Targeting HPV‐infected cervical cancer cells with PEGylated liposomes encapsulating siRNA and the role of siRNA complexation with polyethylenimine

Abstract The greatest obstacle to clinical application of cancer gene therapy is lack of effective delivery tools. Gene delivery vehicles must protect against degradation, avoid immunogenic effects and prevent off target delivery which can cause harmful side effects. PEGylated liposomes have greatly improved tumor localization of small molecule drugs and are a promising tool for nucleic acid delivery as the polyethylene glycol (PEG) coating protects against immune recognition and blood clearance. In this study, small interfering RNA (siRNA) was fully encapsulated within PEGylated liposomes by complexing the siRNA with a cationic polymer, polyethyleneimine (PEI), before encapsulation. Formation methods and material compositions were then investigated for their effects on encapsulation. This technology was translated for protective delivery of siRNA designed for human papillomavirus (HPV) viral gene silencing and cervical cancer treatment. PEGylated liposomes encapsulating siRNA were functionalized with the AG86 targeting peptide‐amphiphile which binds to the α6β4 integrin, a cervical cancer biomarker. It was found that both targeting and polymer complexation before encapsulation were critical components to effective transfection.


| I N T R O D U C T I O N
Rapid advancements in genetic technologies have given researchers the ability to target and modify individual genetic events involved in disease progression, providing potential treatment avenues for previously untreatable diseases. 1 E6 and E7 bind to tumor suppressor proteins p53 and pRb, marking them for degradation or blocking their binding sites, thereby preventing apoptosis and driving cellular proliferation. 6,7 With the knowledge of the genetic mechanism of this oncovirus, several siRNA sequences targeting the gene sequences that encode the E6 and E7 proteins have been developed, demonstrating rescue of the p53 and pRb tumor suppression pathways, resulting in cell cycle arrest and apoptosis in HPV-infected cancer cells both in vitro and in vivo. [8][9][10][11][12] In order for the potential of clinical gene therapy to be realized, several key obstacles to efficient in vivo delivery need to be overcome.
For successful transfection and therapy to occur, siRNA must be internalized into the cells and released into the cytosol to mediate gene silencing.
While traversing the blood stream to reach the target tissue, siRNA must avoid degradation by nucleases, recognition by the immune system, and renal clearance. Several technologies have been developed to address each of these barriers to siRNA delivery, including chemical modification, nanoparticle complexation, and addition of targeting moieties. [1][2][3][4][5]13 In this study, we developed cancer gene therapy delivery vehicles composed of targeted PEGylated liposomes encapsulating siRNA. PEGylated liposomes are hollow, spherical phospholipid nanoparticles functionalized with a layer of PEG which have seen clinical success for the intravenous (IV) delivery of chemotherapeutic agents. [14][15][16] PEGylation has been shown to increase blood circulation, minimize immunogenicity and increase tumor accumulation of IV delivered liposomes. [14][15][16] PEGylated liposomes present a potential solution to the toxicities observed from traditional cationic siRNA transfection agents such as PEI and 1,2-dioleoyl-3-trimethylammonium-propane. [17][18][19] Liposomes functionalized with ligands designed to bind to upregulated surface receptors can enhance cellular association and internalization into cancer cells. The a 6 b 4 integrins are upregulated surface receptors associated with metastatic behavior in several cancer types, including cervical cancer. [20][21][22] The AG86 peptide was identified as an a 6 integrin binding ligand, 23 and was investigated here for specificity for the a 6 b 4 integrin and for targeting to HeLa cervical cancer cells. PEI complexation with nucleic acids alone has been shown to aid in endosomal escape through the proton sponge effect, whereby the high buffering capacity of PEI can cause osmotic swelling and rupture of intracellular organelles. [24][25][26] Previously, we have demonstrated successful encapsulation of plasmid DNA (pDNA) complexed with PEI within the aqueous core of PEGylated liposomes. 27 With the addition of a targeting ligand, liposome encapsulated PEI-complexed DNA achieved efficient transfection in colorectal cancer cells. 28,29 We therefore hypothesized that siRNA/PEI complexation could enhance transfection efficiency within PEGylated liposomes. We engineered AG86-functionalized PEGylated liposomes encapsulating PEI complexed siRNA ( Figure 1) as a delivery scheme to address each of the barriers to effective IV gene delivery. Optimal targeting and complexation properties of this vehicle were identified for successful gene silencing of the HPV-E7 gene in cervical cancer cells.
FIG URE 2 Binding and internalization of fluorescent targeted PEGylated liposomes. 0-10 mol% AG86-functionalized, calcein loaded, PEGylated liposomes were delivered at 100 lM lipids to HeLa cells for 3, 6, and 24 hr at 37 8C and binding and internalization was examined by lysing cells and measuring fluorescence. Data are presented as the mean 6 SE (n 5 3, performed in quadruplicate). All p-values from statistical analysis are listed in Supporting Information Table S1 FIG URE 1 Targeted PEGylated liposomes encapsulating siRNA/ PEI complexes. si18E7-674 shown to silence the HPV-E7 gene was complexed with PEI, encapsulated into AG86-functionalized PEGylated liposomes and delivered to a 6 b 4 -expressing HPV-18 containing HeLa cervical cancer cells FIG URE 3 Size (A) and zeta potential (B) measurements of siRNA/ PEI complexes. siRNA/PEI particles were complexed at various N:P ratios, and size and charge were determined. Data are presented as the mean 6 SE (n 5 4-11). * p < 0.01, ** p < 0.001 comparing zeta potential measurements. There was no significant statistical difference for pairs without brackets and 24 hr at 37 8C. As delivery time increased, binding and internalization increased for all targeting peptide concentrations. Increasing the concentration of the targeting peptide results in a nonlinear increase in binding and internalization (Figure 2), which is likely mediated by increased peptide valency resulting in binding avidity. 30,31 Liposomes prepared with 5 mol% AG86 achieved significantly more efficient delivery compared to 0, 2, and 3 mol% peptide at all times (Supporting Information Table S1). Although liposomes functionalized with 9 mol% peptide achieved the highest level of binding, more material is required for their production. 5 mol% peptide was therefore chosen as a sufficient targeting peptide concentration for subsequent gene delivery studies. To further verify that the AG86 peptide was responsible for liposome binding to HeLa cells, the binding of AG86-functionalized PEGylated liposomes was measured after incubation with free AG86 peptide and compared to binding without peptide blocking (Supporting Information Figure S1A). The presence of the free peptide decreased liposome binding by 98%, confirming AG86-mediated binding. The initial discovery of the AG86 peptide demonstrated specific interaction with the a 6 integrin. 23 The presence of a 6 antibodies disrupted 40% of cellular adhesion to AG86 peptide coated surfaces. 23 The a 6 integrin is known to dimerize with either the b 1 or the b 4 integrin, 32 and while blocking with a b 1 specific antibody demonstrated that AG86 binding is not specific for the a 6 b 1 heterodimer, binding to the b 4 integrin was not explored. 23 We therefore investigated the binding interactions of the AG86 peptide using antibody blocking of the AG86-functionalized PEGylated liposomes to cells (Supporting Information Figure S1B). In the presence of anti-a 6 and anti-b 4 integrin antibodies, liposome binding is decreased by 66% and 86%, respectively, verifying the a 6 b 4 integrin as the binding target of the AG86 peptide.

