Molecular Imaging of Chimeric Antigen Receptor T Cells by ICOS-ImmunoPET

This article is featured in Highlights of This Issue, p. 903

Chimeric antigen receptor (CAR) T-cell therapy has shown impressive results in B-cell hematologic malignancies leading to the clinical approval of the first engineered cellular therapy for cancer (1, 2). In contrast to conventional pharmacologic approaches, CAR T cells are living drugs that after administration actively migrate to target tumor-infiltrated tissues, expand, and persist to exert their function efficiently. The development of new CAR T-cell–based therapies and the improvement of existing ones are greatly influenced by our ability to monitor their in vivo dynamics to gain greater insights into the success or failure of this treatment approach. The measurement of circulating CAR T cells in the peripheral blood by flow cytometry and/or PCR is the only immuno-monitoring modality currently available. Some studies have reported an association between the peak and the duration of circulating CAR T-cell levels and the clinical outcome (3–7), while other studies have failed to confirm these results (8, 9), suggesting that such a measure might not accurately reflect the actual biodistribution and persistence of the cells in the body. Such a discrepancy can be particularly relevant when CAR T cells are employed for the treatment of solid tumors for which lack of efficient migration to the tumor site is considered one of the main barriers limiting the efficacy of CAR T-cell therapy (2, 10).

Molecular imaging is an attractive strategy for noninvasive and longitudinal monitoring of CAR T-cell distribution in vivo. To date, several studies using PET imaging have been conducted because of the high sensitivity and quantitative capabilities of this clinically relevant imaging modality. The majority of the preclinical and clinical pilot studies published so far involve the use of reporter genes inserted into the CAR construct for the detection of CAR T cells (11–15). Despite the enormous potential of this indirect labeling approach, major limitations of this strategy are the requirement for the generation and approval of entirely new CAR T cells, specifically designed to incorporate the imaging functionality, and a significant risk of immunogenicity. To circumvent these limitations, an alternative approach that has been tested involves the ex vivo labeling of CAR T cells with radioactive tracers prior to administration (16–18). This approach has the advantage of being applicable to any CAR T cells, including the current commercially available ones, but has several limitations, principally limited temporal resolution, a function of the radioactive decay of the label employed. Even if radioisotopes with long half-lives are employed, imaging is limited to a relatively short time frame after administration and is further compounded by signal dilution as CAR T cells proliferate in vivo.

ImmunoPET is a rapidly expanding area of molecular imaging that employs mAbs and antibody fragments radiolabeled with PET isotopes, thereby combining the ultra-high specificity and affinity of antibodies for cell surface markers with the superior sensitivity of PET (19). Advantageously, the radiolabeled antibodies can be injected at different timepoints after administration of the cells and even repeatedly to allow longitudinal and serial assessment of CAR T-cell persistence. Target molecules evaluated to date to study T-cell immune responses principally include lineage defining molecules, such as CD3 (20, 21), CD4 (22, 23), CD7 (24), and/or CD8 (25, 26), and T-cell surface activation markers, such as OX40 (27), HLA-DR (28), and ICOS (29). Our group and others have evaluated such approaches for monitoring different classes of cancer immunotherapies in preclinical models, including tumor vaccination (27, 29), immune checkpoint blockade (29), and allogeneic hematopoietic stem cell transplantation (28, 30).

In this work, we identified Inducible T-cell COStimulator (ICOS or CD278), a costimulatory molecule upregulated during T-cell activation (31), as a candidate target for CAR T-cell monitoring using immunoPET. Moreover, we assessed the utility of ICOS-targeted immunoPET to monitor CD19-specific CAR T-cell activation, expansion, and homing to target tumor-infiltrated tissues in a murine model of B-cell lymphoma.

We analyzed an RNA sequencing (RNA-seq) dataset we recently published (32) and deposited in NCBI Gene Expression Omnibus, series accession number, GSE136891. RNA-seq data were obtained at serial timepoints from CD19.CD28z CAR T cells generated from naïve (CD45RA⁺, CD45RO⁻, CD62L⁺, CCR7⁺, CD95⁻, and CD122⁻) CD4⁺ or CD8⁺ T cells as detailed in ref. 32.

