Potential dual imaging nanoparticle: Gd2O3 nanoparticle

Gadolinium (Gd) is a unique and powerful element in chemistry and biomedicine which can be applied simultaneously to magnetic resonance imaging (MRI), X-ray computed tomography (CT), and neutron capture therapy for cancers. This multifunctionality can be maximized using gadolinium oxide (Gd2O3) nanoparticles (GNPs) because of the large amount of Gd per GNP, making both diagnosis and therapy (i.e., theragnosis) for cancers possible using only GNPs. In this study, the T1 MRI and CT dual imaging capability of GNPs is explored by synthesizing various iodine compound (IC) coated GNPs (IC-GNPs). All the IC-GNP samples showed stronger X-ray absorption and larger longitudinal water proton relaxivities (r1 = 26–38 s−1mM−1 and r2/r1 = 1.4–1.9) than the respective commercial contrast agents. In vivo T1 MR and CT images of mice were also acquired, supporting that the GNP is a potential dual imaging agent.

same atomic concentration, the number density of nanoparticles is much lower than those of molecular agents 26 . Furthermore, because of Gd's higher X-ray attenuation coefficient 35 , the injection doses of GNPs can be further reduced.
The diagnostic and therapeutic applications of GNPs are summarized in Fig. 1. Although not examined in this study, GNPs can also be applied to NCT for cancers (Fig. 1). Note that NCT has been proved to be powerful for noninvasively treating malignant brain cancers [56][57][58][59] . Two 10 B-chemicals (B 5 boron) have been developed for clinical purpose 59 . However, 157 Gd, with a ,67 times higher thermal neutron capture cross section than 10 B (19.97% natural abundance), is expected to be even more powerful than 10 B-NCT [60][61][62] . This, when combined with the imaging property of GNPs, will make GNPs a powerful theragnostic agent for cancers 11,[27][28][29] , whereas 10 B-chemicals are useful only as therapeutic agents for cancers because they have no diagnostic capability.
This study examines the dual imaging capability of GNPs. To this end, various iodine compound (IC) coated GNPs (IC-GNPs) were synthesized, and their T 1 MRI and CT functionalities were investigated in vitro and in vivo. Four commercial iodine contrast agents were used as surface coating ligands on GNPs to maximize the CT functionality of the nanoparticles because of the iodines in the ICs, as well as to make them water-soluble and biocompatible. To prove the potential of GNPs as a dual imaging agent, the water proton relaxivities, X-ray absorption, and in vitro cellular cytotoxicities were measured, and finally in vivo T 1 MR and CT images of mice were acquired after intravenous injection.

Results
Particle diameter and hydrodynamic diameter. As shown in highresolution transmission electron microscope (HRTEM) images ( Fig. 2a), the diameters of the nanoparticles range from 1 to 3 nm for all the samples. Arrows indicate individual nanoparticles. Gd and I in Sample 2 were mapped onto a high-angle annular dark fieldscanning transmission electron microscopy (HAADF-STEM) image ( Fig. 2b) to see the Gd and I distributions in the nanoparticle. As expected, a dense Gd population at the nanoparticle core and widely spread I over the nanoparticle were observed. The average particle diameters were estimated from log-normal function fits to the observed particle diameter distributions ( Fig. 2c and Table 1). The average hydrodynamic particle diameters were also estimated from log-normal function fits to the observed dynamic light scattering main biomedical applications of GNP are T 1 MRI, CT, and NCT using the magnetic, X-ray absorption, and thermal neutron capturing properties of Gd, respectively. Combined applications include dual imaging, which is considered in this study, and theragnosis of cancers.   Table 1). Very broad X-ray diffraction (XRD) patterns were observed for all the powder samples, likely owing to the ultrasmall particle diameters 63 , whereas sharp peaks corresponding to cubic Gd 2 O 3 , were observed for all the samples after thermo-gravimetric analysis (TGA) (Supplementary Information). This is due to particle growth during TGA treatment 63,64 . The estimated cell constant of the TGA treated samples was 10.81 Å , which is consistent with the literature value for cubic Gd 2 O 3 (510.813 Å ) 65 .
