RNAi provides a direct way to study putative cancer targets; however, specific delivery of therapeutics to the tumor parenchyma remains an intractable problem. We describe a platform for the discovery and initial validation of cancer targets, composed of a systematic effort to identify amplified and essential genes in human cancer cell lines and tumors partnered with a novel modular delivery technology. We developed a tumor-penetrating nanocomplex (TPN) comprised of siRNA complexed with a tandem tumor-penetrating and membrane-translocating peptide, which enabled the specific delivery of siRNA deep into the tumor parenchyma. We employed TPN in vivo to evaluate inhibitor of DNA binding 4 (ID4) as a novel oncogene. Treatment of ovarian tumor-bearing mice with ID4-specific TPN suppressed growth of established tumors and significantly improved survival. These observations not only credential ID4 as an oncogene in 32% of high-grade ovarian cancers, but also provide a framework for the identification, validation, and understanding of potential therapeutic cancer targets.
Genome-scale studies of cancer samples have begun to provide a global depiction of genetic alterations in human cancers, but the complexity and volume of data that emerge from these efforts has made dissecting the underlying biology of cancer difficult, and little is known about the functions of most of the candidates that emerge. For example, in studies of 489 primary high-grade serous ovarian cancer genomes, 1825 genes were identified as targeted by recurrent amplification events (1). Systematic approaches to study the function of genes in cancer cell lines, such as genome-scale, pooled short hairpin RNA (shRNA) screens, offer a means to assess the consequences of the genetic alterations found in s uch genome characterization efforts. We recently used a shRNA-based approach to find genes that are both overexpressed in human primary tumors and also essential for the proliferation of ovarian cancer cells (2). This approach identified 54 overexpressed and essential genes in ovarian cancer and 16 genes in non-small cell lung cancer (NSCLC) that required further validation in vivo. Furthermore, many of these candidates represent targets that are not amenable to antibody-based therapeutics or traditional small molecule approaches. Thus, if one envisions a discovery pipeline that begins with cancer genomes and ends with novel therapeutics, there is clearly a bottleneck at the point of in vivo validation of novel targets.
Achieving silencing in the epithelial cells in the tumor parenchyma is especially critical to study the genetic alterations of interest. RNA interference (RNAi) is a potentially attractive means to silence expression of candidate genes in vivo, particularly for “undruggable” gene products (3-4). However, systemic delivery of small interfering RNA (siRNA) to tumors has been challenging owing to rapid clearance, susceptibility to serum nucleases, and endosomal entrapment of small RNAs, in addition to their inherent inadequate tumor penetration (5). Tumor penetration is also a problem for the delivery of siRNA and shRNA, among other cargo, and is characterized by limited transport into the extravascular tumor tissue beyond the perivascular region (6). This low penetration is thought to arise from the combination of dysfunctional blood vessels that are poorly perfused, and a high interstitial pressure, especially in solid tumors, in part due to dysfunctional lymphatics (7). The leakiness of tumor vessels partially counteracts the poor penetration (the so-called enhanced permeability and retention, or EPR, effect), but the size-dependence and variability of this property can limit its usefulness (7). Desmoplastic stromal barriers can further impede transport of therapeutics through tumors (8-9).
A new class of tumor-penetrating peptides has been described recently, which home to several types of tumors and leverage a consensus R/KXXR/K C-terminal peptide motif [the C-end rule (CendR)] to stimulate transvascular transport and rapidly deliver therapeutic cargo deep into the tumor parenchyma (10-11). These peptides are tumor-specific, unlike canonical cell-penetrating peptides that do not display cell or tissue-type specificity (12), and are able to improve the delivery of small molecules, antibodies, and nanoparticles (13). Despite their promise, this class of peptides has not been successfully co-opted for siRNA delivery, in part owing to the additional challenges of delivering oligonucleotides across cell membranes, out of endosomes, and into the cytosol to achieve gene silencing.
