Assessing the Heterogeneity Level in Lipid Nanoparticles for siRNA

Dec 4, 2012 - Nanomaterial Syntheses. R. Tantra , K.N. Robinson , J.C. Jarman , T. Sainsbury. 2016,25-48. Simultaneous Analytical and Purification Met...
0 downloads 0 Views 379KB Size
Download PDF
Article pubs.acs.org/molecularpharmaceutics

Assessing the Heterogeneity Level in Lipid Nanoparticles for siRNA Delivery: Size-Based Separation, Compositional Heterogeneity, and Impact on Bioperformance Jingtao Zhang,*,†,‡ Yi Pei,‡ Hangchun Zhang,‡ Lei Wang,†,‡ Leticia Arrington,†,‡ Ye Zhang,†,‡ Angela Glass,†,‡ and Anthony M. Leone†,‡ †

Department of Pharmaceutical Sciences and ‡Department of RNA Therapeutics, Merck Research Laboratories, Merck & Co., Inc., West Point, Pennsylvania 19486, United States S Supporting Information *

ABSTRACT: A primary consideration when developing lipid nanoparticle (LNP) based small interfering RNA (siRNA) therapeutics is formulation polydispersity or heterogeneity. The level of heterogeneity of physicochemical properties within a pharmaceutical batch could greatly affect the bioperformance, quality, and ability of a manufacturer to consistently control and reproduce the formulations. This article studied the heterogeneity in the size, composition, and in vitro performance of siRNA containing LNPs, by conducting preparative scale fractionation using a sephacryl S-1000 based size-exclusion chromatography (SEC) method. Eight LNPs with size in the range of 60−190 nm were first evaluated by the SEC method for size polydispersity characterization, and it was found that LNPs in the range of 60−150 nm could be well-resolved. Two LNPs (LNP A and LNP B) with similar bulk properties were fractionated, and fractions were studied in-depth for potential presence of polydispersity in size, composition, and in vitro silencing, as well as cytotoxicity. LNP A was deemed to be monodisperse following results of a semipreparative SEC fractionation that showed similar size, chemical composition, in vitro silencing activity, and cytotoxicity across the fractions. Therefore, LNP A represents a relatively homogeneous formulation and offers less of a challenge in its pharmaceutical development. In contrast, LNP B fractions were shown to be significantly more polydisperse in size distribution. Interestingly, LNP B SEC fractions also exhibited profound compositional variations (e.g., 5 fold difference in N/P ratio and 3 fold difference in lipid composition) along with up to 40 fold differences in the in vitro silencing activity. The impact of LNP size and formulation composition on in vitro performance is also discussed. The present results demonstrate the complexity and potential for presence of heterogeneity in LNP-based siRNA drug products. This underscores the need for tools that yield a detailed characterization of LNP formulations. This capability in tandem with the pursuit of improved formulation and process design can lead to more facile development of LNP-based siRNA pharmaceuticals of higher quality. KEYWORDS: siRNA delivery, lipid nanoparticle, heterogeneity, polydispersity, size-exclusion chromatography, fractionation, nanomedicine development

■

tries.5−7 In contrast to its rapid application in basic research, the development of siRNA therapeutics, however, has been less straightforward due to the difficulty in delivering siRNA to target tissues safely and efficaciously through systemic administration. Many carrier systems have been adopted for delivering siRNA, among which ionizable lipid nanoparticles (LNPs) are showing great promise.8,9 Ionizable LNPs are self-assembled nanoparticles consisting of siRNA and several functional lipids, which include ionizable amino lipids, neutral phospholipids,

INTRODUCTION The discovery of RNA interference (RNAi) for facile gene silencing in conjunction with the identification of vast gene targets from human genome sequencing offers tremendous opportunity for biomedical research, drug discovery, and therapeutics development.1,2 The initial application of RNAi is in the field of biological research and drug discovery, as evidenced by the rapid adoption of RNAi as an important research tool in understanding gene function, identifying, and validating drug targets.3,4 Due to their ability to address broad genome space that is inaccessible to small molecule or biologic drugs, small interfering RNA (siRNA), an important RNA molecule capable of triggering RNAi mechanism, was widely anticipated to become a new class of therapeutic modality that could revolutionize the pharmaceutical and biotech indus© 2012 American Chemical Society

Received: Revised: Accepted: Published: 397

September 19, 2012 November 15, 2012 December 4, 2012 December 4, 2012 dx.doi.org/10.1021/mp3005337 | Mol. Pharmaceutics 2013, 10, 397−405

