Targeted gadofullerene for sensitive magnetic resonance imaging and risk-stratification of breast cancer

Molecular imaging of cancer biomarkers is critical for non-invasive accurate cancer detection and risk-stratification in precision healthcare. A peptide-targeted tri-gadolinium nitride metallofullerene, ZD2-Gd3N@C80, is synthesised for sensitive molecular magnetic resonance imaging of extradomain-B fibronectin in aggressive tumours. ZD2-Gd3N@C80 has superior r 1 and r 2 relaxivities of 223.8 and 344.7 mM−1 s−1 (1.5 T), respectively. It generates prominent contrast enhancement in aggressive MDA-MB-231 triple negative breast cancer in mice at a low dose (1.7 µmol kg−1, 1 T), but not in oestrogen receptor-positive MCF-7 tumours. Strong tumour contrast enhancement is consistently observed in other triple negative breast cancer models, but not in low-risk slow-growing tumours. The dose of the contrast agent for effective molecular MRI is only slightly higher than that of ZD2-Cy5.5 (0.5 µmol kg−1) in fluorescence imaging. These results demonstrate that high-sensitivity molecular magnetic resonance imaging with ZD2-Gd3N@C80 may provide accurate detection and risk-stratification of high-risk tumours for precision healthcare of breast cancer.

P recision medicine requires accurate detection and characterisation of tumours for tailoring personalised therapies to improve healthcare of cancer patients. Non-invasive sensitive imaging of tumour markers is essential for accurate cancer detection and characterisation of tumour aggressiveness 1 . Magnetic resonance imaging (MRI) is a powerful clinical imaging modality that provides high-resolution three-dimensional images of soft tissues. Magnetic resonance (MR) molecular imaging or molecular MRI of cancer biomarkers can facilitate non-invasive cancer detection and characterisation, image-guided interventions and therapeutic efficacy assessment in cancer precision medicine 2,3 . High spatial resolution of MRI allows early detection of tumours as small as a few hundred cells, a few hundred microns in size 4 . Paramagnetic materials, including stable Gd(III) chelates and superparamagnetic iron oxide nanoparticles, are often used as contrast agents to increase T 1 and T 2 relaxation rates of water protons in the tissues of interest and to produce contrast enhancement for diagnostic imaging 5,6 . Despite the continuous efforts in developing novel MRI contrast agents, there is a serious lack of safe and effective-targeted contrast agents for high-sensitivity molecular MRI in clinical practice.
MR signal enhancement is determined by the extent of increase in the water-proton relaxation rate, which is proportional to both the concentration and relaxivity of the contrast agents. Currently, the design of targeted MRI contrast agents mostly focuses on increasing their local concentration around the molecular targets, e.g. using nanoparticles with high payloads of paramagnetic materials 7 . Various nanosized-targeted contrast agents, including superparamagnetic iron oxide, Gd(III) and Mn(II)-containing nanoparticles [8][9][10] , have shown promise in MR molecular imaging. However, because of their low relaxivity, a relatively large dose is required to produce detectable signal enhancement. High dose of nanoparticle contrast agents may cause unintended side effects due to their slow excretion and accumulation in normal tissues 11,12 . Recently, a new strategy of targeting abundant biomarkers in the tumour microenvironment using small peptide conjugates of clinical Gd(III)-based MRI contrast agents has been developed 4 . Although this approach facilitates faster excretion of the targeted contrast agents in molecular MRI 13 , a relatively high dose of these agents is still required to generate detectable signal enhancement.
