, Iron oxide crystals coated in carbohydrates such as dextran are extremely effective agents for direct labeling of cells. The most effective particles are small ones with core diameters in the 50–180 nm range, also known as superparamagnetic iron oxide nanoparticles (SPIOs). In fact, the first cell tracking on MRI was demonstrated using SPIOs three decades ago in the early 1990s (13, 14). In these studies, blood cells were labeled by co-incubating cells with SPIO in culture media [alternatives include transfection agents or electroporation (15)]. Labeled blood cells were then readily identified by virtue of a negative contrast induced by a very strong local disturbance in the local magnetic field. Because this field disturbance extends over a volume an order of magnitude larger than that occupied by SPIOs, the area of dark contrast seen on MRI grossly overestimates that truly occupied by cells. This so-called “blooming artifact” is a double-edged sword, however: high detection sensitivity for relatively low cell numbers [10³ cells (16)] is achievable and desirable, but obliteration of signal in surrounding tissue precludes precise validation of cell targeting, especially in small and narrow structures of the brain, spinal cord, and heart. Nonetheless, owing to high detection sensitivity and ease of labeling, a plethora of SPIO cell-tracking applications ensued in the years and decades following, resulting in SPIO-direct labeling as the most adopted approach of all available MRI cell tracking methods. Applications spanning the tracking of immune cells (17), cancers cells (18), neural stem cells (19), cardiac stem cells (20), and smooth muscle cells (21), to name a few, abound, even today. Figure 2 illustrates tracking mesenchymal stem cells in the rat brain.

In addition to ex vivo cell labeling, in situ cell labeling is also possible with the use of ultrasmall SPIOs (USPIOs) with diameters <50 nm [which, incidentally, are not as efficacious as SPIOs for direct labeling (22)]. An intravenous injection of the iron oxide is administered to the subject, and phagocytic cells (e.g., macrophages and monocytes) will preferentially take up the iron oxide particles. In this way, inflammation foci rich with infiltrating phagocytes can be identified. Stroke, myocardial inflammation, and atherosclerosis are just a handful of many inflammatory conditions where in situ cell labeling with iron oxides have shown value (23–25). However, because of its non-specificity amongst different phagocytes and limited utility beyond phagocytic cells, in situ cell labeling remains a niche approach that confers value strictly in MRI of inflammation.

Paramagnetic gadolinium metal ions (Gd³⁺) have also been investigated for labeling and tracking cells on MRI. Methods for labeling are analogous to those for SPIOs: co-incubation, electroporation, and transfection. Unlike SPIOs, Gd-labeled cells emit a positive, or bright, contrast due to enhanced longitudinal relaxation of water. The extent of contrast enhancement is restricted to the volume occupied by labeled cells; there is no “blooming artifact” as is seen with SPIOs. Therefore, positive contrast greatly improves the precision of cell targeting and eliminates any possibility of signal obliteration in surrounding critical tissue structures. However, detection sensitivity is also lower compared to negative contrast methods, with some studies reporting 10⁴ cells (26) at the minimum detection threshold – at least 10 times higher than for SPIOs. While cell tracking with Gd-labeling has been reported for many cell types – stem cells (27), endothelial and muscle cells (21, 28), neural progenitors (29), and cancer cells (30) (Figure 3) – it is a much less common technique compared to SPIOs simply due to its lower sensitivity of detection. Large macromolecular Gd-based agents with higher relaxation efficiencies have been proposed for improving sensitivity; examples include gadolinium rhodamine dextran (31), gadolinium oxide nanoparticles (32), and gadofullerenes (33). Yet, despite these advances, Gd-based cell labeling remains relatively scant. A plausible explanation may be the cytotoxicity of Gd³⁺. As a metal foreign to the human body, Gd³⁺ has been reported to lower cell proliferation and increase reactive oxidative species (34, 35); as these are acute effects, the long-term impact on cell function remains unknown.

Manganese (Mn²⁺) is another paramagnetic metal ion that generates positive contrast on MRI. Many cell labeling applications involving Mn²⁺ utilize the free ionic form in MnCl₂. Interestingly, the first application of manganese-enhanced MRI (MEMRI) was not for tracking cells but for distinguishing ischemic from healthy myocardium in dogs (36). The concept was straightforward: the divalent metal Mn²⁺ would enter viable cardiomyocytes through membrane calcium channels (37) but could not enter dead cells. Since that seminal paper, MEMRI has been used to visualize heart viability and, more commonly, neuronal connections in the brain and central nervous system (38–40).

