|Home | About | Journals | Submit | Contact Us | Français|
This article is a brief survey of preclinical in vivo cell tracking methods and applications using perfluorocarbon (PFC) probes and fluorine-19 (19F) MRI detection. Detection of the 19F signal offers high cell specificity and quantification abilities in spin-density weighted MR images. We discuss the compositions of matter, methods, and applications of PFC-based cell tracking using ex vivo and in situ PFC labeling in preclinical studies of inflammation and cellular therapeutics. We will also address potential applicability of 19F cell tracking to clinical trials.
This article provides a brief survey of emerging methods and applications of MRI cell tracking using perfluorocarbon (PFC) based cell labels and fluorine-19 (19F) detection. Cells are the fundamental building blocks of any organ system. For small animal studies, there are many options available for tracking cells in their native environment, especially using various fluorescent and bioluminescent probes and reporters. However, there remains a great unmet need for cell tracking technologies that have the potential for clinical translation. There are several non-invasive diagnostic imaging modalities that are routinely used in humans including various radioisotope methods, MRI, computed tomography, and ultrasound. Adopting existing diagnostic imaging modalities to visualize cells in the body is a complex problem, but there is a strong rationale for undertaking this technical challenge. Non-invasive imaging of the dynamic trafficking patterns of populations of immune cells can play an important role in elucidating the basic pathogenesis of major diseases such as cancer and autoimmune disorders. Other cell populations, such as tumor or stem cells, can be tracked using MRI to provide insight into metastatic processes, cell engraftment and differentiation, and tissue renewal. Moreover, cells are increasingly being used as therapeutic agents to treat genetic and neurological disorders, as well as chronic conditions such as autoimmunity and cancer. A common need for virtually all cell therapies at the development stage is a non-invasive way to detect and quantify the cell biodistribution following injection. Non-invasive imaging of cell trafficking is capable of providing critical feedback regarding modes of action of the cells, optimal routes of delivery and therapeutic doses for individuals. On the regulatory side, emerging new therapies, such as those using immunotherapeutic and stem cells, are slow to gain regulatory approvals partly because clinical researchers are challenged to verify where the cells go immediately after inoculation and where they migrate to days and weeks later. Cell tracking can potentially provide this information and may help in lowering regulatory approval barriers.
Intimately related to cell trafficking is inflammation and the inflammatory response. Biomedical research organizations commonly build research programs around major diseases where inflammation is a hallmark. Prevalent inflammatory diseases include, for example, arthritis, asthma, atherosclerosis, cancer, diabetes, chronic obstructive pulmonary disease (COPD), inflammatory bowel disease (IBD), infection, multiple sclerosis, and organ transplant rejection. The progression of these diseases can often be slow, and the effectiveness of treatment can be observed only after days, weeks or months. Thus, there is a strong unmet need for inflammation-specific diagnostics, as well as inflammation surrogate biomarkers that permit therapeutic developers to determine efficacy quickly, quantitatively, and in a longitudinal fashion. A related need entails pharmacological safety profiling to detect ‘off target’ inflammatory side effects in pre/clinical drug trials. A non-invasive, image-based biomarker could potentially fill these unmet needs. Vital imaging can accelerate the ‘go/no go’ decision making process at the preclinical and clinical trial stages, and can facilitate smaller, less costly trials by enrolling fewer patients. Imaging can potentially yield quantitative data about inflammation severity and time course in the anatomical context. The highest value imaging biomarker would have broad utility for multiple diseases and be applicable from mouse-to-man, thereby minimizing validation studies.
In an effort to devise next generation cell tracking probes, there has been significant recent interest in the use of 19F-based cell labels employing PFC molecules. Detection of 19F offers the advantage that there is no background signal from the host’s tissues and only labeled cells are detected. Moreover, quantification of the PFC probe is directly related to the signal intensity in 19F spin-density weighted MRI or magnetic resonance spectroscopy (MRS), thereby yielding a marker of the apparent number of cells in regions of interest, or alternatively, inflammation severity. The detection of PFC labeled cells using 19F MRI/MRS is different from prior art using metal-ion-based contrast agents. In the case of the latter, one detects the presence of the paramagnetic contrast agent indirectly via its effect on T1, T2, and/or T2* of surrounding protons in mobile water. The PFC acts like a ‘tracer’ agent rather than a contrast agent since 19F magnetic resonance directly detects the 19F spins associated with the labeled cells.
Broadly, there are two approaches for labeling cells for tracking using PFC reagents. In the first approach, isolated cells of interest are labeled ex vivo (i.e., in culture) using a PFC formulated as an emulsion. Following transfer to the subject, cells are tracked using 19F MRI/MRS (Fig. 1). The fluorine contained in the labeled cells yields cell-specific images, with no background, and these can be used to quantify apparent cell numbers at sites of accumulation. This approach, called ‘in vivo cytometry’ (1) is capable of tracking a wide variety of cell types in a cell-specific manner. Alternatively, the cell biodistribution can be accurately measured in panels of intact tissue specimens using conventional high-resolution 19F NMR spectroscopy measurements (1,2).
