Imaging of T-cell Responses in the Context of Cancer Immunotherapy

Abstract

Immunotherapy, which promotes the induction of cytotoxic T lymphocytes and enhances their infiltration into and function within tumors, is a rapidly expanding and evolving approach to treating cancer. However, many of the critical denominators for inducing effective anticancer immune responses remain unknown. Efforts are underway to develop comprehensive ex vivo assessments of the immune landscape of patients prior to and during response to immunotherapy. An important complementary approach to these efforts involves the development of noninvasive imaging approaches to detect immune targets, assess delivery of immune-based therapeutics, and evaluate responses to immunotherapy. Herein, we review the merits and limitations of various noninvasive imaging modalities (MRI, PET, and single-photon emission tomography) and discuss candidate targets for cellular and molecular imaging for visualization of T-cell responses at various stages along the cancer–immunity cycle in the context of immunotherapy. We also discuss the potential use of these imaging strategies in monitoring treatment responses and predicting prognosis for patients treated with immunotherapy.

Background

Therapies that harness the immune system to mount effective antitumor responses have revolutionized cancer therapy. Several mAb-based therapies that target cytotoxic T lymphocyte–associated antigen 4 (CTLA-4) and programmed cell death receptor 1 (PD-1) or its ligand PD-L1 have been approved by the FDA and European Medicines Agency (EMA) in the past decade (1, 2). Blocking these inhibitory pathways that suppress the function of effector T cells can effectively restore endogenous antitumor immunity (3, 4). Cell-based immunotherapies, such as chimeric antigen receptor (CAR) T cells, have also demonstrated remarkable tumor responses (5). Nevertheless, substantial proportions of patients are resistant to or acquire resistance to these immunotherapies (6–8). Based on the initial success in some patients but also the lack of response in many, there is now a major ongoing effort to explore the potential to improve efficacy and overcome resistance by combining multiple immune-based therapies and/or combining immunotherapy with other treatment modalities (7, 8).

The rationale of immunotherapeutic strategies focuses on inducing or enhancing cytotoxic T lymphocyte (CTL) infiltration into and function within tumors (9, 10). Currently, biopsy and peripheral blood analyses are recognized as the “gold standards” to assess immune responses following immunotherapy (11). However, these methods do not reflect the dynamic spatiotemporal aspects of immune responses (11). Therefore, development of noninvasive imaging modalities for visualization of T-cell responses is of clinical importance to evaluate immune responses in patients and facilitate design of more effective anticancer immunotherapies.

Clinical imaging modalities, such as CT, MRI, single-photon emission tomography (SPECT), and PET, allow a view of whole organs noninvasively and longitudinally, and are incorporated into clinical routines for staging patients at baseline and for monitoring tumor responses during treatment. CT measures tumor volume and characterizes the enhancement patterns after the injection of contrast agents. It is widely used in guiding clinical decision-making and developing treatment plans but contributes little information on specific immunologic pathways critical for the efficacy of immunotherapy. With the use of specific contrast agents or radiotracers that target specific molecules, pathways, and cellular components, the other imaging modalities have been developed to assess T-cell responses in patients with cancer during immunotherapy (12, 13). These noninvasive clinical imaging technologies show high potential to become integral tools for further developing anticancer immunotherapy.

The cancer–immunity cycle (9) provides a conceptual framework for this review to illustrate the utility of visualizing T-cell responses in vivo using different imaging approaches (Fig. 1). First, we introduce the cancer–immunity cycle with emphasis on targets and processes relevant for imaging purposes. Next, we address some of the questions regarding these immunologic processes raised by current clinical immunotherapy experiences. Finally, we discuss the potential and requirements of various imaging modalities to address these questions, and how imaging technologies might advance in vivo evaluation of T cells during immunotherapy.

Figure 1.

Imaging T cells in the cancer–immunity cycle. The cancer–immunity cycle begins when endogenous anticancer immune responses kill cancer cells, causing antigens to be released (1). These antigens are captured, processed, and presented by antigen-presenting cells (APC; 2), leading to priming and activation of antigen-specific T cells (3). The T cells traffic to and infiltrate into the tumor (4, 5), where they recognize and then kill cancer cells (6, 7), starting the cycle again. As shown here, different clinical imaging modalities are applicable for imaging T-cell response at different points in the cancer–immunity cycle.

The Cancer–Immunity Cycle and In vivo Imaging

The cancer–immunity cycle conceptually captures a series of ordered events that are necessary for an effective anticancer immune response (9). Endogenous anticancer immune responses are initiated by the release of cancer cell–associated antigens that are captured and processed by antigen-presenting cells (APC) such as dendritic cells (DC). These APCs can prime and activate the effector functions of antigen-specific T cells. The effector T cells traffic to the tumor where they can, at least in some instances, infiltrate the tumor microenvironment (TME) and execute cytotoxic effector functions. Killing of the cancer cells releases additional cancer cell–associated antigens, resulting in the induction of new waves of effector cells in subsequent revolutions of the cycle (9, 10).

In the process of T-cell activation and expansion, several cell-surface markers are increasingly expressed by T cells (4), and glycolysis and nucleic acid metabolism are upregulated as well (14, 15). Targeted imaging of early activation markers and metabolic reprogramming can potentially serve as indicators of successful induction of immune responses. Checkpoint molecules such as CTLA-4 and PD-1 are also typically upregulated on activated T cells, and when engaged by their ligands, they act to tone down the activation state of T cells to balance between host tissue integrity and effective cytotoxicity.

Expressions of several receptors and their ligands are essential for T-cell trafficking to and infiltration of tumors. For instance, INFγ released by memory T cells enhances the recruitment of circulating T cells into tumors (16). One major barrier to successful adoptive cell therapies for solid tumors is the limited accumulation of ex vivo–expanded tumor-infiltrating lymphocytes (TIL), following systemic administration. The persistence and distribution of T cells determine the balance between immunotherapeutic efficacy and potential adverse events, and therefore, need to be evaluated by noninvasive imaging approaches. From an imaging perspective, three critical issues need to be considered when seeking to track small numbers of therapeutic T cells. First, the injected T cells should be marked with tags that do not decay and are not diluted upon cell division (17). Second, the label should have no detrimental effects on the viability or function of the cells (18, 19). Finally, highly sensitive modalities that provide a wide view of the whole body, such as SPECT, PET, and MRI, are required.

