Letter
`
`pubs.acs.org/NanoLett
`
`Optimization of Lipid Nanoparticle Formulations for mRNA Delivery
`in Vivo with Fractional Factorial and Definitive Screening Designs
`‡
`†,‡
`‡,§
`∥
`∥
`Kevin J. Kauffman,
`J. Robert Dorkin,
`Jung H. Yang,
`Michael W. Heartlein,
`Frank DeRosa,
`‡
`‡,⊥
`and Daniel G. Anderson*,†,‡,#,∇
`Faryal F. Mir,
`Owen S. Fenton,
`†
`‡
`David H. Koch Institute for Integrative Cancer Research, §Department of Biology,
`Department of Chemical Engineering,
`⊥
`∇
`#
`Department of Chemistry,
`Institute for Medical Engineering and Science, and
`Harvard MIT Division of Health Science and
`Technology, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, United States
`∥
`Shire Pharmaceuticals, Lexington, Massachusetts 02421, United States
`*S Supporting Information
`
`ABSTRACT: Intracellular delivery of messenger RNA (mRNA) has
`the potential to induce protein production for many therapeutic
`applications. Although lipid nanoparticles have shown considerable
`promise for the delivery of small
`interfering RNAs (siRNA), their
`utility as agents
`for mRNA delivery has only recently been
`investigated. The most common siRNA formulations contain four
`components: an amine-containing lipid or
`lipid-like material,
`phospholipid, cholesterol, and lipid-anchored polyethylene glycol,
`the relative ratios of which can have profound effects on the
`formulation potency. Here, we develop a generalized strategy to
`optimize lipid nanoparticle formulations for mRNA delivery to the
`liver in vivo using Design of Experiment (DOE) methodologies including Definitive Screening and Fractional Factorial Designs.
`By simultaneously varying lipid ratios and structures, we developed an optimized formulation which increased the potency of
`erythropoietin-mRNA-loaded C12-200 lipid nanoparticles 7-fold relative to formulations previously used for siRNA delivery. Key
`features of this optimized formulation were the incorporation of 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine (DOPE) and
`increased ionizable lipid:mRNA weight ratios. Interestingly, the optimized lipid nanoparticle formulation did not improve siRNA
`delivery, indicating differences in optimized formulation parameter design spaces for siRNA and mRNA. We believe the general
`method described here can accelerate in vivo screening and optimization of nanoparticle formulations with large
`multidimensional design spaces.
`KEYWORDS: Lipid nanoparticle, mRNA, design of experiment, nucleic acid, in vivo
`
`N ucleic acids have tremendous therapeutic potential to
`
`modulate protein expression in vivo but must be
`delivered safely and effectively. Because the delivery of naked
`nucleic acids results in poor cellular internalization, rapid
`degradation, and fast renal clearance,1,2 lipid nanoparticles
`(LNPs) have been developed to encapsulate and deliver nucleic
`acids to the liver. Most notably, the field has seen orders-of-
`in the delivery of 21−23
`magnitude potency advances
`nucleotide-long double stranded small
`interfering RNAs
`(siRNAs) due in part to the creation of new synthetic ionizable
`lipids and lipid-like materials.2 Whereas some of these novel
`lipids were synthesized with rational design approaches by
`systematically varying the lipid head and tail structures (e.g.,
`DLin-KC2-DMA, DLin-MC3-DMA, L319),3−5 other materials
`were discovered by creating large combinatorial
`libraries of
`lipid-like materials (e.g., C12-200, cKK-E12, 503O13).6−8
`When formulated into LNPs, these amine-containing ionizable
`lipids and lipid-like materials electrostatically complex with the
`negatively charged siRNA and can both facilitate cellular uptake
`and endosomal escape of the siRNA to the cytoplasm.6,9 In
`particular, the ionizable lipid-like material C12-200 has been
`
`widely used to make siRNA-LNP formulations for various
`therapeutic applications in vivo to silence protein expres-
`sion.10−12
`In addition to the ionizable material, three other excipients
`are also commonly used to formulate LNPs: (1) a
`phospholipid, which provides structure to the LNP bilayer
`and also may aid in endosomal escape;2,13 (2) cholesterol,
`which enhances LNP stability and promotes membrane
`fusion;14,15 and (3) lipid-anchored polyethylene glycol
`(PEG), which reduces LNP aggregation and “shields” the
`LNP from nonspecific endocytosis by immune cells.16 The
`particular composition of the LNP can also have profound
`effects on the potency of the formulation in vivo. Several
`previous efforts to study the effect of formulation parameters
`on siRNA-LNP potency utilized the one-variable-at-a-time
`method,17,18 in which formulation parameters were individually
`
`Received:
`June 23, 2015
`Revised: October 12, 2015
`
`© XXXX American Chemical Society
`
`A
`
`DOI: 10.1021/acs.nanolett.5b02497
`Nano Lett. XXXX, XXX, XXX−XXX
`
`Moderna Ex 1019-p. 1
`Moderna v Arbutus
`
`

`

`Nano Letters
`
`Letter
`
`Figure 1. Formulation of lipid nanoparticles. Lipid nanoparticles (LNPs) are synthesized by the mixing of two phases: (1) a four-component ethanol
`phase containing ionizable lipid, helper phospholipid, cholesterol, and lipid-anchored PEG; (2) an acidic aqueous phase containing mRNA.
