Systemically delivered antibody-labelled magnetic iron oxide nanoparticles are less toxic than plain nanoparticles when activated by alternating magnetic fields

Objective: Toxicity from off-target heating with magnetic hyperthermia (MHT) is generally assumed to be understood. MHT research focuses on development of more potent heating magnetic iron oxide nanoparticles (MIONs), yet our understanding of factors that define biodistribution following systemic delivery remains limited. Preclinical development relies on mouse models, thus understanding off-target heating with MHT in mice provides critical knowledge for clinical development. Methods: 8-week old female nude mice received a single tail vein injection of bionized nanoferrite (BNF) MIONs or a counterpart labelled with a polyclonal human antibody (BNF-IgG) at 1mg, 3mg or 5mg Fe/mouse on day 1. On day 3, mice were exposed to an alternating magnetic field (AMF) having amplitude of 32, 48 or 64 kA/m at ~145 kHz for 20 minutes. 24 hours later, blood, livers and spleens were harvested and analysed. Results: Damage to livers was apparent by histology and serum liver enzymes following MHT with BNF or BNF-IgG at doses ≥3 mg Fe and AMF amplitudes ≥48 kA/m. Differences between effects with BNF vs BNF-IgG at a dose of 3 mg Fe were noted in all measures, with less damage and increased survival occurring in mice injected with BNF-IgG. Necropsies revealed severe damage to duodenum and upper small intestines, likely the immediate cause of death at the highest MHT doses. Conclusions: Results demonstrate that the MION coating affects biodistribution, which in turn determines off-target effects. Developments to improve heating capabilities of MIONs may be clinically irrelevant without better control of biodistribution.


INTRODUCTION
Magnetic iron oxide nanoparticles (MIONs) have been a focus of biomedical research and applications because their magnetic properties enable imaging and therapy [1][2][3][4]. Colloidal suspensions of iron oxide have been used as parenteral anemia therapies since the 1930s, and extensive preclinical and clinical experience has established MION biocompatibility and safety for continued development for applications in medicine [5][6][7][8][9][10].
In 2010, magnetic hyperthermia (MHT) with MIONs achieved clinical validation for treatment of glioblastoma [11]. MHT for treatment of cancer is accomplished by exposing the tumor containing MIONs to an alternating magnetic field (AMF). Depending upon the AMF frequency and amplitude, and magnetic properties (total anisotropy energy and timedependent relaxation of magnetic moments) of the MIONs, they can generate heat energy through magnetic hysteresis loss [12,13].
To realize more robust treatment efficacy, considerable effort has been devoted to developing MIONs having magnetic properties optimized for greater heating potential per unit mass of material, i.e. specific loss power (SLP) [13][14][15][16]. For MHT treatments that rely upon direct percutaneous delivery of MIONs to the tumor, higher heating capability can provide great benefit [17,18]. However, for MHT treatments that involve heating a region of tissue following systemic delivery (i.e. intravenous) the potential therapeutic advantages of higher SLP MIONs can be outweighed by the toxicity induced by overheating normal tissues and organs [19]. Systemic delivery of 'targeted' MION constructs is considered preferred for many imaging and therapy applications to achieve clinical benefits such as reduced invasiveness of procedure and treatment of occult disease [20,21]. In particular, systemic delivery of 'tumor-targeting' MIONs has potential to address imaging and treatment of disseminated or metastatic cancer, which is responsible for cancer-related mortality; however, nanoparticle targeting to the cancer site must be efficient to reduce the concentration of MIONs in off-target healthy tissues [13,19,[22][23][24][25].
As with most nanoparticle-based imaging and therapy constructs, systemic delivery yields deposition in non-target healthy tissues and organs, particularly in liver [25]. Indeed, for most nanoparticle-based therapeutics currently approved or in clinical trials, only about 1% of the injected material accumulates in the tumor(s) with the majority accumulating elsewhere [26,27], and thus defining toxicity [25,[28][29][30].
For MIONs having magnetic properties (i.e. total anisotropy energy) favoring high specific loss power with AMFs, activation by AMF following systemic delivery has potential to cause heat damage to normal organs and tissues, particularly liver and spleen. On the other hand, both liver and spleen are highly perfused with blood, providing robust (organ-specific) cooling that is generally considered to be more effective than in tumor. Furthermore, spatial control of activating AMFs can reduce the potential for damage to normal organs, but this can be particularly challenging in mouse models which represent an overall small volume. Many research groups have limited access to shaped fields, which has demonstrable mitigating effects [22,23]. It thus remains an open question whether MION constructs optimized for high loss power will manifest obvious toxicity in mouse models when activated by AMFs. Furthermore, the extent to which altering the biodistribution of nanoparticle constructs with a targeting moiety, e.g. antibody, mitigates toxicity by providing a different organ deposition profile remains untested [31].
We sought to determine the degree of acute toxicity caused by heating systemicallyadministered MIONs with AMFs by comparing MHT-related toxicities of bionized nanoferrite (BNF) plain (unconjugated) MIONs [32][33][34][35][36][37] and MIONs labeled with an antibody (BNF-IgG) [38][39][40] at varying concentration and AMF amplitude. The AMF inductor used produces a homogeneous flux density in a volume that encompassed the entire mouse. These are intended to be benchmark studies that provide a useful reference for future preclinical MHT research. We used healthy mice to ascertain conditions that may represent excessive toxicity precluding MHT treatment in order to develop appropriate MHT treatment combinations for future studies in cancer tumor-bearing mice. The antibody used for this purpose was a non-therapeutic polyclonal human (anti-)IgG which, absent a specific disease model, was appropriate for a general assessment of potential differences in toxicity arising from altered organ deposition [38][39][40]. A human anti-IgG antibody was used as a representative antibody ligand that might be conjugated to a nanoparticle for preclinical testing and development of a nanoparticle-antibody conjugate therapy in mouse models.
described [17,[32][33][34][35][36][37], as has their utility for tumor localization in vivo with antibodies [38][39][40]. Nanoparticle size and zetapotential data were provided by the manufacturer and are listed in Supplementary Materials for reference. Human polyclonal IgG was purchased (R&D Systems, Minneapolis, MN) and provided to micromod for conjugation with BNF nanoparticles to produce BNF-IgG using methods previously described [40]. BNF nanoparticles were suspended in sterile water and BNF-IgG nanoparticles were in PBS to provide biocompatible suspension.

