Keywords

7.1 Introduction

Identification of tumor-associated surface markers plays a fundamental role in the design strategy for tumor-targeted nanoparticles (NPs) [1]. NP conjugation with a ligand molecule of high specificity for the particular biomarker of interest constitutes the basis for the mechanism of active tumor targeting [2, 3]. Optimal ligand conjugation design confers these NPs with a greater ability to facilitate tumor-specific NP binding and payload uptake than passive targeting mechanisms which rely solely on the enhanced permeation and retention (EPR) effect in which NPs accumulate in the tumor through the enhanced leakiness of tumor vasculature [4].

Several classes of tumor biomarkers have already been identified and used in the development of targeted NPs [1,2,3, 5]. These include: (1) receptors for vitamin uptake such as the folate receptor α, β (FAR α , FAR β ) [6, 7], biotin receptor [8, 9]; (2) an integrin family of receptors such as α v β 3 [10]; (3) prostate-specific membrane antigen (PSMA) receptor [11, 12]; (4) growth factor receptors including HER2 [13], epidermal growth factor receptor (EGFR) [14], fibroblast growth factor receptor (FGFR) [15]; insulin receptors [16]; and (5) the transferrin receptor [17]. Each of these tumor biomarkers is overexpressed in one or more types of tumors and engages in receptor-mediated endocytosis [18], which serves as the specific route for the internalization of targeted NPs.

Riboflavin receptors (RFRs) belong to the class of vitamin uptake receptors which show promising potential for tumor-targeted applications [19, 20]. In an earlier study, Low et al. [21] investigated the cellular uptake mechanism of riboflavin (RF)-conjugated bovine serum albumin (BSA) in several human tumor cell lines. Uptake of this conjugate occurred at a rate greater than that of unmodified BSA, and the process was RFR-dependent and specific. Such facilitated protein uptake was attributed to RFR-mediated endocytosis, and highlighted the potential of using a RFR-targeted strategy for enhancing specific delivery. A research group led by Swaan, P.W. later also demonstrated the receptor-mediated uptake of a RF-rhodamine dye conjugated form [22, 23] in human cell lines.

Recently, we [24,25,26,27,28] and others [29,30,31,32] have started the development of RFR-targeted delivery platforms, and have conducted several proofs of concept studies in vitro and in vivo for their validation in tumor-targeted delivery. In this review article, we aim to describe our approaches in the design of RF-conjugated NPs, and provide several lines of evidence supportive of their significance and potential as a novel platform for tumor-targeted delivery. The purpose of this chapter is primarily to address the current lack of reviews and perspectives focused on RFR-targeted applications. Other established tumor biomarkers such as FARs [33, 34], integrin α v β 3 [10], PSMA receptor [35], HER2 [13, 36], and EGFR [14, 37] are already extensively reviewed elsewhere and thus are introduced only minimally here. We believe that this review provides a timely coverage of the various aspects important to RFR-targeted drug and gene delivery, and will serve as an invaluable resource in the design of RFR-targeted nanoplatforms.

7.2 Riboflavin Receptors and Ligands

In cellular metabolism, RF (vitamin B2) is required in the biosynthesis of flavin-based redox cofactors including flavin mononucleotide (FMN) and flavin adenine dinucleotide (FAD). However, its hydrophilicity (logP = −1.46; P = partition coefficient = [RF]octanol/[RF]water) makes it unable to passively diffuse across hydrophobic cell membranes [38]. Thus, its cellular availability depends on the uptake mechanism mediated by its receptors.

7.2.1 RFRs

A group of multiple proteins is involved in the cellular uptake of RF, consisting of RF carrier, RF transporter, and RF binding protein (RFBP). In this article, these proteins are collectively referred to as riboflavin receptors (RFRs) (Table 7.1). They are expressed as both soluble and membrane-bound isoforms [23, 39,40,41], and are involved primarily in cellular trafficking and uptake of RF. RFRs display high RF affinity as illustrated by RFBP (K D ≈ 1 nM) [41]. Recent studies suggest that RFRs are overexpressed in certain malignant cells including human breast and prostate tumors, which implicate RFRs as a class of tumor biomarkers [19, 20, 32]. Thus, these cells displayed the unique ability to take up riboflavin or its macromolecular conjugates, which is indicative of their expression of RFRs [20, 21, 26, 32, 42]. These cells include KB carcinoma [21, 26, 42], LnCap (prostate cancer) [20, 32], SK-LU-1 and A549 (lung cancer) [21], and SK-OV (ovarian cancer) [21]. In addition, a class of ATP-dependent RF transporters is involved in the subcellular accumulation of RF in certain cancer stem cells which are resistant to anticancer chemotherapeutic agents, suggesting their potential as a biomarker for these cells [43].