A stiff and dense extracellular matrix in the TME can, depending on specific architectural features, form a physical barrier to prevent T-cell infiltration (20, 21). Several cells in the TME, including cancer-associated fibroblasts, myeloid-derived suppressor cells, and regulatory T cells, express immune inhibitory molecules (e.g., PD-L1) and release immune suppressive mediators (e.g., TGFβ, PGE2, IDO, and adenosine), which restrain infiltration of T cells and inhibit the effector functions of those CTLs that do infiltrate the tumor (1, 22). Based on the nature and extent of T-cell infiltration in the TME, cancers can be classified as inflamed (tumor-cell nests infiltrated by immune cells), excluded (tumor-cell nests surrounded by but not infiltrated by immune cells), and immune deserts (no or sparse T cells in the tumor) categories (23, 24). Clinical trials have shown that T-cell infiltration status is highly associated with patients’ prognosis (25) and treatment responses to immunotherapy (26). Moreover, complex interactions during T-cell activation and eventual tumor-cell killing determine the immunotherapy response with a variety of patterns (27, 28). Therefore, imaging T-cell infiltration of tumors and antitumor effector function using high specificity and quantitative imaging modalities has the potential to allow clinicians to predict tumor responses to immunotherapy and guide tailored treatment. In this respect, highly specific mAbs/antibody fragments with matched biological half-lives are preferred in designing such contrast agents.

In vivo Imaging of T Cells by Targeting Indicators of Metabolic Reprogramming

During activation, T cells undergo metabolic reprogramming. The adaptive metabolic pathways, therefore, have potential as targets of imaging tracers to distinguish between active and nonactive T cells. For example, deoxycytidine kinase (dCK), the key enzyme in the cytosolic deoxyribonucleoside (dN) salvage pathway, imaged with 1-(2′-deoxy-2′-[¹⁸F]fluoroarabinofuranosyl) cytosine ([¹⁸F]FAC), shows high sensitivity to local immune activation and specificity to mature populations of T cells such as effector CD8⁺ T cells (29, 30). The ¹⁸F-labeled clofarabine ([¹⁸F]CFA) probe for dCK exhibits positive correlations between radiotracer accumulation and dCK expression in leukemia cells, and these can be abrogated by treatment with a dCK inhibitor (31). More importantly, [¹⁸F]CFA PET/CT also demonstrates preferential accumulation in tissues with high dCK expression such as hematopoietic bone marrow and secondary lymphoid organs (31, 32), revealing the potential of [¹⁸F]CFA PET as an integral biomarker for monitoring treatment responses. In patients with glioblastoma (GBM) vaccinated with tumor lysate–pulsed DCs and anti–PD-1 therapy, [¹⁸F]CFA PET tracers accumulate in both tumor and secondary lymphoid organs, showing the promising use of [¹⁸F]CFA PET in noninvasively localizing and quantifying immune responses induced by immunotherapies (33).

Deoxyguanosine kinase (dGK) is another rate-limiting enzyme in the dN salvage pathway (34), which can be targeted for therapy and PET imaging in cancer. A PET tracer targeting dGK, 2′-deoxy-2′-[¹⁸F]fluoro-9-β-D-arabinofuranosylguanine ([¹⁸F]F-AraG), shows preferential accumulation in activated T cells (35, 36). In addition, [¹⁸F]F-AraG PET imaging enabled the visualization of donor T-cell activation prior to symptoms in a murine model of acute graft-versus-host disease and showed favorable biodistribution in healthy humans, highlighting the usefulness of [¹⁸F]F-AraG in imaging activated T cells (37).

Successful therapeutic cancer vaccination induces lymphocyte proliferation, during which thymidine kinase 1 (TK1) is upregulated and incorporates the thymidine analog into newly synthesized DNA (38). 3′-Deoxy-3′-[¹⁸F]fluorothymidine ([¹⁸F]FLT), a thymidine analog that can measure the activity of TK1, is a promising target to quantify cell proliferation in this regard (38). [¹⁸F]FLT was used as a PET tracer to assess antigen-specific immune responses after administration of DC vaccines in a clinical study (39). [¹⁸F]FLT uptake profoundly increased in the lymph nodes of patients with melanoma treated with small numbers of DCs (4.5 × 10⁵ cells), whereas [¹⁸F]FLT uptake was absent in control lymph nodes treated with saline or DCs without antigen (39). Moreover, de novo immune responses could be visualized by increased [¹⁸F]FLT uptake early after primary vaccination and persisted for up to 3 weeks. More promisingly, the magnitude of [¹⁸F]FLT uptake correlated with the level of antigen-specific proliferative responses of T cells in peripheral blood (39). Hence, PET imaging offers a sensitive tool to discriminate responding from nonresponding patients treated with anticancer vaccines (39).

Blockade of the coinhibitory molecule CTLA-4 enhances T-cell proliferation, tumor targeting, and regression. [¹⁸F]FDG and [¹⁸F]FLT PET, which measure glucose metabolism and cell replication, respectively, have been performed in patients with advanced melanoma receiving the CTLA-4–blocking antibody tremelimumab to quantify cell proliferation in secondary lymphoid organs after CTLA-4 blockade (40). Although an absence of [¹⁸F]FDG uptake and increased [¹⁸F]FLT uptake were observed in the spleen as a result of immunotherapy, the number of patients in that study was insufficient to correlate the alterations of [¹⁸F]FLT uptake in the spleen to either tumor responses or adverse events (40). Therefore, [¹⁸F]FDG PET imaging for immunotherapy response assessment is still unclear (41). Some studies suggest a positive role for predicting early progression using [¹⁸F]FDG PET imaging (42, 43); however, tumor cells are frequently associated with abundant glycolysis, confounding the interpretation of the [¹⁸F]FDG signal at the tumor site. Despite some success in assessing immune-related adverse events (44), the potential for using [¹⁸F]FDG PET imaging to quantify early metabolic switches after immunotherapy still needs to be validated (45).

