Duplex End Breathing Determines Serum Stability and Intracellular Potency of siRNA-Au NPs

Abstract

Structural requirements of siRNA-functionalized gold nanoparticles (siRNA-Au NPs) for Dicer recognition and serum stability were studied. We show that the 3′ overhang on the nucleic acids of these particles is preferentially recognized by Dicer but also makes the siRNA duplexes more susceptible to non-specific serum degradation. Dicer and serum nucleases show lower preference for blunt duplexes as opposed to those with 3′ overhangs. Importantly, gold nanoparticles functionalized with blunt duplexes with relatively less thermal breathing are up to 15 times more stable against serum degradation without compromising Dicer recognition. This increased stability leads to a 300% increase in cellular uptake of siRNA-Au NPs and improved gene knockdown.

Introduction

RNA interference (RNAi) is a potent gene regulatory mechanism that leads to the inhibition of target genes in a sequence dependent manner.¹ This process is triggered by double stranded RNAs, known as small interfering RNAs (siRNAs),² that are recognized by the RNA-induced Silencing Complex (RISC). Dicer, an endoribonuclease, binds and cleaves long dsRNA³ to generate siRNA, 18-22bp in length. RISC then binds to the siRNA leading to passenger strand cleavage.⁴ The complementary strand, known as the guide strand, is used by the active RISC to find the target sequence along the messenger RNA (mRNA) by complementary base interactions. Finally, RISC acts as an endonuclease to cleave the target mRNA, preventing its translation into the corresponding protein. This mechanism of selectively controlling protein expression has sparked interest in harnessing RNAi for disease management and treatment by delivering siRNAs targeting specific proteins of interest to cells.

However, the delivery of small RNAs to cells must overcome several challenges to be effective as a therapeutic. The RNA must be stable against extracellular nucleases which non-specifically degrade small RNA sequences.^5-7 It must be able to enter the cells without the aid of toxic agents or untenable techniques, such as electroporation⁵ and microinjection.⁸ Finally, it must be able to participate in the natural pathway of RNAi. Unmodified siRNAs degrade within several minutes while in blood and serum.^9-10 Numerous modifications to the phosphate backbone and the 2′ position of the sugar have been used to stabilize RNA against non-specific nucleases.^11-12 However, these modifications can have negative side effects, such as non-specific protein binding and increased toxicity,¹³ decrease participation in the RNAi pathway and intracellular activity,¹⁴ and have unknown long term effects. Cellular delivery of siRNA has been realized with complexation/conjugation with auxiliary chemical reagents^15-17 or via electroporation.¹⁸ However, these techniques lead to undesired toxicity and are difficult to translate to more complex animal model organisms and clinical settings.¹⁷ Nonetheless, recent reports using lipid complexation have shown improved stability¹⁹ resulting in in vivo applications in animals^20-21 and humans.²²

Nanomaterials are becoming promising constructs for drug delivery, disease detection, molecular diagnostics and intracellular gene regulation.^23-35 In 1996, we introduced the concept of the polyvalent nucleic acid nanoparticle conjugate,³⁶ and subsequently identified several unique properties associated with it^37-40 and developed it for many in vitro molecular diagnostic^41-43 and intracellular gene regulation applications.⁴⁴ In 2009, we devised a procedure for synthesizing polyvalent siRNA functionalized gold nanoparticles (siRNA-Au NPs) and explored their utility for intracellular gene regulation.⁴⁵ Through these studies, we have discovered that these hybrid nanomaterials have unique composite properties that make them attractive for biological applications.⁴⁶ They have been shown to enter cultured cells (over 50 different cell lines) without the need to use transfection agents, such as cationic lipids. They stabilize the conjugated siRNAs against degradation by serum nucleases without any chemical modifications of the siRNA.⁴⁵ They exhibit significantly less induction of innate immune response compared to their molecular counterparts.³⁷ While nucleic acids and siRNAs, in particular, have been conjugated to gold nanoparticles^47-48 or quantum dots,^{18, 49} detailed studies of the structural features that impart nuclease stability and Dicer recognition on nanoparticles have not been conducted. We have conducted such a study that allows us to determine the features of siRNA-Au NPs that are critical to their stability in serum and recognition by Dicer.