Here, we designed a siRNA delivery vehicle that was tumor-penetrating and modular, so it could be easily assembled in a single step to accommodate different payloads to various genes of interest. We envision that such a technology would enable a platform wherein novel targets can be identified using structural and functional genomics and subsequently rapidly “credentialed” both in vitro and in vivo. Follow-up studies could then identify the mechanism of action underlying the observations and establish (and ultimately prioritize) novel oncogenes as therapeutic targets. To achieve this goal, we have combined a systematic effort to identify genes that are both essential and genetically altered in human cancer cell lines and tumors with the development and deployment of a novel tumor-specific and tissue-penetrating siRNA delivery platform.
To facilitate the identification of genes that are essential in specific cancer types, we initiated Project Achilles, a large-scale effort involving genome-scale, pooled shRNA screens in human cancer cell lines (2). Recent efforts to characterize the genomes of primary high-grade serous ovarian cancer have revealed 63 recurrent regions of copy number gain and 50 regions of copy number loss, each containing several genes (1). To identify genes that are both recurrently amplified and essential in ovarian cancers that harbor increased copy number of these genes, we quantified the distribution of shRNA proliferation scores among all shRNAs for each amplified gene (Fig. 1A). We identified 206 cases in which shRNAs targeting the amplified gene were significantly depleted (P<0.05, z-score), including known ovarian cancer oncogenes KRAS, AKT1, BCL2L1, and ERBB3, and newer candidates for ovarian cancer, including the inhibitor of DNA binding 4 (ID4) transcriptional regulatory protein and the S-phase kinase-associated protein 2 (SKP2) E3-ubiquitin ligase (table S1).
We selected ID4, a helix-loop-helix (HLH) transcriptional regulator, for further study because recent genomic analyses have indicated that the chromosomal region containing ID4 (6p22) is amplified in 32% of high-grade serous ovarian cancers (1) (Fig. 1B). Also, ID4 is overexpressed in the majority of primary ovarian cancers, but not in normal ovary, fallopian tube, and other tissues (Fig. 1C; fig. S1A). In addition, by examining the transcript levels of ID4 in a large panel of cancer cell lines, we found that ID4 was frequently overexpressed in the majority of ovarian cancer cell lines and cell lines derived from other cancer lineages, such as endometrial cancer, breast cancer, and glioblastoma (fig. S1B).
After identifying ID4 as a candidate oncogene in human samples, our next step was to establish preclinical models to credential the oncogenic potential of ID4, specifically by investigating the relationship between amplification, expression, and essentiality of ID4 in a panel of human cancer cell lines. First, we found that multiple shRNAs that did not alter the expression of ID1, ID2, and ID3, but suppressed ID4 (fig. S1C), significantly inhibited the proliferation of 9 out of 11 ovarian cancer cell lines and two glioblastoma cell lines tested (range 53-92%) (Fig. 1D). Ovarian cancer cell lines OVCAR-8, OVCAR-4, and CaOV-3, which harbor increased copy number (Fig. 1E) and exhibit overexpression of ID4 (Fig. 1D), died by apoptosis after ID4 suppression (fig. S1D). By contrast, 7 cell lines that express comparatively lower ID4 levels showed little inhibition of proliferation after ID4 suppression (range 0-44%) (Fig. 1D).
We then tested whether ID4 was an oncogene by investigating its ability to induce cell transformation. Expression of oncogenic HRASV12 or co-expression of known RAS effectorsrenders immortalized human epithelial cells tumorigenic in vivo (14). Therefore, to investigate the role of ID4 in human ovarian epithelial cells, we similarly created an ovarian surface epithelial cell line expressing the SV40 Large T and small t antigens, hTERT, and MEKDD (IOSE-M cells) (fig. S1E), and used this cell line to identify genes that promote tumorigenicity. Expression of ID4 at levels found in ovarian cancer cell lines did not affect the cell cycle progression (fig. S1F) but dramatically increased the rate of tumor formation in vivo compared to cells expressing a control vector (Fig. 1F). Addition of ID4 that harbored mutations in the HLH domain, or the addition of ID1, ID2, or ID3, each failed to confer significant tumorigenicity (Fig. 1F).