Molecular Pharmaceutics

Article

and poly(ethylene glycol) (PEG)-lipids.10 Through formulation optimization and extensive structure−activity relationship (SAR) study of ionizable amino lipids, potency of LNPs for siRNA delivery has been improved to an unprecedented level while tolerability is generally well-maintained.11,12 The increased understanding of LNPs enables the clinical evaluation of these nanomaterials in diseases such as hypercholesterolemia, transthyretin-mediated amyloidosis, and liver cancer.13 Despite progress, the field could still advance from a more complete understanding of the origin of toxicity,14,15 the mechanism of action,16 and the relationship of these pharmacologic properties to product physiochemical attributes (e.g., polydispersity in size, morphology, internal composition, surface properties, etc.). In particular, the lack of understanding and, to a greater extent, the lack of control in the heterogeneity of pharmaceutical products can cause significant problems in maintaining the quality and consistency of a product. Furthermore, heterogeneity also raises concern about the existence of subpopulations that might dominate the activity or toxicity,17−19 thereby complicating the interpretation of any pharmacology or toxicology study. siRNA containing LNPs are commonly observed to have a range of distribution in particle size and morphology.20 Our goal was to thoroughly characterize LNPs to better understand size and compositional polydispersity. We also aimed further to evaluate the impact of subpopulations on bioperformance. Due to its cationic, hydrophobic, and fragile nature (a soft selfassembled lipid particle) and relatively large size (∼100 nm), physical characterization of LNPs without disruption of their structure is challenging. Consequently, most studies have been focused on the low-resolution bulk measurements that provide little understanding on the heterogeneity in the formulations. Several past studies were conducted for the separation and characterization of DNA-containing cationic lipoplex particles, revealing poorly formed particles and heterogeneity in sizes.17−19,21,22 Recently, we reported a high-resolution sizeexclusion chromatography (SEC) method with multiple detections that can be used to assess the LNP distribution in size, molecular weight, and siRNA loading amount.23 However, most of the above approaches are restricted to the analytical scale. Therefore, they may not be readily scaled-up for a preparative-scale separation to enable the product development and the in vitro and in vivo characterization of fractions. SEC is a commonly used approach to assess the size polydispersity in polymers and protein therapeutics.24 It separates based on hydrodynamic diameter, and the flow path differences that result from the size-dependent entry of analytes into the pores of stationary phases. Due to the ease of operation and scale-up, it was widely used for the analysis and preparative purification of protein therapeutics. In recent years, it was also used for the separation and characterization of neutral liposome particles.25−28 Here, we developed a SEC-based semipreparative scale separation method for siRNA containing LNPs. Using this method, we fractionated two representative LNPs (A and B) and studied the heterogeneity in the size, composition, and in vitro silencing activity, as well as cytotoxicity of these nanoparticles. We demonstrated that, although both LNPs were assembled by the same process and have similar bulk properties, they differed greatly in size and compositional distribution after fractionation. Further in vitro silencing activity and cytotoxicity test of LNP A and B SEC fractions also revealed significant differences among fractions. To the best of the authors’ knowledge, this is the first report describing the

presence of compositional and bioperformance heterogeneity in siRNA containing LNPs.

■

EXPERIMENTAL SECTION Materials and General Methods. Amino lipids and siRNA used in the study were synthesized in Merck Research Laboratories. The duplex siRNAs used in this study target the ApoB lipoprotein and consist of a guide strand (3′-5′: AUUUCAGGAAUUGUUAAAGUU) and a passenger strand (3′-5′: CUUUAACAAUUCCUGAAAUTT). Distearoylphosphatidylcholine (DSPC) and cholesterol were obtained from Avanti Polar Lipids (Alabaster, AL). PEGylated lipid (PEG 2000-DMG) was obtained from NOF Corporation (White Plains, NY). All commercial lipids were used as received without further purification. Trifluoroacetic acid, reduced Triton X-100, Tris salt, and sodium perchlorate were obtained from Sigma-Aldrich (Milwaukee, WI). SYBR gold was obtained from Molecular Probes (Eugene, OR). Dulbecco’s phosphate buffered saline solution (PBS) was obtained from HyClone laboratories (Logan, UT). All other buffers used in the study were prepared from the acid or sodium salt form of the components. Dynamic light scattering (DLS) and zeta potential measurement were conducted on a Zetasizer Nano (Malvern Instruments, UK), and results were reported as the average of triplicate measurements along with standard deviations. DLS was conducted in PBS, and a 173° scattering angle was used; the zeta potential measurement was conducted in 10 mM phosphate buffer pH adjusted to 7.4. The percent of siRNA encapsulated inside LNPs was determined using a SYBR gold fluorimetric method. Briefly, the fluorescence intensity of LNP solutions in the presence of SYBR gold (free siRNA) was compared to the fluorescence intensity of LNP solutions in the presence of both 0.5% Triton X-100 and SYBR gold (total siRNA), and the difference between these two values were the amount of siRNA encapsulated inside LNPs. The SYBR gold fluorescence was read on a SpectraMax M5 fluorescence plate reader (Molecular Devices, Sunnyvale, CA) with the excitation wavelength (λex) set at 485 nm and the emission wavelength (λem) set at 530 nm. Assembly of Lipid Nanoparticles (LNPs). LNPs were assembled using an adapted procedure described before.10 In general, lipid components (i.e., amino lipid, DSPC, cholesterol, and PEG 2000-DMG) were premixed in an organic solvent (typically ethanol), and siRNA was dissolved in an aqueous solution. Except when mentioned otherwise, the mole composition of these lipids is 58:30:10:2 (amino lipid/ cholesterol/DSPC/PEG 2000-DMG), and the N/P ratio between amino lipid and siRNA is kept at 6. Two streams were mixed through a tee mixing process, and the resulting solution went through a dialysis or diafiltration process to remove the organic solvent. The final LNP assemblies were in PBS and stored at 5 °C. They were generally characterized within a week of assembly. The amount of siRNA encapsulation in these LNPs is above 90% for all formulations according to the SYBR gold fluorimetric method. Chemical Analysis of the Lipid and siRNA Concentration in LNPs. The siRNA concentration in LNPs was determined by a gradient strong anion-exchange (SAX) method using a Dionex DNAPac PA200 (4 × 50 mm) column. To free the encapsulated siRNA, all LNPs were first treated with 0.5% Triton X-100; the resulting mixtures were then subject to SAX separation on an Agilent 1100 high-performance liquid chromatography (HPLC) system (Agilent Technologies, 398

dx.doi.org/10.1021/mp3005337 | Mol. Pharmaceutics 2013, 10, 397−405

Molecular Pharmaceutics

Article

Table 1. Summary of Bulk Composition, Hydrodynamic Diameter, Surface Charge, and Encapsulation Data for LNPs A and B lipid compositions (mol %)

particle diameter

LNP

siRNA

amino lipid

cholesterol

DSPC

PEG-DMG

N/P

Z-avg (nm)

PDI

zeta potential (mv)

encapsulated siRNA (%)

A B

ApoB ApoB

55.5 (lipid 1) 55.1 (lipid 2)