In the second approach, PFC emulsion is directly injected intravenously (i.v.). Following injection, the emulsion droplets are taken up by cells of the reticuloendothelial system (RES), including monocytes and macrophages. These ‘in situ’ labeled leukocytes participate in inflammatory events resulting in 19F accumulation at inflammatory loci. The 19F MRI signal can be quantified by integrating the signal in regions of interest, and this integral is linearly proportional to the macrophage burden (3,4).
In this article we provide a brief survey of the compositions, methods, and applications of PFC-based cell tracking using in vivo cytometry and in situ cell labeling. We describe a sampling of applications in preclinical studies. We will also address potential applicability of 19F cell tracking to clinical trials.
PFCs have been studied for more than 40 years and are among the most biologically inert organic molecules ever produced (5). There are no known enzymes that metabolize fluorocarbons in vivo (6), and they do not degrade at typical lysosomal pH values (7,8). These molecules contain C-F covalent bonds that are among the strongest known, a consequence of fluorine’s high electronegativity. PFCs are generally highly hydrophobic and are often highly lipophobic. Consequently, PFCs have a tendency to segregate when placed in an aqueous environment and are immiscible in tissues; these property are partially responsible for PFCs being biological inert and nontoxic in vivo even at very high doses. Clearance of PFC agents from the body generally occurs via RES uptake and lung exhalation (9).
Most PFCs have the ability to readily dissolve oxygen, carbon dioxide and nitrogen. Early biological applications of PFCs exploited the high oxygen solubility (10). PFCs have been studied for decades as artificial oxygen transport vehicles and blood substitutes (11) for human use. Consequently, the safety profile of PFCs in man is well characterized.
The notion of using PFCs as a 19F tracer agent has been around since the beginning of MRI/MRS (12–14). Early PFC imaging applications included their use as angiographic agents (12). There is strong rationale for 19F MRI in certain applications because of fluorine’s very low background biological abundance. Additionally, the 19F isotope has 100% natural abundance and a spin-½ nucleus. The gyromagnetic ratio of 19F differs from 1H by only approximately 6% and has a relative sensitivity of 0.83. The sensitivity of PFCs as MRI reagents is highly dependent on its chemical structure. Optimal molecules should have a simple 19F NMR spectrum, ideally having a single, narrow resonance to maximize sensitivity and minimize chemical shift image artifacts. Additionally, a short 19F spin-lattice relaxation time (T1) and a long spin-spin relaxation time (T2) is desirable to promote rapid data acquisition times using conventional fast-imaging pulse sequences.
A review of suitable PFC molecules for 19F MRI is described elsewhere (15,16). PFC molecular structures generally fall into several classes, including, branched, linear, cyclic and linear polyether. Perfluorooctyl bromide (PFOB), a linear PFC, was used early on as a 19F MRI agent (17). PFOB is hydrophobic, but displays finite lipophilicity due to covalently-bound bromine, which enhances clearance rates from the body. The PFOB 19F NMR spectrum has eight peaks, one for each CFn moiety. To minimize chemical shift artifacts when using PFOB, MRI pulse sequences often incorporate pre-saturation RF pulses on undesired resonance peaks prior to readout. PFOB is of most PFCs used in biomedicine in that it displays multiple 19F peaks, which significantly compromises sensitivity, as generally only one single peak can be used for imaging, and the other peaks must be suppressed. Interestingly, Giraudeau et al. (18) has described the use of frequency-selective refocusing pulses in multiple spin-echo sequence to suppress J-coupling between the CF3 and CF2 groups and associated signal loss, yielding a 4-to-6-fold increase in sensitivity over traditional gradient-echo (GRE) and chemical shift imaging (CSI) acquisitions. Significant improvement in MRI sensitivity can be achieved using the macrocyclic PFCs such as the perfluoro-15-crown-5 ether (PCE) molecule with 20 chemically equivalent fluorine atoms and a single NMR resonance. PFCs commonly used for MRI, such as PFOB and PCE, have intrinsically long T1 relaxation times (e.g., >1 s) and this can limit the MRI data acquisition rate and ultimate sensitivity. More recently, linear perfluoropolyether (PFPE) polymers have been employed as a cell tracking reagent (1,19–21). The linear PFPE is sensitive due to its simple 19F NMR spectrum (1,19,22) with a single major resonance, >40 chemically equivalent fluorine, and a small T1/T2 ratio (16). Moreover, linear PFPE has endgroups that are amenable to chemical modification to provide additional functionality, for example, to enable covalent conjugation to bright fluorescent dye molecules for dual-modality detection (19).
