Combinatorial synthesis of chemically diversecore-shell nanoparticles for intracellular deliveryDaniel J. Siegwarta,b, Kathryn A. Whiteheada,b, Lutz Nuhna, Gaurav Sahaya, Hao Chenga,b, Shan Jianga,b, Minglin Maa,c,Abigail Lytton-Jeana, Arturo Vegasa,c, Patrick Fentona,c, Christopher G. Levinsa,c, Kevin T. Loveb, Haeshin Leea,Christina Corteza, Sean P. Collinsa, Ying Fei Lia, Janice Janga, William Querbesd, Christopher Zurenkod,Tatiana Novobrantsevad, Robert Langera,b,c,e, and Daniel G. Andersona,b,c,e,1

aDavid H. Koch Institute for Integrative Cancer Research, Massachusetts Institute of Technology, Cambridge, MA 02139; bDepartment of ChemicalEngineering, Massachusetts Institute of Technology, Cambridge, MA 02139; cDepartment of Anesthesiology, Children’s Hospital Boston, Harvard MedicalSchool, Boston, MA 02115; dAlnylam Pharmaceuticals, Inc., 300 Third Street, Cambridge, MA 02142; and eHarvard-MIT Division of Health Sciences andTechnology, Cambridge, MA 02139

Edited* by Alexander M. Klibanov, Massachusetts Institute of Technology, Cambridge, MA, and approved June 20, 2011 (received for review May 2, 2011)

Analogous to an assembly line, we employed a modular design forthe high-throughput study of 1,536 structurally distinct nanoparti-cles with cationic cores and variable shells. This enabled elucidationof complexation, internalization, and delivery trends that couldonly be learned through evaluation of a large library. Using roboticautomation, epoxide-functionalized block polymers were combi-natorially cross-linked with a diverse library of amines, followedby measurement of molecular weight, diameter, RNA complexa-tion, cellular internalization, and in vitro siRNA and pDNA delivery.Analysis revealed structure-function relationships and beneficialdesign guidelines, including a higher reactive block weight frac-tion, stoichiometric equivalence between epoxides and amines,and thin hydrophilic shells. Cross-linkers optimally possessed ter-tiary dimethylamine or piperazine groups and potential bufferingcapacity. Covalent cholesterol attachment allowed for transfectionin vivo to liver hepatocytes in mice. The ability to tune the chemicalnature of the core and shell may afford utility of these materialsin additional applications.

Chemically diverse nanoparticles form the basis of a number ofemerging applications in materials science (1–3), including

drug and gene delivery vehicles (4, 5). However, it is difficultto predict the optimal chemical and physical properties for deliv-ery of a specific drug or biomolecule. A key challenge in nano-particle development is the need to perform distinct synthesis,characterization, and formulation steps before performancecan be evaluated. Although combinatorial methods have drivensmall molecule drug discovery (6), they have been less exploredfor the discovery of polymers (7–12) in part due to these inherentchallenges. A less studied area in high-throughput (HT) polymericnanoparticle methodology (13–16) is the facile incorporation ofdiverse chemical groups into polymers with defined architectures.By applying HT robotic techniques to the controlled cross-linkingof block copolymers, we were able to prepare a library of core-shellnanoparticles for nucleic acid complexation and delivery with greatvariability in the chemical nature of the protonizable amine-basedcore. The evaluation of a large library revealed that internalizationand/or complexation alone is not sufficient for silencing and thatcertain chemical functionalities including tertiary dimethylamineor piperazine groups with potential buffering capacity may be ad-vantageous for polymer-based delivery.

Advances in the controlled synthesis of polymers with func-tional group tolerance [particularly controlled/living radicalpolymerization (CRP) (17–20) methods that enable control overMW and architecture] and in robotic technology (8) motivatedus to explore an experimental design based on modules where96-well plates are shuffled between different HT instrumentsfor synthesis, characterization, and screening. This process allowsfor examination of key physical and chemical properties in rela-tion to performance. siRNA (21, 22) was selected as an initialagent for delivery; despite its marked potential to treat various

diseases, safe and effective delivery remains the most significantbarrier to broad clinical use (23). We previously utilized HTmethods for the synthesis and screening of materials for deliveryincluding lipids (15, 24) for siRNAdelivery and linear poly(β-ami-no ester)s (PBAEs) (14, 25–27) for pDNA delivery. HT synthesisof star polymers (13) and linear polymers using CRP methods(10, 11, 19) has been reported but not applied for gene therapywhere cationic charges are required to bind nucleic acids throughelectrostatic interactions. Modification of linear polymers forpDNA delivery has been described (12, 28); however, to ourknowledge such polymers have not been shown to effectivelydeliver siRNA. There is a growing interest in sub-100 nm particlesfor delivery (4) due to the increased ability of smaller particles toaccess capillaries, penetrate tissues, and cross cell membranes.Because the possible size range for stable liposomes is generally>50 nm (29), and because star shaped polymers (13, 30, 31) anddendrimers (32) are considerably smaller, we selected a core-shellsystem that allows for greater size range control and the ability toeasily incorporate a multitude of different cationic groups intothe core. Distinct polymer-based nanoparticles may have betterstability than lipid-based systems and greater potential for attach-ment of targeting ligands (33).

