Lowering the cost of magnetic-resonance imaging (MRI) could revolutionize how doctors diagnose and screen for many diseases.
In a study published in the journal Science Advances, a researcher at Massachusetts General Hospital (MGH) and colleagues in Australia identify the missing piece needed to generate high-quality imaging using low-cost MRI scanners, which could expand the role of this powerful technology in medicine.
A typical MRI scanner carries a price tag of up to $3 million or more, which is why the machines are primarily found in imaging clinics and are typically unaffordable for hospitals in remote areas with small patient populations.
Finding a way to make MRI scanners more affordable has long been the objective of physicist Matthew Rosen, Ph.D., director of the Low-field MRI and Hyperpolarized Media Laboratory at the MGH Martinos Center for Biomedical Imaging.
“The focus of my lab is to deconstruct the MRI scanner,” says Rosen.
The cost of an MRI scanner is largely driven by its superconducting magnet, explains Rosen: The stronger the magnetic field it produces, the more expensive the machine.
A typical MRI machine generates a magnetic field of 1.5 Tesla (T), though increasingly machines reach 3 T.
However, a new generation of portable “low-field” MRI scanners that operate at 0.064 T and cost between $50,000 and $100,000 has recently become available.
While low-field MRIs are growing in popularity, radiologists frequently ask Rosen if an injectable contrast agent is available to improve the images they produce. Doctors occasionally inject patients with a contrast agent based on the heavy metal gadolinium before performing a conventional MRI to enhance image quality, though concerns about long-term toxicity now limit that practice.
For gadolinium to be used with low-field MRI, a doctor would have to administer 1,000 times more than the amount approved by the Food and Drug Administration (FDA), explains Rosen.
To solve the problem, physicists David Waddington, Ph.D., the lead author of the Science Advances study, and Zdenka Kuncic, Ph.D., both from the University of Sydney, suggested testing superparamagnetic iron oxide nanoparticles (SPIONs) as a contrast agent with low-field MRI.
SPIONs are safe and approved by the FDA for treating some cases of anemia, or iron-deficiency, but they have another desirable quality. “SPIONs essentially amplify low magnetic fields,” says Waddington, noting that SPIONs are 3,000 times more magnetic than conventional MRI contrast agents.
In the study, healthy lab rats were scanned with Rosen’s homemade ultralow field (ULF) MRI (0.0065 T), then injected with SPIONs and rescanned. A comparison of pre- and post-injection images shows a striking difference, with kidneys, livers, and other organs glowing brightly following administration of SPIONs.
While SPIONs need to be approved by the FDA for use as contrast agents, doctors can use them now “off label” with low-field MRI.
Rosen and Waddington believe the combination of portable low-field MRIs and SPIONs will bring this valuable imaging technology to emergency rooms, intensive care units, and doctors’ offices for routine screenings.
Waddington and Kuncic are also investigating the use of specially coated SPIONs that could allow MRI to be used for detecting malignant tumors. “This is an enabling technology that will make low-cost MRI a reality,” says Rosen.
Superparamagnetic iron oxide nanoparticles (SPION) are widely used for biomedical applications, including magnetic resonance imaging (MRI), magnetic particle imaging (MPI), magnetic fluid hyperthermia (MFH), separation of biomolecules, and targeted drug and gene delivery [1,2,3].
This widespread list of applications not only results from the magnetic properties of SPION, but also from the capability of synthesizing them in different sizes and shapes.
For all of the above applications, SPION should ideally have a high magnetization value, a size below 100 nm and a narrow size distribution [4, 5].
SPION are typically based on Fe3O4 and/or Fe2O3. They can be synthesized using various methods, such as co-precipitation [5, 6], thermal decomposition , sol–gel , microemulsion , hydrothermal , and electrochemical synthesis .
The co-precipitation technique is among the most successful, most commonly employed and most cost-effective methods for high-yield synthesis. However, strategies are needed to overcome the most important limitation of this method, i.e. the very broad particle size distribution of the resulting SPION mixture [5, 6].
In this study, we describe a straightforward, easily implementable and broadly applicable centrifugation protocol to obtain relatively monodisperse SPION from a polydisperse starting mixture prepared using the co-precipitation technique.
As a result of their refined size distribution, the obtained optimized SPION dispersions showed substantially improved performance in MRI, MPI and MFH compared to the crude starting formulation, as well as to commercial SPION products, such as Resovist® and Sinerem®.
