U1 Adaptor Oligonucleotides Targeting BCL2 and GRM1 Suppress Growth of Human Melanoma Xenografts In Vivo.

U1 Adaptor is a recently discovered oligonucleotide-based gene-silencing technology with a unique mechanism of action that targets nuclear pre-mRNA processing. U1 Adaptors have two distinct functional domains, both of which must be present on the same oligonucleotide to exert their gene-silencing function. Here, we present the first in vivo use of U1 Adaptors by targeting two different human genes implicated in melanomagenesis, B-cell lymphoma 2 (BCL2) and metabotropic glutamate receptor 1 (GRM1), in a human melanoma cell xenograft mouse model system. Using a newly developed dendrimer delivery system, anti-BCL2 U1 Adaptors were very potent and suppressed tumor growth at doses as low as 34 µg/kg with twice weekly intravenous (iv) administration. Anti-GRM1 U1 Adaptors suppressed tumor xenograft growth with similar potency. Mechanism of action was demonstrated by showing target gene suppression in tumors and by observing that negative control U1 Adaptors with just one functional domain show no tumor suppression activity. The anti-BCL2 and anti-GRM1 treatments were equally effective against cell lines harboring either wild-type or a mutant V600E B-RAF allele, the most common mutation in melanoma. Treatment of normal immune-competent mice (C57BL6) indicated no organ toxicity or immune stimulation. These proof-of-concept studies represent an in-depth (over 800 mice in ~108 treatment groups) validation that U1 Adaptors are a highly potent gene-silencing therapeutic and open the way for their further development to treat other human diseases.Molecular Therapy - Nucleic Acids (2013) 2, e92; doi:10.1038/mtna.2013.24; published online 14 May 2013.

Introduction

The ability to silence a single target gene with minimal toxicity and few off-target effects (OTEs) would offer the medical community new tools to combat a wide variety of diseases. Several hybridization-based technologies have been tested over the past 30 years that utilize synthetic oligonucleotides to suppress expression of a specific target gene. In broad terms, these technologies are based on a gene being targeted either by small interfering RNAs (siRNAs) which exploit the natural cellular RNA interference machinery[1,2] or by an antisense oligonucleotide (ASO) that can function via a variety of mechanisms, including RNase H-mediated cleavage of RNA, steric hindrance of mRNA translation, splice site switching, and miRNA antagonists.[3,4,5] ASOs and siRNAs have struggled to produce commercially viable therapeutics, in spite of initial excitement and large-scale investment[6] presenting an opportunity for alternative gene-silencing technologies.

U1 Adaptors are a recently invented gene-silencing technology that exploits the natural ability of the U1 small nuclear ribonucleoprotein (snRNP) splicing factor to inhibit gene-specific polyA site activity of the target gene, a regulated nuclear pre-mRNA processing step obligatory for nearly all RNA Polymerase II genes.[7,8,9,10,11,12] A U1 Adaptor is a synthetic oligonucleotide (typically 28–33 nucleotides) comprised of a 5′ target domain (TD), which binds to the target pre-mRNA, and a 3′ U1 domain (U1D), which binds to the 5′-end of the U1 small nuclear RNA subunit of U1 snRNP.[7,8] Tethering of the U1 snRNP to a target pre-mRNA blocks maturation leading to reduced levels of mature mRNA. The U1D sequence is common to all U1 Adaptors and is defined by the U1 snRNP; design and chemical modification patterns have already been optimized.[7] In contrast, the TD sequence is target-specific and hence unique to each U1 Adaptor. Like all other gene knockdown technologies, empiric testing is required for site selection and optimization. Extensive medicinal chemistry studies have been done in ASO and siRNA systems to find chemical modifications that improve nuclease stability, enhance potency, and reduce OTEs.[3,5,13,14,15,16,17] Unlike siRNAs or RNase H-mediated ASOs, U1 Adaptors do not interact or function with any cellular enzymes (such as RNase H, Dicer, Argonaut 2, etc.) and thus can be made entirely using modified components, such as 2′-O-Methyl RNA (2′OMe) or locked nucleic acids (LNAs), with or without phosphorothioate (PS)-modified internucleotide linkages.[7] All oligonucleotide-based silencing technologies have associated toxicities and U1 Adaptors are no exception, as evidenced by a recent report that showed significant OTEs when used at high doses.[18] In a previous report, use of U1 Adaptors at a lower dose gave effective and specific silencing with few OTEs.[7]