To address these issues, we prepared a library of nanoparticlescomposed of cationic cross-linked nanogel cores and variable shellswith precise control over particle size, chemical composition, andarchitecture. siRNA can complex to or inside of the core, while thearms provide a protective shield. To achieve the desired structure,we employed a method of cross-linking block copolymers (34, 35)prepared by reversible addition-fragmentation chain transfer(RAFT) polymerization (18). For cross-linking, we sought an effi-cient and versatile reaction that would permit library formation.While other click reactions (36) might have been suitable, we chosering-opening of epoxides with amines (37–39) due to the availabilityof a large number of amines that can incorporate cationic chargeinto the core, and the efficiency of this catalyst-free reaction. Bychanging the physical properties of the core and shell, these parti-cles could also find application in the delivery of proteins and drugs,assembled multilayers (40), electronics (41), as stabilizing particles

Author contributions: D.J.S., R.L., and D.G.A. designed research; D.J.S., K.A.W., L.N., G.S.,H.C., S.J., M.M., A.L.-J., A.V., P.F., C.G.L., K.T.L., H.L., C.C., S.P.C., Y.L., J.J., W.Q., and C.Z.performed research; D.J.S., K.A.W., L.N., G.S., H.C., S.J., M.M., A.L.-J., A.V., C.G.L., K.T.L.,H.L., W.Q., T.N., R.L., and D.G.A. analyzed data; and D.J.S. and D.G.A. wrote the paper.

Conflict of interest statement: R.L. is a shareholder and member of the Scientific AdvisoryBoard of Alnylam. D.G.A. is a consultant with Alnylam. R.L and D.G.A have sponsoredresearch grants from Alnylam. Alnylam also has a license to certain intellectual propertyinvented at MIT. W.Q., C.Z., and T.N. are employed by Alnylam.

*This Direct Submission article had a prearranged editor.1To whom correspondence should be addressed. E-mail: dgander@mit.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1106379108/-/DCSupplemental.

12996–13001 ∣ PNAS ∣ August 9, 2011 ∣ vol. 108 ∣ no. 32 www.pnas.org/cgi/doi/10.1073/pnas.1106379108

Dow

nloa

ded

by g

uest

on

Aug

ust 1

, 202

0

in polymer blends (42), and catalytic scaffolds and supramolecularhosts (43, 44).

Results and DiscussionLibrary Design and Synthesis. The rationale for employing the HTcore-shell design was to create a cationic core to facilitate siRNAcomplexation, with variation in the nature of the protonizable amine,and a shell with variation in polymer length and chemical properties.Although many techniques exist for the formation of nanoparticles(31, 45–48), we requiredmethods that were amenable to parallel HTsynthesis. For this reason, we employed a method of reacting epox-ide groups with amines (37–39, 49–51) that avoids the use of salts,surfactants, and mixed solvents. Instead of using homopolymers andrandom copolymers (38), we chose block copolymers that couldform the core-shell morphology. The basic reaction for the synthesisof core-shell nanoparticles involved block copolymers consisting ofone nonreactive block and one epoxide-functionalized block, suchthat the core could be cross-linked with a variety of amines (Fig. 1).These amines varied greatly in structure (linear, ring, branched),reactivity (1–4 reactive amine groups per molecule; 1°, 2°, and 3°amines within the molecule), and MW (small molecule, dendrimer,and polymer). Each amine could form a distinct and structurally un-ique nanoparticle starting from one block copolymer.

We synthesized a series of different block copolymers to serve asprecursors for nanoparticle formation (Table S1). Block copoly-mers possessing poly(oligo(ethylene oxide) methacrylate) (POEO-MA) with different lengths of the PEO side chain, may increaseblood circulation time due to the PEO shell of the resulting nano-particle. Anionic, cationic, zwitterionic, and hydrophobic blockswere also evaluated as shells. Utilizing the 96-well plate footprint,each block copolymer was reacted with 96 amines to form a set ofcore-shell nanoparticles. The HTapproach enabled randomization

of the core and shell for elucidation of the key chemical andmaterial properties for intracellular delivery. To facilitate rapidsynthesis, characterization, and screening, we developed a methodwhere a microtiter plate was passed through a series of modules(Fig. 2). Periodic analysis by prescreening may lead to an improve-ment in chemical selection. In this study, we optimized the aminelibrary by replacing the poorer performing amines with new can-didate amines during the process. Automation of this process ac-celerated not only material handling, but also data manipulationsand reaction calculations. The MW, density, drawn chemical struc-ture, and dissolved concentration were entered into Library Studioto enable rapid calculation of reaction stoichiometry, taking into ac-count the MW, density, structure (number of amines), and concen-tration. Through preliminary studies, we found that a mole ratio of1∶1 of epoxides to 1° or 2° amines and a concentration in dimethylsulfoxide (DMSO) of 100 mg∕mL (approximately 300 mM)was op-timal for nanoparticle formation with limited macroscopic gelation.

