Cytosolic delivery of siRNA by ultra-high affinity dsRNA binding proteins

Abstract Protein-based methods of siRNA delivery are capable of uniquely specific targeting, but are limited by technical challenges such as low potency or poor biophysical properties. Here, we engineered a series of ultra-high affinity siRNA binders based on the viral protein p19 and developed them into siRNA carriers targeted to the epidermal growth factor receptor (EGFR). Combined in trans with a previously described endosome-disrupting agent composed of the pore-forming protein Perfringolysin O (PFO), potent silencing was achieved in vitro with no detectable cytotoxicity. Despite concerns that excessively strong siRNA binding could prevent the discharge of siRNA from its carrier, higher affinity continually led to stronger silencing. We found that this improvement was due to both increased uptake of siRNA into the cell and improved pharmacodynamics inside the cell. Mathematical modeling predicted the existence of an affinity optimum that maximizes silencing, after which siRNA sequestration decreases potency. Our study characterizing the affinity dependence of silencing suggests that siRNA-carrier affinity can significantly affect the intracellular fate of siRNA and may serve as a handle for improving the efficiency of delivery. The two-agent delivery system presented here possesses notable biophysical properties and potency, and provide a platform for the cytosolic delivery of nucleic acids.

where NA is Avogadro's constant 4. The degradation rate of the siRNA-carrier complex was set to equal that of the more stable component (i.e., the more stable species protects the lesser stable species from degradation, as observed with p19 and siRNA (13)). Association rate of siRNA and carrier in the extracellular space

!"#
Dissociation rate of siRNA and carrier in the extracellular space Varied (part of simulation) Dissociation rate of siRNA and carrier in the endosome Set to equal !"" Dissociation rate of siRNA and carrier in the cytoplasm Set to equal !"" Association rate of siRNA and receptor-bound carrier in the extracellular space Set to equal !" Dissociation rate of siRNA and receptor-bound carrier in the extracellular space Set to equal !"" Dissociation rate of siRNA from the receptor-bound complex in the endosome Set to equal !"" Association rate of the carrier and receptor in the extracellular space 1.8 × 1. Measured experimentally unless otherwise noted. The extracellular space was modeled as a pH 7.4 environment, and the endosome pH 5.5.
2. Based on compartment pH. The affinity of p19 does not significantly differ between the two pHs (15). 3. As siRNA loading did not affect receptor binding by the targeted carrier (Figure 2b), we assumed that p19 and the EGFR binder function independently within the targeted carrier and that siRNA binding kinetics would not be affected by the carrier's association state to EGFR. The off-rates were estimated as described in Model Construction: Receptor-mediated uptake of siRNA.

Figure S1. Model equations
Model construction RNA interference. The model for RNA interference developed here was adapted from Bartlett et al. (3) to create a simplified representation. As with the former model, the total number of RISCs available to form activated RISCs was assumed to remain constant, and dissociation of activated RISC was chosen to be negligible (3). We additionally assumed that dissociation of the activated RISC/mRNA complex is negligible and that the target mRNA is immediately cleaved to liberate RISC. The association rate between activated RISC and mRNA was fitted using experimental data from the literature correlating the number of siRNAs in the cytoplasm and the resulting degree of knockdown ( Figure S2). Receptor-mediated uptake of siRNA. Receptor-mediated internalization implemented by Hackel et al. (16) was simplified to create a net-internalization model capturing the essence of receptor-mediated uptake. As the targeting Fn3 used in this study (clone E6) was demonstrated not to affect surface EGFR levels (16), we assumed that surface EGFR levels are maintained at steady-state throughout the simulation and binding of the targeted carrier does not affect the rate of internalization. Non-specific uptake of siRNA was negligible in the dynamic concentration range (Figure 3d), and receptor-bound carriers were allowed to capture free siRNA as observed on the yeast surface. To represent the media change that was performed after 6 hours during the transfection experiments, the concentrations of free species in the extracellular space were set to zero at 6 hours. This effectively removed the siRNA source in the system. To simplify model simulations, the association rate of the carriers were fixed at 1 × 10 5 M -1 s -1 (14), such that their affinity was varied by changing the dissociation rate only. With this onrate, an effective off-rate was fitted for p19-E18 using experimentally measured uptake kinetics in Figure 4a (Figure S3). The fitted off-rate closely matched experimentally reported values (17), although the effective affinity of p19 calculated in this manner was lower than measured values. We speculate that the strong temperature-dependence of p19's affinity or the presence of a nucleic acid competitor in the culture medium may have caused this discrepancy. Figure S3. Receptor-mediated uptake of the p19-E18/siRNA complex. Open circles represent experimental data, and solid lines represent simulation results.

