Functionalization of Cellulose Nanocrystals with PEG-Metal-Chelating Block Copolymers via Controlled Conjugation in Aqueous Media.

Elongated nanoparticles have recently been shown to have distinct advantages over their spherical counterparts in drug delivery applications. Cellulose nanocrystals (CNCs) have rodlike shapes in nature and have demonstrated biocompatibility in a variety of mammalian cell lines. In this report, CNCs are put forward as a modular platform for the production of multifunctional rod-shaped nanoparticles for cancer imaging and therapy. For the first time, PEGylated metal-chelating polymers containing diethylenetriaminepentaacetic acid (DTPA) (i.e., mPEG-PGlu(DPTA)18-HyNic and PEG-PGlu(DPTA)25-HyNic) are conjugated to CNCs to enable the chelation of radionuclides for diagnostic and therapeutic applications. The entire conjugation is based on UV/vis-quantifiable bis-aryl hydrazone-bond formation, which allows direct quantification of the polymers grafted onto the CNCs. Moreover, it has been shown that the mean number of polymers grafted per CNC could be controlled. The CNCs are also fluorescently labeled with rhodamine and Alexa Fluor 488 by embedding the probes in the polymer corona. Preliminary evaluation in a human ovarian cancer cell line (HEYA8) demonstrated that these CNCs are nontoxic and their penetration properties can be readily assessed in multicellular tumor spheroids (MCTSs) by optical imaging. These findings provide support for biomedical applications of CNCs, and further in vitro and in vivo studies are warranted to evaluate their potential as imaging and therapeutic agents for cancer treatment.

Introduction

Nanomedicines, owing to their many advantages over small-molecule chemotherapeutics, have been pursued extensively for application in cancer therapy. A number of nanoformulations have been approved for clinical use (e.g., Doxil and Abraxane) and have been demonstrated to extend the circulation lifetimes and reduce the commonly observed toxicities of many chemotherapeutic drugs.[1] The physicochemical characteristics of nanocarriers have been shown to have a profound effect on their biological performance.[2−6] In particular, the morphology of nanocarriers has been found to influence drug loading capacity, solid tumor penetration, and in vivo blood circulation lifetime.[7−9] For example, Discher’s group found that wormlike micelles (filomicelles) formed from PEGylated diblock copolymers had higher drug loading and longer circulation lifetimes than their spherical counterparts.[5] As well, Chauhan et al. demonstrated that the tumor penetration of quantum dot-based rod-shaped particles is greater than that of similar spherical particles.[10] Although these findings highlight the potential of rodlike nanoparticles in drug delivery, there are a limited number of reports of materials that enable robust production of rod- or cylinder-shaped nanocarriers.

Cellulose nanocrystals (CNCs), which are plant-derived, glucose-based elongated nanoparticles, offer a unique advantage over other platforms by enabling a facile manipulation of particle dimensions through the variation of the source of cellulose and experimental conditions employed for preparation.[11] Besides their oblong geometry, CNCs have shown to be nontoxic,[12,13] biocompatible, and biodegradable,[14−16] rendering them ideal for drug delivery applications.[17] Furthermore, the surface of CNCs exhibits abundant hydroxyl groups, which can be employed directly in covalent or noncovalent binding[18,19] or be further modified to enable attachment of drugs and imaging agents in addition to active targeting moieties to achieve site-specific delivery to the desired site.[20] However, the majority of reported surface modifications of CNCs, including oxidation,[21] esterification,[22] silylation,[18] urethanization,[23] and polymer grafting,[24,25] involve harsh organic solvents and laborious solvent exchange steps. These steps make the process less green and undesirable for pharmaceutical applications because of concerns about toxicity from the residual solvent.[14−16]

Here, we describe a green synthetic method for generating multifunctional CNCs that are decorated with water-soluble block copolymers (BCPs) for therapeutic and imaging applications in cancer research. To generate BCP–CNC conjugates, we explored a “grafting-onto” approach via bis-aryl hydrazone (BAH) formation between the terminal pyridylhydrazine group on the BCP and benzaldehyde groups on the surface of CNCs. The grafting-onto approach described here offers multiple advantages over other methods of CNC modification.[26] Because the CNCs and BCPs are independently synthesized and modified prior to coupling, this approach enables a thorough characterization of each component and can be expanded to generate a library of BCP–CNC conjugates with a diverse range of properties and applications. As well, the coupling chemistry employed in the present study allows for monitoring of the reaction progress by the increase in the absorbance of the BAH group at 354 nm.[27−29] We demonstrate that the UV absorption property of BAH not only allows the quantification of the mean number of polymers per CNC[30,31] but also enables precise control of the polymer grafting density and consequently the loading of therapeutic agents. Furthermore, the CNC and BCP modification steps as well as the coupling reaction are performed in aqueous solutions, eliminating the need for organic solvents.

In the current study, fluorescent BCP–CNC conjugates were prepared by grafting biocompatible BCPs onto the surface of fluorescently labeled CNCs. The BCP employed in this study consists of an outer PEG block that provides “stealth” properties for in vivo applications[32] and an inner poly(glutamic acid) (PGlu) block that introduces functional groups for the conjugation of metal-chelating diethylenetriaminepentaacetic acid (DTPA) for future applications in radiotherapy (Scheme ).[33] As a proof-of-concept evaluation of their potential as imaging agents, the penetration properties of these BCP–CNCs were investigated in a 3D multicellular tumor spheroid (MCTS) model of ovarian cancer. Overall, the findings of this study demonstrate that the “grafting-to” method of CNC modification offers a robust platform for the preparation and characterization of BCP–CNC conjugates. We think of these as stepping-stone experiments for the development of multifunctional theranostic agents based on which the second-generation CNC derivatives will be designed and developed to understand the factors that affect their cellular and in vivo behavior (e.g., pharmacokinetics and biodistribution). For example, the chelator DOTA (1,4,7,10-tetraazacyclododecane-1,4,7,10-tetraacetic acid) binds more tightly to lanthanide ions than DTPA, and thus, corresponding DOTA derivatives will be useful as carriers of 177Lu3+ for radiotherapy[34] and as carriers of Gd3+ for magnetic resonance imaging.[35,36]

Scheme 1

Design of a BCP–CNC Conjugate for the Delivery of Radionuclides to Tumors

Grafting densities of 330 mPEG-PGlu(DTPA)18 and 185 mPEG-PGlu(DTPA)25 molecules were achieved per CNC with an average length of 170 nm.

Results and Discussion

Surface Modification of CNCs

In our experimental design, we wished to convert some of the −OH groups on the surface of the CNCs to primary amines and then use these as sites of attachment for metal-chelating diblock copolymers (MC-BCPs) that were synthesized independently. This constitutes a grafting-onto approach for the construction of the CNC–BCP structures.[37−39]

Our starting material was a commercial CNC suspension, which was filtered (0.45 μm filter) to remove any possible aggregates. It was characterized by transmission electron microscopy (TEM) as shown in Figure a. The image analysis of 100 individual CNCs gave an average length 170 ± 30 nm and a width of 12 ± 2 nm, which are in good agreement with the literature values reported for CNCs from the same source.[40]

Figure 1

TEM images of (a) pristine CNCs and (b) amine-modified CNCs (CNC–NH2).

Amine groups were introduced to the surface of CNCs following the general procedure reported by Dong and Roman[41] and is shown in Scheme . The first step involved the modification of the CNC surface with epoxy functional groups by reacting the CNC suspension in a 1 M sodium hydroxide solution with epichlorohydrin at 60 °C. In the second step, the epoxy rings were opened with ammonium hydroxide to introduce primary amino groups. Because our goal was to attach multiple copies of MC-BCPs to the CNCs, we aimed to optimize the reaction conditions to maximize the number of amino groups per CNC (CNC–NH2). To address the concerns about low stability of the epoxy group under long reaction times in the presence of 1 M NaOH, the epoxy-ammonia reaction was optimized by carrying out the reaction with epichlorohydrin for various times ranging from 40 min to 20 h.

Scheme 2

Illustration of the Preparation of CNC–NH2

After amination, the resulting CNC–NH2 was treated with fluorescein isothiocyanate (FITC), and the fluorescein content of the CNC samples was determined by measuring the absorbance at 490 nm (Figure S1).[41] A calibration curve (Figure S2b) was generated for FITC in ammonium hydroxide at pH = 11 containing 0.02 wt % CNCs. As shown in Figure , the maximum degree of amination was obtained from the 1 h epoxidation reaction, which gave a FITC density of 37 μmol FITC per gram of CNC. Using the value of the nanocrystal width of 12 nm measured from TEM images of filtered CNCs and assuming a CNC density of 1.6 g/cm3,[42] we found approximately 680 FITC molecules per CNC.

Figure 2

The incorporation of FITC (μmol per gram of CNC) into CNC–NH2 samples obtained from CNC samples reacted with epichlorohydrin for different times.

