Nanoparticle architecture preserves magnetic properties during coating to enable robust multi-modal functionality

Magnetic iron oxide nanoparticles (MIONs) have established a niche as a nanomedicine platform for diagnosis and therapy, but they present a challenging surface for ligand functionalization which limits their applications. On the other hand, coating MIONs with another material such as gold to enhance these attachments introduces other complications. Incomplete coating may expose portions of the iron oxide core, or the coating process may alter their magnetic properties. We describe synthesis and characterization of iron oxide/silica/gold core-shell nanoparticles to elucidate the effects of a silica-gold coating process and its impact on the resulting performance. In particular, small angle neutron scattering reveals silica intercalates between iron oxide crystallites that form the dense core, likely preserving the magnetic properties while enabling formation of a continuous gold shell. The synthesized silica-gold-coated MIONs demonstrate magnetic heating properties consistent with the original iron oxide core, with added x-ray contrast for imaging and laser heating.

A schematic of the chemistry and coated particle structure, and summary of samples prepared and measurements conducted are provided in Fig. 1, (AuSi-MION, 3) and in Table 1, respectively.
The nanoparticles chosen for this study were citrate-stabilized dense polycrystalline core of magnetic iron oxide, previously described as JHU MIONs (1, Fig. 1) 29,48,49 . JHU MIONs were prepared by high-gravity controlled precipitation 50 with thermal aging, and stabilized with citric acid 48 . Physical structures of the samples used in the current studies were measured by dynamic light scattering (DLS), transmission electron microscopy (TEM), and small angle neutron scattering (SANS) 51 . The MIONs comprise a stable aqueous suspension of ~55 nm (polydispersity index (PDI) of ~0.17) nanoparticles containing agglomerated iron oxide crystallites (6-11 nm diameter) to form a dense polycrystalline core (Figs 2a,c and 3). As previously reported the dense polycrystalline core of JHU MIONs exhibits an inverse spinel-type structure and comprises a mixture of magnetite (Fe 3 O 4 ) and maghemite (ɣ-Fe 2 O 3 ), as determined by Mössbauer spectroscopy 29 .
For silica coating, a modified Stöber process was developed to control the thickness of the silica layer by stoichiometric addition of tetraethylorthosilicate (TEOS), and formation of the silica coating was confirmed by The most interesting observation of MION physical transformation with coating occurs following the initial addition of silica. Small angle neutron scattering (SANS) has proven to be a powerful probe of the internal structural (1 nm to 200 nm) and magnetic properties of nanostructured ensembles that are not readily apparent from bulk analysis 29 . Careful choice of the experimental conditions, i.e. sample (iron oxide nanoparticles) and solvent (H 2 O or D 2 O), enables scattering contrast variations to highlight specific features such as the magnetic scattering, or elemental compositions of individual layers 29 . Here, SANS revealed the coating evolution from a comparison among correlated fitted results for all three nanoparticle constructs (Fig. 3a). Robust solutions to model fitting of SANS data (Fig. 3a,b) were only possible when data from multiple complementary techniques, such as TEM (Fig. 2c) and DLS (Fig. 2a), were combined to identify the combination of SANS geometric models appropriate for each particle construct (See Supporting information for details) 53 . Note that SANS, DLS, and TEM provided no evidence consistent with the formation of pure silica or pure gold nanoparticles.
Fits to SANS data obtained from the uncoated JHU MIONs (Fig. 3a) proved sensitive to the dimensions of a flat ellipsoid model (Fig. 3b), which contains clusters of small iron oxide crystallites ( Fig. 2c (1)). The individual crystallites within each core were indistinguishable because their scattering length densities (SLDs) are comparable, yielding negligible contrast. TEM and SANS data measured from the Si-MION sample clearly demonstrated formation of a robust silica layer encasing the iron oxide cores ( Fig. 2c (2) and Fig. 3a), as expected. However, analysis of the SANS data demonstrated that the silica also diffused into the dense polycrystalline core to surround individual iron oxide crystallites (Fig. 2c, part (2)). Surprisingly, a geometrical construct consistent with the SANS data ( Fig. 3b) comprised clusters of iron oxide crystals that were coated by a layer of mixed iron oxide/ silica and then by a thin (~2 nm) layer of silica only. These layers are distinguishable with SANS because silica and iron oxide possess very different SLDs, thus increasing SANS sensitivity to the individual iron oxide crystallites that form the core. The AuSi-MION TEM (Fig. 2c (3)) showed a continuous, electron-dense coating covering the entire surface of each particle consistent with a distinct gold shell. Again, SANS models were consistent with elliptical geometries larger than either the precursor intermediates (Fig. 3b), indicating addition of a coating layer, though the local encasement of the iron oxide crystallites with silica was preserved. Use of the shape parameters obtained from SANS fitting enabled reconciliation of differences observed between the DLS and SANS data interpretation. When ellipsoidal objects were considered with interpreting the DLS model, namely that DLS presumes a spherical particle and is most sensitive to the median dimension of an ellipsoid, agreement between DLS and SANS ellipsoid models resulted (see Supplemental information).