`
`Table 1. Library A, B, and C Formulation Parameters
`
`parameter
`C12-200:mRNA weight ratio
`phospholipid
`
`original formulation
`5:1
`DSPC
`
`Library A
`Library B
`2.5:1 to 7.5:1
`7.5:1 to 12.5:1
`DSPC, DSPE
`DSPC
`DOPC, DOPE
`DOPE
`C12-200 molar composition
`50%
`40% to 60%
`30% to 40%
`35%
`phospholipid molar composition
`10%
`4% to 16%
`16% to 28%
`16%
`cholesterol molar composition
`38.5%
`21.5% to 55.5%
`28.5% to 51.5%
`46.5%
`PEG molar composition
`1.5%
`0.5% to 2.5%
`2.5% to 3.5%
`2.5%
`aPhospholipid abbreviations: DS = 1,2-distearoyl-sn-glycero- (saturated tail), DO = 1,2-dioleoyl-sn-glycero- (Δ9-cis unsaturated tail), PC = 3-
`phosphocholine (primary amine headgroup), PE = 3-phosphoethanolamine (quaternary amine headgroup).
`
`Library C
`5:1 to 25:1
`DOPE
`
`varied to maximize LNP potency; this approach, however, does
`not allow for examination of potentially important second-
`order interactions between parameters. Inspired by statistical
`methodologies commonly used in the engineering and
`combinatorial chemistry literature,19,20 we chose to utilize
`Design of Experiment (DOE) to better optimize LNP
`formulations for nucleic acid delivery. Using DOE, the number
`of
`individual experiments required to establish statistically
`significant trends in a large multidimensional design space are
`considerably reduced, which is particularly relevant for the
`economical screening of LNP formulations: in vitro screens are
`often poor predictors of in vivo efficacy with siRNA-LNPs,21
`and it would be both cost- and material-prohibitive to test large
`libraries of LNP formulations in vivo.
`To demonstrate the application of DOE to LNP formulation
`optimization in vivo, we formulated LNPs with a different type
`of nucleic acid than siRNA. Recently, messenger mRNA
`(mRNA) has been investigated for
`therapeutic protein
`production in vivo, including applications in cancer immuno-
`therapy, infectious disease vaccines, and protein replacement
`therapy.22,23 Unlike plasmid DNA, mRNA need only access the
`cytoplasm rather than the nucleus to enable protein translation
`and has no risk of inducing mutation through integration into
`the genome.24 Because there are inherent chemical and
`structural differences between mRNA and siRNA in terms of
`length, stability, and charge density of the nucleic acid,25 we
`hypothesized that LNP delivery formulations for mRNA may
`require significant variation from those developed for siRNA
`delivery. We further hypothesized that formulated mRNA may
`
`pack differently and with different affinity into nanoparticles
`than siRNA. To optimize LNP formulation parameters
`specifically for mRNA delivery, we developed a novel strategy
`in which we used DOE methodologiesincluding both
`Fractional Factorial and Definitive Screening Designsto
`synthesize several smaller LNP libraries to screen in vivo.
`Using the formulation conditions of the original siRNA-LNPs
`as a starting point, each successive generation of library was
`designed to improve protein expression based upon the
`parameters in the previous library that were found to correlate
`with improved efficacy. Through this approach, we aimed to
`develop an optimized C12-200 LNP with increased protein
`expression over the original LNP formulation.