PFCs are water insoluble and immiscible in cell membranes. Consequently, cell tracking applications require formulation of PFC into a biocompatible emulsion. A significant body of art on formulating stable PFC emulsions has been reported in the context of developing artificial blood substitutes (5,6,23). In these applications, emulsions must be stable in vascular circulation for extending periods of time, a property that is also desirable for in situ labeling of inflammatory macrophages for imaging. In contrast, ex vivo cell labeling demands alternative formulations that enable efficient and rapid intracellular uptake of PFC emulsion in culture media, including cells with an intrinsically-low phagocytic phenotype, such as stem cells and T cells.
Ideally, an emulsion for MRI cell tracking should have a small, uniform droplet size, ideally <200 nm. Large emulsion droplets can affect cell activation phenotype after labeling, for example in sensitive dendritic cells (DCs) (24). For ex vivo cell labeling, wash steps are employed to remove excess emulsion that is not taken up by cells after the labeling period; in the case of large emulsion droplets, the wash step is ineffective because centrifugation spins the emulsion down with the cells and excess emulsion cannot be discarded with the supernatant. Moreover, a tightly distributed droplet size of low dispersity helps ensure uniform labeling within a cell population.
A detailed description of the materials and methods associated PFC emulsion formulation for imaging applications is described elsewhere (16,25). Formulation generally employs surfactant(s) to stabilize the colloidal suspension and provide a practical shelf-life. Importantly, the choice of excipient may also alter biological characteristics of the emulsion, for example, to promote rapid uptake of PFC into cells in culture, or to enable a long circulation time in the blood stream.
Excipients may also be used to modify the 19F NMR characteristics of the PFC molecules. For example, gadolinium chelates have been attached to the emulsion droplet surface to reduce the 19F T1 relaxation time, thereby imparting sensitivity enhancement by shortening data acquisition times (26). However, there is empirical evidence that the gadolinium chelate does not stay associated with the PFC emulsion droplet once inside the intracellular milieu (27).
Minimal excipient mass is desirable so that low amounts of MRI-inactive material is delivered inside the cell. Also, excipients should be non-toxic to cells and not modify phenotype, morphology or biological function of the labeled cells. Common emulsifiers include phospholipids (e.g., egg yolk phospholipids) and pluronics (e.g., poloxamer 188 or F68) (5). In one example, a ‘self-delivering’ and stable emulsion for ex vivo cell labeling was formulated using linear PFPE, pluronic F68 and linear short chain polyethylenimine (19). The formation of stable, small particle, colloidal suspensions of PFC and excipients requires high energy processing methods. Methods include, for example, high pressure, high shear homogenization (e.g., microfluidization) and sonication. The most common degradation mechanism for PFC emulsions is Ostwald ripening, a molecular diffusion phenomena that results in a gradual growth of the larger particles at the expense of smaller ones (28,29) and represents a major challenge in formulating stable PFC emulsions (30–32). Additional details describing emulsion formulations are described elsewhere (16).
There is increasing emphasis on non-invasive multi-modal imaging. Dual-mode PFC emulsions that can be detected by both 19F and fluorescence have been described in several studies (19,33–35) In preferred formulations, the PFC molecule is directly conjugated to a fluorescent dye prior to emulsification to ensure coincident MRI and fluorescence signals (Fig. 2) (19). Alternatively, lipophilic fluorescent dyes can be added to the surfactant when phospholipid-based excipients are employed (36). In preclinical studies, these dual-mode reagents enable one to positively identify the fate and phenotype of labeled cells following 19F MRI, days and weeks after cell transfer using flow cytometry, fluorescence microscopy, and/or in vivo optical imaging (21). In one example, Janjic et al. (19) developed dual-mode PFC emulsions by direct conjugation of dye molecules in the visible spectral range to linear PFC and formulated a self-delivering emulsion for ex vivo cell labeling. A linear correlation was observed between the 19F NMR signal and the fluorescence signal in labeled cells. Mouse primary T cells were labeled by the dual-mode PFC emulsion (Fig. 2a). In vivo 19F imaging showed hot-spots in multiple lymph nodes, which was validated by fluorescence detection via optical microscopy of CD4+ cells in the isolated lymph nodes and by flow cytometry (Fig. 2b). Near-infrared dyes have also been incorporated into PFC emulsions enabling in vivo imaging at moderate tissue depths (37). Moreover, fluorescent PFC emulsions have stand-alone utility as an optical-only cell labeling probe due to their brightness, low toxicity, long retention time in cells, and the existence of self-delivering formulations.