In this context, it is important to keep in mind that not the centrifugation protocol per se, but the eventual development of a SPION formulation with a very well-defined size and with a very narrow size distribution (and its consequent more optimal use for diagnostic and therapeutic purposes) is the objective of our work.
Thus far, no systematic study has been published on SPION size-isolation via sequential centrifugation, and no systematic analysis is available in which the performance of five size-isolated SPION sub-fractions (and clinically/commercially relevant controls) is head-to-head compared in MRI, MPI and MFH setups.
SPION preparation and size-isolation
Prototypic citrate-coated SPION were prepared via the standard co-precipitation technique, under nitrogen atmosphere [5, 6] (see “Experimental” section for details).
Based on this highly polydisperse starting batch, which we refer to as the “crude sample”, five sequential rounds of centrifugation were performed to obtain much more monodispersed SPION subfractions.
To this end, as depicted schematically in Fig. 1, the crude sample was transferred into 1.5 ml Eppendorf tubes and centrifuged at 14,000 rpm for 20 min. The resulting 1 ml of supernatant was collected and referred to as the “C1 sample”.
Subsequently, 0.1 ml of the bottom compartment in the Eppendorf tube that contained the largest nanoparticle fraction was resuspended in water. The obtained dispersion was then again centrifuged, the top 1 ml was collected as the “C2 sample”, and the bottom 0.1 ml was again resuspended and re-centrifuged.
These steps were sequentially repeated to obtain five fractions of relatively monodisperse SPION samples. These fractions are referred to as C1–C5. The crude starting mixture, Resovist® and Sinerem® are referred to as C, R and S, respectively.
Multiple systematic experiments were performed to identify the optimal centrifugation speeds and times for obtaining monodispersed SPION with well-defined sizes. The optimum conditions for size-isolation are presented in Fig. 1.
The production efficiencies of the size-isolated fractions C1, C2, C3, C4 and C5 were approximately 7, 29, 23, 18 and 11%, respectively.
Despite the large number of previous publications describing the synthesis of iron oxide nanoparticles, the tools and technologies for their size separation are relatively limited. Techniques employed to control average particle size and polydispersity can be based on the use of magnetic/electric fields, porous media, and mass- and density-based purification [12,13,14].
Fortin and colleagues for instance synthesized citrate-coated nanocrystals of maghemite and cobalt ferrite by alkaline co-precipitation, and size-sorted the nanoparticles by successive electrostatic phase separation .
Magnetic field-flow fractionation (MFFF) uses a homogeneous external magnetic field applied orthogonal to the direction of flow, to achieve efficient separation of particles .
Nonmagnetic size-exclusion chromatography (SEC) is another frequently used method for size separation of iron oxide nanoparticles. The fractions separated by SEC and MFFF have similar size distributions.
However, MFFF is faster and has a higher capacity [12, 16]. In addition to the above techniques, differential magnetic catch-and-release (DMCR) has recently been established to size-sort magnetic nanoparticles. DMCR, like MFFF, relies on an external magnetic field to separate magnetic species .
High-gradient magnetic separation (HGMS) is a column flow method used to isolate iron oxide nanoparticles from a nonmagnetic medium . Capillary electrophoresis (CE) is used for the separation of colloidal nanoparticles in an electric field.
CE requires specialized equipment, because of the high electric field. Electrical field-flow fractionation (ElFFF) separates iron oxide nanoparticles based on their size and electrophoretic mobility but without the drawbacks of CE [12, 16].
As compared to the above techniques, the here presented centrifugation method is somewhat more time- and labor-intensive, but it is also easier to perform and more broadly applicable, because it does not require specialized equipment.
Almost all SPION formulations were found to be biocompatible. Additional file 1: Figures S2–S4 document the observed cytotoxicity for the crude, C1–C5, Resovist® and Sinerem® samples studied by XTT, LDH and ROS assays. XTT analysis at iron concentrations of 0.1 and 1.0 mM showed no significant differences in the viability of NIH3T3 cells upon incubation with the samples C1–C5 as compared to Resovist® and Sinerem®.
Interestingly, at iron concentrations of 5 and 10 mM, XTT-based viability assessment indicated that all monodispersed samples except for C1 had an even higher biocompatibility than Resovist® and Sinerem® (Additional file 1: Figure S2).
