Nanoparticles That Do Not Compete with Endogenous Ligands – Molecular Characterization In Vitro, Acute Safety in Canine, and Interspecies Pharmacokinetics Modeling to Humans

Abstract

A vast majority, if not all of the receptor-mediated drug delivery systems utilize nanoparticles that are conjugated to physiological mimic ligands, with testing restricted to in vitro and rodent models. In this report, we present for the first time, a full spectrum characterization of transferrin receptor 1 (TfR1)-targeted polymeric nanoparticles (abbreviated, P2Ns-GA) that do not compete with endogenous transferrin, and serve as a versatile platform for oral drug delivery. Based on endocytosis inhibitors and receptor knockdown, the cellular uptake of P2Ns-GA is clathrin-mediated and dependent on TfR1 expression, but other trafficking mechanisms, particularly those involving caveolae/lipid rafts, can also play a role. The utility of P2Ns-GA in promoting the oral bioavailability of encapsulated com-pounds is demonstrated with a hydrophobic polyphenol, urolithin A (UA). When compared against plain UA or UA in ligand-free nanoparticles, UA-loaded P2Ns-GA led to markedly higher plasma concentrations among healthy canines, with no adverse health effects observed after oral dosing. Finally, a semi-mechanistic pharmacokinetic model was developed using both rat and dog datasets to quantitatively evaluate the effect of P2Ns-GA on oral bioavailability of UA. The model was allometrically scaled to humans to simulate clinical pharmacokinetics of plain UA and UA-loaded P2Ns-GA following oral administration.

Introduction

The conversion of therapeutic injectables into oral dosage forms is undoubtedly advantageous, as it not only simplifies medical supervision, but also facilitates patient adherence to treatment regimens, leading to better clinical outcomes. While most existing drugs can be formulated orally, there are critical instances in which the oral route is not viable. For example, drugs belonging to Biopharmaceutical Classification System (BCS) class III or IV, (including peptides and many lipophilic compounds), exhibit very poor or erratic absorption by the gastrointestinal tract (GIT), with nearly all of the administered dose susceptible to pH- and enzyme-associated breakdown, aggregate formation, P-glycoprotein/MRP2 efflux, and intestinal CYP450-mediated inactivation [1–5]. Over the past three decades, an increasing amount of attention has been paid to nanoparticle delivery systems, which are engineered with physicochemical characteristics that provide the encapsulated drugs protection from as well as passage across the GIT [6, 7]. However, despite a plethora of basic research surrounding these drug carriers, transitioning to clinical trials has been slow, and there is currently no regulatory approval of human or veterinary versions for oral delivery [8]. Given the difficulty in developing this promising technology, it may be worthwhile to reconsider key aspects of nanoparticles, in particular the traditional design choices underlying cellular uptake.

Enabling consistent and robust gastrointestinal absorption remains one of the most challenging obstacles facing nanomedicine research. While simple diffusion across the GIT is possible, targeted or active transport offers the potential for better dose regimens, thereby addressing inherent toxicity profiles of the encapsulated drugs without compromising efficacy [8–10]. To accomplish this, nanoparticles are typically conjugated to ligands, many of which consist of commonly ingested nutrients (or their close derivatives) that are recognized by receptors expressed on the intestinal epithelium [11–13]. Ligands that have been investigated for receptor-mediated drug delivery range from molecules such as folic acid and vitamin B₁₂ (cobalamin), to peptides such as transferrin, lectin, RGD tripeptide, and the F_c region of immunoglobulin G [14–19]. Upon receptor binding, the conjugated nanoparticles are actively taken up on the apical side by clathrin or caveolin-coated vesicles, and a fraction of which are subsequently released into the capillaries or lacteals on the basolateral side through transcytosis [8]. The use of natural ligand functionalities on nanoparticles has some clear advantages such as well-established target specificity and affinity, but also drawbacks that include greater tendency for degradation (for peptide ligands) as well as more complicated scale-up synthesis (for high molecular weight ligands) [11, 12]. However, a far more serious problem inherent to nearly all ligands used to date is that their endogenous analogues can be present in the GIT at much higher concentrations. This ultimately results in the competitive occupation of cognate receptors, thereby interfering with receptor-mediated nanoparticle uptake and causing low or unpredictable delivery of the drug cargo [20]. To resolve this fundamental problem, a paradigm shift away from the use of physiological mimic ligands may be required.