One of the longstanding goals in the field of cellular and molecular MRI is targeted imaging of cells of a particular phenotype in vivo. This goal is often stymied by the fact that MRI is intrinsically insensitive compared to other techniques (e.g., positron emission tomography, PET). Targeting cellular epitopes in situ with an MRI probe requires overcoming the quadruple requirements of efficient tissue access of the imaging probe following intravenous injection, high binding specificity, high cellular epitope density, and high payload of the MRI-detectable moiety (38,39). However, in many instances, one can sidestep these limitations using ex vivo cell labeling, where one optimally labels cell populations of interest that have been pre-sorted for the desired phenotype in culture, followed by (re)introduction of the labeled cells into the subject. This approach is particularly feasible for cellular therapies that are already routinely cultured outside the body prior to infusion into the patient; a cell labeling reagent can be just another ‘factor’ that is added to cells prior to injection. Other cell types, such as autologous leukocytes obtained from peripheral blood leukapheresis procedures, bone marrow harvests, cells recovered from tissue biopsies, or cell lines, are also amendable to ex vivo cell labeling for MRI. Although 19F MRI has been around for decades (15), the notion of using PFCs for cell tracking and quantification in a cell-specific manner, i.e., in vivo cytometry, is a recent development. Ex vivo 19F cell labeling, was first proposed in 2005 (36). The fluorine contained in the labeled cells yields cell-specific images, with no background, that can be used to quantify apparent cell numbers at sites of accumulation.
In vivo cytometry was first used to visualize dendritic cells (DCs) in mouse (36) (Fig. 1a). Several T cell studies have used 19F cell tracking to examine early inflammatory events in a rodent model of type-1 diabetes (1) (Fig. 1b), an acute inflammation model (21), and the biodistribution of Muc1-specific lymphocytes in inflammatory bowel disease (IBD) (2). Other studies have employed 19F MRI cell tracking to visualize stem cells in vivo in rodent models (Fig. 1c) (33,40–42). The use of PFC labeling for cancer cells has also been reported (43). Additionally, primary human DCs relevant to immunotherapeutic clinical trials have been labeled with PFC, biologically characterized, and imaged in an immune compromised mouse (44).
While certain cells having a phagocytic phenotype, such as immature DCs, readily take up PFC emulsion droplets in culture, other non-phagocytic cell types such as stem cells and T cells, do not. For the latter cell types, transfection methods are needed, for example, by employing a premix step of the emulsion with a cationic transfection agent prior to addition to the cell culture (36). However, transfection agents are sometimes toxic to sensitive cell types such primary lymphocytes, can alter cellular phenotype, and often not suitable for use in human health. An important PFC reagent innovation has been the development of self-delivering emulsion formulations that do not require the use of a transfection agent and a premix step (19) and results in overall improvements in cell viability and ease of use. Current formulations of PFC reagents can efficiently label cells by simple co-incubation over a period of several hours. The degree of labeling can readily be assayed by acquiring a 19F NMR spectrum of a sample consisting of a cell pellet of known cell number and a fluorine reference compound with a disparate chemical shift, where the mean 19F/cell is calculated from the integrated areas of the PFC and reference peaks (1). This simple assay method commonly yields cell loading values ranging from 1011 to 1013 19F/cell (19).
The PFC emulsions used for ex vivo cell labeling have been rigorously tested and appear to be biologically safe, presenting no observed adverse effects to viability or function in cells. Numerous studies have investigated the impact of PFC cell labeling on cellular phenotype and function in primary immune cells using a variety of sensitive in vitro assays, for example in the context of murine DCs (36), T cells (21), and stem cells (33,40–42). The most detailed in vitro study to date involved PFC labeled primary human DCs (44); cells were assayed for viability, maturation phenotype, cytokine production, T cell stimulatory capacity, and chemotaxis (44), and no difference in these parameters was observed between labeled and unlabeled cells. More recently, primary human CD34+ hematopoietic stem cells were labeled with PFC and shown to retain their pluripotency and ability to reconstitute the full leukocyte repertoire in vivo in mice that underwent immunoablation (45).
In PFC labeled cells having a mitotic phenotype, cell division and subsequent dilution of the intracellular label, can potentially limit long-term cell tracking studies of itinerant cells and/or decrease the accuracy of cell quantification. Death of labeled cells can lead to dispersion of the reagent and loss of 19F signal. Potentially, the PFC droplets can also be transferred to resident phagocytes (e.g., macrophages). If a large number of these labeled phagocytes remain in a region of interest, false positive signals could result. These caveats are the same for many commonly used imaging modalities where a tracer material is utilized, for example with SPECT using Indium-111 probes and with various nanoparticle probes (SPIO, NIR, Q-Dots, etc.). To rigorously overcome detection of false positive results from live versus dead cells, a genetically-encoded reporter gene may be employed, for example for MRI, using the ferritin gene (46–48) or a CEST-type agent (49) which are degraded by proteolytic enzymes upon cell death.
In many instances, a key question that arises in the early development of cell therapies is the cell biodistribution post-transfer. NMR of fixed tissue samples is an alternative method to detect and quantitate ex vivo labeled cells (1,2), and has the advantage of high sensitivity. A quantitative cell biodistribution assay can be used to characterize delivery efficacy, cell engraftment, and possible off-target cellular niches. Conventional methods to assay cell biodistribution in excised tissue panels, such as histology and flow cytometry of single cell suspensions, represent a laborious obstacle toward gaining this information that lacks quantitative rigor. Alternative methods may rely on the use of radioisotopes to label cells prior to infusion. However, radioisotope labeling may cause changes to cell function, especially to those cells that are exquisitely sensitive to irradiation such as stem cells (50). Moreover, the radioisotope lifetime may limit long-term cell detection and quantification (51,52).
