Advances in using MRI probes and sensors for in vivo cell tracking as applied to regenerative medicine

Magnetic resonance imaging (MRI; see Box 1 for a brief history on its development) is a non-invasive imaging technique that allows the visualization of the internal structures of the body in health and disease; it has thus been used as a diagnostic tool, with a wide range of medical applications, for more than 30 years. The principle of MRI is based on manipulating the magnetic properties of the protons and neutrons contained in atomic nuclei present in the body (most commonly, those found in the atoms of hydrogen). The motion of these nuclei produces a small magnetic moment (see Box 2 for a glossary of terms). When a body is placed in the magnetic field of the MRI scanner, the magnetic moment of these nuclei aligns with the direction of the magnetic field. A radiofrequency (RF) pulse is then applied to the body in the scanner, which excites the nuclei such that there are transitions between lower and higher energy spin states. Once the RF pulse is given, the nuclei return to their equilibrium state (a process called relaxation), releasing their absorbed extra energy and emitting an RF signal. This signal is detected by the scanner’s RF coils and is then used to generate a detailed image of the body’s tissues. By using MRI contrast agents (see the following sections, Boxes 2 and 3, and Table 1 for more details), the contrast of this image, and so the visibility of specific body structures, can be improved. Subsequent advancements in this field made it possible to use these contrast agents to label specific cell types and thus to monitor cells at the molecular level.

The application of MRI to cellular imaging in vivo has proved particularly useful in the field of regenerative medicine research, where it allows the tracking of engrafted cells and the monitoring of their physiological responses in a non-invasive manner. Over the past two decades, stem cells have been increasingly used as potential therapies for different disease conditions, particularly those in which cell replacement can restore the normal function of tissue or organs subsequent to their damage or degeneration. For example, as reported in the NIH public clinical trials database (http://www.clinicaltrials.gov; accessed 26 January, 2015; only open studies included, unknown status excluded), 1502 clinical trials at different phases are currently using stem-cell-based therapies to treat various disease conditions, e.g. myocardial infarct, neurodegenerative diseases and autoimmune diseases.

Based on the increasing numbers of cell-replacement therapies, it has become imperative to monitor non-invasively the engraftment of cells in vivo to determine the overall safety and efficacy of these approaches. For example, two FDA-approved cord blood products, ‘Hemacord’ (manufactured by New York Blood Center, Inc.; www.fda.gov; Submission Tracking Number: BL 125397/0) and ‘HPC-Cord Blood’ (manufactured by Clinimmune Labs, University of Colorado Cord Blood Bank; www.fda.gov; Submission Tracking Number: BL 125391/0) are being used for hematopoietic stem cell replacement therapies. Both cell therapies are systemically delivered, non-specific, and rely on the engraftment of an extremely large number of cells (recommended minimum dose: 2.5×10⁷ nucleated cells/kg body weight), with the assumption that enough cells will find their way to the target sites. Only non-invasive imaging renders it possible to evaluate the homing of such cells in vivo, thereby confirming the efficacy of such a therapeutic approach.

In certain conditions, such as in neurodegenerative disorders and in type 1 diabetes, poor engrafted cell survival, due to a hostile host tissue microenvironment or to inadequate nutrient support (Robberecht and Philips, 2013; Thomas et al., 2013), is a major barrier to the success of stem-cell-based therapies. Thus, the capacity to determine, in real time, the physiological state of engrafted cells, in terms of their gene expression and enzymatic activity, and to monitor their cell survival, will enable clinicians to better assess the therapeutic outcome of such treatments. This might, in turn, lead to the development of more-effective therapies.

MRI is an ideal tool for the in vivo tracking and ‘sensing’ of engrafted cells because of its ability to image deep inside tissue and to gather accurate anatomical and physiological information with high temporal resolution and sensitivity (Srivastava and Bulte, 2014). MRI could also be used to monitor alterations in cell function, tissue damage and changes in the dynamics of the biological processes that are associated with certain diseases (Haris et al., 2014; Yoo and Pagel, 2006). This use of MRI for non-invasive cell tracking first emerged from the use of MRI to label immune cells (Bulte et al., 1992; Bulte et al., 1993), and was followed by the first clinical application of MRI cell tracking to label and follow the fate of anti-tumor dendritic cells, used as cancer vaccines (de Vries et al., 2005).