| Characterization of siRNA/PEI complexes
Before encapsulation in targeted liposomes, anionic siRNA was complexed with the cationic polymer PEI to form nanoparticles. siRNA was complexed at several different nitrogen:phosphate ratios (N:P) to investigate the effect of N:P ratio on particle size, charge, liposomal encapsulation yield and transfection efficiency. None of the particle sizes are significantly different between different N:P ratios ( Figure 3A). 33 Representative histograms for each N:P ratio are included in Supporting Information Figure S2. As N:P ratio was increased from 2 to 8 and more positively charged polymer was added during complexation, the zeta potential increased from 211 to 13 mV ( Figure 3B). The larger standard error of the average particle size for siRNA/PEI complexes at N:P 5 4 is indicative of the higher size polydispersity observed from these complexes, as is commonly seen at N:P ratios that produce particles approaching a neutral charge or at the transition from negative to positive zeta potential associated with increasing N:P. [34][35][36] 2.3 | Isothermal calorimetry (ITC) exploring siRNA/PEI complexation Microcalorimetric titrations of PEI into siRNA solutions were performed to monitor the thermodynamic properties associated with the formation of siRNA/PEI complexes. Titrations were performed over the entire range of N:P ratios investigated for transfection ( Figure 4A, C). The calorimetry results over this N:P range showed complete saturation of siRNA with PEI between N:P 1 and 2. Others have observed DNA/PEI saturation between N:P of 2-3 using branched PEI of similar size, 34,37,38 and between 1 and 2.5 for siRNA/PEI. 39 In order to better observe the transition from free to fully complexed siRNA, titrations were also performed spanning N:P ratios of 0-2 ( Figure 4B,D). A "one set of sites" model was used to calculate binding affinity (K 5 3.1 3  34,38 An n of 0.57 corresponds to an siRNA:PEI ratio of 25 and a negative to positive charge ratio of 1.7:1. This deviation from a charge ratio of 1:1 could be explained by a difference in linear intercharge spacing between siRNA (0.17 nm) 40 and PEI (0.25-0.35 nm). 41 Normalizing to the charge ratio, it is reasonable to expect the interaction of siRNA and PEI to resemble the interaction of DNA with PEI. The binding of siRNA and PEI in distilled water was characterized using similar thermodynamic parameters in the literature, however an n of 2.26 was identified, requiring more PEI molecules for condensation of 1 siRNA molecule. 39 This discrepancy could be caused by differing buffer conditions. 40 In Figure 4A   or uncomplexed siRNA (N:P 5 0) were delivered to HeLa cells and cell viability was assayed 24 hr later. As Figure 9 shows, the nanoparticles with the complexed siRNA decreased cell proliferation by 30%, while those encapsulating the uncomplexed siRNA had no effect on cell proliferation. 2.5 nM siRNA delivered using targeted PEGylated liposomes achieved 68% mRNA silencing, as shown in Figure 6, however protein expression and subsequent phenotypic effects are not quantitatively predicted by mRNA expression and may account for the observed 30% cell toxicity. 43,44 Our results are in agreement with previous findings where it was shown that silencing of the HPV-E7 gene using RNAi promoted 80-90% mRNA silencing and resulted in 40-60% inhibition of cell proliferation at early time points. 12,45 Therefore, continued doses of siRNA may be necessary to achieve more potent cytotoxicity effects. The cytotoxicity from the individual components of the targeted PEGylated liposomes encapsulating siRNA/PEI was also investigated. Empty targeted PEGylated liposomes and siRNA/PEI complexes of a non-specific sequence were delivered to the HeLa cells either free in solution or encapsulated in the targeted liposomes. As Supporting Information Figure S4 shows, none of the isolated components exhibited significant cytotoxicity compared to the untreated control, thus concluding that the toxicity observed in Figure 9 was the result of transfection of the si18E7-674 siRNA sequence.