Field-dependent magnetization measurements of the JHU MION constructs demonstrated that magnetization saturation (M s ) of AuSi-MIONs was reduced to about 30% of the uncoated JHU MIONs M s , when normalized to total solid content (Fig. 2b). This decrease is expected, as the gold and silica provide only a diamagnetic contribution which was not subtracted and the additional mass of the gold and silica are expected to reduce M s accordingly. An additional contribution to the decreased M s (see Supplemental Materials) originates from background contributions that cannot be properly accounted because of silica intercalation. Thus, precise comparisons of magnetization among the samples is precluded, however it is possible to extract general features from a comparison. When examining the coercivity at 5 K (see Supplemental Material), there is an initial increase from 24 kA/m for the JHU-MIONs to 32 kA/m for the Si-MIONs, which can be most readily attributed to the rigid encapsulation of the MIONs in silica. After coating with Au, however the coercivity returns to its previous value of 24 kA/m. These results suggest that the Au coating has a modest effect on the magnetic properties of the MIONs. Next, we look at the impact of the silica and gold coating on the functionality of the nanoparticles, beginning with imaging (both MRI and x-ray CT contrast). Using T 2 -weighted magnetic resonance sequences, MR contrast capabilities of the MION series (1, 2 and 3) were compared (Fig. 4a). As expected, T 2 relaxivity increased linearly with iron concentration, which is consistent with reported observations of JHU MIONs 49 . This was true for all three constructs, and again supports the magnetometry results which indicate minimal change to the magnetic  For reference, water is assigned a value of 0 HU, and clinical contrast agents typically measure >100 HUs. The JHU MIONs and Si-MIONs were typically <100 HU. Adding x-ray opacity to magnetic iron oxide nanoparticles has significant benefit for imaging-guided therapy (hyperthermia) applications because tissue concentrations required, >1 mg Fe/g tissue, often produce artifacts with magnetic resonance which is more sensitive to the magnetic moments of magnetic iron oxide nanoparticles making it difficult to reliably image tissue concentrations >0.1 mg Fe/g tissue. A magnetic iron oxide construct having both x-ray opacity and significant responsiveness to an alternating magnetic field provides significant benefit for imaging-guided magnetic hyperthermia.
The responsiveness of MIONs to magnetic fields contributes to their therapeutic potential as well as imaging. This property provides several routes to enhance therapy by enabling remote localization (conceptually illustrated in Fig. 5b) of nanoparticles in vivo through placement with external static (time-invariant) gradient magnetic fields, or by generating heat when a region containing MIONs is exposed to alternating magnetic fields (AMFs). AuSi-MIONs (purple) show potential for use in magnetic location control because they migrate when exposed to a permanent (static) magnet (Fig. 5a).
Magnetic nanoparticle-mediated hyperthermia can be used to treat deep-tissue tumors by remote activation of MIONs with AMFs, which are not attenuated by tissue 8 . On the other hand, AMF-induced power deposition in tissue can occur, generating off-target heating if large volumes are exposed to AMFs having high frequency and amplitude 8,54 . JHU MIONs demonstrated suitable heating at low-amplitude (<10 kA/m peak) AMFs (Fig. 6a), measured using methods previously reported 48,49,55 . Coating with gold had only a modest effect on the measured heating efficiency (Fig. 6a).
