PET tracer tracks activated T cells during STING agonist immunotherapy trial

A radioactive tracer designed to light up activated T cells on a PET scanner behaves in a remarkably predictable way, even in patients receiving an experimental immunotherapy combination, according to a new pharmacokinetic study published in EJNMMI Research. Its central message is technical but consequential: tumor uptake of the tracer is best captured by a two-tissue irreversible compartment model, and the most reliable simplified measure is a ratio normalized to tracer levels in plasma rather than whole blood.

The tracer, 2-deoxy-2-[18F]fluoro-9-b-D-arabinofuranosyl guanine, was developed by Cellsight Technologies to image activated T cells specifically. Because activated tumor-infiltrating lymphocytes are considered the driving force of anti-tumor immunity, a scan that could quantify them noninvasively would address what the authors describe as an urgent need for clinical biomarkers of immunotherapy’s mechanistic efficacy. Before [18F]F-AraG can serve that role, however, its behavior in the body must be characterized rigorously, and compartment modeling is considered the reference standard for that task. The chosen model and its macro parameters, the net influx rate Ki and the volume of distribution VT, in turn determine which simplified uptake metric, such as a standardized uptake value or a tumor-to-blood ratio, is most appropriate for routine use.

Earlier work from the same group had established the optimal pharmacokinetic model in patients with non-small cell lung cancer: an irreversible two-tissue compartment model, abbreviated 2T3K, with Ki as the preferred parameter, and tumor-to-blood and tumor-to-plasma ratios as the best simplified metrics. But tumor type and treatment can introduce biological variation that alters a tracer’s kinetics.

The imaging substudy enrolled four adults with advanced solid tumors who had exhausted standard therapy options, screened between November 2022 and April 2023. All were participants in the parent phase 1 trial of the STING agonist GSK3745417 combined with the anti-PD-1 antibody dostarlimab, a combination designed to provoke a pro-inflammatory interferon response in dendritic cells and then unleash T cells by blocking the PD-1 checkpoint. Eligibility required measurable disease by RECIST v1.1 and excluded autoimmune disease, active infections and immunosuppressive medications. Patients received GSK3745417 weekly and dostarlimab every three weeks; three patients received 0.1 mg of the agonist while the fourth received 0.2 mg after dose escalation in the parent trial.

Each patient underwent three dynamic 90-minute [18F]F-AraG PET/CT scans on a long axial field-of-view Biograph Vision Quadra system: one within a week before treatment started, one week after the first agonist dose, and six weeks after starting the agonist and five weeks after starting dostarlimab. Arterial blood samples collected during each scan measured radioactivity in whole blood and plasma and tracked [18F]F-AraG metabolites, generating the plasma input function needed for quantitative modeling. Tracer administration was well tolerated, with no adverse events reported after injection. The mean injected activity was 187.9 MBq, with a mean mass dose of 0.94 micrograms.

The modeling compared three candidate models fitted by nonlinear least squares regression, the single-tissue reversible 1T2K, the two-tissue irreversible 2T3K and the two-tissue reversible 2T4K, each with a blood volume fraction. Model selection relied on the Akaike Information Criterion, visual inspection of fit quality and the reliability of the estimated parameters. Across 55 tumor regions, the 2T3K model had the lowest AIC in 40 cases, the reversible 2T4K in 14 and the 1T2K in just one. Tumors favoring the reversible model showed a very low k4, suggesting minimal true reversibility, and the higher coefficient of variation for 2T4K-derived VT compared with 2T3K-derived Ki made the irreversible model the more stable choice overall.

