Introduction

Arsenic (As) has a long history as both drug and poison [1,2,3]. The most common therapeutic form of arsenic, arsenic trioxide, As2O3 (ATO), can enter cells via passive transport as As(OH)3 [2, 4]. Among other diseases, cancer has been a target of arsenic drugs. One blood cancer has been successfully treated with ATO, but neither ATO nor other arsenic drugs have been clinically successful in treating solid cancer tumors. A stumbling block has been that, to penetrate solid tumors, higher, systemically toxic dosages of ATO are required. These higher ATO concentrations are also difficult to maintain due to rapid renal clearance. Synthetic nanoparticles containing various formulations of ATO were found to enter cells via endocytosis and to improve antitumor efficacy while attenuating systemic toxicity in murine xenograft models [5,6,7,8,9].

ATO and all other current arsenic drugs contain reduced forms of arsenic, typically As(III). Another common form of inorganic arsenic, arsenate (H2AsO4−/HAsO42− at neutral pH), contains As(V), and arsenate is much less systemically toxic than ATO [2]. Arsenate could conceivably be used as a reducible “prodrug”, but we have found no reports of the therapeutic use of arsenate (or any other As(V) compound).

We constructed a platform for cancer cell-targeted arsenate delivery based on the 24-subunit iron storage protein, human H-ferritin (HFn) (Fig. 1a) [10]. The ~ 8-nm hollow cavity of HFn (Fig. 1b) can be artificially loaded with up to ~ 3000 irons as an amorphous ferric oxyhydroxide polymer ([FeO(OH)]n). Our strategy, diagrammed in Fig. 2, is based on the observation that ferritin can be artificially co-loaded with ~ 1700 irons as [FeO(OH)]n and ~ 800 arsenates per 24-mer [11]. (The arsenate loading does not occur without iron.) In iron-only loaded ferritins the [FeO(OH)]n can be mobilized by reduction to Fe2+ [10], and, based on redox potentials, arsenate is easier to reduce than [FeO(OH)]n [12, 13]. The hypoxic environment of many tumors could, therefore, promote reduction of arsenate to the more toxic ATO. Another potential arsenic release mechanism stems from the observation that intravenously injected human heavy chain ferritin (HFn) is taken up by lysosomes [14], within which the acidic pH and proteolytic activity are known to promote HFn degradation and to facilitate iron release [15].

Fig. 1
figure 1

Structural features of Hfn. a Outer surface with the 24 identical subunits distinguished by color and magenta color highlighting one of six sets of four C-termini surrounding a fourfold rotation axis. b Cross section of the 24-mer showing the ~ 8-nm internal cavity. Drawings used coordinates from Protein Data Bank entry 3ajo [16]

Full size image
Fig. 2
figure 2

Arsenic and iron prodrug delivery strategy

Full size image

HFn has been used for delivery of drugs and imaging agents to tumors [17]. When injected intravenously, HFn was shown to localize to mouse tumor xenografts from commercial cancer cell lines [14]. This localization was attributed to the enhanced permeability and retention effect and to HFn’s affinity for transferrin receptor 1, Trf1 [14, 18]. Targeting of HFn to other receptors that are specifically overexpressed in particular types of cancers could conceivably increase the chances of achieving clinically effective therapies. For HFn, a simple targeting strategy has been to extend the protruding N- or C-terminal ends of the 24 identical subunits with known tumor targeting peptides [17, 19,20,21,22,23]. The epidermal growth factor receptor (EGFR) is known to be hyperexpressed in many cancers [24], including a significant percentage of triple-negative breast cancers, which are particularly aggressive [25,26,27]. GE11 designates a 12-residue peptide, which, when attached to synthetic nanoparticles, was shown to target but not activate EGFR and to induce uptake of the nanoparticles by endocytosis [28,29,30]. For binding to EGFR, the GE11 peptide has typically been attached to the nanoparticle via its N-terminal end. As diagrammed in Fig. 2, we, therefore, extended the GE11 peptide from the protruding C-terminal ends of the Hfn subunits (Fig. 1a) and tested the Hfn-GE11’s ability to deliver arsenic and iron to two breast cancer cell lines.

Materials and methods

Reagents and general procedures

Reagents and buffers were of the highest grade commercially available. All reagents, protein, and media solutions were prepared using water purified with a Millipore ultrapurification system to a resistivity of 18 MΩ to minimize trace metal ion contamination. Unless otherwise stated, all chemicals were purchased from Sigma-Aldrich, LLC, Saint Louis, MO, USA.