`EPO mRNA Delivery with Original siRNA-Optimized
`LNP. The formulation process for synthesizing LNPs is
`described in Figure 1. The organic phase containing the lipids
`was mixed together with the acidic aqueous phase containing
`the nucleic acid in a microfluidic channel,26 resulting in the
`formation of mRNA-loaded LNPs. We chose to use unmodified
`mRNA coding for erythropoietin (EPO), a secreted serum
`protein that has previously been successfully translated in
`vivo.25,27 It has further been recently reported28 that LNP-
`delivered unmodified EPO mRNA is more potent than EPO
`mRNA with pseudouridine and/or 5-methylcytidine modifica-
`tions in vitro and in mice. To establish a baseline from which to
`improve, EPO mRNA was first formulated into LNPs using the
`formulation parameters previously published6 for
`original
`siRNA delivery in vivo (Table 1). The formulation was dosed
`intravenously at 15 μg of total mRNA per mouse and resulted
`
`B
`
`DOI: 10.1021/acs.nanolett.5b02497
`Nano Lett. XXXX, XXX, XXX−XXX
`
`Moderna Ex 1019-p. 2
`Moderna v Arbutus
`
`

`

`Nano Letters
`
`Letter
`
`Figure 2. Efficacy results of LNPs in Libraries A, B, and C. (a) Serum EPO concentration 6 h post-intravenous injection of 15 μg total mRNA for
`each formulation in Libraries A and B, including the original formulation (data presented as mean + SD, n = 3). (b) A statistically significant trend of
`increasing serum EPO concentration was observed with increasing C12-200:mRNA weight ratio and with DOPE phospholipid for Library B
`formulations, independent of the other formulation parameters. Furthermore, a statistically significant second-order effect was observed between
`DOPE and increasing weight ratio, as indicated by the larger relative slope of the DOPE best-fit line compared to the DSPC best-fit line. (1 data
`point = 1 mouse) (c) Serum EPO concentration 6 h post-intravenous injection of 15 μg total mRNA for formulation B-26 and Library C, which had
`similar formulation parameters as B-26 with differing C12-200:mRNA weight ratios. (Data presented as mean + SD, n = 3.)
`
`in an average EPO serum level of 963 ± 141 ng/mL at 6 h
`post-injection.
`Optimization of mRNA LNPs with Design of Experi-
`ment. Some previous efforts
`to optimize nanoparticle
`formulations have involved varying each of
`the important
`parameters individually and then possibly combining each
`optimized parameter
`for an overall optimized formula-
`tion.17,18,29 Because pilot experiments
`suggested strong
`second-order effects between parameters in our system, we
`chose instead to vary all five independent parameters
`simultaneously. In an attempt to maximize EPO expression in
`mice and thereby optimize the C12-200 LNPs for mRNA
`delivery, we chose to simultaneously vary the C12-200:mRNA
`weight
`ratio,
`the phospholipid identity, and the molar
`composition of the four-component LNP formulation. Three
`additional phospholipids structurally similar to DSPC but with
`differing head groups (primary vs quaternary amine) and tail
`saturation (saturated vs Δ9-cis unsaturated) were incorporated
`into the LNP formulations.
`Library A: Definitive Screening Design. We designed the
`first library, Library A, to be centered around the original
`siRNA-optimized LNP formulation parameters (Table 1). With
`four three-level quantitative factors (C12-200:mRNA weight
`ratio and three independent formulation molar compositions)
`and one four-level qualitative factor (phospholipid type), this
`
`large five-dimensional design space required DOE to reduce the
`number of formulations (3 × 3 × 3 × 3 × 4 = 324) to a
`reasonable number for in vivo experiments. An initial library of
`14 formulations (coded A-01 through A-14, see Table S1 for
`parameters) was created using a Definitive Screening Design, a
`recently described economical DOE in which main effects are
`not confounded with two-factor interactions and nonlinear
`correlations can be detected.30 The purpose of this first screen
`was to sample the large design space in a controlled fashion to
`eliminate unimportant formulation parameters and/or find a
`local maximum in efficacy from which a second-generation
`library could be generated.
