Magnetic resonance imaging, gadolinium neutron capture therapy, and tumor cell detection using ultrasmall Gd2O3 nanoparticles coated with polyacrylic acid-rhodamine B as a multifunctional tumor theragnostic agent

Monodisperse and ultrasmall gadolinium oxide (Gd2O3) nanoparticle colloids (davg = 1.5 nm) (nanoparticle colloid = nanoparticle coated with hydrophilic ligand) were synthesized and their performance as a multifunctional tumor theragnostic agent was investigated. The aqueous ultrasmall nanoparticle colloidal suspension was stable and non-toxic owing to hydrophilic polyacrylic acid (PAA) coating that was partly conjugated with rhodamine B (Rho) for an additional functionalization (mole ratio of PAA : Rho = 5 : 1). First, the ultrasmall nanoparticle colloids performed well as a powerful T1 magnetic resonance imaging (MRI) contrast agent: they exhibited a very high longitudinal water proton relaxivity (r1) of 22.6 s−1 mM−1 (r2/r1 = 1.3, r2 = transverse water proton relaxivity), which was ∼6 times higher than those of commercial Gd-chelates, and high positive contrast enhancements in T1 MR images in a nude mouse after intravenous administration. Second, the ultrasmall nanoparticle colloids were applied to gadolinium neutron capture therapy (GdNCT) in vitro and exhibited a significant U87MG tumor cell death (28.1% net value) after thermal neutron beam irradiation, which was 1.75 times higher than that obtained using commercial Gadovist. Third, the ultrasmall nanoparticle colloids exhibited stronger fluorescent intensities in tumor cells than in normal cells owing to conjugated Rho, proving their pH-sensitive fluorescent tumor cell detection ability. All these results together demonstrate that ultrasmall Gd2O3 nanoparticle colloids are the potential multifunctional tumor theragnostic agent.