Experimental Section

Synthesis of gold nanoparticles and RNA

Citrate stabilized gold nanoparticles (13 ± 1 nm) were prepared using the Frens method⁵⁰ and rendered RNase-free by treatment with 0.1% diethylpyrocarbonate (DEPC, Sigma) followed by autoclaving to inactivate DEPC. RNA oligonucleotides were synthesized using TOM - RNA reagents (Glen Research) on a MerMade 6 system (Bioautomation) using manufacturer recommended cleavage and deprotection protocols. All oligonucleotides were purified using reverse-phase high performance liquid chromatography (RP-HPLC) on a Varian Microsorb C₁₈ column (10 μm, 300 × 10 mm) with 0.1 M triethylammonium acetate (TEAA), pH 7 using a 1%/min gradient of 100% CH₃CN at a flow rate of 3 mL/min, while monitoring the UV signal of the nucleic acids at 254 nm. The aqueous buffer was DEPC treated (0.1% v/v) and autoclaved before use. After purification, the oligonucleotides were lyophilized and stored at −80° C until use. The following sequences that target enhanced green fluorescent protein (eGFP) mRNA were made (all listed 5′ to 3′): Phosphate - ACC CUG AAG UUC AUC UGC ACC ACC G (Hexaethyleneglycol)₂ – Propyl thiol, CGG UGG UGC AGA UGA ACU UCA GGG UCA, Fluorescein - CGG UGG UGC AGA UGA ACU UCA GGG UCA, CGG UGG UGC AGA UGA ACU UCA GGG U, Fluorescein - CGG UGG UGC AGA UGA ACU UCA GGG U, CGG UGG UGC AGA UGA ACU UCA GGG UCA, Fluorescein - CGG UGG UGC AGA UGA ACU UCA GGG UCA, CGG UGG UGC AGA UGA ACU UCA GGG U*C*A, Fluorescein - CGG UGG UGC AGA UGA ACU UCA GGG U*C*A, CGG UGG UGC AGA UGA ACU UCA GGG U*C*A, Fluorescein - CGG UGG UGC AGA UGA ACU UCA GGG U*C*A, Phosphate - CGG UGG UGC AGA UGA ACU UCA GGG U (Hexaethyleneglycol)₂ – Propyl thiol, ACC CUG AAG UUC AUC UGC ACC ACC G, Fluorescein - ACC CUG AAG UUC AUC UGC ACC ACC G. Bold letters indicate Locked Nucleic Acid (Exiqon) residues and * indicates phosphorothioate substitution.

Synthesis of siRNA-Au NPs and RNA loading measurements

Gold nanoparticles were functionalized with siRNA duplexes as described previously.⁴⁵ To measure the surface coverage of RNA on nanoparticles, the concentration of fluorescein-labeled siRNA functionalized nanoparticles was determined by measuring the intensity of the surface plasmon resonance band at 524nm by UV-Vis spectroscopy and using the Beer-Lambert law. The nanoparticles were then dissolved with potassium cyanide (5 mM final), and the fluorescence of the sample was measured on a FluoDia T70 plate reader (Photon Technology International). This fluorescence was compared to a standard curve generated using the same fluorescein-labeled RNA sequence. The measured concentration of RNA was divided by the measured concentration of nanoparticle to derive the average RNA loading per nanoparticle.

Dicer activity and nuclease assays

Dicer activity assays were performed using published protocols in a semi-clear 96-well, PCR plate using a FluoDia T70 plate reader (Photon Technology International).⁴⁰ Briefly, fluorescein containing siRNA-Au NPs (5 nM final) were mixed with reaction buffer⁵¹ (20% glycerol, 60mM Tris-HCl, 10mM MgCl₂ and 0.5mM DTT, 1mM ATP) and allowed to equilibrate at 37°C. Recombinant human Dicer (Invitrogen) was added to the assay well, followed by 40 μL of mineral oil (Bio-Rad) layered on top to prevent evaporation. Degradation of the RNA by Dicer was monitored by measuring an increase in the fluorescence signal as the fluorescein (excitation = 480 nm, emission = 530 nm) was dissociated from the gold nanoparticles. Stability of siRNA-Au NPs against serum degradation was measured under identical assay conditions with heat inactivated fetal bovine serum (FBS, 10% final, Mediatech) substituted for Dicer. The reactions with Dicer and 10% FBS were monitored for four hours at 30 second intervals, and the activity rates were determined by calculating the slope of the fluorescence curve from the first 30 points. All substrates were tested at least three times in each assay to ascertain variability. The amount of substrate catalyzed was calculated by first determining the fluorescence signal generated from complete dehybridization of the fluorescein-labeled RNA from a 5nM solution of Au NPs and setting this level of fluorescence to the maximum level of fluorescence possible for a 5nM solution of the substrate (fluorescein modified siRNA-Au NPs). Fluorescence level in buffer alone was assigned the value of 0 nM substrate catalyzed. These two values were used to construct a linear plot for comparison with the unknowns.