The 180 kDa, homodimeric transferrin receptor 1 (TfR1) is a type II transmembrane protein abundantly expressed on the epithelial villi of duodenum and jejunum [21]. While its purpose in the GIT is likely pleiotropic, TfR1 has been shown to mediate the apical-to-basolateral shuttling of iron-bound transferrin (its physiological ligand), akin to a classic nutrient receptor [22, 23]. The possible exploitation of TfR1 for nanoparticle-based oral drug delivery had been demonstrated using transferrin as well as 7-peptide, a competitive, synthetic ligand for TfR1 derived from phage display [19, 24]. Our lab, on the other hand, reported that gambogic acid (GA), a xanthanoid found in Garcinia hanburyi resin, could non-competitively target nanoparticles to TfR1, even in the presence of excess transferrin, 7-peptide, or anti-TfR1 immunoglobulin G [20, 25]. In rodents, GA-conjugated nanoparticles have significantly elevated the oral bioavailability of cyclosporin and urolithin A, as well as enhancing their therapeutic efficacy for lupus and acute kidney injury disease models, respectively [25–28]. Based on in silico docking simulations, the non-competitive nature of GA/TfR1 binding can be explained by GA occupying an interface between the extracellular apical and protease-associated domains of TfR1, which is topologically distinct from any transferrin-binding sites [29]. Despite these data, very little pertaining to how GA-conjugated nanoparticles gain entry into cells, and the extent to which such entry depends on TfR1 expression, is known.

In this report, we offer an in-depth characterization of P2Ns-GA, (a next-generation polymeric nanoparticle conjugated to GA), and its performance relative to P2Ns, a ligand-free control nanoparticle. Following an investigation into the cellular transport mechanisms governing these nanoparticles, we conducted a multi-species in vivo analysis of P2Ns-GA and P2Ns. To the best of our knowledge, almost all animal experiments involving nanoparticles have been conducted in mice or rats, even though the oral bioavailability correlation between rodents and humans is weak (R² < 0.3) [30]. While no animal can supplant actual human testing, there is a pressing need for preclinical trials to incorporate both small and large mammals (e.g. rodents along with dogs or pigs), where consistent performance of a nanoparticle system among species of diverse anatomical, absorptive, and metabolic characteristics can be more informative with respect to its efficacy as a drug vehicle [31–33]. Using the polyphenol compound urolithin A (UA), we assessed the effect of P2Ns-GA and P2Ns on UA oral bioavailability in rats and beagle dogs. This interspecies pharmacokinetics (PK) dataset allowed us to construct an allometrically scaled model to predict UA biodistribution in humans.

Results and Discussion

To determine how P2Ns-GA and P2Ns undergo cellular internalization, various pharmacologic agents that inhibit components of endocytic pathways were applied. NHS-fluorescein labeled P2Ns-GA or P2Ns nanoparticles were added to Caco-2 cells with or without pre-exposure to ethylisopropyl amiloride (EIPA), chlorpromazine (CPZ), filipin, or methyl-beta-cyclodextrin (MβCD) [34]. The resulting cellular fluorescence was measured by FACS to evaluate the magnitude of nanoparticle binding or uptake. In untreated Caco-2 cells, P2Ns-GA achieved a 25-fold increase in mean fluorescent intensity (MFI) over P2Ns, consistent with our previous findings (Figure 1 A and F) [25]. EIPA, which inhibits macropinocytosis by disrupting the plasma membrane H⁺/Na⁺ gradient, did not have a large impact on the MFI of cells incubated with either P2Ns-GA or P2Ns, leading to approximately 19% decrease in the former and 36% increase in the latter (Figure 1 B and F). In contrast, both CPZ and filipin virtually abolished the MFI differences between P2Ns-GA and P2Ns, indicating that clathrin-mediated endocytosis (blocked by CPZ) and caveolae/lipid-raft endocytosis (blocked by filipin) were primarily responsible for P2Ns-GA binding/uptake (Figure 1 C, D and F). The cycling of many membrane receptors (e.g. insulin receptor and EGFR) can be executed by either clathrin- or caveolin-dependent pathways [35–37]. TfR1 however, utilizes the clathrin-dependent pathway exclusively, and its intracellular trafficking is known to be insensitive to filipin treatment [37]. By showing a significant sensitivity to filipin, our results suggest that other receptors besides TfR1 are involved in the binding/uptake of P2Ns-GA.

Figure 1.

Involvement of clathrin-mediated endocytosis, caveolae, lipid raft, and pinocytosis in the cellular trafficking of nanoparticles. Using Caco2 cell culture, and a selection of drugs that selectively target various endocytosis pathways, FACS analysis of P2Ns-GA and P2Ns cellular association was carried out with (a) no drug, (b) Ethylisopropyl-amiloride (EIPA), (c) chlorpromazine (CPZ), (d) Filipin complex, and (e) Methyl-beta-cyclodextrin (MβCD). (f) Mean and standard error of MFI values are plotted (n=3).

Interestingly, MβCD led to an enhancement of MFI (43% increase) for P2Ns-GA incubated cells, despite being known to inhibit the activities of both clathrin- and caveolin-mediated endocytosis by depleting membrane cholesterol (Figure 1 E and F). However, since MβCD has no effect on receptor recycling, the increased MFI could be a side effect of P2Ns-GA binding to a higher density of receptors that accumulated at the plasma membrane under endocytosis blockage [38]. While all drugs tested, (with the exception of MβCD), had a negative effect on cellular binding/uptake of P2Ns-GA, the opposite was true for P2Ns. Both CPZ and filipin resulted in large MFI increases for P2Ns, at 380% and 680%, respectively (Figure 1 A–E). These endocytosis inhibitors have been reported in literature to indirectly enhance the compensatory activities of non-specific endocytosis, such as macropinocytosis [34, 39, 40]. Therefore, it is likely that the ligand-free nanoparticle undergoes internalization primarily through non-specific, receptor-independent means, some of which become upregulated with drugs that block receptor-mediated endocytosis.