HFn expression plasmids

Expression plasmids containing genes encoding either HFn or HFn-GE11 were synthesized by GenScript, Inc. (Piscataway, NJ, USA). The encoding nucleotide sequences were inserted into the 5′NdeI and 3′BamHI restriction sites of the E. coli expression plasmid, pT7-7 [31]. The HFn-GE11 plasmid encoded the amino acid sequence: MTTASTSQVRQNYHQDSEAAINRQINLELYASYVYLSMSYYFDRDDVALKNFAKYFLHQSHEEREHAEKLMKLQNQRGGIFLQDIKKPDCDDWESGLNAMECALHLEKNVNQSLLELHKLATDKNDPHLCDFIETHYLNEQVKAIKELGDHVTNLRKMGAPESGLAEYLFDKHTLGDSDNESGGGSGGGTGGGSGGGYHWYGYTPQNVI, listed as the N- to C-terminal HFn sequence followed by a 15-residue glycine-rich spacer [19] (italicized) connected to the GE11 sequence (bolded).

Protein overexpression, purification, and characterization

Escherichia coli BL21(DE3) competent cells (Invitrogen) were transformed with either the HFn- or HFn-GE11-encoding plasmids. The proteins were expressed from 1 L cultures of these transformed strains in Luria–Bertani broth containing 100 mg/L ampicillin (LB/amp) at 37 °C. When the OD600 of the 1 L cultures reached 0.6–0.8, 1 mL of 100 mg/mL of isopropyl-beta-d-thiogalactoside was added to induce protein expression. These 1 L cultures were incubated for an additional 4 h with shaking at 37 °C. Cells were then harvested by centrifugation at 4 °C and frozen at − 80 °C.

For isolation of HFn, the thawed cell pellet was resuspended in 25 mL of 50 mM 3-(N-morpholino)propanesulfonic acid (MOPS), pH 7.3, containing 250 mM NaCl, and 2 mM β-mercaptoethanol, pH 7.3. The resuspended cells were lysed on ice by sonication and cellular debris was removed by centrifugation. The supernatant was heated at 60 °C for 15 min, followed by 30 min centrifugation at 20,000g to remove precipitate. The supernatant was loaded on S200 Sephacryl® gel filtration XK16/100 column (GE HealthCare Life Sciences) that had been pre-equilibrated with 50 mM MOPS, pH 7.3, 250 mM NaCl, 2 mM β-mercaptoethanol. Protein content of eluted fractions was assessed by glycine sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). The HFn-containing fractions were pooled and concentrated by buffer exchange into 50 mM MOPS, pH 7, and stored at − 80 °C.

For isolation of HFn-GE11 after cell lysis and centrifugation, the supernatant was treated with 50 µg/mL of DNAse and RNAse and 5 mM MgSO4 at room temperature for 30 min, filtered through a 0.22-µm pore membrane, then loaded into a 5-mL HiTrap Q anion-exchange column (GE Healthcare Life Sciences) equilibrated with 50 mM MOPS, pH 7.3. The column was washed with 50 mL of 50 mM MOPS, pH 7.3, and the protein was eluted with 200 mL of a two-step linear gradient of 0–0.2 M NaCl and 0.2–1 M NaCl in the same buffer. Protein content of eluted fractions was assessed by SDS-PAGE, and fractions containing HFn-GE11 were pooled and concentrated in a 100 K molecular weight cut-off Amicon® centrifugal filter (Millipore). The concentrated protein was loaded onto a S200 Sephacryl® XK16/100 size exclusion column (GE HealthCare Life Sciences) that had been pre-equilibrated with 50 mM MOPS, pH 7.3, 250 mM NaCl, 2 mM 2-mercaptoethanol and eluted at 0.5 mL/min with the same buffer. Protein content of the eluted fractions was assessed by SDS-PAGE. The appropriate fractions were pooled and concentrated by buffer exchange into 50 mM MOPS, pH 7.3, and stored at − 80 °C.