We carried out the long-term colloidal stability study in physiological conditions (i.e., pH 5 7.0, 1 mM glucose, 1 atm, and room temperature), and we found that the colloidal stability maintained for a week for all sample solutions. After IC-GNPs settled down, however, they could be readily re-dispersed again as stable colloids in solutions. Photographs of 1 mM Gd sample solutions showing stable colloidal dispersions in physiological conditions are provided in Fig. 2e. We also estimated the free Gd 31 ion concentrations liberated from IC-GNPs in aqueous sample solutions with 1.0 mM Gd, but the free Gd 31 concentrations were below the detection limit of the ICP-AES (i.e., ,0.1 ppm Gd) in all sample solutions for one month.
Surface coating results. Surface coating of GNPs with ICs was investigated by recording Fourier transform infrared (FT-IR) absorption spectra (Fig. 3a). The absorption peaks from ICs in the samples showed that the GNPs were successfully coated with ICs in all the samples. Overall, the absorption peaks of all the samples were broad and overlapped with neighboring peaks. The assignments of some important peaks are provided in the Supplementary Information. A sufficient surface coating of GNPs with ICs is crucial because both water-solubility and biocompatibility are necessary for biomedical applications. For example, Gd may cause nephrogenic systemic fibrosis when it is released in the body during circulation 66 . The amount of surface coating on each sample was estimated in weight percent (%) (5P) from the mass drop in its TGA curve (Fig. 3b).
Here, an initial mass drop due to moisture desorption between room temperature and ,105uC was subtracted in estimating P. Then, the grafting density (5s) 67 , which corresponds to the average number of ICs coated per unit surface area of a GNP, was estimated using the bulk density of Gd 2 O 3 (57.407 g/cm 3 ) 68 , P, and the average particle diameter (5d avg ). By multiplying the estimated s by the nanoparticle surface area (5pd avg 2 ), the average number (5N) of ICs coated per GNP was estimated using N 5 spd avg 2 . Finally, the average number (5NI) of iodines per GNP was estimated by multiplying N by the number of iodines per IC (i.e., NI 5 3N for Samples 1 and 3, and 6N for Samples 2 and 4). The estimated P, s, N, and NI values are provided in Table 1 and plotted as a function of the IC mass in Fig. 3c. ICs with a larger mass had smaller P, s, N, and NI values, indicating that fewer molecules of an IC with a larger mass were coated on each GNP. This is because a more massive IC occupies a greater volume on the GNP surface.
An IC is bonded to a GNP through its functional group, such as COOH, NH 2 , or OH. This bonding corresponds to a hard acid [Gd(III)]-hard base (functional group of IC) type of reaction [69][70][71] . The bonding strength of these functional groups to the GNP is likely to be in the order COOH . NH 2 . OH in triethylene glycol solvent. According to the FT-IR absorption spectra, the initially coated triethylene glycol on the GNPs was replaced by the ICs used in this study, indicating that the OH is the weakest bonding group among the above three functional groups. The strongly bonded COOH generally shows a red-shift with respect to free COOH, as observed in many cases 63,[72][73][74][75] . The red-shift in this study was observed to be 74-75 cm 21 in Samples 1, 2 and 3, as indicated by arrows in Figs. 3a(1) to 3a(3). Among the two functional groups COOH and NH 2 in Sample 1, the more strongly bonding COOH group was bonded to the GNPs, as indicated by the red shift of the bonded COOH in Sample 1 [ Fig. 3a(1)]. The IC in Sample 4 has many OH groups, so it is likely that any OH groups that are geometrically accessible to the GNP can be bonded to it. This is likely an entropy-driven replacement reaction of the initially coated triethylene glycol on GNPs with iodixanol, which has many OH groups. Because of the geometrical difficulty of both COOH groups in Samples 1 and 2 accessing the GNP, only one of them is likely bonded to the GNP. Figure 4 shows the most probable bonding structures between the ICs and GNPs in the four samples according to these results. The elemental analyses of surface coated ICs on GNPs were also studied using both the elemental analyzer (EA) and the X-ray photoelectron spectrometer (XPS). The EA (C, H, O, N) elemental analyses (Supplementary Information) show that the total weight percents (i.e., P) are 45.2, 38.0, 55.1, and 37.2 for samples 1, 2, 3, and 4, respectively. These values are roughly consistent with the respective P values estimated from TGA curves, as shown in Fig. 3d. The differences in P between TGA and EA data are likely owing to moisture because the moisture content of EA data could not be eliminated, whereas it was eliminated in TGA data (Fig. 3b). The XPS spectra also clearly showed the fingerprint transitions of iodine [i.e., 618.8 eV (3d 5/2 ) and 630.2 eV (3d 3/2 ) in electron binding energy (EBE) scale] in all samples (Fig. 3e), confirming the surface coated ICs on GNP surfaces in all samples. The full scan XPS spectra with transition assignments in all samples are provided in Supplementary Information.
