P

Pn

LGa

b

a

ARRAA

KBNIA

1

abbi

Pg�hibtdTadmNp

0h

International Journal of Pharmaceutics 460 (2014) 73– 82

Contents lists available at ScienceDirect

International Journal of Pharmaceutics

j o ur nal ho me page: www.elsev ier .com/ locate / i jpharm

harmaceutical Nanotechnology

oly(�-benzyl-l-glutamate)-PEG-alendronate multivalentanoparticles for bone targeting

aura de Miguela,∗, Magali Noiraya, Georgiana Surpateanub, Bogdan I. Iorgab,illes Ponchela

Univ. Paris Sud, UMR CNRS 8612, Institut Galien, 92296 Châtenay-Malabry Cedex, FranceInstitut de Chimie des Substances Naturelles, CNRS UPR 2301, Centre de Recherche de Gif-sur-Yvette, 1 Avenue de la Terrasse, 91198 Gif-sur-Yvette, France

r t i c l e i n f o

rticle history:eceived 29 August 2013eceived in revised form 21 October 2013ccepted 28 October 2013vailable online 6 November 2013

eywords:

a b s t r a c t

Hydroxyapatite (HAP), a highly specific component of bone tissue, is the main target in order to impartosteotropicity. Bone targeted nanoparticles can increase the strength of the interaction with HAP throughmultivalency and thus constitute a valuable strategy in the therapeutics of skeletal diseases. PBLG10k-b-PEG6k-alendronate nanoparticles (∼75 nm) were prepared by a simple nanoprecipitation method. Thecalcium affinity (KCa

+2 = 1.8 × 104 M−1) of these nanoparticles was evaluated using isothermal titrationcalorimetry. The multivalent interaction with HAP surfaces (KHAP) was studied by fluorescence and was

10 −1

one targetinganoparticles

sothermal titration calorimetrydsorption

estimated to be 1.1 × 10 M , which is more than 4000 times stronger than the reported monovalentinteraction between alendronate and HAP surfaces. Molecular modeling suggests that the number ofbinding sites available at the HAP surface is in large excess than what is required for the whole surfacecoverage by alendronate decorated nanoparticles. The lower calcium affinity of these nanoparticles thanfor HAP allows calcium bound nanoparticles to interact with HAP, which yields a deeper understandingof bone targeted carriers and could potentially improve their bone targeting properties.

© 2013 Elsevier B.V. All rights reserved.

. Introduction

Multivalent interactions, simultaneous binding of multiple lig-nds into one biological entity are involved in many different

iological processes. Once the recognition has been established, theinding strength will depend on the possible number of individual

nteractions and will be enhanced compared to a single monovalent

Abbreviations: HAP, hydroxyapatite; PBLG, poly(�-benzyl-l-glutamate);EG, polyethylene glycol; DMF, dimethylformamide; BLG-NCA, �-benzyl-l-lutamate-N-carboxylic anhydride; TFA, trifluoroacetic acid; MeO-PEG5k-NH2,-methoxy-�-amino poly(ethylene glycol)5000, NHSPEG6k-NHS, �,�-Bis-N-ydroxysuccinimide-poly(ethylene glycol)6000; bnz, benzyl; FITC, fluorescein

sothiocyanate; NHS, N-hydroxysuccinimide; DCTB, trans-2-[3-(4-tert-utylphenyl)-2-methyl-2-propenylidene]malonitrile; CF3COOK, potassiumrifluoroacetate; EtOH, ethanol; DEE, diethyl ether; MeOH, methanol; DMSO,imethyl sulfoxide; TEA, triethylamine; DCC, N,N′-dicyclohexylcarbodiimide;HF, tetrahydrofuran; COOH, carboxylic acid; NCA, N-carboxylic anhydride; ATR,ttenuated total reflection; 1H NMR, proton nuclear magnetic resonance; CDCl3,euterated chloroform; TEM, transmission electron microscopy; MALDI-TOF,atrix-assisted laser desorption/ionization time-of-flight mass spectrometry;MR, nuclear magnetic resonance; ITC, isothermal titration calorimetry; PBS,hosphate buffer saline; UV, ultraviolet; DPn, degree of polymerization.∗ Corresponding author. Tel.: +33 1 46 83 57 12; fax: +33 1 46 83 57 12.

E-mail address: laura.de-miguel@u-psud.fr (L. de Miguel).

378-5173/$ – see front matter © 2013 Elsevier B.V. All rights reserved.ttp://dx.doi.org/10.1016/j.ijpharm.2013.10.048

interaction (Mammen et al., 1998). As a result, site specific drugdelivery is expected to increase therapeutic concentrations locally,simultaneously minimizing dose and side effects. For this purpose,nanoparticles can display multivalency through the conjugation ofa discrete number of targeting ligands to their surface in order toenhance their target affinity (Carlson et al., 2007; Kiessling et al.,2000).

The design of bone targeted systems is of great importance inthe therapeutics of skeletal diseases. HAP, (Ca10(PO4)6(OH)2), isspecific to bone – except for teeth and pathological calcificationsand thus constitutes an ideal target for the design of osteotropicsystems. Biphosphonates as well as other molecules such as acidicoligopeptides or tetracyclines have shown to bind effectively toHAP (Wang et al., 2005). Biphosphonate structure presents twophosphonate groups sharing a common carbon atom (P C P),which are responsible for the high affinity for HAP; the mostimportant interactions involving the calcium ions of HAP. For asystemically administered system to attain the bone mineralizedtissue, it has to cross the blood-bone barrier (Shea and Miller,2005) including the fenestrated capillaries with pore sizes up to

80 nm (Howlett et al., 1984; Shea and Miller, 2005). Biphospho-nates have been largely used to confer osteotropic properties todrugs by conjugation to small molecules (Hirabayashi et al., 2001),to linear macromolecular carriers (Wang et al., 2005), to proteins

7 urnal of Pharmaceutics 460 (2014) 73– 82

(aad

(abmcti

cntiw

2

2

(rcgafCS

M

2

oP-wwwbPsdDcasTpoit

aamwoDo

4 L. de Miguel et al. / International Jo

Gittens et al., 2005) or to nanoparticles (Park et al., 2003; Rossnd Roeder, 2011). However, only little research on the bindingffinities of these systems with bone tissue or HAP has been con-ucted so far.

In this work we present nanoparticles prepared from poly�-benzyl-l-glutamate) (PBLG), a synthetic polypeptide that candopt rigid �-helix structures and shows biocompatibility and goodiodegradability due to the degradable amide bond in its poly-eric structure (Oh et al., 1995). Homopolymers and amphiphilic

opolymers of PBLG can assemble into small nanoparticles (lesshan 80 nm in diameter) by a simple nanoprecipitation method ast has been shown in our group (Barbosa et al., 2007).

The aim of this study was to prepare bone targeted nanoparti-les under 80 nm derived from PBLG, PBLG10k-b-PEG6k-alendronateanoparticles, and to evaluate their binding affinity with HAP crys-als and with calcium ions, which are mostly responsible for thenteraction with HAP, deepening into the interaction of alendronate

ith HAP.