| Cell apoptosis induced by different formulations
The oncogenic behavior exhibited by the E7 viral oncogene results from its ability to bind pRb, a tumor suppressor protein that participates in apoptosis and cell cycle regulation. 8,12,46 An apoptosis assay was therefore performed to investigate the role of apoptosis in the cytotoxicity observed from delivery of E7-specific siRNA encapsulated within targeted PEGylated liposomes as shown in Figure 9.  Figure 9 and Supporting Information Figure S4. The results also highlight our approach of using AG86-functionalized liposomes that bind to a 6 b 4expressing HeLa cells, along with the si18E7-674 that is shown to be specific for the HPV-18 containing HeLa cells, as the formulation had no effect on the HPV-negative C33A cells. Furthermore, the lack of apoptotic signal from the delivery of targeted liposomes that were either empty or loaded with a complexed control siRNA demonstrates the safety of our delivery system itself. of this lipidoid library were used to predict and design highly efficient siRNA delivery vehicles. The optimal PEGylated siRNA lipidoid nanoparticles achieved more than 95% protein silencing in vivo in hepatocytes. 52 Interestingly, surface pKa of a particular lipidoid was a critical parameter for predicting transfection efficiency for a lipidoid nanoparticle. 52 The SNALPs are a well-studied non-targeted lipid based delivery vehicle which has seen great success in preclinical studies. In one study, SNALPs were used to silence polo-like kinase 1 in tumors while abrogating activation of innate immune response and reducing tumor size by 75%. 53 Transferrin targeted cyclodextrin particles have also successfully delivered siRNA to tumors in vivo, and an accumulation/function study again revealed similar tissue accumulation for targeted and non-targeted nanoparticles, but enhanced transfection associated with targeting. 54 Preclinical results from SNALP and cyclodextran nanoparticle development motivated clinical trials for oncogene silencing and cancer treatment. 49 With these advances, many biodistribution challenges have been addressed, and often the key barrier lies in specific cellular uptake and appropriate intracellular release. 13,55 Previous work within our group has shown that pDNA condensed with PEI can be fully encapsulated within neutral PEGylated liposomes  42 It was found that there was no significant statistical difference in encapsulation yield between targeted PEGylated liposomes encapsulating siRNA/PEI complexed at different N:P ratios, while silencing increased with increasing N:P up to N:P 5 6. Others have also observed an increase in siRNA transfection using PEI with increasing N:P up to a certain ratio, followed by a decrease in transfection efficiency. The increase in transfection with increasing N:P ratio was attributed to higher uptake, and the decrease attributed to increasing cytotoxicity at higher N:P ratios. 24,33,36,39 Since binding and internalization of the targeted PEGylated liposomes is largely driven by the targeting peptide as shown before and in Figure   2, 29 and not by the presence of PEI in the liposomes, this explanation is insufficient in our case. Therefore, we speculated that the increase in siRNA silencing with an optimal N:P ratio of the encapsulated siRNA/ PEI was more likely caused by an optimization of the local buffering capacity within the liposome. [59][60][61] Since targeted PEGylated liposomes encapsulating siRNA/PEI complexes at N:P 5 6 showed no difference in binding and internalization compared to liposomes encapsulating uncomplexed siRNA (N:P 5 0), as shown in Figure 8, but achieved 3.7fold higher silencing efficiency (Figure 6), this indicates that the PEI complexation of siRNA in these gene delivery vehicles may have improved efficiency either through carrier release, endosomal escape, or a combination of both. PEI mediated endosomal release could minimize immunogenicity, as it has been found that the endosomal acidification process is crucial to siRNA induction of the interferon and cytokine response. 62 Active targeting has been shown to significantly improve transfection for several other gene delivery vehicles. 50