Gold-shell nanoparticles have demonstrated non-radiative heating potential via surface plasmon resonance when excited by a laser 35,56,57 . This mechanism of heating differs from magnetic nanoparticle-mediated heating, offering alternate therapeutic options with gold-coated MIONs. Laser heating of JHU MIONs and AuSi-MIONs was compared using a 5.5 W laparoscopic laser (780 nm) by illuminating MION-containing solutions and measuring temperature with optical fiber temperature probes (Fig. 6b). The specific absorption rates (SAR) of AuSi-MIONs (709 Wg −1 Fe) and JHU MIONs (127 Wg −1 Fe) were estimated after normalizing by iron content 55 .  To complement the in silico measurements, we conducted pilot in vivo tests. CT imaging and magnetic heating were assessed in mice bearing subcutaneous LAPC-4 human prostate tumor xenografts on their thigh (Fig. 7). Mice received intra-tumor injections of either MION or saline (negative control) solutions. CT contrast of tumors injected with AuSi-MIONs was enhanced and consistent with results obtained from gel phantoms, whereas neither the saline nor JHU MIONs provided measurable contrast (Fig. 7a). AuSi-MIONs remained in the vicinity of the tumor for up to 13 days, as determined in a fourth mouse (Fig. S2). Following imaging with CT, mice were exposed to an AMF using methods described previously 48,58 . Tumors injected with either nanoparticle construct heated by 4°C within 2-3 min (Fig. 7b) of exposure to AMF, while only minor heating due to eddy currents, occurred in the tumor injected with saline control. Following AMF therapy, mice were sacrificed and tumors were excised and processed for histology. Histological examination of tumors stained with hematoxylin and eosin  (H&E) (Fig. 8) revealed that the JHU MIONs (brown) and the AuSi-MIONs (purple) could be visualized (row I), although confirmation from more specific stains was needed. Stains for iron oxide and gold deposits in tissue, Perls' (Prussian blue) reagent (row II) and silver enhancement (row III), respectively, confirmed the presence of those species in tumor tissue samples. We note that the AuSi-MION tumor sample also stained blue with  The control shows no iron oxide or gold present. Tissues from the mouse injected with JHU MIONs show iron oxide particles in the H&E stain, iron staining (blue) with Prussian blue and no response to the silver enhancement stain. Tissues from the mouse injected with AuSi-MIONs show a dark purple color from the gold nanoparticles in the H&E stain and iron staining (blue) with Prussian blue. Dark black staining of the AuSi-MIONs by the silver enhancement stain, which only stains metallic gold or silver, can be seen in row III. Whole tumor images are composites created from separate 4x images; magnified images were obtained at 20x. Perls' reagent, perhaps because sample preparation degraded the gold coating, exposing iron oxide. On the other hand, only the sample obtained from tumor injected with AuSi-MION stained positive (black color) with silver enhancement stain (row III), demonstrating the presence of gold 59 . Further detailed characterization is needed with additional numbers of animals to confirm therapeutic potential of heating and correlate quantitative analysis with imaging and histology.
In summary, we report the synthesis and physical characterization of a magnetic iron oxide nanoparticle construct in which magnetic properties of iron oxide cores are not unduly impacted, and new functionality is added by silica then gold coating to achieve multi-modal imaging and heating capability using a single nanoparticle system. Comprehensive physico-chemical characterization with multiple techniques, including SANS, DLS, TEM, and magnetometry, confirmed a continuous gold layer and revealed unexpected silica intercalation with the dense polycrystalline MION cores. This silica intercalation promoted formation of a continuous gold coating while preserving the magnetic behavior. The sensitivity of magnetic resonance can be used to detect low concentrations of the Au-Si-MIONs in tissue, while the gold shell enables x-ray visualization of higher nanoparticle concentrations needed for therapeutic applications. Magnetic properties sufficient for remote localization with gradient magnetic fields and heating with alternating magnetic fields were demonstrated, as well as additional heating capability with laser activation. This work demonstrates for the first time a continuous gold-silica coating of magnetic iron oxide nanoparticles in which the magnetic properties are preserved sufficient to retain significant hysteresis heating capability. We also note the critical need for accurate and correct physical and magnetic characterization of nanostructured materials, free of assumptions typically encountered when coating. The identification of the key structural characteristics responsible for the robust performance of this MION formulation can be exploited for future applications and development.

Experimental Methods
We identify certain commercial equipment, instruments, or materials in this article to specify adequately the experimental procedure. In no case does such identification imply recommendation or endorsement by the National Institute of Standards and Technology, nor does it imply that the materials or equipment identified are necessarily the best available for the purpose.