. Materials and methods

.1. Materials

DMF extradry Acroseal (99.8%) and benzylamine Acroseal+99.5%) and BLG-NCA were purchased from Acros, and IsoChemespectively and were used as received. TFA (99%) was pur-hased from Sigma–Aldrich. All other regeants were of analyticalrade and used directly. MeO-PEG5k-NH2 (MW PEG = 5000 g/mol)nd �,�-Bis-NHS-PEG6k (MW PEG = 6000 g/mol) were purchasedrom Iris Biotech GmbH. Alendronate was purchased fromhemos GmbH. DCTB and CF3COOK were purchased fromigma–Aldrich.

ethods

.2. Synthesis of PBLG-polymers

Different derivatives of PBLG were synthesized by ringpening polymerization of BLG-NCA: PBLG25k-bnz, PBLG10k-bnz,BLG40k-b-PEG5k, PBLG40k-b-PEG6k-alendronate, PBLG10k-b-PEG6kalendronate and PBLG40k-FITC. Briefly, n milimoles of BLG-NCAere weighed in a glove box under inert atmosphere (argon) andere dissolved in DMF at a concentration of 0.5 M. The solutionas stirred for 10 min and the initiator, a solution in DMF of

enzylamine or previously dried MeO-PEG5k-NH2 in the case ofBLG40k-b-PEG5k, was added with an argon-purged syringe. Theolution was stirred at 30 ◦C and bubbled with argon several timesuring 5–7 days. Polymers were obtained by precipitation in coldEE and dried under vacuum. The evolution of the reaction wasontrolled by infrared spectroscopy, by following the disappear-nce of BLG-NCA peaks and the appearance of those of PBLG. Aecond purification step, consisting of dissolving the precipitate inHF, reprecipitating in cold DEE and drying under vacuum, waserformed. For copolymers containing the hydrophilic PEG blockr the FITC fluorophore, the purification step also included wash-ng the precipitates three times with MeOH in order to eliminatehe excess of unreacted PEG or FITC.

The synthesis of PBLG-b-PEG6k-alendronate copolymer waschieved based on a slightly modified carbodiimide chemistrypproach (Ozcan et al., 2011) and it is detailed in Supporting Infor-ation. �,�-Bis-NHS-PEG6k was dissolved in DMSO and reacted

ith a solution of alendronate in water with 3.5 equivalents

f TEA at a ratio of 1/1.5 overnight at 40 ◦C (reaction mediumMSO/water 60/40). The reaction was precipitated in a mixturef EtOH/DEE, dialyzed with a membrane of a molecular cut-off

Fig. 1. Structure of the PBLG-b-PEG6k-alendronate copolymer.

of 3500 Da during 72 h to eliminate unreacted alendronate andlyophilized. A further activation of the COOH groups was con-ducted in anhydrous DMF, in the presence of DCC (5 equivalents)and NHS (5 equivalents), overnight under argon atmosphere at25 ◦C. After filtration with 0.45 �m filters, it was precipitated inDEE and dried under vacuum. Next, reaction between NHS-PEG-alendronate and the amine groups of the PBLG block, synthesizedusing the benzylamine as initiator molecule as described above, at a[NHS-PEG-alendronate]/[PBLG] molar ratio of 1.5 was performed inanhydrous DMF at 40 ◦C for 24 h. The product was purified by pre-cipitating in DEE and washed three times with MeOH to eliminatethe any unreacted PEG (Fig. 1).

For the PBLG-FITC polymers, close to the end of the reaction, asolution containing three times more moles of FITC than of the ini-tiator (benzylamine) was added and after 24 h, PBLG-FITC polymerwas obtained by precipitation as described above (Segura-Sanchezet al., 2010) (see Supplementary Information).

2.3. Characterization of PBLG polymers

2.3.1. Fourier transform infrared spectroscopy (FTIR)Infrared measurements were performed on a Fourier transform

Perkin-Elmer 1750 infrared spectrometer using the ATR system toconfirm the absence of NCA auto-polymerization, to follow theevolution of the polymerization reactions and to determine thesecondary structure of the polymers.

2.3.2. Nuclear magnetic resonance (NMR)1H NMR spectra were recorded on a Bruker AC 300 spectrome-

ter. 1H NMR spectra of the polymers were recorded in CDCl3 + 15%TFA. For the PBLG40k-b-PEG5k block copolymer, where polymeriza-tion initiator was the mPEG-NH2, the molar mass of the polymercould be determined by integration of the benzyl protons belong-ing to the PBLG block and the ethylene protons of the PEG block andtaking into account the mass of the initiator. For the PBLG-b-PEG6k-alendronate copolymers, the efficacy of the reaction between thetwo blocks could be determined. Phosphorus nuclear magneticresonance (31P NMR) spectra were recorded on a Bruker AC 300spectrometer for the PEG-alendronate to quantify the alendronateusing KH2PO4 as an internal standard.

2.3.3. Matrix-assisted laser desorption/ionization time-of-flight(MALDI-TOF) mass spectrometry

A Voyager DE-STR MALDI-TOF mass spectrometer (AB Sciex, LesUlis, France), equipped with a 337-nm pulsed nitrogen laser (20 Hz)and an Acqiris® 2 GHz digitizer board, was used for all experi-ments. Mass spectra were obtained in linear positive ion mode withthe following settings: accelerating voltage 20 kV, grid voltage 75%of accelerating voltage, extraction delay time of 150 ns. The laser

intensity was set just above the ion generation threshold to obtainpeaks with the highest possible signal-to-noise (S/N) ratio with-out significant peak broadening. All data were processed using theData Explorer software package (AB Sciex). DCTB was used as the

urnal

mc

2

i5wviwENa

tc�Pcad

N

wstrh

2

2

nHnP6iwtw5oc

2

cPHpbefptBmptfo

L. de Miguel et al. / International Jo

atrix for MALDI-TOF experiments and CF3COOK was used as theationizing agent.

.4. Nanoparticle preparation and characterization

Nanoparticles were prepared following a modified nanoprecip-tation method previously described (Barbosa et al., 2007). Briefly,

mL THF solutions of the pure or mixtures of PBLG derivativesere added dropwise to 10 mL water and stirred for 10 min. Sol-

ents were evaporated by a standardized protocol under vacuumn the rotavapor at 40 ◦C. Nanoparticles size and zeta potential

ere determined by DLS using a Zetasizer 4, Malvern Instruments.xperiments and measurements were always made in triplicate.anoparticles were observed by means of TEM at 120 kV with neg-tive staining.

For the alendronate decorated fluorescently labeled nanopar-icles, the number of PBLG �-helices forming a nanoparticle wasalculated as the ratio between the nanoparticle volume and PBLG-helix volume. Nanoparticles were considered to be spheres andBLG �-helices were considered to be rods, whose lengths werealculated by taking into account the projected segment length of

single aminoacid unit (=0.15 nm), the number of residues, and theiameter of 1.6 nm (Klok et al., 2000).

˛helix/np = 4r3np

3(0.75 h˛helixA r2˛helixA

+ 0.25 h˛helixB r2˛helixB

)(1)

here rnp is the radius of nanoparticles obtained from TEM mea-urements from 100 nanoparticles, r˛helixA and h˛helixA arehe radius and the length of the PBLG-b-PEG-alendronate helicesespectively, and r˛helixB, h˛helixB are those for the PBLG-FITCelices.