Nanoparticle-Mediated Delivery in Vitro. As an initial screen, weevaluated the ability of every nanoparticle to complex RNA(Fig. 3A). Automated methods were developed that allowed forevaluation of 50 microtiter plates per week. The degree of entrap-ment is expressed as a percentage based on measured fluores-cence, where higher values in the figure indicate strongercomplexation. For nanoparticles with PEO shells (blocks A–J),a range of complexation was observed that predominantly corre-lated with the amine. Structurally analogous amines showed thatmore protonizable nitrogens resulted in better complexation. Forexample, the linear polyamine 301 complexed well in most cases,whereas the linear aliphatic diamine 112 with similar length didnot. Nanoparticles with anionic shells (blocks K and L) repelledthe anionic phosphate backbone of siRNA, resulting in almost

Fig. 1. Combinatorial synthesis of core-shell nanoparticles. (A) Epoxide-containingblock copolymers and amines were used tosynthesize a combinatorial library of hairycore-shell nanoparticles. (B) Synthesis occursvia the ring-opening reaction of epoxides byamines forming β-hydroxyl groups and cross-linking points separated by the functionality(X) contained in the amine. Each nanoparti-cle is named by a letter corresponding to thestarting block copolymer and a number cor-responding to an amine.

Siegwart et al. PNAS ∣ August 9, 2011 ∣ vol. 108 ∣ no. 32 ∣ 12997

APP

LIED

PHYS

ICAL

SCIENCE

S

Dow

nloa

ded

by g

uest

on

Aug

ust 1

, 202

0

no measurable complexation. In contrast, nanoparticles with acationic shell (block M) were able to complex siRNA in everycase, implying that the siRNA is located in the shell (and possiblyin the core) of these nanoparticles, whereas the siRNA is onlycomplexing with the core of the other nanoparticles.

The libraries were then screened for the ability to productivelydeliver siRNA to a HeLa cell line that stably expressed bothfirefly and Renilla luciferase (Fig. 3B). Cells were treated with fire-fly luciferase-targeting siRNA(siLuc)-nanoparticle complexes andthe ratio of firefly to Renilla luciferase expression was measuredafter 24 h. Reduction of only firefly luciferase demonstrates non-cytotoxic specific silencing, whereas reduction in expression of bothluciferase proteins indicates toxicity or other nonspecific effects.Following the modular HTapproach, the complexes were formedin 96-well plates by simple mixing. Because the size, shape, andcharge of the nanoparticles vary greatly in this library, we also eval-uated them for pDNAdelivery. Compared to siRNA, which is rigidand linear, pDNA is a few hundred times longer, circular, andmoreflexible. HeLa cells were transfected with pDNA coding for luci-ferase, and luciferase activity was measured after 24 h (Fig. 3C).

When considering the maximum siRNA knockdown for oneamine core across all block copolymers, those nanoparticles thatwere expected to increase in charge as the pH drops with the transi-tion from outside to inside of the endosome were more likely toeffectively deliver siRNA. We modeled the microspecies pKas ofthe GMA-amine products in the core of the nanoparticles, estimat-ing the relative buffering capacity that may be involved in endo-somal escape (Table S2). The model system calculations suggestthat 62 of the 102 amines will not gain charge during the expecteddrop in pH within the endosomal compartment. The remaining40 amines are expected to exhibit buffering. Across all block types,80 different amines enabled>30% luciferase silencing in vitro. Con-sidering only the criteria used to segregate buffering from nonbuf-fering amines, a significantly higher proportion of the modeledbuffering amines (77%) are represented amongst the effectivegroup than nonbuffering amines (14%) (Fig. S1).

Analysis of the results also revealed several trends regardingthe block copolymer structure in nanoparticles capable of deliver-

ing siRNA in vitro. Considering nanoparticles with POEOMAshells (blocks A–J), nanoparticles formed from block C showedthe highest knockdown. This was likely due to an optimal blockratio with a short POEOMA block [degree of polymerization(DP) 12] and a longer glycidyl methacrylate (GMA) block (DP135). Block D, which also had a higher weight fraction of PGMA,demonstrated effective knockdown, although it was less capableof delivering pDNA. In contrast, nanoparticles formed fromblock H, which possessed a lower weight fraction of POEOMA(DP 48) compared to PGMA (DP 17) were unable to deliversiRNA or pDNA. This trend of weight fraction is likely relatedto resulting charge density, particle formation, and to surfacePEO presentation. J had a long GMA block that resulted insome macroscopic gelation. F, which was a poorly formed blockcopolymer with high polydispersity, was not effective, suggestingthat defined block architecture is important for nanoparticleformation. Many G nanoparticles could deliver siRNA due totheir less dense core (via copolymerization of 2-hydroxyethyl

Fig. 3. In vitro screening of core-shell nanoparticles. (A) The degree of RNAentrapment was quantified using the RiboGreen assay. The average percentof complexation is shown. (B) HeLa cells stably expressing both firefly andRenilla luciferases were treated with firefly targeting siRNA-nanoparticlescomplexes. The average percent reduction in firefly luciferase activity aftertreatment is shown. (C) HeLa cells were treated with pDNA(luc)-nanoparticlecomplexes. Activity using LF2000 was standardized to 100. The average per-cent activity versus LF2000 after treatment is shown. To provide apicture of the capability of all nanoparticles, the complexation, siRNA deliv-ery, and pDNA experiments were performed in triplicate at a weight ratio of50∶1 (nanoparticle:siRNA).