In summary, imaging for alterations in T-cell metabolism has the potential to aid in assessment of the immune responses at early time points in the context of immunotherapy. However, most tumors are characterized by metabolic plasticity in which various metabolic pathways are exploited to fuel the energy demands of tumor cells; in this regard, signal interpretation at tumor sites will be a formidable challenge. A possible solution for imaging intratumorally is to target specific T-cell surface receptors (Table 1).

Table 1.

Representative protein/peptide/small-molecule–based T-cell targeting imaging probes.

Target	Modality	Tracer	Type	Species	Pros and cons	Stage	Reference
Metabolic reprogramming
dCK	PET	[18F]FAC	Small molecule	Murine	Pros: better selectivity for lymphoid organs; Cons: rapid catabolism	Preclinical	(29, 30)
	PET	[18F]CFA	Small molecule	Murine/human	Pros: high specificity for dCK and favorable biodistribution; Cons: possible false-negative scans (competitive uptake by endogenous nucleosides)	Clinical	(31–33)
dGK	PET	[18F]F-AraG	Small molecule	Murine/human	Pros: low background and favorable kinetics	Clinical	(37)
TK1	PET	[18F]FLT	Small molecule	Murine/human	Pros: high sensitivity and specificity	Clinical	(38, 39)
Cell-surface activation markers
IL2	SPECT	99mTc-HYNIC-IL2	Cytokine	Human	Pros: short plasma half-life; Cons: expression of IL2 varied at cellular level	Clinical	(50, 51)
	PET	[18F]FB-IL2	Cytokine	Murine	Pros: short plasma half-life and good correlations with IHC staining	Preclinical	(52)
CXCR4	PET	[64Cu]Cu-AMD3100	Monoclonal antibody	Human	Cons: potential overlap in the signals between immune cells and other sources	N/A	(58)
OX40	PET	[64Cu]Cu-DOTA-AbOX40	Monoclonal antibody	Murine	Potential cons: heterogeneous expression in each cell	Preclinical	(59, 60)
ICOS	PET	[89Zr]Zr-DFO-anti-ICOS	Monoclonal antibody	Murine	Potential cons: ICOS+ T cells may not be implicated in all models or therapy types	Preclinical	(61, 62)
Cell-surface lineage markers
CD3	PET	[89Zr]Zr-DFO-CD3	Monoclonal antibody	Murine	Pros: directly monitor immune response; Cons: receptor present on both CD8+ and CD4+ cells, probe dilution by cell proliferation, long plasma half-life and toxicity	Preclinical	(63)
CD4	PET	[89Zr]Zr-malDFO-GK1.5 cDb	Cys-diabody	Murine	Pros: relatively rapid clearance and less dependent on tissue structure and perfusion	Preclinical	(67)
CD8	PET	[89Zr]Zr-malDFO-169 cDb	Cys-diabody	Murine	Pros: comparable stability, reduced radiation exposure, and high image contrast with minimal adverse effects	Preclinical	(65, 68)
	PET	[89Zr]Zr-Df-IAB22M2C	Minibody	Human	Pros: safe and has favorable kinetics	Clinical	(69)
TCR	PET	[64Cu]Cu-cOVA-TCR	Monoclonal antibody	Murine	Pros: efficient internalization and accumulation of the tracer into cells	Preclinical	(70)
	PET	[89Zr]Zr-Df-aTCRmu-F(ab’)2	Fab′2	Human	Pros: higher specific activities with the exploitation of the reshuttling TCR; Cons: possible induction of signaling and probe dilution by cell proliferation	Preclinical	(71)
Effector molecules
PD-L1	SPECT/CT	111In-anti-PD-L1	Monoclonal antibody	Murine/human	Pros: retained immunoreactivity after radiolabeling, high affinity, and high contrast; Cons: lower resolution, sensitivity, and lack of quantification compared with PET tracers	Preclinical	(72)
PD-1	PET	[64Cu]Cu-DOTA-PD-1	Monoclonal antibody	Murine	Cons: slow accumulation in peripheral tissue, thus multiple day-acquisition protocols	Preclinical	(73)
	PET	[89Zr]Zr-DFO-nivolumab	Monoclonal antibody	Human	Pros: accurate reflection of antibody biodistribution in vivo; Cons: no established correlation with molecule expression levels	Clinical	(74)
CTLA-4	PET	[64Cu]Cu-DOTA-anti-CTLA-4	Monoclonal antibody	Murine	Pros: accurate reflection of antibody biodistribution in vivo	Preclinical	(75)
Granzyme B	PET	[68Ga]Ga-NOTA-GZP; [18F]AlF-mNOTA-GZP	Peptide	Murine/human	Pros: less dependent on tissue perfusion and possibly better reflection of actual molecule expression levels	Preclinical	(76–78)
IFNγ	PET	[89Zr]Zr-anti-IFNγ	Monoclonal antibody	Murine	Cons: no established correlation with molecule expression levels	Preclinical	(79)

Abbreviations: N/A, not applicable; TCR, T-cell receptor.

In vivo Imaging of T Cells by Targeting Cell-Surface Activation Markers

DC-based therapies have been developed to induce immune responses by delivering activated DCs loaded with tumor-associated antigens (46, 47). Although successfully utilized to confirm the delivery of DC vaccines, anatomical focused imaging modalities, including MRI and ultrasound, have offered limited capacity to detect and quantify T-cell responses (48). As such, sensitive and quantitative imaging techniques (SPECT and PET) with the use of sufficiently specific tracers are potentially preferable modalities for visualizing and characterizing T-cell activation.