Introduction
The superior physical properties of Gd (see Abbreviations) applicable to various biomedical applications such as T 1 MRI, 1,2 CT, 3,4 and GdNCT [5][6][7] provide us with the opportunity to synthesize tumor theragnostic agents using one element of Gd in various chemical forms.This condition of Gd simplies the design and synthesis of tumor theragnostic agents, and minimizes the production cost.The effectiveness of Gd is most signicant in the form of ultrasmall Gd 2 O 3 nanoparticle colloids (nanoparticle colloid ¼ nanoparticle coated with hydrophilic ligand) owing to their high Gd-density per nanoparticle colloid.An additional advantage of ultrasmall Gd 2 O 3 nanoparticle colloids over molecular Gd-chelates is their large surface area available for additional functionalization through surface-conjugation.Ultrasmall Gd 2 O 3 nanoparticle colloids also have advantages over conventional large Gd 2 O 3 nanoparticle colloids in biomedical applications owing to their higher number density, higher colloidal stability, higher performance, and renal excretion ability.
In this study, we present monodisperse and ultrasmall Gd 2 O 3 nanoparticle colloids as a multifunctional tumor theragnostic agent.The nanoparticle colloid is composed of three components: ultrasmall Gd 2 O 3 nanoparticle, PAA, and Rho (Fig. 1a).First, PAA was used as surface-coating ligand to make the ultrasmall Gd 2 O 3 nanoparticle colloids stable and non-toxic.It is an excellent surface-coating biocompatible polymer with numerous carboxylic groups per polymer, 8 with one carboxylic group per monomer unit.An additional functionalization of the nanoparticle colloids can be easily performed by conjugating a functional molecule such as drug, tumor-targeting ligand, or dye (such as Rho in this study) to one of its carboxylic groups.The remaining carboxylic groups can be used for surface coating.The other two components are described below.
The main component, i.e. ultrasmall Gd 2 O 3 nanoparticle has both diagnostic and therapeutic functions as mentioned before, and among them, those applicable to T 1 MRI [9][10][11] and GdNCT 7,12,13 were investigated in this study.First, an important focus of this study was the application of ultrasmall Gd 2 O 3 nanoparticle colloids to GdNCT in vitro.6][17] 157 Gd (natural abundance ¼ 15.7%) has the highest thermal neutron capture cross-section of 257 000 barns among stable radionuclides, which is $67 times higher than 3840 barns of 10 B (natural abundance ¼ 19.7%) currently used for BNCT. 5,13,14However, common elements in the body, such as H, O, C, N, S, Na, Cl, K, Ca, and Fe, have tiny thermal neutron capture cross-section values. 5,14Therefore, GdNCT does no harm to our body unless high thermal neutron beam irradiation doses are used.The tumor therapy by GdNCT occurs by nuclear reaction products.Therefore, for successful GdNCT, Gd-chemicals should reside inside or close to tumor cell nuclei because of the short path lengths (a few mm) of both IC and ACK electrons, the nuclear reaction products of the 157 Gd(n,g) 158* Gd nuclear reaction. 5,14hese electrons can damage DNA inside tumor cell nuclei, thus killing tumor cells.The Gd-chemicals applied to GdNCT to date include commercial Gd-chelates such as Gd-DTPA and Gd-DOTA, [18][19][20][21] fullerene-encapsulated Gd, 22 GdCo@carbon nanoparticles, 23 hybrid Gd 2 O 3 @polysiloxane nanoparticles, 12 and various nanocomposites containing commercial Gd-chelates such as chitosans, 24,25 liposomes, 26,27 calcium phosphate polymeric micelles, 28,29 and ethyl cellulose microcapsules. 30Nanocomposites accumulated more in tumor cells than did Gdchelates in vivo 25,28,31,32 and in vitro. 27,31,33,34However, these nanocomposites are generally too large (d > 30 nm) for intravenous administration.Therefore, they were usually administrated either intratumorally 24,25,32,35 or intraperitoneally. 30,31owever, ultrasmall nanoparticle colloids (d < 3 nm) can circulate freely through capillary vessels and be excreted through the renal system aer intravenous administration. 9,36,37hey may be able to target tumor cells aer conjugation with targeting ligands.Therefore, ultrasmall Gd 2 O 3 nanoparticle colloids will be extremely useful for GdNCT applications and thus were applied to GdNCT in vitro in this study to investigate their effectiveness as GdNCT agent before in vivo application.Their performance was compared to that of Gadovist (Gadobutrol, Bayer Schering Pharma, Germany), a commercial molecular T 1 MRI contrast agent.Second, ultrasmall Gd 2 O 3 nanoparticle colloids can be applied to T 1 MRI.To demonstrate this, water proton relaxivities and in vivo T 1 MR images in a nude mouse aer intravenous administration of the sample solution were investigated in this study.
Lastly, pH-sensitive uorescent tumor cell detection of ultrasmall Gd 2 O 3 nanoparticle colloids was explored in this study.To this end, Rho was conjugated to surface-coating PAA through amide bond because Rho shows pH-dependent uorescent intensity in the visible region (l em ¼ 582 nm), which is useful for its use as a pH-sensitive tumor cell detection probe, [38][39][40] as well as for use as an FI agent. 40,41Rho tints red in acidic conditions but has no color in basic conditions, and thus exhibits a stronger uorescent intensity at more acidic conditions.The origin for this is the spirolactam ring opening reaction in acidic conditions.Therefore, ultrasmall Gd 2 O 3 nanoparticle colloids will show stronger uorescent intensities in tumor cells than in normal cells because the pH of tumor cells is more acidic than that of normal cells, which is nearly neutral. 42Therefore, combining all the applications mentioned above (Fig. 1b), the ultrasmall Gd 2 O 3 nanoparticle colloid can serve as a multifunctional tumor theragnostic agent as demonstrated in this study.