In recent years, great progress has been made in the development of novel MRI sensors to monitor the different cellular functions of engrafted cells. In this Special Article, we describe recent advances in the development of MRI probes and sensors that are used for cell tracking and for detecting cellular functions in vivo. We also discuss the limitations of these approaches, their associated safety issues and the opportunities for their further improvement.

Contrast agents (see Table 1) are exogenous compounds that are administered intravenously or orally prior to MRI in order to improve the visibility of body structures for clinical diagnostic purposes. These agents change the relaxation rate of protons in water, creating a change in signal on MRI (see Box 3). However, contrast agents can also be used to pre-label cells ex vivo before transplantation, which is the most commonly used approach in MRI-based cell tracking.

There are different ways to incorporate contrast agents into living cells, such as by, for example, the use of transfection agents (Frank et al., 2002) and the use of translocation peptides. In this section, we discuss the main types of magnetic resonance (MR) contrast agents, how they function and their applications in clinical settings, as well as in experimental cell-tracking and regenerative approaches.

Paramagnetic MR contrast agents (Table 1) are widely used in clinical MRI. Gadolinium (III) (Gd³⁺) chelates (see Box 2) are the most effective paramagnetic contrast agents, owing to their seven unpaired electrons. The unpaired electrons of Gd³⁺ create a magnetic moment that increases the T1 of the surrounding water proton spins, creating ‘positive’ contrast on a T1-weighted scan (see Box 3). As a research tool, Gd³⁺ has been used to label and track different types of stem cells, such as hematopoietic progenitor cells, monocytic cells, endothelial progenitor cells and mesenchymal stem cells in cell transplantation studies in small animals (Agudelo et al., 2012; Guenoun et al., 2012; Hedlund et al., 2011). Because they are not nanoparticles, the cellular uptake of Gd³⁺ chelates occurs by pinocytosis (a non-specific form of endocytosis in which small particles present in the extracellular fluid are internalized into cells) or via electroporation (see Box 2). However, overall, the low sensitivity of these contrast agents and their low uptake by cells is one of the main barriers to cell labeling with Gd³⁺. Different methods, e.g. transfection using transfection agents (lipofectin and lipofectamine) (Rudelius et al., 2003) or coupling of the contrast agent to a membrane-translocation peptide (13-mer HIV-tat peptide) (Bhorade et al., 2000), can be used to increase the uptake of paramagnetic chelates. In the last few years, several advanced paramagnetic contrast agents with improved uptake efficiency, e.g. Gd hexanedione, liposomal-based Gd nanoparticles, and gadofluorine (Nejadnik et al., 2012b), have been developed and successfully used for cellular imaging (Ghaghada et al., 2009; Tseng et al., 2010).

However, the major limitation to the use of Gd³⁺ contrast agents is the long-term presence of Gd³⁺ in cells, which might lead to toxicity, either to the cells themselves or to the surrounding host tissue. This has been reported in the case of Gd³⁺ chelates that are not rapidly cleared in patients with impaired kidney function, causing severe fibrosis and even death (Thomsen, 2006). In addition, in stroke models, as for example in a rat model of middle cerebral artery occlusion (MCAO), transplantation of neural stem cells labeled with gadolinium-rhodamine dextran (GRID) did not significantly improve the therapeutic outcome. To the contrary, T2-weighted MRI of MCAO rats monitored for 1 year revealed that the presence of GRID-labeled cells slightly increased the infarct size in the treated rats when compared with MCAO rats that were not treated with cells, whereas the lesion size was decreased by 35% in animals that were treated with cells that did not have gadolinium labeling (Modo et al., 2009). The exact mechanism of gadolinium toxicity is not known, but dechelated (free) gadolinium is believed to be taken up in the bone marrow, where it might interfere with metal ion hemostasis of bone marrow stem cells.

A different class of paramagnetic contrast agents is represented by the so-called PARACEST (paramagnetic chemical exchange saturation transfer) agents (see Boxes 2, 3; Table 1) (Zhang et al., 2001). Because different metals induce different chemical shifts (see Box 2 for a glossary of terms) in resonance frequencies, PARACEST MRI can be used to track two different types of cell populations simultaneously by labeling them with two different PARACEST agents (Aime et al., 2005). As further proof-of-principle, the two PARACEST agents Yb- and Eu-HPDO3A have been successfully used in vivo to track two cell populations (macrophages and melanoma cells) (Ferrauto et al., 2013). One of the limitations of PARACEST agents, however, is their intrinsically low contrast sensitivity; to detect sufficient contrast typically requires concentrations ranging from 1 to 10 mM. At such concentrations, these paramagnetic metal agents might become toxic when present in the body for prolonged periods, as in the case for Gd³⁺ described above.