Magnetic properties. The mass-corrected magnetization (M) (emu/ g) versus the applied field (H) (i.e., M -H) curves (25 # H # 5 tesla) at a temperature T of 5 or 300 K, and the mass-corrected zero-fieldcooled (ZFC) M versus T (i.e., M -T) curves (5 # T # 300 K) at H 5 0 Oe are shown in Figs. 5. The M -H curves at T 5 5 and 300 K (Fig. 5a) show that both the coercivity and the remanence are zero in all the samples (i.e., no hysteresis occurs). This lack of hysteresis and the absence of a magnetic transition down to T 5 5 K in the M -T curves (Fig. 5b) show that all the samples are paramagnetic down to 5 K, like bulk Gd 2 O 3 [76][77][78][79] . From the M -H curves, the net M values of the GNPs were estimated for all the samples (Table 1) Table 1 are lower than those (i.e., 190-200 emu/g) 64 of the previous measurements. These are owing to the overestimated net Gd 2 O 3 masses from TGA curves because iodines were not completely removed from the samples during TGA analyses. This was confirmed from iodine signals in XPS spectra of the TGA analyzed samples (Supplementary Information). This is likely because solid iodine compounds with either oxygen or gadolinium were formed during TGA analyses.
Relaxometric and X-ray absorption results. The longitudinal water proton relaxivity (r 1 ) should be large, and the r 2 /r 1 ratio should be close to one (where r 2 is the transverse water proton relaxivity) for T 1 MRI contrast agents because the r 2 value is theoretically always greater than the r 1 value 30,31,41 , and the X-ray absorption should be strong for CT contrast agents. As discussed below, GNPs satisfy all these conditions. The r 1 and r 2 values of aqueous sample solutions were estimated from the slopes of plots of the inverse longitudinal (T 1 ) and transverse (T 2 ) relaxation times (i.e., 1/T 1 and 1/T 2 ), respectively, as a function of the Gd concentration (Fig. 6a). The estimated r 1 and r 2 values of all the samples ( Table 2) were larger than those of Gdchelates, and their r 2 /r 1 ratios were also close to one (i.e., between 1.0 and 2.0). All the sample solutions also showed clear dose-dependent contrast enhancements in their longitudinal (R 1 ) and transverse (R 2 ) map images (Fig. 6b). These results show that all the samples are potential T 1 MRI contrast agents with better relaxometric properties than commercial contrast agents. The differences in r 1 and r 2 values between samples can be attributed to the ICs used for surface coating because all the samples have similar GNP diameters. The wateraccessibility to the core GNP depends on the ligand-size 81 . In general, fewer water molecules can access the core GNP for a large ligand coating, reducing the r 1 and r 2 values 81 . The observed r 1 and r 2 values showed this trend ( Fig. 6c and Table 2), supporting the suggestion that the differences in r 1 and r 2 values between samples are due to the ICs.
Next, the X-ray absorption at an X-ray source voltage of 70 kV is plotted as a function of the Gd (or I) concentration (Fig. 6d); the observed X-ray absorption of all the samples is stronger than those of commercial contrast agents Omniscan and Ultravist because both Gd and I absorb X-ray radiation in the samples, whereas only I absorbs X-ray radiation in Ultravist and only Gd absorbs X-ray radiation in Omniscan. Here, water is a reference with 0.0 Hounsfield units (HU). X-ray absorption phantom images acquired at 100 mM Gd (or I) are also shown in Fig. 6d; the phantom images of all the samples are brighter than those of Omniscan and Ultravist. These results show that all the samples are potential CT contrast agents with X-ray absorption values higher than those of commercial contrast agents. More X-ray absorption phantom images are provided in the Supplementary Information. The observed X-ray absorption powers are in the order Sample 1 . Sample 2 < Sample 4. This is because the core GNP is the same for all the samples and because the number (NI) of iodines per GNP is in the same order: NI (Sample 1) . NI (Sample 2) < NI (Sample 4) ( Table 1). Here, the X-ray absorption of Sample 3 was not measured owing to its low concentration, but it is expected to be comparable to that of Sample 1 because NI (Sample 1) < NI (Sample 3) (Table 1).