Synthesis of Rho-PAA
To prepare Rho-PAA, Rho-NH 2 was rst synthesized according to the known method (Fig. 2a). 43,44Briey, a mixture of 5 mmol Rho, 20 mL ethanol, and 3 mL ethylenediamine was reuxed overnight until the reaction mixture lost its red color.Ethanol was evaporated and the crude mixture was then washed with triple distilled water, extracted with CH 2 Cl 2 , and dried by Na 2 SO 4 .The organic phase was concentrated and puried using a silica gel column (DCM : CH 3 OH ¼ 95 : 5 as eluent) to obtain Rho-NH 2 in a light-yellow solid form (1.86 g, 76.8% yield).Formation of Rho-NH 2 was conrmed from its NMR and FT-IR absorption spectra (Fig. S1 and S2, † respectively).
Rho-PAA was synthesized according to the known method 45 with a slight modication (Fig. 2b).0.5 mmol PAA, 0.25 mmol DCC, and a catalytic amount of 4-DMAP were dissolved in 10 mL dry THF.Then 0.1 mmol Rho-NH 2 dissolved in dry THF was slowly added to the PAA solution, which was further magnetically stirred at room temperature for 2 days.THF was later removed by evaporation and the crude product was washed with dichloromethane and puried by dialysis (MWCO ¼ 1000 Da) against triple distilled water for 1 day.Formation of Rho-PAA was conrmed from its FT-IR absorption spectrum, and pHdependent solution and uorescent solution colors (Fig. S3 and S4, † respectively).The PAA used has $twenty-ve -COOH groups per polymer.Calculations based on the mole ratio of PAA : Rho-NH 2 ¼ 5 : 1 used in the synthesis, showed that one of ve PAA molecules was conjugated to one Rho, and the others were free PAA.Therefore, the nal product was a mixture of PAA and Rho-PAA in the mole ratio of 4 : 1.This was conrmed by an FT-IR absorption spectrum of the product that was overwhelmed by PAA peaks (Fig. S3 †).The product (PAA + Rho-PAA) was obtained as a solid with a red color owing to Rho-PAA.The pH-dependent solution and uorescent solution colors were similar to those of free Rho: an aqueous solution of the product tinted red and exhibited strong uorescence in the red region aer irradiation at 365 nm with a mercury lamp, and had no color and no uorescence at basic pH values like free Rho (Fig. S4 †).
Synthesis of ultrasmall Gd 2 O 3 nanoparticle colloids (coating material ¼ a mixture of PAA and PAA-Rho in the mole ratio of PAA : PAA-Rho ¼ 4 : 1) In a three-neck round bottom ask, a mixture of half of the above mixture (PAA + Rho-PAA) ($0.25 mmol) and 2 mmol GdCl 3 $xH 2 O were magnetically stirred in 20 mL TEG for 2 h at 60 C under atmospheric conditions until a clear red precursor solution was obtained (Fig. 2c).In a separate beaker, 10 mmol NaOH in 10 mL TEG was prepared and then added slowly to the precursor solution.Owing to the increment of pH to 9-11 aer addition of the NaOH solution, the mixed solution lost its red color.The solution was magnetically stirred at 110 C for 12 h and then cooled to room temperature.The solution was transferred to a 500 mL beaker and 400 mL ethanol was added.The solution was magnetically stirred at room temperature for 10 minutes and kept in refrigerator for 3 days until the ultrasmall Gd 2 O 3 nanoparticle colloids settled to the bottom of the beaker.The supernatant was decanted and the remaining solution was lled with 400 mL ethanol.This process was repeated 5 times.The product solution was washed 5 times with an adequate amount (<10 mL) of triple distilled water to remove ethanol by evaporation.A portion of the product was dried in air to powder form for various characterizations and the rest of the product was diluted with triple distilled water to prepare a nanoparticle colloidal suspension ($30 mM Gd).

General characterizations
The particle diameters were measured using an HRTEM (Titan G2 ChemiSTEM CS Probe, FEI, Hillsboro, OR, USA) operated at an accelerating voltage of 200 kV.For measurements, one drop of diluted nanoparticle colloids in triple distilled water was dropped onto a carbon lm supported by a 200-mesh copper Fig. 3 The in vitro GdNCT experimental procedure.grid (PELCO no.160, Ted Pella, Inc., Redding, CA, USA) and placed on a lter paper using a micropipette (2-20 mL, Eppendorf, Hamburg, Germany).The copper grid with the sample was le to air dry for 1 h at room temperature.The copper grid with the sample was then mounted inside the HRTEM for measurement.
The crystal structure of the powder samples before and aer TGA was measured using a powder XRD spectrometer (X-PERT PRO MRD, Philips, Eindhoven, The Netherlands) with unltered CuKa (l ¼ 1.54184 Å) radiation.The scanning step and scan range in 2q were 0.033 and 15-100 , respectively.
The surface coating of the nanoparticles with a mixture of PAA and Rho-PAA was investigated by recording FT-IR absorption spectra with an FT-IR absorption spectrometer (Galaxy 7020A, Mattson Instruments, Inc., Madison, WI, USA).For measurements, the powder samples were dried on a hot plate at $40 C for a week to remove moisture.Pellets of dried powder samples were prepared in KBr.FT-IR absorption spectra were recorded in the range of 400-4000 cm À1 .
The amount of surface coating on the nanoparticle surface was estimated by recording a TGA curve using a TGA instrument (SDT-Q 600, TA Instruments, New Castle, DE, USA).Because organic compounds burn out below 400 C, a TGA curve was scanned between room temperature and 900 C under air ow.The amount of surface coating was estimated from the mass drop in the TGA curve aer subtraction of the initial mass drop between room temperature and $105 C owing to water and air desorption.
The Gd concentration of the nanoparticle colloids suspended in triple distilled water was determined using an ICPAES (IRIS/AP, Thermo Jarrell Ash Co., Franklin, MA, USA).The colloidal suspension was pre-treated with acids to completely dissolve the nanoparticle colloids in solution before measurement.

