Clinical imaging in regenerative medicine

Abstract

In regenerative medicine, clinical imaging is indispensable for characterizing damaged tissue and for measuring the safety and efficacy of therapy. However, the ability to track the fate and function of transplanted cells with current technologies is limited. Exogenous contrast labels such as nanoparticles give a strong signal in the short term but are unreliable long term. Genetically encoded labels are good both short- and long-term in animals, but in the human setting they raise regulatory issues related to the safety of genomic integration and potential immunogenicity of reporter proteins. Imaging studies in brain, heart and islets share a common set of challenges, including developing novel labeling approaches to improve detection thresholds and early delineation of toxicity and function. Key areas for future research include addressing safety concerns associated with genetic labels and developing methods to follow cell survival, differentiation and integration with host tissue. Imaging may bridge the gap between cell therapies and health outcomes by elucidating mechanisms of action through longitudinal monitoring.

Introduction

Many tissues and organs in the human body, such as the heart, brain and spinal cord, cannot regenerate in response to disease or trauma; damage leads not to restoration of structure and function but to an inflammatory response and scar formation. Regenerative medicine aims to achieve functional recovery of damaged tissues by providing specific cell populations, alone or incorporated in biomaterial scaffolds, that enhance the body’s intrinsic healing capacity¹. The field has seen considerable progress in several areas, including development of new sources of transplantable cells and improved approaches to test the safety and efficacy of experimental therapies. However, many difficult challenges remain. Transplantation into diseased tissues is a stressful experience for cells. Most cells leak out from the injection site or die through multiple mechanisms². The hardy survivors have to migrate, proliferate and self-organize into a tissue, integrate functionally with the host parenchyma and recruit a vascular supply to support their long-term survival and function. Transplanted cells are often immature and are required to mature in situ. All of these processes must occur in a damaged tissue environment that is hostile to transplanted cells. After excessive cell loss, most tissues produce scars, such as a fibro-collagenous scar in the heart or a glial scar in the central nervous system. The scarring response is one of the banes of regenerative medicine because it blocks integration of stem cell grafts with surrounding host tissue. Other hostile factors in damaged tissues include an ischemic environment that limits oxygen and nutrient delivery, and acute and chronic inflammation, which generates reactive oxygen species and injurious cytokines.

With a few notable exceptions, such as pluripotent stem cell–derived cardiomyocytes, dopaminergic neurons and pancreatic islet cells, most cell therapies have not acted like the ‘bricks and mortar’ repair cells we originally imagined. The limited functional restoration that has been achieved may be elicited by paracrine mechanisms that are poorly understood. Rational improvement of the design of cell therapies and of clinical protocols for administering them will require a deeper understanding both of the cells and of their fate after transplantation as they interact with an inhospitable environment. Some of these questions can be elucidated through imaging. Each organ in the body has unique imaging characteristics that make it possible to differentiate pathological and regenerative responses to therapy. Many aspects of impaired organ function can be readily identified by clinical imaging, including gross morphological changes (e.g., ruptures, cystic cavities, geometric remodeling); volume loss (e.g., atrophy); hemorrhage or iron by-products; thrombosis; hypoperfusion; increased vascular permeability and resulting edema; metabolic shifts compared to surrounding parenchyma; oxygenation; fibrosis; and loss of function. Such data can provide an indirect measure of graft function and shed light on therapeutic mechanisms. However, direct tracking of the fate and function of transplanted cells in humans is at an early, experimental stage and is subject to substantial limitations. Imaging approaches for cell therapies based on the action of secreted paracrine factors is fundamentally different from imaging of cell replacement and requires the ability to track the structural and functional changes within the host organ.

This review explores the benefits and limitations of clinical imaging technologies for applications in regenerative medicine. We discuss imaging modalities that may have a role in clinical trials within a decade rather than early-stage, experimental technologies or approaches that are limited to animal research. We apologize to colleagues whose work is not included, and we refer interested readers to recent reviews^{3, 4, 5, 6, 7}. We begin with a short primer on the capabilities of clinical imaging, which is subject to more constraints than the same technologies used in animal models. Next, we discuss cell-labeling approaches that are in the clinic or in development, and their applications for tracking transplanted cells and understanding the host environment and the response to therapy. These issues are then explored in detail for three tissues—the heart, pancreatic islet cells and the brain—chosen because they are among the most advanced in human translation and because, at least in some instances, they include cell grafts that survive long term in the host environment. We propose recommendations to researchers interested in incorporating imaging in their studies and conclude with a discussion of important areas for future research.

Capabilities and limitations of clinical imaging in regenerative medicine

Imaging is used routinely in clinical practice both to provide secondary endpoints that complement primary health outcomes and to investigate the mechanisms of therapeutic successes and failures. In early-stage clinical trials that are not powered for endpoints such as mortality or rehospitalization, imaging is often the primary ‘surrogate’ endpoint—for example, to assess structural and functional responses of the tissue or to monitor adverse reactions such as immunogenicity and tumorigenicity⁸. The main clinical imaging modalities in use today are magnetic resonance imaging (MRI), computed tomography (CT), positron emission tomography (PET), single photon emission tomography (SPECT) and ultrasound, as well as multimodality methods that rely on co-registration of images (Box 1). These technologies have improved considerably in recent years and now provide sophisticated multimodality imaging, shorter acquisition times and lower radiation doses. The latest clinical scanners offer new opportunities in regenerative medicine for assessing the tissue composition of organs, monitoring transplanted cells and evaluating the effects of therapy on tissue function. They are beginning to be applied to non-invasively assess the survival, migration, biodistribution, and differentiation of transplanted cells and the underlying mechanisms of cell therapy⁹. However, imaging in humans is inherently more constrained than imaging in animals owing to limits on radiation, image acquisition times and, more importantly, limits on the use of genetically modified cells for long-term cell tracking. These limits translate into reduced sensitivity, specificity and ability to monitor changes over time. The benefits and challenges of clinical imaging in regenerative medicine are summarized in Box 2.

Box 1. Clinical imaging modalities.

A limited number of medical imaging technologies are currently approved for clinical use. We provide brief descriptions of them below.

Magnetic resonance imaging

MRI uses magnets (from 0.5–7 Tesla magnetic field strength) to polarize the hydrogen nuclei in water molecules in human tissues. Combinations of time-varying gradient magnetic fields and pulse sequences of radio frequency waves provide the spatial distribution of signals emitted from protons, which are displayed as high-resolution, multidimensional images. MRI does not use ionizing radiation and can be performed serially over time. It is the most versatile clinical imaging modality, with superior sensitivity in detecting morphology, pathology and function.

X-ray computed tomography

CT uses computer-processed X-rays to produce tomographic images. The disadvantage is the use of ionizing radiation, which damages DNA. Two new approaches under development rely on more exotic forms of radiation: proton beams and synchrotron radiation. Proton CT records the position, direction and energy loss from a proton beam as it traverses a patient’s body, providing a more detailed image of the body’s density and requiring less radiation exposure. CT images based on synchrotron X-rays have much higher photon energies than conventional X-rays, reducing radiation dose.

Positron emission tomography

PET is a tomographic technique that produces images of functional processes in the body through detection of biologically active positron-emitting radio-tracer, such as fluorine-18, attached to a small molecule. The short lifetime of radioligands requires fast image acquisition the same day as tracer synthesis. Sensitivity is very high and there is no limit on tissue penetration depth, so PET can be used to track cells expressing reporter proteins. PET scans are usually co-registered with MRI or CT images.

Single photon emission tomography

SPECT is a tomographic technique that uses gamma rays to detect radioactive probes, such as technetium-99m (^99mTc) and indium-111 (¹¹¹In), injected into the bloodstream. Sensitivity is very high and there is no limit on tissue penetration depth. There is only one FDA-approved SPECT agent (¹¹¹In oxine). SPECT can report whole-body biodistribution of injected radiolabeled cells but does not provide high-resolution anatomical information that allows a precise delineation of cellular location. The main limitation of this method is signal decay; labels cannot be tracked over weeks, the signal diminishes as cells divide and the radio-labeled components are metabolized. Ionizing radiation may cause DNA damage, so exposure to radioactivity must be closely monitored.