.5. Multivalent binding affinity for HAP surfaces

.5.1. In vitro HAP binding assayThe multivalent interactions of PBLG10k-b-PEG6k-alendronate

anoparticles and PBLG40k-PEG6k-alendronate nanoparticles withAP were studied by means of fluorescence. Multifunctionalanoparticles containing a PBLG-derivate and a fluorescent one,BLG-FITC, at a ratio of 3:1 wt in 0.1% of poloxamer Pluronic® F8 were incubated in a HAP suspension in PBS at a pH of 7.4 dur-

ng 1 and 18 h. The suspensions were centrifuged at 3000 rpm andere observed under UV light (� = 365 nm). Pellets were washed

hree times with 1% poloxamer in PBS at pH 7.4. Supernatantsere quantified by a Perkin Elmer Luminescence spectrometer LS

0B at room temperature (�excitation = 495 nm, �emission = 525 nm) inrder to determine the amount of nanoparticles bound to HAP. Aalibration curve was plotted for each type of nanoparticle.

.5.2. Binding isothermsDifferent concentrations of multifunctional nanoparticles

ontaining PBLG10k-b-PEG6k-alendronate or PBLG10k-bnz, andBLG40k-FITC in 0.1% of Pluronic® F 68 were incubated in aAP powder suspension in PBS at a pH of 7.4 during 18 h androceeded as described above. Binding isotherms of PBLG10k--PEG6k-alendronate in the presence and absence of CaCl2 (10quivalents of CaCl2 per equivalent of alendronate, pre-incubatedor 3 h prior to the assay) were plotted as the mass of nanoparticleser mass of HAP added versus the initial nanoparticle concentra-ion. PBLG10k-bnz nanoparticles were used as a negative control.inding isotherms could be described by Langmuir isotherm. Lang-uir isotherm assumes an ideal model where adsorption takes

lace at specific homogenous sites within the adsorbent, no interac-ions existing between adsorbed molecules and only a monolayer isormed at the maximum adsorption, molecules of adsorbate beingnly adsorbed on the free surface of the adsorbant (Langmuir,

of Pharmaceutics 460 (2014) 73– 82 75

1916). The binding data were plotted and expressed in concen-tration as a function of alendronate, PBLG10k-b-PEG6k-alendronatecopolymer and number of nanoparticles considered as multivalentobjects, since a single nanoparticle could present simultaneouslymany alendronate molecules, all of them being available for inter-action with HAP. Data was fitted by the Langmuir linear regressionmethod, as follows:

C

Q= C

Qmax+ 1

KHAPQmax(2)

where Q is the mass of alendronate or PBLG10-PEG-alendronatecopolymer or number of nanoparticles bound per mass of HAP,Qmax is the maximum surface binding, KHAP is the multivalent equi-librium binding constant (M−1) and C the initial concentration ofnanoparticles. The plot of 1/Q versus 1/C allowed to determine theQmax and the KHAP, expressed either in alendronate, in PBLG10k-b-PEG6k-alendronate copolymer or in PBLG10k-PEG6k-alendronatenanoparticles. The concentration of alendronate available innanoparticles was calculated taking into account the amount ofalendronate estimated by NMR.

When binding data were expressed as a function of the numberof particles, the latter was calculated assuming that nanoparticleswere spheres and the density of PBLG lyophases available in theliterature was taken into account (1.271 g/cm3) (Shiau and Labes,1989) as follows:

Nnp = MPBLG

dPBLG4/3�r3(3)

where MPBLG is the molecular weight of PBLG estimated by MALDI-TOF, dPBLG is the density of PBLG and r3 is the hydrodynamic radio ofPBLG10k-b-PEG6k-alendronate nanoparticles measured by dynamiclight scattering. To calculate the affinity constant (KHAP) in M−1

expressed in nanoparticles, the number of Avogadro (NA) was used.

2.6. Binding affinity for calcium ions

The interaction of alendronate displayed by PBLG10k-b-PEG6k-alendronate nanoparticles with calcium ions was studied by ITC(Microcal Inc., USA). The cell (1.2 mL) was loaded with the nanopar-ticle suspension and stirred at 264 rpm by a rotating syringe,which was filled with a 4 mM solution of CaCl2, PBLG10k-b-PEG6k-alendronate nanoparticle suspensions at 0.3 mM concentration ofpolymer corresponding to 0.075 mM concentration of alendronatewere prepared. The concentration of alendronate in nanoparti-cles was calculated taking into account the amount of alendronateestimated by NMR. The titration of PBLG10k-b-PEG6k-alendronatenanoparticles involved 30 injections of 10 �L of the calciumsolution during 200 s at an interval of every 600 s. Experimentswere carried out at 25 ◦C. Blank nanoparticles, i.e. not containingalendronate groups, were prepared from PBLG10k-bnz and PBLG40k-b-PEG5k and were used as a negative control. A positive control wasmade with an aqueous solution of PEG-alendronate. As the inter-action with calcium ions is a monovalent interaction, the affinityconstant was assumed to remain the same for PEG6k-alendronate insolution and when PEG6k-alendronate was inserted at the surface ofPBLG10k-b-PEG6k-alendronate nanoparticles. The affinity constantdetermined in the PEG-alendronate ITC experiment was obtainedconstraining the stoichiometry to 1 and it was used to determine

the effective stoichiometry of the PBLG10k-b-PEG6k-alendronatenanoparticles and calcium ions interaction. Additionally, the inter-action of an aqueous solution of alendronate with calcium ions wasperformed and used as a reference.

7 urnal

2

d(ccoS

aewpftM2ttof(essctceb

odac

seaVwbwSorsocn

3

3

p∼aoPsa∼

6 L. de Miguel et al. / International Jo

.7. Molecular modeling

The three-dimensional structure of HAP (as a .cif file) wasownloaded from the Inorganic Crystal Structure DatabaseLeventouri et al., 2003) (ICSD, http://icsd.ill.eu/icsd/index.html,ode icsd 98873) and manually curated. Packing structures werereated with Mercury 3.0 (Macrae et al., 2008) with sizesf 10 × 10 × 10 and visualized within Maestro interface fromchrodinger Suite (Schrödinger, 2011).

The three-dimensional structures of alendronate were gener-ted using CORINA v3.44 (CORINA, version 3.44) as individualntities. Alendronates were modeled as zwitterionic molecules,ith one negative charge on each phosphonate group and oneositive charge on the protonated amino group. A stochastic con-ormational search, based on Monte Carlo sampling, was employedo generate various conformations, using the Mixed torsional/Low-

ode sampling method in the MacroModel module (Macromodel,011) of Schrodinger Suite with improved setup options as follows:he maximum number of steps was set to 10,000, all calcula-ions used 1000 steps per rotatable bond and an energy cut-offf 42 kJ/mol above the global energy minimum. Redundant con-ormers were eliminated by imposing a root mean square deviationRMSD) cutoff of 2.0 A. Unless otherwise noted, only the low-st energy conformation for each compound was retained. Theearches were done using the water continuum model, withtandard settings as below: OPLS 2005 force field with suppliedharges, dielectric constant 1.0, van der Waals cut-off 8.0 A, elec-rostatic cut-off 20.0 A, and hydrogen bond cut-off 4.0 A. Theonformational search produced a number of 24 unique conform-rs, for which all possible distances between two oxygen atomselonging to the bisphosphonate group were measured.