In 2006, Aoki et al. (41) reported the first cell labeling study with MnCl₂ on human natural killer cells and cytotoxic T cells. At a labeling concentration of 0.5 mM MnCl₂, these cells maintained their in vitro killing capacity and demonstrated no cytotoxicity. A few applications in direct labeling of human embryonic stem cells (42) and prostate cancer cells (43) followed in the years since. In situ labeling of breast cancer cells with MnCl₂ has also been demonstrated, showing greater metal uptake in more aggressive cancer cell lines (44, 45). However, compared to labeling with SPIO and Gd³⁺ compounds, applications with MnCl₂ are far fewer. This may be simply attributed to less fervent effort in developing Mn-containing compounds for cell labeling. Interestingly, in situ labeling of tumors by manganese porphyrins was discovered even earlier, in the early 1990’s. Brain tumors in animals were shown to enhance significantly against background tissue with the administration of Mn(III)TPPS (46, 47). More recently, there has been a resurgence of interest in using porphyrin as a chelator for more stable metal binding to reduce potential toxicity (48–52), as well as other formulations involving Mn²⁺ [e.g., manganese oxide nanoparticles (53, 54)]. Figure 4 illustrates the application of Mn²⁺ direct labeling of human breast cancer cells and Mn³⁺ direct labeling of human embryonic stem cells.

The direct cell labeling methods described hereto all involve exploiting the inherent T2 (SPIO) or T1 (Gd³⁺ and Mn²⁺/Mn³⁺) contrast specific to the contrast agent. Unfortunately, everything in the body has a characteristic T2 and T1 contrast, which can and does appear isointense to labeled cells. This ambiguity as to the source of contrast is especially complex with T2-weighted cell tracking with SPIOs, because many endogenous sources of dark contrast exist (e.g., microbleed and hemorrhage, iron accumulation, air-tissue interface, air pockets) that are indistinguishable from labeled cells. In contrast, endogenous sources of bright T1 signal are far fewer (except for fat, which can be attenuated with fat suppression techniques).

Fluorine-based cell tracking eliminates this ambiguity entirely, as endogenous sources of mobile fluorine is well below the detection limit [< 10⁻³ μmol/g wet tissue weight (55)]. For this reason, it is preferable to metalated contrast agents for cell labeling. In 2005, Ahrens et al. (56) reported the first ¹⁹F-cell tracking, demonstrated for immunotherapy cells. Areas of positive contrast were readily identified as labeled cells only, and signal intensity scaled quantitatively with cell number. However, imaging with fluorine-based agents requires two separate images: ¹⁹F images acquired using a dedicated coil and ¹H images acquired with a water proton coil – the former localizes cells, while the latter maps out anatomy. Because ¹⁹F image must be overlaid on anatomical ¹H MRI image for cell localization, care must be taken to ensure image co-registration. This requirement for image overlay is a potential pitfall, as any mis-registration can potentially localize labeled cells in a different and wrong anatomy. Nonetheless, since the first application in 2005 ¹⁹F-cell tracking has been applied in monitoring inflammation (57–59), neural (60) and hematopoietic (61) stem cells, T cells in diabetes (62), and cancer cells (63). Perfluorocarbons such as perfluoropolyether and perfluoro-15-crown-5-ether are the most common ¹⁹F contrast agents for cell labeling (64). Figure 5 illustrates ¹⁹F tracking of phagocytic tumor-associated macrophages in a mouse model of murine breast cancer.

Despite the high specificity and quantitative nature of ¹⁹F cell tracking, sensitivity is inherently low, much lower than SPIO, Gd, or Mn-based cell tracking approaches (65). The minimum detectable cell number varies from study to study, depending on the scanner field strength, pulse sequence, and cell type. On average, the minimum detectable cell number is over 10⁶ cells per voxel (66), and this threshold barely changes even when high-SNR sequences (e.g., 3D bSSFP) are used to reduce the voxel size by 100 times, with a concomitantly longer acquisition (up to 1 h imaging) (65). From a practical standpoint, the sensitivity limit severely restricts the appeal of ¹⁹F cell tracking in applications where cell numbers are modest.