Fig. 2. Modular design for the synthesis, characterization, and screening ofa library of core-shell nanoparticles. (A) The synthesis of 1,536 core-shell na-noparticles was carried out on a Symyx fluid-handling robot inside of glassvials in a 96-well plate format. (B) The nanoparticles were purified and fil-tered through a HT filtering apparatus into a standard 96-well tissue cultureplate. (C) The MW and particle size of the polymers were measured by a HTGPC system. (D) The same plate format was used for an RNA complexationassay, carried out on a Tecan cell culture robot. HT in vitro screens for siRNAand pDNA delivery were performed. (E) Biodistribution of C227 (left twoimages) and C80 (right two images) are shown. GPC, siRNA complexationdata, and in vitro siRNA and pDNA delivery results with standard deviation(s.d.) for nanoparticles C32-C94 (positions A1-A12 in the plate) are presented.

12998 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1106379108 Siegwart et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 1

, 202

0

methacrylate) but were not superior to the best nanoparticlesformed from C. Nanoparticles formed from cationic blocks (Hand M) and anionic blocks (K and L) did not effectively deliversiRNA. Some nanoparticles formed from I could deliver siRNAbut not pDNA, possibly due to the shorter lengths of the blocksand correspondingly smaller size of the nanoparticles. The oppo-site trend was also observed, where some nanoparticles with azwitterionic shell (N) or a hydrophobic shell (O) could deliverpDNA but not siRNA. To a small extent, a few nanoparticles withanionic shells (K and L) mediated transfection of pDNA; how-ever, the anionic nanoparticles were largely ineffectual.

The performance of the top nanoparticles for siRNA deliveryis detailed in Fig. 4A. In parallel experiments, these nanoparticleswere as effective as Lipofectamine 2000 (LF2000), including C77,C80, C206, and C227, which were more effective at the given dosein the presence of serum. The best performing nanoparticles wereable to effectively complex siRNA. Dose response data (Fig. 4B)revealed an increase in gene silencing as the dose increased. Inthe top 40 nanoparticles that silenced >30% of firefly expression,five amines appeared in more than one nanoparticle (77, 80, 94,208, and 222). The top 20 nanoparticles all had amines with 1–2reactive groups. Having more than two reactive amines possiblyled to interparticle cross-linking. E206 was the best nanoparticlefor pDNA due to long block lengths and large size, where the cellsexpressed 625% more luciferase compared to cells transfectedwith LF2000 (Fig. 4C).

To examine the effect of amine structure on cellular internaliza-tion, we quantified the uptake after 1 h of incubation with each ofthe C-based nanoparticles complexed with fluorescent siRNA(Fig. 5) and used image analysis to identify key structural trends.Nanoparticles containing tertiary amines were more effectivelyinternalized than nanoparticles with secondary amines. Within ter-tiary alkyl amines, dimethyl (80) > diethyl (222) > dibutyl (223),and within secondary alkyl amines, methyl (211) > ethyl (210). Inaddition to uptake, siRNA complexation and silencing efficiencywas also improved for nanoparticles with pendant dimethyl tertiaryamines. Similarly, nanoparticles with piperazine (227) and pyrroli-dine (77) structures showed enhanced complexation, delivery, andinternalization, as compared to morpholine (224) and imidazole(94) groups. Although complexation remained high for C224, theinternalization decreased slightly and the siRNA silencing abilitydecreased greatly compared to optimal C227. Increased cellular up-take was observed for linear polyamines with internal methyl sub-stitutions (e.g., 235 versus 234; 127 versus 125). PEG-based diaminecross-linkers (103-1, 206, 207, 208) enabled efficient nanoparticle

formation, siRNA complexation, and high internalization. However,only C206 exhibited high siRNA silencing ability, possibly due thehigher hydrophobicity and methyl group adjacent to the amines.Some nanoparticles (e.g., 207) mediate efficient internalization,but that alone is not sufficient for silencing. In general, incorpora-tion of tertiary amines led to an improvement in complexation,cellular internalization, and siRNA silencing efficiency.

Characterization of the Core-Shell Nanoparticle Library. The entirecore-shell nanoparticle library was characterized using HT GPC.In order to decrease the total analysis time frommonths to weeks,we utilized a GPC equipped with a single flow injection polymeranalysis column and three detectors capable of measuring bothMW and particle size. GPC traces, absolute MW, Rh, and Rg forall nanoparticles appear in Table S3. Preliminary studies were car-ried out before the full library was synthesized. To achieve the great-est flexibility in choice of monomers, we used cumyl dithiobenzoateas the RAFTchain transfer agent (CTA) in most cases. Conditionswere investigated and optimized for the controlled homopolymer-ization of OEOMA and GMA (Table S4 and Fig. S2). The semi-logarithmic plot of lnð½M�0∕½M�Þ versus time for the polymerizationof GMAwas linear, indicating a constant concentration of radicals.Mn also increased linearly with conversion. Chain extension ofOEOMA from PGMA was examined, yielding block copolymerswith a clean GPC shift in MW, and polydispersity below 1.3. GMAwas also polymerized from POEOMAmacro-CTAs. Kinetic studiesshowed trends of a controlled polymerization. In advance of form-ing nanoparticles, model ring-opening reactions with small mole-cules were carried out. At 1∶1 molar equivalents of epoxide toamine, reaction with piperidine and sodium azide resulted in com-plete opening of the epoxides (Fig. S2). To examine the effect ofblock length on nanoparticle formation, we synthesized a seriesof block copolymers that systematically varied in weight % ofOEOMA and GMA blocks (Table S1). These block copolymerswere reacted with different amines, and the particle size of the re-

Fig. 4. In vitro core-shell nanoparticle-mediated delivery trends. (A) RNAcomplexation and luciferase silencing after delivery of nanoparticle:siRNAcomplexes (50∶1, wt∕wt) is expressed for the top performing nanoparticles.(B) siRNA dose response for the top nanoparticles. (C) Complexation andluciferase activity after delivery of nanoparticle:pDNA complexes (50∶1,wt∕wt) is expressed for the top performing nanoparticles. (A–C) n ¼ 4; s.d.is expressed.