Similar oxygen–oxygen distance measurements were carriedut on the exterior surfaces of HAP. The comparison of these twoistance data sets showed that four values (4.12 A, 4.75 A, 4.80 And 5.02 A) were common to both HAP surfaces and alendronateonformers.

Selected alendronate conformers were inserted on the HAPurface using the matching oxygen–oxygen distances identifiedarlier, the biphosphonate moiety facing the HAP surface and themino-containing positive end stretched in the opposite direction.arious HAP-alendronate complexes were designed, in accordanceith the global symmetry of the HAP exterior surfaces and with the

iphosphonate coordination states in alendronate. Steric clashesithin the complexes were removed by energy minimization using

chrödinger’s Maestro framework, with all HAP atoms frozen inrder to keep the intact crystal structure. Visual inspection of theesulting HAP-alendronate complexes using PyMol (DeLano, 2006)howed that the amino-alkyl substituent in alendronate is mainlyriented vertically on the HAP surface. Larger HAP-alendronateomplexes (10 × 10 × 10) were then built by extending the crystaletwork on both surfaces using symmetry rules.

. Results

.1. Synthesis and characterization of PBLG polymers

Polymerization reactions were followed by FTIR. The com-lete disappearance of the anhydride signals at ∼1850 cm−1,1775 cm−1, ∼920 cm−1 corresponding respectively to the C5 Ond C2 O and C O C of the anhydride confirms the completionf the polymerization reaction (see Supplementary Information).

olymers were characterized by NMR, FTIR, and MALDI-TOF. FTIRpectra of all polymers show an �-helix secondary structure, char-cterized by the amide I, amide II, and amide III at ∼1655 cm−1,1550 cm−1 and ∼1260 cm−1 respectively (Barbosa et al., 2007;

of Pharmaceutics 460 (2014) 73– 82

Cauchois et al., 2013; Segura-Sanchez et al., 2010). A criticalDPn of 18 was established above which the �-helix structureis increasingly favored versus the �-sheet secondary structure(Papadopoulos et al., 2004). In the present study, the minimal DPn

of the synthesized polymers was 38, and as a result they were onlycomposed of � helical secondary structures (Table 1). Differencesbetween theoretical and the experimental DPn determined by theMALDI-TOF technique and between DPn established for PBLG40k-PEG5k by NMR and MALDI-TOF, suggest that MALDI-TOF techniqueunderestimated the molar mass of polymers, probably due to themolar mass discrimination effect (Favier et al., 2004). Alendronatecontent in PBLG-b-PEG-alendronate was estimated by 31P NMR(see Supplementary Information) and 1H NMR, quantifying theefficacy of the reactions. It was found that 3.6% of PBLG40k-PEG6k-alendronate copolymer chains contained a molecule of alendronatewhereas that of 5.3% for the PBLG10k-PEG6k-alendronate copolymerchains.

3.2. Characterization of nanoparticles

Nanoparticles prepared from one or from a mixture of differentPBLG derivates could be easily obtained by a simple nanoprecipi-tation method without the use of any additional surfactant. In allcases, nanoparticles were characterized by a size under 80 nm witha narrow size distribution and thus nanoparticles were likely to passthrough the bone-blood barrier and reach the bone tissue. They allshowed a negative zeta potential, which varied depending on thepolymer or polymer mixtures. The more negative zeta potential ofPBLG10k-b-PEG6k-alendronate nanoparticles could be attributed tothe presence of alendronate, which holds two negative charges atpH 7.9 (Table 2).

Nanoparticles prepared from the PBLG10k-bnz and PBLG10k-b-PEG6k-alendronate, and their mixtures with PBLG40k-FITC in a 3:1proportion, presented rather a spherical form, as revealed by theTEM images, whereas it can be observed that nanoparticle shapebecame increasingly ellipsoidal as the molar mass of the PBLGincreased (Cauchois et al., 2013), as shown also for PBLG25k-bnznanoparticles and PBLG40k-b-PEG6k-alendronate (see Supplemen-tary Information).

Calculations suggested that on an average each individual fluo-rescently labeled PBLG10k-b-PEG6k-alendronate nanoparticle couldbe formed by the self-assembly of approximately 1100 �-helixchains and thus, approximately 60 molecules of alendronate werepresent on each nanoparticle. For fluorescently labeled PBLG40k-b-PEG6k-alendronate nanoparticles, each individual nanoparticlecould contain approximately 190 �-helix chains and thus the num-ber of alendronate molecules per nanoparticle was estimated to bearound 7. Thus, it is observed that the molecular weight of the PBLGblock had a considerable influence in the number of ligands theoret-ically available for interactions with HAP. Both the high solubilityof alendronate in water combined with the rigidity of the auto-assembling PBLG blocks favors the localization of the alendronatemolecules on the nanoparticle surface.

3.3. Multivalent binding affinity for HAP

3.3.1. In vitro HAP binding assayThe multivalent interaction of PBLG10k-b-PEG6k-alendronate

nanoparticles with HAP was studied by means of fluorescence todetermine if alendronate decorated nanoparticles could bind toHAP surfaces. It is a multivalent interaction because each nanopar-ticle is decorated with multiple alendronate molecules on its

surface, which interact at multiple sites with one HAP surfacewith various calcium ions available on its structure. This assayrevealed the effective binding of alendronate nanoparticles to HAP(Fig. 2). PBLG10k-b-PEG6k-alendronate nanoparticles, decorated

L. de Miguel et al. / International Journal of Pharmaceutics 460 (2014) 73– 82 77

Table 1Molar mass of the polymers determined by 1H NMR and MALDI-TOF.

PBLG-polymers DPna DPn

b Mn (g/mol) PIc

PBLG25k-bnz 136 106c 23407c 1.13PBLG10k-bnz 45 38c 8431c 1.16PBLG40k-b-PEG5k 228 156c 235d 39366c 56465d 1.23PBLG40k-b-PEG6k-alendronate 228 190c 42415c 1.17PBLG10k-b-PEG6k-alendronate 45 38c 8711e –PBLG40k-FITC 228 163c 36155c 1.14

a Theoretical.b Of the PBLG block.c Determined by Maldi-TOF.d Determined by NMR.e Determined from the PBLG10k-bnz.

Table 2Size and zeta potential of nanoparticles determined by DLS.