Alternating Magnetic Field (AMF) System
The alternating magnetic field (AMF) system used in this study has been previously described [35,[41][42][43]. Briefly, it comprises three main components: (1) the inductor coil; (2) external impedance matching network, and, (3) the power supply. The power supply was a 120-kW induction heating system providing alternating current with variable frequency between 135 and 440 kHz, (PPECO, CA, USA). Stable oscillation at 140-160 kHz was achieved by adjusting capacitance in the matching network (AMF Life Systems Inc., MI, USA). As previously reported the AMF system was calibrated using a field probe (AMF Life Systems, Inc., MI, USA) and field amplitude was measured in the coil center before each trial. The induction coil itself heats because it is a conductor carrying a high current load, particularly at high amplitude. The AMF components and inductor coil were cooled using a closed-loop circulating water system maintained at 26 ± 2°C during operation. A polycarbonate cylindrical water jacket with separate temperature-controlled circulating water was placed inside the coil to provide a temperature-regulated chamber for mouse heating [42].

Particle SLP measurements
Specific loss power (SLP) measurements to estimate heating potential of the MIONs were performed according to methods previously described [13,35]. Briefly, 2 mg nanoparticle sample suspended in 1-ml suspending medium (PBS or water) was placed in 5-ml polystyrene test tubes and inserted into the sample holder within the solenoid induction coil. Temperatures were measured at 1-sec intervals with fiber optic probes (FISO, Technologies, Quebec City, Canada) and measurements from a water blank containing 1 mL of distilled water (or PBS) were also taken at each power setting, and subtracted from sample temperatures to correct for calorimeter heat capacity [13,35,44]. Samples were tested at several applied AMF amplitudes from 16 to 64 kA/m. From temperature data, the SLP can be estimated using the following expression [13,44], where m is the mass of iron in the sample, C the specific heat capacity of the sample

Mice
Seventy-eight female athymic nude mice 5-6 weeks old and weighing between 21-25 g were used (Charles River Laboratories, Frederick, USA). Mice were housed in sterilized filter-topped cages and maintained in a pathogen-free animal facility, and provided with a normal diet and water ad libitum. All animal care and experiments were conducted at an Association for Assessment and Accreditation of Laboratory Animal Care-accredited facility in Johns Hopkins University School of Medicine, using procedures approved by Johns Hopkins Animal Care and Use Committee [45]. Female nude mice were selected for their relevance to other ongoing studies.

Experimental design
At age 7-8 weeks, mice were randomly assigned into one of nine groups (N = 4-5 in each) and injected with one of PBS, BNF or BNF-IgG MIONs via tail vein. Figures 1A [19,[22][23][24]38], and AMF amplitude settings were chosen to span a range (low to high) of values generating modest to high heating determined from the measured SLP of the nanoparticles ( Figure 1C). 48 hours after injections, mice were exposed to AMF for 20 minutes under anesthesia at one of four amplitudes. Exposures to AMF with nanoparticle combinations were designed to provide statistically significant sampling in the following magnetic hyperthermia (MHT) dose categories -0 (control), low (either low MION or low AMF), medium (intermediate MION and AMF), and high (at high MION and high AMF) ( Figure 1). Mice were euthanized 24 hours after AMF exposure (i.e. 72 hr after injection), and some mice died before the study endpoint. At sacrifice, blood was collected from hearts into heparinized tubes for liver function tests and iron quantification by inductively coupled plasma mass spectrometry (ICP-MS). Livers and spleens were harvested and sectioned for histology and for iron quantification by ICP-MS, respectively. For histology, tissues were fixed in 10% formalin, paraffin embedded and stained with hematoxylin and eosin (H&E) and Perl's reagent (aka Prussian blue).