Chemical exchange saturation transfer (CEST) is an advanced MRI method for detecting low concentrations of compounds. It achieves this by taking the compound of interest, saturating its ¹H protons, allowing this saturation to be transferred from the compound to water, and repeating this exchange at least 100 times (to amplify water signal changes by 100 times). In the context of cell tracking, cells can be directly labeled with a CEST agent (most commonly a PARACEST agent), injected in vivo, and tracked over days. In this regard, it is very similar to the other direct cell labeling methods available for MRI. However, spatial resolution is poor, with voxels exceeding 2 mm in plane (67), and SNR is much lower than that for T1 or T2-weighted imaging (68), thus requiring long imaging times to boost signal above the noise floor. These limitations associated with low spatial resolution and SNR may explain the relatively sparse literature on CEST-based cell tracking. Yet, a few notable exceptions do exist: tracking of murine breast cancer (69) and myoblasts (67). Figure 6 illustrates myoblast tracking in mice.

Direct labeling of cells with an exogenous contrast agent is the most widely used cell labeling method. As seen above, a wide variety of T2, T1, and CEST-based contrast agents have shown utility for labeling and tracking cells in vivo. Despite ease of use and simplicity, direct labeling has several critical limitations. As alluded to previously, SPIO-induced dark contrast can be easily mistaken for endogenous dark-contrast sources, such as bleeds and air pockets, complicating the specificity of cell identification. A second limitation is the inability to distinguish living labeled cells from dead labeled cells or phagocytic immune cells that have taken up contrast agent released by dead cells (70). This limitation is seen with all MRI contrast agents, but it is especially problematic for larger particles to which macrophages have high affinity, such as SPIO and ¹⁹F-based agents. A third limitation is the challenge of quantifying absolute cell numbers in T2-and T1-based ¹H imaging. For SPIO-labeled cells, there exists a linear relationship between iron concentration and cell number over a very narrow concentration range at low iron concentrations. T1-based agents, such as Gd³⁺ chelates and Mn²⁺ nanoparticles, provide a much larger linear range and at higher concentrations due to the absence of signal dropout. The fourth limitation is potential cellular toxicity from the contrast label, which is not only agent-specific but cell type-specific. While it is impossible to enumerate the safe dosing levels for all labeling agents across all cell types at different stages of maturity, literature has reported reduced endothelial cell proliferation at 0.1 mM Gd-oxide over 24 h (21), 50% neuronal cell death at 0.05 mM MnCl₂ over 120 h (71), and reduced chondrocyte expression (72) and neural stem cell motility (73) at 25 ug/mL SPIO. It is also important to remember that even if toxicity is drastically attenuated via different chemical formulations, there has been no systematic study on their long-term stability and toxicity risks, which is an especially important consideration for Gd³⁺-based labels. In contrast, a lethal dose of perfluorocarbon has not been reported, with unaltered viability confirmed even at 20 mg/mL perfluorocarbon (74, 75). The fifth and perhaps most critical limitation is the lack of long-term cell tracking capability. As implanted cells proliferate and migrate in vivo, the total amount of contrast label on a per cell basis gets diluted over time. Kustermann demonstrated in murine embryonic stem cell that iron oxide-labeled cells underwent 4 replication cycles before signal diluted appreciably (76). This dilution phenomenon is observed with all direct cell labeling methods and, depending on the cell type, restricts in vivo cell tracking to days (18) or weeks (77) post-cell transplantation.

Ferritin is a universal iron-storage protein found in all mammalian cells, helping to regulate iron release in a controlled manner. Its application as an MRI reporter gene was first described in 2005 by the laboratory of Ahrens (8). An epithelial lung carcinoma cell line transfected with the ferritin transgene showed over 60x the background ferritin levels and exhibited over twice the T2 relaxation rate of wildtype control cells due to higher accumulation of supplemental iron. In vivo, transduced neurons and glial cells in the mouse brain displayed hypointensity on T2*-weighted MRI out to 39 days without the need for exogenous iron supplementation. Since this seminal study, no fewer than 70 reports have emerged utilizing ferritin for cell tracking. Example applications include: tracking stem cell delivery to the mouse heart (81), monitoring melanoma cells in lymph nodes in mice (82), and detecting neuronal differentiation in stem cells (83). Implementation of the ferritin method varies from study to study, with some opting for stable integration of the ferritin transgene (9, 84), some supplementing with exogenous iron to increase signal (9, 85), and some relying on endogenous iron store for contrast (81, 84).