Fig. 5. In vitro uptake screening. (A) Cellular internalization of selected na-noparticles after 1 hr of incubation is demonstrated by HT automated con-focal microscopy. HeLa cells were exposed to all C-based nanoparticlescomplexed with Alexa-594 siRNA (50∶1, wt∕wt). Twenty different fields wereimaged and a representative image of nanoparticle/siRNA complex is pre-sented (Alexa 594 is pseudocolored green). Additional images appear inFig. S4. (B) The total fluorescence was quantified and plotted. False positiveswere identified by visible aggregation (e.g., C208 included for reference) orsignificant background and were removed from the graph. Interestingly, C60nanoparticles interact with the cell membrane but did not enter cells, whileC80 nanoparticles were effectively internalized (see zoom insets).

Siegwart et al. PNAS ∣ August 9, 2011 ∣ vol. 108 ∣ no. 32 ∣ 12999

APP

LIED

PHYS

ICAL

SCIENCE

S

Dow

nloa

ded

by g

uest

on

Aug

ust 1

, 202

0

sulting nanoparticles was measured by DLS. Desired particle size(10–100 nm) was obtained for most block copolymers (15–85weight % GMA). Atomic force microscopy (AFM) images of na-noparticles prepared by the reaction of FF with the linear aliphaticdiamine 110 showed uniform particles with a diameter of 34 nm(Fig. 6A). Polyamines were also able to efficiently cross-link theblock copolymers to form the hairy nanoparticle shape. The reac-tion of R or S with 126 formed particles with a diameter of about21 nm, as measured by DLS, with the diameter measured by AFMslightly smaller (22.9 nm) for R126 than for S126 (25.4 nm) due tothe slightly shorter GMA block (Fig. S3). To verify the core-shellstructure, transmission electron microscopy (TEM) was employed.An examination of R126 (Fig. 6B) and C227 nanoparticles(Fig. S3E) revealed a higher electron density in the core, comparedwith the shell, indicating the core-cross-linked morphology. Themeasured diameter of C80 and C227 nanoparticles by DLS andTEM was in agreement (Table S5 and Fig. S3). In addition to theHT internalization studies, we examined cellular internalizationby synthesizing FITC-labeled nanoparticles via reaction of the iso-thiocyanate group with dangling amines to covalently label thecore. At higher magnification, C80-FITC and C227-FITC can beclearly visualized inside endosomes that are the hallmark for endo-cytic trafficking (Fig. 6 C and D). Because size and surface chargecan also affect cellular uptake, the diameter and zeta potential oftop performing nanoparticles were also measured (Table S5).

In Vivo Core-Shell Nanoparticle-Mediated Delivery of siRNA to Hepa-tocytes in Mice. Based on results from the in vitro screen, weexamined the utility of top performing nanoparticles in facilitat-ing siRNA delivery in vivo by employing the mouse Factor VIIgene silencing model (15, 24). Factor VII is a useful gene targetfor the evaluation of hepatocyte-specific delivery because it isonly produced in the cells of the liver parenchyma, possesses arelatively short plasma half-life, and is secreted into the blood,enabling facile protein quantification. siRNA-directed againstthe blood clotting Factor VII was complexed with nanoparticlesat a weight ratio of 10∶1 (NP:siRNA) and delivered intrave-nously. Mouse body weight was monitored over the duration ofthe experiment because body weight loss can indicate toxicityassociated with nanoparticle treatment.

Since in vivo delivery to hepatocytes poses more challengesthan the in vitro delivery to HeLa cells, we investigated theinclusion of cholesterol because covalent attachment of choles-terol to siRNA has been shown to improve in vivo delivery tothe liver (52, 53). Of all the nanoparticles tested, C80 showedthe most improvement after cholesterol attachment and was fo-cused on for this study. 10 Mole % of the epoxide groups werereacted with N 0-cholesteryloxycarbonyl-1,2-diaminoethane (318),followed by addition of 80 to stoichiometrically cross-link the re-maining epoxide groups. Covalent attachment of cholesterol toC80 nanoparticles enabled 40% silencing of Factor VII (Fig. 7A).Noncovalent encapsulation of unmodified cholesterol had nomeasured effect. Delivery of control siGFP using these nanopar-ticles did not result in any measurable silencing, suggestingspecific delivery to the liver and silencing of Factor VII. No ap-preciable weight loss was observed, indicating minimal toxicity.To probe the biodistribution of these nanoparticles with andwithout cholesterol in vivo, a single bolus dose of 1 mg∕kg Cy5-

labeled siRNA-nanoparticle complex was administered via tailvein injection (Fig. 7 B and C). C80-Cy5-siRNA nanoparticleswere observed in the liver, kidneys, and lungs. The attachment ofcholesterol resulted in increased liver accumulation.