Nanoparticles Size (nm) ± SDa Polydispersity index � Potential (meV) ± SD

PBLG25k-bnz 58 ± 0.4 0.14 −29 ± 1.3PBLG10k-bnz 70 ± 0.3 0.13 −23 ± 1.4PBLG40k-b-PEG5k 52 ± 2.8 0.15 −24 ± 0.5PBLG40k-b-PEG6k-alendronate 57 ± 1.7 0.16 −27 ± 0.8PBLG10k-b-PEG6k-alendronate 76 ± 1.3 0.17 −34 ± 1.2PBLG25k-bnz: PBLG40k-FITC 3:1 68 ± 2.1 0.17 −18 ± 1.3PBLG10k-bnz: PBLG40k-FITC 3:1 56 ± 1.2 0.16 −17 ± 0.9PBLG40k-b-PEG5k: PBLG40k-FITC 3:1 52 ± 1.1 0.13 −20 ± 1.0

667

wcaoaonb

a21PaoR

Ffnn6

PBLG40k-b-PEG6k-alendronate: PBLG40k-FITC 3:1

PBLG10k-b-PEG6k-alendronate: PBLG40k-FITC 3:1PBLG10k-b-PEG6k-alendronate: PBLG40k-FITC 3:1 pre-incubated with CaCl2

ith approximately 60 molecules of alendronate per nanoparticle,ould bind totally to HAP after 18 h whereas for PBLG40k-b-PEG6k-lendronate nanoparticles, which contained around 7 moleculesf alendronate per nanoparticle, HAP bound particles were ofbout 80%. Control nanoparticles showed no binding in the casef PBLG25k-bnz nanoparticles. In the case of PBLG40k-b-PEG5kanoparticles, a non-specific interaction was observed, as showny the 20% of HAP bound nanoparticles after 18 h.

The binding of alendronate decorated nanoparticles to HAP was kinetic process, as described already for alendronate (Leu et al.,006) and other molecules (Kandori et al., 2000; Moreno et al.,984; Rill et al., 2009). Indeed it was a rapid process since 85% ofBLG10k-b-PEG6k-alendronate nanoparticles were already boundfter 1 h, which was similar but slightly quicker than previously

bserved for biphosphonate-tagged gold nanoparticles (Ross andoeder, 2011).

ig. 2. Degree of binding expressed in percentage of nanoparticles bound to HAPor PBLG40k-b-PEG6k-alendronate nanoparticles and PBLG10k-b-PEG6k-alendronateanoparticles after 1 and 18 h compared to PBLG25k-bnz and PBLG40k-b-PEG5k

anoparticles. Nanoparticles at a concentration of 1.5 mg/mL in 0.1% Pluronic® F8 were incubated with 0.4 g of HAP suspension in PBS buffer.

3 ± 1.3 0.12 −20 ± 1.69 ± 0.5 0.15 −38 ± 1.20 ± 0.8 0.18 −16 ± 1.2

3.3.2. Binding isothermsThe multivalent interaction between PBLG10k-b-PEG6k-

alendronate nanoparticles and HAP was estimated frombinding isotherms experiments. Binding isotherms ofPBLG10k-b-PEG6k-alendronate and calcium pre-incubated PBLG10k-b-PEG6k-alendronate nanoparticles were studied and compared toPBLG10k-bnz nanoparticles, used as a negative control.

No differences were observed between the binding isotherm forPBLG10k-b-PEG6k-alendronate in the presence or in the absence ofcalcium ions (Fig. 3). Both followed the Langmuir model. This couldsuggest that calcium ions in solution could be easily exchangedwith calcium ions belonging to the HAP structure. The plot of C/Qversus C expressed in alendronate, in polymer and in nanoparti-cles using a Langmuir model exhibited correlation coefficients R2

of 0.99, where Qmax and KHAP could be determined and are shownin Table 4.

Fig. 3. Binding isotherms of PBLG10k-b-PEG6k-alendronate nanoparticles and cal-cium pre-incubated PBLG10k-b-PEG6k-alendronate nanoparticles with HAP in PBSpH 7.4 versus PBLG10k-bnz nanoparticles. Experimental data were fitted to a Lang-muir model using Eq. (1).

78 L. de Miguel et al. / International Journal of Pharmaceutics 460 (2014) 73– 82

Table 3Binding parameters for PBLG10k-b-PEG6k-alendronate nanoparticles and for PEG6k-alendronate and alendronate in solution with calcium ions determined by isothermaltitration calorimetry.

Alendronate PEG6k-alendronate PBLG10k-b-PEG6k-alendronate nanoparticles

Chi2/DoF 99 90 238Stoichiometry, N 0.5a 1.0a 1.0 ± 0.05b

KCa+2 (M−1)c 1.8 × 104 ± 7 × 102 1.8 × 104 ± 1.7 × 103 1.8 × 104

�H (kcal mol−1) 11 × 102 ± 34 16 × 102 ± 36 13 × 102 ± 57�S (cal deg−1 mol−1) 23 25 24

or PEG

a−ann

ITC experiments were carried out to determine the affinity of

FP

a Imposed stoichiometry for calculations.b Calculated stoichiometry using the experimental affinity constant determined fc Expressed in alendronate concentration.

Not surprisingly, calcium pre-incubated PBLG10k-b-PEG6k-lendronate nanoparticles had a less negative zeta potential (from26.7 meV to −16.2 meV), which did not affect nanoparticle inter-ction with HAP. This supports the fact that the interaction of

anoparticles were driven by the alendronate targeting moiety, andot by the zeta potential.

ig. 4. Typical ITC integrated heat data profiles obtained from the binding interaction of CBLG25k-bnz and Np PBLG40k-b-PEG6k; (b) (up-right) Np PBLG10k-b-PEG6k-alendronate (c

6k-alendronate in solution.

3.4. Binding affinity for calcium ions by isothermal titrationcalorimetry

the interaction between the alendronate molecules in PBLG10k-b-PEG6k-alendronate nanoparticles and calcium ions, as it is the most

aCl2 with: (a) (up-left) nanoparticles (Np) PBLG10k-b-PEG6k-alendronate versus Np) (down-left) PEG6k-alendronate solution (d) (down-right) alendronate solution.

L. de Miguel et al. / International Journal of Pharmaceutics 460 (2014) 73– 82 79

Table 4The binding parameters of PBLG10k-b-PEG6k-alendronate nanoparticles for HAP determined by the Langmuir model using the Langmuir linear regression method.

Np PBLG10k-b-PEG6k-alendronate R2 Qmax (moles/g HA) KHAP (M−1)

99

−9 7

99

99

iwsfecPtctz(anCPnpicI(cnsttTama

3m

ssmFctosiiM(aaMaoamdfo

Expressed in alendronate 0.Expressed in polymer PBLG10k-b-PEG6k-alendronate 0.Expressed in nanoparticles 0.

mportant interaction involved in the interaction of alendronateith HAP. PBLG10k-b-PEG6k-alendronate nanoparticles clearly

howed a specific interaction with calcium ions that was not seenor negative control nanoparticles. Such interaction resulted inxponentially decreasing heats of interactions while adding cal-ium ions to the nanoparticle suspension. A positive control withEG6k-alendronate in solution was performed so as to mimic theype of interaction occurring with nanoparticles, since alendronateonjugation with PEG chains (see Fig. 1) was likely to differ fromhat observed with free alendronate molecules. Alendronate on itswitteronic form is known to form 2:1 complexes with calcium ionsFernandez et al., 2003). However, alendronate moiety in PEG6k-lendronate in solution or exposed on nanoparticle surface had twoegative charges, where it could be depicted that interaction witha2+ would form a 1:1 complex. Interactions between alendronate,EG6k-alendronate in solution and PBLG10k-b-PEG6k-alendronateanoparticles with calcium ions had similar thermodynamicarameters (�H and �S), which confirmed that the interaction

nvolved was likely to be similar. All interactions were entropi-ally driven and could be described by a one-site model (Fig. 4).nteractions were strong, as suggested by the binding constantsKCa

+2), and slightly stronger for the 1:1 complex alendronatealcium (in PEG6k-alendronate and PBLG10k-b-PEG6k-alendronateanoparticles) than for the 2:1 complex (in calcium-alendronateolution). The heat signal variations were weaker in the case ofhe titration of PBLG10k-b-PEG6k-alendronate nanoparticles due tohe limited concentration of alendronate, but were still exploitable.he stoichiometry for PBLG10k-b-PEG6k-alendronate nanoparticlesnd calcium interaction was found to be 1,04 indicating that allolecules of alendronate were exposed on the nanoparticle surface

nd available to interact with calcium ions (Table 3).