Ultrasound

Images are generated by the reflections and echos of an oscillating sound wave on tissues. Ultrasound provides images in real time (no processing delay after an acquisition), can be brought to a sick patient’s bedside, is substantially lower in cost than other imaging modalities and does not use ionizing radiation. Ultrasound is widely used in the clinic and in research to evaluate the structure, function and blood flow of organs. Sufficient image resolution for cell tracking requires high frequencies, but the depth penetration of high-frequency ultrasound waves is limited. Therefore, ultrasound is not widely used to track transplanted cells.

Multimodality imaging

Multimodality imaging involves co-registration of images obtained from different scanners. It combines high-sensitivity but low-resolution methods, such as PET and SPECT, with high-resolution anatomical images acquired with MRI or CT scanners. Examples include MRI/PET, PET/CT, SPECT/CT and ultrasound/CT.

Box 2. Benefits and challenges of imaging in regenerative medicine.

Diagnostic and interventional imaging can be incorporated into all aspects of cell therapy research and clinical trials. The main benefits and challenges are summarized below.

Subject selection

Imaging is essential for determining whether subjects meet inclusion or exclusion criteria for entry into clinical trials. Imaging results should be part of inclusion and exclusion criteria for subjects entering into the study. Investigators should verify that accepted subjects would tolerate imaging studies.

Subject evaluation

Diagnostic imaging is essential for defining the extent of pathology and organ dysfunction to be treated. Interventional imaging can aid evaluation of organ vascular anatomy and flow.

Route of administration

Imaging is essential for selecting the best route of administration (e.g., direct implantation, intravascular, intranasal, endoscopic, surface application).

Patient safety

Imaging of cell homing and dispersion is useful for evaluating potential toxicity, graft-versus-host disease, tumor formation and undesirable changes in the host tissue.

Host microenvironment

Multimodality imaging is valuable for assessing whether the targeted region can support the cell graft.

Cell tracking

A challenge is to develop labeling approaches that uniquely identify cells over the desired temporal and spatial scales.

Cell dosing

A challenge is to use clinical imaging to determine the potency, volume of distribution and lifespan of the cell product, and to develop a protocol for administration to patients. Novel labels to track the fate and function of cell products will bring mechanistic insights into clinical trials.

Label dilution and transfer

Loss of exogenous labels by cell division limits long-term tracking. A challenge is to develop agents that remain within cells, turn off upon cell death or are not transferred to host cells such as macrophages. A change in the imaging signature of the label may indicate metabolism or cell engraftment. A decrease in the signal volume may indicate cell death or migration away from implantation site.

Mechanism of action

A challenge is to develop imaging strategies that can reveal the mechanisms of action of transplanted cells (e.g., cell replacement or paracrine effects).

Cell dose

The possibility of imaging transplanted cells depends in part on their number. The optimal cell dose required for a therapy is determined empirically, guided by the normal cell content of the tissue or organ (Fig. 1) and the cell deficit in individual patients. It is important to remember that all organs have reserve capacity and that disease ensues only when cells are depleted below a critical threshold. This means that full regeneration is not necessary for substantial improvement in organ function or symptoms.

Figure 1.

The human body contains ~3.7 × 10¹³ cells¹⁰.

The cell number and weights of individual organs^160,161 provide a baseline for understanding the numbers of cells that may be needed for replacement therapies and the associated challenges for imaging.

When cell products are injected intravenously, a typical dose is 1–5 × 10⁶ cells/kg (or 70–350 × 10⁶cells in a 70 kg person). In a 5-liter blood volume containing ~10¹⁰ nucleated cells¹⁰, this amounts to an initial dilution of 1 cell in 10³–10⁵ blood cells, making initial tracking of transplanted cells akin to finding a needle in a haystack. As the cells distribute throughout the body, they localize primarily to the lungs, liver, spleen or marrow spaces, with relatively few cells homing to diseased sites^{11, 12,13}. The alternative to intravenous injection is direct injection into tissues. In these therapies, doses may be smaller than ~10⁵ (e.g., for repair of the cornea or the retinal pigment epithelium¹⁴) or larger than ~10⁹ cells (e.g., to rebuild a heart). To be detectable by clinical imaging, at least ~10⁴–10⁶ cells must engraft in a specific location in the body.

Below we discuss in detail the prospects for imaging transplanted cardiomyocytes, pancreatic islets and neural cells. The left ventricle of the heart has ~20 million cardiomyocytes per gram of tissue¹⁵, or ~4 billion in total. A myocardial infarction that causes heart failure can kill ~25% of these cells¹⁶, or ~1 billion cardiomyocytes. Cell therapy for diabetes involves similar numbers of cells. A healthy human pancreas has about ~1 million islets of Langerhans, each containing ~2,000 cells, of which 65–80% are insulin-producing beta cells, for a total of 1.3–1.6 × 10⁹ beta cells. The dose of beta cells needed to dispense with insulin therapy can be estimated from current protocols for transplantation of cadaveric islets into the liver. Patients generally receive two infusions of 5,000–11,000 islet equivalents (IEQ) per kilogram (or 0.7–0.9 × 10⁹ beta cells IEQ/75 kg human) per infusion¹⁷. After transplantation through the portal vein, islets are lodged as single islets or islet clusters in the capillaries of the liver, where they start secreting insulin. However, the majority of these cells die soon after transplantation, and optimized therapies may require fewer cells.

The projected cell dose needed to treat neurological conditions varies widely, mostly depending on whether the aim is to replace cells or to deliver therapeutic factors. For instance, the putamen region of a healthy brain has ~7.61 ng/mg of dopamine¹⁸ supplied by ~500,000 dopaminergic neurons from the substantia nigra. In a cell therapy for Parkinson’s disease, transplantation of only ~138,000 fetal neural cells showed efficacy, increasing dopamine release from 25% to 62% of normal levels¹⁹. However, replacement or regeneration of an anatomical region, such as the putamen, may require ~7 × 10⁶ cells^{20, 21}.

Whether one is transplanting cardiomyocytes, islet cells, neural cells or other cell types, the ability to detect them ultimately depends on the cell density in the target tissue and the sensitivity and specificity of the imaging technique (Table 1).

Table 1.

Characteristics of clinical imaging modalities

Modality	Applications	Labels (exogenous, endogenous reporter gene)	Main characteristics	Depth of penetra- tion	Sensitivity	Image resolution (voxel size)	Acquisition time	Injected reagents	Clearance of label after cell death	Disadvantages
MRI	Anatomy Pathology Metabollsm Chemical exchange Physiology Function Intervention Cellular Molecular	Iron oxide particles Gd chelates Microcapsules with fluorine sodium carbon Ferritin Lysine-rich protein protamine	Versatile High soft tissue contrast Metabolite concentrations Perfusion Characterization of the microenvironment Short- and medium-term cell tracking Graft size	No limit	Moderate	~<1–3 mm3	Microseconds to minutes	Stable Commercially available	Slow	Not compatible for patients with implants Acoustic noise Specialized coils (19F 23Na, 13C)
PET	Metabolism Physiology Function Cellular Molecular	18FDG 18FHBG 18FDOPA	Graft proliferation Tissue viability Inducible cell death Longitudinal serial imaging Differentiation possibilities	No limit	High	~3–5 mm3	Minutes to hours	Labile Commercially available, Custom synthesized	Rapid Depends on T1/2 Isotope	Ionizing radiation Biohazardous labels
SPECT	Metabolism Physiology Cellular Molecular	Radioisotopes: Tc-99m In-111 NIS Dopamine receptor	Distribution of injected cells	No limit	High	~5 mm3	Minutes to hours	Labile Commercially available Custom synthesized	Rapid	Ionizing radiation Limited short-term cell tracking
X-ray, CT	Anatomy Pathology Intervention Vascular delivery Biopsy	Microcapsules Barium Gold nanoparticles	Multimodal pairing with PET, SPECT Bone, lung	No limit	High	<1 mm3	Seconds	Stable Commercially available	Slow	Ionizing radiation Not suitable for soft tissue imaging
Ultrasound	Anatomy Pathology Intervention Vascular delivery Biopsy	Microbubbles Microcapsules Liposomes	Real-time fast imaging	Trade-off with spatial resolution	Moderate	1 mm	Seconds to minutes	Stable Commercially available	Rapid	Tradeoff between depth of penetration and spatial resolution Limited field of view

Limits of detection

Clinical imaging is a macroscopic technology and has limitations when applied to the detection of transplanted cells. The cell density in the target tissue corresponds to the number of labeled cells per voxel, the size of which varies with the imaging technique. The sensitivity and specificity of the imaging technique is a function of the concentration of the labeling agent in the cells and the contrast-to-noise ratio on the acquired images²². Uptake or expression of labels and their retention over time varies for different cell types and labels. Imaging contrast can be lost or diluted with cell division. Transplanted labeled cells, or free label released by dying cells, can be engulfed by macrophages, confounding interpretation of the imaging signal. Phagocytosis by macrophages may begin minutes to hours after transplantation and can continue for months, depending on the lifespan of the graft. Imaging of single cells^{23, 24} or small clusters (~50 cells)^{22,25, 26, 27} of labeled cells has been reported in animal models (using 7 T²³ and 11.7 T²⁴ MRI scanners) but in clinical imaging, in the best-case scenario, at least 600 labeled cells per voxel are needed to produce a detectable signal²⁸. However, tracking of single or small numbers of cells is unnecessary and, for the foreseeable future, unattainable by clinical imaging.