In vitro cytotoxicity measurements
The in vitro cytotoxicity of the nanoparticle colloids was measured using a CellTiter-Glo Luminescent Cell Viability Assay (Promega, Madison, WI, USA).The ATP was quantied using a Victor 3 luminometer (Perkin Elmer, Waltham, MA, USA).Three cell lines, i.e.DU145, NCTC1469, and U87MG, were used.Each cell line was seeded onto a separate 24-well cell culture plate and incubated for 24 h (5 Â 10 4 cell density, 500 mL cells per well, 5% CO 2 , and 37 C).Five dilute solutions were prepared by the dilution of the concentrated colloidal suspension with a sterile PBS solution.Each of the test cells was treated with $2 mL of each diluted sample solution and the nal Gd concentrations in the treated cells were 10, 50, 100, 200, and 500  mM Gd.The treated cells were then incubated for 48 h.Cell viabilities were measured twice to obtain the average cell viabilities, which were then normalized with respect to that of untreated control cells (0.0 mM Gd).

Relaxivity and map image measurements
The T 1 and T 2 relaxation times and the R 1 and R 2 map images were measured using a 1.5 T MRI scanner (GE 1.5 T Signa Advantage, GE Medical Systems, Chicago, IL, USA) equipped with a knee coil (MSK-Extreme, ONI Medical Systems, Inc., Wilmington, MA, USA).Five aqueous dilute sample solutions (1.0, 0.5, 0.25, 0.125, and 0.0625 mM Gd) were prepared by diluting the concentrated colloidal suspension with triple distilled water.These dilute solutions and triple distilled water were then used to measure both the T 1 and T 2 relaxation times and the R 1 and R 2 map images.The r 1 and r 2 water proton relaxivities of the sample solution 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 conducted using an inversion recovery method.In this method, the TI was varied at 1.5 T and the MR images were acquired at 35 different TI values in the range from 50 to 1750 ms.The T 1 relaxation times were then obtained from the nonlinear leastsquare ts to the measured signal intensities at various TI values.For the measurements of T 2 relaxation times, the Carr-Purcell-Meiboom-Gill pulse sequence was used for multiple spin-echo measurements.A total of 34 images were acquired at 34 different TE values in the range from 10 to 1900 ms.The T 2 relaxation times were obtained from the nonlinear least-square ts to the mean pixel values for the multiple spin-echo measurements at various TE values.

Animal experiment
This study was performed in accordance with the Korean guidelines and approved by the animal research committee of Kyungpook National University.
In vivo T 1 MR image measurements in a nude mouse In vivo T 1 MR images were acquired using the same MRI scanner used for relaxometry measurements.For imaging, an ICR (Institute of Cancer Research, USA) mouse ($30 g) was anesthetized with 1.5% isourane in oxygen.Measurements were made before and aer administration of the nanoparticle colloidal suspension into the mouse tail vein.The administration dose was typically $0.1 mmol Gd per kg.Aer measurement, the mouse was revived from anesthesia and placed in a cage with free access to food and water.During measurement, temperature of the mouse was maintained at $37 C using a warm water blanket.The parameters used for the measurements were as follows: external MR eld ¼ 1.5 T; temperature ¼ 37 C; NEX ¼ 4; FOV ¼ 6 mm; phase FOV ¼ 0.5; matrix size ¼ 256 Â 192; slice thickness ¼ 1 mm; spacing gap ¼ 0.5 mm (coronal) and 2.0 mm (axial); pixel bandwidth ¼ 15.63 Hz; TR ¼ 500 ms; and TE ¼ 13 ms.