While GA has been shown to bind TfR1 in multiple in vitro assays, it is still unclear whether TfR1 is the only binding partner for GA in mammalian cells, and therefore the primary basis for improved P2Ns-GA binding/uptake [20, 29, 41]. To elucidate the role of TfR1, transfection of dicer substrate siRNA (DsiRNA) was carried out in order to transcriptionally silence all three isoforms of TfR1 in non-neoplastic, human intestinal cells (FHs74Int).

FHs74Int cells were used because they were more amenable to transfection than Caco-2, and expressed significantly more TfR1 (Figure S1A) [42]. The transfection efficiency of DsiRNA was validated to be near 100% using TYE₅₆₃ fluorescent tracer (Figure S1B), and at 72 hours post-transfection, both TfR1 mRNA and protein appeared to be expressed at less than 10% of baseline levels (Figure S1 C and D). FITC labeled P2Ns-GA or P2Ns were then incubated with normal and knockdown (KD) cells to assess nanoparticle binding/uptake efficiency using FACS and fluorescent microscopy. Our results showed that a majority KD cells (75%) still bound and internalized P2Ns-GA, compared to 85% for normal cells (Figure 2A). The difference in MFI values between KD and normal cells was more striking however, with a decrease of about 35% for P2Ns-GA in KD cells (Figure 2B). Similar changes were noted under microscopy, which showed about 45% lower P2Ns-GA binding/uptake with KD cells compared to normal cells, based on corrected total cell fluorescence (CTCF) levels (Figure 2C). By contrast, TfR1 knockdown did not alter binding/uptake for P2Ns, suggesting that GA/TfR1 interaction was an important aspect of nanoparticle binding and internalization by cells. Notwithstanding the near total absence of TfR1, P2Ns-GA still resulted in 3–4 times higher cellular fluorescence compared to P2Ns. These results agreed with the outcomes of prior endocytosis inhibitor experiments, and pointed to the existence of one or more non-TfR1 receptors that could be involved in the intracellular trafficking of P2Ns-GA.

Figure 2.

Role of transferrin receptor (TfR1) in the binding/uptake of nanoparticles with normal intestinal cells (FHs74Int). (a) FACS histograms of cell suspensions (FHs74Int) incubated with fluorescent P2Ns-GA and P2Ns showed decrease in cell binding/uptake with knockdown (TfR1 siRNA). (b) Examples of FACS histograms with P2Ns-GA or P2Ns in normal or knock down cells were shown, along with mean and standard error of MFI values associated with each (n=3). (c) Fluorescent microscopy of nanoparticles with adherent cells using P2Ns-GA (Top) and P2Ns (Bottom), with or without knockdown using TfR1 siRNA. Dotted areas indicate an area of approximately 3–4 cells. Mean and standard error of corrected total cell fluorescence (CTCF) calculated from the fluorescent images were plotted. Student t test was used for statistical analysis, *** p < 0.001.

To ascertain the effects of GA-conjugation on in vivo drug delivery, P2Ns-GA and P2Ns were separately formulated with UA to yield P2Ns-GA-UA and P2Ns-UA, respectively, whereas plain UA, solubilized in carboxymethylcellulose, served as the nanoparticle-free control (Figure 3A). As a dietary metabolite of pomegranate juice, UA has garnered intense scientific interest in the last few years owing to its remarkable anti-inflammatory, anti-oxidant, and mitophagy-inducing properties, but the bioavailability of this compound remains poor under conventional drug delivery technologies [43–45]. A major obstacle facing nanoparticle-based medicine is the scale-up of chemical synthesis and the production of consistently high-quality formulations for administering to animals larger than rodents [9, 10]. Here, we sought to demonstrate that our previous methodology to prepare P2Ns-UA and P2Ns-GA-UA could be adopted at gram-scale. Polymer synthesis showed predictable GA conjugation peaks at 6.5–6.6 ppm and 5.4–5.5 ppm under proton NMR as well as good size distribution under gel permeation chromatography (Figure 3 B and C). The assembled polymeric nanoparticles also gave excellent dynamic light scattering profiles, and looked consistently spherical under atomic force microscopy (Figure 3 D and E). The pharmacokinetics (PK) of P2N-GA-UA, P2Ns-UA and UA had already been reported in rats, but we decided to further evaluate the oral bioavailability of these formulations in healthy beagle dogs. Apart from their importance in veterinary medicine, dogs have been routinely subjected to preclinical pharmacological testing due to well-established gastrointestinal similarities with humans, including GIT motility patterns, intestinal pH profiles, MRP2 expression/distribution, and activities of some CYP450 metabolizing enzymes [32, 33].

Figure 3.