.5. Analysis of alendronate complexation at HAP-surfaces byolecular modeling

Molecular modeling has been used to investigate the pos-ibilities of interactions of alendronate moieties with the twourfaces of HAP crystals. Four modes of interaction of alendronateolecules with HAP surfaces 1 and 2 were identified as shown in

ig. 5. The alendronate interacts with HAP not only by chelation ofalcium ions but also by hydrogen bonds with oxygen atoms fromhe crystalline structure. The distances between the oxygen atomsf alendronate in the selected conformers are compatible with theame distances in the HAP structure, which allows for an optimalnteraction between these two entities. These oxygen–oxygennteratomic distances are of 4.12, 4.80 and 5.02 A respectively for

odels 1, 2 and 3 on Face 1, and of 4.75 A for Model 4 on Face 2Fig. 5). For Model 1, two calcium atoms coordinated by oxygenre situated at 2.48 and 2.83 A from the first atom and respectivelyt 2.52 and 2.83 A from the second oxygen on HAP surface. Forodel 2, distances of 2.81 A are measured between the oxygen

toms responsible for binding and the common calcium ion. Twothers calcium ions coordinate these oxygen atoms and are foundt 2.33 A. According to Model 3, the binding of alendronate is

ediated by two calcium ions that coordinate oxygen, within

istances of 2.32 and 2.74 A for the first oxygen and 2.58 and 2.89 Aor the second. Alendronate coordination on HAP’s face 2 involvesxygen atoms positioned at a distance of 4.75 A. These atoms are

4.4 × 10 1.6 × 107.1 × 10−9 9.6 × 107

6.4 × 10−12 1.1 × 1010

coordinated with two calcium ions, positioned at 2.50 and 2.30 Afrom the first, and at 2.50 A from the second oxygen.

The alendronate interaction with HAP is enhanced by the inter-action between the three OH groups of alendronate with thecrystalline structure: two hydrogen bonds for Model 2 and threehydrogen bonds for the other models. The areas of external sur-faces of HAP were measured in order to determine the specificsurface for the interaction with alendronate and for comparisonwith the experimental data. The HAP structure that we generatedcontains 1000 units (10 × 10 × 10 units from the crystal cell) withareas of 9427 A2 for the first surface (face 1) and 7544 A2 for thesecond (face 2) (see Supplementary Information). According to ourmodels, a number of 121–132 alendronate molecules can theo-retically interact with the first face (corresponding to interatomicoxygen–oxygen distances of 4.80 A, 5.02 A and 4.12 A). Similarly, upto 110 alendronate molecules can interact with the second surface(face 2), corresponding to the oxygen–oxygen distance of 4.75 A(see Supplementary Information).

4. Discussion

The design of efficient bone targeted delivery systems dependson many aspects, an important one being the affinity of the drugcarriers to their target. Self-assembly of amphiphilic copolymers toform nanoparticles represents a convenient way to display multi-ple targeting units to their surface, likely to result in multivalentinteractions with their targets. Multivalent interactions occur fre-quently in nature and are known to be much more stronger thancorresponding monovalents interactions (Mammen et al., 1998).

PBLG copolymers have been selected for preparing bone tar-geted nanoparticles due to an attractive set of other characteristicsincluding their biocompatibility and degradability. Ring openingpolymerization was found to be convenient to obtain copolymerswith low polymolarity index (PI < 1.3). In order to attain bone,HAP targeting has been the selected strategy as it is highly spe-cific of bone tissue, existing only also in teeth and pathologicalcalcifications. Alendronate has been successfully coupled to PBLGcopolymers which could self-assemble to form small nanoparticlesso as to impart bone osteotropicity.

For bone-targeted nanoparticles to reach bone tissue, a pro-longed time of circulation is required. The small hydrodynamicdiameter of the nanoparticles (50–75 nm, depending on the com-position) was likely to favor a prolonged bloodstream circulation(Alexis et al., 2008). In order to avoid premature clearance bythe reticuloendothelial system, pegylation of nanoparticles is apreferred method (Owens and Peppas, 2006). The hydrophilic-ity and flexibility of the surface PEG chains is widely used toprevent the adsorption of opsonins and the subsequent phagocy-tosis and clearance by the reticuloendothelial system. Thus, a PEGabove 5 kDa, MeO-PEG5k-NH2,was selected to synthesize PBLG40k-b-PEG5k block copolymer. To ensure that molecules of alendronatecould be exposed and available on the surface, NHS-PEG6k-NHSwas chosen for the synthesis of the bone-targeting copolymer bothas a spacer and to confer stealth properties to the bone-targeted

nanoparticles.

As previously reported by our group, PBLG blocks adopt �-helix structures resulting in quite rigid rods able to self-assembleand to form the core of nanoparticles. Nanoparticles made from

80 L. de Miguel et al. / International Journal of Pharmaceutics 460 (2014) 73– 82

2 (M

Pwhws

e

Fig. 5. HAP-alendronate complexes on face 1 (Models 1–3) and on face

BLG-b-PEG6k-alendronate copolymers had a core-shell structure,here the flexible PEG chains would form an external flexibleydrophilic shell, while the rigidity of the hydrophobic PBLG block

ould mechanically favor alendronate presentation on the particle

urface.First, the ability of alendronate decorated nanoparticle to

ffectively bind to HAP was checked by a fluorescence method.

odel 4) (a) (left) view of the whole model and (b) (right) detailed view.

The possibility of assembling different PBLG-derivates to formmultifunctional nanoparticles was previously reported by ourgroup (Martinez-Barbosa et al., 2009) and it was confirmed with

this assay, where each PBLG10k-b-PEG6k-alendronate nanoparticledisplayed at the same time its fluorescent and its bone bind-ing property. It can be inferred from this experiment that thebinding between PBLG10k-b-PEG6k-alendronate nanoparticles and

urnal

Picune

nia(a(cHaemimi

iHstt

cciPediaaicwcwm

tAatdasosmanic1mteaot6

L. de Miguel et al. / International Jo

BLG40k-b-PEG6k-alendronate nanoparticles with HAP is a specificnteraction, not driven by the negative zeta potential of nanoparti-les. The absence of binding for PBLG10k-bnz nanoparticles alloweds to conclude that the affinity of PBLG10k-b-PEG6k-alendronateanoparticles for HAP is due to the alendronate moiety and toxclude the differences in shape, evidenced by the TEM.