Recent work suggests that the majority of ovarian cancers arise from the fallopian tube (FT) epithelium rather than the ovarian surface epithelium (15). As such, we used immortalized FT epithelial cells (16) and assessed whether ID4 expression also induced cell transformation in these cells. As we found for ovarian epithelial cells, ID4 induced anchorage-independent growth in FT cells (Fig. 1G). These observations show that ID4 is amplified, essential, and transforming in ovarian epithelial cells.
To explore the mechanism of action by which ID4 induces transformation, we performed gene expression profiling in immortalized ovarian surface epithelial (IOSE-M) cells that overexpress ID4 or a control vector. Among the top 5 gene sets, we found highly significant enrichment of 3 gene sets upregulated by expression of the oncogenic Nucleoporin 98-Homeobox A9 (NUP98-HOXA9) fusion protein (17) (Fig. 2A; fig. S2). These 3 gene sets represent short-term (6 h) and long-term (3 and 8 days) upregulation after induction of NUP98-HOXA9. ID4 expression in IOSE-M cells consistently increased expression of multiple HOXA genes, including HOXA9, HOXA7, and HOXA3 (Fig. 2B); whereas suppression of ID4 led to significantly reduced levels of these same genes (Fig. 2C; fig. S3A). To determine whether HOXA9 was required for ID4-induced cell transformation, we introduced three previously validated HOXA9-specific shRNAs (18) or control shRNAs targeting green fluorescent protein (shGFP) into ID4-overexpressing IOSE-M cells (fig. S3B). Suppression of HOXA9 only slightly (~30%) reduced cell proliferation (fig. S3C), but significantly inhibited ID4-induced anchorage-independent growth (Fig. 2D) and reduced the rate of tumor formation in vivo compared to control shGFP (Fig. 2E).
Finally, to confirm these observations in human ovarian cancers, we analysed the expression profiles from primary ovarian tumors by TCGA project (1) and found that many genes within the NUP98-HOXA9 expression signature (TAKEDA_TARGETS OF NUP98-HOXA9 FUSION_10D_DN) were significantly downregulated in tumors that express low levels of ID4 (Fig. 2F). The gene set was originally defined 10 days after expression of NUP98-HOXA9. Human tumors that harbored ID4 amplifications showed significantly increased expression of a gene set that was downregulated by p21WAF1/CIP1 or TP53 tumor suppressor (P21_P53_ANY_DN) (19), indicating a suppressed p21WAF1/CIP1 transcriptional activity in ID4-amplified tumors (Fig. 2G). These observations demonstrate that expression or amplification of ID4 regulates the HOXA9 and CDKN1A transcriptional programs to disrupt the normal regulation of proliferation and differentiation.
To assess whether ID4 functions as a therapeutic target for ovarian cancer, we designed a siRNA delivery platform that permits access to specific cancer cells. To this end, we synthesized a library of tandem peptides capable of tumor-homing and penetration (with a fixed cyclic domain) and siRNA delivery across cell membranes (variable linear domains for electrostatic binding) (Fig. 3A; table S2). We selected the cyclic nonapeptide LyP-1 (CGNKRTRGC) as the tumor-penetrating domain for all tandem peptides because it binds p32 (p33/gC1q receptor/HABP1), a mitochondrial protein whose expression is elevated on the surface of stressed tumor and tumor-associated cells in a wide range of tumor types (20). Once bound, LyP-1 is proteolytically processed by endogenous proteases to reveal a cryptic CendR motif that activates tissue penetration of associated payloads and results in higher therapeutic accumulation (11, 21).
We postulated that a tandem peptide consisting of an N-terminal cell-penetrating peptide (CPP) with a C-terminal LyP-1 separated by a 4-glycine linker domain would retain membrane translocation and tissue-penetrating properties, condense siRNA into nanoparticles, and selectively target tumor cells expressing p32 at the surface. We created a library of 18 linear membrane-translocation domains that contained both polycationic peptide sequences [oligoarginines of varying net charge, TAT (48-60), and the HSV-1 tegument protein VP22] and amphipathic sequences [Penetratin (PEN) and Transportan (TP)]. Inspired by nonenveloped RNA viruses (22), a myristoyl group was added to the amino terminus to facilitate interactions with membrane lipids (Fig. 3B; fig. S4A). We first confirmed that peptides from the library bound siRNA electrostatically to form stable nanocomplexes in water (diameter ~80 nm), phosphate-buffered saline (PBS, diameter ~200-400 nm), and serum (diameter ~200-400 nm) (fig. S4, B and C). The majority of peptides were noncytotoxic, with the exception of highly cationic peptides 12R and 15R (fig. S4D).