Excised tissue samples containing PFC labeled cells can readily be assayed using NMR. Tissue panel preparation involves fixation, weighing, and the acquisition of 1-dimensional 19F spectra. An aliquot of a 19F reference solution (e.g., trifluoroacetic acid) is generally added to the tissue specimen, preferably placed in a small sealed capillary alongside the tissue mass. The tissue specimen plus reference solution should fit entirely within the receptive field of the NMR coil (~1–2 cm along the z-axis) in order to detect every 19F spin for accurate quantification. This sample preparation may degrade the spectral linewidth compared to conventional high-resolution NMR practices, but as only the integrated peak area is used and the spectrum is simple, linewidth is of limited concern. The total number of cells in the tissue sample is calculated from knowledge of the 19F/cell following labeling (described above), the integral of the 19F spectrum of the tissue sample, and the integral of a fluorine reference compound added to the NMR tube (3). Normalization of the measured cells/sample by the specimen weight yields the average tissue cell density. In preferred implementations of these methods, a 10 mm broadband probe is used. For example, in murine studies a 10 mm NMR tube can accommodate any intact organ (e.g., liver and brain) without the need for manual segmentation. If tissue panels involved large numbers of samples, commercially-available robotic or automated NMR sample changers can be employed to accelerate measurement throughput. The PFC compounds are stable over time, thus the samples can be safely stored until the time of measurement or shipped to appropriate instrumentation without loss of measurement accuracy. As the NMR tissue preparation and measurement is non-destructive, the same tissues can be processed for histology to refine the tissue analysis as required.
Inflammation is a protective mechanism initiated by the immune system to remove foreign pathogens or injurious stimuli and start the healing process. Cascades of biological events are involved in an inflammatory response, which includes but not limited to recognition of pathogen, activation of immune cells, and secretion of inflammatory mediators (53). As a consequence of these events, there is often recruitment and translocation of leukocytes from circulation to sites of inflammatory loci. In vivo visualization of inflammatory hot-spots advances our understanding of the inflammatory response, provides significant diagnostic value for a wide range of diseases, and can be used as a surrogate marker to monitor therapeutic interventions.
Following systemic PFC administration to a subject, emulsion droplets are initially taken up in situ by circulating phagocytes in the blood stream. These cells are a subgroup of leukocytes that include primarily monocytes, macrophages, neutrophils, and DCs; these cell types are tasked with engulfing and removing cellular debris, foreign substances, and pathogens. Macrophages are monocytes in their matured form, and these cells preferentially engulf PFC emulsion in situ (54). To a much smaller degree, other cells types, including neutrophils (55,56), DCs, B lymphocytes (57), microglia (58) can also uptake PFC in situ.
The fluorine-tagged leukocytes participate in inflammatory events in vivo. When labeled cells accumulate in sufficient amounts at sites of inflammation, they become detectable by 19F MRI. In situ labeling has been widely used for 19F MRI of phagocytes in a multitude of preclinical models of human disease (3,4,13,34,55–57,59–64). Unlike in vivo cytometry, in situ labeling of leukocytes with PFC emulsion has the advantage that it does not require procedures to collect immune cells, handle them ex vivo and reintroduce them back into the subject. In situ labeling has the disadvantage that the precise phenotype of labeled cells is unknown and may involve multiple cell types. Also, the amount of 19F/cell taken up by cells in situ is not known, thereby hindering cell quantification in vivo. However, inflammation quantification in spin-density weighted images can be performed and involves counting the number of apparent fluorine atoms in regions of interest, with aid of a calibrated 19F reference (e.g., capillary filled with PFC) placed alongside the subject in the field of view. The number of fluorine atoms present in regions of interest is approximately linearly proportional to the macrophage burden (3,4).
The notion of using in situ PFC cell labeling and 19F MRI to detect inflammation has been around since the first demonstration of 19F MRI. Early 19F MRI studies visualized 19F hot-spots in rat liver, tumor, and abscess following PFC infusion (13). In other studies (64), PFOB was administered intravenously to a mouse that received abdominal radiation, and a decreased 19F uptake was detected in the spleen indicating impaired phagocytic function of resident macrophages as a consequence of acute radiation damage (64). Ratner et al. (63) detected 19F signal accumulation in a mouse tumor after intravenous delivery of PFOB, presumable due to tumor-associated macrophages. In the late 1990’s, 19F MRI with in situ cell labeling was used to monitor inflammation associated with autoimmune disease in the context of experimental allergic encephalomyelitis (EAE), an animal model of multiple sclerosis (62). In these experiments, perfluoro-15-crown-5 ether (PCE) emulsion was infused into EAE rats, and macrophage recruitment was detected in the tissue in proximity to the midbrain, medulla, and the cervical spinal cord.