Toxicity of magnetic hyperthermia
For AMF exposure, each mouse was anesthetized with ketamine/xylazine given intraperitoneal. Three fiber optic temperature probes (FISO, Inc., Quebec, Canada) were used to measure mouse and water jacket temperatures. One probe was affixed to skin of upper abdomen, in the region near liver using surgical tape. A second probe was inserted 1 cm into the rectum (rectal). The third probe was placed on the inner surface of the water jacket to monitor sample environment temperature. No temperature probes were inserted into the body cavity to directly measure liver or spleen temperatures, in order to avoid disturbing the 24-hr endpoints (liver and spleen function assays, tissue [Fe] assessed with ICP MS) and to minimize invasiveness of procedures.
After the probes were in place, each mouse was placed on a holder constructed from a 50-mL conical centrifuge tube. The mouse in the holder was inserted into the water jacket and the mouse was positioned in the modified solenoid coil [40]. Once the mouse was in place, the AMF generator was turned on. Mice were exposed to AMF for 20 min.

Tissue harvest
500μl of terminal blood samples were collected by cardiocentesis from mice deeply anesthetized by isoflurane anesthesia. Approximately 400μl of blood was collected in BD Vacutainer tubes with polymer gel content for serum separation and used for liver function enzyme analysis. Another 100μl of blood was collected in heparin coated blood collection tubes (BD vacutainer, USA) and processed for iron concentration determination. Then the livers and spleens were harvested after sacrifice and sections were processed. For each liver, the left lateral lobe was processed for histopathology and the remaining lobes (left medial, right medial, right lateral and caudate lobes) were processed for iron concentration determination by ICP-MS. Spleens were randomly divided into two approximately equal sections (transverse plane), one each for histopathology and ICP-MS, respectively.

Histopathology and quantification
Sections were prepared from formalin fixed paraffin embedded tissues and were stained with H&E to evaluate tissue morphology and quantify damage. Perls' reagent was used to qualitatively confirm the presence and tissue distribution of ferric (Fe +3 ) ions from the injected MIONs. All stained tissue slides were digitized by scanning in an Aperio ScanScope At or CS system (Aperio, Vista CA) at 20X or 40X magnification for H&E and Perls' reagent, respectively. Grading of liver necrosis was done on H&E stained microscopic tissues. Slides were scanned at 20X magnification, and necrotic areas were defined as hepatocyte cellular architectural disruption, pale staining on H&E, and nuclear loss/lysis (pyknosis or karyorrhexis). Digital analysis of 20X scanned sections of liver were analyzed by ImageScope, in which the necrotic areas were gated and percentage of necrotic liver was calculated (necrotic area/total liver area × 100). Liver necrosis was then divided into 4 classifications: "none" = 0% area of necrotic liver, "mild" = 1-15%, "moderate" = 16-55%, "severe" = 56-75%. No animals had necrosis that encompassed >75% of the liver. Animals died during the procedure, since these animals often didn't have histologic lesions in the liver due to the extreme peracute nature of the injury. Necrotic areas and Perls' reagent positive areas in liver sections were quantified by Aperio ImageScope analysis [40] and degree of necrotic region was divided into four levels defined above.

Iron concentration measurement by ICP-MS
The tissue samples were processed and measured by ICP-MS as previously described [19]. Each sample was transferred to a 7-ml Teflon microwave digestion vessel (Savillex Corporation, MN, USA) to which was added 1ml optima-grade HNO 3 (Fisher Scientific, MD, USA). The vessel was sealed and placed into a 55-ml Teflon microwave digestion vessel (CEM Corporation, NC, USA), to which 10 ml of ultra-pure H 2 O (Millipore Corporation, MA, USA) was added. The 55 ml vessel was sealed and assembled according to the manufacturer's protocol. The assembly was then placed into a MARS5 Xpress microwave (CEM Corporation, NC, USA), where the tissue samples were digested using the following single-stage ramp-to-temperature microwave method: 15 min ramp to 130°C, with a hold of 30 min. After 2 hrs cooling, each sample was transferred to 15-ml tube and diluted 1:1000 with nitric acid to a final concentration of 10ppb/ml as an internal standard to monitor instrument drift during analysis. For every batch of 22 tissue samples, three samples of Seronorm™ Trace Elements Whole Blood (Sero AS, Billingstad, Norway) and two reagent blanks (for quality control) were digested and analyzed. The total iron content of the tissue samples was measured using an Agilent 7500ce inductively coupled plasma mass spectrometer (Agilent Technologies, CA, USA). Each measurement was blank-corrected using the average iron value of the reagent blanks, multiplied by the dilution factor, and adjusted based upon the recovery of iron from Seronorm. A nine-point calibration curve (0, 1, 5, 10, 50, 100, 500, 1000 and 2000 μg/l) was obtained. The analytical limit of detection was calculated by multiplying the standard deviation of the lowest detectable calibration standard (1μg/l) by three. For samples with values below the analytical limit of detection, one-half of the limit of detection was substituted.

