In 2013, the journal Science named cancer immunotherapy as the “breakthrough of the year” based on targeted approaches using chimeric antigen receptor (CAR) T-cells. CD19-specific CAR T-cell therapy has revolutionized the treatment landscape for patients with relapsed B-cell acute lymphocytic leukemia (1,2). Similar successes have not been seen in patients with solid tumors, in part because of inconsistent expression of specific tumor antigens and physical impediments to T-cell trafficking, for example, for passing the blood–brain barrier or reaching metastatic disease in bone marrow. Other factors interfering with prolonged response to CAR T-cell therapy include loss of target antigen and T-cell exhaustion. These resistance mechanisms may be alleviated by CARs targeting 2 or more tumor antigens (NCT03019055), more sophisticated T-cell engineering techniques, or combination therapies of CARs with various immune checkpoint inhibitors. So far, it has remained unclear why certain cell therapies succeed and even provide durable clinical responses (1,2) whereas others fail (3).
Ideal monitoring of CAR T-cell therapies should include the ability to track T-cell migration, engagement with the antigen-bearing tumor cells, as well as T-cell expansion and persistence at the tumor site—all essential steps for therapeutic efficacy. Imaging studies might perhaps also enable timely intervention to avoid potentially lethal systemic toxicity. Current clinical methods to monitor the infused cells include serum profiling of cytokines associated with T-cell activation, direct enumeration of tumor-specific T-cells in peripheral blood, and (repeated) tumor biopsies. Overall, the in vivo activation dynamics of engineered immune cells remain incompletely understood, as no means of real-time monitoring of the intratumoral milieu currently exists.
PRINCIPLES OF CELL TRACKING
Cell tracking by imaging requires the direct or indirect labeling of cells. Direct labeling is relatively simple and has been used clinically for several decades, for instance, with 99mTc or 111In-labeled leukocytes. More recently, 89Zr-oxine has been used for cell labeling and PET imaging (4). As a major disadvantage, the label is diluted over time when cells divide, resulting in decreased amounts of label per individual cell, narrowing the time window for cell tracking. In addition, direct cell labeling does not permit visualization of cell proliferation, cell activation, or cell death. So far, no clinical CAR T-cell studies using direct cell labeling have been reported.
Indirect labeling uses a reporter gene that is introduced into the genome of the cell and then translated into a protein (e.g., enzyme or cell surface receptor), engaging a radiolabeled probe. Stable expression of a reporter gene theoretically enables serial imaging of labeled cells over a prolonged time. As the reporter gene is passed on to cell progeny, imaging of expanding cell populations is possible, providing information about cell viability. One prerequisite is that the physical half-life of the isotope label matches the pharmacokinetics of the transplanted T-cells. Nevertheless, under clinical conditions, reporter gene imaging remains challenging and success stories are rare (5,6). This is partly related to constraints in vector design and vector delivery, as well as the biologic and potentially immunogenic effects of reporter gene products on the engineered cells.
CLINICAL IMAGING
Herpes simplex virus type 1 thymidine kinase has been the most widely used reporter gene and is currently used to track CAR T-cells targeting interleukin-13 receptor α2 and prostate-specific membrane antigen in clinical studies (NCT00730613, NCT01082926, NCT01140373). Yaghoubi et al. reported the noninvasive detection of cytotoxic T lymphocytes in a patient with recurrent glioblastoma; cytotoxic T lymphocytes were engineered to express herpes simplex virus type 1 thymidine kinase and an interleukin-13 CAR, enabling PET imaging with the reporter probe 18F-FHBG (5). The patient underwent gross tumor resection followed by serial autologous CD8+ CAR T-cell infusions (cumulative T-cell dose of 1 × 109) into the postsurgical cavity. Three days after the last T-cell dose, 18F-FHBG was given intravenously, followed by PET imaging. PET/MR fusion images showed tracer accumulation at the postsurgical site, and also in the contralateral hemisphere, at a nonresected tumor site. Recently, these authors reported a small clinical trial in 7 patients with recurrent glioma, using the same approach. Although the imaging signal was relatively low (with SUVs below 1.0), this did not limit T-cell detection, because there was negligible background activity in normal brain. However, such low background activity cannot be expected in other body regions. Nonspecific radiotracer uptake was also noted in 1 patient (positive baseline scan before cytotoxic T-lymphocyte administration) (6).