`Out of 14 formulations in Library A, two formulations (A-02
`and A-09) resulted in higher EPO serum levels (6445 ± 1237
`and 2072 ± 302 ng/mL,
`respectively) than the original
`formulation (Figure 2a). Although the results from Library A
`were insufficient to deduce statistically significant effects for
`EPO production in vivo, there were statistically significant (p <
`0.05) orthogonal trends (Figure S2). We hypothesize that the
`increased encapsulation efficiency with increasing C12-
`200:mRNA weight ratio (Figure S2a) is caused by better
`complexation of more positively charged ionized C12-200 lipid
`with negatively charged mRNA. We also observed decreased
`LNP size with increasing PEG composition (Figure S2b), a
`phenomenon that has been previously observed in the
`
`C
`
`DOI: 10.1021/acs.nanolett.5b02497
`Nano Lett. XXXX, XXX, XXX−XXX
`
`Moderna Ex 1019-p. 3
`Moderna v Arbutus
`
`

`

`Nano Letters
`
`literature18,31 and has been speculated to be caused by
`increased lipid bilayer compressibility and increased repulsive
`forces between liposomes.32 The two top-performing for-
`mulations of Library A (A-02 and A-09) possessed similar
`attributes:
`increased weight ratio (7.5:1 vs 5:1),
`increased
`phospholipid content (16% vs 10%), and either DSPC or
`DOPE as the phospholipid; moreover, A-02 had decreased
`C12-200 content (40% vs 50%) and A-09 had increased PEG
`content (2.5% vs 1.5%).
`Library B: Fractional Factorial Screening Design. A more
`robust second-generation library, Library B (coded B-15 to B-
`32, Table S1), was generated using a L18-Taguchi Fractional
`Factorial Design29 with new parameter ranges which shifted in
`the direction of the two top-performing LNPs from the first
`library (Table 1). Out of 18 formulations in Library B, 11
`formulations resulted in higher EPO serum levels than the
`original
`formulation (Figure 2a). The top-performing for-
`mulation was B-26 with an average serum EPO concentration
`of 7485 ± 854 ng/mL. A standard least squares linear
`regression model was applied to the data from Library B, and
`several statistically significant factors were found with respect to
`efficacy (Table S2). Several second-order effects were found to
`be statistically significant as well, including the second-order
`interaction between DOPE and C12-200:mRNA weight ratio as
`shown by the best-fit line (p < 0.05) for DOPE in Figure 2b.
`Additional description of the statistical model and significant
`effects may be found in the Supporting Information (Table S2,
`Figure S1).
`that
`trend from Library B was
`The most apparent
`formulations with DOPE as the phospholipid resulted in
`significantly higher EPO production than formulations with
`DSPC, the original phospholipid (Figure 2b). In fact, the
`presence of DOPE in the formulation was the single strongest
`predictor of
`in vivo efficacy in our study. Whereas DSPC
`contains a quaternary amine headgroup and a fully saturated
`tail, DOPE contains a primary amine headgroup and a tail with
`one degree of unsaturation. It has been reported that conical
`lipids, such as DOPE, tend to adopt the less stable hexagonal
`phase, while cylindrical lipids, such as DSPC, tend to adopt the
`more stable lamellar phase.33 Upon fusion with the endosomal
`membrane, LNPs containing DOPE may reduce membrane
`stability, ultimately promoting endosomal escape.34,35 Another
`possible explanation involves their different encapsulation
`efficiencies: independent of other varying formulation param-
`eters, formulations with DSPC entrapped mRNA on average
`significantly better than DOPE (51% vs 36%), so it may be
`possible that the stronger complexation of mRNA to lipid in
`DSPC LNPs hinders the subsequent decomplexation of mRNA
`from lipid once inside the cell, thus inhibiting translation of the
`mRNA to protein.
`Library C: Maximizing Lipid:mRNA Weight Ratio with
`DOPE. As was initially hypothesized, we observed several
`second-order effects on EPO production between formulation
`parameters in Library B, most notably the synergistic effect
`between increasing the C12-200:mRNA weight ratio along with
`the use of DOPE as the phospholipid (Figure 2b). In an effort
`to further increase in vivo potency, a third and final library was
`generated (Library C, Table 1) to exploit
`this discovered
`second-order effect. The top-performing formulation (B-26)
`from Library B was reformulated with C12-200:mRNA weight
`ratios varying from 5:1 to 25:1 (coded C33−C38, Table S1).
`Surprisingly,
`increasing the weight ratio only increased the
`serum EPO concentration up to a certain point (Figure 2c); it
`
`Letter
`
`appears that increasing the weight ratio beyond 10:1 confers no
`significant efficacy advantage in vivo. Because no significant
`increases in EPO production were observed beyond 10:1 and
`to mitigate any concerns with possible lipid toxicity caused by
`increased lipid doses, we chose the 10:1 C12-200:mRNA
`weight
`ratio (C-35) as the final mRNA-optimized LNP
`formulation (Table 2).
