Folate-mediated intracellular drug delivery increases the anticancer efficacy of nanoparticulate formulation of arsenic trioxide

Arsenic trioxide (As₂O₃) is a potent clinical agent for the treatment of acute promyelocytic leukemia and is rapidly moving toward designation as a frontline agent (1, 2). It also shows significant activity in relapsed/refractory multiple myeloma (3). The mechanism of action is not fully understood; however, As₂O₃ is involved in induction of differentiation, apoptosis, and angiogenesis (3). Recent studies ofAs₂O₃ show anticancer activity against a variety of solid tumor models and tumor cell lines, including liver, gastric, ovarian, cervical, breast, prostate, renal, and bladder cancer (4–6). Clinical response of solid tumors to As₂O₃ in many cases, however, has been poor compared with that of hematologic cancers, such as acute promyelocytic leukemia (4, 7), and much higher As₂O₃ dosages are required for solid tumors (8). High doses are accompanied by severe side effects, including peripheral neuropathies, liver failure, and cardiac toxicity, thus limiting their clinical utility (4, 9). Previous pharmacokinetic studies have shown that plasma arsenic is eliminated rapidly with a half-life of ∼12 hours after i.v. administration (10, 11), which may partially account for the limited activity of As₂O₃ in solid tumors. To improve the therapeutic index of the drug and expand its clinical utility to solid tumors, an effective delivery system is needed that can increase its accumulation at tumor sites, diminish off-target toxicity, and extend the circulation time of the active agent in blood (12). Encapsulation in liposomes can improve the therapeutic efficacy of numerous agents by reducing their side effects and increasing the drug concentration in tumors through the enhanced permeability and retention effect (13). Liposomal doxorubicin (Doxil), for instance, has an improved therapeutic index and safety profile over its parent drug, doxorubicin, in the clinic (14); however, analogous preparations of liposomal As₂O₃ face a number of challenges.

Instability has been a limiting feature in the few published accounts of liposomal or polymeric carriers for As₂O₃: substantial amounts of the drug are lost within a few hours under physiologic conditions or over a few days under storage conditions (15, 16). In neutral aqueous solutions of As₂O₃, the uncharged As(OH)₃ species (which readily diffuses across lipid membranes) is predominant (17, 18). Recently, we have developed a novel process for encapsulating nanoparticulate forms of As₂O₃ (18) into 100-nm liposomes with high density (270 mmol/L), excellent retention, and long shelf life (>6 months at 4°C). The method allows As₂O₃ and coencapsulated metal ions to form pH-sensitive nanoparticles within liposomes. Importantly, the pH dependence of the nanoparticle assembly enables these liposomes to unload the drug cargo when the pH is lowered to 6 or 5, as encountered in cellular endosomes and lysosomes (19, 20). Lipid encapsulation further reduces the toxicity of the nanoparticulate As₂O₃ in vitro (18). These novel arsenic liposomes present a platform for subsequent conjugation with targeting ligands, such as folic acid and antibodies, which would further enhance tumor uptake. The vitamin folic acid (FA) has a high binding affinity to folate receptors (FR) on the cell surface of many human tumors (21). FR is a tumor marker because it is highly overexpressed in malignant tissues of epithelial origin relative to normal tissues (21). Folic acid has hence emerged as an effective targeting ligand for selective delivery of attached imaging and therapeutic agents to cancer tissues (22).

In this study, we have applied the nanoparticulate formation process (18) to coencapsulate As₂O₃ with transition metal ions (Ni²⁺ and Co²⁺) into 100-nm folate-targeted liposomes, the bilayer of which contains a small amount (0.3 mol%) of DSPE-PEG₃₃₅₀-folate [i.e., folic acid conjugated with polyethyleneglycol (MW 3350)–derivatized distearoylphosphatidylethanolamine]. This assembly is shown to be robust and to undergo triggered drug release under mildly acidic conditions. Confocal microscopy, quantitative drug analysis, and cytotoxicity studies reveal that these folate liposomal arsenic agents are efficiently taken up by tumor cells in a receptor-mediated process and have significantly enhanced anticancer efficacy relative to the parent drug As₂O₃.