Table 7.1 Properties of riboflavin receptors (RFRs)
Full size table

RFRs share several structural and functional similarities with FARs. Both RFRs and FARs belong to the family of folate binding proteins which are glycosylphosphatidylinositol-anchored surface receptors [38]. The receptors exhibit a high degree of homology in their amino acid sequences [44] and have similar secondary structure [45] in their ligand binding domains. RFRs, like FARs are taken up along with their bound ligand by the cell through endocytosis [21, 23, 46], the mechanism responsible for the uptake of NPs following their cell surface binding to the receptor.

Despite such similarities between these two important vitamin uptake receptors, RFRs play a distinct role in RF transport and cellular uptake and exist as more diverse forms such as soluble carriers, transporters, and membrane-bound proteins [23, 39, 40, 45]. As summarized in Table 7.1, RFRs also show clear differences in their tissue distribution and the types of malignant cells they are associated with including cancer stem cells [43]. Such distinct features of RFRs suggest an important opportunity for targeting specific malignant cells which are otherwise not addressable by use of other tumor biomarkers. In addition to their ligand role, certain types of RF analogs display potent cytotoxic activity due to their ability to competitively interfere with the cellular functions of flavin cofactors [48], and offer additional benefits in the therapeutic applications of RF ligand-conjugated nanoplatforms [25].

7.2.2 Riboflavin Ligands

The primary targeting ligand for RFR-targeted platform design is RF which is the endogenous ligand for these receptors. It is made up of two structural units—an isoalloxazine and a (D)-ribose, each modifiable for ligand conjugation (Fig. 7.1). In addition, there are a number of structural homologues to RF which are referred to as RF antagonists or antimetabolites. These include roseoflavin, cofactor F420 [47], and 2(4)-imino-4(2)-amino-2,4-dideoxyriboflavin [25, 47,48,49]. Each of these retains the ability to bind RFRs, but lacks the requisite functional activity required for the biosynthesis of RF-associated cofactors.

Fig. 7.1
figure 1

Riboflavin and selected riboflavin antagonists which are either naturally occurring (roseoflavin, cofactor F420, lumiflavin) or synthetic (quinacrine, chlorpromazine)

Full size image

To be considered ideal for the design of RFR-targeted platforms, the ligand should provide certain sites amenable for linker installation, allow easy synthetic modification, and lack any functional activity for stimulating cell growth. In an effort to identify such ligands, we searched RF-mimicking small molecules in the SciFinder® database, and identified a set of candidate molecules that include perphenazine, chlorpromazine, quinacrine, and chloroquine (Fig. 7.1) [25]. Each of these is structurally less complex than RF but contains a flat tricyclic heterocycle that mimics the isoalloxazine head of RF.

We investigated the binding interaction of these small molecules with chicken RFBP using isothermal titration calorimetry (ITC), and determined their dissociation constants (K D) as summarized in Table 7.2. In general, these molecules bound with lower affinity than RF in the order of RF (5.0 nM) > lumiflavin (61 nM) > quinacrine (264 nM) > chloroquine (2100 nM). Two other tested compounds, perphenazine and chlorpromazine, lacked detectable binding affinity despite their structural similarities based on the tricyclic heterocycle. Both lumiflavin and quinacrine do not cause undesired positive trophic effects as those associated with the function of RF which has been shown to stimulate tumor cell growth. Despite their lower affinity, it is anticipated that NPs conjugated with this lower affinity ligand will still have the ability to bind RFRs on the cell surface with high avidity constants via multivalent binding interactions [50,51,52]. Use of these RF-mimicking molecules in the design of RFR-targeted platforms constitutes a subject of follow-up studies.

Table 7.2 Binding affinity and thermodynamic parameters of RF antagonists to riboflavin binding protein (RFBP) in PBS buffer, pH 7.4
Full size table

7.3 Application of RF-Conjugated Dendrimers

7.3.1 Linker Design

An X-ray crystal structure was determined for RF in complex with chicken RFBP at a resolution of 2.5 Å [45]. This serves as the basis for rational linker design by providing several insights on the position and orientation of the linker needed in the design of a RF-linker construct. First, the xylene domain of its isoalloxazine unit is stacked between aromatic planes in the ligand binding cleft and is not ideal for linker attachment. However, the opposite face (N-3 position) of the same isoalloxazine head is relatively open for linker modifications. This is illustrated by an earlier study, in which 3-carboxymethylriboflavin (1, Fig. 7.2) [25, 53, 54], a RF derivative with a carboxylic acid extended out from the N-3 position, retained its affinity for RFBP which allowed its use in the detection of RFBP in milk products [53, 54]. Synthesis of this riboflavin linker construct at the N-3 position (1) as described in literature [25, 53, 54] is conveniently achieved in three consecutive steps that comprise of the exhaustive acetylation of (−)-riboflavin, the N-alkylation of 2′,3′,4′,5′- tetra-O-acetylriboflavin to the ethoxycarbonyl methyl derivative, and complete removal of ester protecting groups by acidic hydrolysis (Fig. 7.2).