`
`Table 2. LNP Characteristics of C-35 Compared to the
`Original Formulationa
`
`optimized formulation
`(C-35)
`10:1
`DOPE
`35%
`16%
`
`46.5%
`2.5%
`
`7065 ± 513
`102
`0.158
`43
`
`original
`formulation
`5:1
`DSPC
`50%
`10%
`
`38.5%
`1.5%
`
`962 ± 141
`152
`0.102
`24
`
`C12-200:mRNA weight ratio
`phospholipid
`C12-200 molar composition
`phospholipid molar
`composition
`cholesterol molar composition
`C14 PEG 2000 molar
`composition
`serum EPO (ng/μL)
`diameter (nm)
`polydispersity index (PDI)
`mRNA encapsulation efficiency
`(%)
`6.96
`7.25
`pKa
`−5.0
`−25.4
`zeta potential (mV)
`aPhospholipid abbreviations: DSPC = 1,2-distearoyl-sn-glycero-3-
`phosphocholine, DOPE = 1,2-dioleoyl-sn-glycero-3-phosphoethanol-
`amine, Serum EPO reported as mean ± SD (n = 3) 6 h after 15 μg of
`total mRNA intravenous injection into mice.
`
`Evaluation of Methodology. Although only 14% (2 of 14)
`of the Library A formulations resulted in increased potency
`compared to the original parameters, 61% (11 of 18) of the
`Library B formulations and 100% of Library C formulations (6
`of 6) did so (Figures 2a,c). This suggests that formulation
`parameters can be optimized and are critically important for
`efficient mRNA delivery with C12-200 LNPs. Furthermore, the
`increasing percentage of formulations that performed better
`than the original in each subsequent library demonstrates the
`predictive success of the generated statistical models (Table
`S2). A flowchart of the complete methodology we developed
`for in vivo nanoparticle optimization can be found in Figure S3.
`Characterization of mRNA-Optimized LNP. The opti-
`mized formulation C-35 had the following formulation
`parameters: 10:1 C12-200:mRNA weight
`ratio with 35%
`C12-200, 16% DOPE, 46.5% cholesterol, and 2.5% C14-
`PEG2000 molar composition. The average efficacy of C-35
`with 15 μg of total EPO mRNA injection in vivo, 7065 ± 513
`ng/mL, was increased over 7-fold compared to the original
`traditional LNP formulation (963 ± 141 ng/mL). C-35 was
`further characterized and compared to the original formulation
`with regard to size, polydispersity, encapsulation efficiency, and
`pKa (Table 2). No significant morphological differences were
`observed between the two formulations with transmission
`electron microscopy (TEM) (Figure S4). Although others have
`reported increases
`in siRNA nanoparticle potency with
`decreasing size,36 we found no such trend with all 38 mRNA
`formulations tested in our LNP system. Jayaraman et al.4 found
`that pKa was an important characteristic in predicting the
`efficacy of liver-targeting siRNA LNPs with an optimal pKa of
`between 6.2 and 6.5. It appears that in our C12-200 mRNA
`
`D
`
`DOI: 10.1021/acs.nanolett.5b02497
`Nano Lett. XXXX, XXX, XXX−XXX
`
`Moderna Ex 1019-p. 4
`Moderna v Arbutus
`
`

`

`Nano Letters
`
`Letter
`
`Figure 3. Efficacy and biodistribution of original and C-35 formulation with Luc mRNA. (a) Efficacy of original and C-35 LNP formulations
`synthesized with mRNA coding for luciferase in three organs of interest as measured by total flux from luminescence 6 h after intravenous injection
`of 15 μg total mRNA. (Data presented as mean + SD, n = 3). (b) Representative biodistribution image of luciferase expression for original and C-35
`LNP in seven organs as measured with an IVIS imaging system 6 h after intravenous injection of 15 μg of total mRNA.