Thus, the design and development of better contrast agents with high relaxivities is essential to significantly improve the sensitivity of molecular MRI and to reduce the local concentration and dose of the contrast agents, which will minimise potential dose-dependent toxic side effects. Recently, gadofullerenes have emerged as a novel class of paramagnetic materials with  [14][15][16][17] . For example, hydroxylated tri-gadolinium nitride metallofullerene Gd 3 N@C80 possesses an r 1 relaxivity that is~20 times that of conventional Gd-based contrast agents (GBCAs) 5,18 . In addition, Gd(III) ions are encapsulated in the fullerene cage, preventing the release of toxic free Gd(III) ions into the body 19 . The small size (ca. 1 nm) of gadofullerenes also allows for complete clearance from the body via renal filtration. Therefore, targeted gadofullerenes are promising targeted MRI contrast agents that can address the limitations and toxicity of the existing MRI contrast agents for safe and sensitive molecular MRI in clinical practice. Extradomain-B fibronectin (EDB-FN) is a marker for epithelial-to-mesenchymal transition (EMT), a biological process associated with tumour invasion, metastasis and drug resistance [20][21][22][23] . EDB-FN is highly expressed in the extracellular matrix of many types of aggressive human cancers 24,25 . Clinical evidence shows that EDB-FN overexpression is associated with poor prognosis of a variety of cancers [26][27][28] . Thus, it is a promising target for cancer detection and characterisation with molecular MRI. We have identified a small peptide ZD2 (Cys-Thr-Val-Arg-Thr-Ser-Ala-Asp) for specific targeting to EDB-FN 24 . We have shown that Gd(HP-DO3A) modified with linear ZD2 peptide (Thr-Val-Arg-Thr-Ser-Ala-Asp) can be used to characterise prostate cancer aggressiveness in MRI 29 .
In this work, we synthesise a high-relaxivity-targeted contrast agent by conjugating a small peptide ZD2 to hydroxylated Gd 3 N@C80, ZD2-Gd 3 N@C80, for sensitive molecular MRI of breast cancer. The targeted contrast agent has a superior T 1 relaxivity, about 20 times higher than the conventional Gd-based MRI contrast agents. The effectiveness of the agent for sensitive detection of aggressive tumours and risk-stratification is tested in multiple aggressive triple negative breast cancer (TNBC) and low-risk breast cancer models in mice. MRI with the targeted contrast agent at significantly reduced doses produces strong signal enhancement in aggressive TNBC tumours, not in slowgrowing low-risk breast tumours. The targeted contrast agent ZD2-Gd 3 N@C80 is effective for sensitive molecular MRI for the detection and risk-stratification of aggressive breast cancer.

Results
Gd 3 N@C80 was first oxidised with succinic acid peroxide and NaOH to introduce carboxyl and hydroxyl groups on the fullerene cage surface ( Fig. 1) 30 . MALDI-TOF mass spectrometric analysis of the hydroxylated Gd 3 N@C80 suggested an estimated structure of Gd 3 N@C80(OH) 18 (CH 2 CH 2 COOH) 6 . Some of the carboxyl groups in the hydroxylated Gd 3 N@C80 were then converted into amines, followed by reaction with maleimidoopfp, yielding maleimido-Gd 3 N@C80 for conjugation with thiolbearing ZD2 peptide (Fig. 1). An excess of ZD2 peptide was used to conjugate to maleimido-Gd 3 N@C80. The N-H and C-N peaks in Fourier transform infrared spectroscopy (FTIR) (Fig. 2a) indicated successful conjugation of the ZD2 peptide. The structure was also characterised with MALDI-TOF mass spectrometry, indicating approximately one ZD2 peptide was conjugated to each Gd 3 N@C80.