Measuring knockdown by real time polymerase chain reaction and eGFP fluorescence

Mouse endothelial cells stably expressing eGFP (C166-eGFP) were purchased from ATCC and cultured as recommended. For a typical experiment, cells were grown in a tissue culture treated plate (Corning) at 5000 cells/well and various siRNA-Au NPs were added to the cells in serum containing media at a solution concentration of 10nM. After three days of incubation, the cells were washed twice with phosphate buffered saline (Hyclone), trypsinized and counted using the Viacount reagent (Millipore) on a Guava Easycyte Mini. The cells were processed with the Cell-to-cDNA™ kit (Ambion) and the relative levels of eGFP mRNA were quantified using the SYBR Green rt-PCR mix (Roche) on a Lightcycler 2.0 system (Roche). ApoB was used as an internal control. The expression levels of eGFP and percentage knockdown were calculated using comparative Ct method. Primers used for rt-PCR were GFP forward (5′ CCA CAT GAA GCA GCA CGA CTT 3′), GFP reverse (5′ GGT GCG CTC CTG GAC GTA 3′), ApoB forward (5′ CAC GTG GGC TCC AGC ATT 3′) and ApoB reverse (5′ TCA CCA GTC ATT TCT GCC TTT G 3′). All primers were purchased from IDT DNA Technologies. Knockdown of eGFP protein was measured by fluorescence quantification at room temperature using a published protocol.⁴⁴ Statistical analyses between individual groups were performed by a Student’s t-Test; a P value of < 0.05 was considered statistically significant.

Cellular uptake studies

The average number of gold nanoparticles per cell for each siRNA-Au NP substrate was determined using Inductively coupled plasma mass spectrometry (ICP-MS, Thermo-Fisher) following published protocols.⁵² Briefly, C166-GFP cells were cultured in the presence of 10 nM solution concentration of siRNA-Au NPs. After 72 hours, the cells were washed with PBS, trypsinized and counted. The cells were centrifuged (750 rcf, 10 minutes) to remove the media, resuspended in a solution of 3% hydrochloric acid in nitric acid, and incubated at 60°C overnight to digest the samples. The digests were diluted in a matrix consisting of 2% HNO₃, 2% HCl and 1 ppb Indium (internal standard). The gold content of the cell digests was determined by comparison with a standard curve (0 to 200ppb Au) made from a gold standard (Sigma) prepared in the same matrix. The number of nanoparticles in each sample was calculated based on the concentration of Au found in the sample.

Results

Dicer and serum nucleases show preference for the 3′ 2-base overhang of the antisense strand

To investigate the structural basis of serum stability and Dicer activation on nanoparticles, we examined anti-eGFP Dicer substrate RNAs of 25-27 base pairs^53-54 immobilized on 13 nm gold nanoparticles using a fluorescence recovery assay.⁴⁰ Gold nanoparticles were functionalized with fluorescein-labeled siRNA (siRNA-Au NPs) following published procedures,⁴⁵ and RNA loading per nanoparticle was measured. All samples were found to have 50±4 duplexes per nanoparticle. To measure Dicer activity or serum stability, recombinant human Dicer (1 unit) or FBS (10% final) was added to a solution containing 5nM of siRNA-Au NP substrates in the reaction at 37°C, and the fluorescence increase was monitored (Figure 1). Initially, we tested two siRNA-Au NP substrates, one with a 2-bp overhang on the 3′ terminus of the antisense strand (termed overhang substrate) to mimic the widely used molecular siRNAs and one with a blunt end (termed blunt A-U substrate) (Scheme 1). The overhang contains C and A in the penultimate and terminal positions to maximize Dicer recognition.⁵⁵ The sense strand with a 5′ phosphate group, an ethylene glycol spacer, and a 3′ propylthiol modification was used to attach the duplexes to gold nanoparticles.