We retrospectively analyzed ex vivo ICOS expression on human CD19-28z CAR T cells in prospectively collected data obtained from 31 patients receiving commercial axicabtagene ciloleucel at Stanford University (Stanford, CA) for relapsed or refractory diffuse large B-cell lymphoma (DLBCL). Mass cytometry analysis was performed as described previously (33). The mass cytometry panel assessed expression of 33 surface or intracellular proteins relevant to T-cell function in blood collected on day 7 (peak expansion) and on day 21 (late expansion) post–CAR T-cell infusion (33). ICOS expression was evaluated on live CAR-expressing T cells identified by anti-idiotype antibody provided by Dr. Laurence Cooper (University of Texas MD Anderson Cancer Center, Houston, TX; ref. 34). Written informed consent was provided by all patients enrolled, and the study was approved by the Stanford University Institutional Review Board (Stanford, CA). Experimental procedures were carried out in accordance with the ethical principles of the Declaration of Helsinki.

BALB/cJ mice were purchased from the Jackson Laboratory. Firefly luciferase (Luc)⁺ transgenic BALB/c mice have been reported previously (35) and were bred in our animal facility at Stanford University (Stanford, CA). All procedures were performed on sex-matched animals between 8 and 12 weeks of age and approved by Stanford University's Administrative Panel for Laboratory Animal Care/Institutional Animal Care and Use Committee in compliance with the guidelines of humane care of laboratory animals.

CAR T cells specifically recognizing the murine CD19 (mCD19) molecule and including a CD28 costimulatory domain were generated as reported previously (36, 37). mCD19 CAR stable producer cell line (37) was kindly provided by Dr. Terry J. Fry (University of Colorado School of Medicine, Aurora, CO). T cells were enriched from BALB/c mouse spleen single-cell suspensions using the mouse Pan T Cell Isolation Kit II (Miltenyi Biotec) according to the manufacturer's protocol. T cells were activated for 24 hours with Dynabeads Mouse T-Activator CD3/CD28 (Life Technologies) in the presence of human IL2 (30 U/mL) and murine IL7 (10 ng/mL; PeproTech) in RPMI1640 media supplemented with 10% heat-inactivated FBS, 1 mmol/L sodium pyruvate, 2 mmol/L glutamine, 100 U/mL penicillin, and 100 μg/mL streptomycin at 37°C with 5% CO₂. Activated T cells were then transduced with mCD19 CAR by culturing them for 48 hours in RetroNectin-coated plates loaded with supernatant harvested from the stable producer line 48 hours after culture. Dynabeads were then removed, and after washing, cells were rested for 24 hours in fresh medium containing IL2 and IL7 before use. Transduction efficiency was measured by flow cytometry after protein L staining (38). Cell numbers were adjusted on the basis of transduction efficacy (50% on average) before in vitro or in vivo use.

Murine CD19.28z (mCD19.28z) CAR T cells were cocultured with luciferase-transduced A20 cells (A20^luc; ref. 39) at different ratios adjusted on the basis of transduction efficiency in culture medium consisting of RPMI1640, supplemented with l-glutamine (2 mmol/L), penicillin (100 U/mL), streptomycin (0.1 mg/mL), 2-mercaptoethanol (5 × 10e-5 mol/L), and 10% FBS. After 24 hours of culture, d-luciferin (PerkinElmer) was added at 5 μg/mL and incubated for 5 minutes at room temperature before imaging using an IVIS Spectrum Imaging System (PerkinElmer). For ICOS expression analysis, CAR T cells cultured with A20^Luc+ at 1:1 ratio or unstimulated controls were harvested and analyzed by flow cytometry.

We employed a systemic B-cell lymphoma mouse model that we have reported previously, in which tumor cells infiltrate secondary lymphoid organs and bone marrow (BM; ref. 39). Briefly, CD19-expressing A20 cells were purchased from ATCC in 2017, expanded for three passages, and cryopreserved until use. A20 cells were thawed, cultured for maximum a week, resuspended in PBS, and injected intravenously (2.5 × 10e5 cells/mouse) by tail vein into sublethally (4.4 Gy) irradiated Thy1.2⁺ BALB/c recipient mice. Seven days after tumor injection, Luc⁺ CAR T cells (1 × 10e6 transduced cells/mouse adjusted on transduction efficiency) or equivalent numbers of in vitro expanded untransduced Luc⁺ T cells were administered by retro-orbital intravenous injection. In tumor homing experiments, A20^Luc+ cells were employed and mice were left untreated.