To evaluate nanocomplex-mediated gene silencing, we treated HeLa cells expressing destabilized GFP with siGFP bound to either tandem peptides or lipofectamine. The top three carriers—6R, 9R, and TP—silenced GFP expression 40-60% (Fig. 3C). Flow cytometry analysis indicated that gene suppression was dose-dependent and heterogeneous across the cell population, with 50-70% of cells exhibiting near-complete gene suppression (fig. S4E), with variation likely owing to heterogeneity in p32 expression. GFP suppression can underreport the magnitude of silencing endogenous gene targets, such as amplified oncogenes (23); thus, we used this assay to compare peptides, but not as a quantitative measure of efficiency of RNAi of oncogenes. We chose nanocomplexes consisting of siRNA bound to TP-LyP-1 for further analysis (hereby referred to as the tumor-penetrating nanocomplex, or TPN), as it had the highest efficacy in delivering siRNA to cells, was potent relative to other tandem peptides (>25% gene suppression at 25 nM siRNA), and has a known secondary structure (24), in contrast to poly-Arg peptides.
Next, we examined the specificity of receptor targeting and gene suppression. Uptake of LyP-1–targeted nanocomplexes was significantly higher than control nanocomplexes bearing ARAL (ARALPSQRSR), a peptide that has the same net charge as LyP-1 (fig. S4F). LyP-1 targeting was specific, as shown by competition assays with free LyP-1 (fig. S4G) and in cells lacking the p32 receptor (fig. S4H). We did not detect statistically significant alterations in GFP expression in HeLa cells exposed to untargeted control nanocomplexes (UCN) carrying siGFP (TP-ARAL/siGFP) or targeted TP-LyP-1 carrying a scrambled siRNA sequence (TP-LyP-1/siUT) (Fig. 3D). In contrast, when we used TP-LyP-1 carrying siGFP, we observed over 40% suppression of GFP expression.
OVCAR-4 and OVCAR-8 human ovarian cancer cell lines have elevated surface p32 expression (fig. S5); therefore, these cell lines were compatible for testing TPN-mediated delivery of ID4-specific siRNA (siID4). When ID4 was suppressed in these cell lines in vitro using TPN (Fig. 4A), we observed a decrease in cell viability (Fig. 4B) and an increased rate of apoptosis (fig. S5; Fig. 4C). We therefore selected the OVCAR-4 and OVCAR-8 ovarian cancer cell lines for our preclinical evaluation of ID4 in vivo.
For in vivo applications, an ideal carrier should extend the half-life of siRNA in serum by protecting it from nuclease degradation, yet readily dissociate from siRNA once in the cytosol of target cells. We used a gel-shift assay to determine an optimal siRNA-to-peptide complexation molar ratio of 1:20, at which free siRNAs were encapsulated into TPN, likely owing to charge-based interactions between the peptide and the siRNA backbone (Fig. 5A). We found that TPN extended the stability of siRNA to >12 h in murine serum (Fig. 5A), and the size of TPN remained stable in the presence of serum over the course of 10 hours (Fig. 5B). Furthermore, we observed unpackaging of TPN upon exposure to endolysosomal pH (5.3-6.5) (Fig. 4D).
We assessed whether TPN enabled receptor-specific, tumor-penetrating siRNA delivery to established tumors (Fig. 5C), by examining their in vivo homing behavior in mice bearing subcutaneous human melanoma tumors (MDA-MB-435 xenografts). After intravenous administration, most siRNAs were cleared renally, as indicated by bladder accumulation of siRNA over 8 hours (fig. S6A). However, we also observed that TPN was cleared more slowly than naked siRNA (Fig. 5D) and distributed to tumor tissues within 30 minutes (fig. S6B). We noted similar pharmacokinetics for intravenous and intraperitoneal injection of TPN, suggesting the potential for multiple routes of administration (Fig. 5D; fig. S6A).