There were also concerns about CNC integrity during amination of CNCs because hot solutions of strong bases have been used to degrade CNCs.[43] However, a TEM image of the CNC–NH2 sample obtained after the 1 h reaction with epichlorohydrin (Figure b) shows that the morphology remained the same as that of the pristine CNCs, and an analysis of 100 individual CNCs resulted in similar width and length values.

Synthesis of the mPEG-MC-BCP

Our synthetic design is presented in Scheme . It begins with the ring-opening polymerization of γ-benzyl-l-glutamate N-carboxyanhydride (BLG-NCA) using the methoxy-PEG-amine derivative (MeO-PEG48-NH2, number-average degree of polymerization (DPn) = 48 as determined by 1H NMR, Figure S3) as the initiator. We refer to the poly(γ-benzyl-l-glutamate) diblock copolymer obtained as mPEG48-PBLG-NH2. Two polymer samples with different DPn values (DPn = 18 or 25) were obtained, resulting from different molar ratios of monomer (BLG-NCA) to initiator. The values of DPn were determined using 1H NMR by comparing the end group methyl protons on MeO-PEG to the benzyl methylene signals in the backbone as described in the Supporting Information (see Figures S4 and S5). For instance, DPn = 18 of mPEG-PBLG-NH2 was calculated by comparing the integration of the signal at 3.23 ppm (end group methyl) to that at 5.03 ppm of benzyl methylene. These values of DPn for both polymers were in good agreement with the [monomer]/[initiator] feed ratios. Gel-permeation chromatography (GPC) analysis showed relatively narrow molecular weight distributions for both polymers (D̵ = 1.12 and 1.21 for polymers with DPn of 18 and 25, respectively). Thiourea was used as part of the initiating complex for the preparation of mPEG-PBLG18-NH2 but was omitted for mPEG-PBLG25-NH2 because it gave signals in the 1H NMR that overlapped with those of the pendant benzyl group (Figure S4). This may account for the broader polydispersity of the DPn = 25 sample.

Scheme 3

Synthesis of mPEG-PGlu(DTPA)-HyNic (n = 18 or 25)

The next step in the synthesis involved tBoc protection of the amino group at the polymer chain end to enable the transformation of the pendant groups. The extent of tBoc functionality was calculated by comparing the integration of the 1H NMR signal of the tBoc group to that of backbone benzyl ester groups (Figures S6 and S7). The calculated tBoc functionality for mPEG-PBLG18-Boc and mPEG-PBLG25-Boc was 80% and 94%, respectively.

The mPEG-PBLG-tBoc polymers were then treated with an excess of ethylenediamine (EDA) at room temperature (RT) to convert the benzyl groups to aminoethylamide pendant groups.[44] This reaction was carried out for different times ranging from 3 to 5 h to optimize both the amine conversion and preservation of the tBoc end group. By comparing the integration of the tBoc protons at 1.44 ppm to that of the signal at 4.34 ppm (backbone methine), the surviving tBoc functionality for mPEG-PGlu(EDA)18-Boc in this step stayed the same for the reaction times of 3 h (1H NMR spectrum not shown) and 5 h (Figure S8). The amine conversion, however, increased from 90% to 97% with increasing reaction time from 3 to 5 h. Thus, a reaction time of 5 h was chosen as the optimum condition. The surviving tBoc functionality for mPEG-PGlu(EDA)25-tBoc was calculated to be 73% for the reaction time of 5 h (Figure S9). For the attachment of DTPA-chelating groups, we followed the procedure described in Majonis et al.,[45] who employed an excess of DTPA prereacted with the coupling agent 4-(4,6-dimethoxy-1,3,5-triazin-2-yl)-4-methylmorpholinium chloride (DMTMM) to form DTPA monoamide groups. By 1H NMR, we found that the pendant amine groups of the polymer were fully functionalized with DTPA to within the ±5% error in measuring NMR signal integrations (Figures S10 and S11). At this stage, the tBoc group was removed using trifluoroacetic acid (TFA) in a mixture of TFA/water (1:1, v/v).[29,44] The 1H NMR spectra of the product showed a complete disappearance of the peak for the tBoc group (Figures S12 and S13).

To introduce the HyNic group onto the polymer chain end, the polymers with exposed amino end groups were treated with succinimidyl 6-hydrazinonicotinate acetone hydrazone (S-HyNic) at pH 8.0 (Scheme ). The appearance of 1H NMR signals in the range of 6.6–8.8 ppm indicated that the HyNic group was successfully attached to the polymer backbone (Figures S14 and S15).[30,31] However, the quantification of HyNic groups by 1H NMR was not reliable because the signals were too weak to integrate accurately. Instead, we examined the reaction of the polymers with 4-formylbenzoic acid (4FB-acid) to form a hydrazone. Hydrazone formation was monitored by UV/vis spectroscopy at 354 nm, taking advantage of the known molar extinction coefficient (ε354 nm = 29 000 M–1 cm–1) of the product.[46] In this way, we calculated a HyNic content of 73% for mPEG-PGlu(DTPA)18-HyNic and 52% for mPEG-PGlu(DTPA)25-HyNic (Figure S16).

CNCs Decorated with 4-formylbenzamide (4FB)

In our design (Scheme ), we planned to introduce 4FB groups onto the surface of the CNCs and then use these groups to react with the HyNic groups on the BCP chain ends via bis-aryl hydrazone (BAH) formation. To introduce 4FB groups, CNCs with amino groups at the surface were treated with an excess of N-succinimidyl-4-formylbenzamide (S-4FB) as illustrated in Scheme . The number of 4FB linkers was quantified via hydrazone-bond formation with 2-hydrazinopyridine at pH 5.0.[31] This process was followed by UV/vis spectroscopy, and the increase of absorbance at λ = 350 nm (ε350 = 18 000 M–1 cm–1) was plotted against time. In the example shown in Figure A, we found 23.1 μmol 4FB/g CNCs, which corresponds, for our CNC sample, to an average of 620 molecules per CNC. This value is consistent with the finding reported above of 680 −NH2 groups per CNC as determined in the reaction with FITC.

Scheme 4

Surface Modification of CNCs with 4FB Groups Followed by the Attachment of the Diblock Copolymers via Bis-Aryl Hydrazone Formation

Figure 3

(A–C) Following hydrazone-bond formation by UV/vis measurements (total volume 400 μL). (A) The coupling reaction between CNC–4FB (0.0625 wt %) and 2-hydrazinopyridine (5 mM) monitored at λ = 350 nm, ε350 = 18 000 M–1 cm–1. Coupling reactions monitored at 354 nm (ε350 = 29 000 M–1 cm–1) of CNC–4FB (0.0625 wt %) with large excesses of (B) mPEG-PGlu(DTPA)18-HyNic (290 μM) and (C) mPEG-PGlu(DTPA)25-HyNic (190 μM). Experiments were performed in sodium acetate buffer at pH 5.0, and the data were collected every 30 s. TEM images of the CNCs obtained after the covalent attachment of (D) CNC–PGlu(DTPA)18-mPEG and (E) CNC–PGlu(DTPA)25-mPEG.

In subsequent reactions, trace amounts of fluorescent dye molecules were attached to the CNCs at the same time that the 4FB groups were attached. In these reactions, we used a mixture of the dye derivative (either 5/6-carboxy-tetramethyl-rhodamine succinimidyl ester Rh552-NHS or Alexa Fluor 488-NHS) with S-4FB and carried out the reaction as described above. The resulting dye-labeled CNC suspensions were then thoroughly washed and dialyzed to remove the unreacted dye in preparation for the polymer conjugation reaction. We refer to the dye-labeled 4FB-modified CNCs (CNC–4FB) samples as Rh552–CNC–4FB and A488–CNC–4FB, respectively.

Alexa Fluor 488 is an anionic dye with spectroscopic properties similar to those of FITC but with greater photostability for imaging studies. Tetramethylrhodamine is a cationic dye. It is known that anionic (e.g., FITC) and cationic (e.g., rhodamine-B isothiocyanate (RBITC)) dyes attached to the surface of CNCs can affect their interaction with cells and in some instances impart toxicity to cells.[47] As a result, in our design, these dyes are to be bound adjacent to the CNC surface and thus shielded by the PEG corona. As we show below, neither dye-labeled CNCs, following grafting of the BCP, were toxic to HEYA8 cells (a human ovarian cancer cell line) and both behaved similarly in studies evaluating penetration into MCTSs.

The Grafting-to Reaction: Covalent Attachment of the BCPs to the CNCs

For the CNC–4FB sample, we examined the polymer conjugation reaction by UV/vis monitoring of BAH formation at λ = 354 nm.[30] The concentration of the CNC–4FB was diluted to 0.1 wt % with sodium acetate buffer (pH 5.0), and an excess of polymer (mole ratio HyNic:4FB = 15:1) in sodium acetate buffer was added. As shown in Figure B,C the reaction was slow, with the product peak growing over 20 h. The reaction was stopped at this point, and the extinction coefficient of the BAH groups (ε354 = 29 000 M–1 cm–1) was used to calculate the yield of the polymer attached to the CNCs based on the change in absorbance. In this way, we found 12.3 μmol mPEG-PGlu(DTPA)18 and 6.9 μmol mPEG-PGlu(DTPA)25 per gram of CNCs after a reaction time of 20 h. As shown in the TEM images in Figure D,E, the polymer conjugation reaction did not affect the size or shape of the CNCs and preserved the rodlike shape of the pristine CNCs. The BCP attachment corresponds to an average of 330 polymer molecules per CNC for CNC–PGlu(DTPA)18-PEG and 185 polymer molecules per CNC for CNC–PGlu(DTPA)25-PEG. These values are included in Table . For the shorter polymer, approximately half of the 620 4FB linkers available on the CNC surface served as sites of polymer attachment, whereas for the longer polymer, in this particular reaction, about 30% of the 4FB linkers reacted with the polymer.