Fig. 7.2
figure 2

Synthesis of riboflavin linker constructs. Each linker (in blue) is installed at the N-3 (1, 2) or N-10 position through the terminal hydroxyl group of (D)-ribose (3)

Full size image

In order to further validate 3-carboxymethylriboflavin as the linker construct in RFR-targeted delivery platforms, we investigated its binding affinity to RFBP by surface plasmon resonance (SPR) spectroscopy using a biosensor chip prepared by immobilization with 2 which contains a spacer (3-aminopropyl) at the carboxylic acid terminus of 3-carboxymethylriboflavin. This amine-terminated riboflavin derivative 2 was prepared by the EDC-mediated amide conjugation of 3-carboxymethylriboflavin 1 with a mono N-Boc protected propanediamine (Fig. 7.2). RFBP bound to the surface in a dose-dependent manner (Fig. 7.3). Its binding was also ligand-specific, as the RFBP adsorption was competitively blocked by co-injecting RF, quinacrine and 2. This SPR study validated the compatibility of the linker installation made at the N-3 position of RF.

Fig. 7.3
figure 3

Surface plasmon resonance (SPR) spectroscopy using a riboflavin (2)-immobilized CM5 sensor chip. a Binding of RFBP in PBS (pH 7.4). Inset a Scatchard plot; b Competitive binding experiments of riboflavin binding protein (RFBP) with 2. aInjection of 2 alone (220 μM) without RFBP; c, d Plot of fractional inhibition (F = 1 − (RU [I]/RU [I]=0)) as a function of the ligand (RF, 2) or competitive inhibitor (quinacrine). Adapted with permission from [25]. Copyright © 2011, American Chemical Society

Full size image

Second, the (D)-ribose unit is largely exposed to the aqueous medium and makes a minimal contribution to receptor binding. Thus, its external accessibility in combination with its flexible configuration makes the sugar unit suitable for linker installation as illustrated by a glutarate linker attached at the terminus of the (D)-ribose through an ester linkage (3, Fig. 7.2). This riboflavin derivative 3 contains a glutarate moiety attached through an ester linkage at its hydroxyl group of the (D)-ribose unit. It was prepared by heating a mixture of riboflavin and glutaric anhydride in a mixture of pyridine and DMSO. This coupling reaction might occur regioselectively at the primary hydroxyl group as suggested by other similar conjugation reactions of riboflavin reported elsewhere [21, 46], possibly because the primary terminal position is sterically less hindered than those secondary alcohols located adjacent to the bulky isoalloxazine head.

7.3.2 Dendrimer Conjugates Designed for RFR-Targeted Drug Delivery

We developed RF-targeted delivery platforms with a generation 5 (G5) poly(amidoamine) (PAMAM) dendrimer (diameter 5.4 nm) [55]. This PAMAM dendrimer has a globular shape with a large number of dendritic branches (theoretically 128 for G5), each terminated with a primary amine which is amenable to conjugation with a targeting ligand or a drug molecule. The use of this G5 dendritic polymer provides several key benefits for biomedical applications [56, 57] as it displays biocompatibility, is characterized by favorable pharmacokinetic properties such as extended duration of circulation, and lacks immunogenicity [58,59,60,61].

Two types of conjugates were designed that include G5(RF)6.3(FITC)1.3 5 and G5(RF)2.5(MTX)3.9 6 (Fig. 7.4). First, G5(RF)6.3(FITC)1.3 is a fluorescently labeled conjugate that has a mean of 6.3 RF molecules and 1.3 fluorescein isothiocyanate (FITC) molecules attached on the dendrimer surface prepared for confocal microscopic imaging of its cellular uptake. Here, the ligand attachment was made conveniently through an ester bond formed between a primary hydroxyl group on the (D)-ribose unit of RF and a glutaric acid spacer presented on the dendrimer surface. Second, G5(RF)2.5(MTX)3.9 6 is a drug conjugate that carries covalently attached methotrexate (MTX) as the payload. MTX potently inhibits dihydrofolate reductase in the cytoplasm with a K i value of 4.8 pM [62], leading to strong inhibition of cell growth.

Fig. 7.4
figure 4

Structure of a fifth generation (G5) poly(amidoamine) dendrimer G5(NH2) (4), and synthesis of two representative riboflavin (RF)-conjugated dendrimers, G5(RF) n=6.3(FITC) p=1.3 (5) and G5(RF) n=2.5(MTX) m=3.9 (6). n, m and p: each refers to a mean number of RF, methotrexate (MTX) or fluorescein isocyanate (FITC) covalently attached to the dendrimer polymer, respectively

Full size image

Assessment of the cellular binding and uptake of conjugate 5 was performed in RFR(+) KB cells. These cells belong to a subline of cervical tumor cells that showed receptor-mediated uptake of RF and RF-dye conjugates [21, 46]. Our flow cytometry analysis showed dose- and incubation time-dependent binding of the conjugate (Fig. 7.5) [26]. When incubated with other human cancer cell lines, this RF conjugate also showed significant fluorescence intensity in these cell lines including IGROV-1 (ovarian) and SCC15 (head and neck) like in KB cells. This is supportive of conjugate binding and uptake by these tumor cells, some of which have been validated earlier for their overexpression of RFR on the cell surface [21].