In vitro GdNCT experiments
The GdNCT experiments were conducted using the cyclotron (MC50, Scanditronix, Sweden) and irradiation facilities installed at the Korea Institute of Radiological & Medical  Science (Fig. S5 †).It was operated at 35 MeV and 20 mA with 9 Be target (diameter ¼ 17 mm) to generate thermal neutron beam.The GdNCT experimental procedure is provided in Fig. 3.As shown in Fig. 3, the experiments consisted of 4 stages: cell culture, cell treatment with ultrasmall Gd 2 O 3 nanoparticle colloids (i.e.sample) and Gadovist, thermal neutron beam irradiation, and clonogenic assay.Two 6-cell well plates were used for cell culture.The cell wells in each plate were divided into 3 groups (i.e., control, sample, and Gadovist) such that each group occupied 1 cell well in each plate.At the cell culture stage, U87MG tumor cells were seeded on 3 cell wells in each plate and incubated for 24 h (5 Â 10 4 cell density, 5% CO 2 , and 37 C).At the cell treatment stage, control cells were untreated (0.0 mM Gd), sample cells were treated with the sample (0.5 mM Gd), and Gadovist cells were treated with Gadovist (0.5 mM Gd).Both treated and untreated (i.e.control) cells were incubated for 24 h and then the treated cells were washed with PBS solution 3 times to remove free nanoparticle colloids and Gadovist.At the irradiation stage, the right cell well plate in Fig. 3 was irradiated with a thermal neutron beam for 12 min, corresponding to a radiation dose of $6 Gy, and the le one was not irradiated.At the clonogenic assay stage, 500 and 1000 cells in each cell well in the two 6-cell well plates were transferred to Petri dishes (diameter ¼ 6 mm).These 2 cell numbers were used to verify consistency of cell viabilities.Twelve or 6 sets of cell dishes in total were prepared this way for clonogenic assay (see the bottom in Fig. 3).Here, each set in each group consisted of unirradiated (0 min) and irradiated (12 min) cell dishes for the 2 cell numbers.All 6 sets of cell dishes were incubated for 2 weeks to allow colonial formations.The cell viabilities were analyzed using a clonogenic assay protocol 46 and measured by cell counting.All cells spent the same time from the cell culture to clonogenic assay.

pH-dependent uorescent spectral measurements
Fluorescent spectra of the ultrasmall Gd 2 O 3 nanoparticle colloids suspended in triple distilled water at different pH values were recorded using a uorescent spectrophotometer (Cary Eclipse, Agilent Tech., Santa Clara, CA, USA).The nanoparticle colloidal suspensions (2 mM Gd) at different pH values were prepared by addition of a small amount of 1 M HCl or 1 M NaOH solution to the sample solution while monitoring the pH value of the sample solution with a digital pH-meter equipped with a glass electrode (CyberScan pH 10, Eutech Instrument, Vernon Hills, IL, USA).For measurements, each sample solution was placed in a cuvette (Sigma-Aldrich, 4 mL) with four optically clear sides, which was put in a dark chamber of the spectrometer.A mercury lamp was used at l ex ¼ 562 nm.
Fluorescent microscopy image measurements in normal and tumor cells: pH-dependent uorescent tumor cell detection DU145 tumor and NCTC1469 normal cells were seeded on 24well culture plate at a density of 1.0 Â 10 4 and 1.5 Â 10 4 per well, respectively, and incubated for 24 h in similar conditions used for cytotoxicity measurements.The cells were treated with an ultrasmall Gd 2 O 3 nanoparticle colloidal solution (2.5 mM Gd, 2 mL), which was prepared by diluting the concentrated colloidal suspension with a sterile PBS solution, and then incubated for 24 h.Aer washing the treated cells with serumfree media twice, uorescence microscopy images of the treated and untreated (i.e.control) cells were captured using a uorescence microscope (IX 51, Olympus, Japan) with a mercury lamp at l ex ¼ 562 nm.

Particle diameter and crystal structure of ultrasmall Gd 2 O 3 nanoparticle colloids
As shown in HRTEM images (Fig. 4a(I-VI)), the monodisperse and ultrasmall particle diameter (d) ranged from 1.0 to 2.5 nm with the d avg of 1.5 nm which was estimated from a log-normal function t to the observed particle diameter distribution (Fig. 4b).XRD patterns before and aer TGA showed that the asprepared powder sample was amorphous owing to ultrasmall particle diameters, but the TGA-treated powder sample showed a cubic structure owing to particle growth, 47 with a cell constant of a ¼ 10.82 Å (Fig. S6 and Table S1 †), which is consistent with the literature (JCPDS card no.43-1014). 48In the previous experiment, 49 we observed that free Gd 3+ ion concentration in solution owing to dissociation of the Gd 2 O 3 nanoparticles coated with ligands was below the detection limit of the ICPAES.Therefore, the dissociation of the Gd 2 O 3 nanoparticle colloids will be negligible in vitro and in vivo.