Dipalmitoylphosphatidylcholine (DPPC), 1,2-dipalmitoyl-sn-glycero-3-phosphoethanolamine-N-(Lissamine rhodamine B sulfonyl; ammonium salt; DPPE-Rh), and 1,2-dipalmitoyl-sn-glycero-3-phosphoethanolamine-N-[methoxy(polyethylene glycol)-2000] (ammonium salt; DPPE-PEG₂₀₀₀) were purchased from Avanti Polar Lipids. 1,2-Distearoyl-sn-glycero-3-phosphoethanolamine-N-[folate(polyethylene glycol)-3350] (DSPE-PEG₃₃₅₀-Folate) was synthesized as previously reported (23). Cholesterol (Chol), nickel(II) acetate [Ni(OAc)₂], cobalt(II) acetate [Co(OAc)₂], sodium chloride, acetic acid, folic acid, paraformaldehyde, HEPES, MES, and Sephadex G-50 were obtained from Sigma. RPMI 1640 was from Invitrogen-Life Technologies. Eagle's MEM and fetal bovine serum (FBS) were from the American Type Culture Collection (ATCC). l-Glutamine, penicillin-streptomycin, and PBS were from MEDIATECH. Folate binding protein antibody (anti-LK26) was from Abcam, Inc., and Alexa Fluor 488 goat anti-mouse IgG was from Invitrogen.

Liposome compositions used in this study were as follows: (a) folate-targeted liposomes, DPPC/Chol/DPPE-PEG₂₀₀₀/DSPE-PEG₃₃₅₀-Folate = 53/45/1.7/0.3 mol%; (b) nontargeted liposomes, DPPC/Chol/DPPE-PEG₂₀₀₀ = 53/45/2 mol%; (c) rhodamine (Rh)-labeled folate-targeted liposomes, DPPC/Chol/DPPE-PEG₂₀₀₀/DSPE-PEG₃₃₅₀-Folate/DPPE-Rh = 52.5/45/1.7/0.3/0.5 mol%; (d) Rh-labeled nontargeted liposomes, DPPC/Chol/DPPE-PEG₂₀₀₀/DPPE-Rh = 52.5/45/2/0.5 mol%. The lipid mixtures in chloroform were evaporated using a rotary evaporator and then placed under high vacuum overnight to remove any residual solvent. The resulting dried lipid films were hydrated in 300 mmol/L nickel acetate [Ni(OAc)₂] or cobalt acetate [Co(OAc)₂] aqueous solutions and subsequently subjected to 10 freeze-and-thaw cycles (freezing in ethanol/dry ice bath and thawing in 50°C water bath). The hydrated liposomes were sequentially extruded at 50°C to 60°C with a manual mini-extruder (Avanti Lipids), through a series of polycarbonate filters of pore size ranging from 0.4 to 0.1 μm. Extruded liposomes were fractionated on Sephadex G-50 columns equilibrated with a buffer composed of 150 mmol/L NaCl and 20 mmol/L HEPES (pH 6.8). The Ni(OAc)₂ or Co(OAc)₂ encapsulated liposomes with or without folate conjugation, f-Lip(Ni), Lip(Ni), f-Lip(Co), or Lip(Co), were then incubated with a As₂O₃ solution at 50°C for 2.5 h. After cooling to room temperature and removal of extraliposomal As₂O₃ with Sephadex G-50 columns, the concentrations of lipids (P), encapsulated As, and M (Ni or Co) in the excluded fractions were determined with an inductively coupled plasma optical emission spectrometer (ICP-OES; Vista MPX). The molar ratios of As/lipid, M/lipid, and As/M were calculated and used to assess loading efficiency. The As and Ni- or Co-coloaded liposomes with or without folate targeting, f-Lip(Ni, As), Lip(Ni, As), f-Lip(Co, As), or Lip(Co, As), had the molar ratios of 0.45 ± 0.08 As/lipid, 0.60 ± 0.1 M/lipid, and 0.75 ± 0.03 As/M (M = Ni, Co). Because the As and M (Ni or Co) species formed solid M(HAsO₃) nanoparticulates within liposomes (18), we can also express the loading as a total number of moles of As per liter of intraliposomal volume. For 100- to 200-nm-scale liposomes, the encapsulated volume is ∼1.5 L/mol phospholipid (24). The 0.45 As/lipid molar ratio of arsenic liposomes corresponds to ∼300 mmol/L arsenic within one liposome. Rhodamine (Rh)-labeled liposomes with or without folate targeting had similar loading efficiencies. The mean liposome sizes (115 ± 20 nm) were determined by dynamic light scattering on a ZetasizerNano ZS (Malvern Instruments). Visualization of folate-targeted liposomes f-Lip(Ni, As) by transmission electron microscopy is shown in Fig. 1A. Transmission electron microscopy samples were stained with 4% uranyl acetate and air-dried before imaging at 200 kV, magnification ×40,000 (Hitachi HF2000, Hitachi High-Technologies Corporation).