Table 1

Characteristics of mPEG-PGlu(DTPA)18-HyNic and mPEG-PGlu(DTPA)25-HyNic and Their Precursor Polymers

sample	MnNMRa (kDa)	MnGPCb (kDa)	D̵b	grafting densityc (μmol/g of CNCs)	grafting densityc,d (chains per CNC)
mPEG-PBLG18-NH2	5.9	6.6	1.1
mPEG-PBLG25-NH2	7.4	9.1	1.2
mPEG-PGlu(DTPA)18-HyNic	12	7.6	1.2	12.3	330
mPEG-PGlu(DTPA)25-HyNic	16	11	1.3	6.9	185

On the basis of DPn from end group analysis by 1H NMR. For the DTPA-containing polymer, this calculation assumes that all of the DTPA groups were in the protonated form.

From GPC analysis with refractive index (RI) detection. For the mPEG-PBLG polymers, GPC analyses were run in 1-methyl-2-pyrrolidinone (NMP) and analyzed using polystyrene standards. For the mPEG-PGlu(DTPA)-HyNic polymers, GPC analyses were run in phosphate buffer ([PB], 0.2 M KNO3, pH 8.5) with poly(methacrylic acid) (PMAA) standards.

Determined from UV–vis measurements of the HyNic–4FB conjugation reaction (Figure ).

Calculated from the grafting density in μmol/g of CNCs with reference to the mean surface area and density of the CNC sample.

In Figure S17, we provide details of the reactions of mPEG-PGlu(DTPA)25-HyNic with Rh552–CNC–4FB and A488–CNC–4FB. The UV–vis spectra show that the dye makes only a small contribution to the sample absorbance even at λmax (ΔA ≈ 0.02). This level of dye content is sufficient to make the CNCs highly fluorescent, but it is small enough that it does not interfere with monitoring bis-aryl hydrazone formation in the reaction of the HyNic polymers with the dye–CNC–4FB derivatives. Unsurprisingly, the reaction of the polymer with the dye-labeled CNCs occurred over the same timescale as the reaction of the polymer with the dye-free CNC–4FB.

Kinetics of the HyNic–4FB Conjugation Reaction

The plots of ΔA vs reaction time shown in Figure provide insights into the reaction kinetics of BAH formation. In both reactions, the pyridylhydrazine derivatives were in a large excess relative to the CNC–4FB. For the reaction of CNC–4FB with 2-hydrazinopyridine (Figure A), the reaction approached completion after ca. 25 min. In contrast, the polymer–HyNic reactions were much slower and exhibited a marked slowing down. None of these growth curves for the increase of absorbance fitted a simple exponential form. They could, however, be fitted with a stretched exponential profilewhere β is the stretch exponent in the range of 0 < β ≤ 1, ΔA0 = ΔA (t ≅ ∞) and k is the characteristic relaxation rate. Stretched exponentials are often characteristic of a distribution of reaction rates.[48] The fitting curves are shown in Figure S18 for the reaction with CNC–4FB, and the fitting parameters are included in Table .

Table 2

Reaction of HyNic–4FB with CNC–4FB, 2-Hydrazinopyridine, and the HyNic–BCPs: Concentrations Used and the Fitted Stretched-Exponential Parameters β, ΔA0, and ka

sample	conc. (mM)	ΔA0	β	k (min–1)
2-hydrazinopyridine	5	0.37	0.63	0.35
mPEG-PGlu(DTPA)18-HyNic	0.29	0.22	0.74	3.4 × 10–3
mPEG-PGlu(DTPA)25-HyNic	0.19	0.12	0.68	4.0 × 10–3

All experiments were performed at RT.

A few more comments are in order about the reaction kinetics. Dirksen and Dawson[49] have shown that the reaction of 6-hydrazinopyridine and aromatic aldehydes follows simple second-order kinetics. One predicts that in the presence of a large excess of one of the reactants, the reaction would follow pseudo-first order kinetics. In the reaction of the mPEG-PGlu(DTPA)-HyNic polymer samples with a large excess of 4FB-acid, however, we found more complicated (nonexponential) behavior (Figure S16). This result is likely a consequence of the fact that the hydrazine group in the polymer–HyNic is protected as acetone hydrazone, which first hydrolyzes in the reaction to form the 6-hydrazinopyridine and then reacts with 4FB-acid.

Controlling the Number of Polymer Molecules Per CNC

The basic idea of the experiments described in this section is that one could monitor the extent of the reaction of mPEG-PGlu(DTPA)-HyNic with CNC–4FB at 354 nm and stop the reaction at various end points. In this way, one could control the mean number of polymer molecules per CNC. The extent of polymer coupling to the CNCs can be calculated from the ΔA value. More important from the perspective of our intended applications, information about the mean number of polymer molecules per CNC can be obtained by saturating the DTPA chelator with trivalent metal ions and measuring the metal-ion content of the CNC–polymer hybrids. Here, we take advantage of the strong binding of lanthanide ions with this chelator. For example, the binding affinity of Tb3+ with DTPA-monoamide can be estimated to be on the order of 1022 M–1 at physiological pH.[50] The analysis of the sample by inductively coupled plasma-mass spectrometry (ICP-MS) enables the metal content and the polymer content of the BCP–CNC conjugates to be determined.

We tested this idea by running seven parallel reactions of mPEG-PGlu(DTPA)25-HyNic with Rh552–CNC–4FB under identical conditions, monitoring one of the reactions at 354 nm. The individual reactions were stopped at different times ranging from 10 to 1400 min. After each time point, the reaction was quenched by quickly diluting with NaHCO3 buffer (pH = 9.0). Then, the resulting mixture was dialyzed against water followed by centrifugation to remove the excess polymer, after which the samples were prepared for Tb3+ loading. Using the rhodamine-labeled CNC–4FB suspension enabled us to account for the loss of the sample after purification by comparing the UV/vis signal of rhodamine (λmax = 552 nm). The details are presented in the Supporting Information.

The results of these experiments are shown in Figure . The continuous curve shown in black describes the mean number of mPEG-PGlu(DTPA)25-hydrazone groups per CNC calculated from the UV/vis measurements monitored over 25 h at 354 nm. The red dots and accompanying error bars represent triplicate ICP-MS measurements from which we also calculated the mean number of polymers per CNC. This calculation assumes that a Tb3+ ion was bound to every DTPA group. Previous experiments from our laboratory with Gd3+ ions and somewhat different metal-chelating polymers demonstrated that the metal ions became bound to all of the DTPA pendant groups.[45]

Figure 4

Polymer–CNC conjugate formation in the reaction between the rhodamine-labeled Rh552–CNC–4FB and mPEG-PGlu(DTPA)25-HyNic expressed as the time evolution of the mean number of polymer molecules per CNC. The black line represents points obtained by monitoring bis-aryl hydrazone formation through its UV/vis absorbance at 354 nm. The red points refer to individual reactions quenched at the indicated time points. For these samples, the DTPA groups on the polymers were saturated with Tb3+ ions, and the Tb content of each was measured by ICP-MS.

It is gratifying to find such good agreement between the two sets of experiments. The main conclusion to draw from these experiments is that monitoring the reaction via hydrazone formation allows one to stop the reaction when the desired number of polymers, on average, is attached to the CNCs. At the end of the reaction, the polymer–CNC conjugates can be purified, and they maintain their colloidal stability and structural integrity.

In Vitro Evaluation of CNC Cytotoxicity and Penetration into MCTSs

The CNCs prepared in this study were labeled with two different fluorescent probes (rhodamine and Alexa Fluor 488), which enables the assessment of penetration into MCTSs of a human ovarian cancer cell line (HEYA8) by confocal microscopy. Although nanoparticles have demonstrated benefits over conventional chemotherapy, the heterogeneous distribution of nanoparticles as a result of poor tumor penetration remains one of the major challenges limiting the efficacy of nanomedicine-based cancer therapy.[51] Because of the characteristic features of the tumor microenvironment, namely, the dense extracellular matrix, as well as the abnormal vasculature and high interstitial fluid pressure, the distribution of nanoparticles is largely limited to areas close to the blood vessels and in the tumor periphery.[51,52] As such, evaluation of tumor penetration is key to understanding the transport properties that ultimately impact the efficacy of nanoparticle-based drug delivery systems. By the conjugation or incorporation of an imaging agent, penetration properties of nanoparticles have been evaluated in vitro and in vivo.[53−57] In particular, 3D MCTSs, which mimic the physiological gradients (e.g., pH, oxygen, nutrient, and proliferation) present in a solid tumor,[58−61] provide a robust model for the evaluation of nanoparticle penetration.[57,61]

Prior to penetration studies, the cytotoxicity of Rh552–CNC–4FB–HyNic-PGlu(DTPA)25-mPEG and A488–CNC–4FB–HyNic-PGlu(DTPA)25-mPEG was assessed in monolayer cells to determine the appropriate concentration for penetration studies in MCTSs. The evaluation of cell viability using the acid phosphatase (APH) assay showed that both Rh552–CNC–4FB–HyNic-PGlu(DTPA)25-mPEG and A488–CNC–4FB–HyNic-PGlu(DTPA)25-mPEG are nontoxic over the range of concentrations studied up to 0.0035 wt % (Figure S19), and thus, this concentration was chosen for penetration studies.