Oligomeric structure

Hydrodynamic radius of HFn and HFn-GE11 was determined by dynamic light scattering (DLS) using a Zetasizer Nano ZS (Malvern Instruments). The oligomeric structure of AsFe–HFn-GE11 was characterized by transmission electron microscopy (TEM). A small aliquot of a solution containing AsFe–HFn-GE11 (50 nM in 24-mer) in 50 mM MOPS, pH 7.4, was deposited onto an ultrathin holey carbon-coated copper grid (Ted Pella) and negatively stained with uranyl acetate. TEM was performed on a JEOL-2010F microscope operating at 200 kV. The image shown in Fig. 3 was obtained using a primary magnification of 150,000×.

Fig. 3
figure 3

Time and ascorbate concentration dependence of iron release (a) and arsenic release (b) from AsFe–Hfn-GE11. Experiments were conducted at 37 °C in anaerobic Leibovitz’s L-15 medium

Full size image

Iron and arsenic loading of HFn and HFn-GE11

The as-isolated HFn and HFn-GE11 reproducibly contained less than 10 irons/24mer and is referred to as “empty”. Iron-loaded HFn-GE11 (Fe–HFn-GE11) was prepared following a published procedure [32], which reproducibly resulted in an iron content of ~ 2500 irons per HFn 24-mer. The AsFe–HFn-GE11 was prepared using a procedure similar to that described for equine ferritin [11]. All manipulations were performed at room temperature. A 0.02 mM ferrous ammonium sulfate solution was freshly prepared in deoxygenated ultrapure water under a N2 atmosphere. This stock solution was used to add Fe2+ to an aerobic solution of 1 μM HFn-GE11 24-mer in 50 mM MOPS, 150 mM NaCl, 2 mM Na2AsO4, pH 7.4. Fe2+ was added sequentially in 200 equivalents/24-mer and incubated at room temperature for 15–20 min between additions (to allow iron to be incorporated and oxidized in the protein). Any precipitation outside the protein was removed by centrifugation (5500×g for 10 min). This process was repeated to achieve a total mol ratio of 2000 Fe2+/HFn or HFn-GE11 24-mer in a final volume of 5 mL. Excess reagents were removed by passing the protein solution over a desalting column equilibrated with 150 mM NaCl. The iron- and arsenic-to-protein mol ratios were determined by the quantification of iron and arsenic using inductively coupled plasma-atomic emission spectroscopy (ICP-OES) and of protein using the bicinchoninic acid colorimetric assay (Pierce™ 660 nm).

Reductive iron/arsenic release

All experiments were conducted at 37 °C with no attempt to exclude air. A series of stock solutions was prepared by diluting 1:10 vol% of concentrated AsFe–HFn-GE11 in 150 mM NaCl into one of Leibovitz’s L-15 medium (Gibco® L-15, ThermoFisher Scientific), 50 mM MOPS, pH 7.4, 50 mM 2-(N-morpholino)ethanesulfonic acid, pH 6.5, or 50 mM sodium acetate (NaAc), pH 5. After dilution all solutions contained AsFe–HFn-GE11 at 0.1 µM 24-mer. The measurements of iron and arsenic release were initiated by adding small volumes of a concentrated stock sodium ascorbate solution to 1 mL portions of HFn-GE11 stock solutions to achieve concentrations of 0, 0.2, 2, or 20 mM sodium ascorbate. The sample vials were placed in a shaking incubator at 37 °C for up to 24 h. At 2, 4, 6, 18, or 24 h after the additions of ascorbate, vials were removed from the incubator, and the protein solution was passed through a 30 kDa molecular weight cutoff centrifugal filter. The iron and arsenic in the flow-through was analyzed by ICP-OES. This procedure was repeated for separate 1 mL portions of each stock protein solution for each time point listed in the results.

Cell lines

Human mammary cancer cell lines, MDA-MB-231 (ATCC® HTB-26™) and MCF-7 (ATCC® HTB-22™), were purchased from the American Type Culture Collection. The MDA-MB-231 cells were cultured in Leibovitz’s medium (Gibco® L-15, ThermoFisher Scientific) supplemented with 10% fetal bovine serum (Gibco) at 37 °C in an aerobic atmosphere incubator. MCF-7 was cultured in Minimum Essential Medium Eagle (MEME, ATTC 30-2003) without l-glutamine supplemented with 10% fetal bovine serum (Gibco), 2.0 mM Glutamax, 1.0 mM sodium pyruvate (Gibco), 0.1 mM non-essential amino acids (Gibco), 1.5 g/L sodium bicarbonate, at 37 °C in a 5% CO2 atmosphere. Media used for confocal microscopy or cell viability assay was supplemented with penicillin–streptomycin (Pen Strep, Gibco) (100 µg/mL of penicillin, 100 µg/mL of streptomycin).