In vitro cytotoxicity results. Human prostate cancer (DU145) and normal mouse hepatocyte (NCTC1469) cell lines were used as test cells. The cells were incubated with IC-GNP samples for 48 hours. The cell viabilities of all the samples were good for a tested Gd concentration range of up to 500 mM (Figs. 7a and b). Therefore, all the samples are biocompatible. This is because all the ICs used are commercial CT contrast agents and because all the samples are sufficiently coated with ICs, as shown by the TGA data (Fig. 3b).
In vivo results: T 1 MR and CT images. Sample 1, with the largest r 1 value among the four samples, was used to acquire in vivo T 1 MR images of mice from the Institute of Cancer Research (ICR), USA at an MR field of 1.5 tesla. Approximately 0.1 mmol Gd/kg was injected into a mouse tail vein, and T 1 MR images were acquired before and after injection (Figs. 8a and b). Appreciably positive (or brighter) contrast enhancements were observed in the mouse liver (labeled L), kidneys (labeled K), and aorta (labeled A) after injection, but returned almost to the initial contrast (i.e., the contrast before injection) in the liver 90 minutes after injection (Fig. 8a), and in the kidneys and aorta 15 minutes after injection (Fig. 8b), because nanoparticles were excreted from the respective organs. The signal to noise ratios (SNRs) of the regions of interest (ROIs) in the liver and in the cortex and medulla of the kidney (indicated with small dotted circles) are plotted as a function of the time after injection (Fig. 8c), showing a decrease in the SNR in both the liver and kidney with time as a result of excretion of nanoparticles from the respective organs. These results show that the sample solution is a potential T 1 MRI contrast agent. More T 1 MR images of Sample 1 are provided in the Supplementary Information. Sample 1, with the largest X-ray absorption among the four samples, was also used to acquire in vivo CT images at an X-ray source voltage of 70 kV. Approximately 0.53 mmol Gd/kg was injected into an ICR mouse tail vein, and in vivo CT images were acquired before and after injection (Fig. 8d). This injection dose is much smaller than 2-6.4 mmol I/kg, which was used for iodine contrast agents in an ICR mouse 82 . Brighter contrast enhancements were observed in the mouse bladder (labeled B) after injection, and they maintained up to more than 210 minutes after injection. The X-ray absorption of the ROI in the bladder (indicated with a small dotted circle) was plotted as a function of the time (Fig. 8e); the contrast reached the maximum  value at ,30 minutes after injection and then decreased, indicating that the sample solution was excreted through the bladder as urine. This result shows that the sample solution is a potential CT contrast agent. A similar result was obtained for Sample 2 and is provided in the Supplementary Information. We performed additional in vivo CT and MR imaging experiments to compare one of our samples (i.e., Sample 4) with commercial MRI (i.e., Omniscan) and CT (i.e., Ultravist) contrast agents at the same injection doses. In vivo MR coronal images of a rat obtained with the Sample 4 and Omniscan at 12 hours after intravenous injection into its tail are shown in Fig. 9a(i) and (ii), respectively. Compared to Omniscan, the Sample 4 showed stronger signal enhancements in the liver. This can be clearly seen in the mean signal to noise ratio (SNR) plot in Fig. 9a(iii). In a similar way, in vivo CT axial images of a rat obtained with the Sample 4 and Ultravist at 3 minutes after intravenous injection into its tail are shown in Fig. 9b(i) and (ii), respectively. Compared to Ultravist, the Sample 4 showed stronger signal enhancements in the liver. This can be clearly seen in the mean signal intensity plot in Fig. 9b(iii). The injection doses were ,0.64 mmol Gd/kg for MR image measurements and ,0.53 mmol Gd/kg for CT image measurements. Sprague Dawley (SD) rats (4 weeks, male) were used.