Samples of f-Lip(Ni, As) and f-Lip(Co, As) were kept at 4°C or 37°C at different pH values with lipid concentrations of 1.0 mmol/L. An extraliposomal buffer of 150 mmol/L NaCl and 20 mmol/L HEPES was used for maintaining the pH at 7.4; for pH 5.0 or pH 4.0, additional 20 mmol/L MES or acetic acid was added, respectively. For serum samples, liposomes were mixed with FBS in a volume/volume ratio of 2:8 (80% serum) at pH 7.4, 37°C. At various time points, aliquots were applied to a Sephadex G-50 column to remove the arsenic and metal species that had leaked from liposomes. The drug-to-lipid molar ratios in the excluded liposome fractions were determined as described above. The drug release percentage (%) was calculated as [(r_o−r_i)/r_o] × 100%, where r_o is the initial As(M)-to-lipid molar ratio and r_i the As(M)-to-lipid molar ratio at a specific time point (M = Ni, Co). The results are compared in Fig. 1B and Supplementary Fig. S1.³

KB (human nasopharyngeal epidermal carcinoma cells, FR⁺), HeLa (human cervical carcinoma cells, FR⁺), and MCF-7 (human breast carcinoma cells, FR⁻) cells were purchased from ATCC and maintained in a humidified atmosphere containing 5% CO₂ at 37°C following the protocols provided by ATCC. Cells were then cultured in folate-deficient RPMI 1640 with 10% FBS and 1% penicillin-streptomycin for a period (KB, 2 wk; HeLa, 2 mo; MCF-7, 24 h) before each experiment.

Cells were stained with folate binding antibody (anti-LK26; refs. 25, 26) for 30 min at 4°C. Control cells were incubated in medium only. After removal of excess antibody by PBS washing, cells were treated with the secondary antibody, Alexa Fluor 488 goat anti-mouse IgG, incubated for 30 min at 4°C, and washed with PBS. Cells were then suspended in PBS-buffered 0.5% paraformaldehyde for flow cytometry analysis using a Beckman Coulter Epics XL-MCL instrument (Beckman Coulter, Inc.). The fluorescence intensity observed by flow cytometry is correlated to the amount of FR antigen on the cell surface. A histogram of the fluorescence intensity was plotted and the median fluorescence intensity per cell for each cell type, KB, HeLa, and MCF-7, are compared in Supplementary Fig. S2.³