When ferritin overexpression is achieved via stable integration and is modest, around 2–5 times baseline, cell function and integrity is preserved, but at the cost of low sensitivity. According to some studies, the sensitivity is so low that even with exogenous iron supplementation, contrast change is negligible (9, 84). Figure 7 illustrates the much lower sensitivity of this technique compared to direct labeling with SPIOs. Furthermore, the necessity of resorting to T2*-weighted MRI to achieve sensitive cell detection is a drawback, as image distortion and low SNR are intrinsic to T2*-weighted imaging. Therefore, despite abundant evidence that ferritin overexpression is safe and non-cytotoxic, ferritin-cell tracking remains an inferior choice to direct labeling with SPIOs due to its very low sensitivity.

In 2020, Szulc et al. reported a highly sensitive, T1-weighted MRI approach to overcome the low sensitivity of conventional “dark” ferritin-based cell tracking (9). In that study, human embryonic kidney (HEK) cells were stably transfected with the ferritin transgene, and a modest two-fold protein overexpression was attained. No impact on cell viability, proliferation, and metabolism was observed at this low overexpression level. Both in vitro and in vivo mouse experiments confirmed that ferritin-overexpressing cells, when exposed to supplemental MnCl₂, exhibited a bright contrast and large T1 reduction on MRI, sustained for 5 days. Chemistry analysis and microscopy revealed the formation of manganese-ferritin nanoparticles inside cells and their eventual degradation by normal cellular pathways. This study not only uncovered Mn²⁺ as an alternative metal to iron that could be sequestered in ferritin protein, but also boasted much higher sensitivity, SNR, and image quality from utilizing T1-weighted MRI. Ongoing investigation of the “bright-ferritin” platform shows promising results in the tracking of human embryonic stem cells and differentiated cardiomyocytes for cardiac therapy (86). Figure 8 compares the contrast and longevity of signal amongst three methods: bright-ferritin, conventional dark ferritin with and without iron supplementation, and DMT-1, to be discussed next.

The divalent metal transporter-1 (DMT-1) is a plasma membrane protein that transports ferrous iron and some, not all, divalent metals ions across the plasma membrane (87). Metals that are transported by DMT-1 include manganese, cobalt, copper, and zinc; however, calcium, also a divalent metal, is not (88, 89). While most studies of DMT-1 revolve around understanding iron transport and metabolism, Bartelle et al. were the first to report its use for indirect cell tracking (10). A number of different cell lines, including HEK cells and murine glioma cells, were stably transfected to overexpress DMT-1 by up to 3-fold. This overexpression of DMT-1 was accompanied by an over 3-fold increase in T1 relaxation rate in vitro when 0.3 mM MnCl₂ was supplemented in culture media for 1 h. As the authors noted, the observed change in T1 relaxivity was greater than the 1.8-fold increase in T2 relaxation rate from labeling ferritin-overexpressing HEK cells at 1 mM Fe for 24 h (90). Consistent with the higher efficiency of DMT-1 relative to ferritin coupled with iron, in vivo results in glioma cells implanted in mice confirmed a bright contrast was sustained relative to control cells beyond 24 h.

A handful of DMT-1 cell tracking papers have since been published after the 2013 discovery, focused on tracking experimental HEK or human stem cells in vivo (9, 91). Both studies demonstrated a clear demarcation of bright contrast where the cells resided. In the HEK cell tracking study in mice, it was further demonstrated that bright signal could be recalled on demand with a simple administration of MnCl₂ (9). This demonstration is one of the rare literature evidence where MRI signal from remaining viable reporter cells can be recalled, as needed, in a longitudinal fashion – what is dubbed true “long-term” cell tracking. Interestingly, the authors also compared the bright contrast obtained from DMT-1 against that from bright ferritin: they discovered that the latter provided a T1 relaxivity of 17.7 mM⁻¹ s⁻¹, where DMT-1 provided only half the relaxivity at 7.1 mM⁻¹ s⁻¹.

The MRI reporter genes described thus far all involve metallic substances that are either paramagnetic or superparamagnetic. This is limiting in that only one cell type can be tracked, precluding the possibility of monitoring multiple labeled cell populations as found in fluorescent methods. In 2007, Gilad et al. described the concept of transfecting cells with a lysine-rich protein (LRP) encoding vector (11). Radiofrequency irradiation at the amide proton frequency results in exchange with water protons (i.e., CEST effect). LRP overexpression showed no toxicity on viability or metabolic rate. In vivo mouse MRI at 11.7 Tesla demonstrated an 8% increase in CEST signal in LRP-overexpressing xenografts, whereas control xenografts exhibited a 3% increase in CEST signal. This approach is attractive in that additional frequency-selective reporters may be designed to label different cell populations. Nonetheless, as with all CEST-based method, low SNR remains a substantial barrier to widespread adoption.