Comparison between samples. Based on r 1 values and X-ray absorption powers, the in vitro T 1 MRI and CT capability evaluations of all the samples are summarized in Fig. 10. As shown in Fig. 10, it seems that the combined capabilities of T 1 MRI and CT are roughly in the order of Sample 1 < Sample 3 . Sample 2 < Sample 4. Here, the X-ray absorption of Sample 3 (slashed column) is drawn according to the expectation that it will be comparable to that of Sample 1 because the core is the same GNP and because NI (Sample 1) < NI (Sample 3) (Table 1). Overall, GNPs with a smaller (i.e., less massive) IC coating showed larger r 1 values and higher X-ray absorption. This is because more water molecules can access the core GNP for a smaller IC coating, providing the larger r 1 values 81 , and because GNPs with a smaller IC coating had larger NI values (Table 1), giving higher X-ray absorption, which explains the observed r 1 values and X-ray absorption results in Fig. 10. However, all the samples showed larger r 1 values than commercial T 1 MRI contrast agents because of the GNPs as they contained, and higher X-ray absorption than commercial iodine contrast agents because of both the Gd and I as they contained. Therefore, all the samples should be potential dual imaging agents in T 1 MRI and CT.

Discussion
Gd is the only element that possesses such diverse and remarkable properties, which are useful for theragnosis for cancers. Therefore, GNPs with the large amount of Gd per GNP will be a powerful theragnostic agent for cancers ( Fig. 1) 27,28,61 . The only shortcoming of Gd is its toxicity 66 . Therefore, GNPs should be coated with watersoluble and biocompatible ligands. Note that this theragnosis might be difficult using the conventional molecular Gd-chelates, because of their low Gd concentrations that could be delivered to cancers 60 .
This study reports the dual imaging capability of GNPs in T 1 MRI and CT. To this end, four IC-GNP samples were synthesized. Four commercial iodine contrast agents were used as surface coating ligands on GNPs (d avg 5 ,2.0 nm) to enhance the CT functionalities of the nanoparticles using iodines in the ICs, as well as to make them water-soluble and biocompatible. To evaluate the dual imaging capability of GNPs, the r 1 values and X-ray absorption of all the samples were measured, and in vivo T 1 MR and CT images of mice were finally obtained.
Overall, GNPs with a smaller (i.e., less massive) IC coating showed larger r 1 values and higher X-ray absorption (Fig. 10). This is because more water molecules can access the core GNP for a smaller IC coating, providing the larger r 1 values 81 , and because GNPs with a smaller IC coating had larger NI values (Table 1), giving higher X-ray absorption. However, all the samples showed larger r 1 values than commercial T 1 MRI contrast agents, and higher X-ray absorption than commercial iodine contrast agents. Therefore, all the samples should be potential dual imaging agents in T 1 MRI and CT.
The dual imaging capability of GNPs was finally confirmed by in vivo T 1 MR and CT images. That is, positive (or brighter) contrast enhancements in both T 1 MR and CT images were observed in mice after intravenous injection (Fig. 8). Finally, in vivo comparisons between one of IC-GNP samples and commercial MRI and CT contrast agents were made. More enhanced contrasts in both MR and CT images at the same injection doses were observed, showing the superiority of IC-GNPs to the commercial contrast agents (Fig. 9).
In summary, owing to unique magnetic and X-ray absorption properties of Gd, and a dense population of Gd per GNP, GNPs showed an outstanding dual imaging capability in T 1 MRI and CT without additional functionalization, which is better than the respective commercial contrast agents. This, when combined with the NCT property of GNPs, will make GNPs a potential theragnostic agent for cancers, which will be investigated in the future.
One-pot synthesis of IC-GNP samples. Four IC-GNP samples were prepared by coating four different types of ICs on GNPs. The ICs used were 5-amino-2,4,6triiodoisophthalic acid, iodipamide, diatrizoic acid, and iodixanol, which are all commercial CT contrast agents (Table 1 and Fig. 4). The IC-GNP samples were synthesized in one-pot using the procedure shown in Fig. 11. Three separate solutions were prepared: (i) a precursor solution made of 5 mmol of GdCl 3 ?xH 2 O in 25 mL of triethylene glycol, (ii) a NaOH solution made of 15 mmol of NaOH in 10 mL of triethylene glycol, and (iii) an IC solution made of 5 mmol of IC in 10 mL of triethylene glycol (in the case of iodipamide, five drops of DMSO were also added to the solution to completely dissolve the iodipamide). The precursor solution was heated to 60uC with magnetic stirring under atmospheric conditions until the precursor was completely dissolved in the solvent. The NaOH solution was then added to the precursor solution. The mixed solution was magnetically stirred at 180uC for 4 hours. For surface coating, the solution temperature was then lowered to 60uC and an IC solution was added slowly to the above solution. The temperature of the solution was again raised to 110uC, and the solution was magnetically stirred for an additional 12 hours. To wash the IC-GNP samples with triply distilled water, the solution was then cooled to room temperature and transferred to a 1 L beaker containing 500 mL of triply distilled water. It was then magnetically stirred for 10 minutes and stored for a week to let the IC-GNP samples settle to the bottom of the beaker. The clear supernatant was decanted and the remaining sample was again diluted with 500 mL of triply distilled water. This washing process was repeated three times. A half volume of each sample was dried in air to obtain powder samples for various characterizations, and the remaining half volume was diluted with triply distilled water to obtain solution samples.
