Current status and future prospects of nanomedicine for arsenic trioxide delivery to solid tumors

Despite having a rich history as a poison, arsenic and its compounds have also gained a great reputation as promising anticancer drugs. As a pioneer, arsenic trioxide has been approved for the treatment of acute promyelocytic leukemia. Many in vitro studies suggested that arsenic trioxide could also be used in the treatment of solid tumors. However, the transition from bench to bedside turned out to be challenging, especially in terms of the drug bioavailability and concentration reaching tumor tissues. To address these issues, nanomedicine tools have been proposed. As nanocarriers of arsenic trioxide, various materials have been examined including liposomes, polymer, and inorganic nanoparticles, and many other materials. This review gives an overview of the existing strategies of delivery of arsenic trioxide in cancer treatment with a focus on the drug encapsulation approaches and medicinal impact in the treatment of solid tumors. It focuses on the progress in the last years and gives an outlook and suggestions for further improvements including theragnostic approaches and targeted delivery.

studies, side effects attributable to ATO lead to Grade 5 events 31 and treatment 29 respectively study discontinuation 37 in patients with solid tumors.
Well-known adverse effects of ATO, not only in patients with solid tumors but in APL patients as well, are QTc prolongation, 15,38 dermatological conditions like rashes or hyperkeratosis, 7,29 neurotoxicity, 15 and transaminase elevation. 14,15 What is more, the carcinogenic potential of arsenic compounds has been pointed out (see Martinez et al. 39 for a review) and carcinogenicity of ATO is considered an "important potential risk" by the European Medicines Agency (EMA). 12 The poor clinical outcome in solid tumors stands in contrast to the antiproliferative, proapoptotic effect of ATO in many solid cancers in preclinical in vitro and in vivo models. For this circumstance, different explanations are conceivable. First, as APL is a hematologic malignancy, intravenously administered ATO is located where it needs to act: in the blood. It does not need to accumulate at a specific tumor site nor pass the blood-brain barrier (BBB), as it does when acting on a solid (brain) tumor. Therefore, insufficient concentrations of ATO reaching the tumor site are considered the main obstacle in the treatment of solid tumors. 40 Second, a priori or acquired resistance towards ATO has been described in APL patients [41][42][43] and seen in solid tumor cell lines 44,45 and is likewise imaginable in solid tumors in the clinic.

| Nanoparticles as an approach to overcome the shortcomings of ATO in solid tumors
In recent years, the advent of nanoparticles as novel drug delivery systems (DDSs) has offered new possibilities for improved delivery of chemotherapeutic drugs, for example, by increasing their bioavailability, decreasing their effects on healthy tissue, or enhancing their uptake by tumor cells (see Sun et al. 46 for a review). Utilizing DDSs for ATO delivery has been proposed as an efficient tool to eliminate some of the drawback of ATO use in therapy, such as (i) rapid clearance of ATO and its products from the blood, 47 and (ii) low specificity. Due to rapid clearance, a therapeutic dose of ATO is not reaching the tumor sites and a simple dosage increase of ATO is not feasible due to its systemic toxicity. However, utilizing DDSs offers an attractive approach to foster the antitumor effects of ATO and to possibly overcome the limitation of the insufficient enrichment at the tumor side while reducing its adverse effects. Moreover, nanotechnology offers the possibility of tailoring the DDDs to target different types of solid cancers specifically, for instance, by attaching specific targeting ligands to the carrier surface.
Different strategies of ATO delivery to solid tumor entities have been examined over the past several years.
They differ regarding the encapsulation strategies, the kind of carrier material used and the type of tumor the ATO-formulation aim to target. In this review, we focus on the newest development over the last few years and highlight some of the older studies, which were reviewed previously. 48 2 | STRATEGIES OF ATO ENCAPSULATION ATO (As 2 O 3 ) is an amphoteric oxide (i.e., a compound able to react both as a base and as an acid) and its aqueous solutions are weakly acidic (H 3 AsO 3 ). ATO dissolves readily in alkaline solutions and forms arsenites with the following pKa values: H 2 AsO 3 − (pK A1 = 9.22), HAsO 3 2− (pK A2 = 12.10), and AsO 3 3− (pK A3 = 13.40). 49 Tables 1 and 2 and discussed in detail in the following chapter.

| MATERIALS USED AS DDSs OF ATO
DDSs can be classified based on the type of material which forms the nanocarrier as organic, inorganic and hybrid.
Each of these groups has its advantages and disadvantages. For instance, liposomes often feature a low drug loading capacity and instability during storage. Meanwhile, mesoporous silica nanoparticles (MSNs) can load the drug efficiently due to the porous structure and extremely high specific surface area, but often exhibit a high burst drug release during systemic circulation. The pros and cons of each material class are shown and discussed in the text below.

| Organic
DDSs of ATO based on organic materials are summarized in Table 1 and include liposomes, proteins, dendrimers, and polymer nanoparticles.