Following 24 h of incubation with Rh552–CNC–4FB–HyNic-PGlu(DTPA)25-mPEG or A488–CNC–4FB–HyNic-PGlu(DTPA)25-mPEG, the penetration of CNCs into MCTSs was quantitatively evaluated by obtaining optical slices at three depths (30, 60, and 90 μm) from the spheroid surface. For each optical slice, the distribution of CNCs was assessed by quantifying the fluorescence intensity in three equally spaced concentric regions (i.e., periphery, intermediate, and core). As shown in Figure , at a 30 μm distance from the surface of spheroids, CNCs are homogeneously distributed across the three regions (periphery, intermediate, and core) of the MCTSs. With increasing distance from the spheroid surface, a progressive decrease in CNC penetration from the periphery to intermediate and core regions was observed. At 90 μm, the concentration of CNCs was significantly higher in the periphery compared to that in the intermediate and core regions with very low levels of fluorescence detected in the core. Using two different fluorescent probes (i.e., rhodamine and Alexa Fluor 488), similar penetration trends were observed, suggesting that these fluorescently labeled CNCs provide a robust platform for assessing the penetration of CNCs for drug delivery applications. The influence of the particle shape on penetration into MCTSs and cellular endocytosis will be evaluated in future studies by in vitro comparison of the rodlike CNCs to their spherical counterparts.

Figure 5

(a) MCTSs of HEYA8 imaged by confocal microscopy with CellMask green plasma membrane staining (top) or under bright field (bottom). Evaluation of CNC penetration into MCTSs of HEYA8 following 24 h of incubation with (b) Rh552–CNC–4FB–HyNic-PGlu(DTPA)25-mPEG and (c) A488–CNC–4FB–HyNic-PGlu(DTPA)25-mPEG at 37 °C. The total fluorescence intensity per unit area was calculated for the periphery, intermediate, and core regions of each MCTS. Data represent mean ± SD in each area (n = 4 MCTSs). Statistically significant differences (p < 0.05) between each region are denoted by (*). Representative images of (d) Rh552–CNC–4FB–HyNic-PGlu(DTPA)25-mPEG and (e) A488–CNC–4FB–HyNic-PGlu(DTPA)25-mPEG penetration into MCTSs at 30, 60, and 90 μm depths from the MCTS surface. The scale bars are 100 μm.

Conclusions

We report the synthesis of PEG-MC-BCPs with two different numbers of repeat units containing DTPA (mPEG-PGlu(DTPA)18-HyNic and mPEG-PGlu(DTPA)25-HyNic). The polymers were characterized by 1H NMR and grafted onto the CNC–NH2 via bis-aryl hydrazone conjugation. UV/vis spectroscopy demonstrated that the conjugation reaction was efficient, with grafting density of 12.3 μmol/(g CNC) for mPEG-PGlu(DTPA)18-HyNic and 6.9 μmol/(g CNC) for mPEG-PGlu(DTPA)25-HyNic. The mean number of polymer molecules that could be attached per CNC was found to be 330 for the shorter polymer and 185 for the longer polymer. By varying the reaction time in the presence of excess BCP, the mean number of polymer molecules attached per CNC ranged from 40 to 170. We showed that the polymer grafting of CNCs via bis-aryl hydrazone formation is controllable, which is, to the best of our knowledge, the first report on the controllable polymer grafting of CNCs.

These conjugates could be covalently labeled with anionic (Alexa Fluor 488) or cationic (tetramethylrhodamine) dyes in such a way that the dyes were shielded by the PEG corona of the BCPs. Importantly, as shown, the conjugation of the dyes did not impart toxicity to the CNCs.

This finding is significant in drug delivery applications as manipulating the degree of conjugation could control the loading amount of therapeutic agents, which is in our case the radionuclide, and thus meet the different demands of treatments.

Experimental

Materials

A suspension of CNCs (obtained by acid hydrolysis of wood pulp, 0.99 wt % sulfur content) was purchased from the USDA Forest Product Laboratory (USA). The reagents and solvents are DTPA (98%, Aldrich), EDA (Sigma-Aldrich), di-tert-butyl dicarbonate (tBoc2O, Sigma-Aldrich), l-glutamic acid γ-benzyl ester (BLG, Sigma-Aldrich), 2-hydrazinopyridine (Sigma-Aldrich), ammonium hydroxide (28%, Sigma-Aldrich), epichlorohydrin (99%, Fluka), N,N-dimethylformamide (DMF, 99.8%, Sigma-Aldrich), dimethyl sulfoxide (DMSO, 99.5%, Sigma-Aldrich), and sodium hydroxide (NaOH, 97%, Sigma-Aldrich). FITC (Fisher Scientific), NHS-rhodamine (ThermoFisher Scientific), Alexa Fluor 488-NHS (Alexa Fluor 488 carboxylic acid succinimidyl ester, Life Technologies) were used without further purification unless otherwise noted. Water was purified through a Milli-Q water purification system (18 MΩ cm). All buffers were prepared in our laboratory. DMTMM (Acros Organics, 99+%), Millipore Amicon spin filters (15 or 4 mL, 3, 10, or 50 kDa molecular-weight cutoff (MWCO) and 5 mL, 50 kDa) were purchased from Fisher Scientific. MeO-PEG-NH2 (M = 2000 (2K)) was purchased from Jenkem as the HCl salt and converted to the free amine by treatment with triethylamine. The analysis of this sample by 1H NMR (see Figure S3 below) gave a DPn = 48. BLG-NCA was synthesized as described previously.[44] Succinimidyl 4-formylbenzoate (S-4FB) and succinimidyl 6-hydrazinonicotinate acetone hydrazone (S-HyNic) were synthesized as described in Grotzky et al.[31] 4FB-acid was purchased from Sigma-Aldrich. S-HyNic and S-4FB are available commercially from Solulink, USA, www.solulink.com.

Instrumentation

TEM

Sample morphology was monitored by TEM using a Hitachi H-7000 instrument operating at 100 kV. CNC samples were diluted to 0.01 wt % with water, followed by sonication (50 W sonication bath, 23 °C) for 10 min. A droplet was placed onto a 200 mesh carbon–copper grid. The excess suspension was wicked off with filter paper. Two droplets of 2% (w/v) aqueous uranyl acetate solution was then deposited onto the dried sample with 1 min of drying, and the excess staining agent was removed by touching a piece of filter paper to the edge of the TEM grid. Images were recorded using a TEM operated at a voltage of 100 kV. TEM images were analyzed using the software ImageJ (National Institutes of Health, USA).

Ultraviolet-Visible Spectroscopy

UV/vis spectroscopy was carried out on a Cary Series 300 UV–vis spectrophotometer (Agilent Technology, USA). All time-dependent UV/vis spectra were recorded at RT with a 70 μL UV–cuvette cell (l = 1.0 cm).

ICP-MS

ICP-MS measurements were carried out on an ELAN 9000 (PerkinElmer) ICP/MS for 159Tb analysis. All samples were diluted with 2% HNO3 until ppb concentrations were obtained. The standard solutions were prepared by a series of 10-fold dilutions of terbium standard solution (10 mg/L, 2% HNO3, PerkinElmer) in triplicate to obtain three standard solutions (0.1, 1, and 10 ppb of Tb).

Surface Modification of CNCs

Introduction of Epoxy Groups and Their Transformation to Amine Groups

A CNC sample (100 mL, 11.9 wt %) was diluted to 3.0 wt % with water and sonicated for 10 min at RT. Then, the suspension was filtered through a syringe filter (0.45 μm) to remove any aggregates. The solids content was found to be 2.65 wt % by freeze-drying an aliquot of the filtered CNC suspension.

To introduce surface epoxy groups, an aliquot of the filtered CNC suspension (10 mL, 2.65 wt %, 265 mg of CNC solids) was mixed with 25 mL of NaOH (1 M). This solution was then treated with epichlorohydrin (5.8 mL, 0.28 mol/g CNC) and heated to 60 °C.

To investigate the effect of time on the modification efficacy, several epichlorohydrin reactions were run in parallel according to the above procedure and stopped at different times (0.67, 1.0, 2.0, 4.0, 6.0, or 20 h). After each given time, the reaction mixture was centrifuged (4000 rpm, 10 min) and resuspended in water multiple times until the pH was below 12 (pH meter, Mettler Toledo).