The XTT findings were confirmed using the LDH assay (Additional file 1: Figure S3). At iron concentrations of 0.1 and 1 mM, no changes in NIH3T3 membrane damage were noted for C1–C5 as compared to Resovist® and Sinerem®, while at iron concentrations of 5 and 10 mM, LDH values (and membrane damage) were lower than for Resovist® and Sinerem® (again except for the smallest-sized batch C1).
In line with this, analysis of ROS production in NIH3T3 cells showed that there was no significant change in the ROS content of cells exposed to the monodispersed samples C1–C5 as compared to the crude sample, Resovist® and Sinerem® (Additional file 1: Figure S4).
Together, these results demonstrate that all monodispersed samples except for C1 have negligible toxicity. The higher cytotoxicity associated with the smallest particles is assumed to result from a more rapid and more extensive cellular uptake, as well as from a relatively larger surface area [19,20,21].
SPION stability in physiological media
All size-isolated SPION samples showed excellent stability in DI water (see columns 4 and 5 of Additional file 1: Table S1; demonstrating stable dispersion up to 6 months). This can be attributed to the highly negatively charged surface of the SPION.
All SPION formulations also showed high colloidal stability in physiological media, i.e. in fetal bovine serum (FBS) and in bovine serum albumin (BSA). The monitoring of the samples by visual inspection up to 24 h implied the absence of aggregation of SPION (see Additional file 1: Figures S5a and S6a).
In line with this, the hydrodynamic diameters and PDI obtained using DLS for 2, 6 and 24 h of incubation in physiological media did not show significant changes in size and size distribution (see Additional file 1: Figures S5b, c, S6b, c and Table S1).
In good agreement with our findings, Yu et al. synthesized two different types of SPION with different surface coatings: tetramethylammonium hydroxide-coated SPION (T-SPION) and citrate-coated SPION (C-SPION). The C-SPION showed robust stability in biological media, while T-SPION aggregated rapidly in all media evaluated .
Magnetic resonance imaging
All SPION samples showed excellent performance as contrast agent for magnetic resonance imaging (MRI).
Figure 4 and Additional file 1: Figures S8–10 show T1- and T2-weighted MR images and quantification of key MRI parameters for the crude, C1–C5, Resovist® and Sinerem® samples [i.e. relaxivities (r1, r2), relaxation rates (1/T1, 1/T2) and relaxivity ratios (r2/r1)]. Figure 4 indicates that all newly prepared samples, i.e. both the monodisperse and the polydisperse SPION, have transverse relaxivities (r2) greater than Resovist® and Sinerem®. Interestingly, while the crude starting mixture and Resovist® were both highly polydisperse, the r2 value of the former was found to be two times higher than that of the latter.
After sequential centrifugation, the r2 values of the monodisperse SPION gradually increased up until the third round of centrifugation. The C3 sample with 13.1 ± 2.2 nm core size possessed the most optimal MRI capabilities, with an r2 value of 434 mM−1 s−1. It produced 3.3 and 5.5 times more contrast in T2-weighted imaging than Resovist® (130 mM−1 s−1) and Sinerem® (79 mM−1 s−1), respectively. A number of studies have demonstrated that the core size, the size distribution and the magnetization of SPION are key factors influencing the transverse relaxation rate (1/T2) [15, 30]. The trend for the r1 values for the samples C1–C5 was found to be similar to that observed for the r2 values.
The efficiency of a T2 contrast agent relies on the r2/r1 ratio in addition to the r2 value . In this context, it is important to note that for all size-isolated samples, it can be concluded that there is a specific enhancement of the r2/r1 ratio in comparison to Resovist® and Sinerem® (Additional file 1: Figure S10), confirming the suitability of these samples for T2-weighted MR imaging.
Saraswathy and colleagues synthesized citrate-coated iron oxide nanoparticles with a similar coating and with a similar core size as C3 sample. They employed this SPION formulation for in vivo magnetic resonance imaging of liver fibrosis. The values for r1 and r2 were 2.69 and 102 mM−1 s−1, respectively .
Comparing the r2/r1 value of their formulation (i.e. 37.9) to that of our C3 sample (i.e. 84.4) exemplifies the usefulness and the potential added value of our sequential size-isolation protocol.
Smolensky et al. investigated the effect of multiple parameters, including particle size and shape, temperature and magnetic field strength, on the longitudinal and transverse relaxivities of iron oxide nanoparticles. According to their findings, r2 values increased linearly with increasing core size (from 4.9 to 18 nm), while r1 values remained relatively constant for particles with core sizes larger than 8 nm .