Single-cell suspensions were prepared from spleen and BM in PBS containing 2% FBS. Extracellular staining was preceded by incubation with purified FC Blocking Reagent (Miltenyi Biotec). Cells were stained with FITC anti-CD45.1 (clone A20), BV785 anti-ICOS (clone C398.4A) or appropriate isotype control (clone HTK888), APC anti-Thy1.1 (clone OX-7), APC/Fire750 anti-CD19 (clone 6D5), BV421 anti-CD4 (clone GK1.5), BV605 anti-CD3 (clone 17A2), and BV650 anti-CD8 (clone 53-6.7). Dead cells were excluded using Fixable Viability Dye eFluor 506 (eBioscience). All antibodies were purchased from BioLegend. Samples were acquired on a BD LSR II Flow Cytometer (BD Biosciences), and analysis was performed with FlowJo 10.5.0 Software (Tree Star).

For in vivo bioluminescence imaging (BLI), mice were injected with d-luciferin (10 mg/kg, i.p.) and anesthetized with 2% isoflurane in oxygen. Imaging was conducted using an IVIS Spectrum Imaging System (PerkinElmer) and data were analyzed with Living Image Software version 4.1 (PerkinElmer), or imaging was conducted using an Ami Imaging System (Spectral Instruments Imaging) and data were analyzed with Aura Software (Spectral Instruments Imaging).

ICOS-targeted mAb (ICOS mAb, clone:7E.17G9, BioXCell) was modified with the bifunctional chelator, deferoxamine (DFO/p-SCN-Bn-Deferoxamine, Macrocyclics). Briefly, unconjugated mAb was prepared at 1 mg/mL in PBS (pH = 7.4), and buffer exchanged with PBS solution adjusted to pH 8.8–9.0 using 1 mol/L Na₂CO₃. Following this, 10-fold molar excess DFO was added to the ICOS mAb solution and the conjugation was allowed to proceed for 1-hour at 37°C. The mixture was subsequently thoroughly buffer exchanged using PBS (pH = 7.4) to remove unreacted DFO using a 2 mL vivaspin centrifugal concentrator with a 50K cutoff (Sartorius). The concentration of the final DFO-ICOS mAb conjugate was determined by Thermo Scientific NanoDrop One Microvolume UV-Vis Spectrophotometer.

For ⁸⁹Zr labeling, 37 MBq (∼1 mCi) of ⁸⁹Zr-oxalate (3D imaging) was diluted in 0.5 mL of HEPES buffer (0.5 mol/L) to ensure a pH range of 7.0–7.5 followed by the addition of approximately 166 μg of DFO-ICOS mAb. Radiolabeling was carried out for 1 hour at 37°C with shaking, after which the radiolabeled antibody, ⁸⁹Zr-DFO-ICOS mAb, was purified using a 7K MW cut-off Zeba Spin Desalting Column (Thermo Fisher Scientific) centrifuged for 1 minute at 1,000 × g. The radiochemical yield and purity were determined via instant TLC (iTLC). Samples (2 μL) of the radiolabeling reaction and purified radiolabeled antibody were spotted into silica-impregnated radio iTLC plates, run with 50 mmol/L EDTA (pH = 4.5), and developed in a Phosphor-plate Reader (PerkinElmer). Radiolabeled antibody remained at the origin (R_f = 0), while free ⁸⁹Zr moved with the solvent front (R_f = 1). A final radiochemical purity of 99% was achieved, along with a final specific activity of 6 μCi/μg. The final formulation was prepared in PBS.

Mice were anesthetized using isoflurane (2%–2.5% for induction and 1.5%–2% for maintenance) delivered by 100% oxygen. ⁸⁹Zr-DFO-ICOS mAb (45 μCi ± 3.6 μCi, 7.5 μg ± 0.6 μg) was administered intravenously in untreated and Luc⁺ untransduced T-cell or CAR T-cell–treated A20 tumor–bearing mice. Radiotracer administration corresponded to day 5 post–CAR T-cell injection. At 24 and 48 hours after ⁸⁹Zr-DFO-ICOS mAb tracer administration, mice underwent imaging with static 20-minute PET scans followed by 10-minute transmission scans for PET attenuation correction on the Inveon Dedicated PET Scanner (DPET, Siemens), and CT scans on the GNEXT (Sofie Biosciences) for anatomic coregistration of the PET data. PET image reconstruction and image analysis were conducted as described previously (27).