The cytokine IL2 binds with high affinity to IL2 receptors highly expressed on activated T cells (49). Use of radiolabeled IL2 for SPECT imaging has been reported in the context of metastatic melanoma (50, 51). Following injection of [^99mTc]Tc-succinimidyl-6-hydrazinopyridine-3-carboxylate-IL2 (^99mTc-HYNIC-IL2), both the primary melanoma and most melanoma metastases showed radiotracer accumulation. Treatment with ipilimumab or pembrolizumab resulted in increased uptake in some lesions, whereas other lesions demonstrated decreased uptake. However, due to limited sample size and histologic analyses, no conclusions could be drawn about a potential relationship between extent of T-cell infiltration and level of ^99mTc-HYNIC-IL2 uptake. To provide higher sensitivity and spatial resolution, PET imaging probes, such as ¹⁸F-labeled IL2 tracer, have been developed and are currently being translated to the clinic. In tumor-bearing mice, irradiation alone or in combination with viral-vector–based immunization using recombinant Semliki Forest virus (rSFV) encoding a fusion protein of E6 and E7 derived from human papilloma virus type 16 (rSFVeE6,7) increased uptake of N-(4-[¹⁸F]fluorobenzoyl)-IL2 ([¹⁸F]FB-IL2) 10- and 27-fold compared with no irradiation, respectively (52). In contrast, the administration of a CXCR4 antagonist to inhibit activated T-cell infiltration into the tumor reduced the uptake of [¹⁸F]FB-IL2 (52).

CXCR4 and its ligand CXCL12 are key factors in lymphocyte trafficking and mediate angiogenesis and cell proliferation (53). Radiolabeled CXCR4 antagonists have been utilized in PET imaging for multiple myeloma (54), non–small cell lung cancer (NSCLC; ref. 55), and cardiovascular diseases (56, 57). However, potential overlap in the signals of radiolabeled CXCR4 ligands, [⁶⁴Cu]Cu-AMD3100, between immune cells and other sources, such as tumor cells and red blood cells (58), hinders its further use for imaging in the context of immunotherapy.

OX40 is another biomarker of activated T cells (59). PET imaging based on a ⁶⁴Cu-conjugated antibody specific for murine OX40 was applied to quantify T-cell activation in A20 lymphoma tumors, tumor-draining lymph nodes, and spleen following in situ vaccination with CpG oligodeoxynucleotide (59). Tumor uptake of this PET tracer on day 2 was capable of predicting tumor response on day 9 after local CpG treatment, indicating that this approach is a promising candidate for predicting and monitoring anticancer immune responses (59). Similarly, in A20 tumor–bearing mice treated with intratumoral injection of CpG, the enhancement of OX40 expression could be visualized by whole body small animal PET imaging after tail-vein administration of [⁶⁴Cu]Cu-1,4,7,10-tetraazacyclododecane-N,N,N, N-tetraacetic acid-OX40 ([⁶⁴Cu]Cu-DOTA-OX40; ref. 60). Based on these promising findings, radiolabeled humanized OX40 agonist mAbs for PET imaging are currently under evaluation for clinical translation. A new promising T-cell activation biomarker, inducible T-cell costimulatory receptor (ICOS), has been radiolabeled with ⁸⁹Zr to develop a PET radiotracer for predicting and monitoring T-cell–mediated immune responses to different types of cancer immunotherapies (61, 62).

In summary, imaging early makers of T-cell activation can provide key insights of successful induction of immune responses. However, expression of activation markers cannot indicate cytotoxic effector functions, which might limit interpretation of results using these probes. Moreover, the expression of activation markers is temporally regulated, hampering direct correlation of signal intensity derived from PET or SPECT with cell numbers (Table 1).

In vivo Imaging of Specific T-cell Populations by Targeting Cell-Surface Lineage Markers

CD3 is a part of the T-cell receptor (TCR) complex that serves as a pan–T-cell marker. CD3-based PET imaging provides an accurate and quantitative assessment of T-cell infiltration, which is helpful for predicting and monitoring T-cell responses. In colon cancer xenograft models treated with anti–CTLA-4, murine anti-CD3 conjugated with desferrioxamine (DFO) chelate was radiolabeled with ⁸⁹Zr to measure T-cell infiltration, and increased [⁸⁹Zr]Zr-DFO-CD3 uptake in the tumor correlated with subsequent tumor regression, defined as reduced tumor volume (63). Moreover, radiotracer-labeled intact antibodies typically accumulate at a slow rate in peripheral tissues, which may detract from their utility.

Antibody fragments exhibit enhanced distribution kinetic and more rapid clearance compared with intact antibodies (64). To date, ⁸⁹Zr-labeled anti-CD4 and anti-CD8 cys-diabodies have been designed to track the respective populations of T cells in some preclinical models. Due to comparable stability and reduced radiation exposure, copper-64 (⁶⁴Cu) was utilized to label anti-CD8 cys-diabody for T-cell tracking and quantifying TILs in tumor-bearing mice treated with anti–PD-1 and CpG (65). More importantly, microdoses of the antibody fragments used for PET imaging could offer high image contrast with minimal adverse effects on T cells in these studies (65). This approach also demonstrated utility in evaluating ensuing T-cell responses and intratumoral distribution after adoptive cell transfer, anti–PD-L1 treatment, and combination treatment with antibodies specific for TNF receptor superfamily member 9 (4-1BB) and tumor antigen–targeted T-cell bispecific molecules (64–68). Notably, PET/SPECT tracers radiolabeled with engineered bispecific proteins that target the 4-1BB molecule on T cells and a tumor antigen (CD19–4-1BBL) or tumor stromal marker (FAP-4–1BBL) could increase the possibilities to track T cells after immunotherapy (66). So far, engineered humanized CD8-targeted antibody fragment, IAB22M2C, has been radiolabeled with ⁸⁹Zr as a PET/CT probe to successfully target CD8⁺ T-cell–rich tissues in a first-in-human prospective study, showing high clinical translational potential for its usefulness for in vivo imaging and to optimize immunotherapy (69).