Fluorescent dye labeling of HFn and HFn-GE11

Either empty or AsFe–HFn-GE11 (0.5 µM 24-mer in 1 mL) in 50 mM MOPS, 250 mM NaCl, pH 7.4, was reacted with either Alexa Fluor™ 488 C5 maleimide (ThermoFisher) or Cy5® maleimide (Lumiprobe GmbH) at 20 mol equiv dye/HFn monomer at room temperature for 4 h as per the manufacturer’s instructions. The HFn subunit contains three Cys residues that could potentially react with the maleimide dyes. Unbound fluorophore was removed by multiple passages over Hi-Trap desalting columns. Conjugated dye per HFn 24-mer was quantified using absorbance spectroscopy and the manufacturer’s published extinction coefficients to be reproducibly 20–22 dyes/24-mer.

Confocal fluorescence microscopy

The binding of either empty HFn-GE11 or AsFe–HFn-GE11 to live MDA-MB-231 cells were visualized on a Zeiss 710 confocal microscope in the University of Texas at San Antonio Biophotonics core. The C32 cells were grown in Nunc™ glass bottom Petri dishes as described above (106 cells per dish). The cells were treated with either Cy5-labeled empty HFn-GE11 or Cy5-labeled AsFe–HFn-GE11 to achieve a concentration of 100 nM 24-mer and incubated for various times between 0 and 12 h at 37 C under 5% CO2. The Fe2+ imaging dye, Rhonox-1, was synthesized and applied to the cells according to the published methods [33], followed by treatment with immunostaining fluorescent probes (NucBlue® Live Cell Stain (Hoechst 33258) for nuclear membrane, and LysotrackDND99™ (ThermoFisher) for lysosomes, and CellMask Red™ (ThermoFisher) for plasma membrane according to the manufacturer’s guidelines. After staining, the cells were washed three times with DPBS, followed by the addition of Live Cell Imaging media (ThermoFisher). All dishes were imaged on the Zeiss 710 microscope using a 63 × objective.

EC50

MDA-MB-231 or MCF-7 cells were trypsinized, divided into 100,000 cells/sample, incubated for 1 h with various concentrations of Alexafluor 488-labeled HFn or HFn-GE11 and then incubated for another 20 min with propidium iodide for the quantification of cell viability. Cells were washed three times with cold phosphate-buffered saline (PBS) pelleted by centrifugation and resuspended by vortexing in Dulbecco's phosphate-buffered saline (DPBS) with 0.1% Triton X-100. For each sample, 20,000 cells were counted and analyzed using the LSR-II flow unit in the University of Texas at San Antonio Flow Cytometry Core Facility. All measurements were done in triplicate. Compensation correction, gating, and statistical data were derived using the FlowJo software. The EC50 values were determined by fitting a single dose–response function available in Origin (OriginLab®) to the data.

Cytotoxicities

Cell viabilities were assayed by a standard fluorescence method. Approximately 20,000 cells were seeded and cultured in the growth medium and atmosphere specified above for each cell line in black-sided, optical clear bottom, 96-well microplates. After 12 h, the medium in each plate was replaced with 100 µL fresh Pen Strep-supplemented medium containing various amounts of HFn-GE11, Fe–HFn-GE11 or AsFe–HFn-GE11 to achieve final concentrations between 0 and 100 nM 24-mer. The plates were placed in 37 °C incubators under either aerobic (MDA-MB-231) or 5% CO2 atmospheres (MCF-7) for 48 h. The medium was then decanted, and cells were washed twice with Hank’s balanced salt solution. Cell viability was measured using the CyQuant NF® assay (Life Technologies™). The fluorescent substrate was added as per the manufacturer’s instructions, the plate was incubated for 30 min at 37 °C, and the fluorescence emission measured at 520 nm (excitation at 480 nm) on a SpectraMax® plate reader. All experiments were performed in multiples of eight. Parallel control experiments were conducted by treating the cells with ATO, sodium arsenate, or ferrous ammonium sulfate in place of the AsFe–HFn-GE11. IC50 values were determined by fitting of a single dose–response function to the data in Origin®. Significant differences in the fitted IC50 values were assessed using the ANOVA test (p < 0.05).