Model construction (continued)
Endosomal release. A first-order rate of endosomal release was fitted using the uptake and silencing data of p19-E18. In the absence of experimental evidence suggesting otherwise, we assumed that the rate of escape is equal for all species. This assumption is consistent with the PFO pore (25-30 nm (18)) being larger than all three free species (6-8 nm; Figure 2c). Considering competition from multiple intracellular proteins, including nucleases, which contain binding domains for dsRNA (19), we assumed that siRNA rebinding to its carrier is unlikely once dissociated inside the cell. The resulting simulation dependably recapitulated the qualitative behavior of silencing ( Figure S4).   Table S5). Silencing simulated using these off-rates and experimental data (Figure 3a) were in good agreement, with the silencing potencies continually improving with higher affinity. Figure S6. Expression yields of affinity-matured p19 clones. The p19 clones were expressed in E. coli as sumo-fusions and purified as described in Materials and Methods. The yields reported here are from a single batch. Figure S7. Non-reduced SDS-PAGE of p19 fusion proteins. The expected molecular weight of the p19-E18 and p19-E6 monomers are 27 kDa and 31 kDa, respectively. Dimerization may occur when the embedded cysteine in p19 is exposed following denaturation.     Figure S11. Trypsinization removes surface-bound siRNA. Fluorescently labeled siRNA (Seq I) was loaded onto the carriers p19-E18 or p19 N15K,G16R -E18, and incubated with the adherent A431-d2EGFP cell line on ice to prevent endocytosis (1 hour at 20 nM). The incubation was performed either before or after trypsinizing the cells, followed by flow cytometry. Washes were performed between steps with cold PBSA. The trypsinization step completely removed surface-bound siRNA from both carriers (columns 1-2), which may be due to potential cleavage of p19. It is unknown whether trypsin stripped the surface-bound carriers as well. The carriers/siRNA complex bound to cells as expected if not exposed to trypsin (columns 3-4). Carriers with very high affinity ( k off SP = 10 −6 s −1 ) sequester siRNA in the cytoplasm, preventing the release of free siRNA and its subsequent incorporation into RISC. In contrast, carriers with very low affinity ( k off SP = 10 −2 s −1 ) are unable to internalize siRNA.

Expression and purification of p19
P19 expression. The p19, p19-E6 and p19-E18 constructs were cloned into the vector pE-SUMO (LifeSensors) and transformed into chemically competent Rosetta 2 (DE3) E. coli (Novagen). An overnight starter culture was diluted 1:100 into Terrific Broth (TB) containing antibiotics (50 mg/L kanamycin and 34 mg/L chloramphenicol) and incubated at 37°C for several hours. When the OD 600 reached 1.9 -2.1, the culture was induced with 1 mM IPTG and transferred to 20°C for overnight expression. Cell pellets were harvested by centrifugation at 5000 xg for 20 min at 4°C, then stored at -20°C until purification.
P19 purification. Bacterial pellets were resuspended in cold sonication buffer (2x PBS containing 3% glycerol, 1% triton X-100 and EDTA-free protease inhibitors (Roche), pH 7.5) and lysed by sonication on ice. Following centrifugation at 20,000 xg for 60 min, the supernatant was incubated with cobalt resin (Clontech) for 1 hr at 4°C. The resin was then transferred to a gravity flow column and washed with increasing concentrations of imidazole (2x PBS containing 5 mM, 10 mM and 15 mM imidazole, pH 7.5). A wash step with 2x PBS containing 1 M NaCl was included in-between the steps with 5 mM and 10 mM imidazole, to aid in the removal of non-specifically bound nucleic acids. Sumo-tagged p19 was eluted with 2x PBS containing 250 mM imidazole and concentrated to a smaller volume. The sumo tag was removed from p19 by overnight digestion with sumo protease (added at a 1:40 ratio by mass) at 4°C while dialyzing into 1x PBS. Cleaved sumo and sumo protease, both of which contain his tags, were removed by incubation with cobalt resin (Clontech) for 1 hr at 4°C. The flow-through containing cleaved p19 was collected and buffer-exchanged into 20 mM bis-tris containing 10 mM NaCl, pH 6.5. Anion exchange chromatography (AEX) was performed using a HiTrap Q HP anion exchange column (GE Healthcare Life Sciences) with an increasing salt gradient from 10 mM to 500 mM NaCl. A small peak absorbing strongly at 260 nm separated early, likely containing fugitive nucleic acids. The dominant peak eluting at approximately 200 mM NaCl was collected and concentrated for subsequent size-exclusion chromatography (SEC). Preparatory SEC was performed using a HiLoad 16/600 Superdex 75 pg column (GE Healthcare Life Sciences) in PBS, where p19 eluted mostly as a monomeric peak. Final samples in PBS were sterile-filtered then flash frozen in single-use aliquots, and stored at -80°C.

Parameter measurements
Steady-state GFP expression levels. The number of GFP molecules expressed in untreated A431-d2EGFP cells was calculated using AcGFP/EGFP flow cytometer calibration beads (Clontech) on an Accuri C6 cytometer (BD Accuri Cytometers). Measurements were performed during multiple passages, the median value of which was used.
Net internalization rate of EGFR. An EGFR binder engineered on the fibronectin scaffold (clone EI3.4.3 (16)) was labeled with Alexa Fluor 647 following manufacturer's instructions (Life Technologies). The labeled binder was incubated with A431 cells at 20 nM in complete media for 0 -25 hours in a reverse time course. Incubation was performed either in the presence or absence of excess competitor (unlabeled EI3.4.3) to determine the degree of non-specific uptake, if any. The background-subtracted fluorescence at each time point was fitted to obtain a firstorder rate of internalization.