The peptide-targeted tri-gadolinium nitride metallofullerene ZD2-Gd 3 N@C80 showed complete water solubility (Fig. 2a), which is essential for further clinical development. The r 1 and r 2 relaxivities of hydroxylated Gd 3 N@C80 and ZD2-Gd 3 @NC80 were determined at 1.5 Tesla. The hydroxylated Gd 3 N@C80 had r 1 and r 2 relaxivities of 57.1 and 98.5 mM −1 s −1 per Gd ion(III), respectively (Fig. 2b). ZD2-Gd 3 @NC80 had superior r 1 and r 2 relaxivities of 223.8 and 344.7 mM −1 s −1 per molecule or 74.6 and 114.9 mM −1 s −1 per Gd(III) ion (Fig. 2c), respectively . The increased relaxivities of ZD2-Gd 3 @NC80 may be attributed to slower tumbling rate and increased rotational correlation time of the targeted agent due to the increased size after peptide conjugation. The r 1 relaxivity of ZD2-Gd 3 @NC80 is almost 20 times that of clinical Gd(III)-based contrast agents, including Gd-DTPA and Gd(HP-DO3A) 5 . The high r 1 relaxivity of the hydroxylated Gd 3 N@C80 is attributed to strong magnetisation of the hydroxyl protons on the cage by the encapsulated Gd (III) ions and rapid exchange rate of the protons with water protons in the surrounding bulk. This superior relaxivity is critical to improving the sensitivity of contrast enhanced MRI, especially T 1 -weighted MRI, for molecular imaging at low doses on most clinical scanners with relatively low magnetic field strengths (1.5 and 3 Tesla). At 7 Tesla, the r 1 relaxivities of hydroxylated Gd 3 N@C80 and ZD2-Gd 3 N@C80 were 24.68 and 24.78 mM −1 s −1 per Gd(III) ion, respectively (Supplementary Fig. 1). It appears that the slower tumbling rate and increased rotational correlation time of ZD2-Gd 3 N@C80 had less effect on improving the relaxivities at the high magnetic field strength. ZD2-Gd 3 N@C80 had an average diameter of 2.8 nm, as determined by transmission electron microscopy (TEM) (Fig. 2d) and dynamic light scattering (DLS) (Fig. 2e), which is smaller than the renal filtration threshold. This small size is necessary for rapid extravasation, target binding for effective molecular MRI and the elimination of the unbound agent from systemic circulation via renal filtration. EDB-FN expression was determined in three TNBC cell lines (MDA-MB-231, Hs578T and BT549) and three oestrogen receptor (ER)-positive breast cancer cell lines (MCF-7, ZR-75-1 and T47D) using quantitative real-time polymerase chain reaction (qRT-PCR). The mRNA levels of EDB-FN in all the TNBC lines were significantly higher than those in the ER-positive cell lines ( Fig. 3a and Supplementary Fig. 2). Western blot analysis also revealed abundant expression of EDB-FN in MDA-MB-231 tumours and negligible expression in MCF-7 tumours (Fig. 3b). In matrigel-based three-dimensional (3D) culture, MDA-MB-231 cells were able to form large spherical structures, indicating the aggressive nature of the TNBC cells (Fig. 3c). In contrast, MCF-7 cells formed smaller cell clusters with limited matrigel invasion, suggesting the low invasiveness of the cells (Fig. 3c). The MDA-MB-231 spheres incubated with ZD2-Cy5.5 showed strong fluorescence intensity under confocal microscopy, indicating high expression of EDB-FN protein and strong binding of the peptide probe in these spheres. In comparison, MCF-7 cells showed lower ZD2-Cy5.5 binding and weaker fluorescence intensity (Fig. 3c).
To further validate the in vivo targeting specificity of the ZD2 peptide, ZD2-Cy5.5 was intravenously injected at a dose of 0.5 µmol kg −1 into the MDA-MB-231 and MCF-7 tumour-bearing mice. Figure 3d shows the fluorescence images of ZD-Cy5.5 in the tumours and other major organs at 3 h after injection. Fluorescence intensity was greater in the MDA-MB-231 tumours than the MCF-7 tumours and normal tissues and organs.
Relatively high fluorescence signal intensity was seen in the liver and kidneys, because the probe is mainly excreted via these organs. EDB-FN expression and ZD2-Cy5.5 binding in the two tumour models were also verified by correlating immunofluorescence staining of EDB-FN and Cy5.5 fluorescence imaging (Fig. 3e). MDA-MB-231 tumour sections were rich in EDB-FN expression and showed strong Cy5.5 fluorescence, whereas little EDB-FN expression and ZD2-Cy5.5 binding were seen in the MCF-7 tumour sections. Microscopically, the aggressive MDA-MB-231 TNBC tumour sections displayed a higher cell density than the ER-positive MCF-7 tumours (Fig. 3f). Taken together, these results validate the strong positive correlation of EDB-FN expression with tumour aggressiveness and the specific binding of ZD2 peptide to the highly abundant EDB-FN in the MDA-MB-231 tumours.