Cells were plated, 24 to 48 h before each experiment, on 22-mm coverslips inside six-well plates. Cells were exposed to the Rh-labeled liposomes at 37°C for various times at a lipid concentration of 40 μmol/L and an arsenic concentration of 18 μmol/L. After drug medium removal, cells were washed with PBS × 4 and fixed with PBS-buffered 4% paraformaldehyde at 20°C for 7 min, then washed with PBS × 1. Next, the coverslips were mounted on slides coated with PBS. Microscopic visualization of cells was done using a Zeiss confocal laser scanning microscope (Carl Zeiss LSM 510). For rhodamine (Rh), maximum excitation was obtained from the 543-nm line of a He-Ne laser, and fluorescence emission intensities >570 nm were observed using a long-pass barrier filter LP-570. A water immersion objective, C-Apochromat 63 × 1.2 W corr. (Zeiss), was used. Cells were also imaged by light microscopy using differential interference contrast.

Cells were plated, 24 h before each experiment, in six-well plates at 500,000 cells per well. Cells were exposed to 10 μmol/L arsenic as free drug or within liposomes for various times at 37°C. The same Ni and Co concentrations (13 μmol/L) of f-Lip(Ni) and f-Lip(Co) were used as in f-Lip(Ni, As) and f-Lip(Co, As). After washing with PBS to remove nonassociated drugs, cells were released from tissue culture plates with 0.05% trypsin/0.02% EDTA (Invitrogen), followed by 3 × PBS washing (centrifugation, 500 × g, 5 min). A sample was taken for cell number determination through Guave ViaCount Assay (27),⁴

Cells were plated, 24 h before each experiment, in 48-well plates at a density of 30,000 to 60,000 cells/mL, 0.2 mL per well. Cells were incubated with drugs continuously for 96 h, or exposed to drugs for 3, 12, and 24 h at 37°C, then washed with PBS × 2 and further incubated up to 96 h in drug-free medium. The same Ni and Co concentrations of Ni(OAc)₂, Lip(Ni), Co(OAc)₂, and Lip(Co) were used as in f-Lip(Ni, As) and f-Lip(Co, As). Cell viability was determined by Guave ViaCount (27),⁴ using a Guava EasyCyte Mini flow cytometer (Guava Technologies). Cell growth rates were expressed as a function of drug concentration on a logarithmic scale. The IC₅₀ values (the drug concentration required for 50% inhibition of cell growth) were determined by fitting to a sigmoidal dose-response curve using Origin 6.0 (Microcal Software, Inc.). The mean values and SDs are based on two to four independent experiments.

KB/MCF-7 cocultures were prepared by plating a mixture of HeLa and MCF-7 cells in folate-deficient medium on 22-mm coverslips inside six-well plates. After 24 h, KB/MCF-7 cocultured cells were exposed to Rh-labeled folate liposomal arsenic [f-Lip(Ni, As)-Rh] at a lipid concentration of 40 μmol/L for 3 h at 37°C, washed, and visualized by light and confocal microscopy (Fig. 2J). The micrographs were compared with those of KB (Fig. 2E) and MCF-7 (Fig. 2I) alone.

Folate-targeted liposomal arsenic agents were efficiently loaded with a high density of drug, corresponding to an As/lipid = 0.45 molar ratio. The obtained agents were stable under 4°C storage conditions with <2% drug release within 1 week and <20% within 6 months at pH 7.4 (Supplementary Fig. S1A).³ When f-Lip(Ni, As) was kept at 37°C with 80% FBS, 20% arsenic was released after 24 hours (Supplementary Fig. S1C). Drug release is stimulated by lowering the solution pH (Fig. 1B; Supplementary Fig. S1D).³ After 24 hours at 37°C, 80% arsenic was released from f-Lip(Ni, As) and 95% from f-Lip(Co, As) at pH 4.0, compared with 15% and 13% release at pH 7.4, respectively. For both f-Lip(Ni, As) and f-Lip(Co, As), 20% to 30% of the arsenic was released after 24 hours at pH 5.0.