More recently, in situ PFC labeling has been applied to a broad spectrum of inflammatory disorders in animal models [for recent reviews see (65,66)]. Flögel and colleagues (57) reported macrophage visualization in ischemia models of the myocardium and cerebrum following intravenous PCE emulsion infusion. Ebner et al. (56) reported monocyte/macrophage trafficking into sites of lipopolysaccharide-induced pulmonary inflammation and described a strong correlation between the intensity of 19F signal and the severity of lung inflammation (56). Hertlein et al. (55) reported the trafficking of phagocytic cells to the murine thigh as part of a bacteria-induced host immune response (Fig. 3). Longitudinal image comparisons were made using both PFC emulsion and iron-oxide nanoparticles in the same model. It was shown that PFC could delineate abscess boundaries in both the acute and chronic phases, but iron-oxide was not able to clearly define the lesion in the acute phase. In another study (60), a neuroinflammatory response was visualized in the peripheral nervous system using a lysolecithin injection into the rat sciatic nerve. Following PFC injection, macrophage recruitment was observed focally at sciatic nerve lesions. In other studies (59), a rodent collagen-induced rheumatoid arthritis model was used to show a strong correlation between disease severity, as observed by quantification of 19F uptake in affected limbs, and conventional joint caliper measurements. To modulate disease severity, glucocorticoid therapy was administered to the rodents.
Non-invasive visualization of the inflammatory response associated with organ transplant rejection is potentially a key use of PFC and MRI, as there is a strong motivation to supplant invasive ‘gold-standard’ biopsy procedures. Potentially, the magnitude of 19F signal in the rejecting organ scales with organ rejection stage. Several studies (34,61) applied in situ PFC labeling to image organ rejection. Hitchens et al. (34) quantified the 19F signal in rat heart and kidney allograft rejection models and showed that the patterns of signal corresponded to rejection histopathology (Fig. 4). Comparisons with the use of iron-oxide nanoparticles were also performed in the same models. Hitchens reported that both labeling methods showed strong correlation between MRI signal and the graft rejection status. However, the complex hypointensity patterns in T2*-weighted images hinders unambiguous detection of iron-labeled cells and complicates the process of rejection quantification and scoring. In another study, Flögel and colleagues showed quantitative analysis of 19F signals is more sensitive to rejection in the early stage than conventional makers such as 1H anatomical images, functional MRI scans, or palpation scores (61).
Inflammatory bowel disease (IBD) has also been the subject of in situ PFC labeling studies. Kadayakkara et al. (4) showed that macrophage activity associated with IBD in a IL-10−/− mouse model can be non-invasively detected and scored using 19F MRI. The bowel is notoriously difficult to image using conventional 1H due to the complexity of the background contrast and 3D anatomy. In this study, the 3D patchy inflammation was imaged in an 8 minute acquisition time (Fig. 5). Validation studies used immunofluorescence, qRT-PCR and in situ macrophage ablation, to demonstrate that the PFC was localized within macrophages and that the cell quantity was reflected in the magnitude of the colon 19F signal. Additionally, the impact of putative therapeutics used to treat colitis was reflected in the magnitude of the 19F signal measured in vivo.
As described above (see ‘NMR cytometry’), conventional liquid-state NMR instrumentation can also be used to rapidly and quantitatively assay inflammation in intact, excised tissue samples. Discovery and preclinical studies often rely on histological analysis of a panel of tissues to examine the extent of inflammation, and these procedures are often viewed as time consuming, expensive, and a bottleneck in research and development. As an example of the applicability of NMR methods, Ahrens et al. (3) used a rat model of EAE displaying pronounced CNS inflammation. This study manually segmented the fixed rat brain and spinal cord into 15 segments and subjected the intact tissue samples to 19F NMR analysis. The results displayed the inflammation profile along the entire length of the CNS. The results for each sample were expressed as the ‘inflammatory index’ with units of fluorine atoms per tissue weight, which was shown to be proportional to the macrophage burden via PCR analysis. Overall, 19F NMR can be used to dramatically reduce the time to evaluate macrophage involvement in a wide variety of acute and chronic inflammatory models. Importantly, liquid-state NMR spectrometers that are 19F-capapble are ubiquitous in laboratory research centers, thus these methods can be widely utilized for inflammation research.
Fluorine-based cell tracking directly detects the density of 19F spins associated with labeled cells, which is generally dilute compared to 1H in biological tissue. Importantly, 19F cell tracking does not demand a high 19F signal-to-noise ratio (SNR). Because there is negligible 19F background, any 19F signal detected is from labeled cells. Unlike 1H anatomical imaging, where one relies on its high SNR to resolve anatomical detail and organ definition, the 19F image only needs to display localized ‘pools’ of cells at low SNR (e.g., <5) and low resolution; the high-resolution 1H underlay provides the detailed anatomical context.