Biodistribution studies showed blood clearance through liver and spleen accumulation likely owing to reticuloendothelial uptake. TPN also distributed to the lung (fig. S6C and D), a common uptake site for passively targeted nanoparticles owing to interactions with the first capillary bed following intravenous injection (25). After clearance from circulation (4 h), we found an over three-fold increase in the tumor fluorescence of TPN in MDA-MB-435 xenografts compared to UCN (Fig. 5E). Similar results in tumor-targeting were obtained in mice bearing OVCAR-8 ovarian tumor xenografts (fig. S6E).
Although bulk tumor accumulation provides a global measure of targeting, it does not provide spatial information on the extravascular and tumor interstitial availability of TPN, which ultimately determines their efficacy. We thus performed histological characterization of TPN penetration in xenografted subcutaneous OVCAR-8 tumors at various time points and found that these TPN exhibited an initial intravascular distribution and subsequently extravasated into the tumor interstitium (Fig. 5F). The fraction of TPN that were sequestered beyond the intravascular space was significantly higher than that of UCN (Fig. 5F). Furthermore, TPN accumulated in the tumor parenchyma to a similar degree as nanocomplexes bearing a different tumor-penetrating motif, iRGD (CRGDKGPDC) (10). Non-penetrating, targeted nanocomplexes (TP-RGD4C) (26) or the commercial reagent lipofectamine failed to show a similar distribution (Fig. 5G). We also found TPN also co-localized with p32-positive tumor cells, as expected, but failed to detect UCN or naked siRNAs in tumors (fig. S6F). This pattern of distribution is similar to previously reported LyP-1 peptide (21) and LyP-1 targeted nanoparticle homing (27).
We next tested whether ID4 was essential for tumor maintenance through TPN-mediated delivery of siID4 in established human ovarian cancer xenografts in vivo. Mice harboring subcutaneous OVCAR-4 tumors were injected intravenously or intraperitoneally with TPN/siID4 every 3 days for 25 days. Control animals with tumors were injected intravenously with saline, TP-LyP-1 carrier without siRNA, or TPN/siGFP. Repeated, systemic administration of the TPN/siID4 resulted in 80-90% decrease in ID4 mRNA relative to treatment with TPN/siGFP (Fig. 6A), with a corresponding suppression of tumor growth by 82% in IV-injected and 87% in IP-injected animals (Fig. 6B, D). Tumor growth was static in siID4-treated cohorts for 30 days after the termination of TPN treatment (Fig. 6B). In contrast, TPN/siGFP or TP-LyP-1 carrier had no effect on ID4 and did not prevent further tumor growth. We also found that both methods of TPN/siID4 injection induced a significant increase in cyclin-dependent kinase inhibitor 1A (CDKN1A, p21WAF1/CIP1) transcript levels (Fig. 6C), a known target of ID4 in mediating cell-cycle arrest (28). Residual TPN/siID4-treated tumors exhibited significant apoptosis (Fig. 6E).
We then tested whether TPN/siID4 could induce therapeutic activity in mice harboring disseminated tumors. We established orthotopic OVCAR-8 tumors, confirmed intra-abdominal dissemination three weeks after implantation, and then initiated intraperitoneal injection of TPN every 3 days for 30 days. The tumor burden in mice that received TPN/siID4 remained low compared to controls (Fig. 7A; fig. S7A), and 80% of the recipients of TPN/siID4 (IP) survived more than 60 days, even after the nanocomplex treatments ceased at day 50 (Fig. 7B). After 40 days, disseminated tumors and hemorrhagic ascites were present in control cohorts (fig. S7B), but there was no visible tumor lesions in 4 out of 5 TPN/siID4–treated mice upon necropsy, indicating tumor regression. Histological analysis of the single visible remnant tumor in a siID4-treated animal revealed significant reduction in ID4 levels relative to tumors from the three control cohorts and increased apoptosis in the tumor parenchyma (Fig. 7C; fig. S7C).