Serum enzyme detection
Serum was separated from fresh clotted whole blood by centrifugation (5000 rpm for 10 min). Enzyme assays were carried out within 24 hrs of serum collection. Biochemical markers: alanine aminotransferase (ALT), aspartate aminotransferase (AST), alkaline phosphatase (ALP) and lactate dehydrogenase (LDH) activity were measured spectrophotometrically using a VetACE™ automated clinical chemistry instrument (Alfa Wassermann Inc., NJ, USA).

Acute toxicity grading
Mice were monitored after MHT treatment for general health and ranked using the following scheme: 1 -Normal: well groomed, active and overall good condition; 2 -Abnormal gait/ motility: Not well groomed, awkward gait and mildly agitated; 3 -Depressed: moves slowly, walks slowly and lethargic; 4 -Alive but unresponsive: slow to move or unresponsive when coaxed, dyspnea; and, 5 -Dead.

Liver necrosis grading
Portions of liver tissues showing necrosis after exposure to MHT with damage level were divided into five groups. 1 -None: no liver necrosis was found 0%; 2 -Mild: liver necrosis area was 1-15%; 3 -Moderate: liver necrosis area was 16-55%; 4 -Severe: liver necrosis area was 56-75%. Animals that died during the procedure were given a 5 -Acute dead, because histopathologic changes were often not apparent in the liver with this extreme peracute injury.

Statistical Analysis
Statistical analysis from 4-5 independent experiments was performed. Throughout this work we used pairwise t-test and adjusted p-value using Holm-Bonferroni method for multiple comparisons, because of the elevated data dispersion found in our ICP-MS and liver function analysis data sets. All tests were performed using the Prism 6 suite (GraphPad Software).

Physical characteristics of MIONs and SLP measurement
BNF (plain) and BNF-IgG MION physical characteristics provided by the manufacturer are shown in Table 1. BNF MIONs had a mean z-averaged hydrodynamic diameter of 97 ± 4 nm, and BNF-IgG mean diameter was 131 ± 11 nm. Other measured physical properties are provided in Table 1. SLP estimates obtained from calorimetric measurements showed no significant differences in amplitude-dependent heating between BNF and BNF-IgG MIONs confirming no alteration of magnetic properties occurred with antibody labeling Figure 1C.

Rectal and skin temperature measurements
Upon systemic, e.g. intravenous delivery, MION deposition generally occurs primarily into the liver and other organs of the mononuclear phagocytic system (MPS) following clearance from blood circulation [25,[38][39][40]. BNF MIONs can generate a large amount of heat when exposed to an AMF (>200 W/g Fe), hence heat generated at sites of accumulation can be measured at a distant location because heat is transferred by physical and biological processes (i.e. blood perfusion) [12,13,17,[34][35][36]38]. To obtain non-invasive estimates of heating in liver, we placed a surface temperature probe on the skin of the upper abdomen/ lower thorax (i.e. chest) in the vicinity of the liver of each mouse. As shown in Figure 2, measured temperature rise correlated with combinations of increased MION concentration and high AMF amplitude. Chest temperatures increased only slightly, similar to PBS + 64 kA/m (control) group, when mice were treated with low MHT conditions (Figure 2A-C). When exposed to intermediate MHT conditions with BNF, BNF at 3 mg Fe and 48 kA/m (BNF3+48 kA/m), temperatures rose to 41°C-43°C as shown in Figure 2D. We measured a steady increase in chest temperatures from ~36.5°C to ~49°C for mice treated with the same BNF dose of 3 mg Fe, but at higher AMF amplitude, BNF3+64 kA/m ( Figure 2E). Conversely, chest temperatures measured from mice treated with the same dose of BNF-IgG (3 mg Fe) and field amplitude (BNFI3+64 kA/m) increased to ~43°C. This is a mean temperature rise of 6°C less than that measured from BNF MIONs. On the other hand, there was a similar measured temperature rise, 36.4°C-51°C, for both BNF and BNF-IgG MIONs when the dose was high (5 mg Fe) with moderately high AMF (48 kA/m) ( Figure 2F).
Interestingly, measured rectal temperatures showed little difference among the groups, except when the highest concentration of MIONs and highest AMF amplitude were used (Figure S1A-E). Mice receiving BNF5+48 kA/m MHT, demonstrated a significant rise of rectal temperature to 39.19 °C -42.04 °C when injected with either BNF-IgG or BNF, respectively as shown in Figure S1F.