Although this first clinical study is pivotal for the field, it highlights the current limitations in CAR T-cell imaging: detection of a small number of labeled cells in a large volume is confounded by the combination of low imaging signal from labeled cells and nonspecific background activity. This will be even more challenging when CAR T-cells are administered systemically rather than via intracavitary methods. The lack of a corresponding positive imaging signal could be limited by lack of sophisticated imaging modalities, suboptimal tumor-specific binding of T-cells and receptor engagement, or absence of T-cell activation, migration, and expansion. Studies can also be compromised by the immunogenicity of a xenogeneic (nonhuman) reporter gene (7).
IMAGING MODELS
Potential immunogenicity of the reporter gene product may be avoided when using human reporter gene systems (e.g., the human sodium iodide symporter, human norepinephrine transporter, human somatostatin receptor subtype 2, human deoxycytidine, and thymidine kinase mutants (8), and prostate-specific membrane antigen) or using artificial constructs, such as a membrane-bound cell-surface antibody fragment binding small metal chelates (9). For instance, the human sodium iodide symporter reporter gene has been used for CAR T-cell tracking in prostate cancer mouse models (10). However, sodium iodide symporter is naturally found in stomach, salivary glands, lactating breast, and thyroid tissue, limiting the specificity of this approach (11). Expression of human somatostatin receptor subtype 2 has been evaluated in adoptively transferred CAR T-cells targeting intercellular adhesion molecule 1, which is overexpressed in anaplastic thyroid cancer, and imaged with 68Ga-DOTATOC PET/CT (12). However, human somatostatin receptor subtype 2 internalizes on binding of the ligand, potentially interfering with T-cell function, and is physiologically expressed on immune cells, again limiting the specificity of this approach (13). Human deoxycytidine kinase double mutant has been tested in prostate-specific membrane antigen-targeted human CAR T-cells as a pyrimidine-specific PET reporter/suicide gene for imaging with 18F-FEAU (14).
Recently, our group has tested a membrane-bound cell-surface antibody fragment using small metal chelates (e.g., DOTA) as a reporter probe, which is a purely synthetic molecule with no natural analogs. Modification by naturally occurring enzymes or competition for target binding by endogenous molecules is thereby avoided. Moreover, only cells modified with this construct should bind the radio-hapten, thus enabling exquisite specificity (9).
NOVEL CONSTRUCTS
CAR T-cells are currently generated using randomly integrating vectors, including γ-retroviral vectors, lentiviral vectors, and DNA transposons, to insert the CAR complementary DNA in the T-cell genome (15). Although effective, this approach may result in variegated gene expression, clonal expansion, oncogenic transformation, and transcriptional silencing (16–18). The emergence of targeted nuclease clustered regularly interspaced short palindromic repeats (CRISPR) provides a new means to specifically disrupt endogenous genes or target transgene delivery to chosen locations (19–21), potentially resulting in greater gene expression (22). A higher density of expressed reporter molecules should facilitate detection of the modified cells in vivo. “Armored” CAR T-cells that also affect the tumor immune microenvironment may lead to greater efficacy of treatments (23).
REAL-TIME APPLICATION
Other T-cell imaging strategies take advantage of specific T-cell surface markers, such as CD8 (NCT03107663) (24). Another strategy is based on imaging T-cell activation with 2′-deoxy-2′-18F-fluoro-9-β-d-arabinofuranosylguanine, a radiolabeled analog of the DNA nucleoside guanine (NCT03142204) (25). The utility of these methods for CAR T-cell tracking has yet to be evaluated. Finally, serial imaging studies with probes binding to T-cell–inhibiting and –activating molecules, such as CTLA-4, PD-1/PD-L1, LAG-3, IL-2R, and OX40 (26–28), may enable understanding of interactions between CAR T-cells and the tumor microenvironment and, potentially, provide real-time response assessment.
CONCLUSION
Noninvasive imaging is a tool that remains to be exploited for assessing the systemic kinetics and functions of CAR T-cells in humans. The field is ripe with novel approaches currently in development.
DISCLOSURE
No potential conflict of interest relevant to this article was reported.
Footnotes
Published online May 3, 2019.
- © 2019 by the Society of Nuclear Medicine and Molecular Imaging.