A Perspective on Cell Tracking with Magnetic Particle Imaging

Magnetic Particle Imaging (MPI) is a new imaging modality that sensitively and specifically detects superparamagnetic iron oxide nanoparticles (SPIONs). Many labs have been developing cellular magnetic resonance imaging (MRI) tools using both SPIONs and fluorine-19 (19F)-based contrast agents for numerous important applications, including tracking of immune and stem cells used for cellular therapies. SPION-based MRI cell tracking has very high sensitivity, but low specificity. SPIONs produce negative contrast in MRI, or signal voids. SPIO is not directly detected by MRI, but indirectly through its relaxation effects on protons, therefore, it is not possible to reliably quantify the local tissue concentration of SPION particles and cell number cannot be determined. 19F based cell tracking uses perfluorocarbons (PFC) to label cells. The number of 19F atoms can be directly measured from 19F MR images and related to cell number. 19F MRI has high specificity, but low sensitivity. MPI cell tracking displays great potential for overcoming the challenges of MRI-based cell tracking allowing for both high cellular sensitivity and high specificity and quantification of SPIO labeled cell number. In this paper we describe nanoparticle and MPI system factors that influence MPI sensitivity and resolution, quantification methods and give our perspective on testing and applying MPI for cell tracking.


In Vivo Cell Tracking
Cellular therapies are now recognized as effective medicines for treatment of a specific subset of cancers as well as a growing list of autoimmune, degenerative, genetic, and infectious diseases. Cell therapy refers to the administration of immune cells (such as T cells or dendritic cells) which are used to treat cancer, and stem cells (such as mesenchymal stem cells or pluripotent stem cells) which have numerous potential applications including treatment of amyotrophic lateral sclerosis, diabetes, graft versus host disease, kidney disease, liver disease, multiple sclerosis, myocardial disease, osteoarthritis, Parkinson's disease, spinal cord injury and stroke. Numerous cell therapy clinical trials are underway.
Despite the immense promise of cell therapies, clinical results have been highly variable because of variations in cell source, preparation, and route of administration/implantation methodology. Many critical questions about the presence, numbers, persistence, and delivery of cells remain unanswered. For example, in cancer immunotherapy, the magnitude of an anti-tumor response is proportional to the quantity of antigen presenting cells that reach a target lymph node, [1][2][3] and therefore, it is crucial to know whether the injected cells have migrated to the target, and how many. In stem cell therapy, the survival and persistence of cells at the implant site informs on whether a patient may need repeat dosing or other interventions. These types of questions must be answered to improve the safety and success of cell therapies and for it to realize its full potential.
In vivo cellular imaging tools have the potential to answer these questions. The ideal imaging modality for monitoring cell therapies would be non-invasive, non-ionizing, sensitive enough to allow detection and quantification of a few hundred cells, specific and, importantly, quantitativeproviding a measure of cell number.
MRI has been widely used for in vivo cell tracking. Cellular MRI uses contrast agents for labeling specific cells, thereby enhancing their detectability. 4,5 The most often used agents for cell tracking with MRI are magnetite (Fe3O4)-based superparamagnetic iron oxide nanoparticles (SPIONs). The presence of SPIONs in cells causes a distortion in the magnetic field and leads to abnormal signal hypo-intensities in iron-sensitive images. 4 Areas containing SPION-labeled cells therefore appear as regions of low signal intensity on MRI images, creating negative contrast.
The iron label has minimal impact on cell function or phenotype at a wide range of iron loading levels. 14 This technique is highly sensitive, permitting the imaging of single cells in vivo 15 , under ideal conditions. There are, however, several limitations of iron-based MRI cell tracking. The first is low specificity due to other low-signal regions in MR images; such as the air-filled lungs or a region of hemorrhage. Although ultra-short echo time imaging methods have been developed for producing positive contrast from iron-labeled cells, these too have similar problems with specificity. Second, quantification of iron-induced signal loss is complicated since the measure of the signal void volume is not linear with the number of cells.
Fluorine-19 ( 19 F) MRI with perfluorocarbon (PFC) nanoemulsions to label cells has been also used for cell tracking. 16 18 The main limitation of 19 F cell tracking is low sensitivity; thousands of labeled cells per voxel are required. Fluorine sensitivity improves with higher field strengths and preclinical studies have reported in vivo detection of as few as 1000 cells per voxel. 19 However, the first human clinical trial at 3 Tesla (T) demonstrated a cellular detection limit between 1 and 10 million cells. 20 Magnetic Particle Imaging (MPI) is a new imaging modality that directly detects SPIONs. 21,22 MPI is built around a gradient magnet system. Two opposing electromagnets form strong gradient magnetic fields (in the order of T), and a field free region (FFR) is created in the position where these gradient fields cancel out. The gradient field (also known as the selection field) saturates the magnetization of all SPIONs except for those SPIONs at the FFR, which experience no magnetic field. The FFR is shifted over an imaging volume, by changing the current through the electromagnets, to produce an image. When the FFR traverses a location containing SPIONs, the SPIONs magnetization changes nonlinearly in response to a secondary sinusoidal excitation magnetic field (in the order of mT). This change in SPIO magnetization induces a voltage which is detected via a receiver coil, and the resulting signal can be assigned to the instantaneous FFR location to reconstruct the final image. 23 The voltages induced are linearly proportional to the number of SPIONs at the FFR location, enabling quantification of SPIONs.
MPI cell tracking has the potential to address many of the limitations presented by SPIONand 19 F-based cell tracking. First, since the MPI signal is only generated when the magnetic moments of the SPIONs rotate in response to the applied fields there is no signal from tissue. This imbues MPI with a positive "hot-spot" contrast that provides spatial localization without ambiguity. Second, MPI has high sensitivity. MPI's sensitivity derives from the direct detection of the electronic magnetization of SPION, which is 10 8 times larger than the nuclear magnetization of protons seen in MRI. 24 Third, the MPI signal is linearly quantitative with SPION concentration, and therefore the number of SPION labeled cells can be calculated. 25 The shortcomings of MPI include relatively low spatial resolution, compared to MRI, and the requirement that anatomical images must be acquired separately with a different imaging modality. A comparison of MSC tracking with iron and 19 F based MRI and MPI is shown in Figure 1.