Cell labeling

Cells transplanted in humans cannot be imaged with current technologies unless the cells are first labeled in vitro. There are two main labeling strategies: direct labeling, by nanoparticles or chemical agents, and indirect labeling, by genetic incorporation of reporter genes^{8, 29, 30, 31, 32} (Fig. 2). The ideal cell label for clinical applications would be nontoxic, would be retained by target cells over sufficiently long time periods at high concentration, would correlate stoichiometrically with cell number, would be cleared rapidly after cell death and would raise no long-term safety concerns. Unfortunately, none of the available labeling agents satisfy all of these criteria.

Figure 2.

Tracking cell fate by noninvasive imaging requires either direct or indirect labeling.

(a) Direct labeling. Exogenous labeling with either MRI-based contrast agents or radioprobes for PET or SPECT imaging. Cells take up SPIO or ¹⁹F nanoparticles (Fnp) primarily through endocytosis, whereas ^99mTcHMPAO or ¹¹¹Indium oxine are lipophilic and pass through cell membranes by passive diffusion. FDG is a glucose analog and is taken up though glucose transporter channels on cells. Small molecules can attach to cell surface markers or enter cells through channels. (b) Indirect labeling. Reporter genes introduced into the genome express surface proteins, channels, storage proteins or enzymes that are detectable or that bind detectable probes. HSV1-tk, herpes simplex virus thymidine kinase; NIS, sodium iodide symporter; LRP, lysine-rich protein; SPIO, superparamagnetic iron oxide nanoparticles.

Direct labeling

The most common contrast agents used for direct labeling are nanoparticles or radionuclides^{33, 34, 35}. Nanoparticles suitable for MRI include superparamagnetic iron oxide (SPIO) nanoparticles (i.e., ferumoxides, ferucarbotran-1, ferumoxytol), gadolinium-filled microcapsules and liposomes, perfluorocarbon nanoparticles³⁶ and manganese-based particles³⁷ (Table 1). Advantages of nanoparticle-based MRI cell tracking are a strong signal, allowing high-resolution visualization of the migration and homing of injected cells, as well as relative ease of cell labeling. SPIO nanoparticles can also be detected by ultrasound, although this approach is not widely used³⁸. SPIO-based labeling has been used clinically to track transplanted cells in the brain for up to 7 weeks by MRI^{39, 40}. Labeling cells with perfluorocarbon nanoparticles is a promising approach that is being tested in clinical trials.

Direct labeling with radionuclides for SPECT and PET imaging has excellent sensitivity compared with MRI, CT and ultrasound because the background signal is low, allowing detection of as few as ~10⁴–10⁶ cells/voxel⁴. Radionuclide labeling is readily quantified and can offer even greater sensitivity than MRI, but half-lives are relatively short, and the limited spatial resolution of clinical PET and SPECT requires image co-localization with CT or MRI. Moreover, concerns regarding radiation exposure may limit these approaches⁴¹. Direct labeling with radioisotopes such as¹¹¹Indium (In)-oxine or ^99mTc chelates is used in the clinic for SPECT imaging of inflammation⁴². The relatively long half-life (67 h) and high in vivo stability of ¹¹¹In-oxine provides the option to acquire images after 24 h or more, whereas ^99mTc chelated agents have a relatively short half-life (6 h). ¹⁸F-fluorodeoxyglucose (FDG) also can be taken up and metabolically trapped by the cells and subsequently tracked in vivo^{43, 44}.

Although direct labeling gives strong signals in the first days after cell transplantation, it has several drawbacks for long-term imaging. First, it cannot distinguish live and dead cells^{34, 45, 46}. An ideal contrast agent would dissipate or be metabolized after the death of a labeled cell, but nanoparticles released from dead cells are phagocytosed by macrophages^{47, 48}, which produce imaging signals identical to those of labeled cells. Soluble radioisotopes may be less prone to this artifact; to our knowledge, their uptake by macrophages has not been studied. Second, contrast agents are diluted as cells divide, resulting in a gradual disappearance of the signal^{49, 50, 51}. Third, imaging signals can be difficult to distinguish from background. For example, SPIO nanoparticles are detectable in MRI as areas of decreased signal intensity, and similar signal voids can be caused by iron depositions in the form of hemosiderin from old hemorrhage⁵². Finally, all contrast agents raise concerns of cellular toxicity, particularly at high concentrations. High doses of SPIOs have been shown to inhibit mesenchymal stem cell (MSC) migration, colony formation ability⁵³ and chondrogenesis⁵⁴, although these results are controversial^{55. 111}In-oxine labeling may impair cell proliferation⁴².

Indirect labeling

In indirect labeling, cells are genetically modified to express reporter genes encoding proteins that generate imaging signals, often upon interaction with a molecular probe or substrate that is taken up by cells. Reporter genes incorporated into the genome are propagated by daughter cells, and the imaging signal is stoichiometrically related to live cell mass. Episomal expression of reporter genes is also possible, but dilution of episomal vectors during cell division makes them unsuitable for long-term tracking of mitotically active cells. Reporter-gene imaging is widely used in animal research, but clinical applications for use in tracking cells to treat cardiovascular or central nervous system diseases have been limited by concerns about the safety of genomic integration and about potential immune responses to some of the foreign reporter proteins.

A variety of reporter proteins detectable by different imaging modalities have been developed. Firefly luciferase, detectable by bioluminescence imaging, is often used in small-animal research³; however, absorption of the emitted light precludes its use in most large-animal and human studies. For large-animal and clinical applications, there are several potential options.

The iron-storage protein ferritin is detectable by MRI as voxels with reduced signal intensity^{56, 57,58}. Because ferritin is ubiquitous in most organisms, it may be possible to use it clinically. However, MRI sensitivity to ferritin is lower than to SPIO nanoparticles (work of A.V.N., C.E.M. and colleagues)⁴⁶.

Herpes simplex virus thymidine kinase type 1 (HSV1-tk) is frequently used for PET imaging in large animals^{32, 59}, and it has been incorporated into sensitized, cytolytic T-cells to image metastases in patients with glioblastoma using the ganciclovir analog 18F-9-(4-[¹⁸F]fluoro-3-hydroxymethylbutyl) guanine (¹⁸FHBG)⁶⁰. HSV1-tk phosphorylates ¹⁸FHBG, trapping it within cells. Because mammalian thymidine kinases have low affinity for this reagent, it is possible to distinguish transplanted HSV1-tk cells from surrounding tissues by PET^{32, 59, 61, 62}.

Another approach is to express reporter proteins on the cell surface. Biotinylated peptides expressed on the plasma membrane were imaged with a contrast agent conjugated to streptavidin in rodent models^{63, 64}. The human membrane-protein sodium-iodide symporter (NIS), responsible for iodide uptake in the thyroid, stomach, salivary gland and choroid plexus but minimally expressed elsewhere, was used to follow the survival and distribution of transplanted cells in pigs by hybrid SPECT/CT³². The cells were pre-labeled with iodine-123, permitting immediate tracking of the grafts, and they could be imaged long term by systemic administration of iodine-123. This approach is especially promising for clinical applications because a human protein should not be immunogenic. Radioactive iodine scans are common in clinical medicine and could be adapted to cell therapies.