Survival outcomes after MHT
All mice treated with either BNF or BNF-IgG MIONs at 1 mg Fe with 64 kA/m AMF exhibited no discernible adverse effects, whereas those treated with BNF MIONs at 3mg Fe and 32 kA/m displayed mild effects, but died when AMF amplitude was 64 kA/m (   Figure 3A. To further assess the extent of damage, livers and spleens were dissected and visually inspected before fixation for histology. Livers and spleens from MION injected mice had a much darker appearance than those recovered from PBS injected control mice, due to the presence of MIONs ( Figure 3B, S2A). PBS injected mice exposed to AMF (64 kA/m) also showed no discoloration of livers or spleens, which appeared normal ( Figures 3B, S2A-B). Evidence of damage, however was noted in livers of mice injected with BNF at 3 mg Fe and exposed to either 48 or 64 kA/m (BNF 3+48 kA/m and BNF 3+64 kA/m).
Mice treated with low MHT doses (1mg Fe MION and 64 kA/m; 3 or 5mg Fe and 32 kA/m) showed no visible evidence of liver tissue damage when evaluated with H&E staining. With high MHT dose (3mg BNF with 64 kA/m or 5mg BNF/BNF-IgG with 48 kA/m) some liver sinus congestion was evident ( Figure 4). Liver damage was also observed in mice injected with BNF-IgG and BNF at 3mg Fe+48 kA/m, and in mice following treatment with BNF-IgG at 3mg Fe+64 kA/m ( Figure 4). A wide range of necrotic areas was observed in the livers of mice exposed to moderate MION dose and moderate or high AMF amplitude, specifically BNF-IgG 3mg+64 kA/m ( Figure 4). Necrosis was not observed in livers of mice treated with BNF 3mg+64 kA/m, however all mice in this cohort died during treatment or shortly thereafter, precluding development of identifiable features of heat-related tissue damage. Analysis of the degree of liver necrosis areas showed a correlation with MHT dose ( Figure 5). With low MHT the area of damage to liver was 0-15%. On the other hand, all mice died following treatment with high MHT dose, consequently no liver tissue necrosis was observed because biological processes ceased early. The most interesting effects were observed when mice were treated at MHT conditions 3mg BNF/BNF-IgG MIONs with 48 kA/m. The extent of liver tissue damage differed for mice injected with BNF vs. BNF-IgG when treated at similar field amplitudes (Table 3, Figure 5). Analysis revealed no evidence of damage to spleens for any MHT conditions or controls ( Figure S2, S3).

Quantification of hemosiderin in tissue
Overall, Prussian blue positive regions showed as blue foci in livers and spleens ( Figure  S4A). A dose dependent increase in Prussian blue positive areas was seen in livers and spleens of mice treated with both MION constructs, indicating dominant accumulation ( Figure S4B, C). For intermediate MHT dose (BNF/BNF-IgG 3mg + 48 kA/m and BNF-IgG 3mg + 64 kA/m), a wide-range of necrotic areas in livers could be identified with Prussian blue staining by prominent hemosiderin (i.e. blue stained) foci ( Figure 5A). Qualitatively, more hemosiderin foci were observed in necrotic areas than undamaged areas in livers after MHT in most cases ( Figure 5A); however, quantification of differences in number of hemosiderin foci in necrotic areas was statistically significant only for BNF-IgG 3 mg Fe and 64 kA/m ( Figure 5B). By contrast, spleens of mice injected with MIONs displayed no necrotic areas even with clear evidence of MION accumulation as measured by hemosiderin foci (Figure S4A, C). Interestingly, spleens of mice injected with BNF-IgG showed more evidence of MION accumulation than did spleens of mice injected with BNF ( Figure S4C).