32.0 31.5

10.7 11.4

1.8 2.0

5.8 7.3

67.4 ± 0.8 85.0 ± 0.8

0.087 0.076

3.2 ± 1.1 0.5 ± 1.0

93 95

Scheme 1. Chemical Structures of the Amino Lipid 1 (LNP A) and 2 (LNP B)

structure. A flow rate of 0.5 mL/min was used to prevent excessive pressure to the resin while maintaining relatively high separation efficiency. For semipreparative separation, 2 mL of LNPs A and B was injected on the SEC column using a manual injector. Fractions were collected into 18 mL falcon tubes, and each fraction was of 4 mL. DLS sizes of fractions were measured immediately after SEC separation. In Vitro Silencing and Cytotoxicity Test of the SEC Fractions of LNPs. HepG2-derived cells stably expressing both Renilla and firefly luciferases were maintained in Dulbecco’s modified Eagle’s medium (Invitrogen, Carlsbad, CA) supplemented with 10% fetal bovine serum (Thermo Scientific, Logan, UT), 100 IU/mL penicillin, 100 μg/mL streptomycin sulfate, and 100 μg/mL Hygromycin B at 37 °C and 5% carbon dioxide atmosphere. One day prior to LNP treatment, cells were plated at a density of 20 000 cells/well in 100 μL of complete media on 96-well microplates. On the day of treatment, LNPs were serially diluted in 2-fold steps in PBS. Then 50 μL of media in each well was replaced by 50 μL of LNP solutions. The treatment was continued for 24 h without media change. Then the bioluminescence signals of luciferases were measured by Dual-Glo luciferase assay kit (Promega, Madison, WI). Treatments were performed in triplicate, and the results were analyzed by Prism version 5 (GraphPad software, La Jolla, CA). For silencing activity comparison, the signal of Renilla luciferase was normalized to that of firefly luciferase and then expressed as a percentage of the normalized value of untreated cells. For cell viability comparison, the signal of firefly luciferase was normalized to that of the untreated cells and expressed as a percentage. The equation used for nonlinear regression analysis of the IC50 of silencing activity was log[inhibitor] versus normalized response.

Santa Clara, CA) with a diode array detector. A 5 min linear gradient from 100% A (25 mM NaClO4, 10 mM Tris, 20% ethanol) to 100% B (250 mM NaClO4, 10 mM Tris, 20% ethanol) and a flow rate of 1 mL/min were used. The siRNA amount was determined using a peak area count at UV 260 nm and calculated from an external siRNA standard curve. The concentration of the four lipids in LNPs (amino lipid, DSPC, cholesterol, and PEG 2000-DMG) was determined by gradient reverse-phase ultrahigh-performance liquid chromatography methods using an Agilent Zorbax rapid resolution high definition (RRHD) SB-C8 (2.1 × 50 mm, 1.8 μm particle size) column set at 80 °C. The 4 min gradient method was from 22% A (0.1% trifluoroacetic acid in water) to 100% B (0.1% trifluoroacetic acid in methanol) and generally contained an initial 1 min hold followed by a 2−3 min linear gradient. For LNPs A and B, different slopes of the gradient needed to be used to permit a good separation between the four lipid components. Prior to analysis, LNP samples were generally diluted 20 times in a DMSO/ethanol (50:50, v/v) mixture to permit the complete solubilization of lipids. For low concentration samples, they were used directly without dilution. A Waters ACQUITY UPLC system (Water Corporation, Milford, MA) with a flow rate of 1.4 mL/min was used for separation, and a corona charged aerosol detector ultra (ultraCAD) (ESA Biosciences, Inc., Chelmsford, MA) was used for the lipid detection. The individual lipid concentration was determined from the area count of the elution peak and calculated from external lipid standard curves with a quadratic fit. Mole percentages of lipids were based on the molecular weight of each lipid, and the N/P ratio calculation was based on total nitrogen in the amino lipid vs total phosphate in the siRNA. Semipreparative SEC of LNPs Using Sephacryl S-1000. Sephacryl S-1000 SF resin was obtained from GE Healthcare. It was self-packed in a XK 16/70 column following manufacturer’s recommended procedures. The column volume was determined to be ∼120 mL using a small molecule marker. After packing, column was equilibrated in PBS for at least 10 column volumes before usage. An Agilent 1100 LC system equipped with a degasser, a quaternary pump, an autosampler with 100 μL injection loop, and a UV−vis diode array detector (DAD) was used for the analytical SEC separation. An AKTA explorer equipped with a P-900 pump, a UV-900 multiwavelength detector, and a Frac 950 fraction collector was used for the semipreparative scale SEC separation. In both cases, PBS was used as the eluent to minimize disruption to LNP

■

RESULTS AND DISCUSSION To permit the ability to tune LNP for both extracellular and intracellular trafficking, LNPs generally include multiple lipids, each serving one or several functions. For example, ionizable amino lipid is used primarily to facilitate siRNA encapsulation and membrane interaction. Neutral lipids such as cholesterol, DSPC, and PEG-DMG are used to modulate the nanoparticle physiochemical properties to improve its processability, biocompatibility, and pharmacokinetic properties. Although ionizable amino lipid is generally the central component in SAR studies,29,30 the composition and structure of other neutral lipid components also play a critical role. The loading of siRNA cargo and the formation of nanoparticles are achieved through a 399