Oral delivery of urolithin-A (UA) to healthy beagles using P2Ns and P2Ns-GA, or plain compound. (a) Proton NMR spectra of P2s and P2s-GA polymers, used for the assembly of P2Ns and P2Ns-GA respectively. The presence of GA doublets is found at 6.5–6.6 ppm as well as at 5.4–5.5 ppm. (b) Size analysis of P2s and P2s-GA using gel permeation chromatography. (c) Dynamic light scattering of P2Ns-GA showing both size and polydispersity index (pDI). (d) Atomic force microscopy images of P2Ns-GA and P2Ns (yellow spheres) with a Z-axis height of approximately 230 – 300 nm. (e) Schematic of P2Ns-GA-UA and P2Ns-UA particles. (f) Plasma UA concentrations of both nanoparticle formulations and plain UA in dogs 0.5 – 48 hours after the administration of a single oral dose. AUC (area under curve), T_max (time to reach maximum concentration) and C_max (maximum concentration) were provided in the table.

We chose to deliver to each dog a single, 7.5 mg/kg dose (for P2Ns-GA-UA and P2Ns-UA) or 10 mg/kg dose (for UA) by feeding syringe, followed by multiple blood draws from the femoral vein at set intervals for up to 48 hours. Total serum UA concentration, which included both glucuronate and sulfate metabolites, was measured by LC-MS. After adjusting for dosage, P2Ns-GA-UA led to a 430% increase in area under curve (AUC) and 920% increase in maximum serum concentration (C_max) when compared to UA. Its advantage over P2Ns-UA was also significant, albeit more moderate at 57% and 130% increase for AUC and C_max, respectively (Figure 3F). A battery of blood analyses was also carried out before and after oral nanoparticle dosing to detect any occurrence of acute toxicity. While diets containing gram quantities of UA have been reported as safe in humans, GA on the other hand, is a known apoptosis inducer among cancer cells and high doses of this compound (>120 mg/kg) have exhibited both hepatic and renal toxicities in rats [46–48]. However, neither the biochemistry panel nor CBC revealed any detectable abnormalities in the blood (Figure S2). These results, together with our lab’s previous study showing that long-term ingestion of GA-decorated nanoparticles by rodents (rats and mice) caused no observable tissue injury, suggested that any health concerns associated with these nanoparticles would likely be minimal [26, 28, 29].

In order to characterize preclinical PK of UA and UA-loaded nanoparticles, and to predict oral bioavailability of UA formulations in humans, a semi-mechanistic PK model (SMPK) was developed (Figure 4A). The SMPK is distinct from a majority of PK models describing nanoparticles PK available in literature, as the former treats drug-containing nanoparticles and the free drug as separate entities. The absorption sub-model consists of seven sequential compartments for both drug-containing nanoparticles (N) and the free drug (D) prior to their entry into systemic circulation, in order to generate an advanced gastrointestinal transit simulation similar to other state-of-the-art physiologically based models such as the Discontinuous Oral Absorption Model and the GastroPlus^™ Advanced Compartmental Absorption and Transit Model (ACAT) [49, 50]. Meanwhile, the UA disposition has been captured with the help of a 2-compartment model, and drug containing nanoparticle (N) have been assumed to remain in the central compartment. The equations used for both the absorption and disposition sub-models are detailed in Figure S3. The SMPK was applied to both our dog PK dataset as well as the rat PK dataset published previously, to estimate relevant drug absorption and disposition parameters, a summary of which is provided in Figure S4 [26]. A sequential approach was adopted for model fitting, using datasets generated following administration of UA, P2Ns-UA, or P2Ns-A-UA in animals (Figure 4B). The SMPK model was able to adequately capture all the PK data (Figure 4B), and provided reasonable estimates of model parameters (Figure S4).

Figure 4.

Semi-mechanistic Pharmacokinetics (SMPK) modeling and the allometric simulation of PK in humans. (a) Schematics of the SMPK model showing free drug and nanoparticle compartments as well as critical parameters used by the model. (b) Model fitting of rat and dog PK datasets; (solid line = model simulation, data points = actual UA plasma concentrations from a single oral dose.) (c) Allometrically-scaled human PK simulations showing plasma UA concentrations over 48 hours. Two sets of human simulations were generated using either the dog or rat datasets, each of which resulted in different model parameters. (d) Human PK simulations of a single plain UA dose (250 mg, 500mg, or 1000 mg) based on dog (solid line) and rat (dashed line) datasets, along with actual data points of plasma UA concentrations obtained from a clinic trial of UA.

To extrapolate the PK profile of UA formulations to humans, the SMPK model was allometrically scaled using the commonly accepted values of 70 kg for human body weight, and 0.32 for human gastrointestinal transit constant (K_at) [51]. Overall, there was a 10-fold disparity in the estimated human plasma concentrations of UA when modeled using the rat versus the dog dataset (Figure 4C), which underscored the large, well-known inter-species differences reported in literature [30, 31]. Nevertheless, human PK simulations derived from both rats and dogs agreed on the superiority of P2Ns-GA-UA over other tested formulations. For example, the dog-derived human PK simulation suggested that UA delivered by P2Ns-GA-UA could achieve more than 500% higher bioavailability compared to plain UA based on area-under-curve estimations, an enhancement that was almost identical in magnitude to what was predicted by the rat-derived human PK simulation. Finally, we compared our human PK simulations to a recently published clinical trial involving UA, in which an actual PK dataset of the drug had been made available [52]. After adjusting for dosages, we found that the PK from the 250 mg oral dose given by the clinical trial matched our dog-derived PK simulation for plain UA quite well, whereas higher doses such as 500 mg and 1000 mg were progressively overestimated by our SMPK model (Figure 4D), which suggests a possibility of dose-dependent nonlinearity in the absorption of UA in humans. The rat-derived PK simulation on the other hand, overestimated each of the real-life UA datasets by more than 10-fold.