TCRs are highly attractive imaging targets due to efficient internalization and accumulation of the tracer in cells. With the employment of ex vivo or in vivo T-cell labeling strategies, a ⁶⁴Cu-labeled antibody specific for a TCR recognizing chicken ovalbumin (OVA) was effectively internalized with antigen recognition preserved and minimal impact on cell viability, DNA damage, and apoptosis–necrosis induction (70). Radiolabeling of an anti-mouse TCR F(ab’) 2 fragment, selective for the murine TCR beta domain of a transgenic TCR, was optimized ex vivo to achieve higher specific activities, and then used for in vivo T-cell imaging and engineered human T-cell tracking (71). Of note, PET signals showed strong correlation to total numbers of transgenic T cells detected ex vivo, independent of the engraftment rates observed in different individual experiments (71).

In summary, imaging of cell-surface markers is emerging as an approach for in vivo detection of specific T-cell populations; however, quantification via this approach still requires optimization because the expression of cell-surface markers varies during immunotherapy. Moreover, the expression of these markers may not always reflect T-cell function. Another potentially promising approach is to target inhibitory pathways and other effector molecules either alone or in combination with the targets discussed here (Table 1).

In vivo Imaging of T Cells by Targeting Effector Molecules

PD-1 is induced during T-cell activation. In the setting of cancer, cancer cells and stromal cells upregulate expression of PD-L1 which binds PD-1 on T cells. Quantitative imaging of these proteins may guide selection of patients for mAb-based immunotherapies that target the PD-1/PD-L1 axis, optimize treatment schedules, and design novel combination therapies.

Preclinical studies indicate that ¹¹¹Indium(¹¹¹In)-labeled anti–PD-L1 can detect PD-L1 expression in tumor-bearing immune-competent mice and can be used to monitor alterations in PD-L1 expression induced by treatment (72). ⁶⁴Cu-labeled anti–PD-1 PET radiotracer has also been reported to detect murine PD-1–expressing TILs, in which tracer uptake was found in both lymphoid organs and tumor (73). Recently, a landmark first-in-human study employed ⁸⁹Zr-labeled nivolumab in patients with NSCLC to evaluate PD-1 expression in the tumor before anti–PD-1 immunotherapy (ref. 74; Fig. 2). In that study, [⁸⁹Zr]Zr-DFO-nivolumab uptake correlated with PD-1–expressing lymphocytes in tumor biopsies, and pretreatment [⁸⁹Zr]Zr-DFO-nivolumab uptake was higher in tumors that subsequently responded to nivolumab treatment than those that did not respond, yielding a higher predictive score even than immunohistochemical markers (74).

Figure 2.

⁸⁹Zr-labeled nivolumab can be used to analyze PD-1 expression in tumors. Representative PET images obtained with ¹⁸F-BMS-986192 and ⁸⁹Zr-nivolumab in patients with advanced NSCLC. Tracer uptake was heterogeneous in different organs in the patient (A). ¹⁸F-BMS-986192 and ⁸⁹Zr-nivolumab uptake in tumor lesions, measured as SUV_peak, correlated with tumor PD-L1 expression (B) and PD-1⁺ TILs (C), respectively. Adapted from Niemeijer AN, Leung D, Huisman MC, Bahce I, Hoekstra OS, van Dongen G, et al. Whole body PD-1 and PD-L1 positron emission tomography in patients with non-small-cell lung cancer. Nat Commun 2018;9 (1):4664. doi 10.1038/s41467-018-07131-y (ref. 74).

CTLA-4 targeted therapy is highly effective in some cancers but is often associated with autoimmune-related adverse effects. In this regard, developing a molecular imaging probe to visualize CTLA-4 expression in tumors will be of great clinical relevance. In a preclinical study, ⁶⁴Cu-tracer was conjugated with an anti–CTLA-4 for examining and quantifying CTLA-4 expression in a CT26 tumor-bearing mouse model (75). Accumulation of the radiotracer correlated well with high expression of CTLA-4 on T cells (75).

Granzyme B, a downstream effector of tumoricidal T cells, can be taken as an early biomarker for tumors responding to immunotherapy. Targeting granzyme B can increase specificity in detecting effective immune responses rather than increased presence of T cells (76, 77). Highly specific murine [1,4,7-triazacyclononane-N,N,N-triacetic acid (NOTA)-mGZP] and humanized (NOTA-hGZP) granzyme B peptides have been radiolabeled with ⁶⁸Ga to analyze granzyme B expression in several tumor models, and quantitative correlations were found between PET signal and magnitude of response (76, 77). Using this approach, granzyme B expression could be detected with high specificity, and immunotherapy-treated responders could be distinguished from nonresponders with excellent predictive ability (76). In another recent study, a ¹⁸F-labeled murine granzyme B peptide, [¹⁸F]AlF-mNOTA-GZP, was developed to stratify immune checkpoint immunotherapy response in two syngeneic models of colon cancer, CT26 and MC38, showing good predictive ability and correlating well with changes in tumor-associated T cells (78).

IFNγ is another promising imaging target for mapping active antitumor immunity. An ⁸⁹Zr-labeled murine anti-IFNγ PET tracer was developed to detect increased levels of IFNγ and monitor immune responses following HER2/neu vaccination in spontaneous salivary and orthotopic neu⁺ mouse mammary tumor models. Specificity was confirmed in a model of induced T-cell anergy where CD8⁺ T cells infiltrate the tumor but upregulate PD-1. In this model, IFNγ-tracer uptake did not exceed isotype control, consistent with a lack of antitumor T-cell activity. Hence, targeting of soluble cytokines such as IFNγ may provide additional tools and insights into the function of immune cells in situ (79).