Surface coating results with a mixture of PAA and Rho-PAA
The surface coating of a mixture of PAA and Rho-PAA on the nanoparticle surface was demonstrated by recording FT-IR absorption spectra (Fig. 5a).As shown in Fig. 5a, characteristic symmetric stretches of C-H at 2920 cm À1 , COO À at 1400 cm À1 , and C-O at 1065 cm À1 , and antisymmetric stretch of COO À at 1550 cm À1 were observed in the sample, conrming the successful surface-coating of the nanoparticles with a mixture of PAA and Rho-PAA.Here, the C]O symmetric stretch of PAA and Rho-PAA at 1704 cm À1 was split into symmetric and antisymmetric stretches of COO À in the sample and red-shied.These are caused by electrostatic bonding between the COO À of PAA and Rho-PAA, and Gd 3+ on the nanoparticle surface (see Fig. 1a for the surface coating structure).1][52] Splitting and red-shi have been observed in many metallic oxide nanoparticles coated with ligands with -COOH groups, [53][54][55][56] supporting our result.
The amount of surface coating with a mixture of PAA and Rho-PAA on the nanoparticle surface was estimated from a TGA curve (Fig. 5b).The initial mass drop of 9.4% between room temperature and $105 C was due to water and air desorption.The mass drop of 64.9% aer this was due to burning of PAA and Rho-PAA in air.The remaining mass of 25.7% was due to Gd 2 O 3 , as conrmed from its XRD pattern (Fig. S6 †).Graing density, 57 corresponding to the average number of ligands coated per nanoparticle unit surface area, was estimated to be $1.5 nm À2 using the bulk density of Gd 2 O 3 (7.41g cm À3 ), 58 the surface-coating amount of 64.9% estimated as above, and the d avg of 1.5 nm determined by HRTEM imaging.The average number of ligands coated per nanoparticle was estimated to be $10 by multiplying the graing density by the nanoparticle surface area (pd avg 2 ), indicating that two Rho-PAA and eight PAA coated each nanoparticle on average.

In vitro cytotoxicity results
It is well-known that gadolinium is toxic. 59As shown in Fig. 6a, uncoated Gd 2 O 3 particles (purchased from Sigma-Aldrich, USA) exhibited high toxicities in both NCTC1469 normal and U87MG tumor cell lines.Gadolinium MRI contrast agents can cause nephrogenic systemic brosis (NSF) if free Gd 3+ ions are deposited in tissues. 60Owing to this, bare Gd 2 O 3 nanoparticles cannot be used for biomedical applications without coating.Therefore, Gd 2 O 3 nanoparticles were coated with hydrophilic and biocompatible PAA and then functionalized with Rho in this study.As shown in Fig. 6b, the ultrasmall Gd 2 O 3 nanoparticle colloids exhibited high cell viabilities such that $93% in DU145, $99% in NCTC1469, and $80% in U87MG cell lines at 500 mM Gd, showing good biocompatibility.
Relaxivities and map images r 1 and r 2 values were estimated to be 22.6 and 29.5 s À1 mM À1 (r 2 / r 1 ¼ 1.3), respectively, from the 1/T 1 and 1/T 2 plots versus the Gd concentration (Fig. 7a).The r 1 value was $6 times higher than those 1,2 of commercial Gd-chelates.This higher r 1 value and r 2 / r 1 ratio (¼ 1.3) which is close to 1, suggested that the nanoparticle colloids would be a powerful T 1 MRI contrast agent.This was conrmed in vitro from their R 1 and R 2 map images, clearly showing dose-dependent contrast enhancements with increasing Gd concentration (Fig. 7b).This higher r 1 value is owing to ultrasmall particle size and hydrophilic surfacecoating with PAA.In vivo T 1 MR images at 1.5 T The effectiveness of ultrasmall Gd 2 O 3 nanoparticle colloids as a powerful T 1 MRI contrast agent was demonstrated by recording T 1 MR images at 1.5 T as a function of time aer intravenous administration.Positive contrast enhancements were clearly observed in the liver, heart, kidneys, and bladder of a nude mouse aer intravenous administration (Fig. 8a).The SNRs are plotted as a function of time (Fig. 8b).The SNRs in the liver, heart, and kidneys initially increased and then slowly decreased with time while the SNR in the bladder increased with time up to 4 h, indicating a slow renal excretion.Renal excretion is due to the ultrasmall particle size of the nanoparticle colloids, as observed by others. 9,38,62However, a slow renal excretion is likely related to the Rho conjugated to surfacecoated PAA.The Rho may bind non-covalently to proteins in blood plasma such as albumin, brinogen, and globulins, as observed for other organic dyes such as FITC and uorescein, 63 which consequently slowed down the renal excretion.The mouse did not die aer in vivo experiment.From the SNR plots in Fig. 8b and using the half-areas of the plots with baseline corrections aer extrapolations to the pre-SNR values, the half-life of the nanoparticle colloids in the heart and liver was estimated to be $155 min.Therefore, the halflife of the nanoparticle colloids in the mouse would be $155 min.