Conclusion

In this study, we have demonstrated that transferrin receptor 1 (TfR1) plays an important role in the cellular uptake of P2Ns-GA, but there appears to be other TfR1-independent uptake mechanisms as well. Based on various inhibitor treatments, both clathrin-mediated as well as caveolae/lipid raft-mediated endocytosis pathways are implicated in the trafficking of GA-conjugated nanoparticles. The involvement of caveolae and lipid-raft is somewhat intriguing, because neither of these membrane microdomains has been associated with the endocytosis and recycling of TfR1. Moreover, unlike clathrin-coated vesicles, caveolae/lipid raft vesicles are more likely to engage in transcytosis and circumvent endosomal/lysosomal degradation, and as a result, mitigate the problem of “easy entrance, difficult discharge” affecting all nanoparticle systems to date [8, 53]. In most cell types, the caveolae/lipid-rafts are predominantly static structures and affixed to the plasma membrane by actin microfilaments, but the binding of certain extracellular ligands, (such as albumin, folic acid, alkaline phosphatase, cholera toxin b-subunit, and SV40 virus, among others), to cognate receptors present on caveolae/lipid rafts triggers the phosphorylation of caveolins and their subsequent endocytosis [53]. It remains to be determined if GA binds to an as yet unidentified caveolae/lipid raft-localized receptor. Interestingly, PEGlyated polyester nanoparticles have been found to preferentially utilize lipid-raft endocytosis, in stark contrast to non-PEGlyated nanoparticles, such as PLGA, which are taken up almost exclusively by clathrin-mediated endocytosis in Caco-2 cells [54, 55]. While both P2Ns and P2Ns-GA incorporate PEG/PLA copolymers at their backbone, the lack of reduction in P2Ns uptake by filipin treatment shows that the presence of caveolae/lipid raft endocytosis observed in our studies may not be a consequence of PEGlyation.

The superiority of P2Ns-GA-UA over P2Ns-UA and plain UA in both rats and dogs is a clear indication that there is a well-conserved, cross-species benefit behind GA-directed, active drug delivery. Since there are numerous physiological differences between the gastrointestinal tracts of dogs and rats, including higher fasting gastric pH (rats), lower density of microvilli (dogs), increased mucin secretion (dogs), lower liquid volume per gut length (rats), and shorter small intestine transit time (dogs), drugs administered to both animals seldom correlate in terms of their oral bioavailability [30, 56]. In our PK studies, we found that the plasma concentrations of UA from all formulations was decreased significantly in dogs compared to rats. Nonetheless, the same degree of advantage (approximately 5-fold) that P2Ns-GA-UA had over plain UA was observed under both animal models. Notably, human PK estimates derived from both species revealed a dual-peak pattern of plasma UA concentration for P2Ns-GA-UA and P2Ns-UA, which was clearly absent for plain UA. The initial plasma UA peak for both P2Ns-GA-UA and P2Ns-UA was early (at 0.5–2 hours post-dosing) and robust, which could be explained by increased stability of the nanoparticle-encapsulated drug as well as additional routes of gastrointestinal uptake that were denied to plain UA[49, 57]. After extrapolating our dog UA dataset to humans using our SMPK model, we were also pleasantly surprised by how closely it matched the only published PK of the drug [52]. Despite their similarity however, there was a key difference between our model, which had the assumption of a single UA dose, and the dataset provided by the clinical trial, which consisted of 28 sequential daily doses of UA, with the PK study conducted immediately after the final dose. Given the fact that UA as well as its glucuronide and sulfate metabolites can accumulate after each daily dosing, the relative bioavailability of UA may be even worse in humans than in dogs, thereby further necessitating the use of nanoparticles to improve oral delivery [58].

Nowadays, nanoparticles have become increasingly prevalent in the fields of cancer medicine, medical imaging, and vaccine [9]. At the same time, the advent of antibody-based biologics has brought about a revolution in targeted therapies that minimize unwanted damage to healthy tissues. However, the combination of nanoparticles with targeted delivery has been an extremely challenging endeavor [9]. Many have posited that the primary obstacle lies with the ligand choices, which can sometimes be far too complex and costly for large-scale synthesis [9, 59]. In the context of oral delivery, a significant number of proposed ligands would have an added stability problem when placed under harsh GI conditions [11]. We have responded to these problems with a nanoparticle system that is simple to scale-up, and conjugated to an unconventional, non-competitive ligand [20]. By demonstrating drug delivery efficacy in dogs, we have shown that the potential of GA-conjugated nanoparticles is not an artifact of in vitro assays or limited to rodent models. Future studies on P2Ns-GA will further evaluate the safety of this system as well as gather a larger body of preclinical evidence to warrant its transition to human trials.