Fig. 7.5
figure 5

a Dose-dependent binding and uptake of 5 G5(RF)6.3(FI)1.3 in KB cells (incubation time = 1, 4 h); b Dose-dependent cytotoxicity of 6 G5(RF)2.5(MTX)3.9 and free MTX in KB cells; c Effect of a competitive ligand (RF) on cell growth. PBS (control), 6 (30 nM), RF (30 μM), and 6 (30 nM) + RF (30 μM); d Effect of a RF-conjugated dendrimer on cell growth. PBS, 6 (30 nM), G5(RF)2.5 (30 μM), and G5(RF)2.5 (30 μM) + 6 (30 nM). Asterisk the p value for each of the data set is <0.005 compared to PBS. Reprinted with permission from [26]. Copyright © 2010, Elsevier

Full size image

In a subsequent study, we investigated the effectiveness of this RFR-targeted conjugate for drug delivery using G5(RF)2.5(MTX)3.9 6. The cytotoxic effect of 6 was determined in KB cells in vitro using an XTT assay (Fig. 7.5b). This conjugate showed potent inhibition of tumor cell growth at low nM doses, and its inhibition activity was incubation time- and dose-dependent. The IC50 value estimated from the dose response curve (at 4 h incubation) for the conjugate was 72 nM, which indicates slightly lower activity than that of free MTX of 48 nM.

In order to verify the mechanism of RF-mediated delivery by 6, we performed a series of ligand competition experiments in RFR(+) KB cells with RF (30 μM, Fig. 7.5c) or G5(RF)2.5 (30 μM, Fig. 7.5d). Here, the addition of RF or its dendrimer conjugate G5(RF)2.5 (which has no MTX attached) alone showed no effect on cell growth. When G5(RF)2.5(MTX)3.9 6 (30 nM) was co-incubated with an excess amount (30 μM) of G5(RF)2.5 (a multivalent ligand competing for RFR), the cytotoxicity of 6 could be effectively blocked due to perhaps competitive occupation of RFRs by the added G5(RF)2.5 which would contribute to the decrease of its intracellular uptake. As a result, the cell growth was restored to 90% from ~50% observed in the absence of G5(RF)2.5.

Co-incubation with free RF failed to show such restoring effects on cell growth. This distinct difference between RF and G5(RF)2.5 is attributable to the high avidity binding of the multivalent conjugate compared to the monovalent RF ligand [51, 52, 63]. Our results of ligand competition experiments are in agreement with a previous uptake study performed with 125I-labeled, multivalent RF-conjugated bovine serum albumin (shortly, 125I-BSA(RF)5) in RFR(+) KB cells reported by Low et al. [21]. Thus, co-incubation of 125I-BSA(RF)5 with free RF (at 10–40 mol excess) resulted in almost no change in its cellular uptake relative to no RF addition, while co-incubation with an unlabeled BSA(RF)5 (at 10 mol excess) led to significant blocking (~70%) of its uptake, evidence supportive of its multivalent tighter binding than monovalent RF.

All of these studies are supportive of the cellular uptake of 6 through a RFR-mediated mechanism and verify the activity of delivered MTX in the cytoplasm in inducing potent cytotoxicity. These studies also point to the potential application of RF-conjugated dendrimers in targeted delivery of a fluorescent imaging molecule and an anticancer therapeutic agent to malignant tumor cells overexpressing the RF receptor.

7.3.3 Dendrimer Conjugates Designed for RFR-Targeted Gene Delivery

We investigated the potential of applying RF-conjugated dendrimers as a new class of nonviral vectors for RFR-targeted gene delivery in tumor cells (Fig. 7.6) [42]. For this approach, we modified the conjugate G5(RF)4.9 by co-attachment of multiple molecules of 3,8-diamino-6-phenylphenanthridinium (DAPP) which has the ability to intercalate into DNA, thus forming polyplexes with dsDNA. The resulting dendrimer G5(RF)4.9(DAPP)6.9 has dual functional motifs, one for targeting RFRs on the cell surface and the other for anchoring a DNA payload.

Fig. 7.6
figure 6

A schematic illustrating the concept for dendrimer vectors with DNA intercalation motifs for RFR-targeted gene delivery. a–c Transfection of luciferase plasmid DNA (pLuc) in FAR(+) and RFR(+) KB cancer cells via dendrimer polyplexes, each made with 7 G5(FA)8.6, 8 G5(DAPP)5.4, 9 G5(FA)8.6(DAPP)5.4, 10 G5(RF)4.9, and 11 G5(RF)4.9(DAPP)6.9. a, b Box and whisker plots showing the distribution of luminescence at dendrimer-to-plasmid (D/P) ratios of 1:1 (a) and 2:1 (b) at 1 μg of pLuc. Each p value was calculated against plasmid DNA alone (Asterisk). c Luciferase transfection at higher D/P ratio of 4:1 or 8:1. DAPP = 3,8-di-amino-6-phenylphenanthridine. RLU relative light unit. Error bars standard deviation (±SD). Adapted with permission from [42]. Copyright © 2011, American Chemical Society