In vitro GdNCT results
The photos of the 6 sets of cell dishes taken 2 weeks aer colonial formation are shown in Fig. 9.The cell viabilities of U87MG cells were estimated by cell counting.The cell viability of irradiated cells (12 min, Fig. 9) in each set was then normalized with respect to that of the corresponding unirradiated cells (0 min, Fig. 9).The results are plotted in histograms (Fig. 10a), showing consistency between cell viabilities of the two cell numbers.Therefore, we took an average of the cell viabilities for the two cell numbers.Cell death in the control was also observed.This was caused by the high dose of irradiation, as observed by others, 12,19,21,22,27 which should be avoided because cells should not be damaged by thermal neutron beam itself.By subtracting the control cell death from those of sample and Gadovist, the net cell deaths of sample and Gadovist were estimated and plotted in histograms (Fig. 10b).
The sample exhibited an average net cell death of 28.1%, which was 1.75 times higher than that of Gadovist.This suggests a higher cellular uptake of the ultrasmall Gd 2 O 3 nanoparticle colloids than Gadovist, since the GdNCT effect is proportional to Gdconcentration in tumor cells as observed in various nanocomposites. 25,27,28,31,32This higher cellular uptake is critical for in vivo GdNCT experiments, and closely related to surface compositions. 64ble 1 (Contd.) Gd-chemical Gd-chemicals that have been applied to or proposed for GdNCT experiments are provided in Table 1.As given in Table 1, however, studies of GdNCT agents are not rich.[28][29][30] The limitations of the previous studies are as follows.Commercial molecular T 1 MRI contrast agents were not suitable for in vivo GdNCT applications because they lack the ability to target tumor cells.1][32]35 Therefore, GdNCT agents which can circulate freely through capillary aer intravenous administration and then target tumor cells, are needed.In this respect, ultrasmall Gd 2 O 3 nanoparticle colloid synthesized in this study may be one of candidates suitable for this purpose aer conjugation with targeting ligand, which thus will be explored in the future.

pH-sensitive uorescent tumor cell detection
As expected, the aqueous ultrasmall nanoparticle colloidal suspension exhibited stronger uorescent intensities at more acidic pH values (l em ¼ 582 nm, l ex ¼ 562 nm) (Fig. 11) and pHdependent solution colors (Fig. S7 †), similar to free Rho.This pH-dependent uorescent property of the ultrasmall Gd 2 O 3 nanoparticle colloids was applied to tumor cell detection.
Optical and uorescent microscopy images of both DU145 tumor and NCTC1469 normal cells were taken before and aer treatments with Rho and the ultrasmall Gd 2 O 3 nanoparticle colloids.Before treatment (¼control), no uorescence was observed in both DU145 tumor and NCTC1469 normal cells (Fig. 12a).For Rho-treated cells, a brighter uorescence was observed in DU145 tumor cells than that in NCTC1469 normal cells, as expected because the pH of tumor cells is more acidic than that of normal cells which is nearly neutral 42 (l ex ¼ 562 nm) (Fig. 12b).9][40] For nanoparticle colloid treated cells, a stronger uorescence was also observed in DU145 tumor cells than that in NCTC1469 normal cells owing to the conjugated Rho in the nanoparticle colloids (Fig. 12c).This result proves the pHsensitive tumor cell detection ability of the ultrasmall nanoparticle colloids.