Following the completion of the scans 48 hours after injection of the tracer, mice were euthanized and ex vivo biodistribution studies were performed to measure tissue-specific radioactivity, and further corroborate PET findings. Briefly, we collected blood (∼100 μL) via cardiac puncture, as well as the following tissues: heart, spleen, kidney, liver, muscle, femur, tibia, iliac bone, and lumbar spine. Harvested hearts were rinsed in clean water after an incision was made to remove residual blood, and then gently dabbed to remove excess moisture. Tissues were placed in a tube, weighed, and radioactivity was measured using an Automated Gamma Counter (Hidex AMG Automatic Gamma Counter). Tissue-associated radioactivity was normalized to tissue weight and amount of radioactivity administered to each mouse, and decay-corrected to the time of radiotracer injection. Data were expressed as percentage of injected dose per gram of tissue (%ID/g) values.

Murine-specific ICOS mAb (clone 7E.17G9; BioXCell) or isotype control (clone LTF-2; BioXCell) was administered intravenously as a single dose on day 5 after CAR T-cell administration (10 μg in 100 μL of PBS). The dose was determined on the basis of the upper limit of antibody administered during PET imaging studies.

Student two-tailed t test, the Mann–Whitney U test, or the Wilcoxon signed-rank test was used as appropriate to determine statistical significance. BLI data (photons/seconds) are displayed as mean and SD for each group over time and were analyzed by two-way ANOVA with Bonferroni correction. Survival curves were represented with the Kaplan–Meier method and compared by log-rank test. Statistical analyses were performed using Prism 8 (GraphPad Software) and R version 3.5.1 (Comprehensive R Archive Network project (http://cran.us.r-project.org) with R studio version 1.1.453.

In an effort to identify surface activation markers as candidate targets for immunoPET, we analyzed recently published RNA sequencing data obtained from human CD4⁺ and CD8⁺ CD19.28z CAR T cells cultured in vitro (32). The analysis was restricted to a selected list of markers known to be upregulated during activation on the T-cell surface and previously employed as targets for PET imaging (27, 29, 40, 41). As shown in Fig. 1A, transcripts of most of the selected markers were expressed at high levels early after activation (day 7) and decreased at later timepoints (days 10 and 14). ICOS transcription exhibited a different trend, progressively increasing and being sustained at later timepoints in both CD4⁺ and CD8⁺ CAR T cells (Fig. 1A). To further validate ICOS as a candidate target for immunoPET, we next assessed its ex vivo expression on CD19.28z CAR T cells isolated from 31 patients receiving commercial axicabtagene ciloleucel (KTE-C19) for relapsed/refractory B-cell lymphoma, at days 7 and 21 after therapy administration. CAR-expressing T cells identified using specific anti-idiotype staining were compared with circulating T cells not expressing the CAR. ICOS was expressed at significantly higher levels on CAR-expressing (CAR⁺) T cells compared with CAR-nonexpressing (CAR⁻) T cells (Fig. 1B). This difference was observed at day 7, corresponding to the peak of CAR T-cell expansion (median intensity ± SEM: CAR⁺, 26.6 ± 4.7 and CAR⁻, 9.6 ± 0.9; P = 0.00002), and further maintained at day 21 (CAR⁺, 16.3 ± 3.2 and CAR⁻, 9.9 ± 1.0; P = 0.00026; Fig. 1C). In accordance with the RNA-seq data (Fig. 1A), we detected higher ICOS expression on both CD4⁺ and CD8⁺ CAR T cells (Supplementary Fig. S1). Collectively, these results identify ICOS as a promising target for CAR T-cell monitoring by immunoPET.