The chemical exchange saturation transfer (CEST) technique uses radiofrequency saturation pulses to detect molecules containing protons that exchange rapidly with water. The exchanging intracellular protons have a unique offset resonance that distinguishes them from extracellular water protons. Exchange rate and CEST contrast depend strongly on the pH^{65, 66, 67}. Because cell death and inflammation are associated with extracellular acidification, CEST probes can be used as sensors of cell viability⁶⁸. For example, CEST contrast based on amine proton transfer has identified heterogenous regions within gliomas in patients; this technique might be helpful in identifying the infiltration of tumors into surrounding parenchyma and tumor grade⁶⁹. Unique, genetically encoded, CEST-detectable amino acids and proteins (e.g., L-arginine, lysine-rich protein and protamine) enabled imaging of transplanted cells by MRI in experimental models³¹. The threshold for detecting CEST protein–tagged cells by clinical MRI field strengths may be 10⁴–10⁶ cells/voxel²². Recently, a novel reporter probe, 5-methyl-5,6-dihydrothymidine (5-MDHT), was described for detection of HSV1-tk expression with CEST imaging⁷⁰; this opens up the possibility of using MRI for HSV1-tk imaging. The main limitation of CEST imaging is difficulty in distinguishing the reporter from intrinsic macromolecules.

The major disadvantage of indirect labeling for clinical studies relates to safety concerns about the effects of genetic manipulation. Insertion of reporter genes in the genome may alter cell potency, immunogenicity⁷¹, differentiation potential and tumorigenicity⁷². Another issue is reporter-gene silencing, which can be misinterpreted as cell death. As with direct labels, reporter proteins in dead or live cells can be phagocytosed by macrophages, resulting in short-term, false-positive signals^47,48, but this is not an issue for long-term imaging.

Choosing a labeling strategy

In evaluating whether to use direct or indirect labeling, it is important to assess cell viability, cell proliferation and contrast-agent retention³⁰. The choice of label may also depend on the culture conditions for maintaining and expanding the cells and on whether the therapy is autologous or allogeneic. In our view, direct labels are best suited to study events in the first few days rather than months to years after transplantation. A potentially important application of direct labels is real-time MRI guidance of cell delivery, as no other imaging modality provides as precise anatomical information about cell migration and short-term biodistribution in vivo³⁹. In our view, indirect labeling approaches are best used when it is necessary to monitor the fate of implanted cells over weeks, months or years in order to ensure that engraftment has occurred in the targeted tissue. An additional advantage of indirect labeling is the potential to turn a reporter gene on or off and to monitor differentiation of implanted cells into desired functional phenotypes. It should also be noted that genetic tagging of cells requires additional clearance through the Recombinant DNA Advisory Committee, which may delay approval of clinical trials.

Imaging of cell fate and host environment

In general, very little is understood about what happens to cells after they are transplanted and how they interact with the host environment, which makes it difficult to design improvements to the therapy. Recent imaging studies show potential for addressing these questions, but much requires further investigation.

Imaging cell fate

It would be useful to be able to characterize basic phenomena such as cell retention, survival, migration, proliferation, differentiation and integration into the host. The most progress has been made in measuring retention, or biodistribution, of cells soon after transplantation. For example, in patients with dilated cardiomyopathy, biodistribution of ~10^{8 99m}Tc–labeled CD34⁺ hematopoietic stem cells delivered by intracoronary or transendocardial injection was monitored by SPECT. The cells cleared the lungs and localized primarily to the liver and spleen, with relatively few cells homing to the heart⁷³ (Fig. 3).

Figure 3.

Examples of clinical imaging used to identify and track labeled cells in the body.

(a) Coronal SPECT images of chest and upper abdomen obtained 18 h after intracoronary (IC) versus transendocardial (TE) delivery of about 10^{8 99m}TcHMPAO CD34⁺ hematopoietic stem cells in patients with nonischemic dilated cardiomyopathy. IC route of administration clearly shows less retention of cells in the myocardium compared to TE route with both delivery techniques showing the distribution of CD34⁺ cells in liver and spleen (IC > TE). This comparison demonstrates the value of short-term labeling of a cell product and the value in assessing the cell delivery route. Figure from ref. 73 reprinted with permission. (b) PET/CT fused axial multimodal imaging of pig chest performed on clinical scanner. MSC transduced by adenovirus containing cytomegalovirus promoter driving HSV1-tk reporter gene implanted with matrigel into porcine left ventricle (LV) myocardium after thoracotomy (long arrows). T is the chest tube inserted during surgery. 10⁸ human MSCs injected into the myocardium were visualized (short arrows) 4 h after intravenous infusion of 9-(4-18F-3-[hydroxymethyl]butyl)-guanine, a thymidine analog that is phosphorylated by HSV1-tk reporter following uptake by cells. % ID/g is percentage uptake per gram of tissue. Figure from ref. 32 reprinted with permission. (c) Coronal MRI performed at 3 T of liver from a diabetic patient baseline, day 1 and day 7 following portal vein infusion of 4.81 × 10⁵SPIO-labeled islet equivalents demonstrating hypointense voxels throughout the liver (white arrows) that cleared rapidly over 24 weeks of imaging follow-up. The number of hypointense areas in liver decreased by about 50–60% in the first week after infusion, representing rapid clearance of the transplanted islets despite immunosuppression of the patient. Figure adapted from ref. 104 with permission. (d) Coronal and axial, T2-weighted, 1.0 T MRI following implantation of fetal neurons in a patient with Parkinson’s disease. Coronal MRI displays the cell transplantation needle track (arrowheads) following injection of ~3.2 × 10⁶ cells. ¹⁸F fluorodopa PET parametric map fused with MRI performed before (PRE) and 3 years post implantation (PI) of fetal cell neurons that matured with time that resulted in increased uptake of the dopamine analog by the innervated cells in the putamen. Figure adapted from ref. 76 with permission. (e) T2-weighted, axial, 3 T MRI before (PRE) and day 1 after implantation in the left temporal lobe of SPIO-labeled autologous neural stem cells in patient with traumatic brain injury (* is site of injury). Magnified serial T2*-weighted, axial MRI performed on days (D) 1,7,14,21 post implantation (PI) of labeled cells are shown. Four hypointense areas (black arrows) were observed on PI days 1,7,14, 21 around lesion site (*) followed by migration of cells (white arrowhead and arrows) along the border of the damaged area. Hypointense areas where labeled cells were injected cleared over time (D14, D21), and by week 7 PI, areas were no longer visible on MRI. Figure adapted from ref. 40 with permission.

Cell survival in the target organ is often monitored in preclinical studies, for example, using PET imaging with reporter-gene labels⁶². However, few labeled cell–tracking studies have been carried out in humans. The problem of cell death after transplantation remains a major roadblock to regenerative therapies. Here, researchers could contribute by moving reporter-gene studies into large animals and by working with regulators to establish safety criteria for clinical trials of these labels.

Proliferation of transplanted cells can also be assessed using reporter genes because the imaging signal is passed to daughter cells^{8, 74}. The most common approach, demonstrated in animals, is to image serially and to estimate growth kinetics from the rate of change in population size. Repetitive, long-term (up to 5 months) PET/CT studies of MSCs expressing HSV1-tk were conducted with a porcine model of myocardial infarction. Cell proliferation peaked at 33–35 days after injection in peri-infarct regions, but surprisingly, most of the transplanted cells were found in the major cardiac lymphatic vessels and lymph nodes⁷⁵.

Imaging differentiation of transplanted cells has received relatively little attention, and there are interesting opportunities in this area. Promising approaches include genetically integrated stage-specific promoters and labeled small molecules that interact with stage-specific cellular targets. The latter approach is well suited to PET and SPECT if the small molecule can be slightly modified by addition of a radiolabel. For example, the PET agent ¹⁸F-DOPA was used to monitor maturation of fetal neurons transplanted in the brain of Parkinson’s disease patients (Fig. 3) as maturation increased uptake of the dopamine analog⁷⁶. Approximately 10⁴ functioning graft cells were detected with PET images overlaid on MRI scans. However, the use of small-molecule probes requires agents that can cross cellular barriers and vasculature (i.e., the blood-brain barrier) and interact specifically with their targets. The development of small-molecule agents and protocols will help guide clinical trials and should be a priority.