Synthesis of gold colloid suspension. Aqueous sodium hydroxide (1 M, 600 mL) and aqueous THPC
(1.2 mM, 2 mL) were added to 90 mL of deionized water and stirred rapidly for ten minutes. Chloroauric acid (1 wt%, 3.4 mL) was quickly added and the solution immediately turned dark brown. The solution was stored at 4 °C.
Synthesis of AuSI-MIONs. JHU MIONs were coated with silica using a modified Stöber method 52 . JHU MIONs and 30% ammonium hydroxide were added consecutively to a solution of ethanol and water. The nanoparticle mixture was sonicated for 15 minutes followed by addition of TEOS and the flask was placed on a mechanical rocker overnight. The silica-coated particles were washed three times with ethanol by centrifugation to remove excess TEOS. The particles were redispersed in water, APTMS was added, and the solution was mixed overnight on a rocker. Amino-terminated nanoparticles were washed three times in ethanol by centrifugation. The silica surface was seeded with the gold THPC colloid suspension (See Supporting Information). The THPC precursor solution was diluted with aqueous K 2 CO 3 and sonicated for two minutes. Aqueous sodium chloride, 1 M (molL −1 ) and amino-terminated nanoparticles were added to the solution and sonication continued for two minutes. The solution was stored overnight at 4 °C. Gold seeded nanoparticles were washed once by centrifugation with aqueous K 2 CO 3 and three times using a permanent magnet. The particles were redistributed in aqueous K 2 CO 3 , followed by addition of 1% HAuCl 4 solution. The solution was mixed by vortexing for 30 min and hydroxylamine (50% in H 2 O) was added. The mixture immediately turned dark purple. The nanoparticles were washed three times with aqueous K 2 CO 3 using a permanent magnet, resuspended in aqueous K 2 CO 3 and stored at 4 °C. Table 1.

Nanoparticle characterization. A summary of samples prepared and characterization performed is provided in
Dynamic light scattering (DLS). The hydrodynamic diameter of the citrate-stabilized (uncoated) Fe 3 O 4 /γ-Fe 2 O 3 and coated (Si or Au-Si) nanoparticles was measured on a Zetasizer Nano (Malvern Instruments, Worcestershire, UK) in 1.8 mM K 2 CO 3 58 with 0.01 wt % nanoparticles. A refractive index of 1.33 (Fe 3 O 4 ) and 2.42 for DI water were used, and instrument setting was volume mode.
Transmission electron microscope (TEM). A Philips EM 420 transmission electron microscope was used to acquire images 58 . Samples were prepared using 10 μL of nanoparticle suspension in 100 μL of water placed on a carbon coated copper grid (Ted Pella, Inc., Redding, CA) 58 , and dried for 24 hrs at room temperature. Note that images shown were taken of individual nanoparticles well separated from any large clusters.
Magnetometry. Measurements were performed at 300 K from ± 5,570 kA/m (±70,000 Oe) and at 5 K from ± 5,570 kA/m (±70,000 Oe) using a superconducting quantum interference device with vibrating sample magnetometer (SQUID VSM) (Quantum Design, Inc). The 5 K data were measured after cooling either in zero applied magnetic field or in the presence of an applied magnetic field of 5.6 MA/m (70,000 Oe). Samples were loaded into Kel-F liquid capsules (LakeShore Cryogenics), and sealed with epoxy to preserve water during measurement under vacuum. Data are normalized to either total iron content, determined by ferene-s method previously described 60 , or by total solid content.

Small angle neutron scattering (SANS).
Unpolarized SANS data were acquired on the CHRNS 30 m SANS (NG7) instrument at the National Institute of Standards and Technology Center for Neutron Research (NCNR) in Gaithersburg, Maryland. Neutron wavelength was 0.84 nm in transmission. Instrument configurations enabled measurements having scattering vectors (Q) from 3 × 10 −5 to 5 × 10 −1 Å −1 using three detector settings (15 m, 4 m and 1.33 m). Samples were measured in water (H 2 O) at room temperature.