To test the utility of ICOS-immunoPET for in vivo monitoring of CAR T cells in a preclinical mouse model, we went on to assess ICOS expression during mCD19.28z CAR T-cell activation. First, we analyzed ICOS expression during a 24-hour cytotoxic assay incubating mCD19.28z CAR T cells together with the Luc⁺ CD19-expressing murine B-cell lymphoma cell line A20. As expected, mCD19.28z CAR T cells exerted a dose-dependent cytotoxic effect against A20 cells (Fig. 2A). ICOS expression, as assessed by flow cytometry, was significantly upregulated on murine CAR T cells exposed to A20 target cells (1:1 CAR to target ratio) compared with CAR T cells alone (Fig. 2B). We next assessed ICOS expression on mCD19.28z CAR T cells activated in vivo. To better mimic the clinical scenario, we employed a systemic tumor model in which A20 lymphoma cells were injected intravenously (Fig. 2C). In agreement with what we have reported previously (39), by the time of CAR T-cell administration, 7 days after tumor inoculation, A20^Luc+ cells were infiltrating the liver, the lymphoid organs, and the BM (Supplementary Fig. S2). Flow cytometry analysis was performed at day 5 after administration of Thy1.1⁺ CAR T cells into Thy1.2⁺ recipients, a timepoint when CAR T cells had expanded and were detectable within the spleen and the BM, as confirmed by in vivo bioluminescence (Fig. 2D). ICOS expression was significantly higher at the surface of administered Thy1.1⁺ infused cells when compared with endogenous Thy1.1⁻ cells both in the spleen (Fig. 2E) and in the BM (Fig. 2F). To further assess the specificity of ICOS upregulation for CAR-mediated activation of CAR T cells and distinguish it from activation resulting from the lymphopenic context, we used ex vivo expanded untransduced T cells as an additional control. As shown in Supplementary Fig. S2B and S2C, lower percentages and absolute numbers of untransduced T cells were detected within the BM, suggesting preferential homing and/or expansion of CAR T cells in the tumor-infiltrated BM. Phenotypically, BM-infiltrating CAR T cells expressed significantly higher levels of ICOS compared with untransduced T cells (Supplementary Fig. S2C), confirming the specificity of ICOS expression for activated CAR T cells. Collectively, these results confirm that ICOS is specifically expressed on murine CAR T-cell surface during ex vivo and in vivo activation, supporting its potential utility as a biomarker for CAR T monitoring by immunoPET.

We next tested the ability of ICOS-targeted immunoPET to visualize CAR T-cell migration, activation, and expansion during an antitumor response in vivo using ⁸⁹Zr-DFO-ICOS mAb. Employing the aforementioned murine tumor model, ⁸⁹Zr-DFO-ICOS mAb was intravenously injected at day 5 after T-cell administration and PET/CT images were acquired at 24 and 48 hours after tracer injection. The 48-hour timepoint was selected as the one providing superior signal-to-noise ratio. Figure 3A and B show representative volume-rendered technique (VRT) PET/CT images of tumor-bearing mice that either received no treatment (not treated, left), untransduced T cells (middle), or mCD19.28z CAR T cells (CAR T, right). PET/CT images of untreated mice showed that ⁸⁹Zr-DFO-ICOS mAb primarily accumulated in the heart, spleen, and liver, but not in bones, consistent with the biodistribution and clearance of intact antibodies (Fig. 3A and B, left; Supplementary Fig. S3). Mice receiving untransduced T cells, as well as CAR T cells, displayed a tracer biodistribution in highly vascularized organs (heart, liver, and spleen) similar to mice that were not treated. Importantly, we detected significantly higher ⁸⁹Zr-DFO-ICOS mAb PET signals in the bones of CAR T-cell–treated mice compared with both untreated mice and mice treated with untransduced T cells [Fig. 3A and B, right; two-dimensional maximum intensity projection (MIP) images in Supplementary Fig. S3]. The PET signal in the CAR T-cell–treated mice was particularly prominent in the lumbar spine, iliac bones, femur, tibia, and humeral heads (Fig. 3A and B, right; MIP images in Supplementary Fig. S3).

To quantify radiotracer accumulation in specific tissues and corroborate the trends observed in the PET images, we conducted a region of interest (ROI) analysis on multiple tissues guided by CT. We detected a slight, but statistically significant increase in radiotracer accumulation in the heart of CAR T-cell–treated mice compared with mice that were not treated (mean ± SEM, 11.4 ± 0.45 %ID/g vs. 9.1 ± 0.9 %ID/g; P < 0.05; Fig. 3C), while no significant differences were detected when compared with mice receiving untransduced T cells (Fig. 3C). No significant differences between CAR T-cell–treated mice and untreated mice were detected in spleen and liver (Fig. 3C). ROI quantification of PET/CT images confirmed markedly increased radiotracer uptake in bones rich in BM from CAR T-cell–treated mice compared with those of both untreated mice (lumbar vertebrae, P = 0.001; iliac bones, P = 0.0013; femur, P = 0.0016; and tibia, P = 0.0016; n = 8–9 per group; Fig. 3C) and mice receiving untransduced T cells (lumbar spine vertebrae, P = 0.014; iliac bones, P = 0.0012; femur, P = 0.0078; and tibia, P = 0.0056; n = 9 per group; Fig. 3C). As expected, there was no significant signal difference in the muscle, considered background, among the groups (Fig. 3C).