Imaging the host environment

Despite its many limitations for tracking transplanted cells, clinical imaging is indispensable in regenerative medicine for understanding the host environment. Imaging can answer many specific questions about tissues both for pretreatment planning and for longitudinal evaluation of response to therapy (Table 2). For example, what is the status of the targeted region with regard to various pathophysiological processes—inflammation, residual stunned or dormant native parenchyma, cellular and morphological alterations, and poor perfusion? Is the targeted region metabolically active or hypoxic, and can it support transplanted cells? Multimodal imaging of cell proliferation, tissue structure, perfusion, vascularity, permeability and metabolism can help delineate the feasibility of cell therapy and measure therapeutic outcomes. Initial and post-transplant assessment of the target tissue may include evaluation of structural integrity, tissue loss (e.g., atrophy, infarction), cellular composition (e.g., fibrosis, gliosis, percentage of viable cells) and inflammation, as well as underlying tissue infrastructure (extracellular matrix, vascularity), perfusion (e.g., blood flow and volume), oxygenation, acidosis (pH), and metabolic state (ATP, lactate, phosphocreatine content). Measuring hormone and neurotransmitter production, contractility and responses to external stimuli can report on the recovery of organ function. Imaging should also be useful for ensuring that cells are delivered to the optimal site (i.e., non-necrotic regions) at the optimal time and dose.

Table 2.

Interrogating the host microenvironment by clinical imaging

Tissue characteristics	Imaging techniques
	MRI	MRSI	X-ray angiography	CT	SPECT/CT	PET/CT	Ultrasound
Anatomy (structure, volume loss, vasculature)	++++	−	++++	+++	+	+	+++
Vascular permeability	+++	−	−	++	+	+	++
Pathology (e.g., mass)	++++	−	−	++	+	++	+
Compositions (e.g., fibrosis)	+++	−	−	++	++	++	++
Inflammation (acute, subacute, chronic)	+++	++	−	−	++	++	−
Physiology and function	++++	−	−	++	+++	+++	++
Perfusion	++++	−	−	−	−	++
Metabolism	−	+++	−	−	−	++++	−
Oxygenation	+++	−	−	−	−	++	−
Cell viability	−	+	−	−	−	++++	−
Cell identity	−	−	−	−	+++	++++	−

MRSI, magnetic resonance spectroscopy imaging. MRI techniques include blood oxygenation level dependent contrast, diffusion tensor imaging, MR elastography, dynamic contrast enhance MRI. − not useful; + limited to ++++ most useful.

Imaging in selected organs

Here we discuss three organ systems in more depth to illustrate the advances and limitations of noninvasive imaging in regenerative medicine. Studies in heart, islets and brain share common challenges, including cell-seeding efficiency, graft cell death and achieving host integration. In the three settings, combined imaging approaches are useful in delineating organ morphology and function before and after cell therapy. Differences in routes of administration, cell labeling approaches and detection limits, which depend on organ characteristics, underscore challenges specific to each system.

Imaging myocardial regeneration

The heart is one of the least regenerative organs in the body. Whole-heart transplantation is currently the only definitive treatment for end-stage heart failure, and it is limited by organ availability to fewer than 0.1% of heart failure patients. Infarction causes significant loss of cardiomyocytes (often 25% of the left ventricle), which cannot be restored by current pharmaceutical therapies. Stem cells offer the possibility of rebuilding the damaged heart from its component parts. We submit that the following three conditions must be achieved to prove true myocardial regeneration: (i) an increase in the volume of viable myocardium within the infarct zone; (ii) structural integration of the new myocardium with the host tissue, including restoration of myocardial fiber architecture; and (iii) functional integration of the new myocardium with host myocardium—that is, synchronous contraction of new and old host myocardium without conduction delay or arrhythmias. Medical imaging is essential in the evaluation of cardiac repair by regenerative therapy. Measureable end points include heart contractile function, morphology, vascularity, inflammation, infarct size, tissue viability and metabolism. In animal studies with genetically labeled cells, it is possible to demonstrate formation of even small amounts of new myocardium by MRI or PET. In humans receiving genetically unmodified cells, it should be possible to identify new myocardium by MRI if it is large enough (i.e., >2% of the left-ventricle mass) and reasonably localized. Structural alignment of grafted cells with host myofibers can be visualized in humans with diffusion tensor MRI⁷⁷. However, demonstrating functional integration currently requires more-invasive techniques.

Clinical imaging studies have shown that cell therapy in the human heart to date is hampered by low retention of transplanted cells. In patients with acute myocardial infarction, low retention (1.5%) was observed 2 h after intracoronary administration of ¹⁸F-FDG–labeled peripheral hematopoietic stem cells⁷⁸. Intracoronary delivery of ¹¹¹In-labeled peripheral blood–derived progenitor cells led to retention of ~6.9% in acute infarct patients and ~2.5% in a subgroup of chronic infarct patients⁷⁹. Low viability of the infarcted myocardium and reduced coronary flow reserve were significant predictors of pro-angiogenic progenitor cell homing⁷⁹.

Imaging studies have also revealed that retention in the ischemic myocardium depends on the delivery method. In the hematopoietic stem cell trial discussed above⁷³, at 18 h after the procedure, myocardial retention was higher in the transendocardial group than in the intracoronary group. Transendocardial delivery also correlated with greater improvement in ventricular function 6 months after infusion. In a 5-year follow-up study⁸⁰, cell therapy was associated with increased left ventricular ejection fraction, increased 6-minute-walk distance, and decreased N-terminal, B-type natriuretic peptide (a biomarker of heart failure). Left-ventricular ejection fraction improvement was greater in patients with higher myocardial homing of injected cells. These studies demonstrate the advantages of direct implantation compared with intravascular injection for cardiac cell therapies. However, direct implantation has several challenges, including minimizing invasiveness, increasing graft viability and promoting vascularization. Imaging researchers can contribute by developing guidance systems for cell delivery and by designing systems that integrate blood flow dynamics with tissue mechanics.

Systematic comparisons of different routes of administration have been carried out in animal models. MRI and PET imaging of acutely infarcted rats showed that direct injection afforded 14% retention, substantially greater than rates with injection into the left ventricular cavity (3.5%) or intravenous delivery (1.2%)⁸¹. Another biodistribution study found that a substantial number of bone marrow mononuclear cells injected into the rodent myocardium actually migrated to the bone marrow, liver and spleen⁸².

The use of imaging to reveal the low retention of transplanted cells provides a possible explanation for poor clinical results and suggests strategies for improving therapeutic efficacy. Retention has been increased by injecting cells in a hydrogel made of natural or synthetic materials. This approach has been tested in pigs using human MSCs expressing HSV1-tk and PET/CT imaging³²(Fig. 3). Surprisingly, myocardial radiotracer (¹⁸FHBG) uptake was not elevated when 200 × 10⁶human MSCs alone were directly injected into the myocardium, despite the grafts being readily detectable by histology. When the same number of cells was injected in Matrigel, signal-to-background ratio increased to 1.87, and increased further to 8.02 when 600 × 10⁶ cells were administered. The relatively low sensitivity for PET detection in this study may relate to timing: imaging was done the same day as cell transplantation, so delivery of radiotracer required its diffusion into avascular cell clumps. Presumably, a vascularized graft would be more easily detected.

Once cells are successfully retained in the heart, the next challenge is the wave of cell death that is initiated by ischemia, loss of survival signals due to matrix detachment, and the toxicity of reactive oxygen species and cytokines². In practice, one can consider cell loss over the first few hours after injection to be a retention problem, whereas loss after that is likely due to death. Imaging of long-term survival of transplanted cells in the heart has not been tested in humans, but animal studies using indirect labels show the feasibility of monitoring cells for as long as 15 weeks⁸³. For example, cell tracking in the infarcted mouse heart with the MRI reporter gene ferritin showed that the signal attenuation by ferritin-containing grafts persists for at least one month and correlates well with histologically determined graft size^{57, 58} (work of A.V.N., C.E.M. and colleagues). One drawback of ferritin is a relatively low sensitivity, requiring development of complicated, time-consuming imaging protocols. A more sensitive MRI reporter gene, and particularly one that yielded a gain of signal rather than signal attenuation, would be a significant advance in this field.