Organ kinetics proved more varied. The highest tracer uptake appeared in the kidneys, liver and pancreas, with median SUVLBM values at 60 to 70 minutes post injection of 18.1, 12.3 and 8.6 respectively, while immune-related organs showed surprisingly modest signal: bone marrow at 1.4 and spleen at 1.9. Most organs favored the 2T3K model, though the brain and lungs favored 1T2K, and the bone marrow, spleen and thyroid preferred 2T4K, with the marrow and spleen showing declining uptake after 15 minutes. The authors suggest this reversible behavior may reflect dephosphorylation of the tracer’s triphosphate form by SAMHD1, which is highly expressed in bone marrow and spleen, and possibly the contribution of myeloid cells, which ex-vivo experiments showed release the tracer within an hour of uptake regardless of activation state.

The simplified metrics analysis delivered the study’s most practical finding. All tested metrics, including SUV normalized for lean body mass, body weight or body surface area, plus tumor-to-blood, tumor-to-plasma and tumor-to-parent-fraction-corrected-plasma ratios, correlated linearly with Ki. However, SUV measures and the tumor-to-blood ratio showed a patient-specific bias, with observations clustering by patient. Normalizing to plasma activity, using the tumor-to-plasma ratio, corrected this bias. The reason lies in the blood itself: roughly 85 percent of the tracer was cell-bound one hour after injection, and the plasma-to-blood ratio varied considerably between patients. When the team reanalyzed the data using an uncalibrated image-derived input function from the ascending aorta, as other groups have done, the AIC flipped to favor reversible models in most lesions, underscoring why plasma correction matters.

Importantly, no treatment-induced changes in tracer kinetics were observed. The relationship between the tumor-to-plasma ratio and Ki held across baseline, week 1 and week 6, and appeared independent of treatment. Only one of the four patients showed an objective response by RECIST criteria, and differences in preferred models between patients did not track with response or with the more immunogenic tumor types of two participants. The authors caution that the study’s primary limitation is its small sample size, which limits definitive conclusions and may have inflated observed correlations because multiple observations came from the same patients. They also note substantial physiological uptake in organs not typically considered T cell reservoirs, such as the pancreas, thyroid and adrenal glands, whose biological mechanism remains unknown and which must be considered when interpreting scans.

The team frames the work as a significant step toward qualifying [18F]F-AraG PET as a biomarker of immunotherapy efficacy, while emphasizing that studies correlating tracer uptake with actual intratumoral T cell presence are still needed. For clinical applications they recommend a simplified protocol: a static scan at 60 minutes post injection accompanied by at least one blood sample to determine the patient-specific plasma-to-blood ratio. A test-retest reproducibility study in NSCLC tumors and lymphoid tissue is already underway under a separate trial registration, and the group calls for translational work including immune profiling of blood cells and tumor biopsies to clarify the biological processes behind the observed kinetics.

Subject of Research: Pharmacokinetic modeling of the activated T cell PET tracer [18F]F-AraG in solid tumor patients undergoing STING agonist and anti-PD-1 immunotherapy

Article Title: PET pharmacokinetics of [¹⁸F]F-AraG as a marker of T cell activation during immunotherapy with a STING agonist with and without anti-PD-1

Article References: Pouw, J. E. E., Stavenga, M., Wijngaarden, J. E., Oprea-Lager, D. E., Schober, P., Schuit, R. C., van Eeden, S. P. P., Windhorst, A. D., Murillo, M. M., Galette, P., Parker, C. A., Bonasera, T. A., de Gruijl, T. D., Yaqub, M. M., Menke-van der Houven van Oordt Oord, C. W., & Boellaard, R. (2026). PET pharmacokinetics of [¹⁸F]F-AraG as a marker of T cell activation during immunotherapy with a STING agonist with and without anti-PD-1. EJNMMI Research, 16(1), Article 143. https://doi.org/10.1186/s13550-026-01503-y

Image Credits: AI Generated

DOI: 10.1186/s13550-026-01503-y

Keywords: [18F]F-AraG, PET pharmacokinetics, T cell activation, STING agonist, anti-PD-1, immunotherapy biomarker, compartment modeling, tumor-to-plasma ratio, EJNMMI Research, Amsterdam UMC, dose escalation, total-body PET