From ex vivo labeling studies, the amount of 19F loaded in cells typical ranges from approximately 1011 to 1013 19F/cell (19); we speculate that the degree of labeling reported for a given cell type correlates to the overall cell (or cytoplasmic) volume. Given this cell labeling level, it has been shown empirically that the minimum cell detection sensitivity for 19F cell tracking is on the order of 104 to 105 cells per voxel for clinical MRI systems and 103–104 cells per voxel for high-field animal scanners (1,36,41,42,67). Single-voxel in vivo magnetic resonance spectroscopy (MRS) detection is expected to yield even higher sensitivity to sparse cell numbers. For cell detection in fixed samples using liquid-state NMR spectrometers, cell detection sensitivity is even greater, approaching <103 cells per sample (unpublished). Experimental details, such as the 19F T1/T2 ratio of the PFC used, the number of NMR peaks of the PFC molecule, the cell type (i.e., cell size) labeled, image acquisition methods, magnetic field strength, and RF coil configuration determine the actual sensitivity possible for a particular study.
Most small animal MRI scanners can readily be adapted to image 19F with the addition of a suitable 19F/1H coil, and numerous vendors can supply such a coil. The preferred RF coil configurations for multi-nuclear MRI are discussed elsewhere (68–71). Clinical scanners are generally designed to be more specialized for 1H-only applications, but most can be adapted to scan 19F with the addition of a 19F/1H coil. However, some scanners will also require an aftermarket preamplifier, T/R switch and RF power amplifier suitable for 19F. Additionally, clinical 19F MRI practitioners may require a vendor-supplied research license to use/install multi-nuclear imaging capabilities.
Pulse sequences with an intrinsically-high SNR per image acquisition time are generally preferred, such as RARE (56,57), SSFP (14,55), GRE (72,73), and EPI (74,75). These sequences may be combined with parallel imaging techniques (76,77) to further reduce the total imaging acquisition time.
In the future, there is likely to be significant opportunities to improve the data acquisition schemes used to acquire 19F data sets that will greatly accelerate the acquisition speed and cell detectability. To date, 19F cell tracking studies have mostly used conventional pulse sequences (spin-echo, RARE, GRE, etc.) that do not exploit the unique characteristics of the 19F data. The 19F signal appears as isolated pools of signal from cell deposits against a background of pure noise. Generally, the image field of view contains only ‘sparse’ data, where often only a few percent of voxels of the total contain signal. Additionally, the 19F spin-density weighted images are generally devoid of structural information and have minimal high-frequency spatial components; the 1H image underlay provides the anatomical context. Due to the low SNR, long MRI scan times may be required using conventional pulse sequences that may hinder the use of 19F detection in regions of sparse cell numbers or to detect low-grade inflammation.
Recently, compressed sensing (CS) methods (78–83) have been used to reduce MRI scan times in situations where images are expected to contain ‘information-sparse’ features. In CS, k-space is undersampled below the Nyquist criteria and image reconstruction constraints are customized for each application. CS is a 3D imaging approach that can improve the image SNR per unit acquisition time and therefore enhance the ability of 19F cell tracking to detect sparse cell numbers by enabling increased signal averaging within the temporal confines of an imaging session. A recent CS MRI study (84) in the context of 19F cell tracking showed that the method is effective in significantly reducing the image acquisition time by at least 8-fold without seriously effecting image features and 19F spin quantification accuracy (Fig. 6). Our view is that data acquisitions schemes that have reduced k-space sampling (e.g., CS) will be important for future clinical applications of 19F based cell tracking as a means to minimize scan time.
One of the characteristics of 19F cell tracking is that the resulting images can be quantified by integrating the signal hot-spots in spin-density weighted images. Often an external 19F reference standard is used with the scans in the image field of view to provide absolute quantification of the total number of fluorine atoms present in regions of interest. Several factors can affect the accuracy of these integrated signal values. Noise in complex images is normally distributed about zero independently for both the real and imaginary components in each voxel. Because the 19F images are often in the low SNR regime, converting the complex-valued images to magnitude images, which is commonplace in MRI, creates non-normally distributed noise having a Rician distribution (85). Consequently, magnitude images have a non-zero mean pixel value in regions devoid of signal, which introduces a noise-dependent bias to the data when the SNR is low. Thus, ideally one should compensate for the Rician bias by rescaling magnitude intensity values that are close to zero (21,85).
Motion of the subject during the 19F scan is another potential source of 19F quantification inaccuracy. As with 1H scans, the sensitivity of 19F images to motion is dependent on the particular imaging method used. Sequences such as UTE (86), SWIFT (87) or other acquisition scheme (88) are preferred for minimizing motion-related quantification inaccuracies. Several studies have addressed motion correction for 19F imaging based on simultaneous proton scans (73) and/or proton self-navigation techniques (89).