For each sample, immediately after the pH of the epoxy-modified CNCs (CNC–epoxy) dropped below 12, the suspension was transferred to a 100 mL round-bottom reaction flask. The pH of the CNC–epoxy dispersion was adjusted to 12 with 50% (w/v) NaOH. Then, an excess of aqueous ammonia (28%, 10 mL, 5 mL/g of cellulose) was added, and the reaction mixture was heated to 60 °C for 3 h. Excess ammonium hydroxide was removed by three cycles of centrifugation (12 000 rpm, 15 min)–redispersion in DI water (15 mL). The purified amino-modified CNC suspension was then stored at 4 °C.

Quantifying the Amine Content of CNC–NH2 with FITC

Aliquots of CNC–NH2 (2.0 mL) were each mixed with NaHCO3 buffer (2.0 mL, 100 mM, pH = 9.0). Then, a solution prepared by dissolving FITC (0.5 mg, 0.32 mmol/g CNCs) in DMF (200 μL) was added to the NaHCO3-buffered suspensions containing CNC–NH2. The reaction mixture was stirred overnight in the dark and then dialyzed for 5 days with water (water was changed twice daily). When the diffusion of excess FITC from the reaction mixture had ceased according to visual inspection and UV/vis monitoring, the suspension was sonicated (20 min, ice-bath cooling) and filtered through a syringe filter (0.45 μm) to remove aggregates. The final suspension (0.5 wt %) had a pH of 6. The amount of FITC attached to each of the CNC–NH2 samples was determined by UV/vis measurements. The pH of the CNC–NH2 samples was adjusted to 11 with NaHCO3 buffer (100 mM). Free FITC in ammonium hydroxide (28%, pH 11) with a range of concentrations was used to generate a calibration curve (Figure S2). The molar extinction coefficient ε490 for FITC amide was calculated to be 5.2 × 104 M–1 cm–1.

4FB Modification of CNC–NH2

An aliquot of an aqueous suspension of CNC–NH2 (7.5 mL, 66.7 mg) was mixed with PB (4.0 mL, 200 mM, pH = 8.0). A solution of S-4FB (5.0 mg, 10 equiv to −NH2 on CNCs) in DMF (1.0 mL) was added to the suspension. Then, the mixture was stirred for 24 h at RT. Afterward, the suspension was sonicated for 20 min with ice-bath cooling and washed with water (4000 rpm, 10 × 15 mL) using a MWCO 10 kDa spin filter to obtain a suspension of CNC–4FB (solid content, 0.75 wt %).

Quantification of 4FB Groups on CNC–4FB

A suspension of CNC–4FB (0.75 wt %) was diluted to 0.1 wt % with sodium acetate buffer (100 mM, pH 5.0). 2-Hydrazinopyridine (4.6 mg, 0.042 mmol) was dissolved in DMF (770 μL) to make a 50 mM solution. The CNC–4FB suspension (360 μL, 0.1 wt %) and 2-hydrazinopyridine solution (40 μL, 50 mM) were mixed in a quartz cell (l = 1.0 cm). The absorbance of the solution was monitored at RT by UV/vis spectroscopy (λ = 350 nm) with data points taken automatically at 30 s intervals until absorbance appeared to reach a plateau (Figure S20).

BCP Synthesis and Characterization

Synthesis of mPEG-PBLG18-NH2

Methoxy(polyethylene glycol)48-poly(γ-benzyl-l-glutamate)18 was synthesized by the ring-opening polymerization of BLG-NCA (1.2 g, 4.56 mmol) catalyzed by thiourea (0.8 mL, 1 M) with an amine initiator MeO-PEG2k-NH2 (0.4 g, 0.2 mmol) to obtain a polymer with an amine group at one end. The molar ratio of BLG-NCA:initiator was 22:1. The reaction was carried out in 15 mL of DMF for 6 h at 0 °C under a nitrogen atmosphere. Then, the solution was poured into diethyl ether to precipitate PEG-PBLG-NH2. The crude polymer was dissolved in dichloromethane (DCM) and reprecipitated in diethyl ether three times. The PBLG solid was collected and dried under vacuum. Yield = 1.31 g (82%).

1H NMR (DMSO-d6) (Figure S4): δ (ppm, integrated peak areas reported based on the backbone benzyl methylene group) 8.36–7.97 (br, −NH–, 1H, integration = 0.33), 7.42–7.18 (m, phenyl, 5H, integration = 2.55, peak overlapped with a thiourea amine peak), 5.03 (m, −CH2–phenyl, 2H per monomer, integration = 1.00), 4.33–3.82 (br, backbone methine, 1H per monomer, integration = 0.44), 3.23 (s, −OCH3, 3H, integration = 0.08), 2.47–1.70 (m, −CH2CH2–, 4H, integration = 1.82, peak overlapped with the NMR solvent peak). The degree of polymerization was calculated by comparing the integration of the signal at 3.23 ppm (end group −OCH3) to that at 5.03 ppm (backbone benzyl methylene), where DPn = 18, MnNMR = 5.9 kDa; GPC (NMP, RI): MnGPC = 6.6 kDa, D̵ = 1.12.

Synthesis of mPEG-PBLG25-NH2

This polymer was synthesized following a similar protocol as described above. In brief, BLG-NCA (1.5 g, 5.7 mmol) and MeO-PEG2k-NH2 (0.4 g, 0.2 mmol) were used as the monomer and initiator, respectively. The molar ratio of BLG-NCA:initiator was 28:1. No thiourea was used for this reaction. Yield = 1.48 g (78%).

1H NMR (DMSO-d6) (Figure S5): δ (ppm, integrated peak areas reported based on the backbone benzyl methylene group) 8.64–7.49 (br, −NH–, 1H, integration = 0.41), 7.40–7.18 (m, phenyl, 5H, integration = 2.60), 5.05 (m, −CH2–phenyl, 2H per monomer, integration = 1.00), 4.33–3.79 (br, backbone methine, integration = 0.45), 3.26 (s, −OCH3, 3H, integration = 0.06), 2.48–1.60 (m, −CH2CH2–, 4H, integration = 1.75, peak overlapped with the NMR solvent peak). The degree of polymerization was calculated by comparing the integration of the signal at 3.26 ppm (end group −OCH3) to that at 5.05 ppm (backbone benzyl methylene), where DPn = 0.5/(0.06/3) = 25, MnNMR = 7.4 kDa; GPC (NMP, RI): MnGPC = 9.1 kDa, D̵ = 1.21.

Introduction of Terminal tBoc Groups: mPEG-PBLG18-tBoc

A sample of mPEG-PBLG18-NH2 (0.78 g, 0.12 mmol) and tBoc2O (0.5 mL, 2.2 mmol) were dissolved in DMF (8.0 mL), and the solution was stirred at RT overnight. Afterward, the solution was poured into diethyl ether to precipitate the product. The crude polymer was dissolved in DCM and reprecipitated in diethyl ether three times. The mPEG-PBLG18-tBoc solid was dried under vacuum. Yield = 0.54 g (78%).

1H NMR (DMSO-d6) (Figure S6): δ (ppm, integrated peak areas reported based on the backbone benzyl methylene group) 8.30–7.95 (br, backbone, −NH–, 1H, integration = 0.37), 7.38–7.16 (m, phenyl, 5H, integration = 2.56), 5.03 (m, −CH2–phenyl, 2H per monomer, integration = 1.00), 4.41–3.82 (br, backbone methine, 1H, integration = 0.44), 3.23 (s, −OCH3, 3H, integration = 0.08), 2.48–1.71 (m, −CH2CH2CO–, 4H, integration = 1.80, peak overlapped with the NMR solvent peak) 1.52–1.31 (m, t-Boc end group, 9H, integration = 0.20). The functionality of t-Boc was calculated by comparing the integration of the 1H NMR signal at 1.52–1.31 ppm (t-Boc end group) to that at 5.03 ppm (backbone benzyl methylene), where % t-Boc functionality = 0.20/((0.5/18) × 9) × 100% = 80%. The degree of polymerization was determined by comparing the integration of the peak at 3.23 ppm (end group −OCH3) to that at 5.03 ppm (backbone benzyl methylene). We found DPn = 18, which agrees with the value calculated for mPEG-PBLG18-NH2.

mPEG-PBLG25-tBoc

This polymer was synthesized following a similar protocol using mPEG-PBLG25-NH2 (0.99 g, 0.16 mmol) and tBoc2O (1.0 mL, 4.4 mmol) in DMF (8.0 mL). Yield = 0.82 g (83%).