Here, we present the first report of in vivo use of U1 Adaptors by targeting two human genes to suppress growth of human melanoma cells in a mouse xenograft model system. The antiapoptotic human B-cell lymphoma 2 (BCL2) gene has been a frequent target in studies using ASO and RNA interference technologies[19,20,21,22,23,24] and plays a role in many cancers, including melanoma[25,26,27,28,29] that has extraordinary intrinsic resistance to apoptotic cell death commonly induced by anticancer drugs, in part due to elevated levels of BCL2.[28,29] We chose a human melanoma xenograft mouse system for its proven track record in predicting efficacy in clinical trials as evidenced by riluzole that has gone on to show efficacy in phase 0 and 2 human trials.[30] Our second target, the metabotropic glutamate receptor 1 (GRM1) gene has only recently been established as an important factor in melanoma as well as other cancers.[31,32,33] Using our novel tumor-specific dendrimer delivery vehicle, we demonstrate that very low doses of anti-BCL2 or anti-GRM1 U1 Adaptors are sufficient to reduce growth/progression of human melanoma xenografts with little apparent toxicity. These results offer proof-of-concept that U1 Adaptors are an effective gene-silencing therapeutic platform that can suppress tumor growth using doses far lower than expected based on published experience using other oligonucleotide-based methods.[34] These results also lay a foundation for exploiting U1 Adaptors to target other genes as well as a wide variety of other human disorders.

Results

U1 Adaptor silencing of BCL2 in vitro

Twelve anti-human-BCL2 U1 Adaptors were screened for functional potency in C8161 melanoma cells (Supplementary Figure S1a,b). Two U1 Adaptors, BCL2-A and BCL2-B showed strong activity in reducing BCL2 mRNA levels (Supplementary Figure S1b,c) and were used in subsequent studies. LNA- and 2′OMe-modified variants of BCL2-A were compared for activity in C8161 cells (, Supplementary Figure S2a) and the LNA-modified variant BCL2-AL2 showed the highest silencing activity at both the protein (western immunoblots) and mRNA (reverse transcription-quantitative PCR) levels ( lanes 5–6, Supplementary Figure S2b). To demonstrate that the silencing activity of BCL2-AL2 is mediated via a U1 Adaptor mechanism, C8161 cells were transfected with matching control U1 Adaptors having either the inactivated TD or U1D by mutation or by unlinking the TD and U1D into two separate molecules called “half-Adaptors” (). In all cases, these negative controls failed to reduce BCL2 protein () or mRNA (Supplementary Figure S2b) levels, strongly supporting a U1 Adaptor-based silencing mechanism, e.g., one where the U1 Adaptor oligonucleotide must tether the target pre-mRNA to the U1 snRNP. Modification of BCL2-A with three PS linkages at each end to improve exonuclease resistance (BCL2-Aps) or a fluorescent label (Cy3-BCL2-A) did not alter potency compared with the parent compound (Supplementary Figure S2c).

Development of a tumor-targeting dendrimer nanoparticle

The cyclic RGD pentapeptide (RGD) was chosen as a tumor-targeting ligand because it is small, easy to conjugate, and specifically binds the α5β3 splice variant of an integrin cell surface receptor that is overexpressed in a wide variety of cancer cells, including C8161.[35,36] The particle was based on a generation 5 (G5) polypropyleneimine (PPI) dendrimer that was previously used to successfully deliver nucleic acid cargos in mouse xenograft models of human tumors.[37,38,39,40,41] As RGD binds to its receptor with much higher affinity as a dimer,[42,43] the RGD targeting ligand was coupled to the PPI G5 dendrimer in a final 2:1 molar ratio to give RGD-G5 (). Transfection of C8161 cells in vitro with a BCL2-A:RGD-G5 complex demonstrated that RGD-G5 was active as a delivery vehicle, having potency comparable to transfection of BCL2-A with the cationic lipid reagent Lipofectamine 2000 (LF2000) (, Supplementary Figure S3a–c). Surprisingly, an anti-BCL2 Dicer-substrate siRNA that was effective using LF2000 transfection did not suppress BCL2 mRNA levels when used with the RGD-G5 dendrimer, suggesting this delivery vehicle is not compatible with siRNA class reagents (). Use of RGD to PPI G5 coupling ratios of 4:1 and 8:1 resulted in reduced efficacy in vitro (Supplementary Figure S4).