Assessment of the host environment and response to various therapies is conducted routinely in cardiology. Myocardial perfusion and infarct size are measured using US Food and Drug Administration (FDA)-approved gadolinium-chelate contrast agents and MRI, and this approach has been adapted for clinical trials of cell therapies. Clinical trials of autologous c-kit⁺ cardiac cells or cardiosphere-derived cells^{84, 85} transplanted into patients with ischemic heart failure⁸⁶ showed a reduction in the zone of delayed gadolinium enhancement and an increase in viable tissue volume, which were described as possible improvements in heart contractility and shrinkage of the infarct scar. However, the reduction in gadolinium accumulation in the interstitial space and contrast enhancement could also have been the consequence of reduced vessel permeability leading to reduced extravasation or of enhanced lymphatic drainage of injured tissue. Although the increased mass of viable tissue is consistent with regeneration, it is also consistent with hypertrophy of pre-existing cardiomyocytes.

Quantitative imaging, such as diffusion-tensor imaging (DTI), can be used to assess the fiber architecture in remodeled myocardium in animals and humans^{77, 87, 88}. Myocardial fibers have a complicated helical structure that is critical for efficient contractile and conductive function of the heart. DTI tractography has shown a smooth transition in fiber orientation from epicardium to endocardium in the healthy heart and severe disruption of myofiber architecture after infarction in different species (human, sheep, rat)⁸⁸. Preliminary DTI results in a murine infarct model⁷⁷ suggest that this technique has potential for assessing structural integration and alignment of transplanted cells with host myocardium in humans.

Finally, some aspects of functional integration are accessible by imaging. Global heart function, regional contractility and wall motion were imaged by high-resolution MRI after transplantation of human ESC-derived cardiomyocytes to the rat heart⁸⁹ (work of A.V.N., C.E.M. and colleagues). It is important to determine whether the graft is electrically connected with the host tissue and whether transplanted cells create areas of arrhythmia. Recently, a genetically encoded Ca²⁺-indicator fused with green fluorescent protein was used to demonstrate that human ESC–derived cardiomyocyte grafts are electrically coupled with host myocardium of the guinea pig⁹⁰ and of a non-human primate⁹¹ (work of C.E.M. and colleagues), indicating that this cell type meets physiological criteria for true heart regeneration. However, new imaging techniques for evaluating electro-mechanical coupling of grafted cells in patients’ hearts are needed.

In summary, imaging has played an important role in advancing cell-based cardiac repair, providing information on transplanted cell retention, viability, proliferation, graft size, integration with the host tissue and restoration of myofiber architecture, in addition to providing evidence of enhanced cardiac structure and function. What is needed are new imaging probes capable of detecting dynamic changes in grafted cells (i.e., that label live transplanted cells and quickly disappear after cell death) and techniques for genetically labeling cells for long-term tracking in humans.

Imaging pancreatic islets

Diabetes mellitus, a metabolic disease caused by insufficient insulin signaling, is traditionally divided into types I and II, with type I resulting from autoimmune destruction of pancreatic beta cells⁹² and type II resulting from resistance to insulin⁹³. In its later stages, type II diabetes is thought to result in beta-cell death, adding insulin deficiency to the problem of insulin resistance. Transplantation of pancreatic islets or beta cells emerged as a promising treatment for diabetes with the development of the Edmonton Protocol, which allowed patients to attain insulin independence through transplantation of cadaveric islets into the liver⁹⁴. However, insulin independence proved to be transient in most recipients⁹⁵. The decline is attributed primarily to drastic graft loss—up to 60% during the early post-transplant period, as shown by histology in animal models⁹⁶, and nearly 100% in patients with long duration of type 1 diabetes (determined at autopsy)⁹⁷. A noninvasive approach to assess cell viability and function over time is urgently needed to investigate the immunological and nonimmunological factors underlying the loss of transplanted cells⁹⁸, whether the cells are cadaveric islets⁹⁴ or beta cells derived from human pluripotent stem cells^{99, 100, 101, 102}.

In the first clinical trial of labeled islets, no correlation was seen between the number of injected islets and the SPIO signal on MRI images¹⁰³. However, a subsequent trial using ferucarbotran-labeled islets demonstrated a 60% decrease in graft volume one week after transplantation in eight patients by MRI¹⁰⁴ (Fig. 3). Although no independent method was used to confirm this finding, significant C-peptide levels and near-normal HbA1c values were achieved in these patients with a substantial reduction of insulin dose, suggesting that the remaining grafts were functional¹⁰⁴. Two clinical trials using PET imaging of ¹⁸F-FDG–labeled islets showed early engraftment in the liver^{105, 106}. However, only 50–70% of the expected signal was observed, suggesting substantial islet loss during the procedure. Importantly, no side effects attributed to islet labeling by ¹⁸F-FDG were detected.

These clinical trial results, although promising, leave much to be desired. An ideal imaging modality would detect retention in the liver by a quantitative correlation between the number of islets present and the number of hyper/hypointense voxels or ‘hot spots’. This has been demonstrated in small animals¹⁰⁷ (work of A.M. and colleagues) but not yet in humans¹⁰³. The detection limit of intrahepatically transplanted islets using clinical magnetic resonance scanners is not known, although studies in rodents have detected ~75 SPIO-labeled islets scattered throughout the liver using a 4.7 T scanner^{107, 108} (work of A.M. and colleagues) and 200 islets under the kidney capsule using a 1.5 T clinical magnet¹⁰⁹. Long-term survival of islets in the human liver is also difficult to assess. Although the greatest clinical need is for monitoring during the early engraftment period (2–5 wks) when graft loss is the highest, long-term monitoring would provide valuable information about recurrent autoimmunity and rejection and/or the options in modulating the regimen of immunosuppression in response to such recurring events. This is not possible with ¹⁸F-FDG–labeled islets owing to the short half-life of ¹⁸F and rapid washout. Finally, islets transplanted in the liver suffer from hypoxia, which could result in impaired insulin secretion. There are no established clinical methods to evaluate islet graft function noninvasively.

The limitations of current islet imaging approaches are due in part to the unique, complex structure of pancreatic islets, which consist of four cell types (beta cells, alpha cells, delta cells and pancreatic polypeptide-producing cells)¹¹⁰. Labeling of this rather large cluster of cells is nonspecific and occurs only by diffusion. In the case of SPIO nanoparticles, 12–48 h are required to achieve a concentration of 2–12 pg/cell^{103, 108, 111} (work of A.M. and colleagues). In vivo, quantification of SPIO-labeled grafts is complicated by the presence of susceptibility artifacts, the inability to distinguish single islets from islet clusters and low signal-to-noise ratios between the grafts and liver parenchyma.

Several emerging strategies, which have been demonstrated in animals, may help to address the remaining challenges in clinical islet imaging. The issue of islet quantification is being explored using ‘off-resonance’ techniques, such as automatic quantitative ultrashort echo time imaging, which provide positive contrast. A study in rats showed high positive contrast from SPIO-labeled islets on three-dimensional dual echo ultrashort echo time images and suppression of the signal from liver and small vessels, allowing for automatic quantification. Quantification of hyperintense pixels correlated with the number of injected IEQs due to the uniform suppressed background and high contrast and appears to be superior to standard imaging and manual counting methods¹¹².

Gadolinium chelates used in MRI have potential disadvantages, including risk of nephrogenic systemic fibrosis in patients with renal insufficiency¹¹³ and obscuring of the imaging signal in patients with hemosiderosis¹¹⁴. Instead, fluorine-19 MRI of perfluorocarbon-labeled islets appears to be a more promising approach because of lack of toxicity and the absence of background fluorine-19 signal in tissues^{115, 116, 117}. Human islets have been incorporated into microcapsules that contain perfluorocarbon, and the microcapsules were transplanted in mice and visualized by¹⁹F MRI and CT^{118, 119}. Perfluorocarbon labeling of the microcapsules did not interfere with long-term insulin secretion in vivo.