Two FR-positive (FR⁺) tumor cell lines, KB and HeLa, as well as the FR-negative (FR⁻) cell line, MCF-7, were studied for binding of folate liposomal arsenic. FR expression on the cell surface was analyzed by flow cytometry using a folate binding antibody (anti-LK26; refs. 25, 26). Significant FR expression was found for both KB and HeLa cells, but not for MCF-7, with the order of KB > HeLa >> MCF-7 (Supplementary Fig. S2).³ This result is consistent with the previous FR analysis using [³H]folic acid (28). Cellular association of folate liposomal arsenic was examined by confocal microscopy using rhodamine (Rh)–labeled liposomes. As shown in Fig. 2, binding of f-Lip(Ni, As)-Rh to KB cells in folate-free medium was obvious within a 30-minute exposure (Fig. 2A and B), mainly through surface binding, as indicated by the red fluorescence of Rh on the cell surface. Subsequent liposome internalization and accumulation in the cytosol was observed at 1 hour (Fig. 2C) and became more obvious after 2 hours (Fig. 2D and E), with more red punctuate fluorescence around the nuclei. In contrast, nontargeted Lip(Ni, As)-Rh showed little fluorescence above background, indicating little cellular association at 4 hours (Fig. 2G). For free ligand competition studies, 2 mmol/L folic acid was added, resulting in a dramatic decrease of cellular association (both surface binding and internalization) for f-Lip(Ni, As) (Fig. 2F), indicating folate-mediated endocytosis of the nanoparticulate arsenic agents (Fig. 3A). For HeLa cells, a similar endocytosis process of f-Lip(Ni, As) was observed, but at a lower extent (Fig. 2H), probably due to their lower FR level relative to that of KB cells (Supplementary Fig. S2;³ ref. 28). FR-negative MCF-7 cells showed no significant association with f-Lip(Ni, As) at 4 hours (Fig. 2I).

In coculture studies, KB cells were mixed with MCF-7 and maintained in folate-free medium. KB cells exhibit a morphology (round and isolated; Fig. 2E) that is easily distinguishable from that of MCF-7 cells (flat and aggregated; Fig. 2I) by light microscopy. These cocultures were exposed to f-Lip(Ni, As)-Rh for 3 hours at 37°C before confocal microscopy studies. We find that a significant amount of folate-liposomal arsenic is taken up by the KB cells but not by MCF-7 cells (Fig. 2J), indicating that the folate ligand mediates selective targeting of liposomal arsenic.

ICP-MS analysis of cellular arsenic levels revealed that f-Lip(Ni, As) uptake by KB cells was rapid over the first 3-hour incubation at 37°C (Fig. 3B). At 3 hours, f-Lip(Ni, As) uptake by KB cells was about six times higher than those of Lip(Ni, As) and f-Lip(Ni, As) + 2 mmol/L FA and three times higher than that of free As₂O₃. For HeLa cells, uptake of f-Lip(Ni, As) gradually increased within the first 4-hour period, reaching levels significantly higher than those for Lip(Ni, As), f-Lip(Ni, As) + 2 mmol/L FA, and free As₂O₃ (Supplementary Fig. S3B).³ For comparison, there were 6.7 × 10⁸ arsenic atoms taken up per KB cell at 3 hours when treated with f-Lip(Ni, As), which is 2.5 times higher than that taken up by HeLa cells (2.7 × 10⁸ arsenic atoms per cell) and 6.8 times higher than that taken up by MCF-7 cells (0.98 × 10⁸ arsenic atoms per cell; Fig. 3B and C).