In addition to Gd³⁺, manganese (Mn II) is another potentially useful positive contrast agent for T1-weighted MRI (see Box 3 and Table 1). The kinetics and behavior of Mn²⁺ ions in the cell mimic those of calcium (Ca²⁺) ions, because Mn²⁺ ions enter cells through ligand- or voltage-gated Ca²⁺ ion channels (Narita et al., 1990). Therefore, Mn²⁺-enhanced MRI (MEMRI) has been used to study neuronal activity and to visualize neuronal connectivity in different animal models (Silva and Bock, 2008). Few reports exist concerning the use of manganese-based contrast agents in cell tracking. One study used MnCl₂ as a contrast agent for the labeling of lymphocytes as proof-of-principle (Aoki et al., 2006), but no in vivo studies using this approach have been performed. By contrast, manganese oxide (MnO) nanoparticles have been used as an MR contrast agent to detect cells in vivo. One of the advantages of using these nanoparticles in vivo is that they can be combined with superparamagnetic iron oxide (SPIO; discussed below, and see Box 2) particles to detect two cell populations with differential cell contrast (positive versus negative) simultaneously (Gilad et al., 2008a). More recently, silica-coated MnO nanoparticles have been developed as a positive T1 contrast agent for the labeling and MR tracking of mesenchymal stem cells (Kim et al., 2011).

SPIO nanoparticles have a much stronger MR relaxivity and higher sensitivity compared to paramagnetic agents (see Table 1), which makes them generate an MRI signal that is strong enough to visualize a small number of cells (Fig. 1). The presence of thousands to millions of magnetically aligned iron atoms in the SPIO core enables sensitive detection after the cells internalize these nanoparticles. Iron particle size varies from 10–50 nm [very small or ultra-small SPIO (VSOP and USPIO, respectively)] to >1 μm [micrometer-sized iron oxide (MPIO)] (Bulte and Kraitchman, 2004). SPIOs were first utilized to label and track transplanted cells in the rat brain (Hawrylak et al., 1993; Norman et al., 1992). Over the years, iron oxide particles have been developed that have higher relaxivities and more efficient intracellular labeling (Nkansah et al., 2011; Tang and Shapiro, 2011). Non-phagocytic cells cannot internalize SPIO nanoparticles; the most commonly used approach to label such cells is through the use of transfection agents, as originally demonstrated for magnetodendrimers (see Box 2 for more details) (Bulte et al., 2001).

As an example of their clinical use, SPIO nanoparticles have been used to track, using MRI, the homing of injected mesenchymal stem cells to the central nervous system (CNS) of individuals with multiple sclerosis (MS) and amyotrophic lateral sclerosis (ALS) (Karussis et al., 2010). This approach is particularly valuable because SPIO nanoparticle labeling does not affect the differentiation capacity (Nejadnik et al., 2012a; Wang et al., 2011), viability or function (Richards et al., 2012) of transplanted stem cells, except for chondrogenesis by mesenchymal stem cells, where it interferes with forming the extracellular matrix of cartilage (Bulte et al., 2004; Kostura et al., 2004).

In the past, only two clinical-grade, iron-oxide-based agents have been developed and approved for MRI of the liver: (1) SPIO nanoparticles coated with dextran as a polysaccharide (i.e. ferumoxides, also known as Endorem^® in Europe and Feridex^® in the USA) and (2) SPIO nanoparticles coated with low molecular weight carboxydextran (i.e. ferucarbotran, also known as Resovist^®). Although these SPIO formulations have been used for several clinical studies in the past, they are no longer manufactured because of economic considerations (Bulte, 2009). The agents never sold well for their FDA-approved application (i.e. detection of liver tumors) and, over time, abdominal MRI techniques have become more advanced to detect those masses without the need of administering contrast.

A limitation to the use of SPIO nanoparticles is their occasional extracellular deposition in tissues, either by active exocytosis (Cromer Berman et al., 2013) or passive release through the death of transplanted cells (Terrovitis et al., 2008). Deposited iron particles are scavenged by macrophages, which can then generate a false signal on MRI. As a result, the clinical application of these nanoparticles is limited to their short-term use and includes the MR-guided delivery of cells in real time (Barnett et al., 2007; Kraitchman et al., 2003) and the subsequent immediate monitoring of their homing and engraftment. As such, the use of these iron nanoparticles is restricted to assessing the acute retention of labeled cells and their short-term distribution in the body.