The versatility of TPN for direct target comparison was demonstrated by treating mice bearing OVCAR-8 tumors with TPN carrying siRNA against Claudin-3 (CLDN3), a recently reported potential ovarian siRNA target (29). TPN/siCLDN3 led to initial reduction of tumor growth compared to controls treated with saline or TP-LyP-1 carrier alone, but was unable to achieve sustained growth suppression compared to TPN/siID4 treatments. (fig. S7D).
In some cases, apparent therapeutic effects of RNAi may be confounded by non-sequence-specific innate immune responses (30). Therefore, we independently measured serum levels of interferon α (IFN-α), tumor necrosis factor α (TNF-α), and interleukin 6 (IL-6) in immunocompetent mice after administering various TPN-based siRNA formulations, and failed to observe induction of these cytokines in animals that received TPN/siID4. By comparison, mice that received an immunostimulatory siRNA, siβgal-728 (30), complexed to either lipofectamine or TP-LyP-1 were positive controls for non-specific immunostimulation (fig. S8A). Moreover, we did not observe deleterious effects on animal weight in two ovarian tumor models (fig. S8B), or evidence of macroscopic or histological signs of organ toxicity in the OVCAR-8 orthotopic tumor model (fig. S8C).
Current genome characterization efforts will eventually provide insight into the genetic alterations that occur in most cancers, and may define new therapeutic targets. However, most epithelial cancers harbor hundreds of genetic alterations as a consequence of genomic instability. For example, while recurrent somatic alterations occur in a small number of genes in high-grade ovarian cancers, ovarian cancer genomes are characterized by multiple regions of copy number gain and loss involving at least 1825 genes. This genomic chaos complicates efforts to identify biologically relevant mutations critical for tumor maintenance.
To isolate which recurrent genetic alterations are involved in cancer initiation, tumor maintenance, and/or metastasis, functional assays can be performed following systematic manipulation of the candidate oncogenes. We have combined results from Project Achilles (2), a large scale screening effort to identify genes essential for proliferation and survival in human cancer cell lines with genome characterization of high-grade ovarian cancers. Using this approach, we identified an oncogene candidate, ID4, which was amplified in 32% of high-grade serous ovarian cancers. ID4 is overexpressed in a large fraction of high-grade serous ovarian cancers, and ovarian cancer cell lines that overexpress ID4 are highly dependent on ID4 for survival and tumorigenicity. Expression of ID4 at levels corresponding to those observed in patient-derived samples induced transformation of immortalized ovarian and fallopian tube epithelial cells. Our observations credential ID4 as an ovarian cancer oncogene and suggest that ID4 is one of a growing class of lineage-restricted transcriptional factor oncogenes in human epithelial cancers.
Like other ID family proteins, ID4 binds to and regulates the activity of E protein transcription factors such as transcription factors 3 and 12 (TCF3 and TCF12). Indeed, we found that ID4 regulates cell proliferation in part through its effects on p21WAF1/CIP1. In addition, the transcriptional program induced by ID4 in ovarian epithelial cells was similar to that induced by HOXA9. Although it is clear that ID4 regulates the expression of many genes, we found that suppression of HOXA9 abrogated ID4-induced cell transformation. HOXA9 is known to regulate normal hematopoiesis by controlling differentiation, proliferation and self-renewal (31), and overexpression of HOXA9 or HOXA9-NUP98 in hematopoietic precursors promotes the development of leukemias (32). In addition, HOXA genes have been shown to play essential roles in specifying regional differentiation of mullerian duct into oviduct, uterus, cervix, and vagina (33). Although prior work has shown that manipulation of ID1, ID2 or ID3 expression induces compensatory changes in other ID family members, we found that expression of ID4—but not the other ID family members—induced cell transformation in ovarian epithelial cells. Because amplifications involving other ID family members have not been observed in ovarian cancers, it appears that ID4 plays a unique role in the transformation of this tissue type. Collectively, these findings suggest that inappropriate overexpression of ID4 may contribute to transformation by affecting both p21WAF1/CIP1 and HOXA9 transcriptional programs that affect differentiation and proliferation programs in the ovary. Nevertheless, further mechanistic studies will be necessary to understand the biology of the new oncogene ID4 and its downstream target, HOXA9.