We next tested whether molecular MRI with ZD2-Gd 3 N@C80 at 1 Tesla could detect the aggressive MDA-MB-231 tumours and differentiate the aggressive TNBC tumours from MCF-7 tumours in animal models. MR image acquisition was performed with mice bearing MDA-MB-231 and MCF-7 tumour xenografts before and after intravenous injection of ZD2-Gd 3 N@C80 at a dose of 1.67 µmol kg −1 or 5 µmol-Gd/kg, which is 20 times less than the standard dose of conventional clinical contrast agents, such as Gd-DTPA and Gd(HP-DO3A) 31 . Significant signal enhancement was observed in the MDA-MB-231 tumours for at least 30 min after the injection, while little enhancement was observed in the MCF-7 tumours or both tumours injected with non-targeted hydroxylated Gd 3 N@C80 (Fig. 4a, b). Co-injection of 1.67 µmol kg −1 of ZD2-Gd 3 N@C80 with 25 µmol kg −1 of ZD2 peptide significantly reduced the signal enhancement in MDA-MB-231 tumours due to competitive binding to the target. Quantitative analysis revealed that ZD2-Gd 3 N@C80 produced 39 to 45% increase of contrast-to-noise ratio (CNR) in the MDA-MB-231 tumours (Fig. 4b). The smaller size of ZD2-Gd 3 N@C80 facilitates rapid tumour extravasation, target binding, and clearance from circulation, and enables rapid sensitive detection of aggressive TNBC with high CNR. No significant difference was observed in signal enhancement in the normal tissues before and after contrast injection in all the tested groups, except for the excretory organs: kidneys and bladder ( Supplementary Fig. 3). Increased signal intensity in the bladder indicates the excretion of unbound ZD2-Gd 3 N@C80 via renal filtration.
The systemic retention of both ZD2-Gd 3 N@C80 and hydroxylated Gd 3 N@C80 was determined at 1 week post-injection using ICP-OES. The amount of Gd(III) in the body was near the detection limit of the ICP-OES. Both the agents had <0.5% of the injected dose remaining in the tissues and organs at 1 week postinjection (Fig. 4c). The potential release of the free Gd(III) ions was also tested by in vitro transmetallation assay with human serum. No transmetallation of ZD2-Gd 3 N@C80 (<0.12%) was observed ( Supplementary Fig. 4). Studies show that fullerene cages are highly stable and the metal ions are stably enclosed in these cages 18,30 . Metallofullerene cages have been reported to even withstand exposure to β-radiation 18 . The residual endogenous metal ions seen in our study could be the result of the weak complexation of the peptide with the endogenous metal ions. Consistent with previous reports 19,32 , the entrapment of the toxic Gd(III) ions in the fullerene cage prevents their systemic release and tissue interaction, which is critical for the safety of the Gd(III)-based contrast agents in consideration of clinical translation. These results suggest that ZD2-Gd 3 N@C80 has the potential to overcome the reported toxic side effects of the existing GBCAs, caused by the release and retention of free Gd(III) ions in the body.
The potential of ZD2-Gd 3 N@C80 for differentiating breast cancer aggressiveness was further assessed in mice bearing ZR-75-1, Hs578T, T47D and BT549 breast cancer xenografts. The ER-positive ZR-75-1 and T47D xenografts showed a very slow rate of growth and relatively smaller tumour sizes as compared with the Hs578T and BT549 TNBC tumours. At 7 T, injection of 20 µmol Gd/kg ZD2-Gd 3 N@C80 produced greater signal enhancement and significantly higher CNR increase in the fast-growing Hs578T and BT549 tumours than in the slow-growing ZR-75-1 and T47D tumours (Fig. 5). These results demonstrate that ZD2-Gd 3 N@C80 is an effective high-relaxivity-targeted contrast agent that can improve the sensitivity of molecular MRI for the detection and characterisation of aggressive breast tumours.