BCL2-A is active in vivo and uses a U1 Adaptor mechanism

Mice bearing established C8161 subcutaneous xenografts were administered BCL2-A:RGD-G5 or control complexes in phosphate-buffered saline (PBS) via intravenous (iv) tail vein injection. Pilot studies (data not shown) established that effective tumor suppression could be achieved using 1.7 µg of the U1 Adaptor and a 1:1.3 Adaptor:RGD-G5 stoichiometry. Dynamic light scattering measurements indicated the particle size of the injected complex (~3.2 µmol/l U1 Adaptor and ~4 µmol/l RGD-G5 in 1x PBS) was 193 nm with a polydispersion index = 0.18 with little difference in size for different U1 Adaptor sequences (Supplementary Figure S5). Mice were given biweekly injections of various formulations for 3 weeks and the tumors were evaluated (). Tumor-bearing treatment group 2 (TG2) mice that were dosed with a 1.7 µg BCL2-A:2.4 µg RGD-G5 complex (68 µg Adaptor/kg) per injection exhibited ~70% tumor reduction compared with TG1, the vehicle control group (P = 0.004). No significant tumor suppression was seen with the two control U1 Adaptor cohorts, which were treated with either a TD-mutant BCL2-A (TG3) or a U1D-mutant BCL2-A (TG4). Thus disruption of either the TD or the U1D resulted in loss of activity, strongly arguing for a U1 Adaptor mechanism of action. The TG6 mice were dosed with a 1.7 µg BCL2-A:2.4 µg G5 complex, where G5 matches the RGD-G5 vehicle but lacks the RGD ligand. This cohort did not show reduction in tumor mass, demonstrating the importance of the targeting ligand. Use of a fivefold lower dose (0.34 µg BCL2-A:0.48 µg RGD-G5 complex) resulted in no inhibition of tumor growth (TG5), establishing that >0.34 µg Adaptor/dose was needed for tumor suppression.

Protein and RNA extracted from tumors of day 22 TG1-4 mice demonstrated that only the BCL2-A:RGD-G5–treated mice had reduced BCL2 protein () and mRNA (Supplementary Figure S6a). Others have shown that therapeutic-based BCL2 silencing in xenograft mice by either ASOs or siRNAs increases the rate of apoptosis.[19,20,21,23] Immunohistochemical analysis shown in demonstrated that BCL2-A–treated, but not the control animals, showed an increased number of positive, activated Caspase 3 stained cells. Furthermore, immunohistochemical analysis using the cell proliferation marker, Ki67, showed that BCL2-A–treated, but not the control animals, had a reduced number of actively proliferating cells (). Both visual and histopathological inspection of several organs including liver, kidney, spleen, heart, brain, and lung from the TG1-TG6 mice showed no evidence for tissue damage or toxicity (data not shown). None of the mice exhibited overt signs of toxicity such as lethargy, not eating or drinking, loss of body weight or ulceration of the transplanted tumor.

While the above data support that BCL2-A's tumor suppression activity is likely to be mediated through reducing BCL2 mRNA levels, it did not rule out a possible contribution of an extracellular mechanism such as impaired tumor vascularization, a possible side effect of RGD targeting.[35,36,42,43] This was addressed by showing that C8161 cells transfected with BCL2-A:RGD-G5 ex vivo before implantation led to significant tumor suppression (). The results of this ex vivo experiment strongly suggest that tumor suppression detected in vivo is primarily if not completely mediated through an intracellular mechanism and also demonstrate a long-lasting effect as the single ex vivo transfection of BCL2-A led to tumor suppression persisting at least through 23 days in vivo. C8161 cells transfected with RGD-G5:Adaptor complexes all had a similar doubling time and apoptosis rate (data not shown) independent of whether the Adaptor was capable of silencing BCL2 indicating in vitro results do not always predict in vivo activities.

Two chemically modified variants of BCL2-A were studied (BCL2-Aps with terminal PS modification and BCL2-AL2 with LNA residues in the TD) in an attempt to increase potency in vivo (). Although BCL2-Aps (TG9) gave significant tumor suppression as compared with the vehicle control (TG7), it was less active than BCL2-A (TG8). BCL2-AL2 (TG13) showed tumor suppression equal to BCL2-A, an observation that contrasts with its superior activity in vitro, further demonstrating that in vitro results do not always translate to similar outcomes in vivo. Others have also found unpredictable behavior of LNA backbone modifications.[16,17] The other data in demonstrated that a 2:1 RGD:PPIG5 coupling ratio was superior to a 4:1 ratio (compare TG8 with TG10); further, a 1:4 Adaptor:RGD-G5 stoichiometry was less effective than a 1:1.3 stoichiometry (compare TG16 with TG18); a minimally effective dose for BCL2-A was 0.85 µg/injection (34 µg Adaptor/kg).