Fluorine (¹⁹F) is the naturally abundant isotope of fluorine, with a nuclear magnetic resonance (NMR; see Boxes 1, 2) sensitivity of 83% compared with hydrogen (¹H). It is also not radioactive, in contrast to the ¹⁸F isotope, which is used in positron emission tomography (PET; see Box 2) imaging. ¹⁹F MRI has emerged as a novel technology for not only tracking transplanted immune cells, but also endogenous macrophage-type cells that are present in several inflammatory disorders, such as autoimmune myocarditis and inflammatory bowel disease (Ahrens and Bulte, 2013; Ahrens and Zhong, 2013; Temme et al., 2012). Although SPIO nanoparticles are currently the most commonly used agent for MRI cell tracking (Cromer Berman et al., 2011), the relationship between the contrast, free iron and the concentration of SPIO is non-linear, and, therefore, it is difficult to quantify (Lebel et al., 2006; Rad et al., 2007). In addition, the SPIO-induced negative contrast is not truly specific and could be difficult to interpret at times, particularly when other sources of low signal contrast are present, as occurs for example in blood vessels or when a patient is bleeding – following illness or injury – where blood iron accumulates and is converted to a negative iron oxide contrast agent. By contrast, ¹⁹F MRI provides a more accurate, unambiguous detection of labeled cells (given the lack of background signal). Moreover, the relationship between the concentration of the ¹⁹F and signal intensity is directly proportional and linear over a wide range of concentrations, and the signal can be quantified directly from the acquired images (Srinivas et al., 2007). The lack of a detectable background in the ¹⁹F signal in biological tissues leads to higher visibility of the target cells, much like ‘hot spots’ emerging from an empty background (Bulte, 2005).

Perfluorocarbons (PFCs), which have many fluorine atoms with identical chemical shifts, are most commonly used for ¹⁹F MRI cell tracking applications, and they include perfluoro-15-crown-5-ether, linear perfluoropolyethers and perfluorooctyl bromide (Kadayakkara et al., 2014; Ruiz-Cabello et al., 2011). PFCs, being lipophobic and hydrophobic, are formulated into stable nanoemulsions by high-energy sonication for cell labeling (Janjic and Ahrens, 2009). Similar to SPIO-labeled cells, cells are labeled with PFC nanoemulsions ex vivo and then injected into the body (Ahrens and Zhong, 2013). Of note is that PFCs with different chemical shifts can be used to visualize two distinct populations of cells simultaneously in vivo (Partlow et al., 2007). As for regenerative medicine, neural stem cells have been successfully labeled with PFCs, and then visualized following their implantation into the brain (Fig. 2) (Bible et al., 2012; Boehm-Sturm et al., 2011; Ruiz-Cabello et al., 2008). Previous studies have shown that PFC labeling has a minimal effect on cell viability and cellular functions, such as differentiation and proliferation, both in vitro and in vivo (Boehm-Sturm et al., 2011; Ruiz-Cabello et al., 2008; Barnett et al., 2011; Keupp et al., 2011; Partlow et al., 2007).

In addition to tracking cells, PFCs can be used to probe cell function by sensing intracellular oxygen tension. For example, one study combined cell-labeling technology with the oxygen-sensing ability of PFCs to measure the intracellular oxygen tension in brain tumor models (Kadayakkara et al., 2010; Zhong et al., 2013), based on the principle that O₂ molecules change the ¹⁹F T1 relaxation time. This study reported that changes in oxygenation following tumor treatment with chemotherapy and immunotherapy provided a reliable imaging biomarker for evaluating the efficacy of these anti-tumor treatments. This approach could potentially be used to assess possible hypoxia at the site of stem cell transplantation, and to determine its effect on stem cell survival. The promising preclinical data and safety profile of PFCs make ¹⁹F MRI ready for clinical translation; recently, the feasibility of using ¹⁹F imaging of cells in a clinical setting has been demonstrated (Ahrens et al., 2014).

There are, however, limitations to the amount of PFCs that can be incorporated into cells: their size, similar to MPIOs, means that they take up a significant amount of cytoplasmic volume. Thus, the population number of cells to monitor should be above a certain threshold for detection, and that might vary depending on cell type. The minimum amount of cells that can be detected using ¹⁹F-based imaging ranges from 10³ to 10⁵ cells (Ahrens and Bulte, 2013). ¹⁹F MRI also requires the use of separate coils for image acquisition, or use of a multinuclear detector coil, neither of which is readily available in many MRI centers. However, the resonance frequency of ¹⁹F is close to that of ¹H; therefore, existing ¹H coils could be adapted for use as dual-tuned coils, capable of imaging both ¹H and ¹⁹F nuclei.