In summary, direct assessment of the presence, amount, and functionality of specific populations of immune cells in vivo is feasible, and can provide a path to transition from two-dimensional immunohistochemistry to three-dimensional whole-body imaging approaches to support the development of treatment plans and to assess responses to immune-based therapies. However, it should also be noted that imaging cell-surface/intracellular biomarkers (e.g., CD3, CD4, and CD8) and immune checkpoint ligands (e.g., PD-1/PD-L1) has some drawbacks such as background signals found in secondary lymphoid tissues and the provision of limited information on downstream CTL effector function. As such, compared with the cell-surface/intracellular markers, imaging-secreted biomarkers seem to be much more specific to measure ongoing antitumor immunity, providing a predictive tool for noninvasive visualization of immune activity in situ in the context of immunotherapy (Table 1).

In vivo Imaging of Adoptively Transferred T-cell Therapies

The development of adoptively transferred T-cell–based immunotherapies using live cell products has been challenging. The location, distribution, and long-term viability, as well as the fate of therapeutic T cells with respect to cell activation and differentiation, are key factors in determining the efficacy of such therapies. In addition, side effects, including on-target off-tumor activities (80) and potential for cytokine storm (81, 82), have also hindered the development of adoptively transferred T-cell therapies (including CAR-T therapy). Clinical development could be enhanced with increased knowledge about the in vivo distribution and fate of these potentially therapeutic cells. Thus, developing approaches to assess their safety, localization, and expansion at target sites is an important ongoing effort.

In vivo imaging of T-cell–based immunotherapy typically includes two strategies to label cells: (i) direct cell labeling, in which contrast agents coupled to exogenous markers are directly loaded into the therapeutic cells; (ii) indirect cell labeling, involving genetic modification of the cells to express a reporter gene that can be detected in vivo with administration of a contrast agent (83).

In vivo imaging of directly labeled T-cell therapeutics

The basic principle for in vivo imaging of directly labeled T cells requires tracer be retained in or on the cells without perturbing function and needs to accommodate the kinetics of in vivo trafficking without dilution by cell proliferation. There are two major methods for direct cell labeling, including “cell loading” by passive diffusion processes and “cell binding” to surface proteins using chelates. The former typically is determined by membrane surface area versus cell volume ratio, whereas the latter mostly depends on the number of receptor/targets, available amines for chelation. The sensitivity of imaging modalities based on direct labeling relies on the uptake of radioactivity into cells via several mechanisms, such as passive membrane diffusion, binding membrane molecules, or endocytosis.

For “cell loading,” two radiotracers, [⁹⁹mTc]Tc-hexamethylpropyleneamine oxime ([⁹⁹mTc]Tc-HMPAO) and [¹¹¹In]In-oxine, are widely used in several infectious and inflammatory diseases (84, 85). Pioneering clinical studies reported the accumulation of ex vivo–loaded [¹¹¹In]In-oxine–labeled CD4⁺ T cells in Hodgkin lymphomas (86) and adoptively transferred ex vivo–expanded TILs in metastatic melanomas with the use of SPECT imaging (87, 88). In a recent study, hemagglutinin-specific (HA-specific) CTLs were labeled with [¹¹¹In]In-oxine to explore the localization of these antigen-specific CTLs in HA-positive versus HA-negative murine colon cancer models using SPECT imaging (89). The labeled CTLs redistributed to HA-positive tumors and infiltrated into the core of the tumor starting 2 hours after injection, and the increased SPECT signal lasted for up to 120 hours after injection. However, insufficient sensitivity and lack of direct quantification hamper the application of scintigraphy and SPECT in clinical immunotherapy studies (90).

PET imaging should be a promising alternative to overcome the shortcomings of SPECT. With the use of optimized cell-labeling methods (such as lipophilic complexes), ⁶⁴Cu which provides a reasonable half-life (12.7 hours) is increasingly used in PET imaging (91–95). Nevertheless, rapid leakage or efflux of these labeled complexes from cells and robust subsequent uptake in the liver are a challenge to their clinical use. To achieve medium-to-long term in vivo tracking, [⁸⁹Zr]Zr-oxine labeling of some immune cells, such as CTLs, DCs, and natural killer cells, has recently been developed for PET imaging, and this yielded a longer half-life of 78.4 hours (96). In vivo trafficking of [⁸⁹Zr]Zr-oxine–labeled CAR T cells was also investigated in murine GBM and prostate cancer xenograft models using PET imaging (97). In a recent study, γδT cells were also labeled with [⁸⁹Zr]Zr-oxine for in vivo tracking in a breast cancer xenograft mouse model, and they could be detected within 48 hours in tumors with histopathologic confirmation (98). With this approach, radiolabeling CAR T cells and tumor cells with [⁸⁹Zr]Zr-oxine, which occurred with similar efficiency as [¹¹¹In]In-oxine labeling, was shown to yield cells that could be detected for up to 6 and 14 days, respectively, in mice. With better label retention and the availability of compatible manufacturing protocols with good manufacturing practice, [⁸⁹Zr]Zr-oxine has been acknowledged as a promising candidate for clinical cell labeling in the future (99).

An alternative for direct cell labeling, “cell binding”, involves directly binding chelators to cell-surface proteins. This approach circumvents the necessity of disrupting the plasma membrane in the process of cell labeling (18). However, this approach also has an inherent drawback, appreciable efflux of sequestered radioactivity in vivo, therefore possibly requiring higher doses for use in patients. Previously, [⁸⁹Zr]Zr-desferrioxamine-NCS ([⁸⁹Zr]Zr-DBN) was conjugated to protein amines expressed on the surface of human mesenchymal stem cells (18). In this case, cell labeling lasted for up to 7 days without interrupting cell viability and showed good in vivo biodistribution (18). In addition, CAR T cells were loaded with ⁶⁴Cu-labeled pegylated gold nanoparticles ([⁶⁴Cu]Cu-Au-NP) using electroporation, and the labeled CAR T cells could be detected predominantly in the lungs rather than the liver and spleen (100).