Measurements of particle diameter, hydrodynamic diameter, and crystal structure. The particle diameters of the IC-GNP samples were measured using an HRTEM (FEI, Titan G2 ChemiSTEM CS Probe) operating at an acceleration voltage of 200 kV. For the measurements, one drop of each sample dispersed in ethanol was dropped onto a carbon film supported by a 200 mesh copper grid (PELCO No.160, TED PELLA, INC.) placed on a filter paper using a micropipette (Eppendorf, 2-20 mL). The copper grid with the sample was left in air to dry for an hour at room temperature. The copper grid with the sample was then mounted inside the HRTEM for measurement.
The Gd concentration of each sample solution was determined using an inductively coupled plasma atomic emission spectrometer (Thermo Jarrell Ash Co., IRIS/ AP). All the samples were pre-treated with acids to completely dissolve the nanoparticles in solution before measurement.
The hydrodynamic diameters of the IC-GNP samples dispersed in triply distilled water were measured using a DLS particle size analyzer (UPA-150, Microtrac). The sample solution concentration was , 0.05 mM Gd.
The crystal structure of the IC-GNP powder samples before and after TGA analysis was measured using a powder XRD spectrometer (Philips, X-PERT PRO MRD) with unfiltered CuKa (l 5 1.54184 Å ) radiation. The scanning step and scan range in 2h were 0.033u and 15-100u, respectively.
Surface coating analysis. The surface coating of GNPs with ICs was investigated using an FT-IR absorption spectrometer (Mattson Instruments, Inc., Galaxy 7020A). For the measurements, powder samples were dried on a hot plate at ,40uC for a week to remove moisture from them. Pellets of dried powder samples in KBr were prepared, and FT-IR absorption spectra were recorded between 400 and 4000 cm 21 .
The amount of the IC coated on the GNP surface was estimated using a TGA instrument (TA Instruments, SDT-Q 600). Because organic compounds burn out below 400uC, TGA curves for each powder sample were scanned between room temperature and 700uC under air flow. The amount of surface coating for each sample was estimated from the mass drop in its TGA curve after an initial mass drop between room temperature and ,105uC due to water desorption was subtracted.
The elemental analyses of surface coated ICs on GNP surfaces were carried out using both the EA (ThermoFisher, Flash 2000) and XPS (ULVAC-PHI, Quantera SXM). Powder samples were used for both measurements. For XPS measurements, powder samples were loaded onto carbon tapes and the spectra were scanned between 0 and 1200 eV in electron binding energy (EBE) with the accumulation time of 30 to 50 minutes. The EA was used to measure the C, H, O, and N in weight percents, whereas the XPS was used to measure the C, O, N, I, and Gd in atomic percents. Relaxometric measurement. Both the T 1 and T 2 relaxation times and R 1 and R 2 map images were measured using a 1.5 tesla MRI scanner (GE 1.5 T Signa Advantage, GE Medical Systems) equipped with a knee coil (EXTREM). Five aqueous dilute solutions (1, 0.5, 0.25, 0.125, and 0.0625 mM Gd) were prepared per sample by diluting each concentrated sample solution with triply distilled water. These dilute solutions were then used to measure both the T 1 and T 2 relaxation times and R 1 and R 2 map images. The r 1 and r 2 water proton relaxivities of each sample were then estimated from the slopes of plots of 1/T 1 and 1/T 2 , respectively, versus the Gd concentration. T 1 relaxation time measurements were carried out using an inversion recovery method. In this method, the inversion time (TI) was varied at 1.5 tesla and the MR images were acquired at 35 different TI values ranging from 50 to 1750 ms. The T 1 relaxation times were then obtained from the nonlinear least-square fits to the measured signal intensities at various TI values. For T 2 relaxation time measurements, the Carr2Purcell2Meiboon2Gill pulse sequence was used for multiple spin2echo measurements. Then, 34 images were acquired at 34 different echo time (TE) values ranging from 10 to 1900 ms. The T 2 relaxation times were obtained from the nonlinear least-square fits to the mean pixel values for the multiple spin2echo measurements at various TE values.