dx.doi.org/10.1021/mp3005337 | Mol. Pharmaceutics 2013, 10, 397−405

Molecular Pharmaceutics

Article

high-energy mixing of an aqueous siRNA solution and an organic lipid solution, followed by several stages of aqueous solvent exchange for particle maturation.10 This particle formation is a spontaneous process, but it also depends strongly on the energy input and the maturation process due to the mixing kinetics. Therefore, it is commonly observed that final particle properties such as size, distribution, and morphology are heavily influenced by the processing techniques and lipid components. In a previous study, we characterized the polydispersity of LNPs using SEC and multiple detection systems.23 In particular, LNPs A and B were studied in depth using a combination of techniques including DLS, cryo-transmission electron microscopy (TEM), and SEC-UV-MALS-RI. LNPs A and B contain identical siRNA, have similar bulk lipid composition, and were assembled using the same process; they differ in the amino lipidslipid 1 is used in LNP A, and lipid 2 is used in LNP B (Table 1, the structures of the lipids are shown in Scheme 1 and Figure S1 of the Supporting Information). Characterization of these two LNPs by DLS and zeta potential showed that they are slightly positive in surface charge with size in the range of 60−90 nm. Cryo-TEM and encapsulation efficiency testing showed that both LNPs are well-formed solid nanospheres with siRNA encapsulated inside. Despite the similarity in their bulk physicochemical properties, distinct differences were found between LNPs A and B. Size and molecular weight distribution analysis of the two LNPs by SEC-MALS-RI showed that LNP A is relatively monodisperse while LNP B is polydisperse. Importantly, the coupling of an in-line UV−vis spectrophotometer after SEC separation showed that LNPs could potentially differ greatly in the distribution of siRNA loading amount. In contrast to LNP A, which showed small UV spectra differences between different eluting fractions, LNP B demonstrated a large variation in UV spectra with some particles showing no feature of siRNA signal, suggesting the potential presence of compositional heterogeneity. We sought to further investigate these two LNPs and understand the difference in compositional distribution by conducting a semipreparative SEC fractionation study. SEC of LNPs on Sephacryl S-1000. In the previous study, a TSKgel-G6000PWxl-CP column was used for the SEC separation because of its low level of LNP-column nonspecific interaction. This column is a high-resolution analytical column that can sustain high pressure from the use of multiple in-line detectors. However, its sample capacity is low, and fractionation from the column does not provide enough material for studying chemical and biological properties. During a screening of several preparative SEC resins, it is found that sephacryl S-1000 offered the required resolution with relatively low LNP-column nonspecific interactions. Sephacryl resins are constructed from cross-linked dextran and are generally inert to the analytes. The pore size rating of the S-1000 resin is similar to the pore size rating of G6000PWxl-CP column (>100 nm). It was used before for separating and characterizing neutral liposome particles with sizes up to 200−300 nm.31 To study the fractionation properties of LNPs, sephacryl S-1000 was packed into a preparative column with a column volume of 122 mL. Figure 1 shows the UV 260 chromatograms of several differentsized LNPs on the self-packed sephacryl S-1000 column. LNPs C−H had similar compositions to LNPs A and B and were assembled using different processes to produce a set of particles whose average hydrodynamic diameter ranged from 75 to 188 nm. It is shown that, except for LNP H (188 nm), all other

Figure 1. SEC of six LNPs on a 122 mL self-packed sephacryl S-1000 column eluted at 0.5 mL/min. LNPs C−H had similar compositions to LNPs A and B and were assembled using different processes to produce particles of various sizes. UV 260 was used to monitor the elution of siRNA containing LNPs. Each LNP possesses a unique Zavg hydrodynamic diameter as annotated in the figure. Different amounts of LNPs are loaded on the column to give similar UV detector responses.

LNPs have broad elution peaks, and the elution time of the peaks correlates with the hydrodynamic diameter of the LNPs. The increase in elution time for smaller particles is consistent with a separation based on SEC mechanismsmall particles can access a larger fraction of the pores and are retained longer during the elution. It is noted that the broadness of the peaks could be used to qualitatively assessing the size polydispersity in LNPs. A close inspection of the chromatograms shows that, besides the broad later elution peak, a large portion of the LNPs also had a small narrow peak around 90 min. When the average sizes of LNPs are above 138 nm, an increasing amount of LNPs starts to elute in this region; when the average sizes of LNPs are above 151 nm, all LNPs elute in this region, forming a narrow single peak. These results suggest that the small peak around 90 min corresponds to the exclusion limit of this column, in which large particles or aggregates of LNPs that are excluded from resin pores elute out. The results also show that the S-1000 resins are mostly useful for separating LNP particles smaller than 150 nm. The use of other sephacryl based resins produced similar results although they can only provide separation resolution for smaller-sized particles. For example, the use of sephacryl S-500 will exclude LNPs with sizes larger than ∼100 nm, while the use of sephacryl S-300 and S-400 will produce chromatograms where all current LNPs elute in the exclusion limits. In addition to the use of UV−vis as the detection system, we also attempted to use multiple detection systems including multiple-angle light scattering (MALS) and refractive index (RI), aiming to obtain online size and molecular weight distribution information. Although size information could be obtained through the use of MALS, it is generally found that the system setup suffers from unstable RI signals, high backpressures that can damage the resins, and high background light scattering signals. Therefore, further work in this direction was not pursued. Preparative SEC Separation, Fractionation, and Fraction Analysis of LNPs A and B. To analyze the properties of the individual particles in LNPs, we fractionated LNPs A and B using the self-packed S-1000 column and 400

dx.doi.org/10.1021/mp3005337 | Mol. Pharmaceutics 2013, 10, 397−405

Molecular Pharmaceutics

Article

collected fractions for chemical, physical, and biological testing. Figure 2a,b shows the UV 260 chromatograms of LNPs A and

Figure 3. Compositional distribution of individual fractions of LNPs A (a) and B (b) collected after the SEC separation on a self-packed 122 mL sephacryl S-1000 column eluted at 0.5 mL/min. The concentration of amino lipid, cholesterol, DSPC, and PEG-DMG is the mole percentage of each lipid in total lipids; the N/P ratio is the ratio between amino lipid and the total phosphates in siRNA. Figure 2. (a−b) UV 260 chromatograms of LNP A (a) and LNP B (b) separated on a self-packed 122 mL sephacryl S-1000 column eluted at 0.5 mL/min; also shown on the graph are the hydrodynamic diameters of the individual LNP fractions collected after the SEC separation.