Conclusions
Excellent performance of ultrasmall Gd 2 O 3 nanoparticle colloids (d avg ¼ 1.5 nm) in T 1 MRI, in vitro GdNCT, and pHsensitive tumor cell detection was demonstrated.The results, as summarized below, indicate that the ultrasmall Gd 2 O 3 nanoparticle colloids are the potential candidate for use as a multifunctional tumor theragnostic agent.
(1) Hydrophilic and biocompatible PAA was used as a surface-coating ligand that was partly conjugated with Rho through amide bond for pH-sensitive tumor cell detection (mole ratio of PAA : Rho ¼ 5 : 1).
(2) Ultrasmall Gd 2 O 3 nanoparticle colloids suspended in triple distilled water exhibited a very high r 1 value of 22.6 s À1 mM À1 (r 2 /r 1 ¼ 1.3).As a result, highly positive contrast enhancements were observed in T 1 MR images in a nude mouse aer intravenous administration, proving their potential as a powerful T 1 MRI contrast agent.
(3) Ultrasmall Gd 2 O 3 nanoparticle colloids were applied to GdNCT in vitro.They showed signicant U87MG tumor cell death aer thermal neutron beam irradiation (0.5 mM Gd, $6 Gy).The net tumor cell death (corrected by tumor cell death by thermal neutron beam only) was estimated to be 28.1%, which was 1.75 times higher than that obtained using commercial Gadovist.
(4) Ultrasmall Gd 2 O 3 nanoparticle colloids showed stronger uorescent intensities in DU145 tumor cells than in NCTC1469 normal cells owing to conjugated Rho, proving their pHsensitive uorescent tumor cell detection ability.experiments, HC and YC took and analyzed MRI data, ITO and KSC measured and analyzed cytotoxicity and uorescent microscopy image data, KHJ, MHK and YJL performed GdNCT experiments, and GHL wrote the manuscript.

Fig. 1
Fig. 1 (a) Three components (i.e.ultrasmall Gd 2 O 3 nanoparticle, PAA, and Rho) of the ultrasmall Gd 2 O 3 nanoparticle colloid, the role of each component, and the surface coating structure.(b) Three applications of the ultrasmall Gd 2 O 3 nanoparticle colloid investigated in this study.

Fig. 4
Fig. 4 (a(I-IV)) HRTEM images at different magnifications [arrows indicate ultrasmall Gd 2 O 3 nanoparticle colloids and the circled region in (a-III) was magnified in (a-VI)] and (b) a log-normal function fit to the observed particle diameter distribution.

Fig. 7
Fig. 7 (a) Plots of 1/T 1 and 1/T 2 as a function of Gd concentration (the slopes correspond to r 1 and r 2 values, respectively).(b) Dose-dependent R 1 and R 2 map images of aqueous ultrasmall Gd 2 O 3 nanoparticle colloidal suspension.

Fig. 9
Fig. 9 Photos of 6 sets of cell dishes containing U87MG tumor cells 2 weeks after colonial formation: control (0.0 mM Gd), Gadovist (0.5 mM Gd), and sample (0.5 mM Gd). 0 and 12 min indicate thermal neutron beam irradiation time, corresponding to 0 (i.e.no irradiation) and $6 Gy irradiation doses, respectively.All cells spent the same time from the cell culture to clonogenic assay.

Fig. 8
Fig. 8 (a) In vivo T 1 MR images at 1.5 T and (b) SNRs in the heart, liver, kidneys, and bladder in a nude mouse before (¼pre) and 35 min and 4 h after intravenous administration.Labels: Hheart; Lliver; Kkidneys; and Bbladder.

Fig. 10 (
Fig. 10 (a) Histograms of cell viabilities of irradiated U87MG tumor cells normalized with respect to those of the corresponding unirradiated cells for control, Gadovist, and sample.(b) Histograms of net cell deaths obtained by subtracting the normalized control cell death from those of sample and Gadovist.
Gd-chemicals applied to or proposed for GdNCT experimentsGd-chemical Delivery system (particle diameter in nm) Experimental type [in vitro a or in vivo (injection type inserted) Sk-Mel-28 cell culture; (ii) mice bearing Sk-Mel-28 tumor (I.T.) Yes (i) Tumor cell death; (ii) C-26 cell culture; (ii) mice bearing C-26 tumor (I.V.) Yes (i) Tumor cell death; (ii) tumor growth suppression better than Gdmicrocapsule (75-106 mm) Mice bearing Ehrlich ascites tumor (I.P., and SCC-VII cell culture No Higher accumulation than Gd-DTPA in all cells