Cell viability measurement. The cytotoxicity of each sample was measured using a CellTiter-Glo Luminescent Cell Viability Assay (Promega, WI, USA). In this assay, the intracellular adenosine triphosphate was quantified using a luminometer (Victor 3, Perkin Elmer). DU145 and NCTC1469 cell lines were used as test cells. Each cell line was seeded onto a separate 24-well cell culture plate and incubated for 24 hours (5 3 10 4 cell density, 500 mL cells per well, 5% CO 2 , and 37uC). Four dilute solutions were prepared per sample by diluting each concentrated sample solution with a sterile phosphate buffer saline solution. Each test cell was treated by ,2 mL of one of the dilute solutions. The final Gd concentrations in the four treated cells were 10, 100, 200, and 500 mM per sample. The treated cells were then incubated for 48 hours. The cell viabilities of treated cells were measured twice to obtain the average values and then normalized with respect to a control cell that was not treated with a sample solution.
Animal experiment. All the animal experiments using mice were approved by the animal research committee of Kyungpook National University and conducted in accordance with its rules.
In vivo T 1 MR image measurement. In vivo T 1 MR images of mice were acquired using the same MRI scanner used for the relaxometric measurements. ICR female mice with a weight of ,30 g were used for the T 1 MR image measurements. For imaging, a mouse was anesthetized using 1.5% isoflurane in oxygen. Measurements were made before and after injection of a sample solution into a mouse tail vein. The injection dose was typically ,0.1 mmol Gd/kg. After measurement, the mouse was revived from anaesthesia and placed in a cage with free access to food and water. During the measurement, the temperature of the mouse was maintained at ,37uC using a warm water blanket. The parameters used for the measurements are as follows: the H 5 1.5 tesla, the T 5 37uC, the NEX 5 8, the FOV 5 100 mm, the phase FOV 5 0.5, the matrix size 5 256 3 160, the slice thickness 5 1 mm, the spacing gap   samples. Comparison of r 1 values and X-ray absorption of all the samples (y-axis is commonly labeled as value for r 1 value and X-ray absorption). The X-ray absorption of Sample 3 (slashed column) is drawn according to the expectation that it will be comparable to that of Sample 1 because the core is the same GNP and because NI (Sample 1) < NI (Sample 3) as mentioned in the text. Phantom image and X-ray absorption measurements. X-ray phantom images were acquired using a micro-CT scanner (Siemens, Inveon). Four dilute solutions (20,50,80, and 100 mM Gd) were prepared per sample by diluting each concentrated sample solution with triply distilled water. A phantom image of water served as a reference with 0.0 HU, and those of the commercial contrast agents Ultravist (20,50,80, and 100 mM I) and Omniscan (20,50,80, and 100 mM Gd) were also measured for comparison. The X-ray absorption of each dilute solution was estimated in HU with respect to water. The parameters used for the measurements are as follows: the X-ray source current 5 400 mA, the X-ray source voltage 5 70 kV, the imaging time per frame 5 200 ms, and the reconstructed image size 5 512 3 512.
In vivo CT image measurement. In vivo CT images of mice were acquired using the same micro-CT scanner used for the phantom image measurements. ICR female mice with a weight of ,30 g were used for the measurements. The injection dose was typically ,0.53 mmol Gd/kg. For imaging, a mouse was anesthetized using 1.5% isoflurane in oxygen. Measurements were made before and after injection of a sample solution into a mouse tail vein. After measurement, the mouse was revived from anaesthesia and placed in a cage with free access to food and water. The parameters used for measurements are as follows: the X-ray source current 5 400 mA, the X-ray source voltage 5 70 kV, the imaging time per frame 5 200 ms, and the reconstructed image size 5 512 3 512.