`
`system, the in vivo efficacy is not significantly correlated with
`pKa of the LNP, although the slightly lower pKa of C-35 (pKa =
`6.96) compared to the original formulation (pKa = 7.25) may
`partially explain its improved efficacy. The surface charge of the
`LNP may also partially explain differences in efficacy: the
`optimized formulation C-35 is less negatively charged (zeta
`potential = −5.0 mV) than the original formulation (−25.4
`mV). C-35 contains twice the amount of amine-rich ionizable
`lipid C12-200 than the original formulation, which is likely the
`predominant reason C-35 is more positively charged. Although
`one study found no relationship between surface charge and
`hepatocellular delivery in vivo with siRNA-loaded lipid
`nanoparticles,21 other reports have noted that more positively
`charged nanoparticles bind better to negatively charged cellular
`membranes and this electrostatic interaction might facilitate
`uptake.37
`In order to determine whether C-35 would similarly improve
`the efficacy of mRNAs with different lengths, we formulated
`LNPs with firefly luciferase (Luc) mRNA, an mRNA which has
`a coding region roughly three times longer than that of EPO
`mRNA (1653 vs 582 nucleotides). Luciferase protein generated
`by C-35 LNPs was expressed predominately in the liver and
`likewise resulted in a statistically significant, approximately 3-
`fold increase in luciferase expression as measured by liver
`luminescence compared to the original formulation (Figure 3).
`Although LNPs made with Luc mRNA had similar
`encapsulation efficiencies as those made with shorter EPO
`mRNA (Tables 1, S3), we anticipate that significantly longer
`mRNAs would eventually become too large to effectively load
`into LNPs.
`siRNA Delivery with mRNA-Optimized LNP. Having
`optimized the formulation for mRNA delivery, we then wanted
`to examine the potential
`for siRNA delivery with C-35 as
`compared to the original siRNA-optimized formulation. We
`formulated siRNA coding for Factor VII (FVII), a serum
`clotting factor expressed exclusively in hepatocytes, using both
`the C-35 LNP and the original LNP formulation to determine
`their
`relative silencing in hepatocytes. FVII
`levels were
`measured 72 h after intravenous injection of siRNA-loaded
`LNPs ranging from 0.01 mg/kg to 0.1 mg/kg, and there was no
`significant difference between the original and optimized
`formulations at any dose (Figure 4, Table S4) despite having
`significantly different formulation parameters. The ED50 of both
`C-35 and the original
`formulations with FVII siRNA were
`
`Figure 4. Efficacy of original and C-35 formulation with siRNA.
`Efficacy of original versus optimized C-35 formulation made with C12-
`200 and siRNA coding against Factor VII (FVII) protein as measured
`by serum FVII levels 72 h post-intravenous injection of various doses
`of total siRNA. FVII levels were normalized with respect to PBS-
`injected control mice. (Data presented as mean + SD, n = 3.)
`
`approximately 0.03 mg/kg of total siRNA content, consistent
`with previous reports.6
`Interestingly, siRNA-loaded LNPs may be more tolerant than
`mRNA-loaded LNPs of design space differences. Over the past
`decade in the siRNA delivery field, many groups have focused
`on developing new ionizable lipids to increase the potency of
`siRNA-LNPs but have generally used the same standard
`formulation parameters in consecutive studies.3,4,6−8 The
`discovery of new ionizable lipids and lipid-like materials,
`however,
`is an endeavor which is often time- and material-
`intensive,
`requiring large-scale combinatorial
`libraries or
`chemically difficult rational design approaches. Meanwhile, we
`have shown that for one of the most commonly used ionizable
`materials for siRNA delivery, C12-200, merely changing the
`formulation parameters can significantly increase the potency of
`the LNP when loaded with two different mRNAs of varying
`lengths, EPO or Luc (Table 2, Figure 3).
`In this study, we have demonstrated a new general method
`for optimizing previously used siRNA lipid nanoparticle
`technology for a new class of RNA therapeutics and identified
`a lead optimized formulation for mRNA delivery, coded C-35.
`To the best of our knowledge, this study represents the first
`optimization of nanoparticle potency in vivo using Design of
`Experiment principles. Although C-35 significantly improved
`
`E
`
`DOI: 10.1021/acs.nanolett.5b02497
`Nano Lett. XXXX, XXX, XXX−XXX
`
`Moderna Ex 1019-p. 5
`Moderna v Arbutus
`
`

`

`Nano Letters
`
`mRNA delivery with mRNA’s of two different lengths, C-35
`was surprisingly equally as efficacious for siRNA delivery as the
`original siRNA-optimized formulation. We believe that the
`optimized formulations described here may provide a basis for
`further formulation optimization with other mRNA delivery
`materials as well. Furthermore, the generalized approach we
`described for in vivo optimization of multicomponent nano-
`particle formulations may accelerate the discovery of more
`potent formulations with other materials and drug payloads.