To corroborate the PET results, we performed a biodistribution analysis (BioD) using ex vivo gamma counting of the different tissues following the 48 hours PET/CT acquisition. BioD analysis confirmed higher radiotracer levels in the heart of CAR T-cell recipients compared with untreated mice (P = 0.026), but not with mice receiving untransduced T cells (Fig. 4). No difference among the groups was detected in blood, liver, and kidney (Fig. 4). In corroboration with the higher ICOS levels detected by FACS, BioD analysis revealed significantly higher tracer uptake in spleens from CAR T-cell–treated mice compared with the untreated group (P = 0.0033) and the untransduced T-cell group (P = 0.024; Fig. 4). When analyzing the bones rich in BM, we confirmed significantly higher tracer uptake in bones from CAR T-cell–treated mice compared with mice not treated (lumbar vertebrae, P = 0.0015; iliac bones, P = 0.007; femur, P = 0.00063; and tibia, P = 0.00062) or treated with untransduced T cells (lumbar spine vertebrae, P = 0.0036; iliac bones, P = 0.019; femur, P = 0.00016; and tibia, P = 0.00049; Fig. 4). No difference was observed in muscle. Collectively, these results demonstrate the ability to specifically track the presence of CAR T cells at target tissues during antitumor responses in vivo using ⁸⁹Zr-DFO-ICOS mAb immunoPET.

The use of costimulatory molecules as a target for immunoPET bears the theoretical risk of impacting CAR T-cell expansion, persistence, and ultimately in vivo antitumor activity, especially considering that the clone we employed (7E.17G9) blocks the interaction between ICOS and its ligand, ICOSL. To assess the impact of the administration of anti-ICOS mAbs on CAR T-cell homeostasis and function, we injected doses comparable with the maximum PET dose of anti-ICOS (clone 7E.17G9) or of the appropriate isotype control, at day 5 after CAR T-cell administration. The BLI signals derived from expanding Luc⁺ CAR T cells were not significantly different between the anti-ICOS (Fig. 5A, top) and the isotype control groups (Fig. 5A, bottom and B). Similarly, the contraction curve of CAR T cells at later timepoints (days 10–14) did not differ in mice treated with anti-ICOS compared with mice receiving isotype control (Fig. 5A and B). Using overall survival as a measure of CAR T-cell efficacy, we did not observe any significant impact of anti-ICOS administration on the survival of untreated and CAR T-cell–treated mice when compared with mice receiving isotype control (Fig. 5C). Collectively, these results indicate that the administration of anti-ICOS mAb at tracer doses does not affect CAR T-cell homeostasis or antitumor activity.

In this study, we identified ICOS-immunoPET as a promising strategy to noninvasively monitor CAR T cells in vivo during an antitumor response.

In vivo monitoring of CAR T-cell therapy using molecular imaging is an area of intensive investigation. Several groups, including ours, have explored the utility of reporter genes (11–15) or ex vivo radiolabeling approaches (16–18) to allow tracking of CAR T-cell therapies with PET. The use of immunoPET has several advantages over these approaches. First, this strategy can be applied to any CAR T-cell product, bypassing the need to insert a dedicated imaging reporter gene. The feasibility of this approach is a significant advantage considering the time and efforts required for approval of new cellular products. Second, the administration of radiolabeled antibodies at different timepoints after CAR T-cell administration, allows for serial imaging of CAR T-cell dynamics in vivo. This latter aspect might be essential in the clinical setting, where detection of CAR T-cell presence and activation by immunoPET might be employed to guide therapeutic decisions.

A major challenge in immunoPET imaging of adoptively transferred cells comes from the difficulties in identifying appropriate candidate targets. An ideal target should be selectively expressed on the transferred population with minimal expression on endogenous cells. In this regard, targeting surface activation markers (27–29) displays a major advantage over targeting lineage-defining markers (20–26). In this article, using an unbiased approach, we first identified ICOS as a potential immunoPET target on human CAR T cells before transitioning to a murine model for the preclinical assessment of our strategy. Notably, ICOS-immunoPET enabled in vivo tracking of murine CAR T cells injected into immunocompetent mice bearing systemic tumors. We believe that such a model more accurately reflects the clinical scenario compared with widely employed xenogeneic tumor models, in which human CAR T cells are injected into genetically lymphopenic mice. First, it allows the evaluation of the tracer in the presence of an endogenous resident population of T cells. Second, in xenogeneic models, the injected cells are the only targets of antibodies recognizing human molecules in a murine system thus, artificially reducing the background noise coming from the endogenous population.