Total iron recovered from blood, spleens and livers measured by ICP MS
Total iron recovered from livers and spleens was quantified by ICP-MS to support findings from histology. Blood iron was also quantified by ICP-MS to confirm MIONs had cleared from circulation ( Figure 6A). Measured total iron in livers was always significantly higher than that measured in spleens, indicating that more MIONs accumulated in the liver than in the spleen of each mouse ( Figure 6B). Concentrations of recovered iron, calculated using total mass of recovered tissues ( Figure S5) differed among individual livers and spleens (2-20 ug iron/gm dry weight tissue), depending on injected dose of MIONs ( Figure S6). We measured higher iron in spleens of mice injected with BNF-IgG than from spleens of mice injected with BNF ( Figure 6C and Figure S6C). Correspondingly, a lower concentration of iron was measured in livers of mice injected with BNF-IgG than in the livers of mice injected with BNF ( Figures 6B, S6B). Thus, we concluded the presence of the IgG antibody correlated with a different distribution of MIONs between liver and spleen with relatively increased accumulation in spleen and reduced liver accumulation. The total measured tissue mass obtained from individual mice presented no differences among the cohorts for a given tissue ( Figure S5).

Areas of necrosis in liver correlated with MION deposits
A complete analysis of overall liver damage by assessment of necrotic areas revealed a correlation with hemosiderin (Prussian blue) deposits confirming liver damage was likely due directly to MION deposits, which created 'hot spots' when combined with AMF ( Figure  7A-B). Areas of necrosis with BNF-IgG 3 mg Fe and 64 kA/m generated areas of necrosis extending more beyond localized MION deposits than for 48 kA/m AMF.