SÖNKSEN ET AL.
| 381 liposome. During the reaction, protons are released, which react with acetate ions to form acetic acid. Consequently, the weak acid diffuses out of the liposome in exchange for ATO. Both the formation of insoluble metal(II) arsenite complexes and the efflux of acetic acid facilitate the ATO uptake and entrapment in a liposome ( Figure 1).
For such systems of drug encapsulation in liposomes, the term "nanobin" (NB) was proposed. 82 For instance, it was shown that nanobin encapsulation of ATO (NB(Ni, As)) significantly improved pharmacokinetic properties of the drug and led to greater therapeutic efficacy compared with free ATO in an orthotopic model of triple-negative breast cancer. 84 In a follow-up work, the nanobins (NB(Ni, As)) were coated with a pH-sensitive polymer to enable pH-triggered drug release. 85 Nanobins were also used for co-encapsulation of arsenic and platinum drugs. 83 Liposomes can be also functionalized with various targeting ligands to enable ATO delivery to specific cells. For instance, ATO-loaded liposomes were functionalized with folate ligands and their cellular uptake and antitumor efficacy were evaluated in folate receptor (FR)-positive human nasopharyngeal epidermal carcinoma (KB) and human cervical carcinoma (HeLa) cells, as well as FR-negative human breast carcinoma (MCF-7) cells. 86 The uptake of folate functionalized ATO-loaded liposomes by KB cells was three to six times higher than that of free ATO or liposomes without the targeting ligands. Zhang et al. 87 reported on nanobins (NB(Ni, As)) functionalized with urokinase plasminogen activator antibodies to promote targeted delivery to epithelial ovarian cancer cells (in which the urokinase system is overexpressed compared to normal cells). The targeted nanobins showed a fourfold higher uptake in ovarian cancer cells in comparison with nontargeted nanobins.
In the last years, delivering ATO using liposomes has also been studied to examine whether liposomalencapsulated ATO could reduce the drug toxicity and improve the efficacy of ATO in treating human papillomavirus (HPV)-associated cancers. Wang et al. 50 showed that ATO encapsulated into liposomes in presence of Ni(II) ions induced apoptosis and reduced protein levels of HPV-E6 in HeLa cells more effectively than ATO alone. Akhtar et al. 51 altered the properties of liposomes such as size (from 100 to 400 nm) and surface charges and studied their influence on the efficiency of ATO delivery to cervical cancer cells. It was shown that neutral liposomes of 100 nm in size were the best-tested formulation, as they showed the least intrinsic cytotoxicity and the highest loading efficiency.
When Mn(II) ions are used as transitions metal to efficiently encapsulate ATO inside a liposome, drug nanocarrier suitable for magnetic resonance imaging (MRI), and thus theragnostic applications, can be prepared ( Figure 1). 52 The formation of the Mn(II) arsenite precipitate in liposomes generates magnetic susceptibility effects, which can be detected as a dark contrast on T 2 -weighted MRI. When accepted by cells, due to a low pH in endosome-lysosome, the Mn(II) arsenite complex decomposes, which results in a release of the As-drug and Mn(II) ions (i.e., a T 1 contrast agent that gives a bright signal in MRI). The convertible MRI signals (dark to bright) enable to follow not only the ATO delivery but also its release. Moreover, the liposomes were functionalized with phosphatidylserine (PS)-targeting antibodies to enable a specific binding of the nanodrug to PS-exposed glioma cells.

| Proteins
Zhou et al. 88,89 investigated albumin as a DDS for ATO. Albumin microspheres as a DDS for ATO were prepared using a chemical crosslink and solidification method and the synthesis was optimized with regard to the particle size and drug loadings. 88 In another work, ATO-loaded albumin microspheres were functionalized with a transactivating transcriptional activator peptide (i.e., a cell-penetrating peptide) and the nano drug delivery into bladder cancer cells was evaluated. 89 The results indicated that the attached peptide enhanced intracellular permeation of the nano drug by translocating microspheres across the cell membrane.