Next, the affinity between PBLG10k-b-PEG6k-alendronateanoparticles and the HAP surface was studied by adsorption

sotherms. Multivalent interactions are known to enhance theffinity compared to its corresponding monovalent interactionKiessling et al., 2006; Mammen et al., 1998). The multivalentffinity constant for PBLG10k-b-PEG6k-alendronate nanoparticlesKHAP) expressed in alendronate was more than 5 times strongerompared to the monovalent interaction between alendronate andAP reported value in the literature (Nancollas et al., 2006). It waslmost 4000 times more if the multivalent affinity constant wasxpressed in terms of nanoparticle concentrations. This is in agree-ent with the previous works which showed that multivalent

nteractions are 4–9500 times stronger than the correspondingonovalent ones, the highest enhancements seen for weaker

ntrinsic affinities (Tassa et al., 2010).Affinity binding was also studied in the presence of calcium ions

n solution since the main interaction between the alendronate andAP is known to involve the calcium ions of the HAP crystalline

tucture. Interestingly, it was shown that calcium-bound nanopar-icles could still bind to HAP structures with the same affinity as inhe absence of calcium ions.

Then, interaction of PBLG10k-b-PEG6k-alendronate nanoparti-les with calcium ions was studied by ITC. The interaction withalcium ions, KCa

+2, determined by ITC was of 1.8 × 104 M−1 whichs 900 times weaker than the multivalent interaction of PBLG10k-b-EG6k-alendronate nanoparticles with HAP (KHAP = 1.6 × 107 M−1)stimated from binding isotherm. This enhanced affinity of alen-ronate nanoparticles for HAP in comparison with calcium ions

n solution explains why calcium pre-incubated PBLG10k-b-PEG6k-lendronate nanoparticles can still bind to HAP surfaces, this being

competition which is favorable to HAP. After intravenous admin-stration of nanoparticles, they are confronted to physiologicaloncentrations of calcium ions in the blood. Interestingly, thisould imply that even if nanoparticles could interact with cal-

ium ions present in the bloodstream or physiological fluids, thisould be reversible and would not affect their interaction with theineral matrix of the bone tissue due to higher affinity to HAP.Finally, molecular modeling studies were performed so as

o determine the interactions involving alendronate and HAP.lthough the most important interactions between alendronatend HAP involves the coordination of the calcium ions of HAP,he interactions involving alendronate/calcium ions and alen-ronate/HAP crystalline structure were not the same since inddition hydrogen bonding involving the alendronate and the HAPtructure are established (see Fig. 5). The theoretical models devel-ped in this work for the interaction of alendronate with the HAPurface are important in evidencing the detailed structural deter-inants of this interaction, at atomic level, and allow the rational

nalysis of the experimental data showing a good affinity of PBLGanoparticles for the bone tissue. According to molecular model-

ng data, the theoretical number of free alendronate molecules thatan interact with a HAP surface unit of 95 nm2 or 75 nm2 was of32 and 110 respectively, depending on the crystal face. Experi-entally, depending on the molecular weight of the PBLG block,

he number of alendronate decorating molecules could be consid-rably varied. However, irrespective of the situation, binding sites

vailable at HAP surface were in excess compared to the numberf alendronate available at nanoparticle surface. For example, inhe case of the PBLG10k-b-PEG6k-alendronate nanoparticles (about0 alendronate per nanoparticle) only a reduced fraction of these

of Pharmaceutics 460 (2014) 73– 82 81

alendronate moieties was likely to interact with HAP surface dueto nanoparticle size (ranging from 52 to 76 nm in size). Therefore,it could be concluded that numerous sites at the surface of HAPcrystals were available for interactions with alendronate moleculescompared to the number of interactions necessary in the case ofalendronate decorated nanoparticles, which is obviously favorablefor their binding to HAP crystals included in bone matrix.

5. Conclusion

Bone-targeted multifunctional alendronate nanoparticleswhich showed strong affinity for HAP, one major component ofbone extracellular matrix, could be easily prepared. The bind-ing isotherm for PBLG10k-PEG6k-alendronate nanoparticles withHAP surfaces suggested that multivalency of the nanoparticlesresulted in 4000 fold stronger interactions with HAP comparedthe monovalent interaction with free alendronate molecules.The molecular modeling studies yielded a deep insight into thepossible interactions of alendronate with HAP. These interactionscould be considerably enhanced by the ability of the nanoparticlesto develop specific multivalent interactions with HAP, via calciumbinding as well as hydrogen bonding. In this case, it was suggestedthat steric hindrance due to the size of the nanoparticles resultedin a very number of HAP binding sites involved in the interaction,which favors a very strong interaction between alendronate andHAP. Finally, the 900 fold lower affinity of the nanoparticlesfor free calcium ions compared to HAP allowed calcium boundnanoparticle to interact with HAP. This is favorable to maintainthe targeting specificity of the nanoparticles and yields a deeperunderstanding of bone targeted nanoparticle in the body.

Acknowledgements

This work has benefited from the facilities and expertise ofthe Platform for Transmission Electronic Microscopy of IMAGIF(Centre de Recherche de Gif – www.imagif.cnrs.fr)”. We gratefullyacknowledge the European Postgraduate Program from “La Caixa”Foundation for the financial support.

Appendix A. Supplementary data

Supplementary data associated with this article can befound, in the online version, at http://dx.doi.org/10.1016/j.ijpharm.2013.10.048.

References

Alexis, F., Pridgen, E., Molnar, L.K., Farokhzad, O.C., 2008. Factors affecting theclearance and biodistribution of polymeric nanoparticles. Mol. Pharmacol. 5,505–515.

Barbosa, M.E.M., Montembault, V., Cammas-Marion, S., Ponchel, G., Fontaine,L., 2007. Synthesis and characterization of novel poly(�-benzyl-l-glutamate)derivatives tailored for the preparation of nanoparticles of pharmaceutical inter-est. Polym. Int. 56, 317–324.

Carlson, C.B., Mowery, P., Owen, R.M., Dykhuizen, E.C., Kiessling, L.L., 2007. Selectivetumor cell targeting using low-affinity, multivalent interactions. ACS Chem. Biol.2, 119–127.

Cauchois, O., Segura-Sanchez, F., Ponchel, G., 2013. Molecular weight controls theelongation of oblate-shaped degradable poly(�-benzyl-l-glutamate) nanopar-ticles. Int. J. Pharm. 452, 292–299.

CORINA, version 3.44, 2013. Molecular Networks GmbH. Erlanger, Germanyhttp://www.molecular-networks.com

DeLano, S., 2006. The Pymol Molecular Graphic System, Version 0.99. DeLano Scien-tific LLC, Palo Alto, CA.

Favier, A., Ladavière, C., Charreyre, M.-T., Pichot, C., 2004. MALDI-TOF MS inves-tigation of the RAFT polymerization of a water-soluble acrylamide derivative.

Macromolecules 37, 2026–2034.