The cytotoxic effects of various arsenic formulations toward KB, HeLa, and MCF-7 cells in folate-free medium are compared in Table 1 and Fig. 3D to F. At 3 hours, f-Lip(Ni, As) was approximately 28 (for KB cells) and 9 (for HeLa cells) times more cytotoxic than Lip(Ni, As), f-Lip(Ni, As) + 2 mmol/L FA, and free As₂O₃, and showed significant folate-mediated targeting effect. Notably, liposome encapsulation alone generally reduced the cytotoxicity of free As₂O₃ to both KB and HeLa cells (Table 1). After targeting by folate ligands [f-Lip(Ni, As)], the potency of liposomal arsenic was significantly enhanced, with cytotoxicity increased by 28 times (3 hours), 9 times (12 hours), 6 times (24 hours), and 4 times (96 hours) in KB cell cultures, and by 9 times (3 hours), 5 times (12 hours), 3 times (24 hours), and 3 times (96 hours) in HeLa cultures (Table 1). The degree of potentiation of f-Lip(Ni, As) versus Lip(Ni, As) gradually decreased at longer drug exposure times, probably due to nonspecific cellular association of liposomes and/or passive arsenic release, both of which would contribute to the cytotoxicity of nontargeted Lip(Ni, As). For MCF-7 cells, f-Lip(Ni, As) show similar cytotoxic effects to Lip(Ni, As) at 3, 12, and 24 hours (Table 1), without obvious folate-mediated targeting effect. This is consistent with the lowest FR level in MCF-7 cells (Fig. 2I and Supplementary Fig. S2).³

KB cells were exposed to f-Lip(Ni, As), f-Lip(Ni), f-Lip(Co, As), and f-Lip(Co) at 10 μmol/L As and 13 μmol/L Ni or Co for 3 hours at 37°C. Similar levels of cellular uptake of As, Ni, or Co (atoms per cell) were found among different drug formulations (Table 2), consistent with the confocal microscopy visualization (Supplementary Fig. S4).³ This indicates that these four drug formulations have similar efficiency of folate-mediated cellular uptake.

KB cells were exposed to various drug formulations for 3 hours at 37°C, washed, and further incubated for 93 hours. F-Lip(Co, As) gave an IC₅₀ >200 μmol/L As or 260 μmol/L Co, which is 28 times higher than that of f-Lip(Ni, As) (IC₅₀ of 7.1 μmol/L As or 9.2 μmol/L Ni; Table 2 and Supplementary Fig. S5).³ Before arsenic loading, f-Lip(Ni) showed a slight cytotoxic effect with IC₅₀ of 61.4 μmol/L Ni, whereas f-Lip(Co) had an IC₅₀ >300 μmol/L Co (Table 2). A similar result was found for HeLa treatments, with f-Lip(Ni, As) being up to nine times more potent than f-Lip(Co, As) at 3hours (Supplementary Fig. S8A).³ These results indicate that the nickel component within the folate-targeted arsenic liposome serves as an adjuvant that stimulates the anticancer activity of the arsenic drug.

This study shows that nanoparticulate forms of the anticancer drug As₂O₃ can be specifically targeted to cancer cells in a ligand-targeted manner. The nanoparticulate drug is entrapped in folate-tethered liposomes with the aid of coencapsulated transition metal ions and undergoes a folate-mediated endocytosis, leading to highly selective cell killing. This approach significantly increases both the potency and specificity of As₂O₃ to the relatively insensitive solid tumor–derived cells and holds the promise of improving the therapeutic index of the drug and expanding its utility in treatment of a wide variety of cancers.