To credential ID4 as a therapeutic target in human ovarian cancer, we showed that a tandem peptide with both a membrane-translocating domain (Transportan, TP) and a tumor-penetrating domain (LyP-1) condensed siRNA into nanocomplexes (TPN) that, upon systemic administration into ovarian tumor-bearing mice, penetrated the tumor parenchyma and silenced ID4 in cells of interest (Fig. 4C). Treatment with TPN/siID4 suppressed the growth of ovarian tumors and significantly improved host survival. Previous approaches using inducible shRNAs to evaluate potential cancer targets were time-consuming and required extensive manipulation of cell lines in vitro (34-35). Therefore, we favor this modular, tumor-specific, tumor-penetrating approach that delivers siRNA systemically. Importantly, the TPN are biodegradable and did not elicit innate immune responses upon systemic delivery, enabling repeated administration for tumor regression studies.
Other methods for siRNA delivery include covalent coupling of siRNA to carriers, but this formulation process requires specialized chemically-reactive siRNA and purification steps, obviating the modular ‘mix and dose’ paradigm motivated by the volume of emerging genomic targets (36). Targeted siRNA delivery has been accomplished in specialized populations such as immune cells, neurons, and hepatocytes; however, these methods are not generalizable across tissue types. Elsewhere, siRNA has been targeted to tumors, but not in a tumor-penetrating fashion (5). For example, tumor endothelium has been targeted using RGD-specific nanoparticles; however, this approach does not enable silencing in parenchymal tumor cells, such as those identified with genetic alterations in TCGA (1). Passive targeting (e.g. EPR effect) of siRNA in neutral liposomes, in complexes with CPP such as TAT, or transferrin-targeted cyclodextrin has also been described in mouse models and recently in humans (37-38), but tumor penetration remained a challenge (39). Our TPN technology represents an important therapeutic paradigm for systemic targeting of “undruggable” proteins that are essential in parenchymal tumor maintenance.
Although we have focused on ID4, TPN delivery technology can be generalized to other cancers. Approximately 60 of 81 samples across 10 tumor types exhibit surface expression of p32 (20). Our modular siRNA delivery platform can leverage other homing peptides, such as iRGD, to expand the number and types of TPN-targeted parenchymal tumor cells. More generally, ongoing cancer genome characterization efforts combined with comprehensive functional studies will narrow the number of potential cancer targets, but putative oncogenes will still require validation in vivo. TPN offers an efficient platform to verify which genes are critical to cancer initiation and maintenance.
Our studies demonstrate that TPN exhibits both specificity and pharmacokinetic properties that may provide a platform for human siRNA delivery. To reach this goal, additional development is required, such as optimization of circulation time, assessment of efficacy in higher fidelity models of cancer, and reformulation to increase knockdown efficiency and reduce dosing. The path to translating TPN to the clinic will require multi-species toxicity testing, assessment of repeated dosing and potential immunostimulation, and evaluation of the prevalence of the tumor-penetration pathway and p32 in human tumors and metastases.
In summary, we developed a targeted tumor-penetrating nanocomplex capable of precisely delivering siRNA into the tumor parenchyma, and have combined this technology with large-scale methods to credential ID4 as an oncogene target in ovarian cancer. As large scale efforts to characterize all cancer genomes accelerate, this capability illustrates a path to identify genes that are altered in tumors, validate those that are critical to cancer initiation and maintenance, and rapidly evaluate in vivo the subset of such genes amenable to RNAi therapies and clinical translation.
shRNA constructs are available through Sigma Aldrich. Genome-scale pooled shRNA screens in 102 cancer cell lines were performed using a lentivirally delivered pool of 54,020 shRNAs targeting 11,194 genes. Each cell line was infected in quadruplicate and propagated for at least 16 population doublings (2) (Supplementary Methods). The abundance of shRNAs relative to the initial DNA plasmids was measured by microarray hybridization and normalized using dCHIP and GenePattern modules before identifying essential genes.
Copy number alterations were determined by TCGA on 489 primary ovarian tumors using Illumina 1MDUO arrays and recurrently amplified regions were analyzed by GISTIC 2.0 (40) with 99% confidence intervals (Supplementary Methods). For each amplified gene, the distribution of shRNA scores were converted to z-score and shRNAs with a p<0.05 were identified.