To further support that BCL2 is indeed the therapeutic target and that the observed tumor suppression was not due to sequence-specific OTEs, we analyzed a second anti-BCL2 Adaptor (BCL2-B) which targets a different site of the BCL2 gene. BCL2-B also was effective in reducing tumor mass, albeit with lower potency than BCL2-A (), and used a U1 Adaptor mechanism as single domain mutant BCL2-B Adaptors were not active in vivo (). Notably, used an (RGD)x2-containing G5 PAMAM dendrimer (RGD-PAMAM-G5) in place of the PPI core found in RGD-G5 as we wanted to demonstrate that our results were not dependent on a particular class of dendrimer. Indeed a head-head comparison of RGD-G5 and RGD-PAMAM-G5 found they had comparable tumor suppression activity in vivo (Supplementary Figure S6b). Dynamic light scattering measurements indicated the particle size of the injected complex (~3.2 µmol/l U1 Adaptor and ~4 µmol/l RGD-PAMAM-G5 in 1x PBS) was 200 nm with a polydispersion index = 0.147 with little difference in size for different U1 Adaptor sequences (Supplementary Figure S4c). We also demonstrated that BCL2-A's tumor suppression activity is not peculiar to C8161 xenografts as BCL2-A also suppressed UACC903 xenograft tumor growth (Supplementary Figure S6c). As compared with C8161, UACC903 is a more aggressive and genotypically distinct melanoma that harbors a V600E B-RAF mutation found in 70% of all melanomas (C8161 has a wild-type B-RAF).44,45

Anti-GRM1 Adaptors also suppress melanoma tumor growth at low dose

Ectopic expression of GRM1 in melanocytes is sufficient to induce melanocytic cell transformation in vitro and spontaneous melanoma development in vivo; short hairpin RNA-based silencing of GRM1 inhibited tumor cell growth in vitro and in vivo.[31,46] Screening anti-human–GRM1 Adaptors in C8161 cells in vitro identified three candidate U1 Adaptors that reduced human GRM1 at the protein () and mRNA (Supplementary Figure S7a–d) levels. Treatment of C8161 xenograft mice with these anti-GRM1 U1 Adaptors decreased tumor growth and reduced GRM1 protein levels in day 21 excised xenograft tumors compared with the vehicle control (). To demonstrate mechanism of action, the two most potent anti-GRM1 Adaptors (GRM1-Aps and GRM1-Bps) underwent additional analysis as follows, GRM1-Bps uses a U1 Adaptor mechanism as repeating the GRM1-Bps treatment with matching TD-mutant or U1D-mutant control U1 Adaptors gave no tumor suppression () and no reduction of GRM1 at the protein and mRNA levels as compared with RGD-G5–treated (vehicle only) control mice (). TD and U1D mutant Adaptor control xenograft experiments for GRM1-Aps also demonstrated that the reduced tumor volumes were again mediated by a U1 Adaptor mechanism (). Tumor suppression was lost when RGD was replaced with RAD, a well-validated inactive sequence variant of RGD that fails to bind integrin receptors, once again demonstrating that tumor targeting is required. Finally, testing the variant RGD-G5/SPDP vehicle that has a short 10 Å non-PEG–containing linker (in contrast to the 50 Å PEG-linker employed in the original RGD-G5 dendrimer) showed significant reduction in tumor volume, demonstrating that RGD-G5's linker length and composition were not critical for activity.

It was previously shown that short hairpin RNA-based silencing of GRM1 in human melanoma cell xenografts leads to reduced levels of phosphorylated AKT (pAKT).[46,47] Excised GRM1-Aps U1 Adaptor-treated day 17 tumor samples also showed decreased levels of pAKT (Supplementary Figure S8). Immunohistochemical analysis of the excised day 17 tumors from the anti-GRM1 U1 Adaptor-treated xenografts demonstrated increased activated Caspase 3 and decreased Ki67 staining (), similar to the previous observation using anti-BCL2 U1 Adaptors. Both GRM1-Aps and GRM1-Bps U1 Adaptors were also able to suppress tumor progression of UACC903 xenografts, broadening their scope of action to melanoma with mutated B-RAF (Supplementary Figure S9a). In contrast to previous observations using BCL2-A, the presence of PS end-modifications on the anti-GRM1 U1 Adaptors resulted in superior in vivo activity compared with the same oligonucleotides with phosphodiester end linkages (Supplementary Figure S9b).

No toxicity or immune stimulation was observed in U1 Adaptor-treated C57BL6 mice

It was important to examine the effects of U1 Adaptor administration in immune-competent mice (C57BL6) to directly assess whether the U1 Adaptor antitumor effects could in part be based on stimulation of the innate immune system, a complication observed using other gene-silencing therapeutics.[48,49,50] Although a variety of receptors will recognize various forms of DNA and RNA, no receptors are known that recognize 2′OMe RNA or LNA residues, making it unlikely that the U1 Adaptors employed in the present study would trigger immune activation. BCL2-A and GRM1-Aps U1 Adaptor:RGD-G5 complexes were injected into C57BL6 mice; interleukin-12 (IL-12) and interferon-α (IFN-α) levels were measured 4 hours post-injection. Baseline levels of both cytokines were seen for all mice except those in the positive control cohort which were administered poly(I:C), a known toll-like receptor 3 (TLR3) agonist ( and Supplementary Figure S10).[51] In a similarly designed but independent experiment, baseline levels of IL-12 and interferon-α were also observed 24 hours post-injection (Supplementary Figure S11a,b). C57BL6 mice were given the standard 3-week U1 Adaptor treatment regimen (68 µg Adaptor/kg; two times/week) and serum levels of alkaline phosphatase and alanine transaminase were measured to assess for evidence of hepatic injury; both enzymes remained at basal levels throughout the study period (). These same serum samples also contained basal levels of IL-12 and IFN-α levels indicative of no immune stimulation even after 3 weeks of biweekly treatments (data not shown). Six organs from the same mice (liver, kidneys, spleen, heart, brain, and lungs) were excised, stained, and analyzed by the Rutgers-UMDNJ Molecular Histopathological Facility Core. No tissue damage was detected and only rare foci of inflammatory cells were seen in all organ samples. Thus no evidence for toxicity was found for any treatment group.