Full size image

We selected a series of RF-conjugated dendrimers along with other targeted dendrimers that include 7 G5(FA)8.6, 8 G5(DAPP)5.4, 9 G5(FA)8.6(DAPP)5.4, 10 G5(RF)4.9, and 11 G5(RF)4.9(DAPP)6.9. Each was used for preparing a series of polyplexes by complexation with plasmid DNA (pDNA) encoding a luciferase (pLuc) as a reporter gene. Several polyplexes containing DNA at various dendrimer-to-pDNA ratios (D/P) were made, and it was investigated whether such dendrimer polyplexes are effective for gene transfection by performing transfection experiments in FAR(+) and RFR(+) KB carcinoma cells in vitro.

As presented in Fig. 7.6a, b, each of the polyplexes prepared at the D/P ratios of 1:1 or 2:1 showed transfection activities greater than the pristine plasmid used as control. The transfection efficiency varied with dendrimer type such that polyplexes prepared with 10 G5(RF)4.9 or 11 G5(RF)4.9(DAPP)6 gave the highest transfection efficiency at each ratio with statistical significance of p < 0.01 (F test). These were more effective than those polyplexes prepared with FA-conjugated dendrimers. It is notable that 10 G5(RF)4.9 showed such high efficiency despite its lack of DAPP which was presumably needed for DNA anchoring. We postulate that RF alone attached to the dendrimer could play a dual functional role as both a targeting ligand and DNA anchor due to its previously demonstrated ability to intercalate its flat isoalloxazine head between two adjacent DNA base pairs in dsDNA [64, 65]. The efficiency of gene transfection also varied with the D/P ratios while a single best ratio applicable for all polyplexes was not observed. This variation is attributable to the mechanism of nonviral gene delivery [66] in which the ratio as well as nanoparticle type determine the shapes and charge properties of the polyplexes, each playing a critical role in the course of intracellular uptake, DNA release, and nuclear transport. This observation is indicative of the challenges in predicting the optimal structure and function of the polyplex which requires further studies in the future.

The cellular uptake of these polyplexes can occur through either FAR or RFR-mediated mechanism. By confocal microscopy, we imaged KB cells treated under the same conditions as in the transfection experiment. As shown in Fig. 7.7, cells treated with each polyplex showed significant fluorescence which is indicative of their cellular uptake. Most of the fluorescence intensity was predominantly localized in the cytoplasmic area rather than on the cell surface and with only minor fluorescence observed in the nuclei. This is supportive of their intracellular uptake possibly via a receptor-mediated mechanism, as well as the release of pDNA in the cytoplasm rather than nuclear uptake of the polyplex complex. As a comparison, 8 G5(DAPP) which lacks RF was taken up, but less effectively than 11 G5(RF)(DAPP), suggesting the possibility of other mechanisms of uptake such as macropinocytosis and phagocytosis which is dependent on nanoparticle shapes [67].

Fig. 7.7
figure 7

a–e Fluorescence confocal microscopy of various polyplexes taken up by KB cells in vitro. Each sample was prepared using polyplexes made of luciferase plasmid (1 μg/mL) in complex with each of dendrimer conjugates 7 G5(FA)8.6, 8 G5(DAPP)5.4, 9 G5(FA)8.6(DAPP)5.4, 10 G5(RF)4.9, or 11 G5(RF)4.9(DAPP)6.9 at a 2:1 ratio (w/w). In this imaging study, the localization of the dendrimer polyplex was detected by fluorescent emission that is attributed to excitation of dendrimer-attached ligands including folate (a), riboflavin (d), and/or DAPP (3,8-di-amino-6-phenylphenanthridine; b, c, e). Reprinted with permission from [42]. Copyright © 2011, American Chemical Society

Full size image

In summary, we investigated a new concept for targeted gene delivery using RF-conjugated multifunctional dendrimers. This novel platform was highly effective for facilitating gene transfection in specific for RFR(+) mammalian cells.

7.3.4 RFR-Targeted Imaging Methods

As shown above, imaging of RF-conjugated dendrimers in the cell can be performed by confocal microscopy by focusing on fluorescent dye molecules associated with the NP. However, these dye molecules are photounstable and rapidly bleach, leading to reduced resolution and detection capabilities. We developed another imaging modality for investigating the cellular association of RF-conjugated dendrimers [27]. It is based on a gold nanoparticle (AuNP) system which displays unique optophysical properties such as surface plasmon resonance (SPR) absorption, visible luminescence, and SPR scattering effects [68, 69]. With these modalities, the detection of AuNP does not require conjugation with any additional fluorescent labels, and it is conveniently performed under dark field light [70,71,72] and confocal microscopy [73].