Regardless, PET imaging still has intrinsic disadvantages. For example, the half-life of most isotopes relative to the kinetics of T-cell trafficking is insufficient, and spatial resolution of PET is still poor. In this regard, cell tracking using MRI has been increasingly developed in the past decade (101). A highly derivatized cross-linked iron oxide nanoparticle (CLIO-HD) was synthesized to provide highly efficient intracellular labeling of cells, which allowed MRI tracking in vivo at near single-cell resolution with great biocompatibility (101). In this study, CLIO-HD–labeled CD8⁺ CTLs could be detected with a cutoff of two cells/voxel in vitro and three cells/voxel in vivo in live mice using 8.5 Tesla MRI (101). In addition, with this approach, high-resolution images could be obtained to trace the recruitment of CLIO-HD–labeled OVA-specific CD8⁺ T cells to tumor in vivo (101). Moreover, except for ¹H nucleus, MRI can also be performed on other nuclei, such as ¹⁹F, ¹³C, and ²³Na. Most promisingly, ¹⁹F MRI provides a higher sensitivity than ¹³C and ²³Na, high selectivity for detecting exogenous ¹⁹F, and direct fluorine quantification from MR images, and has been successfully employed in cell tracking in vivo (102, 103). In a pioneering study, ¹⁹F MRI was utilized to successfully track activated T cells in vivo over 21 days (104), and follow-up studies confirmed its efficiency for in vitro labeling of autoreactive CD4⁺ and CD8⁺ T cells (105). However, because of cell expansion and possible label loss, the amount of label at the site appeared disparate to actual cell numbers over time (106).

Both direct cell labeling approaches have certain drawbacks, including cytotoxicity. Labels using long-lived radioisotopes will increase radioactive dosing overtime (97) and be retained intracellularly for the apparent lifetime of the cell; thus, cytotoxicity or other adverse effects on cell function must be extensively tested (106). Radiotracer efflux from labeled cells and signal dilution as a result of immune cell proliferation are also critical issues that limit the use of these approaches (107, 108). In this respect, direct labeling of cells with these approaches is not well suited to tracking therapeutic cells that are expected to persist and replicate in vivo (Table 2).

Table 2.

Representative imaging probes for adoptively transferred T-cell therapies.

Method/target	Modality	Tracer	Type	Specie	Pros and cons	Stage	Reference
Direct labeling
Cell loading	SPECT	[111In]In-oxine	Chelating agent	Murine/human	Cons: impaired viability or functionality of T cells, efflux from the cell	Clinical	(86–89)
	SPECT	[99mTc]Tc-HMPAO	Chelating agent	Human	Cons: insufficient sensitivity and lack of direct quantification	Clinical	(85)
	PET	[64Cu]Cu-PTSM, [64Cu]Cu-PEI, [64Cu]Cu-tropolonate	Chelating agent	Murine	Pros: reasonable half-life; Cons: rapid leakage or efflux from cells and robust subsequent uptake in the liver	Preclinical	(91–95)
	PET	[89Zr]Zr-oxine	Chelating agent	Murine	Pros: longer half-life	Preclinical	(96–99)
Cell binding	PET	[89Zr]Zr-DBN	Chelating agent	Human	Pros: without interrupting cell viability and good in vivo biodistribution; Cons: probe dilution due to cell division and radiotoxicity of bone	Preclinical	(18)
	PET	[64Cu]Cu-Au-NP	Nanoparticle	Murine	Cons: electroporation used	Preclinical	(100)
	MRI	CLIO-HD	Nanoparticle	Murine	Pros: near single-cell resolution with great biocompatibility; Cons: ex vivo labeling of cells	Preclinical	(101)
	MRI	19F-NP	Stable isotope	Murine	Pros: high sensitivity and selectivity, direct quantification; Cons: probe dilution due to cell division and probe persistence after cell death	Preclinical	(105, 106)
Reporter gene-based T-cell imaging
HSV-1tk	PET	[18F]FHBG	Enzyme	N/A	Cons: immunogenicity	Clinical	(120)
NIS	PET	[124I]I-, [18F]BF4-, [18F]SO3F-, [18F]PF6-.	Symports sodium ions	Human	Pros: endogenously expressed	Clinical	(118, 127–137)
hSSTr2	PET	[68Ga]Ga-DOTATOC, [68Ga]Ga-DOTATATE	Cell-surface receptor	Human	Pros: endogenously expressed; Cons: potentially cause cell signaling and change proliferation	Preclinical	(138–141)
D2R	PET	[18F]FESP, [11C]C-Raclopride, [11C]C-N-methylspiperone	Cell-surface receptor	Human	Cons: Slow clearance of [18F]FESP and high background in the pituitary gland and striatum due to endogenous expression	N/A	(142, 143)
PSMA	PET	[18F]F-DCFPyL, [18F]F-DCFBC, [68Ga]Ga-PSMA-11	Cell-surface enzyme	Human	Pros: high expression in prostate; Cons: background in organs of excretion route	Clinical	(144, 145)

Abbreviation: N/A, not applicable.

In vivo imaging of T-cell therapeutics using reporter genes

Indirect cell labeling using reporter genes, characterized by ectopic expression of enzymes or transporters that can be passed on and maintained upon cell replication, allows tracking of the localization and monitoring of the survival of transferred cells in vivo (109). Despite the requirement of genetic engineering and high costs associated with this approach, it may still be preferable to direct cell labeling because it does not require ex vivo cell labeling and induces less cytotoxicity. On the other hand, reporter gene–based cell tracking is associated with a higher regulatory burden in view of concerns related to the potential for aberrant viral integration. With the development of new gene-editing strategies (e.g., CRISPR/Cas9 system), gene reporters now could be inserted into some safe harbor locations (such as AAVS1 and Rosa26) to reduce this risk (110).

Various reporter genes have been utilized for in vivo cell tracking using different imaging modalities (111). Despite the usefulness of optical signal-generating protein reporters such as fluorescent and luminescent proteins in preclinical studies, they contribute little to clinical cell tracking because of inherent issues such as imaging depth for humans and potential immunogenicity (111). Owing to high sensitivity and space resolution, reasonable depth penetration, and absolute quantification ability (112–114), SPECT and PET have been extensively used for clinical in vivo tracking of cell-based immunotherapy (115–118).