Discussion

U1 Adaptors were previously shown to be effective triggers of gene silencing in mammalian cells in vitro but their applicability for in vivo use was heretofore untested. The present study offers proof-of-concept that U1 Adaptors can be effective in suppressing expression of targeted genes and slow tumor growth in vivo when combined with a suitable delivery vehicle. Success was observed using the same xenograft mouse system and human cell lines (C8161 and UACC903) that proved predictive for riluzole in preclinical studies that later went on to show efficacy in phase 0 and 2 human trials.[30] The broad utility of the U1 Adaptor approach is underscored by the fact we successfully targeted two genotypically different melanoma cell lines (BRAF wild-type and BRAF V600E mutant) and two different cancer-implicated genes, BCL2 and GRM1. The technology also proved robust as our very first in vivo experiment and all subsequent experiments, encompassing ~108 treatment groups totaling over 800 mice, exhibited a very high potency for both genes, suppressing tumor progression at doses as low as 34–68 µg Adaptor/kg. U1 Adaptors have two distinct functional domains, both of which must be present on the same oligonucleotide to function via a U1 snRNP-based mechanism of action. Suitable negative controls for mechanism of action include use of full-length U1 Adaptors with mutations in either the TD or the U1D or half-Adaptors (isolated TD or U1D) to show that both functional domains must reside on the same molecule.[7] The present study employed all of these mechanistic controls and demonstrated that an intact U1 Adaptor oligonucleotide with functional (i.e., non-mutated) sequences in both the TD and U1D is required to achieve tumor suppression and a gene suppression effect. These controls further support a U1 mechanism of action in both the in vitro and in vivo studies.

Tumor growth can be suppressed by a variety of routes including downregulation of necessary oncogenes, nutritional starvation (disruption of vascular supply), direct chemical cytotoxicity, or immune activation. In the present study, we found evidence for induction of apoptosis and reduced cell division only in tumor-bearing animals treated with intact (non-mutated) anti-BCL2 or anti-GRM1 U1 Adaptors, which correlated with reduction of BCL2 or GRM1 levels in the tumor tissue. Treated animals showed no evidence for either general chemical toxicity from the administered compounds or immune activation. Further, tumor cells treated ex vivo with the U1 Adaptors before implantation showed marked growth retardation in animals that never received iv dosing of the U1 Adaptor:RGD-G5 complex, making it unlikely that disrupted vascularization contributed to tumor suppression.

U1 Adaptors are large, highly charged synthetic nucleic acids with a molecular weight of around 11,000 daltons. Molecules of this type do not readily enter most cell types without assistance, so use of an appropriate delivery vehicle is crucial to the success. The need for efficient delivery, one of the primary challenges of the oligonucleotide therapeutic field, continues to inspire production of an ever-widening variety of delivery systems ranging from cationic lipid or polymer-based nanoplexes, to antibody-protein fusion systems, to encapsulation of the oligonucleotide drug in exosomes or bacterial minicells. With the exception of delivery to liver,[34,52,53] efficacious siRNA-based targeting using low microgram doses is usually not possible with systemic iv dosing and typically is only seen with local administration (e.g., direct intraocular or intratumoral injection). Use of dendrimers to deliver therapeutic oligonucleotides is relatively under-explored, although a recent report successfully used a PAMAM dendrimer to deliver therapeutic doses of multiple Dicer-substrate siRNAs at single digit microgram levels in a humanized mouse model of HIV1.[54] Although the PPI dendrimer-based system employed here was useful in this proof-of-concept study and exhibited no detectable toxicity, it is clearly prudent to explore the compatibility of other delivery systems with U1 Adaptors. It is important to acknowledge the contribution of ligand targeting for the success we observed using the present dendrimer-based systems. The low-dose efficacy achieved here required the RGD targeting ligand and tumor suppression was lost when the RGD ligand was either omitted from the vehicle or substituted with the related but inactive RAD variant.