Materials and Methods

Polymer Synthesis and Nanoparticle Preparation.

Detailed methods are provided in Ganugula et al. 2017 and Zou et al. 2019 [25, 26]. Briefly, polymers were first synthesized from pre-polymeric blocks of PEG and PLA; these were then crosslinked with cyclohexanetetracarboxylic dianhydride (HCDA) to provide periodically spaced carboxyl groups along the polymer backbone. An ethylene diamine linker was then added to the carboxyl group by 1-Ethyl-3-(3-dimethylaminopropyl)carbodiimide (EDC) chemistry, and GA was in turn conjugated to the linker. To obtain fluorescent polymers, NHS-fluorescein instead of GA was conjugated. GA-free polymers were likewise synthesized, consisting of all of the previous components except for GA and ethylene diamine. Nanoparticle assembly utilized the oil-in-water emulsification method. UA-loaded nanoparticles were prepared by adding polymers (with or without GA) and UA to water containing 1% (Di-n-dodecyl)-dimethylammonium bromide (DMAB). Fluorescent nanoparticles were prepared by adding a mixture of 5% fluorescein-conjugated polymers and 95% non-fluorescent polymers to water containing 1% DMAB. The nanoemulsions were collected by centrifugation, suspended in 5% sucrose, and lyophilized. Immediately prior to use, water was added to the freeze-dried nanoparticles, attaining requisite concentrations.

Cell Cultures.

Caco-2 cells (passage 8–10) were seeded at a density of 4.0 × 10⁵ cells/well in a 6-well plate for all experiments (n = 3), and cultured for 72 hours in 2 ml Dulbecco’s Modified Eagle Medium (DMEM/F-12) containing 10% fetal bovine serum (FBS). FHs74Int (passage 3) were seeded at a density of 2.0 × 10⁵ cells/well in a 6-well plate for all experiments (n = 3), and cultured for 96 hours in 2 ml Hybri-Care Medium ATCC 46-X supplemented with 30 ng/ml EGF and 10% FBS; this medium was renewed every 72 hours. All cells were incubated at 37°C and 5% CO₂.

Inhibition of Endocytosis.

CPZ, filipin, EIPA, and MβCD were purchased from Sigma Aldrich (St. Louis, MO). Nanoparticle interaction with Caco-2 cells under different endocytosis modulators was assayed using flow cytometry as follows: Cells were trypsinized, and cell pellets were suspended in complete medium (DMEM/F-12 with 10%FBS). Cell suspensions were pre-incubated with the drug of interest (CPZ 1 μg/ml, filipin 1 μg/ml, MβCD 20mM, and EIPA 40μM) for 1 hour, and fluorescent nanoparticles (P2Ns or P2Ns-GA 50 μg/mL) were added to the medium. After an additional 1-hour incubation, cells were centrifuged and suspended in FACS buffer (PBS with 1% BSA), washed thrice and subjected to flow cytometry. FACS data was analyzed and plotted using FlowJo v10 (Becton Dickinson, Franklin Lakes, NJ).

Transfection of siRNA.

FHs74Int cells at 70% confluence were transfected in triplicates with 10 ng/well of dicer substrate siRNA duplex (5’-GUGAAUGGAUCUAUAGU-3’ and 5’-UGACAAUCACUAUAGAUC-3’) to target all three isoforms of TfR1. Transfection was carried out using Lipofectamine RNAiMAX (Thermo Fisher, Waltham, MA). Scrambled RNA duplex and a fluorescent tracer (TYE₅₆₃-conjugated RNA) were used for negative control and for assessing transfection efficiency, respectively. Predesigned siRNA was purchased as part of the TriFECTa RNAi kit from IDT (Coralville, IA).

Analysis of Nanoparticle/Cell Interaction with TfR1 Knockdown.

72 hours after transfection (96 hours after seeding on 6-well plates), FHs74Int cells were trypsinized, and cell pellets were suspended in serum-free medium (Hybri-Care Medium ATCC 46-X), to which fluorescent nanoparticles (P2Ns or P2Ns-GA 50 μg/mL) were added. After 1-hour incubation, cells were centrifuged, washed thrice and suspended in FACS buffer (PBS with 1% BSA), and subjected to flow cytometry. FACS data was analyzed and plotted using FlowJo v10 (Becton Dickinson, Franklin Lakes, NJ). For analysis using microscopy, FHs74Int cells remained as monolayers on 6-well plates, but the medium was replaced with serum-free medium containing fluorescent nanoparticles (P2Ns or P2Ns-GA 50μg/mL). After 1-hour incubation, the monolayers were washed thrice with PBS, and imaged using EVOS-FL microscope (Thermo Fisher, Waltham, MA). Corrected total cell fluorescence (CTFC) was calculated with the aid of ImageJ, based on the observed fluorescence (integrated density) of a region of interest subtracted by the fluorescence of the background.