1H NMR (DMSO-d6) (Figure S7): δ (ppm, integrated peak areas reported based on the backbone benzyl methylene group) 8.69–7.92 (br, backbone, −NH–, 1H, integration = 0.35), 7.38–7.16 (m, phenyl, 5H, integration = 2.58), 5.01 (m, −CH2–phenyl, 2H per monomer, integration = 1.00), 4.32–3.78 (br, backbone methine, 1H per monomer, integration = 0.54), 3.23 (s, −OCH3, 3H, integration = 0.06), 2.46–1.73 (m, −CH2CH2CO–, 4H, integration = 1.69, peak overlapped with the solvent peak) 1.48–1.28 (m, t-Boc end group, 9H, integration = 0.17). The functionality of t-Boc was calculated by comparing the integration of the 1H NMR signal at 1.48–1.28 ppm (t-Boc end group) to that at 5.01 ppm (backbone benzyl methylene), where % t-Boc functionality = 0.17/((0.5/25) × 9) × 100% = 94%. The degree of polymerization was calculated by comparing the integration of the peak at 3.23 ppm (end group −OCH3) to that at 5.01 ppm (backbone benzyl methylene). We found DPn = 25, which agrees with the value calculated for mPEG-PBLG25-NH2.

Aminolysis of the PBLG Block

mPEG-PGlu(EDA)18-tBoc

EDA (4.0 mL, 140 mmol) was added into a NMP solution (4.0 mL) containing mPEG-PBLG18-tBoc (100 mg, 0.23 mmol repeat units) and 2-pyridone (2HP, 160 mg, 1.7 mmol, 7 equiv to the repeat unit of the polymer), and the solution was stirred at 0 °C for 5 h. The reaction mixture was precipitated with diethyl ether (100 mL), and the solid was dissolved in aqueous acetic acid solution (5% v/v). Then, this solution was washed using a MWCO 3 kDa spin filter with sodium acetate buffer (100 mM, pH 5.0, 4000 rpm, 3 × 15 mL) and followed by water (4000 rpm, 3 × 15 mL). Afterward, the solution was lyophilized to obtain mPEG-P(GluEDA)18-tBoc as the acetate salt (ca. 70 mg, yield = 71%).

1H NMR (D2O) (Figure S8): δ (ppm, integrated peak areas reported based on the backbone methine group) 7.43 (br, residual phenyl group from PBLG, 5H, integration = 0.14), 4.34 (m, backbone methine, 1H per monomer, integration = 1.00), 3.51 (t, J = 6.0 Hz, −NH–CH2–CH2–NH3+, 2H, integration = 2.08), 3.38 (s, −OCH3, 3H, integration = 0.18), 3.15 (t, −NH–CH2–CH2–NH3+, J = 6.0 Hz, 2H, integration = 1.98), 2.39 (m, −CH–CH2–CH2–CO–, 2H, integration = 2.07), 2.17–1.88 (br, −CH–CH2–CH2–CO–, 2H, integration = 1.87), 1.44 (s, t-Boc end group, 9H, integration = 0.41). The functionality of surviving t-Boc was calculated by comparing the integration of the peak at 1.44 ppm (t-Boc end group) to that at 4.34 ppm (backbone methine), where the surviving % t-Boc functionality = 0.40/((1/18) × 9) × 100% = 80%. The conversion of aminolysis was calculated by comparing the integration of the residual phenyl group (7.43 ppm) to that of the backbone methine (4.34 ppm), where the conversion of aminolysis = (1 – 0.14/5) × 100% = 97%.

mPEG-PGlu(EDA)25-tBoc

This polymer was synthesized following a similar protocol. In brief, EDA (12 mL, 0.4 mol) was added into a NMP solution containing mPEG-PBLG25-tBoc (0.5 g, 1.14 mmol repeat units) and 2HP (1.0 g, 10 mmol, 9 equiv to the repeat unit of the polymer). Yield = 0.44 g (88%).

1H NMR (D2O) (Figure S9): δ (ppm, integrated peak areas reported based on the backbone methine group) 7.33 (br, residual phenyl group from PBLG, 5H, integration = 0.17), 4.33 (m, backbone methine, 1H per monomer, integration = 1.00), 3.50 (t, J = 6.0 Hz, −NH–CH2–CH2–NH3+, 2H, integration = 1.96), 3.39 (s, −OCH3, 3H, integration = 0.12), 3.14 (t, −NH–CH2–CH2–NH3+, J = 6.0 Hz, 2H, integration = 1.99), 2.39 (m, −CH–CH2–CH2–CO–, 2H, integration = 2.12), 2.22–1.91 (br, −CH–CH2–CH2–CO–, 2H, integration = 2.00), 1.43 (s, t-Boc end group, 9H, integration = 0.26). The surviving t-Boc functionality was calculated by comparing the integration of the peak at 1.43 ppm (t-Boc end group) to that at 4.33 ppm (backbone methine), where % t-Boc functionality = 0.26/(1/25 × 9) = 73%. The conversion of aminolysis was calculated by comparing the integration of the residual phenyl group (7.33 ppm) to that of the backbone methine (4.33 ppm), giving a conversion of aminolysis of 97%.

Introduction of DTPA Metal Chelators

mPEG-PGlu(DTPA)18-tBoc

A DTPA solution (10 mL containing 2.5 g of DTPA, 6.3 mmol, 30 equiv to each polymeric amino group, pH adjusted with 1 M NaOH to 8.5) was mixed with a freshly prepared DMTMM solution (1.0 mL containing 150 mg of DMTMM, 4 equiv to amine groups on the polymer) in a 50 mL round-bottom flask. The mixture was stirred for 5 min, then was poured into a mPEG-PGlu(EDA)18-tBoc solution (4.0 mL of H2O, containing 50 mg of polymer). The reaction was stirred for 30 min at RT. Afterward, the solution was transferred to a 15 mL MWCO 3 kDa spin filter and washed with water (4000 rpm, 9 × 15 mL). Finally, the solution was lyophilized to obtain mPEG-PGlu(DTPA)18-tBoc (65 mg, 60% yield).

1H NMR (D2O) (Figure S10): δ (ppm, integrated peak areas reported based on the backbone methine group) 7.43 (br, residual phenyl group from PBLG, 5H, integration = 0.13), 4.36 (m, 1H per monomer, integration = 1.00), 3.68–2.95 (br, m, −NH–CH2–CH2–NH– of the EDA pendant group, 18H from DTPA per monomer and −OCH3, integration = 20.76), 2.35 (br, −CH–CH2–CH2–CO–, 2H, integration = 2.17), 2.22–1.88 (br, −CH–CH2–CH2–CO–, 2H, integration = 2.01), 1.44 (s, t-Boc end group, 9H, integration = 0.37). The conversion of aminolysis is 97%, which agrees with the precursor PEG-P(GluEDA)18-tBoc. The functionality of the DTPA group was calculated by comparing the backbone methine signals to the that of EDA associated with DTPA. Each DTPA contains 18 protons from EDA, and the DTPA functionality = (20.76 – 4 – (1/18) × 3)/(22 – 4) × 100% = 92%. The tBoc functionality was calculated by comparing the tBoc signal to the backbone methine signal, giving a t-Boc functionality of 74%.

mPEG-PGlu(DTPA)25-tBoc

This polymer was synthesized following a similar protocol as described above using PEG-PGlu (EDA)25-tBoc (100 mg, 0.23 mmol repeat units), DTPA (3 g, 7.6 mmol, 30 equiv to each polymeric amino group), and DMTMM (300 mg). Yield = 130 mg (65%).

1H NMR (D2O) (Figure S11): δ (ppm, integrated peak areas reported based on the backbone methine group) 7.41 (br, residual phenyl group from PBLG, 5H, integration = 0.17), 4.37 (m, 1H per monomer, integration = 1.00), 3.65–2.88 (br, m, −NH–CH2–CH2–NH– of the EDA pendant group and 18H from DTPA per monomer, integration = 22.37), 2.36 (br, −CH–CH2–CH2–CO–, 2H, integration = 2.00), 2.23–1.79 (br, −CH–CH2–CH2–CO–, 2H, integration = 1.99), 1.44 (s, t-Boc end group, 9H, integration = 0.28). The conversion of aminolysis is 97%, which agrees with the precursor PEG-PGlu(EDA)25-tBoc. The functionality of the DTPA group was calculated by comparing the integrals of backbone methine to those of the EDA associated with DTPA. The functionality of the DTPA group = (22.37 – 4 – (1/25) × 3)/(22 – 4) × 100% = 100%. The t-Boc functionality was calculated by comparing the tBoc signal to the backbone methine signal, giving a tBoc funcitonality of 75%.

Deprotection: Removal of tBoc Groups to Yield the Amino Polymer

mPEG-PGlu(DTPA)18-NH2

mPEG-P(Glu-EDA-DPTA)18-tBoc (47 mg, 0.009 mmol) was dissolved in water (1.0 mL). TFA (1.0 mL, 13 mmol) was added, and the solution was stirred at RT for 2 h. Afterward, the solution was diluted to 15 mL with water and washed using a 15 mL MWCO 3 kDa spin filter (4000 rpm, 8 × 15 mL). Finally, the solution was lyophilized to obtain PEG-P(GluEDA-DPTA)18-NH2 (37 mg, 78%).