Figure 1.

Analysis of siRNA-Au NPs in enzymatic assays with Dicer (A) and 10% FBS (B). In (A), recombinant Dicer shows a preference for the siRNA-Au NP substrates with a 3′ overhang. Substrates with modifications on the 3′ overhang (phosphorothioate, LNA or both) or no overhang show lower Dicer activity. In (B), non-specific degradation in 10% FBS is higher for siRNA-Au NPs containing single-stranded overhangs. Modifications to the overhang region decrease the non-specific degradation slightly. Highest stability against non-specific degradation is imparted by blunt duplexes with distal, thermally stable ends.

Scheme 1.

(A) Sequences and arrangement of siRNA substrate on the nanoparticles used in this study. In some experiments, the 2-bp overhang (in red) was modified by addition of phosphorothioate linkages, LNA phosphoramidites or both and is indicated as overhang P/S, overhang LNA or Overhang LNA + P/S, respectively. In all cases, the sense strand is shown on top. (B) Scheme of a typical Dicer or serum nuclease activity assay. Gold nanoparticles functionalized with fluorophore-labeled siRNAs are incubated in the presence of nucleases at 37°C. The change in fluorescence is measured over time to assess the activity of the nucleases (see Experimental Section).

The results of these experiments demonstrate that Dicer has a preference for siRNA duplexes with 3′ overhangs on the RNA-Au NP (Figure 1). The catalysis rate was approximately twice as fast for the overhang substrates compared with blunt A-U substrates (1.23×10⁻³ nM/min and 5.71×10⁻⁴ nM/min for overhang and blunt A-U substrates respectively, Table 1). These results confirm the studies from the molecular system^55-56 on this enzyme. Interestingly, the overhang substrates were also approximately 40% more susceptible to non-specific degradation in serum compared to the blunt A-U substrates (1.16×10⁻² nM/min vs 8.43×10⁻³ nM/min for overhang vs blunt substrates). Importantly, for siRNA-Au NPs, the non-specific activity from general nucleases was 10 - 15 fold higher than the specific activity from Dicer (Table 1). These results show that the 2-base overhang is responsible for increasing the Dicer recognition of siRNA-Au NPs. However, the same structural feature also makes the siRNA-Au NPs more susceptible to non-specific degradation, presumably due to activity of RNAse A family or similar processive exonuclease.⁵⁷

Table 1.

The rates of catalysis of double-stranded RNA by Dicer and 10% FBS at 37°C. The rates were determined by taking the slope of the activity curves over the first 15 minutes of the fluorescence recovery assay described in the Experimental Section.

	Dicer Activity (nM substrate catalyzed/min)		FBS Activity(nM substrate catalyzed/min)		FBS/ Dicer
	Rate	Std Dev	Rate	Std Dev
Overhang	1.23E-03	1.06E-04	1.16E-02	4.36E-04	9.43
Blunt A-U	5.71E-04	3.79E-05	8.43E-03	1.94E-04	14.77
Overhang P/S	7.68E-04	8.66E-05	7.82E-03	3.64E-04	10.18
Overhang LNA	6.82E-04	2.91E-05	7.80E-03	1.58E-04	11.44
Overhang LNA + P/S	4.61E-04	3.81E-05	7.37E-03	2.82E-04	15.99
Blunt G-C	6.24E-04	1.98E-04	4.99E-04	7.12E-05	0.80

Chemical modifications to the overhang region lower dicer processing and serum degradation