continuous size decrease for the majority of its fractions (F3− F16). These large differences between two LNPs confirm the findings from the in-line size analysis using SEC-MALS in the previous study, which shows that LNP A are more uniform in size distribution compared with LNP B.23 All of the fractions were analyzed for their lipid composition, siRNA concentration, and N/P ratio (Figure 3a−b and Tables S1−2 of the Supporting Information). It is shown that, for fractions accounting for the majority of LNP A mass (i.e., F6 to F16), they demonstrate a uniform lipid chemical composition (Figure 3a). For instance, the relative standard deviation (RSD) of the concentration of lipid 1, DSPC, and cholesterol of these fractions is less than 5%. A majority of the earlier fractions (F6 to F10) appears to have slightly lower PEG-DMG% compared with later fractions, although the difference is not significant (the range is from 1.5% to 2.3%, and the RSD is 14%). Based on lipid 1 and siRNA concentration, we calculated the N/P ratio of the fractions. It is found that earlier fractions (F3−F5) have a significantly higher N/P ratio compared with the rest of the fractions, suggesting that these large particles appear to contain less siRNA cargo compared with other fractions. Nevertheless, it is noted that the fractions accounting for the majority of LNP A masses (F6−F16) have a fairly similar N/P ratio (the range is from 4.7 to 6.3, and the RSD is 10%). Taken

B on the self-packed sephacryl S-1000 column. It is found that the elution profiles of LNPs A and B are similar to the elution profiles of LNPs C-E, showing a small peak in the exclusion volume as well as a broad later elution peak. LNP B was found to have a broader elution profile compared with LNP A, consistent with its higher size polydispersity observed on the TSKgel G6000PWxl-CP analytical column.23 A total of 19 fractions from the preparative sephacryl S-1000 separation were collected between the exclusion limit and the column volume, and they were analyzed by DLS for particle sizes (Figure 2a−b) and HPLC for their chemical compositions (Figure 3a−b). It was shown that, except for LNP fractions in the exclusion limit (F1 and F2), both LNPs (F3−F19) cover a similar hydrodynamic diameter range (55−180 nm). For earlier elution fractions of both LNPs, hydrodynamic diameters of the fractions decreased with the increase in the elution volume, consistent with a separation based on the size-exclusion mechanism. It is noted that the majority of LNP A elute in the fractions F6−F16, among which a significant portion is of the same size (e.g., F12−F16). In contrast, LNP B shows a 401

dx.doi.org/10.1021/mp3005337 | Mol. Pharmaceutics 2013, 10, 397−405

Molecular Pharmaceutics

Article

Figure 4. In vitro silencing activity of the bulk and SEC fractions of LNPs A and B in a dual-luciferase reporter cell line derived from HepG2. (a) Dose−response in vitro silencing activity of the LNPs A and B bulk. The level of target mRNA knockdown is derived from the normalized relative expression ratio between the Renilla luciferase and the firefly luciferase (see the text for detail). The concentration of LNPs is in terms of the siRNA concentration in the LNPs. Experiments were performed in triplicate, and results are reported as an average along with standard deviation. (b) The in vitro silencing activity (IC50) of the LNPs A and B SEC fractions. IC 50 values (nM), the concentration of siRNA needed to cause a 50% reduction in the normalized signal of Renilla luciferase, are obtained from the dose−response silencing experiments (part a and Figures S2 and S4). The error bars are the 95% confidence interval of the calculated IC50 value.

calculated from the sum of the total concentration of lipids and siRNA of the 19 fractions with respect to the total amount of LNP materials loaded to the column (Tables S1−S2 of the Supporting Information). It is shown that for both LNPs ∼ 80% of both lipids and siRNA are recovered after the separation, suggesting that majority of the LNPs could be quantitatively separated and recovered by the column. The data also confirm that the size and chemical composition distribution results in Figure 3 are representative of the overall LNP populations. In Vitro Silencing and Toxicity Assessment of SEC Fractions of LNPs A and B. The successful scaled-up separation of LNPs enables an assessment of the bioperformance of discrete monodisperse fractions and provides an opportunity to understand the impact of subpopulations on potency and toxicity. Prior research in this area was mostly conducted on the bulk materials, and the results could be affected by the heterogeneity level within samples. We sought to determine the silencing activity and toxicity of the SEC fractions of LNPs A and B using a cell-based mRNA knockdown assay. In this assay, the HepG2 cell line was engineered to stably express both firefly luciferase and Renilla luciferase. A target site of an ApoB siRNA was inserted at the 3′ untranslated region (3′ UTR) of Renilla luciferase gene. Consequently the reduction of Renilla luciferase signal provided a readout for the silencing activity of LNPs containing ApoB siRNA. The firefly luciferase signal served as an internal control for the knockdown, and its value also indicated cell viability. For silencing activity comparison, the signal of Renilla luciferase was normalized to that of firefly luciferase and then expressed as a percentage of the normalized value of untreated cells. For cell viability comparison, the signal of firefly luciferase was normalized to that of the untreated cells and expressed as a percentage. Figures 4a, S2, and S4 show the dose−response curves of the bulk and SEC fractions of LNPs A and B. To compare the activities of these LNPs, IC 50 values, defined as the concentration of LNP causing 50% reduction of the normalized Renilla luciferase signal, were calculated and shown in Figure 4b. It is shown that LNPs A and B (the bulk LNPs without fractionation) are potent and have overlapping dose− response curves (Figure 4a) and very comparable IC 50 values