Surface coating and nanoparticle aggregation are also highly important parameters. Blanco-Andujar and coworkers studied the evolution of r2 with SPION aggregate size . In case of small clusters, nanoparticles are homogeneously dispersed in water and protons can readily diffuse between the magnetic cores.
Under these conditions, r2 values gradually increase with hydrodynamic diameter (up to approx. 80 nm). At a size of 80–90 nm, there is no further increase in r2. If the size exceeds 90 nm, r2 values start to decrease with increasing size, due to reductions in surface accessibility and proton exchange rate.
This trend is in line with our results, showing reductions in r2 values when the hydrodynamic diameter goes beyond 70 nm (r2 values for C4 and C5 are 398 and 350 mM−1 s−1, respectively, as compared to 434 mM−1 s−1 for C3).
Magnetic particle imaging
SPION are important tracer materials for magnetic particle imaging (MPI). MPI is a novel and increasingly popular hot-spot imaging technique that can be employed to visualize magnetic nanoparticles with very high temporal and spatial resolution.
MPI is able to provide real-time 3D imaging information on the localization and concentration of magnetic nanoparticles, and it can be employed for multiple medical imaging applications .
The potential usefulness of MPI strongly depends on the availability of size-optimized SPION to generate high quality images. As a matter of fact, MPI contrast generation critically depends on both SPION size and size distribution, since both parameters strongly affect the magnetization response.
Resovist® was originally developed as a contrast agent for MRI. In recent years, it has also been extensively employed for MPI, because of its large magnetic moment. At the moment, Resovist® is the most extensively employed SPION formulation for MPI.
From TEM images, it is known that Resovist® mainly consists of particles with an average core diameter of 5.8 ± 2.5 nm, many of which are agglomerated in clusters (Fig. 2a). It is assumed that these aggregates, which are formed by small elementary particles, are responsible for its good MPI performance .
However, the MPI performance of Resovist® still leaves significant room for improvement. As result of this, in recent years, ever more scientists have started to work on the development of better SPION formulations for MPI [26, 36].
Figure 5a shows the MPI signal-to-noise (SNR) values of the different SPION formulations used in this study, obtained at the 4th harmonic frequency of the drive field. It also shows the full width at half maximum (FWHM) values, and the hysteresis loss determined from the point spread function (PSF) measurements.
To allow for a quantitative comparison, it is generally considered to be sufficient to read the SNR at one harmonic frequency. This is typically the 4th harmonic frequency (Fig. 5a). Additional file 1: Figure S11 shows the SNR values for other harmonic frequencies. To compare the MPI performance of the different samples, SNR values were normalized to the iron concentration inside the probe volume.
The normalized SNR values for C2 and C3 were found to be much higher than for all other samples. At the 4th harmonic frequency, the normalized SNR for C2 was 2.3 and 7.0 times higher than for Resovist® and Sinerem®, respectively.
In addition, FWHM and hysteresis loss analysis showed that C2 and C3 were almost as good as Resovist®. Lower FWHM and hysteresis loss values refer to a higher achievable spatial resolution and to a lower spatial displacement in MPI, respectively.
To exemplify the MPI imaging capabilities of our size-isolated SPION, we fabricated two phantoms. One was an E-shaped phantom (Fig. 5b), serving as a somewhat more complex structure, made up of single tracer-filled dots of 0.5 mm.
The other phantom was V-shaped (Additional file 1: Figure S12a), and consisted of single dots with a diameter of 0.5 mm with an increasing distance between them (2, 3, 4, 5 and 6 mm). Both phantoms were filled with the crude starting mixture, with the C2 sample and with Resovist®, making sure that the iron concentrations were identical. Figure 5c and
Additional file 1: Figure S12b show the line profiles of the voxel intensities along the red marked lines for the E and V phantoms, respectively. It can be seen that the lowest and the highest intensities are obtained with the crude and the C2 sample, respectively. The C2 sample produced signal intensities more than two times higher than those of Resovist®.
From the MPI parameter analysis as well as from the MPI phantom experiments it can, therefore, be concluded that the C2 (and to a lesser extent also the C3) formulation is a useful alternative for Resovist® and suitable contrast agent for MPI.
More information: High-sensitivity in vivo contrast for ultralow-field magnetic resonance imaging using superparamagnetic iron oxide nanoparticles, Science Advances (2020). DOI: 10.1126/sciadv.abb0998