1H NMR (D2O) (Figure S12): δ (ppm, integrated peak areas reported based on the backbone methine group) 7.43 (br, residual phenyl group from PBLG, 5H, integration = 0.10), 4.35 (m, 1H per monomer, integration = 1.00), 4.08–3.10 (backbone ethylene of PEG, EDA, 18H from DTPA, and −OCH3, integration = 31.28), 2.36 (br, −CH–CH2–CH2–CO–, 2H, integration = 2.16), 2.22–1.88 (br, −CH–CH2–CH2–CO–, 2H, integration = 2.16). The disappearance of the peak at 1.44 ppm indicated that the t-Boc group was completely removed. The functionality of the DTPA group was calculated by comparing the backbone methine signals to those of the EDA pendent groups associated with DTPA, where DTPA functionality = 92%.

mPEG-PGlu(DTPA)25-NH2

This polymer was synthesized following a similar protocol as described above using mPEG-PGlu(DPTA)25-tBoc (100 mg) and TFA (2.0 mL) in water (2.0 mL). Yield = 90 mg (90%).

1H NMR (D2O) (Figure S13): δ (ppm, integrated peak areas reported based on the backbone methine group) 4.36 (m, 1H per monomer, integration = 1.00), 4.13–2.98 (backbone ethylene of PEG, EDA, 18H from DTPA, and −OCH3, integration = 26.80), 2.36 (br, −CH–CH2–CH2–CO–, 2H, integration = 2.08), 2.22–1.80 (br, −CH–CH2–CH2–CO–, 2H, integration = 2.06). The disappearance of the peak at 1.44 ppm indicated that the t-Boc group was completely removed. The functionality of the DTPA group was calculated by comparing the backbone methine signals to those of the EDA associated with DTPA, where DTPA functionality = 90%.

HyNic Modification of Polymers

mPEG-PGlu(DTPA)25-HyNic

mPEG-PGlu(DTPA)25-NH2 (55 mg, 3.5 μmol) was dissolved in PB (600 μL, 200 mM, pH = 8.0), and DMSO (2.0 mL) containing HyNic-NHS (28.5 mg, 98 μmol) was added to the solution. Then, the solution was shaken overnight on a vortex mixer. Afterward, the solution was transferred to a spin filter (3 kDa MWCO Millipore Amicon) and washed with water (4000 rpm, 9 × 15 mL). Finally, the aqueous solution was freeze-dried. Yield = 50 mg (90%).

1H NMR (D2O) (Figure S15): δ (ppm, integrated peak areas reported based on the backbone methine group) 8.51 (br, =CH–N=), 8.02 (m, −CH=CH–C−), 7.40 (br, residual phenyl group from PBLG, 5H, integration = 0.01), 6.86 (br, −CH=CH–C−), 4.37 (s, 1H per monomer, integration = 1.00), 4.13–2.98 (backbone ethylene of PEG, EDA, 18H from DPTA, −OCH3, and =C(CH3)2, integration = 30.03), 2.36 (br, −CH–CH2–CH2–CO–, 2H, integration = 2.21), 2.14–1.82 (br, −CH–CH2–CH2–CO–, 2H, integration = 2.01). MnNMR = 15.8 kDa; GPC (aqueous, RI): MnGPC = 11.3 kDa, D̵ = 1.29.

mPEG-PGlu(DTPA)18-HyNic

This polymer was synthesized following a similar protocol as described above using mPEG-PGlu(DTPA)18-NH2 (20 mg, 1.7 μmol) and HyNic-NHS (6.5 mg, 22 μmol). Yield = 18 mg (90%).

1H NMR (D2O) (Figure S14): δ (ppm, integrated peak areas reported based on backbone methine group) 8.42 (br, =CH–N=), 8.05 (br, −CH=CH–C−), 7.42 (br, residual phenyl group from PBLG, 5H, integration = 0.11), 6.97 (br, −CH=CH–C−), 4.36 (s, 1H per monomer, integration = 1.00), 4.04–2.65 (backbone ethylene of PEG, EDA, DPTA, −OCH3, and =C(CH3)2, integration = 31.25), 2.35 (br, −CH–CH2–CH2–CO–, 2H, integration = 2.05), 2.18–1.74 (br, −CH–CH2–CH2–CO–, 2H, integration = 2.08). MnNMR = 11.8 kDa; GPC (aqueous, RI): MnGPC = 7.6 kDa, D̵ = 1.24.

HyNic Group Content of the Polymers

To quantify the HyNic group content of the two BCP samples, aliquots of mPEG-PGlu(DTPA)25-HyNic and mPEG-PGlu(DPTA)18-HyNic were reacted with excess 4FB-acid, and the conversion was monitored in a microcuvette cell (l = 1.0 cm) by UV/vis measurements at λ = 354 nm, with data points taken automatically at 30 s intervals until the absorbance appeared to reach a plateau. A solution of PEG-P(Glu-EDA-DPTA)18-HyNic (340 μM) or mPEG-PGlu(DTPA)25-HyNic (190 μM) was prepared by dissolving mPEG-PGlu(DTPA)18-HyNic (0.8 mg, 0.068 μmol) or mPEG-PGlu(DPTA)25-HyNic (0.6 mg, 0.038 μmol) in sodium acetate buffer (200 μL, 100 mM, pH = 5.0). The 4FB-acid solution (50 mM) was prepared by dissolving the acid (9.3 mg, 0.062 mmol) in DMSO (1.28 mL). Then, 4FB-acid solution (50 mM) was diluted to 5 mM with sodium acetate buffer (100 mM, pH = 5.0). The polymer-HyNic solution (50 μL) was then diluted 10-fold with the 4FB-acid solution (450 μL, 5 mM). The concentration of polymers containing a HyNic end group ([polymer-HyNic]) was calculated from the change in absorbance ΔA354 nm.where ε354 nm is the molar extinction coefficient of the formed bis-aryl hydrazone (BAH) bond (ε354 nm = 29 000 M–1 cm–1)[46] and l is the absorption path length of the UV–cuvette micro cell (l = 1.0 cm). The plots of ΔA354 nm as a function of time for both diblock copolymers are presented in Figure S16, where we calculated a HyNic content of 73% for mPEG-PGlu(DTPA)18-HyNic and 52% for mPEG-PGlu(DTPA)25-HyNic. The time-evolution plots were fitted to a stretched-exponential function, and the fitted parameters are summarized in Table S1.

BCP Attachment to CNC–4FB

CNC–4FB–HyNic-PGlu(DTPA)18-mPEG

A solution of mPEG-PGlu(DTPA)18-HyNic (760 μM) was prepared by dissolving PEG-PGlu(DTPA)18-HyNic (1.8 mg, 0.15 μmol) in sodium acetate buffer (200 μL, 100 mM, pH = 5.0). Then, the CNC–4FB suspension (100 μL) was diluted to 0.1 wt % with sodium acetate buffer (100 mM, pH = 5.0). The polymer solution (150 μL) and the CNC–4FB suspension (250 μL) were mixed in a quartz cell (l = 1.0 cm), and UV/vis absorbance values at RT were recorded every 30 s at 354 nm (Figure S21).

CNC–4FB–HyNic-PGlu(DTPA)25-mPEG

A solution of mPEG-PGlu(DTPA)25-HyNic (500 μM) was prepared by dissolving mPEG-PGlu(DTPA)25-HyNic (2.5 mg, 0.16 μmol) in sodium acetate buffer (320 μL, 100 mM, pH = 5.0). The CNC–4FB suspension (100 μL) was diluted to 0.1 wt % with sodium acetate buffer (100 mM, pH = 5.0). The polymer solution (150 μL) and the CNC–4FB suspension (250 μL) were mixed in a quartz cell (l = 1.0 cm), and the reaction at RT was monitored by UV/vis spectroscopy as described above (Figure S22).

Fluorescent-Labeled CNC–4FB–HyNic–Polymer

Rhodamine-Labeled CNC–4FB

NHS-rhodamine (0.4 mg, 0.8 μmol) and S-4FB (2.8 mg, 8.0 μmol) were dissolved in DMF (80 and 260 μL) to make 0.5 and 1 wt % solutions, respectively. The dye is characterized by λex,max 552 nm/λem,max 575 nm, and we denote the rhodamine-labeled CNC–4FB as Rh552–CNC–4FB. The pH of the CNC–NH2 suspension (2.4 mL, 0.83 wt %) was adjusted to 8.0 with PB (0.5 M, pH 8.0), to which NHS-rhodamine solution (70 μL, 0.5 wt %) and S-4FB solution (240 μL, 1 wt %, 10 equiv to NHS-rhodamine and 5 equiv to amino groups on CNCs) were added. Then, the mixture was shaken overnight at RT. Afterward, the product was washed with PB (0.1 M, pH 8.0) using a 4 mL MWCO 10 kDa Millipore Amicon spin filter (4000 rpm, 6 × 4 mL), followed by water (4000 rpm, 6 × 4 mL).