We believe that the development of this library of core-shell na-noparticles represents an important step forward for the HTsynth-esis of polymers for intracellular delivery. We aimed to movebeyond simple linear polymers by exploring CRP for the formationof complex architectures with great chemical diversity in the coreand shell. This synthetic method enables parallel generation oflarge libraries of chemically distinct core-shell materials and de-creases the time from design to analysis. The common structuralfeatures of these materials suggest certain design criteria for creat-ing future nanoparticles, including (i) a block weight ratio around15% nonreactive block and 85% reactive block, (ii) a stoichiometicbalance between epoxides and amines, (iii) thin water-soluble hy-drophilic shells, (iv) tertiary amines with 1–2 reactive sites, and (v)amines with buffering capacity. Covalent attachment of cholesterolto the nanoparticle was required for in vivo delivery to the liver.Further studies are warranted to extend this technology for thebroadest applications of RNAi therapy and drug delivery. We alsodesire to use what we have learned in this study to aid in the designof fully degradable nanoparticles that satisfy the aforementionedgeneral properties and design criteria. Finally, we believe that themodular design utilized in this report is applicable to many othersystems. By changing the composition of the blocks and cross-linkers, a variety of other nanoparticles could be formed withwide-ranging applications.

MethodsHT Nanoparticle Synthesis. Robotic synthesis was performed on a customizedSymyx Core Module equipped with four different liquid dispensing elements(a single tip, a motor driven gripper with a heated piercing tip, a positivedisplacement tip, and a heated parallel 4-tip), heating and cooling bay ele-ments, a magnetic stirring element, a mass balance, and imaging equipment.Library Studio (Symyx Discovery Tools) was used to design the core-shell na-

Fig. 6. Characterization of core-shell nanoparticles.(A) Dried FF110 nanoparticles on mica showed uniformparticle size around 34nmbyAFM. (B) TEMwas utilizedto visualize the core-shell structure of R126 nanoparti-cles. The insert clearly showed an electron dense cross-linked core, and amore electron loose shell. These TEMimages are unstained. Additional AFM and TEM imagesappear in Fig. S3. Cell internalization of FITC-C80 (C)and FITC-C227 (D) nanoparticles after 1 hr of incubationwas demonstrated by confocal microscopy. Punctategreen fluorescence was observed within the cell.

Fig. 7. In vivo silencing of Factor VII in liver hepatocytes. Nanoparticles werepurified by dialysis, complexed with siRNA at a weight ratio of 10∶1 (nano-particle:siRNA), and delivered intravenously to C57BL/6 mice. Mice received asingle bolus administration of 5 mg∕kg total siRNA via tail-vein injection andFactor VII levels were quantified 48 hr post injection. (A) Covalent attachmentof cholesterol to C80 nanoparticles enabled silencing. Noncovalent encapsu-lation of cholesterol and delivery of control siGFP resulted in no knockdown.Data points represent group mean� s:d. Data points marked with asterisksare significant relative to control treated groups (***, P < 0.0001; t-test, dou-ble-tailed, n ¼ 14). Representative images of isolated organs for Cy5-siRNA-C80 (b) and Cy5-siRNA-C80-10 mol% cholesterol (C) nanoparticle complexesdemonstrated nanoparticle accumulation in the liver, kidneys, and lungsafter 120 min. The covalent attachment of cholesterol to C80 resulted in en-hanced liver accumulation and reduced lung accumulation.

13000 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1106379108 Siegwart et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 1

, 202

0

noparticle libraries. For each nanoparticle, a solution of amine (100 mg∕mL inDMSO) was added to a stirring solution of block copolymer (100 mg∕mL inDMSO) such that the mole ratio of epoxides to amine was 1∶1 and the totalconcentration of reactive components was on the order of 300 mM. Reactionblocks with 96 1 mL glass vials equipped with tumbling stir bars and coveredwith a rubber mat were used. The resulting mixtures were stirred and heatedat 50 °C for 24 h on the deck of the robot.

HT RNA Complexation Assay. The Quant-iT RiboGreen RNA assay (Invitrogen)was performed on a customized Tecan Freedom EVO 200 with 8-Tip, LiHa,MCA-96, and Extended RoMa elements. Fluorescence was measured usinga coupled Infinite M200 multimode monochromator-based reader. The assaywas programmed in EVOware Plus.

HT Confocal Microscopy. 15,000 HeLa cells were plated in black 96-well plates(Greiner Bio-one). Cells were starved for 30 min in the presence of PBS/BSAand pulsed for 1 h at 37 °C with 50 ng of BLOCK-iT Alexa Fluor Red Fluores-cent Oligo (Invitrogen) complexed with the nanoparticle at a nanoparticle/siRNA ratio of 50∶1 (wt∕wt). Cells were chased for 30 min in the presenceof unlabeled nanoparticles and washed with PBS. Cells were then fixed using4% paraformaldehyde and counterstained in PBS containing Hoescht

(2 μg∕mL) for nuclei identification. Cells were then imaged with an auto-mated spinning disk confocal microscope (OPERA, Perkin Elmer). Quantita-tion of internalized fluorescently labeled siRNA was done using Acapellasoftware. After identification of cell location and perimeter, intracellular siR-NA signal intensity over the whole search region (single field) was calculated.Data presented are an average of intracellular intensity from 20 differentfields plotted in Excel. The same defined pattern of fields from eachwell were acquired to eliminate bias during analysis. Quality control wasperformed on all images and large aggregates or nonspecific backgroundseen in rare fields was excluded from analysis.