| Polymers
Nanoparticles Another polymer examined as DDS for ATO was poly(lactide-co-glycolide) (PLGA). 90  reported that the nano drug had a better inhibition and promoted greater lactate dehydrogenase release in comparison to free ATO. In vivo the ATO-NPs induced a significant decrease in the expression of DNA methyltransferases, while the expression of N-terminal-cleaved gasdermin E was upregulated. As a consequence, the nanoparticles inhibited the tumor growth more than free ATO or a control.
Lian et al. 58  Lu at al. 59 reported on a pH-responsive dendrimer based on polyamidoamine (PAMAM) as a DDS of ATO.
The surface of the nanoparticles was functionalized with an αvβ3 integrin targeting ligand to enable targeted delivery to glioma. In in vitro BBB model, the targeting ligand attachment heightened the cytotoxicity of the ATO-loaded nanoparticles, due to an increased uptake by C6 (glioma) cells. In vivo, the tumor volume of C6 glioma-bearing rats was reduced by 61.5 ± 12.3% after intravenous administration of the nano drug, and that was approximately fourfold higher than that of free ATO and twofold higher than that of the nano drug without the targeting ligands.

| Inorganic nanoparticles
As inorganic carriers, two types of materials were intensively studied-materials based on metal (or metal oxide) nanoparticles and silica nanoparticles. The overview of inorganic DDSs for ATO is given in Table 2.

| GdAsO x nanoparticles
As metal nanoparticles for ATO delivery, GdAsO x NPs were proposed. To synthetize such nanodrug, Chen et al. 60 co-precipitated As with Gd in the presence of dextran into GdAsO x NPs. It was proposed that the unloading of ATO from such nanoparticles could be triggered by endogenous phosphate ions present in the plasma and cytosol. In the release process, the arsenite ions would be exchanged by phosphate ions, and thus ATO release could be achieved. Indeed, the in vitro results showed that the nanoparticles gradually "dissolved" into fragments in a phosphate solution. In follow-up studies, the therapeutic effect of GdAsO x NPs on aggressive HCC was studied. 61,93 After administration of the ATO-NPs, arsenic accumulation within tumors was evaluated. It was found that the accumulation of the ATO-NPs was as much as 5%, which was ten times more than when only ATO was administrated. 61  tumor. 62 The nanoparticles were studied both in vitro and in vivo, and the results showed that the ATO-NPs caused severe necrosis via chemoembolization combinational therapy.

| Mesoporous silica nanoparticles
Nanoparticles formed by mesoporous silica have been extensively studied as DDSs not only for ATO. 94 MSNs are a class of inorganic porous material, which comprise open mesoporous channels with a diameter of 0.1-10 nm. Furthermore, their outer surface can be modified by attaching various molecules including targeting ligands for tumor specific drug delivery.
The high material porosity enables high drug loading. However, due to nonspecific drug-material interactions, a burst drug release is often observed. To decrease the burst release and increase ATO loading, two main strategies were reported ( Figure 4A,B). First, enhanced ATO binding via thiol 63,64 or amino functional groups 65,66 anchored on the surface of the mesoporous channels, and second-similar to the strategy for liposomes described above-an encapsulation of ATO in presence of transition metal ions to form insoluble MAsO x complexes. [67][68][69][70] To increase the ATO loading even more, the second approach was applied to hollow MSNs ( Figure 4C). [71][72][73] Thiol group and amino group functionalized MSNs Silica nanoparticles functionalized with thiol groups were used to bind ATO to develop nano drug for treating MDA-MB-231 triple-negative breast cancer (TNBC). 63 The inner and outer surfaces of MSNs were functionalized with thiol groups not only for the ATO binding, but also to conjugate targeting agents to the outer surface. As a targeting ligand, cyclic