Rh552–CNC–4FB–HyNic-PGlu(DTPA)25-mPEG

A solution of mPEG-PGlu(DTPA)25-HyNic (330 μM) was prepared by dissolving a polymer sample (1.6 mg, 0.10 μmol) in sodium acetate buffer (300 μL, 100 mM, pH = 5.0). Then, the Rh552–CNC–4FB suspension (100 μL) was diluted to 0.1 wt % with sodium acetate buffer (100 mM, pH = 5.0). The polymer solution (300 μL) and the CNC–4FB suspension (500 μL) were mixed in a quartz cell (l = 1.0 cm), and UV/vis absorbance values at RT were recorded every 30 s at 354 nm until the reaction appeared to reach a plateau (20 h).

Alexa Fluor 488 Probe Labeled CNC–4FB

Alexa Fluor 488-NHS (0.8 mg, 1.2 μmol) and S-4FB (4.3 mg, 12 μmol) were each dissolved in DMF (140 and 360 μL) to make 0.5 and 1 wt % solutions, respectively. The pH of CNC–NH2 solution (3.6 mL, 0.83 wt %) was adjusted to 8.0 with PB (0.5 M, pH 8.0), to which the Alexa Fluor 488-NHS solution (140 μL, 0.5 wt %) and S-4FB solution (360 μL, 1 wt %, 10 equiv to Alexa Fluor 488-NHS and 5 equiv to amino groups on CNCs) were added. Then, the mixture was shaken overnight at RT. Afterward, the suspension was washed with PB (0.1 M, pH 8.0) using a 4 mL MWCO 10 kDa Millipore Amicon spin filter (4000 rpm, 6 × 4 mL), followed by water (4000 rpm, 6 × 4 mL). We denote this dye-labeled CNC–4FB as A488–CNC–4FB, noting that for this dye, .

A488–CNC–4FB–HyNic-PGlu(DTPA)25-mPEG

A solution of mPEG-PGlu(DTPA)25-HyNic (330 μM) was prepared by dissolving a polymer sample (1.8 mg, 0.11 μmol) in sodium acetate buffer (350 μL, 100 mM, pH = 5.0). Then, the A488–CNC–4FB suspension (100 μL) was diluted to 0.1 wt % with sodium acetate buffer (100 mM, pH = 5.0). The polymer solution (300 μL) and A488–CNC–4FB suspension (500 μL) were mixed in a quartz cell (l = 1.0 cm). UV/vis absorbance values at RT were recorded every 30 s at 354 nm as described above.

Parallel Syntheses of Rh552–CNC–4FB–HyNic-PGlu(DTPA)25-mPEG Samples with Different Mean Numbers of BCPs Per CNC

UV/Vis Measurements

A solution of mPEG-PGlu(DTPA)25-HyNic (330 μM) was prepared by dissolving mPEG-PGlu(DTPA)25-HyNic (11 mg, 0.70 μmol) in sodium acetate buffer (2 mL, 100 mM, pH = 5.0). Then, a Rh552–CNC–4FB suspension (600 μL) was diluted to 0.1 wt % with sodium acetate buffer (3 mL, 100 mM, pH = 5.0). The diluted Rh552–CNC–4FB suspension (0.1 wt %) was then split into seven 500 μL portions. One portion of CNC–4FB (500 μL) was mixed with an aliquot of polymer solution (300 μL) in a quartz cell (l = 1.0 cm), and UV/vis absorbance values at RT were recorded every 30 s at 354 nm until it appeared to reach a plateau. At the same time, each of the six other CNC–4FB suspensions (500 μL) was quickly mixed with the same amount of polymer solution (300 μL) in a quartz cell, and these reaction mixtures were placed outside of the UV/vis spectrometer for monitoring in the dark. The reactions were stopped at different time points ranging from 10 to 1400 min by visually tracking the UV/vis spectrum. After each desired time, one sample was quickly diluted with NaHCO3 buffer (500 mM, pH 9.0) to stop the reaction and later transferred into a dialysis tube (MWCO 50 kDa, 1 mL). The dialysis was continued for 4 days against water and further purified using a spin filter (MWCO 50 kDa, 4000 rpm, 4 × 4 mL water). The purified suspensions of Rh552–CNC–4FB–HyNic-PGlu(DTPA)25-mPEG were then ready for the metal loading experiment.

ICP-MS Measurements on Metal-Loaded CNC–Polymer Conjugates

Aliquots of each sample of Rh552–CNC–4FB–HyNic-PGlu(DTPA)25-mPEG described in the previous paragraph were treated with terbium acetate to create Tb–DTPA complexes. Each sample was then analyzed by ICP-MS to determine the 159Tb content. To prepare these terbium (159Tb)-containing CNCs, terbium acetate hydrate (0.34 mg, 10 equiv of Tb3+ per DTPA units of the polymer) was dissolved in ammonium acetate buffer (1 mL, 20 mM, pH 6.0) and added to each of the polymer samples described above. Each reaction mixture was stirred for 30 min at 37 °C. Afterward, each suspension was transferred into a 4 mL 50 kDa MWCO Millipore Amicon spin filter and washed three times with Tris buffer (Tris 25 mM, NaCl 150 mM, KCl 2 mM, pH 7.4) and then with water.

All samples were diluted with 2% HNO3 until ppb concentrations were obtained. The dilutions were repeated in triplicate and measured with an ELAN 9000 ICP-MS for 159Tb analysis. Three standard solutions were prepared by a series of 10-fold dilutions of terbium standard solution (10 mg/L, 2% HNO3, PerkinElmer) in triplicate to obtain three standard solutions with 0.1, 1, and 10 ppb of Tb. A blank solution (2% HNO3) was also measured in triplicate. A linear calibration curve was obtained, relating the counts and terbium concentration. All concentrations of terbium in ppb were obtained from the calibration curve.

Cell Culture and MCTS Experiments

Human ovarian carcinoma (HEYA8) cells were obtained from M.D. Anderson Cancer Centre (Houston, TX, USA). Cells were cultured in RPMI-1640 supplemented with 10% fetal bovine serum and 1% penicillin–streptomycin and grown at 37 °C in 5% CO2 and 90% relative humidity.

Evaluation of CNC Cytotoxicity by APH Assay

Cells were plated in 96 well plates at a density of 4000 cells/well, and treated with Rho-CNCs or AF488-CNCs over a range of drug concentrations for 24 h. The APH assay was then used to assess cytotoxicity.[57] In brief, following treatment, cells were washed with PBS, and then, 100 μL of freshly prepared reaction buffer (2 mg/mL p-nitrophenyl phosphate in 0.1% Triton X-100/0.1 M sodium acetate buffer (pH 5.5)) was added to each well. Following 2 h of incubation at 37 °C, 10 μL of 1 M sodium hydroxide was added to each well, and the cell viability was determined by measuring the UV absorbance at 405 nm using an automated 96 well plate reader (SpectraMax Plus 384, Molecular Devices, Sunnyvale, CA, USA). Cell viability was determined by normalizing the data to controls: cell viability (%) = Abstreatment – Absmedia/(Abscontrol – Absmedia).

MCTSs

MCTSs of HEYA8 cells were grown as previously reported by our laboratory.[57,61] Briefly, subconfluent cells were trypsinized and seeded at a density of 2000 cells/well onto nonadherent 96-well round-bottom Sumilon PrimeSurface spheroid plates (MS-9096U; Sumitomo Bakelite, Tokyo, Japan). Cells were incubated in complete growth media at 37 °C in 5% CO2 and 90% relative humidity for 4 days.

Penetration of CNCs into MCTSs

The penetration of CNCs labeled with rhodamine (Rh552–CNC–4FB–HyNic-PGlu(DTPA)25-mPEG) and Alexa Fluor 488 (A488–CNC–4FB–HyNic-PGlu(DTPA)25-mPEG) was evaluated by following a previously reported method with modifications.[61] Briefly, MCTSs were grown for 4 days until they reached a diameter of approximately 500 μm. The media (50 μL) was removed from each well and replaced with 50 μL of media solution containing 0.0034 wt % rhodamine-CNCs or AF488-CNCs. Following 24 h of incubation with CNCs, MCTSs were carefully washed with PBS and then transferred onto Nalge Nunc Lab-Tek II #1.5 German Coverglass microscopy slides for imaging. MCTSs were imaged using a Zeiss LSM700 confocal microscope with a 20× objective and Rhodamine or Alexa Fluor 488 filter for Rho-CNCs and AF488-CNCs, respectively. Fluorescence laser gain was determined to minimize bleed-through artifacts and was kept constant for all measurements. CNC penetration was assessed at three different depths by obtaining optical slices at 30, 60, and 90 μm from the MCTS surface. For each optical slice, the fluorescence intensity per unit area was determined for three equally spaced, concentric regions (i.e., periphery, intermediate, and core) using a custom MATLAB algorithm (MathWorks Inc., MA, USA). The fluorescence intensity for each region was expressed as mean ± SD for n = 4 MCTSs.

Statistical Analysis

Statistical analyses were performed using GraphPad Prism (GraphPad Software Inc., San Diego, CA, USA). The statistical significance was determined using one-way analysis of variance (ANOVA) with subsequent multiple comparisons using the Bonferroni correction for three or more groups. p < 0.05 is considered to be statistically significant.