MPI Sensitivity and Resolution
Particle sensitivity in MPI refers to the lowest mass of SPIO detected per imaging unit (e.g., ug/mm3). For MPI cell tracking, the aim is to optimize the cellular sensitivity, which refers to the lowest number of SPION-labeled cells detected per imaging unit (i.e. number of cells/mm 3 ).
The sensitivity of MPI depends on both nanoparticle and scanner specific factors. Nanoparticle factors include the strength of the nanoparticle magnetization (stonger magnetization improves MPI signal), the rate of SPION relaxation at the FFR, and the efficiency of the nanoparticle cell labeling (more iron per cell leads to higher sensitivity). Scanner specific factors include increasing the amplitude of the excitation field, 29 decreasing the gradient strength (at the cost of resolution), and signal averaging. For example, increasing the overlap fraction or combining multiple projections (i.e. 3D imaging) can improve cell detection. An important consideration is that cellular sensitivity in vitro may be enhanced compared to when cells are in vivo. This is due to dispersion of cells from the site of administration and additional factors such as breathing motion.
The current in vitro detection limit is reported at 200 stem cells labeled with ferucarbotran (1000 cells imaged with SNR = 5), 30 however these MPI cellular detection limits should be demonstrated and evaluated more closely. Song et al. 31 reports an in vivo cell detection limit of 250 HeLa cells injected subcutaneously in a bolus and using a custom MPI SPION (refered to as the Janus nanoparticle). There is no doubt that cellular sensitivity will improve with further advances in acquisition strategies and SPION design.
The resolution of MPI is driven primarily by SPION relaxation. Neel relaxation refers to the reversal of the SPION magnetic moment whereas Brownian relaxation refers to the physical rotation of the SPION, in response to the excitation magnetic field used in MPI. [32][33][34] The dominant relaxation mechanism depends primarily on the SPION size and the MPI resolution worsens with increasing Brownian relaxation. Theoretical modeling based upon the Langevin theory of SPIONs predicts that resolution improves with increasing core size. However, Tay et al. 33 have found that improved resolution with increasing magnetic core size follows the steady-state prediction up to approximately 25 nm when the effects of SPION rotational times become significant due to increasing Brownian relaxation. SPIONs above this size range experience increased drag, slowing their magnetization response and limiting resolution. Importantly, this work looked at single core mono-dispersed nanoparticles. Many of the SPIONs used for cell tracking are multi-domain, clustering nanoparticles with more complicated physics.
Resolution is also influenced by the interaction of the nanoparticle and the magnetic field gradient. Stronger gradients increase resolution but at the cost of sensitivity. The spatial resolution of MPI using currently available SPIONs is approximately 1 mm 25 and using optimized iron oxide nanoparticles, MPI resolution of 200 m has been demonstrated. 35 With further development, sensitivity and resolution should increase and significant improvements to cell detection limits will be possible.