To determine the effects of shRNA on cell proliferation, six replicate infections were performed and viability was determined 5 d later (Supplementary Methods). Transformation was determined by introducing ID4 into immortalized IOSE-M (41) and FTSEC-M cells (16) and performing anchorage-independent growth experiments or monitoring tumor formation by injecting cells subcutaneously into BALB/c immunodeficient mice (Charles River).
The tandem peptide library was synthesized via standard FMOC solid-phase peptide synthesis and purified by HPLC (Supplementary Methods). Peptides were mixed with siRNA at a molar ratio of 20:1. Nanocomplexes were incubated over cell culture for 4 h and cells were harvested 48h later for characterization of gene knockdown. For competition with free LyP-1, unlabeled LyP-1 peptide was added for 1 h prior to TPN treatment.
TPN carrying ID4-specific siRNA were administered intravenously or intraperitoneally to NCr/nude mice bearing either subcutaneous OVCAR-4 or intraperitoneal disseminated OVCAR-8 tumors, respectively. Tumor burdens were monitored via caliper measurements or bioluminescence imaging (Supplementary Methods).
Unless otherwise indicated, one-way analysis of variance (ANOVA) was used (GraphPad). A P value < 0.05 was considered statistically significant. Fisher's exact test was used for tumor formation assays. Two-tailed Student's t test was used for pair-wise comparisons. A log-rank test was performed for animal survival studies.
Tumor-penetrating siRNA nanocomplexes credentials ID4 as a therapeutic oncogene target in human ovarian cancer.
We thank H. Fleming, A.A. Chen, P. Sharp, and E. Lander for critical readings of the manuscript (MIT). We thank D. Livingston for the luciferized OVCAR-8 cells and valuable insight into ovarian cancer (DFCI). Funding: Supported by the Howard Hughes Medical Institute (S.N.B.), National Cancer Institute U54 CA119349 (S.N.B.), U54 CA119335 (S.N.B.), R01 CA124427 (S.N.B), R33 CA128625 (W.C.H.), RC2 CA148268 (W.C.H.), U54 CA112962 (W.C.H.), RO1 CA152327 (E.R.), P50 CA105009 (R.D.), NIH U01 CA152990 (R.D.), Starr Cancer Consortium (W.C.H. And S.N.B.), The Marie-D. & Pierre Casimir-Lambert Fund (S.N.B), The Mary Kay Foundation (R.D.), the Sandy Rollman Ovarian Cancer Foundation (R.D.), Canadian Institutes of Health Research Fellowship (A.M.K.), and partially by Cancer Center Support (core) Grant P30 CA14051 from the NCI.
Author contributions: Y.R., H.W.C., G.S.C., B.A.W., J.S.B., D.E.R., W.C.H., and S.N.B. designed the experiments. Y.R., G.V.M., A.A., and S.N.B. designed the peptide screen. Y.R. characterized TPN. G.S.C., J.S.B., B.A.W., D.E.R., and W.C.H. supervised RNAi screens. B.A.W., J.S.B., H.W.C., G.S.C., R.D., P.T., J.P.M., D.E.R., and W.C.H. analyzed the RNAi data. H.W.C., and A.M.K. performed experiments with transformation and shRNAs. B.A.W., H.W.C., and J.S.B. performed expression profiling and analysis. Y.R., H.W.C., W.C.H., and S.N.B. analyzed the integrated RNAi data. J.L. assisted with biodistribution studies. M.S.H. and J.F.L. analyzed the tissue-microarray. V.F. and E.R. provided advice, reagents and analysis on tumor targeting aspects of the work. Y.R., H.W.C., W.C.H., and S.N.B. wrote the manuscript. All authors discussed the results and commented on the manuscript.
Competing interests: E.R. is a shareholder of CendR Inc, which licenses the iRGD peptide. A patent has been filed by M.I.T. describing the tumor-penetrating nanocomplexes.
Data and materials availability: All data files can be assessed at the Integrative Genomics portal (http://www.broadinstitute.org/IGP/home). An MTA is required for distribution of the primary FT epithelial cells (A.M.K).