Real-time PCR.

At 72 hours post-transfection, FHs-74int cells were washed three times in cold PBS buffer, lysed in 0.5 ml Trizol reagent, and scraped from the 6-well plate. RNA was extracted from the Trizol/chloroform solution, and RNA concentrations were assayed with absorbance at 230/260/280 nm using a plate reader (TECAN Infinite^® M200, Morrisville, NC). DNAse digestion and reverse transcription were carried out with iScript^™ gDNA Clear cDNA Synthesis Kit (Bio-Rad, Hercules, CA) using 1μg of RNA. iTaq^™ Universal SYBR Green Supermix (Bio-Rad) was used to amplify target transcripts on the CFX96 Touch Real-Time PCR Detection System (Bio-Rad). The following primers were used: 5’- CATGCTCATCTGGGGACAGG-3’ and 5’- CAGCTTTTCTGCAGCAGCTC-3’ for TfR1; and 5’-AGGTCGGTGTGAACGGATTTG-3’ and 5’- TGTAGACCATGTAGTTGAGGTCA for GAPDH. ΔCt values were calculated with GAPDH normalization, and fold change of the gene of interest was expressed as 2^−ΔΔCt.

Western Blot.

At 72 hours post-transfection, FHs-74int cells were washed three times in cold PBS buffer, lysed in 0.2 mL of RIPA buffer (Thermo Fisher), along with protease/phosphatase inhibitors cocktail (Cell Signaling Technology, Danvers, MA). After assaying for concentration with bicinchoninic acid reagent (Sigma Aldrich), 30μg of protein, diluted in Laemmli buffer with 2-mercaptoethanol, was loaded onto 12% SDS-PAGE. Gel-separated proteins were then transferred to nitrocellulose membrane and blocked for 1 hour in 5% skim milk. Afterwards, primary antibodies against TfR1 (Rabbit monoclonal IgG, D7G9X, Cell Signaling Technologies) or GAPDH (Rabbit monoclonal IgG, D16H11, Cell Signaling Technologies) were added and incubated overnight. The primary antibodies were removed and the membranes washed thrice using TBS with 0.25% tween. The latter was incubated with HRP-conjugated goat anti-rabbit IgG secondary antibodies (Thermo Fisher) for 1 hours, washed thrice using TBS with 0.25% tween, and visualized on ChemiDoc Imaging Systems (Biorad).

Canine Pharmacokinetics.

Animal Use Protocol for the dog study #2017–0279 was approved by the Texas A&M University Institutional Animal Care and Use Committee. Eight healthy purpose-bred juvenile intact male beagles (Ridglan Farms, Mount Horeb, WI) were included in the study and maintained at Texas A&M University, Comparative medicine program facility for duration of the study. All dogs were 4–6 months old throughout the study. Dogs were both single-housed and pair-housed in runs (2.1 m L × 1.2 m W × 3.0 m H). All the dogs used in the study provided controlled light cycle (day/night), temperature (21–22°C), and humidity (55–60%) conditions. Water was available at all times and a maintenance diet was provided twice daily.

Total 12 dogs were divided into three groups (n=4), each dog per group was administered a single oral dose, freeze-dried nanoparticle formulations suspended in water; with group 1 receiving P2Ns-GA-UA (7.5 mg/kg); group 2 receiving P2Ns-UA (7.5 mg/kg); and group 3 receiving plain UA (10mg/kg) suspended in CMC. Blood was collected during this kinetic study at 0.5h, 2h, 6h, 12h, 24h and 48h. Plasma UA concentrations were determined using LC-MS. Detailed safety assessments of oral P2Ns-GA-UA, P2Ns-UA were evaluated pre and 24-hours post dosing for clinical pathology tests such as complete blood count and serum biochemistry.

Rat UA Intravenous Pharmacokinetics.

Male SD rats (200–250 g) were purchased from approved venders (Envigo, Madison, WI), and maintained in the animal quarantine in accordance with Texas A&M Institutional Animal Care and Use Committee (IACUC). All experimental protocols were reviewed and approved by the Texas A&M University IACUC. All animal procedures for this study were carried out in accordance with the relevant IACUC guidelines and approved protocol number IACUC 2017–0219. Plain UA was dissolved in DMSO and injected to rats (10mg/kg) via tail vein for about 125μL/rat (n=4). Blood samples were collected via tail vein at 0.5h, 2h, 4h, 6h, 12h, 24h and 48h post dosing. Serum UA concentrations were determined using LC-MS.

LC-MS of Urolithin A.