In 2014, a novel ‘hot spot’ highly shifted proton (HSP) MRI technique for cell tracking was described that directly detects dysprosium (Dy)- or thulium (Tm)- 1,4,7,10-tetraazacyclododecane-1,4,7,10-tetramethyl-1,4,7,10-tetraacetic acid (DOTMA)-labeled protons inside cells (Bulte, 2014; Schmidt et al., 2014). The principle behind HSP MRI is that certain lanthanides (see Box 2 for a glossary of terms), in particular Tm and Dy, cause a chemical shift of proton resonance frequency, one that is far away (100 ppm) from the water peak (from which the MRI signal is normally collected). When the MRI excitation pulse is applied at that chemical shift, only cells that contain the Dy- or Tm-bound protons will resonate and generate a signal. A customized pulse sequence, termed ultra-short echo time (UTE) MRI, was used to take advantage of the dramatically shortened T1 value of the protons bound to the chelate (Schmidt et al., 2014). When human fibrosarcoma tumor cells were labeled with Tm-DOTMA by electroporation and implanted into the flank of nude mice, the tumor could clearly be detected as a ‘hot spot’ on HSP MRI, and the overlay with conventional T2-weighted MRI allowed its anatomical localization (Fig. 3). HSP MRI is thus somewhat comparable to ¹⁹F MRI, and has a lower approximate cell detection number of 1×10⁴ cells.

Beyond cell tracking, MRI now has the added potential of being used to monitor enzymatic activity, gene expression, metal ion homeostasis and cell death in vivo. In this section, we discuss the most recent advances in the development of MRI probes and sensors to monitor cellular function, and also discuss the potential applications of these approaches to regenerative medicine.

In several preclinical studies, stem cells have been genetically modified to overexpress growth factors and enzymes in order to enhance their therapeutic properties in different disease conditions (Li et al., 2014; Krakora et al., 2013; Liang et al., 2013a; Suzuki et al., 2008; Xie et al., 2007). However, changes in the expression or level of activity of endogenous enzymes can also be associated with different diseases, requiring imaging methods to be developed that can distinguish between the two conditions.

In recent years, several MRI-based sensors have been designed to measure the expression and activity of both endogenous and recombinant enzymes. Most of these studies have been performed in vitro as proof-of-principle, with few in vivo examples as of yet. These sensors can either be substrates of the enzyme of interest conjugated to a contrast agent, or a substrate that changes its chemical exchange saturation transfer (CEST; see Boxes 2, 3) properties upon enzymatic conversion (Fig. 4). The contrast of the MRI signal is then either decreased or increased when the enzyme is present. The enzyme’s substrate can generate T1-weighted (Chen et al., 2004; Duimstra et al., 2005; Louie et al., 2000), T2-weighted (Westmeyer et al., 2010), T₂*-weighted (Hingorani et al., 2013; Liu et al., 2011; Schellenberger et al., 2008), ¹⁹F (Matsuo et al., 2013) or CEST (Airan et al., 2012; Liu et al., 2011) MRI-based contrast.