Long-term tracking of human SPIO-labeled islet grafts for up to 188 days has been shown in diabetic immune-compromised mice^{108, 111}, non-human primates¹²⁰ (work of A.M. and colleagues) and swine¹⁰⁷. The feasibility of this approach for long-term tracking of transplanted islets in humans has already been demonstrated with data going out to 24 weeks¹⁰⁴, although quantitative correlation between the number of infused and remaining islets was lacking and should be addressed in future trials.

The possibility of evaluating graft function by MRI has also been suggested in preclinical studies by quantifying zinc co-released with insulin in response to glucose challenge from endogenous islets using the Zn⁽²⁺⁾-responsive T1 agent, GdDOTA-diBPEN¹²¹. This approach offers the exciting potential for deep-tissue monitoring of beta cell function after islet transplantation in patients.

Finally, combination of several imaging modalities could provide considerably more information regarding graft fate than a single approach. To that end, a trimodal microcapsule for islet labeling containing gold nanoparticles functionalized with DTDTPA (dithiolated diethylenetriaminepentaacetic acid) and gadolinium chelates was developed for co-encapsulation with islets using protamine sulfate as a clinical-grade alginate cross-linker. Microencapsulation decreased islet rejection due to the presence of a semi-permeable membrane that prevents the passage of antibodies but allows the passage of cell-derived factors, including insulin. Intra-abdominally engrafted islets were detectable by multimodal imaging (i.e., MRI, CT and ultrasound)^{114, 122}. Inclusion of perfluorocarbon emulsions in alginate constructs for islet encapsulation^{118, 119} also allows for their detection by ¹⁹F MRI, ultrasound and CT (in the case of perfluorooctyl-bromide) and does not alter the permeability of the capsules or affect islet function. These new approaches not only improved islet protection but also allowed for MRI of transplanted islets with positive contrast and for islet infusions under fluoroscopic and ultrasound guidance, which are common techniques in clinical practice and could ease clinical translation. However, wide clinical applicability of these methods would require the use of approved materials for islet labeling and further optimization.

In summary, the ability to image transplanted pancreatic islets or islet cells provides a unique opportunity for clinicians to monitor their distribution, survival and function in the short and long term. However, algorithms for precise quantification of infused and engrafted islets are still under development, as are probes that are quantitative, nontoxic and long retained by islet cells. Furthermore, the described methods label islet cells and are not specific for beta cells.

Imaging brain repair

Acute brain injury from stroke and chronic neurodegenerative disease are prevalent in our aging societies and increasingly burden healthcare systems. The human brain has little or no regenerative capacity. Although endogenous neurogenesis has been described in the injured human brain¹²³, it is insufficient to promote significant repair. The possibility of cell therapy in the brain arose with the development of methods for MRI-guided stereotactic delivery directly into the parenchyma^{124, 125, 126}.

The most striking results in clinical cell therapy for the brain came in the mid-1990s in trials for Parkinson’s disease¹²⁷. The disease is caused by degeneration of dopaminergic neurons, which have their cell bodies in the substantia nigra and project axons to the caudate and putamen. Implanting fetal neural tissue grafts ‘ectopically’ in the caudate and putamen improved outcomes in patients, whereas implants in the substantia nigra had no benefit^{128, 129}. In cell therapy for stroke, minor improvements in patient outcomes were achieved by brain transplantation of neuroteratocarcinoma-derived neurons^{130, 131}. This approach aimed both to replace lost neurons and to provide trophic factors that promote angiogenesis and plasticity in existing neural networks. Re-myelination was achieved in patients with Pelizaeus-Merzbacher disease through implantation of neural stem cells into the brain¹³². In patients with multisystem atrophy, infusion of unlabeled autologous MSCs into the carotid artery led to increased uptake of ¹⁸F-FDG in various areas of the brain compared to the control group and improvement in functional neurological outcome over a period of 1 year¹³³. Of concern, however, diffusion-weighted MRI detected areas of microinfarction, presumably due to MSC clumping in vasculature. Longitudinal imaging studies are an essential part of cell therapy clinical trials to monitor damage and toxicity.

Efficacious cell therapy requires delivery to the appropriate region of the brain. Thus, imaging of the retention and biodistribution of transplanted cells in the human and animal brain is important for interpreting therapeutic outcomes. In one early-phase clinical study, SPIO-labeled, autologous brain-derived cells were transplanted into a patient with an open brain trauma⁴⁰. MRI at 3T could delineate the damaged parenchyma, the implantation site of the labeled cells, and movement of the label through the parenchyma for 21 days (Fig. 3); the hypointense voxels disappeared at seven weeks after implantation⁴⁰. Imaging has also revealed that intracerebral retention of transplanted cells requires delivery into the parachyma rather than into the circulation¹³⁴. In stroke patients who received ^99mTc-labeled bone marrow mononuclear cells, only a very small proportion of intravenously or intra-arterially injected cells invaded the central nervous system, and most cells were cleared within 24 h¹³⁵. Nevertheless, intravenous cell injection is the most common route of delivery for acute brain injury, with some evidence of therapeutic efficacy^{134, 136}. The lack of intracerebral penetration and retention of MSCs after intravenous injection indicates that any observed efficacy is not mediated by cell replacement.

Imaging of cell proliferation and differentiation has not been demonstrated in the human brain. A major obstacle has been the development of suitable probes that can cross the blood-brain barrier. Potential indicators of integration, such as brain plasticity and cell differentiation, can be assessed by functional MRI¹³⁷ or MR spectroscopy^{138, 139}, respectively. Re-myelination after neural stem cell implantation in patients with Pelizaeus-Merzbacher disease was evident in the restoration of the tissue microenvironment, as indicated by an increase in fractional anisotropy and reduced radial diffusivity on diffusion MRI¹³². The restoration of neurochemical tone, as in the case of Parkinson’s disease, can be determined reliably and quantitatively by PET using the dopamine D2 receptor-binding agent, raclopride^{76, 140}. Reduction of raclopride binding correlates to the presence of endogenous dopamine, and this approach has revealed the function of transplants even 15 years after transplantation¹⁴¹. Indeed, PET imaging of neurotransmitter activity also revealed that graft-induced dyskinesia is mediated by serotonergic neurons contained within the graft^{128, 142}. Paracrine and juxtacrine pathways, such as those discussed for the heart, may also play a role in the clinical benefits. PET ligands targeting receptors of paracrine factors, such as the TrkB receptor, are being developed for brain imaging in rodents¹⁴³, and should eventually allow investigation of paracrine pathways in patients.

As with cell therapies for many organs, imaging of the brain is essential for evaluating the tissue environment before and after treatment. MRI and PET scans can reveal regional damage, structural and functional connectivity, and metabolic and vascular changes^{131, 144, 145}. Incorporation of non-invasive imaging into clinical trials will further benefit the safety assessment (e.g., small bleeds, edema) and quality control (e.g., location of cells) of cell delivery. The host inflammatory response in patients can be monitored using ¹¹C-PK11195 PET, which binds to the peripheral benzodiazepine receptor on activated microglia¹⁴⁶, as well as ultrasmall SPIO uptake into peripheral macrophages detectable by MRI¹⁴⁷. In rodents, an endogenous neural stem cell response to damage and cell therapy can potentially be monitored using the proliferation marker 3′-deoxy-3′-¹⁸F-fluoro-L-thymidine (¹⁸F-FLT)^{148, 149}, whereas regional changes in metabolic activity indicating an increase in cellular density can be assessed in the human brain using FDG-PET¹⁵⁰.

Although systemic cell administration or intracerebral implantation of cell suspensions can potentially improve behavior, repair of large cavities in damaged tissue is still problematic¹⁵¹. In these cases a scaffolding support is required¹⁵²; various biomaterials can serve this purpose, and future studies are needed to determine the clinical efficacy of this approach¹⁵³. The most appropriate cell types may be fetal neural stem cells, which are positionally specified to develop into a particular type of tissue, or pluripotent stem cell–derived neural stem cells, which require positional specification to differentiate into cells of the appropriate brain region. However, the skull restricts delivery of large constructs to the brain. To avoid tissue damage, presurgical planning for each subject based on imaging is essential to determine both the volume and location of injection¹⁵⁴. In one study in rats, imaging enabled visualization of scaffolding material in the brain, although the survival and distribution of implanted cells was not determined¹⁵⁵. Monitoring of scaffold biodegradation and formation of new tissue are other unsolved problems. More sophisticated imaging approaches are needed to visualize the cell and biomaterial components independently¹⁵⁶.