Liver function enzyme analysis revealed damage with MHT
It was evident from analysis of tissues that the liver potentially sustained clear heat induced damage from MHT. Liver function enzymes aspartate transaminase (AST), alkaline phosphatase (ALP), alanine transaminase (ALT) and lactate dehydrogenase (LDH) are well established indicators of liver damage, and revealed MHT dose-dependent elevations ( Figure  8). ALP was relatively unchanged in any of the treatment groups. On the other hand, AST, ALT and LDH, crucial indicators of liver damage, were significantly higher in mice exposed to BNF or BNF-IgG at 3mg Fe with 48 kA/m and BNF-IgG 3mg+64 kA/m treatments ( Figure 8B-D). Blood samples were not collected from mice that died during treatments at high MHT (BNF3+64 kA/m and BNF5+48 kA/m).
Recognizing that accumulation of MIONs to the tumor following systemic delivery is likely to be limited (i.e. <1 mg MION/g tumor) [37], effort has been devoted to development of MIONs exhibiting increased heating potential [3,11]. Yet without similar developments in strategies to effectively reduce deposition to off-target sites, use of higher heating materials following systemic delivery carries considerable risk to increase damage to healthy tissues and organs. While this reality is often acknowledged, studies demonstrating the utility of novel higher-heating MIONs typically ignore consequences of off-target heat-related toxicities.
To ascertain the relative differences of MHT-related toxicity, we used unlabeled starchcoated MIONs, BNF, and BNF MIONs labelled with a polyclonal IgG antibody (BNF-IgG) in healthy nude mice as a representative generalized model. The polyclonal IgG serves as a model for a generalized antibody-targeting strategy; however, this approach possesses inherent limitations to its applicability for a specific disease model. A direct comparison of plain or unlabeled MIONs with a generalized antibody-labeled MION in healthy nude mice (representing the strain often used in many cancer nanomedicine studies), can afford some insights into the nature of off-target effects with high-heating.
In this study an anti-human IgG antibody was used because this choice mimics the type of ligand most likely to be used in preclinical testing of a cancer nanomedicine. We note that both immunodeficient mice and anti-human IgG present inherent limitations to gaining understanding of fundamental (murine) immune-nanoparticle interactions that have recently become a focus in cancer nanomedicine [27,[29][30][31]40,52,53,55,57]. Furthermore, we acknowledge that murine phagocyte, e.g. macrophage, activity is likely to be triggered by cross-species activation of Fc-receptors leading to altered biodistribution that may not represent a scenario in either humans or in mice with a murine antibody. Nevertheless, preclinical testing of cancer nanomedicines is performed extensively in mouse models, which are generally immune deficient because they provide a permissive host to enable grafting of human cancers in order to test (human-relevant) active pharmaceutical ingredients in a living animal model. Despite the obvious limitations, selection of both mouse model and BNF-IgG were intended to provide a baseline of results for reference with future studies of MHT-related toxicities.
We observed no adverse health events or evidence of liver or spleen damage in mice treated with either high concentration BNF/BNF-IgG MIONs with no AMF (0 kA/m) or high amplitude AMF with no MIONs (0 mg Fe). Individually, MIONs or AMF at the doses selected produced no measures of acute toxicity using the selected assays. Long-term effects of chronic exposure or clearance of MIONs were not addressed, and remain an area to be studied further. At the chosen 72-h time point, measurements of blood iron collected from mice injected with MIONs, either BNF or BNF-IgG, were similar to PBS controls, indicating complete clearance from blood. Also, observations of general health, liver and spleen histology measuring necrosis, and liver enzyme analysis of blood were consistent with measurements from PBS controls, indicating generally favorable biocompatibility and no acute adverse effects from MION exposure without AMF. These observations are consistent with other results obtained from polysaccharide-coated MION suspensions.
Differences in deposition to liver and spleen between groups injected with either BNF or BNF-IgG were apparent in measured iron content. As expected, livers showed dosedependent and consistently higher iron accumulation than did spleens indicating the former is the preferred site of accumulation. On the other hand, measured iron from spleens revealed differences between BNF and BNF-IgG, with the latter exhibiting relatively higher accumulation in this organ and correspondingly lower relative concentrations in liver than mice injected with BNF. This difference, also noted in analysis of Prussian blue stained tissue sections, is perhaps an indication that the MION-IgG interacted more strongly with cells in the spleen.
Combined, MION concentration and AMF amplitude are responsible for heat energy deposition in tissues with MHT, and thus acute MHT-related toxicity. Given the generally higher accumulation of MIONs in livers, we expected much of the MHT-related toxicity to be apparent in the livers of mice despite the large size of the organ and blood perfusion. Low dose MHT, i.e. low MION (≤3 mg Fe) and AMF (≤32 kA/m) generated no discernible toxicities, except mild necrosis to portions of some livers observed only on examination of histology slides. On the other hand, high dose MHT, i.e. high MION (>3 mg Fe) and AMF (>48 kA/m), combinations proved generally lethal. This acute toxicity was associated with high rectal temperatures, >40°C, indicating extreme systemic heat stress. Full necropsies were performed in three mice and no obvious damage was observed in organs leaving the conclusion that death resulted from extreme and systemic heat stress leading to neurologic (and potentially other) organ failure. Nevertheless, it is known that heat-related illness can cause other adverse effects to metabolism, acid-base balance and cell permeability leaving open the possibility of other contributing causes of death [58,59].
Despite the heavy iron load in livers we found no conclusive evidence pointing to acute liver toxicity as the immediate cause of death. Rather, the evidence suggested systemic and extreme heat stress caused death. This is important because it suggests that even localized (i.e. tumor-specific) activation of high SLP MIONs can potentially overwhelm the mouse thermoregulatory system and its ability to compensate, leading to elevated core (rectal) temperature and death. At the extreme, the total heat/power deposited to the body becomes potentially lethal; however, spatial and temporal modulation of power can provide sufficient reprieve for thermoregulatory recovery [60][61][62][63]. Spatial control of AMF, as with the hyperthermia module possible with small animal magnetic particle imaging (MPI) devices represents and interesting novel modality with clinical potential [64].
Most interesting were differences in measures of toxicity between BNF and BNF-IgG at intermediate and moderately high MHT doses, i.e. 3 mg Fe @ 48 or 64 kA/m and 5 mg Fe @ 32 kA/m. Varying toxicities were evident from general appearance and health, histology, and liver enzyme panels, however at similar conditions MHT with BNF-IgG was associated with less severe symptoms than MHT with BNF. At the highest AMF amplitude and 3 mg Fe of BNF-IgG, it is notable that an inverse relationship was found between necrotic areas of liver and hemosiderin, i.e. deposits of MIONs, with the former encompassing larger areas of tissue. This is likely due to effects of heat transfer from local hotspots extending the zone of damage clearly beyond the MION locale, as would be expected with higher measured SLP at this amplitude (~350 W/g Fe vs. ~270 W/g Fe). Interestingly, even though spleens showed higher relative iron content in mice injected with BNF-IgG MIONs, no evidence of damage to this organ was observed. It is thus considerable that we note a difference in liver and spleen deposition, and MHT toxicity of MIONs that depends on the MION surface chemistry or coating/ligand properties.
Using measured MION SLP values and tissue concentrations, it is possible to estimate the total heat energy deposited in an organ or tissue from MHT [13,23]. Total energy deposited in livers and spleens was estimated using values obtained from ICP-MS, AMF and MION parameters, and displayed in Figure 9 for each treatment combination and within each organ. Correlating these results with outcome (Tables 2 and 3) then enables an estimation of maximum tolerated energy to differentiate between likely lethal and non-lethal MHT exposure ( Figure 9A). Estimated energy deposited into livers during MHT treatment was in the range ~200 to ~1000J, whereas for the spleen energies were <120J (Figure 9 A-B). With this analysis, we note for a given MION dose (≥3 mg Fe) and AMF, estimated energy deposition from BNF to the liver was consistently higher than from BNF-IgG, and that ~500J deposited to liver may represent an upper limit of tolerance in mice. While no evidence was found that acute liver toxicity was the cause of death, total energy deposition to liver may represent a viable surrogate measure of MHT-related toxicity and tolerance threshold. On the other hand, total energy deposited to spleens was significantly lower, despite the seemingly high MION concentration, thus providing a possible explanation for the absence of damage observed in this organ. These results provide a measure of an upper limit, or estimate of threshold for tolerance with MHT.