together, the chemical composition data show that majority of LNP A appears to be fairly uniform in their lipid composition and siRNA content. Interestingly, in contrast to the relative uniform distribution in LNP A, a large variation in the lipid composition and N/P ratio could be observed in LNP B SEC fractions (Figure 3b). It is shown that, for fractions accounting for the majority of LNP B mass (i.e., F3−F16), the mole percentage of lipid 2 decreases from 66% to 22% with the increase in elution volume. Correspondingly, there is an increase in the mole percentages of the other three lipids for these fractions. For example, PEGDMG% increases from ∼1.2 to ∼4, and DSPC% increases from ∼8 to ∼25 with the increase in elution volume. In the mean time, N/P ratios of these fractions decrease continuously from ∼10 to ∼2. This large change in N/P ratio explains the findings in the in-line UV−vis spectra analysis of LNP B fractions in the previous study, in which a lack of siRNA signal was observed for large particles and a stronger siRNA signal was observed for small particles.23 Taken together, results here show that LNP B is much more heterogeneous in its size distribution and chemical compositions than LNP A. The exact reason for the increased heterogeneity in LNP B is not completely understood at this point. LNPs A and B have similar bulk chemical composition, were assembled by the same process and only differ in their amino lipids. Because of the significantly reduced molecular weight of lipid 2, it is expected to have lower lipophilicity and higher aqueous/ethanol solubility than lipid 1. We speculate that this could lead to less cohesive nanoparticles and/or enhanced interparticle lipid transport during the self-assembly process (i.e., Oswald ripening process). It is noted that the heterogeneity in LNP B also coincides with particle size growth over long-term storage (data not shown). A similar mechanism could also explain why large particles in LNP A also contain less siRNA compared with the smaller sized particles. An important concern in using chromatography to separate lipid-based particles is the nonspecific interaction and irreversible adsorption to the resin, and the resulting low yield from the separation.26 To address this issue, we assessed the yield of the SEC separation by analyzing the total siRNA and lipids recovered after the SEC separation. The yield was 402

dx.doi.org/10.1021/mp3005337 | Mol. Pharmaceutics 2013, 10, 397−405

Molecular Pharmaceutics

Article

bioperformance.32 Therefore, the heterogeneity in the activity of LNP B fractions could be partially explained by the large differences in the N/P ratio or lipid composition of LNP B fractions. However, it should be noted that the in vitro efficacy of LNPs appears to be robust to the changes in particle size and composition in certain ranges. For example, although the size of LNP A fractions decreases from 137 to 56 nm (F5 to F16), their IC 50 values are comparable to each other. For the LNP B fractions F4−F11, their IC 50 values are in a relatively small range (0.9−2.9 nM) despite a 2-fold decrease in N/P ratio and a decrease of lipid 2 percentage from 66% to 51%. A further reduction in N/P ratio and changes in lipid composition from F11 to F16 resulted a dramatic drop in potency (from 3 to 40 nM), suggesting the presence of a critical formulation range to maintain potency. The present results also suggest the possibility of using fractionation as a method to rapidly screen formulation conditions and assess formulation robustness.

(∼3 nM). In contrast, fractions of LNPs A and B manifest significant differences in the in vitro activity (Figure 4b). The IC 50 values of all LNP A fractions cover a small range (1.7−7.1 nM). Close inspection shows that only two earlier fractions (F3 and F4, ∼0.1% siRNA dose) have the high IC 50 values (∼7 nM); the IC 50 values of all other LNP A fractions (F5−F19) are in a very narrow range (1.7−2.6 nM). This contrasts the large IC 50 differences in LNP B fractions, in which ∼40 fold differences in the IC 50 values could be observed (F5 vs F14). It is noted that these large differences are mostly caused by a loss of potency in the later elution fractions (F12−F19), which accounts for ∼22% siRNA dose. In fact, the potency of the majority of LNP B fractions is still in a narrow range (F4−F11, 0.9−2.9 nM). We also evaluated the dose-dependent cell viability to determine the potential differences in the in vitro cytotoxicity of the LNP fractions (Figures S3 and S5 of the Supporting Information). Both LNPs A and B bulk show no cellular toxicity in all the evaluated concentration (up to 112 nM and 160 nM, respectively). In contrast, the LNP fractions of A and B show large differences in their cellular toxicity. It is found that except earlier LNP A fractions (F1−F6) majority of LNP A fractions demonstrate no toxicity in all of the evaluated concentrations (Figure S3). Earlier LNP A fractions (F1−F6) account for only a 0.6% siRNA dose and coincide with a large particle size and high N/P ratio. Further evaluation on LNP B fractions shows that the cytotoxicity data for LNP B fractions are much more complex. For example, all fractions show cellular toxicity in the high concentration range (1−10 nM, Figure S5). This result is a little surprising since LNP B bulk shows no cytotoxicity at similar or even higher concentrations. We speculate that the lack of cytotoxicity in bulk LNP B might be related with the high level of heterogeneity in LNP B fractions. For instance, different fractions of LNP B might cause cellular toxicity through different mechanisms; therefore toxic fractions when mixed together may not reach their toxic concentration. A close inspection of Figure S5 shows that the level of cytotoxicity for LNP B fractions appears to be different. For instance, earlier LNP B fractions (F1−F5) show significantly higher toxicity compared with the later fractions. This result is consistent with the findings in LNP A fractions, suggesting that the cytotoxicity might be related with the large size and high N/P ratio in LNPs. It is further noted that cytotoxicity studied here is deduced from the activity of a control luciferase at 24 h post LNP treatment. It is possible that certain acute cytotoxicity could be induced at earlier time points by some LNP fractions and become insignificant at later time points. Taken together, the results here show that LNP A fractions are uniform with respect to in vitro activity and toxicity, while a large heterogeneity in activity and toxicity is present in LNP B. Therefore, while an assessment of LNP A bulk can describe the properties of LNP A well, a similar assessment on LNP B bulk will be inadequate in understanding LNP B. For example, the change in the LNP B bulk activity value can be caused by either a change in the intrinsic activity level or the polydispersity extent of the formulation. It is further noted that the bioperformance polydispersity results of LNPs A and B appear to be consistent with the chemical composition differences observed in their fractions. For lipid based nucleic acid delivery system, it is commonly observed that the physiochemical properties of the nanoparticles such as composition of lipids, N/P ratio, surface charge, and size could have an impact on