MSNs with MAsO x complexes
To prepare silica nanoparticles with MAsO x complexes, two approaches were reported ( Figure 4B)-(i) loading presynthesized MSNs with a transition metal salt and subsequently with ATO, 67 and (ii) pre-synthesizing MAsO x nanoparticles and coating them subsequently with a shell of mesoporous silica. [68][69][70] The advantage of the first approach is that the MSNs can be combined with other nanoparticles before ATO is loaded. For instance, MSNs were combined with magnetic iron oxide nanoparticles to enable not only ATO delivery but also real-time Another theragnostic agent combining ATO delivery and MRI was reported by Zhang et al., 68 who developed MnAsO x @SiO 2 core-shell nanoparticles. In the synthesis, first manganese arsenite complexes were prepared by a co-precipitation of manganese acetate and aqueous ATO. Then tetraethyl orthosilicate was added to coat the MnAsO x nanocomplexes with a silica shell. In a subsequent step, the nanoparticles were decorated with a pH-low insertion peptide (pHLIP), which was added to target an acidic tumor microenvironment. The targeting ability was Fei et al. 74 prepared hybrid core-shell nanoparticles by coating HSNs (functionalized with amino groups) with a liposomal shell for controlled ATO release. The surface of the nanoparticles was functionalized with Arg-Gly-Asp (RGD)-ligands to enable targeted delivery. In vitro, the ATO-NPs showed good biocompatibility and low toxicity on HepG2, MCF-7, and LO2 cells. Moreover, due to the attached ligand, enhanced cellular uptake and a reduced halfmaximal inhibitory concentration (IC 50 value) of the nano drug could be detected. In addition, the targeting efficiency of the ligand functionalized ATO-NPs was also confirmed in an H22 tumor-xenograft mouse model.

| Hybrid
Hybrid materials consist of at least two constituents at the nanometer or molecular level. Commonly one of these components is inorganic and the other one organic in nature. Many of the materials discussed in the two previous chapters (organic and inorganic materials) and summarized in Tables

| Metal-organic frameworks
MOFs are porous crystalline coordination polymers. They comprise inorganic metal ions (or clusters) and organic ligands. 96 They exhibit outstanding properties including high internal surface area and chemical versatility. They have been suggested as promising materials for many different applications including gas storage, catalysis, and sensing, 97 but also drug delivery. 98 The most prevalent method described in published reports to capture drug molecules in MOFs is via noncovalent interactions. 98 In such cases, upon administration, the drug can easily diffuse from the material, and thus no control over the release is achieved. However, when administrating toxic drugs such as ATO, having a control of the drug release is crucial. Therefore, for ATO delivery, MOFs having possibilities to form a strong interaction with ATO, such as a chemical bond, were proposed and are summarized in Table 2.
F I G U R E 5 Schematic illustration of HSNs loaded with ATO and its drug release with activatable T1 imaging process inside cells enabled by released Mn(II)ions. (Figure reprinted with permission  clusters to prepare a core-shell structure, in which the core would be responsible for imaging via MRI, whilst the shell could function as DDS for treatment of ATRT. Both the imaging and therapeutic activity were demonstrated in vitro.
Another reported MOF for ATO delivery, which also displayed a prominent pH-triggered behavior, was Zn-MOF-74. 78 Zn-MOF-74 consists of Zn(II) ions and 2,5-dihydroxybenzene-1,4-dicarboxylate ligands and when desolvated, it contains a high density of vacant metal sites readily accessible for guest binding. It was shown that ATO could be successfully attached to these sites, and thus a high drug loading could be achieved. Moreover, it has been shown that the drug release, tested in a phosphate buffered saline, could be triggered by a pH change from 7.4 to 6.0. However, no additional biological studies have been reported.

| BENEFITS OF UTILIZING NANOPARTICLES FOR ATO DELIVERY
In addition to the chemical properties and encapsulation strategies of DDSs for ATO, the benefits which the ATOformulations offer are of great interest too. The most recent studies dealing with ATO-NPs can be assigned to five categories regarding the benefit(s) which the nanoparticle formulation(s) is/are supposed to yield: • Improvement of pharmacokinetics, • Targeted delivery via surface modification, • Theragnostic properties, • Enhancement of Transarterial Chemoembolization (TACE), and • Enhancement of BBB crossing.

| Improvement of pharmacokinetics
Since improvement of pharmacokinetics-such as controlled release or prolonged blood circulation half-life-is such a crucial point when it comes to nanomedicine, almost all studies evaluated dealt with this subject in one way or another.