| C O NC LU S I O N S
A significant challenge for cancer therapy lies in the effective discrimination between healthy and tumor tissue. Targeting delivery vehicles to cancer biomarkers in order to improve delivery specificity has been explored in depth, however, off target delivery is a common limitation of this scheme. 69,70 Additional genetically mediated targeting could potentially overcome this obstacle by removing off target effects from the therapy itself. 28,71,72 Each aspect of the targeted PEGylated liposome encapsulating siRNA/PEI complexes, provided a significant deliv-

| Isothermal calorimetry (ITC)
siRNA and PEI were prepared in the same buffer to minimize mixing effects. PEI was then injected at 2 lL increments into the siRNA solution using MicroCal Auto-iTC 200 (Malvern Instruments, Malvern, UK) to measure the power required to maintain a constant chamber temperature. Using the MicroCal Auto-iTC200 software, the resulting power versus time data were integrated to determine heat exchange associated with each injection. The heat of dilution was accounted for by subtracting the heat data obtained from injections of PEI into buffer from the heat data obtained from injecting PEI into an siRNA solution.
The resulting heat curve was fitted to a "one set of sites" binding model, which uses n, the number of binding sites, K, the binding association constant, and DH, the enthalpy of binding, as fitting parameters where all n binding sites have the same K and DH. 38,40 The data were fitted to the following equations: where u is the fraction of sites occupied by the siRNA, X ½ is the free siRNA concentration, X t is the bulk siRNA concentration, M t is the bulk PEI concentration, Q is the total heat content of the solution, DQ i ð Þ is the heat released from the ith injection, V o is the volume of the chamber, and dV i is the injection volume at the ith injection.

| Liposome formation
where C t,ref1 and C t,ref2 are the threshold cycles of the individual reference genes.
HPV-E7 gene silencing was calculated using the following equations. First, the normalized sample threshold cycle, DC t , was determined by subtracting the reference gene threshold cycle (C t,ref ) from the target gene threshold cycle (C t,x ) for each sample: The difference between treated and untreated threshold cycles, DDC t , was then calculated by subtracting the DC t,untreated of the untreated sample from the DC t,treated of the treated sample: 2DDC t 52 DC t;treated 2DC t;untreated À Á The fold silencing that was achieved in the treated sample relative to the untreated sample was calculated as follows: were delivered and allowed to incubate for 24 hr at 37 8C and 5% CO 2 .

| Binding and internalization of siRNA liposomes
Cell viability was then measured using a WST-1 Cell Proliferation Reagent (Roche, Indianapolis, IN) following the manufacturer's protocol.
Absorbance was measured using a Synergy H1 microplate reader (Bio- tek, Winooski, VT). Cell viability was normalized to untreated cells.

| Statistics
ANOVA analysis and Tukey's honest significant difference (HSD) test were performed to calculate p-values and determine statistical significance between means. When only two means were compared within an experiment, student's t-test was used to calculate p-values.

ACKNOWLEDGMENTS
The flow cytometry analysis was performed in the University of

SUPPORTING INFORMATION
Additional Supporting Information can be found the online version of this article at the publisher's website. Table S1. p-values from ANOVA statistical analysis for Figure 1 showing binding and internalization of fluorescent liposomes with varying peptide concentration and incubation times at 37 8C.