■

CONCLUSION Successful development of lipid nanoparticle (LNP) based siRNA therapeutics will require a strong understanding of formulation polydispersity and influence of this property on the product quality and performance. Currently, little work has been done to characterize the polydispersity of LNPs. This article reports semipreparative SEC separation of LNPs and its use in studying the heterogeneity in size, composition, and in vitro bioperformance of LNP subfractions. We demonstrate that LNPs in the size range of 60−150 nm could be well-resolved by the size-exclusion separation on a sephacryl S-1000 resin based column. Two representative LNPs (A and B) with similar bulk properties were fractionated using these methods, and fractions collected were studied in-depth for size, chemical composition, in vitro silencing, and cytotoxicity. LNP A exhibited monodisperse properties, while LNP B is found to be polydisperse. Interestingly, it is found that, of the LNPs studied, those with a narrow size distribution also presented as compositionally monodisperse. The LNP of high size polydispersity also presented substantial compositional polydispersity. Although the in vitro silencing and cytotoxicity assays did not discriminate these two LNPs when tested as bulks, major differences could be observed after fractionation. For example, the IC 50 values of most LNP A fractions are in a narrow range (1.7−2.6 nM), while ∼40 fold differences in the IC 50 values could be observed in the LNP B fractions. From these results, it is concluded that bulk properties alone are incapable of fully describing these materials, particularly in the circumstance of a complex mixture such as LNP B. Therefore, care should be taken in examining and understanding LNP drug products to ensure their quality and consistency. From a drug development perspective, LNPs of narrow size and composition distribution are more desirable and could reduce the complexity and risk in commercialization. Therefore, a characterization of particle polydispersity at the research stage could play a role in facilitating the candidate selection for commercial development. Clearly, the decision for candidate selection should balance the merits of any particular product’s pharmacological and toxicological properties with the barriers to commercial development. The methodology developed and employed here can facilitate both that assessment of commercial developability and serve as a useful tool within the overall control strategy. An alternative approach could be to employ the separation technique as a purification of nanoparticles to remove subpotent or toxic populations. Selection of 403

dx.doi.org/10.1021/mp3005337 | Mol. Pharmaceutics 2013, 10, 397−405

Molecular Pharmaceutics

Article

(12) Zimmermann, T. S.; Lee, A. C. H.; Akinc, A.; Bramlage, B.; Bumcrot, D.; Fedoruk, M. N.; Harborth, J.; Heyes, J. A.; Jeffs, L. B.; John, M.; et al. RNAi-mediated gene silencing in non-human primates. Nature 2006, 441, 111−114. (13) Vaishnaw, A. K.; Gollob, J.; Gamba-Vitalo, C.; Hutabarat, R.; Sah, D.; Meyers, R.; de Fougerolles, T.; Maraganore, J. A status report on RNAi therapeutics. Silence 2010, 1, 14. (14) Abrams, M. T.; Koser, M. L.; Seitzer, J.; Williams, S. C.; DiPietro, M. A.; Wang, W. M.; Shaw, A. W.; Mao, X. Z.; Jadhav, V.; Davide, J. P.; et al. Evaluation of Efficacy, Biodistribution, and Inflammation for a Potent siRNA Nanoparticle: Effect of Dexamethasone Co-treatment. Mol. Ther. 2010, 18, 171−180. (15) Tao, W.; Mao, X.; Davide, J. P.; Ng, B.; Cai, M.; Burke, P. A.; Sachs, A. B.; Sepp-Lorenzino, L. Mechanistically Probing Lipid-siRNA Nanoparticle-associated Toxicities Identifies Jak Inhibitors Effective in Mitigating Multifaceted Toxic Responses. Mol. Ther. 2011, 19, 567− 575. (16) Akinc, A.; Querbes, W.; De, S. M.; Qin, J.; Frank-Kamenetsky, M.; Jayaprakash, K. N.; Jayaraman, M.; Rajeev, K. G.; Cantley, W. L.; Dorkin, J. R.; et al. Targeted Delivery of RNAi Therapeutics With Endogenous and Exogenous Ligand-Based Mechanisms. Mol. Ther. 2010, 18, 1357−1364. (17) MacDonald, R. C.; Pozharski, E. V. Analysis of the structure and composition of individual lipoplex particles by flow fluorometry. Anal. Biochem. 2005, 341, 230−240. (18) MacDonald, R. C.; Pozharski, E. V. Single lipoplex study of cationic Lipoid-DNA, self-assembled complexes. Mol. Pharmaceutics 2007, 4, 962−974. (19) Szoka, F. C.; Xu, Y. H.; Hui, S. W.; Frederik, P. Physicochemical characterization and purification of cationic lipoplexes. Biophys. J. 1999, 77, 341−353. (20) Howell, B. J.; Crawford, R.; Dogdas, B.; Keough, E.; Haas, R. M.; Wepukhulu, W.; Krotzer, S.; Burke, P. A.; Sepp-Lorenzino, L.; Bagchi, A. Analysis of lipid nanoparticles by Cryo-EM for characterizing siRNA delivery vehicles. Int. J. Pharm. 2011, 403, 237−244. (21) Heath, T. D.; Goncalves, E.; Debs, R. J. The effect of liposome size on the final lipid/DNA ratio of cationic lipoplexes. Biophys. J. 2004, 86, 1554−1563. (22) Lee, H.; Williams, S. K. R.; Allison, S. D.; Anchordoquy, T. J. Analysis of self-assembled cationic lipid-DNA gene carrier complexes using flow field-flow fractionation and light scattering. Anal. Chem. 2001, 73, 837−843. (23) Zhang, J.; Haas, R. M.; Leone, A. M. Polydispersity Characterization of Lipid Nanoparticles for siRNA Delivery Using Multiple Detection Size-Exclusion Chromatography. Anal. Chem. 2012, 84, 6088−6096. (24) Barth, H. G.; Boyes, B. E.; Jackson, C. Size exclusion chromatography and related separation techniques. Anal. Chem. 1998, 70, 251r−278r. (25) Lundahl, P.; Zeng, C. M.; Hagglund, C. L.; Gottschalk, I.; Greijer, E. Chromatographic approaches to liposomes, proteoliposomes and biomembrane vesicles. J. Chromatogr., B 1999, 722, 103− 120. (26) Grabielle-Madelmont, C.; Lesieur, S.; Ollivon, M. Characterization of loaded liposomes by size exclusion chromatography. J. Biochem. Biophys. Methods 2003, 56, 189−217. (27) Baeumner, A. J.; Edwards, K. A. Analysis of liposomes. Talanta 2006, 68, 1432−1441. (28) Brandl, M.; Hupfeld, S.; Holsaeter, A. M.; Skar, M.; Frantzen, C. B. Liposome size analysis by dynamic/static light scattering upon size exclusion-/field flow-fractionation. J. Nanosci. Nanotechnol. 2006, 6, 3025−3031. (29) Zhang, J.; Fan, H.; Levorse, D. A.; Crocker, L. S. Ionization Behavior of Amino Lipids for siRNA Delivery: Determination of Ionization Constants, SAR, and the Impact of Lipid pK(a) on Cationic Lipid-Biomembrane Interactions. Langmuir 2011, 27, 1907−1914. (30) Zhang, J. T.; Fan, H. H.; Levorse, D. A.; Crocker, L. S. Interaction of Cholesterol-Conjugated Ionizable Amino Lipids with