SUMMARY AND CONCLUSION
We sought to evaluate acute toxicity due to MHT following systemic delivery of MIONs to estimate the potential for damage by off-target accumulation of high-heating MIONs. A generalized antibody-labeled MION, representing an 'active' targeted moiety was compared against the unlabeled ('passive' targeted) counterpart. Toxicity was noted only when AMF and MION dose combined to produce moderate or high MHT conditions. In the latter case, very high heat loads producing systemic heat stress always proved fatal. At intermediate MHT exposures, differences in toxicity were noted between antibody-labeled and unlabeled constructs with the former being less toxic. Analysis of results suggests that, even for the same total heat/power deposition below a critical threshold, a more spatially distributed energy deposition that reduces the energy load to a single organ, i.e. liver, can be less toxic.
We conclude that with MIONs having SLP values in the range of those used in this study, damage to spleen from MHT is much less likely than is damage to liver, and that MHT toxicity can be reduced by distributing off-target MION accumulation among organs. Further study is warranted to determine optimal treatment strategies with systemic delivery of MIONs for treatment of systemic diseases.

Supplementary Material
Refer to Web version on PubMed Central for supplementary material.

DECLARATION OF INTEREST
C.G. is an employee of micromod Partikeltechnologie, GmbH the manufacturer of the nanoparticles used in this study. R.I. is an inventor on nanoparticle patents. All patents are assigned to either The Johns Hopkins University or Aduro BioTech, Inc. R.I. consults for and is on the Scientific Advisory Board for Imagion Biosystems, a company developing imaging with magnetic iron oxide nanoparticles. He also consults for Magnetic Insight, a company developing magnetic particle imaging technology. All other authors report no conflicts of interest.  A) Athymic nude female mice (7-8 weeks old) were treated with of bionized nanoferrite (BNF) or BNF-IgG nanoparticles vial tail vein injection (n= 4-5/group). After 48 hours, the mice were exposed to 20 min alternating magnetic field (AMF). The mice were euthanized 24 hours later. Blood from the heart was collected into heparinized tubes for liver function tests and iron quantification by inductively coupled plasma mass spectrometry (ICP-MS). Livers and spleens were also harvested and divided into two portions -one for histology and the other for ICP-MS. B) Mice were randomly assigned into one of nine groups with four to five mice per group. The flowchart shows magnetic hyperthermia (MHT) dose level depending on the concentration of nanoparticles and field conditions used.    Mice were sacrificed 24 hours after MHT and tissues were harvested for H&E staining. A representative sample of mouse liver tissue sections showing a wide range of necrotic tissue areas observed following treatment with either BNF 3mg + 48 kA/m, BNF-IgG 3mg + 48 kA/m or BNF-IgG 3mg + 64 kA/m. No differences are seen with low MHT dose groups, compared with controls. Mice exposed to high dose MHT died, and organs harvested less than 1hr after death revealed liver sinus congestion (white arrow). The enlarged images of liver sections show representative normal, necrotic and sinus congestion. A) Blood-ALP levels showed higher variability with BNFI3+48 kA/m and BNFI3+64 kA/m groups when compared to controls. Blood-ALP levels did not show any difference in any of the BNF treated groups when compared to that of controls. All other groups were in the normal range. B) BNF and BNF-IgG3+48 kA/m and BNFI3+64 kA/m groups showed increased AST in blood. All other groups were in the normal range. C) Measured blood-LDH levels were increased following BNF3 + 48 kA/m, BNFI3 + 48 kA/m and BNFI3 + 64 kA/m treatment combinations. D) Measured blood-ALT levels were also increased following treatment with BNF3 + 48 kA/m, BNFI3 + 48 kA/m and BNFI3 + 64 kA/m groups. An increase was also noted in BNF5 + 32 kA/m treated mice. *N.D: Blood samples were not collected due to death of animals in BNF3 + 64 kA/m, BNF5 + 48 kA/m and BNFI5 + 48 kA/m treatment groups.  A) Total energy deposited was estimated using MION SLP values from Figure 1C, mean iron recovered from organs ( Figures 6B and 6C), and AMF amplitude and treatment time (1,200 s kA/m and BNF/BNF-IgG 5mg Fe and 48 kA/m. The dashed horizontal line represents an estimated maximum tolerated total energy deposition to livers using data from Table 3. B) Results of estimates as in A), but for data obtained from spleens. Unlike for livers, iron recovery from spleens was higher for BNF-IgG than for spleens of mice injected with BNF; but, the total iron load in spleens was less thus the overall energy deposited to spleens was significantly lower than that deposited in livers.