This study has demonstrated that the peptide-targeted gadofullerene with superior relaxivities can significantly improve the sensitivity of molecular MRI in detection and characterisation of aggressive breast cancer. We have shown that the superior relaxivity of ZD2-Gd 3 N@C80 markedly improved the sensitivity of MR imaging of the oncoprotein EDB-FN in aggressive breast cancer. MRI with ZD2-Gd 3 N@C80 produces robust signal enhancement in aggressive MDA-MB-231 TNBC tumours at a much lower dose (1.67 µmol kg −1 ) at 1 Tesla, a sensitivity comparable with that of fluorescence imaging of ZD2-Cy5.5 (0.5 µmol kg −1 ) in Fig. 3. The targeted contrast agent also produced strong enhancement in two other TNBC tumour models, not in slow-growing low-risk tumours at a high-field strength (7 Tesla) and reduced dose. These results have demonstrated the ability of the novel-targeted contrast agent for differentiating the fast-growing aggressive TNBC tumours from the slow-growing, non-metastatic and ER-positive breast cancers. Our work pioneers effective non-invasive differentiation between breast tumours of different aggressiveness using contrast enhanced MRI. EDB-FN is also highly expressed in other types of aggressive tumours, including prostate cancer 24 , head and neck cancer 33 and ovarian cancer 34 . Therefore, molecular MRI with ZD2-Gd 3 N@C80 can potentially be used for accurate detection and characterisation of a broad spectrum of aggressive cancers with high sensitivity and superior resolution. The low dose, rapid renal clearance and absence of potential release of free Gd(III) ions of ZD2-Gd 3 N@C80 are advantageous safety features for improving the safety profile of GBCAs in clinical use. Clinical translation of molecular MRI with ZD2-Gd 3 N@C80 has the potential to overcome the limitations of current imaging technologies and to significantly improve the accuracy of early detection and characterisation of high-risk breast cancer for precision healthcare of cancer patients. Synthesis of ZD2-Gd 3 N@C80. Gd 3 N@C80(OH) 18 (CH 2 CH 2 COOH) 6 (6 mg) and mono-Fmoc ethylenediamine (5 equiv) were dissolved in DMF, and then HBTU (5 equiv) and DIPEA (5 equiv) were added. The reaction was stirred at room temperature for 2 h. Piperidine/DMF (20%, v/v) was used to remove the protecting group of Fmoc. Afterwards, the product was precipitated in cold ether to obtain Gd 3 N@C80(OH) 18 (CH 2 CH 2 COOH) x (NH 2 ) y (3) of brown colour. (yield, 63%). Gd 3 N@C80(OH) 18 (CH 2 CH 2 COOH) x (NH 2 ) y (3 mg) was dissolved in DMF (5 mL) and excess maleimido-opfp (20 mg) was added. The reaction continued for 30 min before precipitating in cold ether to give maleimido-containing hydroxylated Gd 3 N@C80. ZD2-Cys peptide (sequence: Cys-Thr-Val-Arg-Thr-Ser-Ala-Asp) was synthesised in solid phase using standard Fmoc-chemistry. Excess ZD2-Cys and Gd 3 N@C80(OH) 18  Synthesis of ZD2-Cy5.5. ZD2 peptide (sequence: Thr-Val-Arg-Thr-Ser-Ala-Asp) was synthesised in solid phase using standard Fmoc-chemistry. After Fmoc removal with 10% piperidine, Fmoc-9-amino-4,7-dioxanonanoic acid (ChemPep, Inc., Wellington, FL, USA) was added to the peptide sequence. After Fmoc removal with 10% piperidine, the resin was washed with DMF/DCM and air-dried, and 10 mg of the dried resin was swelled in DCM for 1 h, followed by reaction with 3 mg Cy5.5-NHS ester (Lumiprobe Corporation, Hallandale Beach, FL, USA) in presence of 5 μL DIPEA. Reaction was stirred overnight at room temperature. Excess Cy5.5-NHS ester was removed by filtration and washing with DMF/DCM 10 mL three times. Peptides were cleaved off resin using TIPS, and precipitated in cold ether. The products were separated from ether by centrifugation at 4000×g. The final product was characterised by MALDI-TOF mass spectrometry ([M+1] + : m/z = 1473.76 (obsd); 1472.78 (calc.)). The product was lyophilised and reconstituted in 500 μL PBS. The concentration of the solution was characterised by measuring absorbance at 650 nm.
