Predicting CAR-T cell immunotherapy success through immunoPET

Summary

Chimeric antigen receptor (CAR) T cell therapy has generated unprecedent advances in the treatment of hematologic cancers, but readily translatable imaging approaches to visualize the in vivo dynamics of CAR-T cells are lacking. Non-invasive positron emission tomography (PET) imaging is the ideal tool to monitor CAR-T cells.

In this issue of Clinical Cancer Research, an article by Simonetta and colleagues (dedicated to the memory of Professor Sanjiv Sam Gambhir) identified the T cell activation marker inducible COStimulator (ICOS) as a potential target for CAR-T cell monitoring in vivo. Following the identification of ICOS as an ideal target, the authors describe for the first time the use of the murine radiolabeled chelator-antibody conjugate ⁸⁹Zr-DFO-ICOS for non-invasive tracking of murine CD19–28z CAR-T cells in a syngeneic model of systemic B-cell lymphoma by immunoPET.(1)

Refractory B-cell malignancies have been remarkably responsive to CAR-T cell therapy. There are currently two FDA-approved CD19-targeted CAR-T cell candidates: tisagenlecleucel (Kymriah) and axicabtagene ciloleucel (Yescarta). However, therapeutic response to CAR-T cells has been largely exclusive to hematologic cancers, although rather unsuccessful in solid malignancies. The inability of CAR-T cells to mount a sufficient anti-tumor immune response in solid tumors has been attributed to the in vivo distribution, target engagement and function of the cell therapeutics. The work of Simonetta and colleagues provides an imaging tool likely able to evaluate this hypothesis and monitor subsequent, improved generations of CAR-T cell-based therapies.

Current approaches to imaging CAR-T cells, both in preclinical and clinical settings, are based on either direct or indirect labeling of therapeutic cells prior to infusion into patients. CAR-T cells can be directly labeled ex vivo. However, central to their mechanism of action is their expansion in vivo, which progressively dilutes the label and prevents accurate long-term monitoring of the therapy.(2) Furthermore, retention of the radiolabel in dead cells may result in a detectable signal from non-viable CAR-T cell populations, impacting the clinical interpretation of the therapeutic response.

Based on a permanent integration of a reporter gene into the CAR-T cell genome, current indirect labeling is able to overcome the limitations of ex vivo labeling. This reporter gene allows for therapeutic cells to be imaged over their entire lifespan and also allows for the imaging of their in vivo expansion. As the reporter expression exclusively occurs in live cells, the strategy provides information on CAR-T cell viability. In a recent study, Keu and colleagues demonstrate that the expression of the reporter gene HSV1-tk in CAR-T cells can be employed as a dual-purpose suicide and imaging reporter gene for the long-term non-invasive detection of CD8+ cytolytic CAR-T cells in patients with recurrent glioma.(3) However, the complexity of transgene expression protocols, the potential for immune rejection of foreign reporters and related regulatory and safety hurdles have significantly slowed clinical translation.

Simonetta and colleagues employed a more readily translatable immunoPET-based technology to monitor CAR-T cell localization and function. ImmunoPET holds several advantages over direct and indirect cell labeling strategies. Notably, it does not involve genetic manipulation, instead combining the targeting specificity of monoclonal antibodies or antibody fragments with the superior sensitivity of PET. Radiolabeling antibodies specifically targeting T cell surface activation markers allows for in vivo monitoring of dynamics, activity and functional state of administered CAR-T cells in patients. Other groups have previously employed this strategy to non-invasively image late-stage T cell activation markers including IFNɣ and granzyme β.(4) Unlike IFNɣ and granzyme β, ICOS is activated upon antigen recognition and can thereby act as a marker to detect the early-stage of T-cell-mediated immune response.(5)

ICOS is primarily expressed on activated CD4+ and CD8+ T cell subsets and its engagement with ICOS ligand is responsible for T cell proliferation, survival and release of IL4, IL10 and IFN. In a study with 31 patients treated with the commercial axicabtagene ciloleucel (KTE-C19, Axi-cel) for relapsed/refractory B-cell lymphoma, Simonetta and colleagues found that ICOS is highly expressed by CAR(+)-expressing T cells relative to non-genetically modified CAR(–) circulating T cells. The disparity in ICOS expression between CAR(+) and CAR(–) T cells indicated its suitability as a target to monitor activated CAR-T cells by ImmunoPET (Fig. 1/left). The authors found that ICOS was also upregulated both in vitro and in vivo during murine CD19.28z CAR-T activation in a murine tumor model of systemic B-cell lymphoma and thus employed this model system to evaluate the viability of ⁸⁹Zr-DFO-ICOS as a non-invasive imaging agent for the assessment of CAR-T cell therapy. To do so, the authors conjugated the anti-ICOS monoclonal antibody to the bifunctional chelator deferoxamine (DFO) and radiolabeled the immuno-conjugate with ⁸⁹Zr (Fig. 1/center). The fully immunocompetent mouse model selected for this study resembles the clinical scenario in which resident T cells are present alongside infused CAR-T cells. It also provides a more clinically relevant model to study the tracer in vivo distribution and background generated by the host non-genetically engineered T cells.

Figure 1.

By targeting T cell surface activation markers with radiolabeled antibodies, immunoPET can be used to visualize the dynamics and functional state of CAR-T cells. COStimulator (ICOS), an inducible T cell activation marker, is expressed on activated CAR-T cells upon antigen recognition and can subsequently inform the early stages of T-cell-mediated immune response (left). The proposed strategy is based on the in vivo labeling of activated T cells with anti-ICOS-DFO-⁸⁹Zr monoclonal antibody (center), enabling non-invasive long-term functional imaging of the administered therapeutic cells in vivo in order to assess therapeutic efficacy and predict response (right).

⁸⁹Zr-DFO-ICOS immunoPET was able to track activated CAR-T cells to sites rich in infiltrating B-cell lymphoma cancer cells. The tracer was also taken up in highly vascularized organs including heart, liver, and spleen. In the future this can be mitigated by using an antibody fragment-based ImmunoPET agent with a more favorable pharmacokinetic profile.

In addition, the anti-ICOS mAb used in the presented study (clone 7E.17G9) is known to block the engagement of ICOS with its natural ligand (ICOSL) and could potentially result in impaired in vivo CAR-T cell expansion, targeting, persistence and ultimately antitumor activity. The authors carefully addressed this aspect and demonstrated that the overall survival (here as a measure of CAR-T cell efficacy) is preserved in the anti-ICOS mAb-treated mice and is comparable to the isotype control.

Overall, the results demonstrate the suitability of ⁸⁹Zr-DFO-ICOS to non-invasively track antigen-specific activation of therapeutic CAR-T cells (Fig. 1/right). The proposed ICOS imaging strategy provides the unique advantage of reporting on the activation as well as the exhaustion of therapeutic CAR-T cells — both crucial for monitoring and predicting patient response to immunotherapy. The ⁸⁹Zr-DFO-ICOS mAb is applicable to all available CAR-T cell products and can be readily translated into clinical practice with prior adaptation to a fully human/humanized version of the ICOS mAb. ICOS expression in human CAR-T cells is known to be heterogenous, thus conferring the potential additional advantage of stratifying responder versus non-responder patients based on CAR-T cell in vivo activation in order to predict therapy success.