Results

As shown schematically in Fig. 2, we extended the C-terminal end of the HFn subunit with GE11 (HFn-GE11). The magenta-colored portion at the center of Fig. 1a shows one of the six locations where we expect four GE11 peptides to protrude from the outer surface of HFn-GE11 [22, 23, 34]. We expressed and isolated HFn and HFn-GE11 using E. coli expression plasmids and standard purification protocols, then loaded the HFn and HFn-GE11 with arsenate and iron to the published levels [11]. We refer to the arsenate/[FeO(OH)]n-loaded protein as AsFe–HFn-GE11. For comparison, we used the non-loaded HFn-GE11(“empty”) and iron-only-loaded protein (Fe–HFn-GE11).

DLS (Fig. S1A) showed that the HFn-GE11 formed a stable monodisperse 24-mer with a slightly larger diameter (16 nm) than that of HFn (13 nm), as expected if the 24 linker-GE11 peptides were protruding from the outer surface of the protein shell. Transmission electron microscopy (Fig. S1B) showed that the HFn-GE11 retained the characteristic spherical shape of ferritin. We were able to load HFn-GE11 with arsenate and iron to the published levels [11] (700–800 arsenic and 1600–1700 iron per 24-mer, quantified by ICP-OES).

Under anaerobic conditions, ascorbate induced little or no iron release from horse spleen ferritin [35, 36]. This observation led us to test the possibility that ascorbate could differentiate the time courses of reductive iron and arsenic release from AsFe–Hfn-GE11. For these tests, we used anaerobic Leibovitz’s L-15 medium, which contains biological factors, such as ~ 2 mM phosphate, which could conceivably displace arsenate, and ~ 1 mM L-cysteine, a potential reducing agent. Figure 3a, b shows that at either 0 or 0.2 mM ascorbate, the latter of which is the normal maximum blood concentration [37], less than 50% of the arsenic and little or no iron was released after 24 h, a time interval within which, according to published work, HFn can localize to mouse xenograft tumors after intravenous injection [14]. Release of iron as Fe2+ from AsFe–HFn-GE11 became evident only at or above 2 mM ascorbate; full release of both iron and arsenic required 20 mM ascorbate and at least 18 h. We found that in the presence of 20 mM ascorbate both arsenic and iron were released more rapidly at lower pHs (Fig. S2). As described below, this pH dependence may be an advantage for intracellular iron and arsenic release.

Flow cytometry results (Fig. 4) showed that HFn-GE11 (EC50 = 47 nM) had an approximately sixfold higher affinity than HFn (EC50 = 280 nM) for the EGFR-hyperexpressing triple-negative breast cancer cell line, MDA-MB-231. Live cell confocal fluorescence microscopy (LCCFM) (Fig. 5) showed that both HFn-GE11 and AsFe–HFn-GE11 were endocytosed by MDA-MB-231 and localized to the lysosomes within 1 h after addition of either protein. Release of iron was observable using an Fe2+-specific imaging dye [33] within 1 h of exposure of cells to the arsenic, iron-loaded protein, whereas no intracellular Fe2+ was observable over the same time period upon treatment with the empty protein.

Fig. 4
figure 4

Flow cytometry for binding of HFn and HFn-GE11 to MDA-MBA-231 cells. Data points are the average of at least three determinations with error bars representing standard deviations

Full size image
Fig. 5
figure 5

LCCFM of MDA-MB-231 cells ~ 1 h after addition of Cy5-labeled empty or AsFe–HFn-GE11 to a concentration of approximately ~ 1 μM. Column headings indicate organelle, Cy5-labeled Hfn-GE11, or Fe2+ fluorescence detection. ‘Co-’ prefix indicates overlay of the indicated pairs of images, demonstrating co-localization

Full size image

LCCFM images at either 2 or 12 h after treatment with AsFe–HFn-GE11 (Fig. 6) show increasing amounts of Fe2+ in the cytoplasm. This observation indicates that the iron in the AsFe–HFn-GE11 core can be reduced after AsFe–HFn-GE11 enters the lysosomes and can then diffuse out. Based on the relative iron and arsenic release rates from AsFe–HFn-GE11 in Fig. 3a vs Fig. 3b, we expect even faster intracellular release of arsenic.