Serum stability is a desirable trait of any siRNA conjugate. Since the 2-base overhang made siRNA-Au NPs more susceptible to non-specific degradation, we hypothesized that by modifying the overhang positions in an attempt to stabilize them, we can increase the serum stability of siRNA-Au NP substrates. While specialty oligonucleotides are available and used by researchers,^{9, 11-12, 19, 58-65} we chose locked nucleic acids (LNA) and structures with phosphorothioates as representative modifications of the pentose sugar and phosphate backbone respectively. The use of LNA bases or phosphorothioate modification has been shown to increase stability of RNA to non-specific degradation.^{61, 66} Researchers have suggested that some modifications, such as phosphorothioate, improve nuclease resistance of siRNA by promoting secondary structures that affect recognition or binding by nucleases.⁶⁷ Additionally, modest LNA and phosphorothioate substitutions have been shown to be compatible with RNAi.^{14, 58, 64, 68} When we modified the 2-base overhang region by changing the terminal two bases of the antisense strand to LNA, inserting phosphorothioate linkages at the 3′ terminus or both, the stability of these substrates against non-specific degradation increased modestly (Figure 1, Table 1), from 1.16×10⁻² nM/min for the unmodified substrate to 7.37×10⁻³ nM/min for substrates containing both LNA and phosphorothioate modifications in the overhang, a 40% increase in serum stability. However, their ability to be processed by Dicer decreased by 60% (1.23×10⁻³ nM/min for the unmodified substrate vs 4.61×10⁻⁴ nM/min substrates containing both LNA and phosphorothioate modifications), indicating that modifications to the antisense strand can affect Dicer recognition and processing. Interestingly, the non-specific nuclease activity remained 10-15 fold higher than Dicer activity meaning that we were trading stability for specificity. These experiments show that modifying the overhang positions had similar effects on the ability of non-specific nucleases and Dicer to recognize and process siRNA-Au NPs.

Serum stability is increased by thermal stability of the distal end of siRNA duplex

We have previously reported that high oligonucleotide density and high local salt concentration near the nanoparticle surface⁴⁰ impart serum stability to oligonucleotides. However, for siRNA-Au NPs, we sought to improve the serum stability while keeping the relative Dicer specificity high. It is known that single-stranded RNA is less stable than double stranded RNA^{12, 65} and that single stranded regions of RNA are potential substrates for processive enzymes of RNAse A family of exonucleases.^{57, 69} In the blunt A-U substrates, the duplex breathing at the solution end could generate single-stranded regions that play a role in the susceptibility of siRNAs to nuclease degradation. For the blunt duplex that we studied, the binding energies (37 °C and 10mM Mg²⁺) of the last five bases of the two ends were computationally predicted by mfold⁷⁰ to be ΔG = −7.4 kcal/mol (G-C end) and −6.1 kcal/mol (A-U end) and the T_ms of the same regions were estimated to be 12.1 °C and <10 °C respectively. We hypothesized that if we reversed the orientation of the duplex on the nanoparticle so that the more stable end was distal to the nanoparticles and therefore interacted with proteins, the serum stability might improve. Indeed, when we changed the orientation of the siRNA on nanoparticles so that the end with a G-C base pair (termed blunt G-C) was presented to the solution and thus enzymes, the serum degradation rate decreased to 4.99×10⁻⁴ nM/min compared with 7.37×10⁻³ nM/min for blunt A-U substrates, a 93% decrease in serum degradation (Figure 1). This same substrate was still recognized by Dicer (6.24×10⁻⁴ nM/min) to an extent similar to the blunt A-U substrate. Importantly, the non-specific nuclease activity was now 25% lower than Dicer activity, compared to 10-15 fold higher in previous substrates (Table 1), suggesting that we had “turned down” the non-specific nucleases without substantially affecting Dicer activity.

The results of these experiments show that in polyvalent siRNA-Au NPs, the stability of the part of the duplex that is oriented away from nanoparticle surface and into the solution plays a key role in its nuclease resistance. In the context of siRNA-Au NPs, the distal or solution end of duplex is more critical for protein binding because the proximal end is essentially inaccessible to proteins due to steric hindrance resulting from nanoparticle functionalization. Of the two blunt siRNA-Au NP conjugates, the conjugates with the more stable distal ends showed 15 fold higher resistance to serum nucleases. The same duplexes, when free in solution, were degraded equally rapidly by 10% FBS (>80% degraded within 15 minutes, data not shown). Importantly, this arrangement maintains Dicer recognition and processing. While we have previously observed that the nanoparticle functionalization can increase the nuclease resistance of oligonucleotides compared to free oligonucleotides,^{40, 45} this is the first study that shows that strand orientation on nanoparticles, and in particular end breathing, has significant consequences on Dicer recognition and selective nuclease degradation.