To date, most reports targeting BCL2 with siRNAs and ASOs in tumor-bearing mice have involved injection of a naked oligonucleotide, precluding comparison with the work presented here. Systemic injection in tumor-bearing mice of 10 mg doses of a naked anti-BCL2 ASO was efficacious leading to phase 3 trials that ultimately failed due to poor efficacy.[2,27] Systemic injection of a naked anti-BCL2 siRNA at 200 µg/kg daily for 24 days led to tumor suppression in human Panc1 (pancreatic) mouse xenografts.[22,23] Of particular interest was a dual functional siRNA that suppressed BCL2 via an RNA interference mechanism while activating RIG-1 expression and an immune response through a 5′ end triphosphate group with a combined efficacy at 50 µg doses delivered on days 3, 6, and 9 in a mouse melanoma model in C57BL6 mice.[55] However, the complex chemistry needed to produce triphosphate siRNAs may preclude the scale-up needed for clinical studies. While the RGD ligand used here was chosen for its simplicity and prior success when used at very low doses to image human tumors growing in mice,[56] there are many other tumor-targeting ligands that may prove superior. Although concern has been raised about the nonspecific effects of RGD when used at high doses,[43] this may be largely alleviated at far lower doses such as the ones reported here.

The achievement of proof-of-concept paves the way for future studies including exploration of pharmacodynamic and pharmacokinetic properties as well as expanded dose response studies and the use of other delivery systems/targeting ligands. Previous in vitro work found that, unlike siRNAs where the most potent siRNA in a pool appears to dominate the functional response,[57,58] the combined use of several U1 Adaptors against the same target increased the level of gene suppression.[7] The use of multiple U1 Adaptors against the same target may permit even lower dosing to be used, which could be beneficial if dose-related toxicity or other OTEs were found in a different system. Further, considering that clinical responses are not durable and relapse is a near-certainty for most cancers using monotherapy, the possibility of simultaneous multi-gene targeting with multiple U1 Adaptors to achieve higher levels of tumor cell apoptosis and longer lasting tumor suppression is an attractive option. Combinational therapy combining U1 Adaptors and small molecule chemotherapeutic agents is also expected to provide enhanced therapeutic benefit, an approach already shown to be successful for siRNAs[23] and ASOs.[21,24,27]

Materials and Methods

Xenografts in immunodeficient nude mice. All animal studies were approved by the Institutional Review Board for the Animal Care and Facilities Committee of Rutgers University. Male nude mice (5 weeks old) were purchased from Taconic (Hudson, NY). Human melanoma cells, C8161 or UACC903, were injected into the dorsal area at 106 cells per site. Unless indicated, all treatment groups comprised 10 mice. Tumors were measured weekly with a Vernier caliper and tumor volume (V; cubed centimeters) was calculated by using the equation V = d 2x·D/2, where d (centimeters) and D (centimeters) are the smallest and largest perpendicular diameters. Once tumors reached ≈10 mm3, the animals were divided randomly into treatment groups so that the mean difference in tumor size between each group was <10%. Mice were killed after ~3 weeks (for C8161) or ~5 weeks (for UACC903) corresponding to when the tumor volume in the vehicle-treated group had reached the maximum size permitted by the institutional review board. The tumor xenografts were excised for further histological and molecular analyses. P values were calculated using Student's two-tailed t-test.

Preparation and analysis of RNA, proteins, and source of oligonucleotides. RNA and protein levels from transfected cells and excised tumors from xenografts were determined by western blotting and reverse transcription-quantitative PCR as previously described,[7,31] except for the immunoprecipitation (IP) western blot (described below). The quantitative PCR primers, siRNAs, and gene-specific U1 Adaptors were manufactured by Integrated DNA Technologies (IDT, Coralville, IA) except for the LNA-containing Adaptors, which were purchased from Exiqon (Vedbaek, Denmark). The Control U1 Adaptor used is a 2′OMe-modified oligonucleotide not complementary to human transcripts 5′-GUCAGAAAUACA CAAUACCUGCCAGGUAAGUAU derived from a reporter transcript described in ref. [7]. Poly(I:C) (VacciGrade) was purchased from Invivogen (San Diego, CA).