Supporting Information. Additional descriptions of materials, instrumenta-tion, modeling, and experimental details, including the synthesis of all blockcopolymers, are available in SI Text.

ACKNOWLEDGMENTS. The authors thank Boris Klebanov, Manos Karagiannis,Christian J. Kastrup, Avi Schroeder, Carmen Barnes, and Robert S. Siegwartfor their help. This work was supported by Alnylam Pharmaceuticals andNational Institutes of Health (NIH) Grants R01-EB000244-27 and 5-R01-CA132091-04. D.J.S. acknowledges postdoctoral support from NIH NRSAaward F32-EB011867.

1. Balazs A, Emrick T, Russell T (2006) Nanoparticle polymer composites: Where twosmall worlds meet. Science 314:1107–1110.

2. Taton T, Mirkin C, Letsinger R (2000) Scanometric DNA array detection with nano-particle probes. Science 289:1757–1760.

3. Shipway A, Katz E, Willner I (2000) Nanoparticle arrays on surfaces for electronic,optical, and sensor applications. ChemPhysChem 1:18–52.

4. LaVan D, Lynn D, Langer R (2002) Moving smaller in drug discovery and delivery. NatRev Drug Discov 1:77–84.

5. Davis ME, Chen Z, Shin DM (2008) Nanoparticle therapeutics: An emerging treatmentmodality for cancer. Nat Rev Drug Discov 7(9):771–782.

6. Geysen H, Schoenen F, Wagner D, Wagner R (2003) Combinatorial compound librariesfor drug discovery: An ongoing challenge. Nat Rev Drug Discov 2:222–230.

7. Brocchini S, James K, Tangpasuthadol V, Kohn J (1997) A combinatorial approachfor polymer design. J Am Chem Soc 119:4553–4554.

8. Boussie T, et al. (2003) A fully integrated high-throughput screening methodology forthe discovery of new polyolefin catalysts: Discovery of a new class of high temperaturesingle-site group (IV) copolymerization catalysts. J Am Chem Soc 125:4306–4317.

9. Xiang XD, et al. (1995) A combinatorial approach to materials discovery. Science268:1738–1740.

10. Hoogenboom R, Meier M, Schubert U (2003) Combinatorial methods, automatedsynthesis and high-throughput screening in polymer research: Past and present.Macromol Rapid Commun 24:16–32.

11. Webster D (2008) Combinatorial and high-throughput methods in macromolecularmaterials research and development. Macromol Chem Phys 209:237–246.

12. Wong S, Sood N, Putnam D (2009) Combinatorial evaluation of cations, pH-sensitiveand hydrophobic moieties for polymeric vector design. Mol Ther 17:480–490.

13. Bosman AW, et al. (2001) High-throughput synthesis of nanoscale materials: Structuraloptimization of functionalized one-step star polymers. J Am Chem Soc 123:6461–6462.

14. Anderson DG, Lynn DM, Langer R (2003) Semi-automated synthesis and screeningof a large library of degradable cationic polymers for gene delivery. Angew ChemInt Ed Engl 42:3153–3158.

15. Akinc A, et al. (2008) A combinatorial library of lipid-like materials for delivery of RNAitherapeutics. Nat Biotechnol 26:561–569.

16. Hook A, et al. (2010) High throughput methods applied in biomaterial developmentand discovery. Biomaterials 31:187–198.

17. Braunecker WA, Matyjaszewski K (2007) Controlled/living radical polymerization:Features, developments, and perspectives. Prog Polym Sci 32:93–146.

18. Moad G, Rizzardo E, Thang S (2005) Living radical polymerization by the RAFT process.Aust J Chem 58:379–410.

19. Zhang H, Marin V, Fijten M, Schubert U (2004) High-throughput experimentation inatom transfer radical polymerization: A general approach toward a directed designand understanding of optimal catalytic systems. J Polym Sci A PolymChem 42:1876–1885.

20. Matyjaszewski K, Tsarevsky N (2009) Nanostructured functional materials prepared byatom transfer radical polymerization. Nat Chem 1:276–288.

21. Fire A, et al. (1998) Potent and specific genetic interference by double-stranded RNA inCaenorhabditis elegans. Nature 391:806–811.

22. Elbashir SM, et al. (2001) Duplexes of 21-nucleotide RNAs mediate RNA interferencein cultured mammalian cells. Nature 411:494–498.

23. Whitehead K, Langer R, Anderson D (2009) Knocking down barriers: Advances insiRNA delivery. Nat Rev Drug Discov 8:129–138.

24. Love K, et al. (2010) Lipid-like materials for low-dose, in vivo gene silencing. Proc NatlAcad Sci USA 107:1864–1869.

25. Lynn DM, Anderson DG, Putnam D, Langer R (2001) Accelerated discovery of synthetictransfection vectors: Parallel synthesis and screening of degradable polymer library.J Am Chem Soc 123:8155–8156.

26. Anderson D, et al. (2004) A polymer library approach to suicide gene therapy forcancer. Proc Natl Acad Sci USA 101:16028–16033.

27. Green J, Langer R, Anderson D (2008) A combinatorial polymer library approach yieldsinsight into nonviral gene delivery. Acc Chem Res 41:749–759.