The sensitivity of cell detection depends on the reporter and imaging modality together with reporter expression levels. With the use of [¹⁸F]tetrafluoroborate-labeled sodium iodide symporter (NIS) PET radiotracer, hundreds or thousands of cancer cells have been effectively detected (119). In another study, up to tens of thousands of effector T cells were tracked using different reporter genes (117). In a clinical study, 9-(4-[¹⁸F]fluoro-3-[hydroxymethyl]butyl)guanine ([¹⁸F]FHBG) PET imaging was performed in patients with GBM treated with adoptive CTLs engineered to express herpes simplex virus 1 thymidine kinase (HSV-1tk) and IL13 zetakine (120). Before the administration of CTLs, small amounts of the radiotracer were taken up in GBM, whereas significant uptake was observed after administration of traceable CTLs (Fig. 3). However, immunogenicity of HSV-1tk was observed in another study (121). In this study, long-term persistence of adoptively transferred HSV-TK–modified donor T cells was not achieved in hematopoietic cell transplantation recipients due to the rapid induction of HSV-TK–specific CD8⁺ and CD4⁺ T-cell responses that lysed the gene-modified T cells (121). Mutant HSV1-tk, engineered by mutagenesis of the HSV1-tk–active site, could improve its enzymatic activity and be used as a potential reporter gene for cell labeling (122–126). However, the development of PET in vivo imaging of immune cells expressing mutant HSV1-tk reporter gene has not yet been reported, possibly because of the issue of immunogenicity.

Figure 3.

[¹⁸F]FHBG PET imaging can be used to image CTLs in GBM. Representative PET images obtained with [¹⁸F]FHBG in patients with GBM before (A) and 1 week after (B) CTL infusions. C, Voxel-wise analysis of [¹⁸F]FHBG SUV in pre- and post-CTL infusion scans. From Keu KV, Witney TH, Yaghoubi S, Rosenberg J, Kurien A, Magnusson R, et al. Reporter gene imaging of targeted T cell immunotherapy in recurrent glioma. Sci Transl Med 2017;9(373). doi 10.1126/scitranslmed.aag2196. Reprinted with permission from AAAS (ref. 120).

Immunogenicity of reporter genes is a general concern for imaging cell-based immunotherapies. To address this obstacle, PET radiotracers of endogenous human reporter proteins have been developed. One promising human reporter protein is the NIS, which is endogenously expressed only in thyroid, salivary, and lachrymal glands. In a recent first-in-human study, adoptively transferred therapeutic cells were traced by the expression of NIS, with PET signals increasing in the cancer after adoptive cell therapy compared with normal organs (127). Moreover, radiotracers will not be trapped by other organs that express NIS; therefore effectively decreasing the tracer doses needed for imaging (128). Thus, NIS is widely utilized to track cells in preclinical studies and CAR T cells in patients (118, 129–137).

Other reporters such as the human somatostatin receptor subtype 2 (hSSTr2), the dopamine type 2 receptor (D2R), and the prostate-specific membrane antigen (PSMA) also have been targeted for cancer and immunotherapy imaging in patients. The hSSTr2 is expressed in the brain, adrenal glands, kidneys, spleen, and multiple cancer types. Preclinical studies confirmed hSSTr2 can be employed in tracking tumor cells (138–140) and CAR T cells (141). Despite high background in the pituitary gland and striatum, D2R is an attractive target for noninvasive imaging for tumor cells (142, 143). PSMA is also a promising candidate for a human genetic reporter based on its selective expression pattern, and some PSMA imaging agents are available for clinical and preclinical applications (144, 145). One PSMA-based reporter-probe system has been used to track anti-CD19 CAR T cells in a Nalm6 model of acute lymphoblastic leukemia, showing the feasibility of this system for noninvasively and repeatably monitoring CAR T-cell disposition (145). Other human reporter genes, such as human thymidine 2 (146) and human cytidine kinases (147), are also potential markers for cell trafficking, but they require complex radiotracer synthesis, which in turn impedes their clinical translation (148).

Also worth considering is the use of reporter genes coupled with MRI. The advantages of MRI over other imaging modalities include excellent tissue resolution, and the combination of anatomical and functional information that can be garnered. Nowadays, several MRI-based reporter genes systems have been developed, such as iron carrier proteins, transferrin, tyrosinase, etc. However, current MRI modalities still lack the ability to sensitively detect reporter gene expressing cells and quantify them at relevant concentrations.

In summary, reporter gene–based imaging is preferred for cell trafficking and monitoring tissue toxicity during adoptive cell therapies. This approach also helps better evaluate the localization and functionality of the T cells, and the temporal and spatial distributions of the T cells relative to the tumor, as well as to determine a patient-specific dosing plan (Table 2).

Future Directions

To visualize the T-cell response in the context of immunotherapy, PET imaging offers most of the desired characteristics (easy quantification, high sensitivity, and availability), but it involves radiation, lacks tissue resolution, and has limited spatial resolution. Multimodal imaging approaches, such as the combination of CT/MRI with PET, provide a possible solution to improve tissue resolution. To overcome the species specificity of mAbs, nanoparticles/small molecules are increasingly being developed and also show sufficient antigen-binding abilities (11). This approach can increase cellular uptake, target specific populations of immune cells, and prolong intracellular retention, and is of interest for integrating with multimodal imaging strategies. More importantly, several preclinical imaging approaches, such as intravital fluorescence microscopy and multiphoton microscopy, are contemporaneously contributing to our knowledge of the conditions required for an effective anticancer immune response, as these technologies provide spatiotemporal insights at the molecular level and with single-cell resolution (149–153). Therefore, translation of these novel in vivo imaging approaches into the visualization of T-cell responses in patients undergoing immunotherapy is an attractive and promising strategy for further development and translation of anticancer immunotherapy.