β-galactosidase (β-gal; encoded by the lacZ reporter gene) is a commonly used enzyme for the detection of transplanted cells by histology. Here, the enzyme converts its substrate, X-gal, to a dense dark-blue-colored product that can be readily detected with conventional light microscopy. Hence, it is not surprising that it has been a widely used platform for the development of MRI-based probes to detect enzymatic activity. As demonstrated early on by Louie et al. (Louie et al., 2000), a galactose-blocked contrast agent can be unblocked by cleavage by cells expressing recombinant β- gal, ‘opening up’ the contrast agent to the nearby protons in water and making it a more effective T1 contrast. Mason and co-workers developed libraries of ¹⁹F-modified substrates for monitoring β-gal activity with ¹⁹F magnetic resonance spectroscopy (MRS) and MRI (Cui et al., 2004; Kodibagkar et al., 2006; Liu et al., 2007; Yu et al., 2012). Other enzymes (either endogenous or genetically engineered) that have been used as targets for responsive MRI probes include: creatine kinase (Haris et al., 2014), matrix metalloproteinase-2 (MMP-2) (Catanzaro et al., 2013; Matsuo et al., 2013), matrix metalloproteinase-9 (MMP-9) (Schellenberger et al., 2008), myeloperoxidase (MPO) (Rodríguez et al., 2010), nitroreductase (Matsuo et al., 2013), urokinase plasminogen activator (uPA) (Yoo et al., 2013), alkaline phosphatase (Westmeyer et al., 2010), cytosine deaminase (Liu et al., 2011), β-glucuronidase (Duimstra et al., 2005), acetylCoA synthetase (Bastiaansen et al., 2013), carboxypeptidase G2 (Jamin et al., 2013; Jamin et al., 2009), transglutaminases (Tei et al., 2010), protein kinase A (PKA) (Airan et al., 2012; Shapiro et al., 2009), caspase-3 (Perez et al., 2002) and others (Razgulin et al., 2011) (see Table 2). Many of these enzymes are associated with different disorders and can serve as disease biomarkers. For instance, a change in the level of MMP can lead to disorders of the growth plate and can contribute to altered skeletal development, to cardiovascular disease, arthritis, cancer and CNS disease (Malemud, 2006). MPO can serve as a predictor for myocardial damage (Brennan et al., 2003) and uPA as a biomarker for pancreatic tumor invasion and metastasis (Yoo et al., 2013). Similarly, creatine chemical exchange saturation transfer (CrEST) MRI can map deficiencies in metabolites produced by necrotic myocardium and detect ischemic myocardium associated with early-stage heart disease (Haris et al., 2014). Non-invasive imaging of these enzymes could help to both predict disease onset and monitor its progression, but could also aid in the development of cell therapies with higher therapeutic benefit.

With recent advances in the field of molecular imaging, it is now possible to non-invasively monitor transgene expression in cells in vivo using genetically encoded reporters. The principle behind this is that a new, foreign gene encoding an MRI reporter protein is introduced into the cell’s genome (either stably or transiently), and is then transcribed and translated to produce a new protein, peptide or enzyme that can affect the MRI contrast in a manner that will enable its detection with MRI. Several landmark studies have utilized such genetically encoded proteins as MRI reporters, including the human transferrin receptor (Koretsky et al., 1996), β-gal (Louie et al., 2000), and proteins that are involved in iron metabolism (Alfke et al., 2003) and storage (Cohen et al., 2005; Cohen et al., 2007; Genove et al., 2005; Zurkiya et al., 2008) (Table 2). An overview of these gene reporter systems has been described in detail in several previous reviews and so is not discussed further here (Airan et al., 2013; Gilad et al., 2007b; Gilad et al., 2008b; Vandsburger et al., 2013).

We have previously demonstrated that the artificial CEST-based lysine-rich protein (LRP) reporter gene can be used to distinguish transplanted rat glioma cells that overexpress the reporter transgene from control cells in vivo (Gilad et al., 2007a). Since then, CEST MRI has been used to detect expression of the herpes simplex virus (HSV) type-1 thymidine kinase (tk) (Bar-Shir et al., 2013c), and to sense cellular signaling using a genetically encoded biosensor (Airan et al., 2012). The key advantage here is that the gene is replicated with each cell division, enabling longitudinal MRI tracking of the dynamics of biological processes. At this early stage of development, nearly all of these studies have been merely proof-of-principle and not yet applied in specific disease settings.

Although there is some overlap between reporter genes and sensors in their function to serve as a beacon, reporters are in general genetically encoded and synthesized by cells, whereas sensors are most often man-made and added to the cells. Importantly, sensors have the capability to sense the extracellular environment and are not limited to act as a beacon for events occurring within or on the surface of the cell. Examples of other recent developments in the field include approaches to express reporters in a specific tissue or cell type [as demonstrated by Bartelle and colleagues, who have taken a creative approach to expressing a multicomponent reporter specifically in endothelial cells (Fig. 5) (Bartelle et al., 2012)]; reporter genes that report on cellular metabolic changes and on the conversion of a pro-drug to an active drug [such as the carboxypeptidase G2 enzyme, a bacterial enzyme with putative therapeutic potential (Jamin et al., 2009)]; and biosensors that detect the activity of proteins, such as PKA (Airan et al., 2012).