Another critical aspect that needs to be taken into account when developing new immuno-imaging approaches is the risk of interfering with CAR T-cell biology. The addition of reporter genes encoding foreign proteins carries a significant risk of immunogenicity (42) that can potentially limit the therapeutic potential of the CAR T cells, a risk that cannot efficiently be evaluated in preclinical xenogeneic mouse models. Moreover, the addition of extra genes to the construct might significantly interfere with CAR expression and/or function. In the case of immunoPET, antibodies may exert biological effects after ligation of their target and, depending on the blocking, agonist, or antagonist nature of the clone employed, they might interfere with CAR T-cell homeostasis and function. In our animal model, we show that the administration of tracer doses of a blocking anti-ICOS does not interfere with CAR T-cell persistence and antitumor effect, suggesting the safety of our approach. Although we cannot exclude any minor transient effects, overall there was no detectable impact on CAR T-cell expansion (as measured by BLI) or function (as assessed by the outcome analysis). Such results will need to be confirmed on human CAR T cells during clinical translation using a fully human/humanized version of the ICOS mAb.

Our study demonstrated the feasibility of utilizing immunoPET with anti-ICOS to track CAR T-cell migration into tumor-bearing mice. The early timepoint studied demonstrates the ability of the CAR T cells to migrate to their target tissue, the BM bearing the CAR-targeted CD19 antigen on infiltrating B-cell lymphoma cells, as well as on endogenous B-cell precursors. This approach is directly applicable to the clinic, yet has several limitations. First, ICOS-immunoPET was affected by a substantial nonspecific background coming from highly vascularized organs containing high amounts of blood such as heart, liver, and spleen. Such background limited the sensitivity of the assays in tumor-containing organs such as the spleen, where a significant difference between CAR T-cell–treated mice and untreated mice was detectable only in ex vivo biodistribution studies, but not by in vivo PET imaging. Such limitation could be circumvented by the use of alternative vector formats that are smaller in size and exhibit rapid pharmacokinetics and are more suitable for imaging, for example, antibody fragments, including minibodies, diabodies, single-chain variable region fragments, and nanobodies, or engineered protein scaffolds (reviewed in refs. 43, 44). Second, the intensity of ICOS expression on human CAR T cells was highly heterogeneous (Fig. 1C), suggesting that the CAR T-cell activation status and, therefore, the sensitivity of our approach might vary between individuals. Such heterogeneity might actually confer additional power to the ICOS-immunoPET strategy, allowing the stratification of patients based on a signal resulting from both CAR T-cell expansion and in vivo activation. Moreover, our analysis of human CAR T-cell phenotype was only performed on circulating CAR T cells recovered from the peripheral blood and might not reflect ICOS expression on CAR T cells at the tumor site. This may be a critical factor especially for CAR T-cell imaging in solid tumors, where the tumor microenvironment might interfere with cell activation. Finally, our preclinical study was not designed to assess the predictive potential of CAR T-cell monitoring by ICOS-immunoPET on animal survival given the high rates of tumor control in our mouse tumor model and the fact that all mice received the exact same number of CAR T cells. Clinical studies would be more appropriate to assess the relationship of site-specific tracer uptake with patient outcomes.

In conclusion, we describe for the first time, in vivo monitoring of CAR T-cell dynamics using immunoPET targeting an endogenous biomarker, a molecular imaging approach that does not require the addition of reporter genes or ex vivo labeling and that is, therefore, potentially applicable to the clinical setting for the study of any commercially available and investigational CAR T-cell products.