In summary, readily available clinical imaging tools used to diagnose or monitor disease can in many cases be applied to measure the therapeutic efficacy of cell therapy in the brain. However, before such therapies could be widely adopted, it will be essential to devise new, robust imaging technologies. Current techniques, such as ⁹⁹Tc SPECT imaging, are adequate to determine short-term, labeled-cell infiltration in the human brain but not to assess precise anatomical distribution or long-term survival (with the exception of dopaminergic neurons). A reasonable goal would be the capability to measure cell location at <1 mm accuracy, cell survival for >1 month, and the interaction of a few hundred transplanted cells with the microenvironment.

Future directions

The path to clinical translation of regenerative therapies is difficult. For researchers intrepid enough to consider jumping into this area, Box 3 offers suggestions on developing programs that include imaging. Incorporating imaging in a cell therapy requires preparation of additional data for Investigational New Drug (IND) submissions. Specific recommendations are presented in Box 4.

Box 3. A framework for implementing imaging in regenerative medicine.

Some specific steps for researchers wishing to incorporate imaging into cell therapies on a path to clinical translation are provided below.

Teamwork

Establish a team that includes stem cell biologists, physiologists, chemists, specialty clinicians, radiologists and imaging scientists to determine the most appropriate imaging strategy.

Regulatory requirements

Learn the requirements by communicating early and often with overseeing regulatory agencies. Clearance from an institutional review board may be required. A consultant familiar with imaging in IND submissions may be helpful.

Standard operating procedures

These are required for production, labeling and handling of cells, data collection, in vivo imaging and analysis, histological verification and clinical outcome measures. Document all results and do not deviate from standard operating procedures once established.

In vivo imaging

Ensure that labeled cells are detectable in vivo with high sensitivity and specificity using clinical scanners. Understanding the number of cells/voxel that migrate to targeted regions is important in determining preclinical doses, schedules and administration routes. It is impossible and unnecessary to track single cells.

Scale up

The ability to move a labeling technique from the bench to a clinical Good Manufacturing Practice facility may be difficult, but it is essential for producing large numbers of cells for clinical trials. Phase 1 trials that include dose escalation should also be used to determine delivery approaches, safety, toxicity and maximum tolerated dose of the labeled cells.

Box 4. Recommendations for preclinical evaluation of imaging methods.

If imaging is to be used in a cell-therapy clinical trial, it is necessary to compile appropriate data for presentation during pre-IND discussions with the regulatory agency. These data should include the following.

In vitro tests of direct or indirect labeling methods should be compared to unlabeled cells and include determination of labeling efficiency; label concentration; rate of cell death; short and longer-term proliferation capacity; differentiation capacity; migration capacity; immunogenicity in a mixed lymphocyte reaction; and surface markers.

It is important to document that labeling does not change cell potency. For cells whose function is secretory, potency can be defined as hormone, neurotransmitter, cytokine/chemokine or growth factor release. For mechanically active cells, such as cardiomyocytes, potency includes electrical and mechanical activity. Gene expression profiling has not been required or routinely recommended to evaluate direct labeling methods because of potential variability between donor cells and uncertainty in how these data correlate with potency.

For indirect labels, it may be necessary to determine the chromosomal location of the label to avoid proximity to oncogenes. If a suicide or therapeutic gene is inserted with the reporter gene, expression of both genes must be documented, along with efficacy of the suicide gene. Long-term passaging of cells is needed to ensure stability of transgenes and lack of malignant potential.

An estimate of the dilution rate of the label over multiple cell divisions can be provided by pulse-chase experiments (i.e., cell labeling followed by expansion).

Sensitivity to the minimal number of cells that can be detected by clinical scanners can be estimated by in vitro serial dilutions of labeled cells if these studies are carried out with clinically relevant voxel sizes and acquisition times.

Preliminary in vivo imaging studies for detecting and tracking cell fate are useful to determine the biodistribution of the cells in normal and experimental models. It is necessary to determine whether preclinical studies use xenogeneic (i.e., human cells into mouse), allogeneic or autologous transplantation. Histological analysis of multiple organs should be performed to assess inflammatory responses and to identify labeled cells in targeted and nontargeted tissues.

Depending on the disease model, dose-escalation studies over three dose levels (typically spanning at least a tenfold range) may be required to determine safety and efficacy of transplanted or infused labeled cells. Proposed cell dosing, dosing schedule, experimental model systems and required validation/toxicity outcomes (i.e., histology, serum chemistries including renal and liver function, blood counts, functional measures) should be included as part of the discussion with a regulatory agency at a pre-IND conference.

There are many examples of experimental labels for tagging cells, but few have made it into clinical trials^{157, 158}, and fewer still are used routinely in the clinic. Researchers who synthesize novel agents do not usually pursue regulatory approval because of lack of funding¹⁵⁸ or insufficient projected markets. However, the demand for imaging technologies to evaluate the host and to track transplanted cells is likely to grow over the coming decades. For those who wish to undertake clinical translation of novel imaging agents, some relevant considerations are summarized in Box 5. Guidelines for bringing a novel imaging agent to the clinic specify rigorous preclinical safety testing and validation for regulatory approval as drugs. Because nanoparticles or MRI contrasts agents may require nanogram to microgram concentrations in cells, their first use in humans must be a phase 1 safety study. In comparison, radiolabeled agents are used in far lower concentrations (femtograms/cell) and are dosed based on radiosensitivity rather than chemical concentration. Introduction into humans can be performed through an exploratory IND submission in a phase 0 clinical trial to determine safety, pharmacokinetics and pharmacodynamics¹⁵⁸. The exploratory trial design is much less costly, allowing rapid screening of radioactively tagged agents developed under good laboratory practice standards.

Box 5. Considerations for the development of novel cell-tracking agents.

Before undertaking the development of a novel imaging agent for cell tracking, investigators should understand the time commitment and funding that are required for clinical translation. The following points should be considered.

• Assemble a multidisciplinary team.

• Confirm that the labeling method is amenable to scale-up of the cell product in a Good Manufacturing Practice facility.

• Ensure high sensitivity of detection by maximizing the concentration of label in cells (consistent with safety) and minimizing the numbers of cells/voxel for detection in vivo.

• Evaluate whether cellular dynamics (cell death, proliferation and interaction with the host), cellular kinetics (movement of cells in the body) and volume of distribution of the cells may affect efficacy.

• Strive to limit transfer of the labeling agent from therapeutic cells to macrophages or bystander cells. Develop methods to turn off the signature of the label after engulfment.

• Resist the temptation to switch to an improved agent or technique after starting preclinical experiments for an IND submission as such modifications will likely require that experiments be repeated, resulting in delays and increased costs.

• Note that regulatory agencies require separate preclinical evaluation of the labeling agent and technique for each cell type and disease model.

A problem common to all contrast agents is that the imaging signal does not distinguish live and dead cells. Reporter genes represent a possible solution, although genetic modification brings additional layers of safety and regulatory concerns. With a concerted effort, it should be possible to assess the risks of knocking reporter genes into safe-harbor loci in only a few years. We encourage researchers, regulatory authorities and funding agencies to work together to solve this tractable problem. An alternative solution would be the development of novel agents that label only live cells, disappearing quickly after cell death. In addition, most cell therapies would greatly benefit from novel ligands or probes with increased imaging specificity to detect dynamic changes in the heterogeneous, injured host environment. Understanding the role of the host environment will be crucial to improving the efficacy of cell therapies.

Many results in animal models are not predictive of the human response. For example, therapeutic results in rodents do not mean that the same number of cells/kg would be safe and efficacious in patients. Similarly, imaging results demonstrated in rodent models do not necessarily translate into the clinic, primarily because of limitations in clinical scanner technology—spatial resolution or voxel size, instrumentation (i.e., radiofrequency coils and magnetic field strength), patient motion and image acquisition times. Investigating teams must be aware that the translation of regenerative therapies from bench to bedside is research, so they are advised to be patient.