The death of therapeutic cells from immunorejection and/or poor access to oxygen and nutrients during the post-transplantation period is an important factor that reduces the efficacy of stem cell therapies. However, the assessment of cell death after in vivo transplantation is challenging. One way to ‘sense’ cell death is to monitor changes in intra- and extracellular pH, a measure of acid-base balance that assesses the effective concentration of hydrogen ions (H⁺). In vivo, this acid-base balance is regulated through various pathways that enable the production and transportation of hydrogen ions so that intra- and extracellular pH are physiologically controlled (Gillies et al., 2004). Because of this control, abnormal pH values can indicate cell death or pathological changes in a tissue, e.g. extracellular pH (pHe) is reduced in tumors (Gillies et al., 1994).

Conventional methods that enable the measurement of intra- or extracellular pH in vitro include the use of pH electrodes (Roos and Boron, 1981) and the use of optical dyes (Li et al., 2006; Thomas et al., 1979). However, several MRI-based methods that are noninvasive, such as MRS, T1 contrast MRI and CEST contrast MRI, allow pH to be measured in live animals. Our group has recently developed a new technique using pH-sensitive CEST agents to monitor the viability of hepatocytes non-invasively in vivo using MRI (Chan et al., 2013). CEST contrast provides an alternative MRI method with which to measure pH (Aime et al., 2002; Liu et al., 2013; Sherry and Woods, 2008; Ward and Balaban, 2000) through changes in the water proton (MRI) signal. This technique offers several advantages; for example, CEST contrast is not constantly ‘turned on’ as T1 relaxation agents are, but is instead switched on through the addition of an appropriate saturation pulse(s) at the start of the imaging sequence, which allows for more-specific detection of contrast resulting from pH changes.

As mentioned above, one of the persisting limitations of stem cell transplantation is their poor survival immediately post- transplantation. To this end, tissue-mimicking scaffolds have been used to house and direct cells to grow, with the goal to improve stem cell survival. These scaffolds can be composed of several materials, including hyaluronic acid hydrogels (Liang et al., 2015; Liang et al., 2013b), collagen (Mertens et al., 2014), fibrin (Barsotti et al., 2011) and chitosan (Gomathysankar et al., 2014). The eventual degradation of biomaterial scaffolds is a key factor to successful cell therapy: would the scaffold persist, it might inhibit the integration of regenerated tissue with the host (Zhang et al., 2014). Recently, approaches have been developed to label scaffolds with SPIO and to monitor their site of implantation, tissue integration and biodegradation (Mertens et al., 2014; Mertens et al., 2015). For scaffolds containing gelatin, such as hyaluronic acid hydrogels, a label-free imaging approach has been developed based on its CEST-enhancing properties (Liang et al., 2015) (Fig. 6). It was shown that gelatin decomposition resulted in a reduced CEST contrast, which was validated using fluorescent labeling of the scaffold. Because no label is required, this approach could be of future clinical relevance.

Metal ions are essential for many biological processes, including for the maintenance of protein structure and for enzymatic activities, as well as in cell signaling and other pathways. The monitoring of metal ion levels is crucial for evaluating the appropriate function and regulation of cellular activity in disease. For example, global alterations in iron levels, including increases in the iron stored by macrophages and microglia, have been observed in the brains of individuals with MS (Stephenson et al., 2014). Similarly, in some neurodegenerative disorders, cellular Ca²⁺ regulation is often compromised, resulting in synaptic dysfunction, impaired plasticity and neuronal degeneration (Mattson, 2007). In ALS, a change in copper (Cu²⁺) homeostasis has been observed (Lovejoy and Guillemin, 2014), which has long been known to occur in Wilson’s disease (Bandmann et al., 2015). In such disorders characterized by irregular metal ion homeostasis, it could be important to evaluate the in vivo effect of cell therapy and cell engraftment on metal ion homeostasis as a read-out for therapeutic failure or success.