A more sensitive, metal-free alternative is changing cellular water permeability. Mukherjee et al. introduced in 2016 a class of MRI reporters based on the human water channel, aquaporin 1 (12), to increase water movement into and out of a cell. No impact on cell viability was noted in the array of cells stably transfected to overexpress aquaporin 1. Water diffusivity increased (i.e., diffusion-weighted signal decreased) across the cell membrane by at least 2-fold for all cell types tested, with no change in T1 or T2 relaxation times, which implies an orthogonal channel for MRI. However, the use of diffusion-weighted MRI also introduces confounding influences from other sources of negative diffusion-weighted contrast. A second reporter gene that also modulates transmembrane water transport is the urea transporter (UT-B). Similar to aquaporin 1, UT-B expression can be increased to effect a proportional elevation in the apparent water exchange rate (92).

All MRI reporter gene cell tracking methods suffer from lower sensitivity of detection compared to direct cell labeling methods. In some instances, the contrast change may not even be detectable, as with iron-supplemented ferritin overexpressing cells. However, it is important to recognize that research in advancing MRI reporter genes has seen only a fraction of the effort applied to direct cell labeling methods. Therefore, many of the current limitations, as discussed in the following, may be surmounted in the not-too-distant future.

Low sensitivity of detection often results from inadequate in vivo exposure to a MR-active agent. For example, when ferritin-overexpressing cells are supplemented with iron, or when ferritin-overexpressing or DMT-1-overexpressing cells are supplemented with Mn²⁺, the in vivo bioavailability of either iron or Mn²⁺ is naturally much lower than it would be in an in vitro direct labeling setting. If one could increase the metal dosing systemically, sensitivity would automatically go up. Yet, this is an impractical solution, as achieving the required dose at the target site can easily lead to toxic overdosing in other organs. A potential solution is to administer the metal supplement locally to the site of interest, but this has limited value once cells have migrated away from the initial injection site.

Potential cytotoxicity or adverse impact on cell function is another concern with gene editing. All the MRI reporters in the literature involve changing metal homeostasis or water diffusion into and out of a cell. At low overexpression of reporter genes, most studies have confirmed that cells are unaffected. At high levels of overexpression, the sensitivity gained is offset by potential toxic effects. The balance between maintaining normal cell function and achieving higher sensitivity of detection is delicate, and it must be determined individually on different cell types and at different stages of maturity.

Finally, there is the risk of gene silencing, a rarely considered topic in MRI cell tracking. Gene silencing refers to the reduced expression of a gene. In the context of MRI reporter genes, a transgene is inserted into the genome with the hope that all the progeny can be tracked on MRI. However, that may not be the case. For instance, if one were to create a stable ferritin-overexpressing stem cell line, they may find that the ferritin transgene is silenced as the cells undergo differentiation and multiple passages. In this way, the “long-term” cell tracking capability that sets MRI reporter gene methods apart, is eventually lost. A recent report described how using CRISPR/Cas9 technology to insert their transgene into the AAVS1 safe harbor locus of human induced pluripotent stem cells did not sustain stable expression (93). In fact, gene silencing occurred during cardiomyocyte differentiation, leading to a decrease in expression from 98.9 to 1.3%. Checking for potential gene silencing is a must for any researcher working with MRI reporter genes.

One of the critical gaps in cancer research is an incomplete understanding of cancer metastasis and how different cancers respond to treatment. The question as to why cancers return in some patients but not in others has led to the concept of cancer stem cells, namely, that sub-population of cells capable of self-renewal and tumorigenicity. While the cancer stem cell population has not been probed specifically in MRI studies, many papers over the past two decades have described tracking the metastasis of cancer cells in experimental animal models to better understand what organs are susceptible (94–96). However, the majority of these studies employ SPIOs, which has limited durability and suffer from signal dilution, especially in rapidly proliferating cancer cells (97). An alternative approach, which has been rarely reported, is in situ labeling of metastatic cancers with MnTPPS, which showed bright contrast in both the primary rat brain tumor and the solitary metastasis (98). Another viable alternative to studying cancer progression and treatment response is to utilize MRI reporter genes that can provide a much longer lasting contrast necessary for studying the migration of metastatic cancer cells. This is an unexplored area of research, and care must be exercised in selecting an MRI reporter gene that does not alter the tumorigenic properties of the specific cancer cell type. Characterization of the relevant baseline protein expression levels is also necessary in the cancer cell of interest.