Stable nanoparticulate formation of liposomal As₂O₃ are efficiently prepared by preloaded liposomes with high concentrations of transition metal acetate salts and subsequent addition of aqueous solutions of As₂O₃ (18). These arsenic “nanobins” have a long shelf life and exhibit attenuated toxicity relative to free As₂O₃ at neutral pH (18). Under mildly acidic conditions such as those found in endosomal/lysosomal compartments, however, these agents undergo triggered arsenic drug release with As(OH)₃ (the active species of As₂O₃; ref. 17) diffusing out of liposomes (18). We show here that this arsenic-loaded system is a robust platform for further optimization by conjugation with specific ligands or antibodies for targeting to specific tumors. Folic acid and its derivatives have been used as effective targeting ligands for treatment of certain tumors (22) because FR is often overexpressed in malignant tissues of epithelial origin, such as ovarian carcinomas (29), but not in normal tissues. Folate targeting in our arsenic liposomal system is achieved with DSPE-PEG₃₃₅₀-Folate (Fig. 1) to present the folate ligand on the liposome surfaces. The presence of a small amount (0.1-0.5 mol%) of DSPE-PEG₃₃₅₀-Folate resulted in the efficient FR-dependent uptake of liposomes (30). Here, we used 0.3 mol% amount of DSPE-PEG₃₃₅₀-Folate, which was added to the original mixture of lipids. After the lipid film was hydrated in a 300 mmol/L Ni(OAc)₂ or Co(OAc)₂ aqueous solution and then extruded through 100-nm filters, and the extraliposomal metal ion was removed by gel filtration, a solution of As₂O₃ was added and allowed to actively load into these folate-targeted liposomes through diffusion of the neutral As(OH)₃ species across the liposomal membrane. We show that the presence of folate ligands did not impair the loading efficiency and stability of liposomal arsenic: high encapsulation efficiency was achieved with As/lipid of 0.45 molar ratio, corresponding to ∼300 mmol/L arsenic within one single folate liposome. The resulting folate-modified agents are homogenous and stable under storage conditions (Fig. 1A). Although the folate liposomal arsenic drug was stably retained (Supplementary Fig. S1A)³ with a long shelf life (<20% release within 6 months at 4°C and pH 7.4), its release was facilitated by lowering the pH to 5 (Fig. 1B and Supplementary Fig. S1D),³ a pH close to that found in endocytic compartments of the folate-mediated endocytosis pathway (19, 20, 31).

We further found that these folate-targeted arsenic-loaded liposomes have efficient cellular uptake and anticancer activities with high specificity. Confocal microscopy studies indicate rapid binding of f-Lip(Ni, As) through surface receptor association followed by endocytosis for KB (FR⁺) and HeLa (FR⁺) cells, but not MCF-7 (FR⁻) cells (Fig. 2). In addition, f-Lip(Ni, As) seemed to be internalized exclusively by KB cells in KB/MCF-7 cocultures (Fig. 2J), suggesting that folate liposomes can facilitate in vivo targeting of As₂O₃ drugs to tumors whose cells overexpress FRs, while preventing the killing of adjacent FR-negative cells in normal tissues. Quantitative analysis further revealed the rapid uptake and substantial accumulation of arsenic within KB and HeLa cells, but not MCF-7 cells when treated with f-Lip(Ni, As) for 3 hours at 37°C (Fig. 3B and C), largely consistent with confocal microscopy findings (Fig. 2). Notably, Lip(Ni, As) had approximately three times lower uptake than free As₂O₃ at 3 hours by KB and HeLa cells (Fig. 3B; Supplementary Fig. S3B).³ As expected, liposome encapsulation reduced the cellular membrane permeability of As₂O₃; however, when such drug-loaded liposomes are linked with folate ligands [f-Lip(Ni, As)] and delivered into tumor cells through folate-mediated endocytosis (Fig. 2E and H), an efficient intracellular unloading of liposomal contents [As(OH)₃ and metal ions] can be achieved (19, 20, 31) and thus accelerate biological actions of drugs (Fig. 3A). This may account for the significantly enhanced cytotoxicity (up to 28-fold for KB and 9-fold for HeLa) of f-Lip(Ni, As) relative to those of free As₂O₃ and Lip(Ni, As) (Table 1).