FTIR and relaxivity measurement. Infrared spectrum of ZD2-Gd3N@C80 was performed using the Cary 630 FTIR Spectrometer (Agilent Technologies, Santa Clara, CA, USA). Lyophilised hydroxylated Gd 3 N@C80 or ZD2-Gd 3 N@C80 was reconstituted to a serial dilution in water. The solutions were pipetted in NMR tubes (500 μL in each tube), and placed in a relaxometer (Bruker) at 1.5 Tesla. For relaxivity measurement at 7 Tesla, NMR tubes containing contrast agent solutions were bundled and placed in a mouse coil in a horizontal 7 Tesla Bruker scanner (Bruker Biospin Co., Billerica, MA). T 1 maps of the solutions were acquired using a previously reported method 29 . T 1 and T 2 values of each solution were measured. The r 1 and r 2 relaxivities of the contrast agent were calculated as the slope of the plot of 1/T 1 and 1/T 2 relaxation rates against the concentrations. Fluorescence imaging. To determine the distribution of ZD2-Cy5.5 in the major organs and tumours, 10 nmol ZD2-Cy5.5 was injected in tumour-bearing mice through the tail vein. At 3 h after injection, mice were killed. Tumours and organs were collected and imaged with CSi Maestro imaging system (Woburn, MA, USA) using the deep red filter sets (exposure time: 1000 ms).
Histological analysis. slice thickness = 2 mm; interslice distance = 1 mm) was then used to acquire the images of the tumours before and 10 min, 20 min and 30 min after contrast injection. Images were exported into DICOM data, which were then processed and analysed using Matlab (Natick, MA, USA). The CNR ratio of tumours in the images was calculated as the difference between tumour mean intensity minus muscle mean intensity, divided by the noise. Three-dimensional images of mice were acquired using a gradient echo T 1 -weighted sequence with the following parameters: TR = 17 ms; TE = 6 ms; flip angle = 15°; FOV = 3.5 cm × 8 cm; matrix size = 128 × 512 × 16; slice thickness = 1.5 mm. Analysis of the change in signal intensity in muscle, heart, liver, kidney and bladder was performed using Matlab. MRI at 7 Tesla was performed on a horizontal Bruker scanner. T 1 -weighted spin-echo sequence with the following parameters was used: FOV: 3 cm; slice thickness: 1.2 mm; interslice distance: 1.2 mm; TR: 500 ms, TE: 8.1 ms; flip angle: 90°; average: 2; matrix size: 128 × 128. The images were similarly analysed as described above.
Biodistribution. The mice injected with the contrast agents were sacrificed 1 week post-injection. Tissue samples were collected, weighed and digested by 1 mL ultrapure nitric acid (EMD Millipore, Billerica, MA, USA) for 7 days. The digested sample (0.5 mL) was diluted to 5 mL with ultrapure water (Milli-Q, EMD Millipore). The solution was centrifuged and filtered using a 0.45 µm filter and the concentration of Gd(III) ions was determined using inductively coupled plasma-optical emission spectrometry (ICP-OES) on a 730-ES ICP-OES system (Agilent Technologies In vitro transmetallation assay. We performed in vitro transmetallation assay according to the protocol reported previously 29 . Briefly, ZD2-Gd 3 N@C80 (0.02 mM Gd/mL) was incubated with human serum for 2 h. According to previous reports 36,37 , transmetallation of Gd with Zn 2+ , Cu 2+ and Ca 2+ bound in serum proteins occurs rapidly and 2 h is sufficient to study the equilibrium of transmetallation. The mixtures were then centrifuged at 4000 rpm and 25°C for 150 min using CF-10 centrifugal filters (molecular weight cutoff 10 kDa). Metal ions in both the upper reservoir and the filtrates were quantified by ICP-OES. Transmetallation with Zn 2+ , Cu 2+ and Ca 2+ bound in serum proteins resulted in increase in free Zn 2+ , Cu 2+ and Ca 2+ in the filtrates. The degree of transmetallation of ZD2-Gd 3 N@C80 with Zn 2+ , Cu 2+ and Ca 2+ ions in serum was evaluated using the percentage of Zn 2+ , Cu 2+ and Ca 2+ ions filtered through the membrane, calculated as Transmetallation (%) = (concentration of ions in the filtrates)/(total ion concentrations before filtration) × 100. Human serum was used a control.
Statistical analysis. All the experiments were performed in triplicates unless stated otherwise. No estimation of sample sizes was performed. No randomisation or blinding was used in animal studies. Data are represented as mean ± s.e.m. Analysis of differences between two groups was performed using Student's t-test assuming equal variance, and the difference was considered significant if P < 0.05.
Data availability. Data available on request from the authors.