`Methods. Lipid Nanoparticle Synthesis. The ethanol phase
`was prepared by solubilizing with ethanol a mixture of C12-200
`(prepared as previously described,6 courtesy of Alnylam
`Pharmaceutics, Cambridge, MA), 1,2-distearoyl-sn-glycero-3-
`phosphocholine (DSPC, Avanti Polar Lipids, Alabaster, AL),
`1,2-distearoyl-sn-glycero-3-phosphoethanolamine (DSPE,
`Avanti), 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC,
`Avanti), 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine
`(DOPE, Avanti), cholesterol (Sigma), and/or 1,2-dimyristoyl-
`sn-glycero-3-phosphoethanolamine-N-[methoxy(polyethylene
`glycol)-2000] (ammonium salt) (C14-PEG 2000, Avanti) at
`predetermined molar ratios. The aqueous phase was prepared
`in 10 mM citrate buffer (pH 3) with either EPO mRNA
`(human erythropoietin mRNA, courtesy of Shire Pharmaceut-
`icals, Lexington, MA), Luc mRNA (Firefly luciferase mRNA,
`Shire), or FVII siRNA (Factor VII siRNA,7 Alnylam). Syringe
`pumps were used to mix the ethanol and aqueous phases at a
`3:1 ratio in a microfluidic chip device.26 The resulting LNPs
`were dialyzed against PBS in a 20 000 MWCO cassette at 4 °C
`for 2 h.
`mRNA Synthesis. mRNA was synthesized by in vitro
`transcription from a plasmid DNA template encoding the
`gene, which was followed by the addition of a 5′ cap structure
`(Cap 1) using a vaccinia virus-based guanylyl
`transferase
`system. A poly(A) tail of approximately 300 nucleotides was
`incorporated via enzymatic addition employing poly-A
`polymerase. Fixed 5′ and 3′ untranslated regions were
`constructed to flank the coding sequences of the mRNA.
`LNP Characterization. To calculate the nucleic acid
`encapsulation efficiency, a modified Quant-iT RiboGreen
`RNA assay (Invitrogen) was used as previously described.38
`The size and polydispersity (PDI) of the LNPs were measured
`using dynamic light scattering (ZetaPALS, Brookhaven Instru-
`ments). Zeta potential was measured using the same instrument
`in a 0.1× PBS solution. Size data is reported as the largest
`intensity mean peak average, which constituted >95% of the
`nanoparticles present in the sample. The pKa was determined
`using a TNS assay as previously described.38 To prepare LNPs
`for Transmission Electron Microscopy (TEM), LNPs were
`dialyzed against water and negative staining was performed with
`2% uranyl acetate. LNPs were then imaged with a Tecnai Spirit
`transmission electron microscope (FEI, Hillsboro, OR).
`Animal Experiments. All animal studies were approved by
`the M.I.T. Institutional Animal Care and Use Committee and
`were consistent with local, state, and federal regulations as
`applicable. Female C57BL/6 mice (Charles River Laboratories,
`18−22 g) were intravenously injected with LNPs via the tail
`vein. After 6 or 72 h, blood was collected via the tail vein with
`serum separation tubes, and the serum was isolated by
`centrifugation. Serum EPO levels were measured using an
`ELISA assay (Human Erythropoietin Quantikine IVD ELISA
`Kit, R&D Systems, Minneapolis, MD). Serum FVII levels were
`measured using a chromogenic assay (Biophen FVII, Aniara
`Corporation, West Chester, OH) and compared with a
`
`Letter
`
`standard curve obtained from control mice. Six hours after
`administration of Luc mRNA LNPs, mice were administered an
`intraperitoneal injection of 130 μL of D-luciferin (30 mg/mL in
`PBS). After 15 min, the mice were sacrificed, and eight organs
`were collected (liver, spleen, pancreas, kidneys, uterus, ovaries,
`lungs, heart). The organs’ luminescence were analyzed using an
`IVIS imaging system (PerkinElmer, Waltham, MA) and
`quantified using LivingImage software (PerkinElmer) to
`measure the radiance of each organ in photons/sec.