28. Sun B, Liu X, Buck M, Lynn D (2010) Azlactone-functionalized polymers as reactivetemplates for parallel polymer synthesis: Synthesis and screening of a small libraryof cationic polymers in the context of DNA delivery. Chem Commun 46:2016–2018.

29. Nagayasu A, Uchiyama K, Kiwada H (1999) The size of liposomes: A factor whichaffects their targeting efficiency to tumors and therapeutic activity of liposomalantitumor drugs. Adv Drug Delivery Rev 40:75–87.

30. Zheng G, Pan C (2006) Reversible addition-fragmentation transfer polymerization innanosized micelles formed in situ. Macromolecules 39:95–102.

31. Gao H, Matyjaszewski K (2009) Synthesis of functional polymers with controlledarchitecture by CRP of monomers in the presence of cross-linkers: From stars to gels.Prog Polym Sci 34:317–350.

32. Lee C, MacKay J, Frechet J, Szoka F (2005) Designing dendrimers for biologicalapplications. Nat Biotechnol 23:1517–1526.

33. Davis M, et al. (2010) Evidence of RNAi in humans from systemically administeredsiRNA via targeted nanoparticles. Nature 464:1067–U1140.

34. Chen D, Peng H, Jiang M (2003) A novel one-step approach to core-stabilized nano-particles at high solid contents. Macromolecules 36:2576–2578.

35. Peng H, Chen D, Jiang M (2005) A one-pot approach to the preparation of organiccore-shell nanoobjects with different morphologies. Macromolecules 38:3550–3553.

36. Kolb HC, Finn MG, Sharpless KB (2001) Click chemistry: Diverse chemical function froma few good reactions. Angew Chem Int Ed Engl 40:2004–2021.

37. Tsarevsky NV, Bencherif SA, Matyjaszewski K (2007) Graft copolymers by a combinationof ATRP and two different consecutive click reactions. Macromolecules 40:4439–4445.

38. van der Ende A, Kravitz E, Harth E (2008) Approach to formation of multifunctionalpolyester particles in controlled nanoscopic dimensions. J Am Chem Soc 130:8706–8713.

39. Kurkuri MD, et al. (2009) Multifunctional polymer coatings for cell microarrayapplications. Biomacromolecules 10:1163–1172.

40. Kim BS, Gao HF, Argun AA, Matyjaszewski K, Hammond PT (2009) All-star polymermultilayers as pH-responsive nanofilms. Macromolecules 42:368–375.

41. Gao H, et al. (2010) Site isolation of emitters within cross-linked polymer nanoparticlesfor white electroluminescence. Nano Lett 10(4):1440–1444.

42. Stenzel M (2009) Hairy Core-Shell Nanoparticles via RAFT: Where are the opportunitiesandwhere are the problems and challenges?Macromol Rapid Commun 30:1603–1624.

43. Hecht S, Frechet J (2001) Dendritic encapsulation of function: Applying nature’s siteisolation principle from biomimetics to materials science. Angew Chem Int Ed Engl40:74–91.

44. Rodionov V, et al. (2010) Easy access to a family of polymer catalysts frommodular starpolymers. J Am Chem Soc 132:2570–2572.

45. Oh J, Drumright R, Siegwart D, Matyjaszewski K (2008) The development of microgels/nanogels for drug delivery applications. Prog Polym Sci 33:448–477.

46. O’Reilly R, Hawker C, Wooley K (2006) Cross-linked block copolymer micelles: Func-tional nanostructures of great potential and versatility. Chem Soc Rev 35:1068–1083.

47. Zhang K, Fang H, Wang Z, Taylor J, Wooley K (2009) Cationic shell-crosslinkedknedel-like nanoparticles for highly efficient gene and oligonucleotide transfectionof mammalian cells. Biomaterials 30:968–977.

48. Matsumoto S, et al. (2009) Environment-responsive block copolymer micelles with a dis-ulfide cross-linked core for enhanced siRNA delivery. Biomacromolecules 10:119–127.

49. Yin X, Stover H (2003) Hydrogel microspheres by thermally induced coacervation ofpoly(N,N-dimethylacrylamide-co-glycidyl methacrylate) aqueous solutions. Macromo-lecules 36:9817–9822.

50. Suzuki D, Kawaguchi H (2005)Modification of gold nanoparticle composite nanostruc-tures using thermosensitive core-shell particles as a template. Langmuir 21:8175–8179.

51. Hantzschel N, Zhang F, Eckert F, Pich A, Winnik M (2007) Poly(N-vinylcaprolactam-co-glycidyI methacrylate) aqueous microgels labeled with fluorescent LaF3 : Eu nano-particles. Langmuir 23:10793–10800.

52. Lorenz C, Hadwiger P, John M, Vornlocher H, Unverzagt C (2004) Steroid and lipidconjugates of siRNAs to enhance cellular uptake and gene silencing in liver cells.Bioorg Med Chem Lett 14:4975–4977.

53. Wolfrum C, et al. (2007) Mechanisms and optimization of in vivo delivery of lipophilicsiRNAs. Nat Biotechnol 25:1149–1157.

Siegwart et al. PNAS ∣ August 9, 2011 ∣ vol. 108 ∣ no. 32 ∣ 13001

APP

LIED

PHYS

ICAL

SCIENCE

S

Dow

nloa

ded

by g

uest

on

Aug

ust 1

, 202

0