| Controlled release of ATO
The most favored approach to achieve controlled release of ATO was to ensure pH-triggered release from the respective nanoparticle. Since acidic pH is a well-known characteristic of tumor tissue, 99 making ATO release SÖNKSEN ET AL.
| 389 pH-dependently, with higher ATO release at lower pH value, ought to provide a kind of tumor-directed ATO delivery while sparing healthy tissue. pH-dependent release was achieved mainly through pH-labile bond respectively attachment between ATO and the nanoparticle. 52,59,[69][70][71][72][73]75,76,78 Other researchers grafted pH-responsive material upon the surface of their nanoparticles to accomplish pH-dependent ATO release. 65,66,68 The degree of pH-selective release differed not only depending on the type of nanoparticle used, but on the exact composition of the respective nanoparticle.
Inorganic phosphate (Pi-)triggered ATO release was another way of obtaining controllable release. All four studies following this approach 60-62,93 used gadolinium-based nanoparticles, in which the arsenic could be exchanged by phosphate ions. Chen et al. 60 reported an outstanding ON/OFF specificity for their GdAsO x nanoparticles, with no arsenic release in the absence of Pi in vitro. Fu et al. 61 and Zhao et al. 62

attempted to introduce
Pi-triggered ATO drug-eluting beats (DEBs) for the improvement of TACE therapy (see below) for HCC. As occlusion of the hepatic artery is a key characteristic of TACE, and intracellular Pi supply is limited upon occlusion, the Pi deprivation slowed down the drug release, avoiding high plasma peak levels of arsenic within the first hours of treatment compared to ATO alone. 62,93 Of note is that none of the studies testing for disturbance of Pi levels in plasma observed lasting changes of the very same. 60,93

| Prolonged blood circulation and sustained release of ATO
In comparison to controlled release that is mediated by a defined trigger, sustained release of ATO eventually aims to prolong the circulation of ATO, allowing sufficient ATO concentrations to reach the tumor site before being metabolized and excreted. Controlled release can also lead to or be accompanied by sustained release. Zhao et al. 73 coated their pH-sensitive, ATO-containing HSNs with GSH and observed a higher retention time in blood, which they attributed to reduced interactions between the GSH-coated nanoparticles and serum proteins. Zhang et al. 68 observed that modifying their nanoparticles with pHLIP not only lead to pH-dependent release of ATO but also prolonged nanoparticle blood circulation in mice. Similar observations were made by Tao et al. 66 as well as by Xiao et al., 65 that both grafted their nanoparticles with the pH-responsive PAA. The in vivo half-life of those PAA-coated nanoparticles was significantly prolonged compared with free ATO. 65,66 Independent from pH-dependency, Lian et al. 58 achieved sustained release in vitro by camouflaging their ATOloaded SANs with RBCM. As RBCM coating reduced the macrophage uptake in vitro and showed higher antitumor effect in vivo, the authors hypothesized that RBCM-SANs could escape the clearance by the immune system, enabling more ATO to reach the tumor site. 58 Two authors used RGD-conjugated nanoparticles as a targeted delivery system (see below) and observed sustained release, namely an enhanced half-time of ATO in vivo compared to uncoated nanoparticles and free ATO. 59,74 Coating with PEG can reduce the uptake of nanoparticles by the reticuloendothelial system, nanoparticle accumulation in the liver and thereby increase the circulation lifetime. 100 By contrast, Chen et al. 60 achieved enhanced arsenic accumulation in the tumor via a different mechanism. Their Pi-triggered nanoparticles showed a 10-fold accumulation of arsenic in the tumor tissue compared to free ATO, which they ascribed to the EPR effect of nanoparticles. 60 The EPR effect was also considered a reason for enhanced uptake of ATO-NPs in the tumor tissue observed by Tao et al. 66 and Huang et al. 69 Another nanoparticle system by Chi et al. 72 lead to almost doubled arsenic uptake compared with free ATO into HCC cells, which the authors speculated might have been due to the rampant metabolism of tumor cells or easier internalization of nanoparticles via endocytosis. 72 Endocytosis was also identified as the most probable mechanism for enhanced uptake of arsenic from ATO-NPs compared with free ATO by Hu et al. 55 ; likewise, they observed a doubling of arsenic concentration.
A similar increase of arsenic accumulation could be observed by Fu et al.,93 in whose study the arsenic level in rabbit VX2 tumors (a model for human HCC) was almost three times higher under treatment with ATO-NPs compared to free ATO. Long-term accumulation was described in a study by Zhao et al.,62 who showed that with their nanoparticles used in the TACE procedure, intratumoral arsenic could be detected as long as seven days after the TACE procedure. Free ATO in turn was close to zero after the same time. 62 The enhanced uptake of arsenic into HCC cells in the study of Zhang et al. 68 showed pH-dependency, wherefore the authors ascribed the accumulation in tumor cells to the pH-triggered release properties of their nanoparticles.