The AuNPs used in this study [27] were spherical in shape and displayed a maximal absorption (λ max) band at 520 nm. The size distribution of AuNPs was measured by atomic force microscopy (AFM), indicating a mean diameter of 13.5 ± 2.2 nm. Their hydrodynamic diameter as determined by dynamic light scattering (DLS) was as large as ~30 nm (Z ave) which points to the contribution of the hydrated diffusion layer surrounding the AuNP core to the measured size. The AuNP was modified to form the core–shell nanocomposite AuNP@dendrimer by its surface modification with dendrimer conjugates 12, 13 G5(RF) n (n = 0, 4) (Fig. 7.8). The surface modification was achieved through Au–S chemisorption between the surface Au and the cyclic disulfide moiety at the terminus of a lipoic amide branch of the dendrimer.

Fig. 7.8
figure 8

SPR scattering imaging for the cellular uptake of dendrimer-coated gold nanoparticles AuNP@G5(RF) n in KB cells. a–c KB cells treated with 50 nM of unmodified AuNP for 2 h were imaged by a SPR scattering or b luminescence. c Co-localization of the signals was confirmed by the overlay. SPR scattering imaging was taken for KB cells treated for 4 h with 80 nM of d AuNP, e AuNP@G5(RF) n=0 or f AuNP@G5(RF) n=5, (inset is a magnified view of the adjacent cells). Scale bar 30.3 μm (inset, 28.8 μm). Signal from the SPR scattering was overlaid with differential interference contrast (DIC) images. Reprinted with permission from [27]. Copyright © 2014, American Chemical Society

Full size image

We first validated two methods for detecting AuNPs which did not contain any fluorescent dyes or chemisorbed dendrimer. As shown in Fig. 7.8a–c, RFR(+) KB cells were incubated with citrate-stabilized AuNPs (50 nM) for 2 h, and were imaged by confocal microscopy via two detection modes including SPR scattering (λ ex 514 nm; λ em 474–506 and 522–570 nm) and luminescence (λ ex at 514 nm; λ em at 550–650 nm). AuNPs associated with the cells were clearly detectable under both detection modes, SPR (Fig. 7.8a) and luminescence (Fig. 7.8b); however, their detection intensities were rather weak due to their low level of cellular uptake.

For enhanced uptake, KB cells were treated at a slightly higher concentration (80 nM) and incubated for a longer period (4 h) with AuNP and AuNP@G5(RF) n (n = 0, 5). Images of the treated cells were acquired in the SPR scattering mode, showing qualitatively clear differences between untargeted NPs (AuNP, AuNP@G5(RF)0) and targeted AuNP@G5(RF)5. The targeted AuNP@G5(RF)5 showed more punctate and localized areas of signal, and their intensity was distinctly greater than those observed otherwise with the bare unmodified AuNP as well as with the AuNP@G5(RF)0) treated cells (Fig. 7.8d–f). We attribute this scattering detection to large aggregates of AuNPs rather than individual AuNPs either bound on the cell surface or internalized.

In summary, our confocal microscopy studies demonstrated the utility of the dual detection modes of SPR scattering and luminescence for the determination of the cellular localization of dendrimer-chemisorbed AuNPs in tumor cells. As covalent modification with fluorophores for detection is sometimes not desirable due to the possible alteration of native activity, these nonfluorescent-based methods of detection may offer a better alternative for imaging applications.

7.4 Biophysical Basis of Multivalent High Avidity

The design principle of targeted NPs involves a multivalent ligand system in which each NP is conjugated with multiple targeting ligands. Thus, the multivalent NP recognizes and binds to a target cell with high specificity and strong binding affinity, which together are referred to as avidity [51, 52, 63, 74]. Unlike affinity which often refers to the strength of monovalent interaction between a single receptor and ligand pair, avidity is a collective property that measures the strength of simultaneous interactions between multiple receptor–ligand pairs [51, 56, 57] (Fig. 7.9).

Fig. 7.9
figure 9

a, b Structures of two RF-conjugated dendrimer series G5(RF) n . In each, the RF ligand is tethered to the G5 PAMAM dendrimer through a linker located at either its N-10 or N-3 position with variable valency (n). c Representative raw ITC data for the interaction between 17 G5(RF)4.5 with chicken RFBP (4 μM) at 25 °C in PBS buffer. d Plot of integrated area under each injection peak for 17. The solid line is an independent model fit to data with parameters n, K D, and ∆H. Inset a model for monovalent receptor–ligand association. Reprinted with permission from [28]. Copyright © 2012, American Chemical Society

Full size image

Multivalent design factors have been extensively investigated by many laboratories including ours by conjugation of small molecule ligands such as carbohydrates [51, 63, 75, 76], folate [77, 78], methotrexate [79,80,81,82,83], vancomycin [84,85,86,87], and oligonucleotide [88] to NP scaffolds based on polymers [75, 89, 90], dendrimers [77, 79, 84, 88], and inorganic nanomaterials [78, 86, 87, 91]. These studies suggest that several factors play a significant role in conferring high avidity and selectivity. These include: (1) use of threshold ligand valence [77, 84, 88]; (2) presentation of two different ligands for co-targeting two distinct receptors on the same cell surface [78, 91]; (3) evaluation of NP sizes and shapes [92] for optimized conformal interactions [75, 93, 94]. It is also notable that an overcrowding or steric effect can occur [89, 95] when too many or bulky ligands are presented on the same surface of a NP which thus can interfere with high avidity binding. This steric interference is especially an issue with large molecule ligands such as antibodies [96]. Here, we investigated the biophysical basis for the specific and high avidity adsorption of RF-conjugated dendrimers to receptor expressing cell surfaces.