Fig. 6
figure 6

LCCFM of MDA-MB-231 either 2 or 12 h after addition of AsFe–HFn-GE11 to a concentration of ~ 1 μM

Full size image

MDA-MD-23 cell viabilities are shown in Fig. 7 and corresponding IC50 values are listed in Table 1. The IC50 of 10 nM 24-mer for AsFe–HFn-GE11 corresponds to an added arsenic concentration of 8 μM, which is similar to the IC50 we obtained for ATO (~ 5 μM As, Fig. 7b; Table 1) under the same conditions and also to values reported in the literature for ATO on the same cell line (2–7 μM As) [5, 38]. Under similar conditions MDA-MB-231 showed little or no sensitivity to the empty HFn-GE11 or to Fe–HFn-GE11 up to 1000 nM 24-mer. The arsenic cargo is thus required for effective cytotoxicity. Moreover, we observed a > eightfold lower cytotoxicity for added ‘free’ inorganic arsenate compared to AsFe–Hfn on an [As] basis (Fig. 7; Table 1). These results indicate a more efficient uptake and/or intracellular reduction of the arsenate prodrug when delivered by the HFn-GE11 cage.

Fig. 7
figure 7

MDA-MB-231 cell viabilities 48 h after addition of various concentrations of a empty, Fe–Hfn-GE11, or AsFe–HFn-GE11 or b ferrous ammonium sulfate (FAS), ATO, or sodium arsenate. Data points are the average of eight determinations with error bars representing standard deviations

Full size image
Table 1 IC50 values for empty Hfn-GE11, Fe–Hfn-GE11, AsFe–Hfn-GE11, ATO or arsenate
Full size table

For the relatively low EGFR-expressing breast cancer cell line, MCF-7, we found that HFn-GE11 bound with an EC50 = 92 nM 24-mer, which was threefold lower than IC50 for the parent HFn (IC50 = 28 nM) (Fig. S3A). The GE11 peptide thus appeared to decrease the affinity of Hfn for MCF-7, possibly by hindering interaction with the Trf1 receptor. MCF-7 showed an order of magnitude lower sensitivity to AsFe–HFn-GE11 (IC50 = 106 nM 24-mer, ~ 85 μM As; Fig. S3B) than did MDA-MB-231. Under our culture conditions, MCF-7 also showed a lower sensitivity to ATO (IC50 = 43 μM As; Fig. S3C) compared to that of MDA-MB-231. Table 1 summarizes the cytotoxicity results for the two cell lines.

Discussion

Our results demonstrate proof-of-concept for a novel approach to targeted cancer cell killing according to the mechanism illustrated in Fig. 2a. The presence of the GE11 peptide on Hfn led to the expected higher affinity for the high EGFR-expressing MDA-MB-231 cell line (Fig. 4) than for the relatively low EGFR-expressing MCF-7 cell line (Fig. S3). As expected, the HFn-GE11 localized preferentially into lysosomes, from which we observed released Fe2+ radiating outward through the cell (Figs. 5, 6). The arsenate/ferric oxyhydroxide “prodrug” cargo within the HFn-GE11 protein shell is thus endocytosed prior to its reductive release of Fe2+. Based on the relative reductive iron vs arsenic release rates from HFn-GE11 (Figs. 3, S2), we infer faster intracellular release and distribution of ATO. This reductive arsenic release scenario is further supported by the striking similarities of the arsenic IC50 values for AsFe–Hfn-GE11 and free ATO on both breast cancer cell lines, and the much higher IC50 values for ‘free’ inorganic arsenate (Figs. 7, S3; Table 1). (Unfortunately, there appear to be no reliable methods for LCCFM imaging of intracellular arsenate or ATO.) These observations together with the rapid uptake of Hfn-GE11 by the cells (Fig. 5) are not consistent with a cytotoxicity mechanism in which (1) arsenate is released from AsFe–Hfn-GE11 outside the cell, (2) the released extracellular arsenate is reduced to ATO, and (3) the ATO is then taken up by the cells. Arsenate loading into Hfn-GE11 was required for effective cell killing. The iron-only-loaded Hfn-GE11 was not detectably cytotoxic up to 1 μM 24-mer (Fig. 7a; Table 1). This “Trojan horse” approach to intracellular delivery of arsenate and reductive release as ATO could conceivably be adapted for the treatment of other cancers with targetable cell or vascular receptors [8, 39]. Hfn has proven to be a versatile scaffold for the delivery of various drugs and imaging cargo to cells [17]. We can thus, envision expansion of our reductive delivery of ATO to anticancer combination therapies using Hfn [3, 40].