Fernandez, D., Vega, D., Goeta, A., 2003. Alendronate zwitterions bind to calciumcations arranged in columns. Acta Crystallogr. C 59, m543–m545.

Gittens, S.A., Bansal, G., Zernicke, R.F., Uludag, H., 2005. Designing proteins for bonetargeting. Adv. Drug Delivery Rev. 57, 1011–1036.

8 urnal

H

H

K

K

K

K

L

L

L

M

MM

M

M

N

Tassa, C., Duffner, J.L., Lewis, T.A., Weissleder, R., Schreiber, S.L., Koehler, A.N., Shaw,S.Y., 2010. Binding affinity and kinetic analysis of targeted small molecule-

2 L. de Miguel et al. / International Jo

irabayashi, H., Sawamoto, T., Fujisaki, J., Tokunaga, Y., Kimura, S., Hata, T., 2001.Relationship between physicochemical and osteotropic properties of bisphos-phonic derivatives: rational design for osteotropic drug delivery system (ODDS).Pharm. Res. 18, 646–651.

owlett, C.R., Dickson, M., Sheridan, A.K., 1984. The fine structure of the proximalgrowth plate of the avian tibia: vascular supply. J. Anat. 139, 115–132.

andori, K., Fudo, A., Ishikawa, T., 2000. Adsorption of myoglobin onto various syn-thetic hydroxyapatite particles. Phys. Chem. Chem. Phys. 2, 2015–2020.

iessling, L.L., Gestwicki, J.E., Strong, L.E., 2000. Synthetic multivalent ligands in theexploration of cell–surface interactions. Curr. Opin. Chem. Biol. 4, 696–703.

iessling, L.L., Gestwicki, J.E., Strong, L.E., 2006. Synthetic multivalent ligands asprobes of signal transduction. Angew. Chem. Int. Ed. 45, 2348–2368.

lok, H.-A., Langenwalter, J.F., Lecommandoux, S.B., 2000. Self-assembly of peptide-based diblock oligomers. Macromolecules 33, 7819–7826.

angmuir, I., 1916. The constitution and fundamental properties of solids and liquids.Part I. Solids. J. Am. Chem. Soc 38, 2221–2295.

eu, C.-T., Luegmayr, E., Freedman, L.P., Rodan, G.A., Reszka, A.A., 2006. Relativebinding affinities of bisphosphonates for human bone and relationship to antire-sorptive efficacy. Bone 38, 628–636.

eventouri, T., Bunaciu, C.E., Perdikatsis, V., 2003. Neutron powder diffraction studiesof silicon-substituted hydroxyapatite. Biomaterials 24, 4205–4211.

acrae, C.F., Bruno, I.J., Chisholm, J.A., Edgington, P.R., McCabe, P., Pidcock, E.,Rodriguez-Monge, L., Taylor, R., van de Streek, J., Wood, P.A., 2008. Mercury CSD2.0 – new features for the visualization and investigation of crystal structures.J. Appl. Crystallogr. 41, 466–470.

acromodel, 2011. Macromodel Version 9.9. Schrödinger LLC, New York, NY.ammen, M., Choi, S.-K., Whitesides, G.M., 1998. Polyvalent interactions in bio-

logical systems: implications for design and use of multivalent ligands andinhibitors. Angew. Chem. Int. Ed. 37, 2754–2794.

artinez-Barbosa, M.E., Cammas-Marion, S., Bouteiller, L., Vauthier, C., Ponchel,G., 2009. PEGylated degradable composite nanoparticles based on mixtures ofPEG-b-poly(�-benzyl-l-glutamate) and poly(�-benzyl-l-glutamate). Bioconju-gate Chem. 20, 1490–1496.

oreno, E., Kresak, M., Hay, D., 1984. Adsorption of molecules of biological interest

onto hydroxyapatite. Calcif. Tissue Int. 36, 48–59.

ancollas, G.H., Tang, R., Phipps, R.J., Henneman, Z., Gulde, S., Wu, W., Mangood,A., Russell, R.G.G., Ebetino, F.H., 2006. Novel insights into actions of bisphos-phonates on bone: differences in interactions with hydroxyapatite. Bone 38,617–627.

of Pharmaceutics 460 (2014) 73– 82

Oh, I., Oh, J., Cho, C., Lee, K., 1995. Biodegradability of poly(�-benzyl-l-glutamate)/poly(ethylene oxide)/poly(�-benzyl-l-glutamate) block copolymerin mice. Arch. Pharm. Res. 18, 8–11.

Owens, D.E., Peppas, N.A., 2006. Opsonization, biodistribution, and pharmacokine-tics of polymeric nanoparticles. Int. J. Pharm. 307, 93–102.

Ozcan, I., Bouchemal, K., Segura-Sanchez, F., Ozer, O., Guneri, T., Ponchel, G.,2011. Synthesis and characterization of surface-modified PBLG nanoparti-cles for bone targeting: in vitro and in vivo evaluations. J. Pharm. Sci. 100,4877–4887.

Papadopoulos, P., Floudas, G., Klok, H.A., Schnell, I., Pakula, T., 2004. Self-assemblyand dynamics of poly(�-benzyl-l-glutamate) peptides. Biomacromolecules 5,81–91.

Park, Y.J., Nah, S.H., Lee, J.Y., Jeong, J.M., Chung, J.K., Lee, M.C., Yang, V.C., Lee,S.J., 2003. Surface-modified poly(lactide-co-glycolide) nanospheres for targetedbone imaging with enhanced labeling and delivery of radioisotope. J. Biomed.Mater. Res. A 67, 751–760.

Rill, C., Kolar, Z.I., Kickelbick, G., Wolterbeek, H.T., Peters, J.A., 2009. Kinetics andthermodynamics of adsorption on hydroxyapatite of the [160Tb] terbium com-plexes of the bone-targeting ligands DOTP and BPPED. Langmuir 25, 2294–2301.

Ross, R.D., Roeder, R.K., 2011. Binding affinity of surface functionalized gold nanopar-ticles to hydroxyapatite. J. Biomed. Mater. Res. A 99, 58–66.

Schrödinger, 2011. Suite 2011: Maestro; Version 9.2. Schrödinger, LLC, New York.Segura-Sanchez, F., Montembault, V., Fontaine, L., Martinez-Barbosa, M.E.,

Bouchemal, K., Ponchel, G., 2010. Synthesis and characterization of function-alized poly(�-benzyl-l-glutamate) derivates and corresponding nanoparticlespreparation and characterization. Int. J. Pharm. 387, 244–252.

Shea, J.E., Miller, S.C., 2005. Skeletal function and structure: implications for tissue-targeted therapeutics. Adv. Drug Delivery Rev. 57, 945–957.

Shiau, C.C., Labes, M.M., 1989. Correlation of pitch with concentration andmolecular weight in poly(�-benzyl glutamate) lyophases. Macromolecules 22,328–332.

modified nanoparticles. Bioconjugate Chem. 21, 14–19.Wang, D., Miller, S.C., Kopecková, P., Kopecek, J., 2005. Bone-targeting macromolec-

ular therapeutics. Adv. Drug Delivery Rev. 57, 1049–1076.