SPIONs for MPI
As described above, both MPI sensitivity and resolution are closely related to the type of SPION, however the ideal SPIONs for MPI are still not known. In early MPI cell tracking, commercially available SPIONs used for MRI were evaluated, including ferucarbotran (Resovist, now manufactured as Vivotrax TM , Magnetic Insight Inc.), and ferumoxtyol. Ferucarbotran has improved MPI characteristics compared to ferumoxytol [36][37][38] , and ferucarbotran has been used in MPI studies of mice to detect mesenchymal stem cells, 36,39 neural stem cells, 40 neural progenitor cells, 30 pancreatic islets, 13 T-cells, 9 and macrophages. 38,41 Although widely used, ferucarbotran is no longer considered optimal for MPI because it has a bimodal size distribution, predominantly containing small cores ~ 5 nm in diameter (70%) with a small fraction (30%) of multi-core aggregates with an effective size of 24 nm. 42 The individual cores are too small to magnetize significantly and so the MPI signal predominately originates from the clustered multi-core structures. Optimizing SPIONs expressly for MPI is emerging as a powerful area of research and will be critical for improving sensitivity and spatial resolution. 33 Approaches to improve MPI sensitivity include increasing the fraction of these larger aggregates 42 or by fractionation of ferucarbotran. 43 Lastly, the synthesis of homogeneously distributed single-core SPIONs with optimized core diameters is being investigated. 44 MPI relaxometry is commonly used as a first step to characterize SPIONs. MPI relaxometry measures the net magnetization of SPIONs, by turning off the selection field and applying a negative magnetic field then a positive field, and back. SPIONs in a sample are driven from a negative magnetic saturation to positive, and vice versa. The output is the derivative of the Langevin function, also called the point spread function (PSF). The signal intensity, or height, of the PSF reflects the sensitivity of the SPION. The full-width half maximum (FWHM) relates to the spatial resolution of the SPION. 45 A narrower tracer response indicates superior spatial resolution and a greater signal intensity per mass of iron indicates superior sensitivity. We have evaluated several commercially available SPIONs (Figure 2). In our experience relaxometry has significant value for testing SPIONs prior to use for MPI as certain SPIONs that may not, in theory, seem suitable for MPI show surprisingly good sensitivity (ie. micron-sized superparamagnetic polystrene beads, MPIO). This can lead to effective labeling of cells, increased iron content in cells or faster labeling. As demonstrated in Figure 3 A-C, the addition of protamine sulfate and heparin significantly improves uptake of ferucarbotran by mesenchymal stem cells. However, several studies have shown that SPION-TA complexes may clump cells together and can lead to higher amounts of free iron in cell preparations. 54 It is also necessary to test whether TAs alter cell functionality or phenotype.
For example, certain TAs reduce the immunopotency of dendritic cells. 55 Importantly, we have recently observed that the MPI characteristics of SPIONs are altered by TAs. Figure 3C-G compares the MPI sensitivity and resolution of samples of ferucarbotran and ferucarbotran mixed with protamine sulfate/heparin. The addition of the TA leads to lower MPI signal and lower resolution images. This requires more study but may result from altered relaxation characteristics due to clustering of cells and iron cores. Therefore, while TAs may be used to achieve higher intracellular iron with the goal of improving MPI sensitivity, this gain may be counteracted by this effect.