7.4.1 Ligand Affinity

We first investigated the thermodynamic basis of monovalent interactions between a receptor and a RF-conjugated dendrimer in solution [51]. One of the key design factors of this dendrimer conjugate is the RF valency which is known to play a significant role in controlling the avidity of the multivalent system. However, little is known about its role in monovalent affinity. Two series of RF-conjugated dendrimers G5(RF) n 14–19 were prepared by the covalent attachment of RF at either its N-10 or N-3 position to the dendrimer (Fig. 7.9). The linker used for RF attachment in each series is composed of a three to five atom spacer. We chose this linker length as our previous cell binding [26] and SPR studies [25] suggested that it is long and flexible enough for the dendrimer-attached RF to be able to dock in the ligand site of RFBP. In each series, the average number (n = valency) of RF attached was varied in order to determine whether the ligand valency plays any role in the monovalent affinity (K D) of the tethered RF ligand in solution.

Several binding parameters for the interaction between RF-conjugated dendrimers and soluble RFBP were determined by ITC in solution as summarized in Table 7.3 [28]. First, free RF binds to RFBP with a binding stoichiometry (n b) of 1:1 (RFBP/ligand) [25]. Each of 14–16 G5(RF) n in the N-10 series showed an n b of ~3–5:1 (RFBP/dendrimer), indicative of partial occupation of all the RF sites on the dendrimer by RFBP. This stoichiometry is in close agreement with the analysis based on a simple sphere model in which approximately six to seven RFBP (d = 4–5 nm) protein molecules can be theoretically accommodated around the surface of a G5 dendrimer nanoparticle (d = 5.4 nm) [55]. In contrast, 17–19 G5(RF) n in which each RF was tethered through a short spacer (3 atom) at its N-3 position showed only 1–2 RFBP occupation per dendrimer. These results suggest that ligand conjugation at the (D) ribose terminus via a longer glutarate linker provides more space and flexibility for more optimal protein accommodation.

Table 7.3 Thermodynamic parameters for the monovalent binding of RF-conjugated dendrimers with RFBP in solution (phosphate buffered saline, pH 7.4) at 25 °C
Full size table

Overall, this ITC analysis suggests a number of new insights in multivalent ligand design. First, we observed an upper limit in the maximal number of ligands that could engage in receptor binding. Control of this limit is dependent on design factors such as linkage position, spacer length, and ligand valency. However, it is notable that the RFBP used here is a monovalent system in solution, and thus different from RFRs presented on the cell surface. Accordingly, the remaining unoccupied ligands on the dendrimer are still available for making opportunistic receptor binding interactions to these RFRs, given their proximity [28].

Second, the dissociation constant (K D) values determined for all dendrimer conjugates were greater by a factor of 93–1110 relative to the K D value of RF (5 nM) [25]. This suggests that the mean affinity of each RF ligand to RFBP is significantly decreased once it is conjugated to the dendrimer surface. The K D values determined for the N-3 linkage series also showed reduced affinity to RFBP, but overall higher affinity by a factor of ~5 than the N-10 series at a similar RF valency. These results clearly signify that contacts made in the binding pocket by the RF ligand with an unmodified (D) ribose moiety are important, and thus contribute to tighter binding in the binding pocket.

Third, the decrease in the affinity of the RF-dendrimer conjugates to RFBP is better understood by examination of thermodynamic parameters based on enthalpic (∆H) and entropic (−T∆S) contributions. Enthalpically, the binding of the N-3 conjugates was much more favorable than the N-10 series. In contrast, the entropic penalty (−T∆S) was more severe for the N-3 conjugates. This implies that the binding of the N-3 conjugates is largely enthalpy driven, and the conjugation of more RF ligands per dendrimer likely results in the higher entropic penalty perhaps due to steric repulsion or congestion [97].

7.4.2 Multivalent Avidity

Our thermodynamic studies above show that the monovalent interaction between the RF-dendrimer conjugate and the RFBP in solution is not enhanced, but is in fact weaker than that of free RF. We then investigated the avidity of the multivalent binding interactions between a RF-conjugated dendrimer and multiple RFBP molecules presented on a surface (a model system of the cell surface). We employed AFM because of its proven ability to quantify multivalent effects in biomolecules and synthetic model systems [98]. In particular, AFM-based dynamic force spectroscopy allows for the precise measurement of the physical forces involved in biomolecular interactions [98].