| Targeted delivery via surface modification
A huge advantage of nanoparticles consists in their modifiable surface. In the past few years, several studies have shown, for instance, that chemical modification not only enabled nanoparticles to increase BBB penetration (see below), but also tuned the toxicity of nanoparticles as drug delivery vehicles. 59 Apart from general diversification of nanoparticle characteristics, surface modification of nanoparticles holds great potential in terms of targeted therapy. Attaching targeting ligands directed towards specific structures on tumor cells or the tumor microenvironment could possibly lead to an enhanced antitumor effect while sparing healthy tissue.
Recently, three studies evaluated nanoparticles modified with RGD for targeted delivery of ATO towards glioma, HCC, and TNBC cells. 59,63,74 RGD selectively binds α v β 3 integrin peptides, which are overexpressed by endothelial cells of the tumor vasculature and tumor cells. 102 Indeed, the authors showed that the tumor uptake of RGD-modified nanoparticles was higher compared to uncoated nanoparticles, which was accompanied by higher antitumor efficacy, namely lower tumor volume, larger area of tumor necrosis in vivo 59,63,74 and longer survival 59,74 compared with uncoated ATO-NPs and ATO alone. Beyond that, Fei et al. 74 confirmed that the transport of their RGD-modified nanoparticles was effectively dependent on α v β 3 integrins.
Another targeting ligand, lactobionic acid, was studied as a coating agent by Song et al. for HCC-directed ATO-NPs. Lactobionic acid is a disaccharide consisting of gluconic acid and galactose. Galactose-binding asialoglycoprotein receptor (ASGPR) is a receptor primarily expressed in the liver and not in other human tissues, therefore it constitutes an interesting target for HCC-directed drug delivery. 103 The authors showed for two different nanoparticle compositions that surface modification with lactobionic acid led to a decreased toxicity of ATO-NPs in normal hepatocytes in comparison to the toxic effect in HCC cells in vitro. 56,57 However, in vivo, only minimal reduction of tumor volume upon treatment with lactobionic acid-modified ATO-NPs could be detected compared with ATO alone. The authors predicated the advantage of lactobionic acid-modified nanoparticles in sparing the healthy tissue compared to free ATO, as confirmed by H&E staining of the liver and kidney. 57 It is of note that the preference for HCC cells is ought to be at least partly mediated by the EPR effect as ASGPR is not only expressed on HCC cells but on normal hepatocytes as well. 103 Folic acid is yet another targeting ligand that aims at a receptor which is overexpressed on the surface of various cancers and has hence been identified as an attractive target for tumor-directed therapy: the folate receptor (see Assaraf et al. 104 for a review). Chi et al. 67  liposomal ATO-NPs to target glioma cells. Their in vitro study revealed that nanoparticle binding to glioma cells was PS-dependent. 52 However, in vivo experiments of this approach are still pending.
Finally, Tao et al. 66 modified their nanoparticles with angiopep-2, a specific ligand of the lipoprotein receptorrelated protein (LRP) receptor. As Glioma and normal brain endothelial cells express LRP receptor on their surface, the authors proposed that functionalization of the nanoparticle surface with angiopep-2 could lead to increased accumulation of ATO in glioma. As a matter of fact, they verified that angiopep-2-modification led to a higher cellular uptake of nanoparticles by glioma and brain endothelial cells. The study revealed that targeted therapy with angiopep-2 was effective in vivo as it was shown by significantly decreased tumor volume, longer survival time and higher accumulation of the nanoparticles in tumor tissue.

| Theragnostic properties
Theragnostics describes the combination of therapy and diagnostics in one system. Visualization of drug-containing nanoparticles by integrating imaging agents into the nanoparticles is an attractive feature as it allows for image- In contrast to the bright T 1 -imaging contrast manganese, iron oxide displays negative enhancement in T 2 -weighted MRI. Ettlinger et al. 77 as well as Chi et al. 67 confirmed that nanoparticles with (superpara)magnetic iron oxide cores could be visualized via MRI. While the biocompatibility of magnetic iron oxide nanoparticles seems to be given, 105 further studies evaluating the in vivo distribution of ATO-NPs with iron oxide cores upon intravenous administration are pending.