F. Simonetta reports grants from American Society for Blood and Marrow Transplantation, Geneva University Hospitals Fellowship, Swiss Cancer League, Fondation de Bienfaisance Valeria Rossi di Montelera, and Dubois-Ferrière-Dinu-Lipatti Foundation during the conduct of the study. Z. Good reports grants from Parker Institute for Cancer Immunotherapy and Stanford Cancer Institute during the conduct of the study, and personal fees from Mubadala Ventures outside the submitted work. D.B. Miklos reports grants from NCI (P01 CA049605, R.S. Negrin, principal investigator) and personal fees and nonfinancial support from Kite-Gilead during the conduct of the study. C.L. Mackall reports grants from NIH during the conduct of the study. C.L. Mackall also reports grants from Bluebird and Obsidian, as well as personal fees from Lyell Immunopharma, Nektar, Neoimmune Tech, Apricity Bio, Roche, BryoLogyx, Unum, Vor, and PACT outside the submitted work; Dr. Mackall also has numerous patents related to CAR T cells pending, issued, and licensed and a patent for CD22-CAR pending, issued, licensed, and with royalties paid from Opus/Juno/Celgene. S.S. Gambhir reports grants from Ben & Catherine Ivy Foundation, Canary Foundation, NIH, and Parker Institute for Cancer Immunotherapy during the conduct of the study. R.S. Negrin reports grants from NIH and Parker Institute for Cancer Immunotherapy during the conduct of the study. No disclosures were reported by the other authors.

F. Simonetta: Conceptualization, data curation, formal analysis, investigation, visualization, methodology, writing-original draft. I.S. Alam: Conceptualization, data curation, formal analysis, investigation, visualization, methodology, writing-original draft. J.K. Lohmeyer: Investigation, methodology, writing-review and editing. B. Sahaf: Resources, data curation, investigation, methodology. Z. Good: Data curation, software, formal analysis, visualization, methodology. W. Chen: Investigation, methodology. Z. Xiao: Investigation, methodology. T. Hirai: Investigation, methodology, writing-review and editing. L. Scheller: Investigation, methodology. P. Engels: Investigation, methodology. O. Vermesh: Formal analysis, visualization, writing-review and editing. E. Robinson: Investigation, methodology. T. Haywood: Investigation, methodology. A. Sathirachinda: Investigation, methodology. J. Baker: Investigation, methodology. M.B. Malipatlolla: Investigation, methodology. L.M. Schultz: Investigation, methodology. J.Y. Spiegel: Data curation, investigation. J.T. Lee: Data curation, software, methodology, writing-review and editing. D.B. Miklos: Resources, funding acquisition, methodology. C.L. Mackall: Resources, supervision, funding acquisition, methodology, writing-review and editing. S.S. Gambhir: Conceptualization, resources, supervision, funding acquisition, project administration, writing-review and editing. R.S. Negrin: Conceptualization, resources, supervision, funding acquisition, project administration, writing-review and editing.

The authors would like to acknowledge the Stanford Center for Innovation in In-Vivo Imaging (SCi³) and in particular, Dr. Frezghi Habte for supporting the preclinical imaging experiments. We are grateful to Rowaid Kellow and Drs. Bin Shen, Gayatri Gowrishankar, and Surya Murty for supporting radiochemistry and in vivo imaging studies. We also thank the Stanford shared FACS facility for their support. We would like to dedicate this work to the memory and legacy of Professor S.S. Gambhir. This work was supported, in part, by funding from the Ben & Catherine Ivy Foundation (to S.S. Gambhir), the Canary Foundation (to S.S. Gambhir), NCI R01 1 CA201719-02 (to S.S. Gambhir), R01 CA23158201 (to R.S. Negrin), P01 CA49605 (to S.S. Gambhir and R.S. Negrin), P5P30CA124435 (to C.L. Mackall), the Parker Institute for Cancer Immunotherapy (to S.S. Gambhir, R.S. Negrin, C.L. Mackall, and Z. Good), the Geneva University Hospitals Fellowship (to F. Simonetta), the Swiss Cancer League (BIL KLS 3806-02-2016 to F. Simonetta), the Fondation de Bienfaisance Valeria Rossi di Montelera (Eugenio Litta Fellowship to F. Simonetta), the American Society for Blood and Marrow Transplantation (New Investigator Award 2018 to F. Simonetta), the Dubois-Ferrière-Dinu-Lipatti Foundation (to F. Simonetta), the Virginia and D.K. Ludwig Fund for Cancer Research (to C.L. Mackall), and a St Baldrick's/Stand Up 2 Cancer Pediatric Dream Team Translational Cancer Research Grant (to C.L. Mackall). Stand Up 2 Cancer is a program of the Entertainment Industry Foundation administered by the American Association for Cancer Research. C.L. Mackall is a member of the Parker Institute for Cancer Immunotherapy, which supports the Stanford University Cancer Immunotherapy Program.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.