this approach would depend on the heterogeneity level of the product, the impact of the subpopulations, the economy of the operation, as well as the availability of scalable purification techniques. With further development, SEC methods described here could serve as a purification approach for polydispersed LNPs to produce high quality products with more monodisperse properties.

■

ASSOCIATED CONTENT

S Supporting Information *

Chemical structures of the lipids, full chemical composition of the LNP fractions, full dose response in vitro silencing activity, and cell viability of the LNP fractions. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*Merck Research Laboratories, Merck & Co., Inc., Department of Pharmaceutical Sciences, West Point, Pennsylvania 19486, United States. E-mail: [email protected]. Tel.: +1 215 652 1286. Fax: 215 993 2265. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS The authors would like to acknowledge Zhong Li for his helpful discussion on chromatographic separation. The authors would also like to acknowledge medicinal chemists in Merck for providing the lipids and pharmaceutical science colleagues for preparing and analyzing LNP assemblies.

■

REFERENCES

(1) Elbashir, S. M.; Harborth, J.; Lendeckel, W.; Yalcin, A.; Weber, K.; Tuschl, T. Duplexes of 21-nucleotide RNAs mediate RNA interference in cultured mammalian cells. Nature 2001, 411, 494−498. (2) Fire, A.; Xu, S.; Montgomery, M. K.; Kostas, S. A.; Driver, S. E.; Mello, C. C. Potent and specific genetic interference by doublestranded RNA in Caenorhabditis elegans. Nature 1998, 391, 806−11. (3) Ashworth, A.; Iorns, E.; Lord, C. J.; Turner, N. Utilizing RNA interference to enhance cancer drug discovery. Nat. Rev. Drug Discovery 2007, 6, 556−568. (4) Kurreck, J. RNA Interference: From Basic Research to Therapeutic Applications. Angew. Chem., Int. Ed. 2009, 48, 1378−1398. (5) Bumcrot, D.; Manoharan, M.; Koteliansky, V.; Sah, D. W. Y. RNAi therapeutics: a potential new class of pharmaceutical drugs. Nat. Chem. Biol. 2006, 2, 711−719. (6) Castanotto, D.; Rossi, J. J. The promises and pitfalls of RNAinterference-based therapeutics. Nature 2009, 457, 426−433. (7) de Fougerolles, A.; Vornlocher, H. P.; Maraganore, J.; Lieberman, J. Interfering with disease: a progress report on siRNA-based therapeutics. Nat. Rev. Drug Discovery 2007, 6, 443−453. (8) Pecot, C. V.; Calin, G. A.; Coleman, R. L.; Lopez-Berestein, G.; Sood, A. K. RNA interference in the clinic: challenges and future directions. Nat. Rev. Cancer 2011, 11, 59−67. (9) Whitehead, K. A.; Langer, R.; Anderson, D. G. Knocking down barriers: advances in siRNA delivery. Nat. Rev. Drug Discovery 2009, 8, 129−138. (10) MacLachlan, I.; Jeffs, L. B.; Palmer, L. R.; Ambegia, E. G.; Giesbrecht, C.; Ewanick, S. A scalable, extrusion-free method for efficient liposomal encapsulation of plasmid DNA. Pharm. Res. 2005, 22, 362−372. (11) Semple, S. C.; Akinc, A.; Chen, J.; Sandhu, A. P.; Mui, B. L.; Cho, C. K.; Sah, D. W.; Stebbing, D.; Crosley, E. J.; Yaworski, E.; et al. Rational design of cationic lipids for siRNA delivery. Nat. Biotechnol. 2010, 28, 172−6. 404

dx.doi.org/10.1021/mp3005337 | Mol. Pharmaceutics 2013, 10, 397−405

Molecular Pharmaceutics

Article

Biomembranes: Lipid Polymorphism, Structure-Activity Relationship, and Implications for siRNA Delivery. Langmuir 2011, 27, 9473−9483. (31) Reynolds, J. A.; Nozaki, Y.; Tanford, C. Gel-Exclusion Chromatography on S1000 SephacrylApplication to PhospholipidVesicles. Anal. Biochem. 1983, 130, 471−474. (32) Lv, H. T.; Zhang, S. B.; Wang, B.; Cui, S. H.; Yan, J. Toxicity of cationic lipids and cationic polymers in gene delivery. J. Controlled Release 2006, 114, 100−109.

405

dx.doi.org/10.1021/mp3005337 | Mol. Pharmaceutics 2013, 10, 397−405