| Enhancement of TACE for HCC treatment
For patients with intermediate-stage HCC, TACE has become a core treatment method. The method combines intra-arterial injection of a chemotherapeutic substance with embolization of tumor feeding vessels. 106 In a randomized trial, it could be demonstrated that TACE using drug-eluting beads (DEB-TACE) leads to a better tumor response with reduced adverse side effects compared with normal TACE. 107 HCC has been the tumor entity prevailing the most recent studies on ATO nanoparticles for drug delivery (see Tables 1 and 2). Therefore, it is only logical that certain studies focused on assessing the value of ATO-nano DEBs (ATO-NDEBs) for TACE. The studies by Fu et al. 93 and Zhao et al. 62 both focused on ATO-NDEBs from which ATO could be released in a Pi-triggered manner (see above). While Fu et al. 93 emulsified their ATO-NPs in lipiodol, which is also used for conventional TACE, Zhao et al. 62 coated their ATO-NPs with dextran. Both authors administered their ATO-NDEBs intraarterially into VX2-tumor-bearing rabbits. They observed high intratumoral arsenic accumulation (see above) and low plasma arsenic levels compared with conventional TACE, indicating that the NDEB formulation prevented the rushing out effect of ATO into the peripheral circulation. 62,93 Moreover, it was demonstrated that the liver and renal toxicity of ATO-NDEB was close to the sham group and much lower than the toxicity of conventional TACE with ATO, confirmed by H&E staining and blood levels of liver and kidney markers. 93

| Enhancement of BBB crossing
The second most prevalent tumor entity used in the evaluation of nanoparticle-based drug delivery of ATO are brain tumors, namely glioma and ATRT (see Tables 1 and 2). As mentioned before, ATO has been shown to be a potent GLI-inhibitor (see above). GLI has been firstly identified to be amplified in human malignant glioma. 108 What is more, a subgroup of ATRT is characterized by an overexpression of GLI. 109 The desire to improve the characteristics of this potentially effective drug by nanoparticle encapsulation is therefore very reasonable. When it comes to brain tumors, the BBB constitutes a limiting factor to successful treatment as most drugs cannot pass it (see Pardridge 110 for a review). This problem has been addressed by evaluating ATO-NPs for transport across the BBB. While both studies showed higher BBB penetration of their modified ATO-NPs in vitro as well as higher antitumor efficacy in vivo, 59,66 the strategies differed. Tao et al. 66 used angiopep-2 as a targeting ligand for LRP receptors, present on both glioma as well as human brain endothelial cells (see above). The competition essay showed that transport of the NPs across the in vitro BBB model veritably relied on the targeting angiopep-2.
Lu et al. 59 in turn coated their ATO-NPs with RGDyC, which is known to interact with integrin receptors expressed on the surface of neutrophils and monocytes. 111 The underlying idea was to target leukocytes in peripheral blood, stimulating phagocytosis of NPs and eventually enabling uptake into the brain across the BBB upon leukocyte recruitment. 59,111 Indeed, their ATO-NPs showed higher efficacy in vivo, but also decreased the cell viability of glioma cells in an in vitro BBB model. The leukocyte targeting therefore cannot be the only explanation for enhanced BBB uptake, which might at least partly be also attributable to the additional PEGylation the authors used. However, the exact mechanisms of RGDyC-mediated BBB crossing remain to be elucidated.

| CONCLUSION AND OUTLOOK
Evidently, the interest in evaluating nanomedicine for ATO delivery to solid tumors has emerged in the last years, especially for HCC and brain tumors. There are many aspects to consider when designing nanocarriers for ATO delivery. It is not just about the loading capacity, but also suitable carrier size, surface properties including an attachment of targeting ligands, options of triggered drug release or combination with imaging agents to form theragnostics. Encouragingly, more and more researchers have taken their nanoparticles to the in vivo stage, supposedly providing a better approximation to the efficacy of NPs than cell culture experiments. However, more data about biodistribution, in vivo safety and stability of NPs have to be gathered before ATO-NPs can be taken to the clinical stage. Given the numerous advances and attempts that have been made in the past few years, we hope that this review can provide an impetus and inspiration for future research on ATO-NPs.

ACKNOWLEDGMENT
The authors gratefully acknowledge financial support by the Else Kröner-Fresenius-Stiftung (project no. 2016_A181).