The raw 2D data obtained from the SANS experiments were reduced via the software program Igor Pro 53 in order to provide the corrected 1D data. The raw 2D data were corrected for empty cell, solvent, and background scattering (blocked beam) as well as detector non-uniformity. The data were plotted as I vs. Q with circular averaging of data at each value of scattering vector, Q. The 1D I(q) data obtained from each instrument configuration were combined into a single file spanning the three q ranges. In order to combine the data from all three q-ranges, the data sets were normalized by absolute intensity and combined using established procedures 51,53 . The result was a complete representation of the scattering from each sample over the range from Q = 3 × 10 −5 to 5 × 10 −1 Å −1 . A variety of (geometrical) model functions were used to fit to the entire scattering curve obtained from each sample with SasView (NIST, Gaithersburg, MD) 53,61 . It was immediately evident that no single geometrical model was adequate to analyze the full SANS data obtained from each sample (Fig. S1a); the complex structure of the nanoparticles required a linear combination of multiple model functions, which was accomplished using SasView (Supplemental Information). Model fitting was constrained using data from other measurements, e.g. DLS and by combining known material properties, e.g. scattering length density. Furthermore, the models were correlated between samples by using the parameters determined in the previous model (e.g., using the JHU-MIONs crystallite sizes in the Si-MIONs fits), so that subsequent models were constrained by the earlier results.
Histology and confocal microscopy. Following CT imaging and/or AMF hyperthermia therapy, mice were sacrificed and tumors were excised. Tumors were fixed for at least 48 hours in 10% formalin solution before being embedded in paraffin. The paraffin blocks were sectioned and stained with hematoxylin and eosin (H&E), Prussian blue, or silver enhancer. H&E and Prussian blue staining were performed by the Molecular & Comparative Pathobiology Histology Core at Johns Hopkins Medical Institute. The silver enhancement kit was used according to the kit instructions (BBI Solutions, Cardiff, UK). The histological sections were examined under an Eclipse 80i microscope (Nikon Instruments, Inc., Melville, NY). Whole-slice images were assembled from multiple images obtained at 4x magnification. Magnified images were obtained with a 20X objective.
Laser heating. Heating rates of JHU MIONs and AuSi-MIONs via laser excitation were compared in solution using a 5.5 W (780 nm) laparoscopic laser directed at the nanoparticle solutions. The increases in temperature were monitored using a FLIR thermal imaging camera and SARs were normalized based on iron content.
X-ray computed tomography imaging. X-ray computed tomography (CT) imaging was performed on gel samples loaded with nanoparticle concentrations ranging 0-7 mg Fe/ml. CT imaging was performed at 65 kV and 0.7 mA with a SARRP (xStrahl Ltd., Surrey, UK) system. Images were reconstructed using 1800 projections and Hounsfield units were calculated for each nanoparticle concentration with ImageJ software.
Alternating magnetic field (AMF) equipment and specific loss power (SLP) measurements. The AMF system and SLP measurements have been previously described in detail 55,[62][63][64] . SLP values were estimated from heating data using the relationship SLP = → C m dT dt / ( / ) Fe t 0 , where m Fe is the mass of iron in the sample for iron oxide-based nanoparticles; C is the heat capacity of the sample (assumed to be that of water or 4.18 J/g °C); and, Δ Δ T t / is the measured rate of temperature rise (ΔT) during the heating interval (Δt) or T n − T n−1 vs. Δt 55,64 .
Pilot in vivo studies. Four male nude mice (Hsd: Athymic Nude-Foxn1 nu , Harlan Labs, Indianapolis, IN) were used in this study. All were 5 to 7 weeks old and weighed about 20 grams prior to treatment with AMF and nanoparticles. Mice were housed in an Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC)-accredited facility in compliance with the Guide for the Care and Use of Laboratory Animals 65 , and procedures were approved by the Johns Hopkins Institutional Animal Care and Use Committee (IACUC). Male nude mice were selected for their relevance to our ongoing studies on prostate cancer therapy. Mice bearing human prostate tumor xenografts were anesthetized using an isofluorane chamber and maintained under anesthesia using a nose cone. Following a pre-injection CT image obtained at 65 kV and 0.7 mA using the Small Animal Radiation Research Platform (SARRP) 66 , study mice were injected with either JHU MIONs or AuSi-MIONs to a target iron concentration of ~5.5 mg Fe/cm 3 tumor 48,67 . Saline control intratumor injection was a comparable volume ~20 μL as used for MION injections. A second CT image was immediately acquired post-injection of the nanoparticles or saline, and the image was reconstructed using ImageJ software. For hyperthermia, equipment and methods were used as described previously 48,58,67 . The AMF system was adjusted for stable oscillation at 150 ± 5 kHz and 40 kAm −1 peak amplitude. Intratumor, rectal, and contralateral skin temperatures were monitored with fiber optic temperature probes (FISO, Inc., Quebec, Canada) 60 . Heating was conducted for 20 minutes.