Quantification of Cell Number
The ability to quantify cell number from MPI images represents a significant improvement for in vivo cell tracking with SPIONs. MPI signal (and the associated iron content) scales linearly with the number of cells present ( Figure 4A). We, and others, have previously described techniques to measure MPI signal from images and determine the iron content in a region of interest (ROI). 13,17,37,38,56,57 Briefly, MPI images are displayed in full-dynamic range and the signal is delinated at the FWHM by referencing a colour look-up table (CLUT). Total MPI signal for the ROI can be calculated by:

MPI signal = Mean signal in ROI (A. U. ) * ROI volume (mm )
A calibration is required to convert measured MPI signal to iron content. The relationship between MPI signal and iron content can be determined by measuring MPI signal from sample(s) of known iron content. A single reference phantom may be included for calibration. 13,38 Construction of a calibration line (MPI signal versus iron content) using multiple iron samples (5)(6)(7)(8)(9)(10) increases the accuracy of this measurement. These samples must be the same SPION (with similar iron content and similar volume) and the MPI images of these samples must be acquired using the same imaging parameters. Figure 4B shows MPI images and a calibration line made by our group for ferucarbotran. The estimation of cell number from MPI images requires a measurement of cellular iron loading. For a sample prepared with a known number of cells, the average iron content per cell can be measured using inductively coupled mass spectrometry (ICP-MS), spectrophotometry, or MPI. These measurements must be acquired using the same cells used for the in vivo imaging experiment. Subsequently, an estimation of cell number from in vivo MPI images can be calculated:

Caution/Challenges in the Quantification and Interpretation of MPI Data
When interpreting measurements of cell number from MPI images it is important to consider factors related to cell labeling which introduce uncertainty. First, the value for iron/cell measured from cell samples is an average value, some cells will contain more iron, some less.
Second, the amount of iron/cell achieved with each cell labeling experiment has a range of values, even for the same labeling protocol, depending on precise timing and culture conditions. Uniform SPION labeling is optimal for cell tracking but is difficult to achieve. Magnetic sorting of cells is one strategy to improve the uniformity of labeled cells but this only removes those cells with low amounts of iron.
There are also factors which may impact the quantification of cell number over time. As cells proliferate in vivo, SPIONs are diluted between cell progeny leading to underestimation of cell number and potential loss of cellular detection over time by MPI 4,58 . Several studies have demonstrated that cells which are nonproliferative, or slowly proliferating in vivo, for example stem cells, can be monitored over longer periods of time. 13,30,36,39 When SPION-labeled cells undergo apoptosis in vivo, they are taken up by phagocytes for clearance by the liver. 4,39 This leads to diminished MPI signal at transplant sites; 36,37 the measurement of MPI signal in the liver can be used to provide a measure of labeled dead cells. 13,18,39 In the interim, bystander labeling of phagocytes may contribute to false MPI at the site of apoptosis. Lastly, the intracellular SPIONs which exist within endosomal compartments may be susceptible to degradation by lysosomes. 59,60 This degradation process can reduce the paramagnetism of iron oxide nanoparticles thereby altering their cellular MPI detectability. 61,62 This is a largely understudied phenomenon; however, it may be reasonable to assume that this effect depends on the biological inertness of the SPION and cellular lysosomal capacity.
When two sources of MPI signal are in close proximity, and are generated from tissues or samples which contain very different iron levels, the detection and quantification of the lower MPI signal can be challenging. This is illustrated in Figure 5. Here, cell pellets of bone-marrow dendritic cells (BMDCs) labeled with ferucarbotran containing 2.5x10 4 or 5.0x10 5 cells were imaged at a distance of 2 cm apart. Samples with equal numbers of SPION-labeled cells could be discerned but when the 5.0x10 4 cell sample is imaged alongside the 2.5x10 4 cell sample the MPI signal from the sample with fewerr cells (or, less iron) is hidden by the strong signal from the sample with more cells. This presents a major challenge for studies where the location of MPI signal is unknown, particularly for in-vivo cell tracking. Proposed solutions for overcoming this challenge include increasing gradient strengths for improved resolution, or using post-processing imaging techniques such as deconvolution filters. This requires further investigation, but is an important consideration for improving cell detection limits. This issue can also arise from unwanted sources of MPI signal. For example, MPI signal is sometimes detected in regions of the digestive system of mice, owing to the presence of iron in mouse feed ( Figure 1C) or at injection sites in mice, resulting from dried blood in the form of hemosiderin (unpublished results). Both of these MPI signal artifacts can be present in mice when no SPION labeled cells are present. If they are close to, or colocalize with, SPION-labeled cells it can be challenging to detect the region of interest.