We hypothesized that the multivalent avidity, which results from multiple, cooperative interactions, should result in a force which is greater to break than the monovalent affinity [51, 52, 63, 99]. For this AFM study, a model system for the cell surface was generated by immobilization of RFBP onto a substrate (an ultra-flat gold surface) as illustrated in Fig. 7.10. Force experiments were performed through the contact approach of an AFM tip coated with G5(RF) n (n = 0, 3, 5) to the RFBP-immobilized substrate followed by retraction to measure the rupture force arising from the recognition interaction. Rupture events were observed over the course of the tip retraction, and the binding specificity was confirmed by addition of a competitive ligand (free RF) (not shown) which led to the block of the rupture events.

Fig. 7.10
figure 10

a An AFM gold (Au) probe tip prepared by surface coating with G5(RF) n (n = 0, 3, 5; linkage at N-3 position), and schematic for dynamic force spectroscopy; b Representative force–distance curves between a G5(RF)5 (13)-coated tip and RFBP covalently attached to an ultra-flat gold surface. Loading rate = 5.7 nN/s. Offset force curves depict rupture events in the a 10–20 pN, b 20–50 pN, and c 50–120 pN ranges; c Dynamic force spectra of G5(RF) n (n = 0, 3, 5) versus loading rate. Square data points represent rupture of unbinding event (a). Circle and triangle points represent unbinding events (b, c), respectively. Adapted with permission from [24]. Copyright © 2014, American Chemical Society

Full size image

For data analysis, rupture forces were extracted by force–distance curves as illustrated for G5(RF)5 in Fig. 7.10, and those forces measured for G5(RF) n were plotted as a function of loading rates as shown in Fig. 7.10D. G5(RF)0 showed only nonspecific, weak interactions as most of its rupture events showed lack of loading rate dependency as typically expected for nonspecific events. However, some events showed a small loading rate dependency which is believed to arise from nonspecific global interactions such as electrostatic and/or van der Waals interactions between the dendritic residues and the RFBP protein. G5(RF)3 showed a loading rate dependency that was markedly different from the nonspecific interactions observed in G5(RF)0. Its rupture forces were in the range of 40–50 pN which may arise from a combination of mono, di- or trivalent interactions given its ligand distribution [100]. G5(RF)5 also showed a loading rate dependency that was different from G5(RF)3. The rupture forces observed from the G5(RF)5-RFBP interactions are higher, and in the range of 70–110 pN, and the upper end of these forces are most likely arising from multivalent binding greater than those in G5(RF)3. It is notable that the rupture forces measured in G5(RF)5 are greater than those in G5(RF)3, and even comparable to the force (~75 pN) reported for a biotin-avidin bond (K D ~ 10−15 M) [101] which constitutes one of the strongest non-covalent interactions.

In summary, this dynamic force spectroscopy study enabled us to quantitatively measure the physical forces involved in the adsorption of RF-conjugated dendrimers to the surface through multivalent receptor binding. RF valency is positively correlated with the magnitude of rupture force for dendrimer adhesion. These results were strongly supportive of its essential role in the design of RFR-targeted NPs.

7.5 Conclusion

RFR plays an essential role in cellular uptake of RF in normal physiology. Its overexpression is, however, observed in a number of cancer cell types and in cancer stem cells [19, 20, 43]. Here, we summarized the proof of concept studies reported by our laboratories which demonstrated that RF or its homologous antagonists such as lumiflavin and quinacrine [25] have a strong potential to serve as ligands for selectively targeting RFRs, specific biomarkers in tumors and cancer stem cell biology. With a rational design approach based on an available RFBP crystal structure [45], we developed linker chemistry which enabled efficient conjugation of RF at its (D)-ribose unit and isoalloxazine head without loss of its binding activity [26, 28]. A series of multifunctional RF conjugates prepared with G5 PAMAM dendrimer were demonstrated as effective nanoplatforms for RFR-targeted delivery in RFR(+) KB cells in vitro using an anticancer therapeutic agent (MTX) and a reporter gene. Imaging methods based on AFM and confocal microscopy in combination with the SPR scattering modality of AuNPs conferred an ability to investigate the receptor-mediated uptake of RF-conjugated dendrimers by tumor cells [27, 42].

Development of nanotechnology for RFR-targeted applications has started only recently following early studies on receptor-mediated uptake of RF and its protein conjugates by tumor cells [21, 23]. Despite its early stage, a number of explorative studies which have been conducted in our laboratories [24,25,26,27,28] and others [29,30,31] are strongly supportive of multiple promising applications. These include delivery of antitumor agents (MTX [26], mitomycin C [31]) by RF-conjugated nanomaterials based on the PAMAM dendrimer [26, 28, 42], AuNP [27], N-(2-hydroxypropyl)methacrylamide (HPMA) copolymer [31], and human serum albumin [21]. In conclusion, we anticipate that RFR-targeted nanotechnology has a strong potential for playing a critical role in the development of new technology and effective nanodevices for tumor-specific delivery and imaging applications.