IPs and western blots. IP of protein from excised tumors was done as follows with all steps done at 4 °C. Protein G beads (catalog 17-0618-01; GE Healthcare, Pittsburgh, PA) were prepared by washing 3x in TNET (20 mmol/l Tris pH 7.4, 150 mmol/l NaCl, 1 mmol/l EDTA, 1% triton X-100), followed by 1 hour TNET + 1 mg/ml bovine serum albumin, followed by three washes in TNET and resuspended as a 50% slurry in TNET. Each IP sample contained 500 µg of tumor protein diluted to 300 µl with TNET containing protease cocktail inhibitors II, III, and IV (Calbiochem catalog nos. 524625, 539134, and 524628; Merck KGaA, Darmstadt, Germany) diluted as per the manufacturer's instructions. To each IP sample in ; lanes 3–6, was added 8 µl anti-BCL2 antibody (Invitrogen catalog 13-8800; Life Technologies, Carlsbad, CA) and 1 µl antitubulin antibody (catalog AB-2/DM1A; Abcam, Cambridge, MA), except for lane 2 that had only antitubulin antibody. After gentle rocking for 4 hours, 20 µl protein G bead slurry (10 µl packed beads) was added and the sample gently rocked for four more hours. The beads were gently pelleted, the supernatant aspirated and the beads washed three times in TNET with a 5-minute rocking/wash. After removing the last of the supernatant, the beads were resuspended in laemmli loading buffer, heated and analyzed by 12% western blot. The upper half of the membrane was probed with antitubulin antibody. To avoid interference with the mouse light chain signal that migrates close to the BCL2 band, the lower half of the membrane was probed with anti-BCL2 from rabbit (SAB4300340; Sigma-Aldrich, St Louis, MO).

Immunohistochemistry. The Tissue Analytical Services at the Cancer Institute of New Jersey performed all the immunohistochemical staining of excised tumor xenografts to detect changes in the number of apoptotic and proliferating cells using the well-known activated Caspase 3 and Ki-67 markers, respectively. The number of stained cells was quantified with a digital Aperio ScanScopeGL system and ImageScope software (v 10.1.3.2028) (Aperio Technologies, Vista, CA) according to the manufacturer's protocol with modifications as described.[59]

Immune response and liver enzyme assays. IL-12 and IFN-α levels in mouse serum samples were measured using specific ELISA kits from Invitrogen (Life Technologies) and PBL Interferon Source (Piscataway, NJ), respectively. The activities of alanine transaminase and alkaline phosphatase in mice serum were determined using the MaxDiscovery ALT enzymatic assay kit and MaxDiscovery AP enzymatic assay kit (Bioo Scientific, Austin, TX) according to the manufacturer's instructions.

Statistics. The number of mice used for each experiment was determined with the help from the Biometrics Facility Core at the Cancer Institute of New Jersey. P values were determined using unpaired Student's two-tailed t-test. Xenograft experiments comprised 10 mice/group except in a few cases where 5–8 mice/group were used. With 10 mice per group, a 35% treatment difference between control and tested group can be detected with 80% power at an α-level of 5% (two-sided test). A P value of <0.05 was considered significant.

Preparation of RGD-G5 nanoparticles. PPIG5 (SyMO-Chem, Eindhoven, Netherlands) was diluted to a 10 mmol/l stock in water using HCl to bring the pH to 7.0 and stored in aliquots at −80 °C. cRGD and cRAD (Peptides International, Louisville, KY), and the SM(PEG)12 and LC-SPDP linkers (Thermo Scientific Pierce, Rockford, IL) were resuspended and stored as per the manufacturer's instructions. Conjugation to make RGD-G5 was as follows: 24 µl of 200 mmol/l SM(PEG)12 in DMSO was added to 6.1 ml of 0.25 mmol/l PPIG5 in aqueous buffer containing 50 mmol/l KCl, and incubated for 1 hour at room temperature, (SM(PEG)12 was threefold excess over PPIG5). Glycine was then added to a final 3 mmol/l concentration to neutralize any unreacted SM(PEG)12 and after 30 minutes the solution brought to 40 mmol/l Tris pH 6.8 and a sixfold excess cRGD added to a final 1.5 mmol/l concentration. After 1 hour of conjugation, excess cysteine was added to neutralize any unreacted groups and the sample extensively dialyzed 48 hours with three changes of buffer against 30 mmol/l Hepes pH 7.9. The final sample was 200 µmol/l RGD-G5 in 30 mmol/l Hepes pH 7.9 and was stored in aliquots at −80 °C. Multiple freeze thawing (>20) had no detectable effect on activity.

Preparation of RGD-G5/SPDP, a short, non-PEG–containing linker version of RGD-G5. RGD-G5/SPDP was prepared as follows: 8 ml of 0.25 mmol/l PPIG5 in 10 mmol/l Hepes pH 7.9, 1 mmol/l EDTA, and 1.5 mmol/l LC-SPDP (added last) was incubated for 2 hours at room temperature. After overnight dialysis with two changes of buffer against 10 mmol/l Hepes pH 7.9, 1 mmol/l EDTA, a pyridine 2-thione release assay was done as per the manufacturer's instructions to determine the conjugation efficiency. Then 500 µl of 5 mg/ml cRGD freshly dissolved in 3% acetic acid was added to 7.3 ml of the activated PPIG5 and the reaction proceeded overnight at 4 °C. After overnight dialysis with two changes of buffer against 10 mmol/l Hepes pH 7.9, 1 mmol/l EDTA, a pyridine 2-thione release assay was done and the final samples stored in aliquots at −80 °C.