Increased stability in serum results in higher intracellular activity

Having designed and prepared a number of siRNA-Au NP substrates with different serum degradation rates, we next studied their ability to knockdown eGFP in cell cultures. C166-GFP cells were treated with the three types of siRNA-Au NPs targeting eGFP mRNA, and control nanoparticles functionalized with non-specific RNA sequence, respectively, in culture media containing 10% FBS at 37°C. Three days following the nanoparticle treatment, the cells were analyzed for eGFP mRNA using real time polymerase chain reaction (rt-PCR) and for eGFP protein using fluorescence protein assays.⁴⁴ They were compared with cells that had not been treated with nanoparticles. We hypothesized that if the siRNA-Au NPs are more stable during the course of the experiment, then improved knockdown may be observed. Indeed, the results of the rt-PCR analyses indicate that the siRNA-Au NPs with the highest stability against serum degradation also showed the highest eGFP knockdown (Figure 2A and 2B). For example, cellular treatment with the nanoparticles functionalized with the overhang substrates resulted in GFP knockdown of 65.6±4.0%, whereas the treatment with nanoparticles functionalized with blunt G-C siRNA resulted in GFP knockdown of 85.5±4.8% as determined by rt-PCR. The results of the protein assay confirmed these findings.

Figure 2.

Quantification of eGFP knockdown in mouse endothelial cells using real-time polymerase chain reaction (rt-PCR) (A) and fluorescent protein assay (B). In both assays, the siRNA-Au NPs that were most stable against serum degradation showed the highest knockdown. The results are average of three independent experiments. In both assays, the results from blunt G-C substrates were found to be statistically significant (p<0.05) in Student’s t-test compared with all the other GFP siRNAs tested.

Additionally, when we measured the cellular uptake of these siRNA-Au NPs, the conjugates with highest stability were also internalized to the greatest extent (Figure 3). For example, the siRNA-Au NPs containing overhang substrates, modified overhang substrates, or blunt A-U substrates had cellular uptake of 6.90 × 10⁵ ± 1.2 × 10⁵, 7.2 × 10⁵ ± 1.8 × 10⁵, and 6.98 × 10⁵ ± 1.1 × 10⁵ nanoparticles/cell, respectively. However, the siRNA-Au NPs functionalized with blunt G-C substrates contained 2.98 × 10⁶ ± 1.90 × 10⁵ nanoparticles/cell, a 300% increase. Significant changes in uptake were not observed for the siRNA-Au NPs containing overhang substrates, modified overhang substrates, or blunt A-U substrates, because the difference in their serum stabilities is relatively small (< 2-fold). The nanoparticle conjugates that showed the highest uptake were 15-20 times more stable compared to the other particles, as determined by serum nuclease activity rate. These results are consistent with our previous observation that increasing the density of oligonucleotides on gold nanoparticles increases their cellular uptake.⁵² In the case of siRNA-Au NPs, as serum nucleases degrade the siRNA, fewer oligonucleotides remain on the particles, resulting in structures that are less likely to be internalized by cells. However, when the siRNA is stabilized on the nanoparticles, the polyvalent structure remains intact over the course of the experiment. This increases the likelihood of binding with cellular proteins, such as scavenger receptors, that mediate cellular uptake⁷¹ thus leading to higher degrees of cellular internalization. The resulting increase in the intracellular concentration of active siRNA-Au NPs results in higher levels of gene knockdown.

Figure 3.

Cellular uptake of siRNA-Au NPs in C166-GFP cells. The data show that the siRNA-Au NPs with highest stability against serum degradation exhibited highest cellular uptake. The high uptake correlates with higher knockdown of target protein.

Discussion

In summary, we have shown that Dicer processing and serum stability of siRNA-Au NPs can vary dramatically with minor changes in nucleic acid structural features, such as distal 3′ overhangs. The serum stability can be increased by as much as 15-fold for siRNA structures with blunt ends of high thermal stability and low duplex breathing. This structural change leads to a modest 2-fold decrease in Dicer recognition and processing. However, the increased serum stability enhances cellular uptake by 300% and thus leads to greater gene knockdown. The observations reported in this article will allow researchers working with nanoparticle (and polyvalent nucleic-acid) -based siRNA delivery methods to improve their substrate design and cellular effects.