UA from plasma samples were quantified on a triple quadrupole mass spectrometer (Quantiva, Thermo Scientific, Waltham, MA) coupled to a binary pump HPLC (UltiMate 3000, Thermo Scientific). MS parameters were optimized for the target compound under direct infusion at 5μL min-1 to identify the selected reaction monitoring (SRM) transitions (precursor/product fragment ion pair) with the highest intensity in negative mode as 227.1–198 m/z for UA and 240.2–148.18 m/z for the internal standard, Salbutamol. Samples were maintained at 4 °C on an autosampler before injection. The injection volume was 10μL. Chromatographic separation was achieved on a Hypersil Gold 5 μm 50 × 3 mm column (Thermo Scientific) maintained at 30 °C using a solvent gradient method. Solvent A was water (with 0.1% formic acid). Solvent B was acetonitrile (with 0.1% formic acid). The gradient method used was 0–4 min (20% B to 80% B), 4–4.1 min (80% B to 20% B) and 4.1–6 min (20% B). The flow rate was 0.5mL min-1. Sample acquisition and analysis was performed with TraceFinder 3.3 (Thermo Scientific).

Pharmacokinetic Modeling -

The developed semi–mechanistic PK (SMPK) model aims to capture the rat and dog PK datasets of UA when administered orally as free drug, or as encapsulated drugs in P2Ns and P2Ns-GA. The equations used for this model are included in FIgure S3, and a summary of all the PK parameters along with their estimated values has been provided in FIgure S4. The absorption sub-model of SMPK model simulates 7 sequential compartments for the nanoparticle (N₀-N₆) and free drug (D₀-D₆). The value of K_at is the first order transit rate constant, which is defined as the inverse of gastrointestinal transit time. K_at is assumed to be the same across all compartments. The rate of drug transfer from N₀-N₆ to D₀-D₆ is defined by K_rel, the first order drug release constant. We assumed that drug release from N₀-N₆ into D₀-D₆ would be an irreversible event. In terms of absorption, we assumed that the free drug would be absorbed in all compartments except D₂, while the unconjugated nanoparticle would be absorbed only in N₀ and N₁. Nanoparticles actively targeted by GA was assumed to be absorbed in N₀-N₁ as well as N₅-N₆ compartments. This assumption was in line with observations that TfR1 expression was significantly higher in both the early and terminal regions of the GIT (i.e. up to the proximal jejunum and colon) [23, 60]. A scaling factor ‘A’ was applied to P2Ns-GA-UA over P2Ns-UA to account for improved absorption due to active targeting. Free drugs are absorbed with a first order rate constant K_a and bioavailability F, whereas nanoparticles are absorbed with a first order rate constant K_aN and bioavailability S. The disposition sub-model has two compartments for the free drug, along with a single systemic compartment for the nanoparticles.

In order to estimate the parameters, a sequential approach was adopted. First, the intravenous and oral PK data for plain UA were fit simultaneously in order to estimate the free drug parameters such as Cl_drug, Cl_d, V₁, and V₂, along with K_a and F in rats. Next, the P2Ns-UA oral PK data was used to fit the model, with the free drug parameters fixed to previously estimated values. This process estimated nanoparticle related parameters such as K_aN, S, K_rel, and K_deg in rats. Finally, after applying the scaling factor ‘A’, the P2Ns-GA-UA oral PK data was used to fit the model in rats. In order to fit the dog data, the disposition parameters, such as Cl_drug, Cl_d, V₁, and V₂, were scaled from rats to dogs using the equations for allometric scaling in FIgure S3. The body weight for dogs was assumed to be 7.0 kg. A similar sequential approach was then taken to estimate other model parameters from the dog PK dataset. Human simulations were undertaken by using the respective model parameters estimated from the rat and dog datasets. Those parameters, such as Cl_drug, Cl_d, V₁, V₂, and K_a were allometrically scaled to humans using fixed body weight of 70.0 kg and K_at of 0.32. The resulting model parameters for human simulations are provided in FIgure S4. To show that the SMPK model is not only adequate to predict PK values in larger organisms, but also in smaller organisms as well, we created a model simulation of UA plasma PK in mouse using the rat and dog datasets (FIgure S5).

Supplementary Material

1

Figure S1. (a) Expression of TfR1 by Caco-2 and FHs-74 int cells assayed using Western blot. (b) The transfection efficiency of FHs-74 int cells was high (>90%), as determined by TEX₆₁₅ fluorescent tracer. The silencing of TfR1 was confirmed real-time PCR (c) as well as by Western blot (d). Error bars indicate standard error, (n = 3).

2

Figure S2. Safety profiles of P2Ns-UA and P2Ns-GA-UA formulations 0 and 24 hours after the administration of a single dose, as analyzed by biochemistry panel (top) and complete blood count (bottom). Serum and whole blood were used for the biochemistry panel and complete blood count, respectively. Values displayed represent mean ± standard error (n=4).

3

Figure S3. Equations used to characterize preclinical animal PK of UA and UA-loaded nanoparticles, and simulation of their human PK.

4

Figure S4. PK parameters obtained from the actual rat and dog datasets using equations provided in FIgure S3, along with allometrically scaled human PK parameters.

5

Figure S5. SMPK model predicted PK of UA and UA-loaded nanoparticles in mouse, which was obtained by allometrically scaling of the PK parameters obtained using the rat and dog datasets. The data here shows that the SMPK model can be used to simulate the PK of nanoparticles in animal species that are both larger and smaller compared to the test organism.