Dynamic light scattering. Particle size of the injected formulation at standard dose either a 1.7 µg (3.2 µmol/l) U1 Adaptor:2.4 µg (4 µmol/l) RGD-G5 complex in 1x PBS or a 1.7 µg (3.2 µmol/l) U1 Adaptor:8 µg (4 µmol/l) RGD-PAMAM-G5 complex in 1x PBS was determined by a DLS apparatus (Malvern Zetasizer; Malvern Instruments, Westborough, MA). Several dilutions of the sample were analyzed by automatic measurement setting comprised of the average of runs, each run having 30 measurements of 5 seconds per measurement.

Preparation of RGD-PAMAM-G4 and RGD-PAMAM-G5. PAMAM-G4 and PAMAM-G5 (Dendritech, Midland, MI) were diluted to a 2 mmol/l stock in water using HCl to bring the pH to 7.0 and stored in aliquots at −20 °C. RGD-PAMAM-G4 was prepared as follows: 1 ml of 0.2 mmol/l PAMAM-G4 in 25 mmol/l Hepes pH 7.9, 1 mmol/l EDTA, 150 mmol/l NaCl, and 1.2 mmol/l LC-SPDP (added last) was incubated for 2 hours at room temperature. After overnight dialysis with two changes of buffer against 25 mmol/l Tris pH 8.0, 1 mmol/l EDTA, and 150 mmol/l NaCl, a pyridine 2-thione release assay was done as per the manufacturer's instructions to determine the LC-SPDP conjugation efficiency. Then 23.2 µl of 25 mg/ml cRGD freshly dissolved in 3% acetic acid was added to 1.3 ml of the activated PAMAM-G4 and the reaction allowed to proceed overnight at 4 °C. A pyridine 2-thione release assay was done to measure cRGD coupling efficiency. After overnight dialysis with two changes of buffer against 1x PBS, the final samples were stored in aliquots at −80 °C. The same protocol was used to prepare RGD-PAMAM-G5.

Calculation of the RGD:PPIG5 ratio. The number of RGD's linked to PPIG5 was determined in several ways. First, the colorimetric bicinchoninic acid protein assay (Thermo Scientific Pierce) that measures reduction of Cu2+ to Cu1+ by the RGD peptide was used. Second, RGD-G5 and RGD-G5/SPDP underwent MALDI-MS analysis using as a matrix 2,4,6-Trihydroxyacetophenone (THAP) in 50% acetonitrile in 50 mg/ml ammonium acetate. Third, the LC-SPDP linker used to make RGD-G5/SPDP permits conjugation efficiency by monitoring pyridine 2-thione release at 343 nm as per the manufacturer's instructions. All of these methods led to close agreement that an average of two RGD's were coupled to each PPIG5.

Cell culture and transfection. C8161 and UACC903 cells were grown in RPMI medium supplemented with 10% fetal bovine serum and antibiotics. Cationic lipid-based transfection with LF2000 was done as previously described.[7] Transfection with RGD-G5 dendrimer was done as follows. For a 6-well plate, a 0.2 ml transfection mix containing 1x PBS pH 7.2 was prepared. RGD-G5 was diluted into 180 µl PBS and then the Dicer-substrate siRNA or U1 Adaptor was added to give a final volume of 0.2 ml in 1x PBS, and the solution gently mixed. After 5 minutes at room temperature, the RGD-G5:oligonucleotide complexes were added to cells that had been overlaid with 1.8 ml of fresh growth media. For larger or smaller scale transfections, the transfection mix was scaled accordingly. After the indicated time, cells were harvested and protein and/or RNA was extracted.

Figure S1. In vitro silencing activity of anti-BCL2 U1 Adaptors. Figure S2. In vitro silencing activity of variants of the BCL2-A Adaptor. Figure S3. In vitro dose response silencing activity of the BCL2-A Adaptor complexed with the RGD-G5 nanoparticle. Figure S4. In vitro silencing activity of various dendrimer:BCL2-A Adaptor complexes. Figure S5. Particle size of various dendrimer:BCL2-A Adaptor complexes. Figure S6. Analysis of BCL2-A activity in C8161 and UACC903 xenografts. Figure S7. In vitro silencing activity of anti-GRM1 U1 Adaptors and matching controls. Figure S8. Western blot analysis to assess the level of pAKT versus AKT in day 17 tumors from anti-GRM1 U1 Adaptor-treated mice. Figure S9. The anti-GRM1 Adaptors suppress growth of UACC903 xenografts and terminal PS bonds in the Adaptors are needed for full activity. Figure S10. U1 Adaptor treatment does not elicit an immune response at 4 hours. Figure S11. U1 Adaptor treatment does not elicit an immune response at 24 hours.