`Statistics. Design of Experiment (DOE) was performed, and
`statistical data were analyzed using JMP software (SAS, Cary,
`N.C.). In this study, statistical significance was defined as p-
`values less than 0.05. Three mice per formulation/dose (n = 3)
`were used for all in vivo experiments. For Library A, a 34 × 22
`Definitive Screening Design30 was used with 4 three-level
`quantitative factors (C12-200 RNA weight ratio, C12-200 mol
`%, phospholipid mol %, and PEG mol %) and 2 two-level
`qualitative factors for phospholipid tail group (DS = 1,2-
`distearoyl-sn-glycero- and DO = 1,2-dioleoyl-sn-glycero-) and
`phospholipid headgroup (PC = 3-phosphocholine and PE = 3-
`phosphoethanolamine). For Library B, a 34 × 21 L-18 Taguchi
`Fractional Factorial Design29 was used with 4 three-level
`quantitative factors (C12-200 RNA weight ratio, C12-200 mol
`%, phospholipid mol %, and PEG mol %) and 1 two-level
`qualitative factor for phospholipid (DSPC or DOPE). To make
`the Standard Least Squares regression model for Library B, a
`full model with all orthogonal and second-order effects was
`generated and subsequently reduced until only statistically
`significant effects remained in the model as determined by
`ANOVA. A posthoc Tukey test was performed using JMP to
`verify that the two levels of phospholipid effect were statistically
`different (p < 0.0001). When comparing means between two
`groups, a Student’s t
`test was used assuming a Gaussian
`distribution and unequal variances. Further details about
`statistics and models used in this study,
`including ANOVA
`results, parameter estimates, residuals, etc., can be found in
`Table S2, Figure S1, and the Supplementary Methods section.
`
`■ ASSOCIATED CONTENT
`*S Supporting Information
`The Supporting Information is available free of charge on the
`ACS Publications website at DOI: 10.1021/acs.nano-
`lett.5b02497.
`Description of Library B statistical model, nanoparticle
`characterization for all LNP formulations (including
`formulation composition, encapsulation efficiency, size,
`polydispersity, and efficacy measurements), additional
`structure/function relationships for Library A, and a
`detailed description of the statistical methodologies used
`including a flowchart (PDF)
`
`■ AUTHOR INFORMATION
`
`Corresponding Author
`*E-mail (D.G.A.) dgander@mit.edu. 500 Main Street, David.
`H. Koch Institute for Integrative Cancer Research, Massachu-
`setts Institute of Technology, Cambridge, MA 02139.
`Author Contributions
`K.J.K. and J.R.D. contributed equally to this work.
`Notes
`The authors declare no competing financial interest.
`
`F
`
`DOI: 10.1021/acs.nanolett.5b02497
`Nano Lett. XXXX, XXX, XXX−XXX
`
`Moderna Ex 1019-p. 6
`Moderna v Arbutus
`
`

`

`Nano Letters
`
`■ ACKNOWLEDGMENTS
`
`This work was supported by Shire Pharmaceuticals (Lexington,
`MA) and the MIT Skoltech Initiative. The authors thank Nicki
`Watson at the W.M. Keck Microscopy Institute (Whitehead
`Institute, Cambridge, MA) for assistance in performing the
`TEM experiments. We also thank Prof. Sumona Mondal
`(Clarkson University, Potsdam, NY) for performing a statistical
`review of the manuscript.
`
`■ REFERENCES
`(1) Whitehead, K. A.; Langer, R.; Anderson, D. G. Nat. Rev. Drug
`Discovery 2009, 8 (2), 129−138.
`(2) Kanasty, R.; Dorkin, J. R.; Vegas, A.; Anderson, D. Nat. Mater.
`2013, 12 (11), 967−977.
`(3) Semple, S. C.; Akinc, A.; Chen, J.; Sandhu, A. P.; Mui, B. L.; Cho,
`C. K.; Sah, D. W. Y.; Stebbing, D.; Crosley, E. J.; Yaworski, E.; Hafez, I.
`M.; Dorkin, J. R.; Qin, J.; Lam, K.; Rajeev, K. G.; Wong, K. F.; Jeffs, L.
`B.; Nechev, L.; Eisenhardt, M. L.; Jayaraman, M.; Kazem, M.; Maier,
`M. a; Srinivasulu, M.; Weinstein, M. J.; Chen, Q.; Alvarez, R.; Barros,
`S. a; De, S.; Klimuk, S. K.; Borland, T.; Kosovrasti, V.; Cantley, W. L.;
`Tam, Y. K.; Manoharan,