Although fluorescent-based dyes for imaging metal ions in living cells are commercially available and widely used, their clinical use is limited by the low tissue penetration of the photons they produce. They are, therefore, unsuitable for in vivo metal ion imaging in the deep tissues of an intact subject and remain clinically unviable. However, recent advances in the field of molecular MRI have resulted in new strategies for the detection of biologically relevant metal ions that are based on the design and synthesis of responsive contrast agents (Que and Chang, 2010). Efforts have been made to design MRI-responsive agents for the following biologically relevant metal ions: Ca²⁺ (Atanasijevic et al., 2006); Zn²⁺ (Esqueda et al., 2009); Mg²⁺ (Hifumi et al., 2007); Fe²⁺ (Livramento et al., 2005); Cu²⁺ (Que and Chang, 2006); Cu⁺ (Que et al., 2012); and K⁺ (Hifumi et al., 2007). Na⁺ can be observed directly using a coil tuned to the resonance frequency of the naturally occurring isotope ²³Na (Madelin and Regatte, 2013). MRI sensors for imaging metal ions have been designed to generate MRI contrast in T1 (Li et al., 2002), T2 (Atanasijevic et al., 2006) and CEST (Trokowski et al., 2005) contrast mechanisms; however, only a limited number of studies so far have produced robust in vivo data. As an example, it was demonstrated that, by administrating a synthetic MRI probe for sensing Zn²⁺, the functionality of insulin-producing β-cells in healthy mice could be determined in vivo (Lubag et al., 2011). Following intravenous glucose injection to stimulate insulin production, the paramagnetic Zn²⁺-responsive T1 agent was administered, resulting in an increase in T1 MRI contrast in healthy mouse pancreas. Because Zn²⁺ release is tightly correlated with insulin release from healthy pancreatic cells, the same experiment performed in diabetic mice (who do not produce sufficient insulin) did not produce any change in MRI contrast. This experiment demonstrates how MRI can reveal cell functionality in vivo by detecting changes in metal ion levels, a method that could, in the future, be used to detect changes in insulin release in islet cell transplantation for the treatment of diabetes.

It remains a challenge, however, to image metal ions that are present at low levels with high specificity. To overcome this, we have demonstrated a novel, metal-ion-imaging approach that combines ¹⁹F MRI with CEST MRI (Bar-Shir et al., 2013a; Bar-Shir et al., 2015) in an approach called (multi)-ion-CEST [(m)iCEST] (see Box 2). Taking advantage of the difference in the ion-specific ¹⁹F NMR chemical shift offset (Δω) values between the ion-bound and free ¹⁹F iCEST probe, we exploited the dynamic exchange between ion-bound and free iCEST probe to obtain MRI contrast. It was demonstrated that a prototype single ¹⁹F iCEST probe could separately visualize mixed Zn²⁺ and Fe²⁺ ions in a specific and simultaneous fashion. A key advantage of this technique is that it allows the detection of low (biologically relevant) concentrations of metal ions by simply reducing the concentration of the ¹⁹F probe (up to a detectable level) – a feature that is not available for ¹H MRI-based approaches. Although the iCEST technique has the potential to be used in vivo, further studies are required to prove the feasibility of this approach.

As pointed out, owing to their metal toxicity once de-chelated, it is highly unlikely that manganese, gadolinium and other lanthanides will have opportunities for clinical translation. However, these contrast agents will continue to aid in the assessment of cell transplantation strategies in experimental settings, which are pivotal as proof-of-principle studies. Tracking cells in patients has only been performed so far with SPIO and perfluorocarbons, and this is likely to remain for some time. Both agents have proven to be safe, with the iron in SPIO being metabolized and reused in the normal body iron pool, whereas fluorine is incompatible with any known biochemical reaction in humans, eventually leaving the body as an inert gas compound. Clinical reporter-gene-based cell imaging using the HSV-tk reporter gene has been implemented using PET and an ¹⁸F-labeled thymidine kinase substrate (Yaghoubi et al., 2009), raising the possibility that similar studies could be done with a CEST MRI substrate (Bar-Shir et al., 2013b; Bar-Shir et al., 2013c). However, there are different regulatory guidelines for using highly sensitive PET probes at a micro-dosing range, as compared to CEST substrates that need to be given at higher doses. Although the PET reporter gene was developed some 20 years ago (Tjuvajev et al., 1995), the CEST reporter gene/sensor field has just begun. CEST agents are based on natural products and do not contain metals; however, they could be involved in immunological or other biological reactions that are at present unknown. Future bio-clearance and dose-toxicity studies will be needed for the further consideration of CEST agents as a clinically viable approach, but so far there have been no indications that clinical translation is prohibited.

MRI-based cell imaging is a versatile technique because it can employ several different types of contrast. It has evolved, in a short period of time, from being a simple cell-tracking approach to an applied method for sensing the physiological state of cells. Except for SPIO- and ¹⁹F-labeled cell tracking performed clinically, the newer functional sensing and reporter gene approaches are still in their infancy, just entering from the test tube phase into preclinical testing. We anticipate that at least a few of these newer probes will be further developed and become commercially available, increasing their potential user group dramatically and allowing the field to grow. Then, a forward path towards clinical translation could become a bright new path ahead.