Another practical concern relates to the use of gaseous anesthesia for animal MRI experiments. Gaseous anesthesia is generally fluorinated, and potentially false positive 19F signals may appear over time if the anesthesia accumulates in tissue. Our experience is that a faint and diffuse 19F MRI signal may appear predominantly in subcutaneous fat regions after extended periods (>2 hour) under gaseous anesthesia. However, in practice, the use of fluorinated anesthesia gases does not cause a serious roadblock for 19F cell tracking. The 19F images should be acquired near the beginning of the imaging session, and if significant signal averaging is employed, one can acquire the center portion of k-space first to minimize any potential 19F signal contamination from the anesthetic. However, to rigorously avoid potential false positive 19F signals from gaseous anesthesia, an option is to use injectable liquid anesthesia, for example, delivered via a mechanical pump and an intraperitoneal catheter (1,36) for the duration of the scanning session.
In addition to cell tracking capabilities, PFC cell labels can also be used as intracellular oximetry probes. In vivo measurement of intracellular oxygen level facilitates our understanding of cell physiology and its response to therapeutic procedures. PFC is known to dissolve paramagnetic oxygen, thereby decreasing the 19F T1 and T2 (75,90,91). Normally T1 is used to measure the partial pressure of oxygen (pO2), as it exhibits a larger response to O2 concentration compared to T2. The 19F relaxation rates of PFC exhibit a highly linear dependence with local oxygen levels in proximity to emulsion droplets. Oxygenation in vivo can be calculated using an in vitro calibration curve correlating NMR relaxation rates with absolute pO2 (92). Compared to traditional oxygen sensing approaches using probe electrodes or ESR (93), 19F MRI oximetry has minimal invasiveness, can be applied longitudinally, and is suitable for deep-seated cell populations.
Much of the work using PFC oximetry focuses on cancerous tissue (94). For example, in prior tumor studies, the PCE emulsion was introduced i.v., or injected directly intratumorally (75,94–99). For i.v. delivery, generally a large PFC dose is injected systemically, and a small fraction of the PFC accumulates at the tumor site, which may be visible several days post-administration. The PFC emulsion droplets are sequestered predominantly in the periphery of the tumors because of leakage in tumor vasculature (98,99), or phagocytosed by tumor-associated macrophages (75,97). Generally, in the above routes of administration PCE emulsion deposits are distributed very non-uniformly in the tumors.
More recent studies (43) have shown that it is feasible to measure intracellular pO2 in a cell-specific manner using in vivo cytometry methods. Kadayakkara et al. (43) investigated PFC labeled 9L glioma cells implanted into the rat striatum. Following inoculation of labeled 9L cells into the striatum, tumor cells were imaged using 19F MRI (Fig. 9a). Single voxel 19F MRS was used to longitudinally measured the mean intracellular pO2 in the tumor implant (43). The same study then monitors changes in pO2 before and after treatment with the chemotherapeutic agent bis-chloroethylnitrosourea (BCNU) (Fig. 9b). A single BCNU dose was sufficient to produce a dramatic and sustained increase in tumor pO2 (Fig. 9b). At the experimental endpoint, histology was used to confirm that the PCE was retained intracellularly in 9L cells via a dual-mode version of the labeling emulsion and fluorescence microscopy (see Ref. (43)); no obvious extracellular PCE could be observed. These data show that intracellular oxygen sensing is feasible in a cell specific manner, with high sensitivity.
The use of 19F cell tracking in clinical trials is still in its infancy. Our view is that in the future in vivo cell tracking will routinely be used to provide a surrogate biomarker for cellular therapeutic clinical trials and for inflammation detection and monitoring, in combination with drug trials. Various types of perfluorocarbons have long been contemplated as oxygen carriers in traumatic surgeries and as artificial blood substitutes (100), and thus a large volume of data exists on the toxicity profile and clearance pathways of this class of molecules. Importantly, in using these products large doses of the PFC are delivered systemically in an effort to maintain satisfactory tissue oxygen levels. In the case of in vivo cytometry, a relatively miniscule quantity of PFC, contained within the transferred cells, is delivered to the subject suggesting that PFC cell tracking will be highly safe for human use. Thus, in these applications the imaging agent may be considered as an excipient to the cell therapeutic, and presumably lower regulatory hurdles are present. In the case of in situ labeling, a much larger PFC dose must be delivered to the subject, thus there are elevated safety concerns. The technical barriers associated with implementation of 19F MRI/MRS on a clinical scanner are surmountable. Prior studies using 19F MRI on a clinical scanner have demonstrated the feasibility of noninvasive characterization of PFC signal intensity at clinical field strength. (69,73,101)
Overall, 19F MRI cell tracking using PFC tracer agents is a rapidly emerging alternative to 1H-based approaches using metal-ion-based contrast agents. Future improvements in sensitivity and functionality of 19F-based cell tracking will require an integrated and interdisciplinary approach that rigorously applies principles of chemistry, biology, spin physics, and image processing. Our view is that these technical undertakings are worthwhile, as work in this field to date has shown that this sub-field of MRI cell tracking can produce unique information about cellular processes in vivo in a diverse range of applications.
We thank Dr. Kevin Hitchens for review of the manuscript. We acknowledge support from the National Institutes of Health grant R01-CA134633, the Pittsburgh NMR Center for Biomedical Research, supported by P41-EB001977, and the Dana Foundation.