Unlike cancer cells, stem cells cannot thrive in inhospitable tissue environments, undergoing massive cell death in the hours and days after transplantation in vivo. A conservative estimate amongst stem cell researchers is a minimum 90% death rate, but over 99% cell death has also been reported in 2000 (99). However, with the availability of immune-compromised animals, drugs for immunosuppression, and new cell delivery vehicles, in vivo cell survival can be improved by an order of magnitude (100) – even 20% cell survival has been reported in the mouse spinal cord 4 weeks after stem cell transplantation (101). Amongst the surviving cell population, cell number will gradually scale up over the course of weeks and even months, but not necessarily to an appreciable number. Furthermore, some cell types, such as cardiomyocytes, do not replicate, and other cell types, such as neural progenitors, will migrate over a larger tissue volume. These nuances complicate cell tracking on MRI, because a low cell number and cell density can be expected, possibly falling below the limit of detection.

One strategy to ensure high detection sensitivity when massive cell death occurs initially is to use a direct cell labeling method. Small molecule-based labeling is recommended for this purpose, and large molecular structures (e.g., SPIOs and ¹⁹F nano-emulsions) should be avoided to avert the transfer of label from dying injected cells to phagocytic cells. This strategy should, at minimum, provide a more truthful depiction of cell survival and distribution in the early days after cell transplantation. Nonetheless, with all exogenous tracers, whether they are small molecules or large nanostructures, the possibility still exists that tracer outside of the intended cell target is being detected.

In the longer term, MRI reporter gene methods are the only option for monitoring proliferation from the remaining viable cell population and tracking their migration in the body. However, given that reporter gene methods generally suffer from low sensitivity, signal may or may not return, depending on the abundance and distribution of proliferating cells. Another key consideration is the anatomy into which cells are injected. If the anatomy consists of small, narrow structures (e.g., spinal cord, thin myocardial walls, renal cortical layers), then a dark-contrast method may obliterate the signal of surrounding critical anatomy. In this case, a positive-contrast reporter gene, such as DMT-1 or bright-ferritin, is extremely valuable. On the other hand, if therapeutic cells are delivered to a more homogenous tissue where less precise targeting can be tolerated (e.g., brain, liver), dark-contrast reporter genes, such as those modulating water permeability or iron sequestration in ferritin, may be exploited.

Broadly speaking, immune cells are readily tracked via in situ cell labeling with SPIOs or ¹⁹F nano-emulsions. These labeling methods are particularly useful, because they label cells circulating inside the body, but the downside is lack of distinction amongst different types of immune cells. With novel cancer immunotherapy, or more specifically T-cell transfer therapy, immune cells are taken from the patient, modified in the lab to attack the cancer more effectively, and re-administered to the patient. To verify tumor homing, long-term viability, retention, expansion, and absence of off-target effects, modified T-cells must be easily differentiated from other immune cells. This requirement stipulates the replacement of non-specific in situ labeling with either a direct or indirect labeling method specific to T-cells. Direct labeling with SPIOs is the most straight-forward approach, but simple incubation may be ineffective, as T-cells are non-phagocytic and do not take up extracellular particles readily. For this reason, a transfection agent can be employed to increase uptake. Furthermore, as T-cells are expected to distribute sparsely after injection, a high label content must be achieved on a per cell basis. Therefore, strategies to achieve high labeling efficiency without compromising T-cell function are crucial. Early reports in this domain described SPIO-direct labeling of cytotoxic T lymphocytes in a mouse glioblastoma model (102) and adoptive cell therapy in a dog prostate cancer model (103) (Figure 9). A more ideal, albeit less sensitive, alternative is MRI reporter gene for T-cell tracking, but no literature report exists. Finally, a quantitative cell tracking method is highly desirable in this application, as optimization of T-cell therapy requires not only correct targeting but also delivering the correct number of cells. In this sense ¹⁹F or T1-weighted imaging methods may be more suitable than T2/T2*-based tracking methods.