We also found that the nature of the coencapsulated transition metal ions inside folate liposomes can influence the efficacy of As₂O₃ (Table 2 and Supplementary Fig. S8).³ After a 3-hour exposure at 37°C, f-Lip(Ni, As) were 28 and 9 times more cytotoxic than f-Lip(Co, As) toward KB and HeLa cells, respectively, although the cellular uptake efficiency of f-Lip(Co, As) was very close to that of f-Lip(Ni, As) (Table 2). Given that both f-Lip(Ni, As) and f-Lip(Co, As) exhibit similar extents of drug release at low pHs (Fig. 1B; Supplementary Fig. S1D),³ the much higher cytotoxicity of f-Lip(Ni, As) versus f-Lip(Co, As) indicates that when the agents are simultaneously delivered through folate-mediated endocytosis (Fig. 3A), Ni²⁺ ions enhance the biological activity of As₂O₃ relative to Co²⁺.

Other agents are known to enhance the antitumor activities of As₂O₃, including ascorbic acid (32), all-trans-retinoic acid (33), l-buthionine-sulfoximine (34), and docosahexaenoic acid (35, 36). These agents are thought to act through modulation of the cellular glutathione (GSH) redox system, which is a critical component of the cellular response to oxidative stress and the induction of apoptosis (9, 37). Agents that lower GSH levels are known to potentiate the apoptotic effect of As₂O₃, whereas those preserving or increasing GSH levels can cause a reduction in the apoptotic effect of As₂O₃ (37, 38). Ni²⁺ ions have been reported to lower GSH levels (39–41), whereas Co²⁺ ions are thought to increase it based on a number of in vitro and in vivo studies (42–44). These findings may be correlated with our observations in this study: Ni²⁺ ions increased the cytotoxic effect of As₂O₃, whereas Co²⁺ ions decreased it (Table 2), after codelivery into tumor cells by folate-targeted liposomes (Fig. 3A). In contrast, such enhancing effects of Ni²⁺ on As₂O₃ cytotoxicity were not found in cells treated with simple aqueous mixtures of As₂O₃ and Ni(OAc)₂ (Supplementary Fig. S9)³ or with nontargeted Lip(Ni, As) (Supplementary Figs. S7 and S8).³ Furthermore, neither free, nontargeted liposomal Ni(OAc)₂ nor empty folate liposomes were toxic within the effective dose range of As₂O₃ (Supplementary Fig. S6),³ warranting the safety of this approach of codelivering As₂O₃ with the Ni²⁺ adjuvant using folate liposome vesicles to efficiently target specific tumors. These findings emphasize the advantage of targeted liposomes as a system for codelivery of an active agent and a nontoxic adjuvant.

In conclusion, we have shown a promising strategy for targeting As₂O₃ to specific tumor cells using folate-tethered liposomes. As₂O₃ can be stably and efficiently loaded into folate liposomes through transmembrane gradients of transition metal ions (Ni²⁺ and Co²⁺). The resulting folate liposomal arsenic drugs show higher anticancer efficacy against FR-overexpressing solid tumor cells, which are relatively insensitive to free As₂O₃. The coencapsulated Ni²⁺ ions contribute to retention of the As₂O₃ drug within the liposomal carrier under physiologic situations and modulate pH-triggered drug unloading in endocytic compartments involved in cellular uptake. They also enhance the anticancer activity of As₂O₃. Future preclinical studies will elucidate the role of enhanced permeability and retention and folate-directed targeting in vivo. Our results provide a rationale for combined Ni/As therapy in further animal studies through the folate-targeted liposomal codelivery system. This novel targeting system will be extended to treatments of other FR-overexpressing tumors, such as ovarian cancer (29) and lymphoma (45), and to improve the therapeutic profile of the drug.

No potential conflicts of interest were disclosed.

We thank Rebecca Marvin for help with ICP-MS; Dr. R. MacDonald for critical reading of the manuscript; and Sang-Min Lee and Drs. S.T. Nguyen, T.K. Woodruff, and J.E. Burdette for stimulating discussions.