Application of imaging toward cell-based immunotherapy

Conventional imaging methodologies routinely used in clinical practice focus on assessment of disease response by providing anatomic (e.g., computed tomography (CT), MRI) or metabolic (e.g., [¹⁸F]-fluorodeoxyglucose (FDG)-PET, functional MRI) information of the tumor but fail to give a complete representation of the complex spectrum of responses to immunotherapy. As the latter is often accompanied by an initial ‘pseudoprogression’, understanding the distinctive features of tumors treated with immunotherapy and the fate of these ‘living drugs’ is paramount to reliably predict responses.

Here we report seminal studies using molecular imaging for the non-invasive preclinical and clinical imaging (tables 1 and 2) of cellular immunotherapies and immune biomarkers, searched using “immunotherapy”, “cancer”, and “imaging” key words and by literature search on PubMed.

Imaging T cell trafficking and persistence

Current clinical trials involving infusion of cell-based immunotherapies are performed without knowledge of their whole-body distribution during treatment. Cell-based therapeutics can potentially redistribute and lead to so-called ‘on-target off-site’ toxicity, causing severe side effects or even death. Non-invasive long-term imaging can be used for the monitoring of their spatiotemporal in vivo distribution, persistence at the tumor site, survival and in situ expansion.

Cytolytic T cells (CTLs). Nuclear imaging with FDA-approved ¹¹¹In-oxine clinical probe16 and ⁸⁹Zr-oxine15 was employed to track the fate of directly labeled CTLs. A good manufacturing practise (GMP)-compatible production of ⁸⁹Zr-oxine and white blood cell radiolabeling has been described and could accelerate its application in in-human cell imaging studies.17 Indirect strategies for in vivo tracking of T cells involve the use of BLI (overviews from Costa et al18 and Rabinovich et al)19 or nuclear-based (overview from Volpe et al) reporters.14 Koehne et al demonstrated for the first time the feasibility of long-term imaging of CTLs by HSV1-tk/[¹³¹I]FIAU- and [¹²⁴I]FIAU-PET.20 As the interplay between CD4+and CD8+ T cells has a significant impact on the kinetics, persistence and antitumor activity of adoptive cell therapies,21 a reporter-based method was developed preclinically to independently track the CD4+ and CD8+ T cells by human norepinephrine transporter (hNET)-afforded -SPECT and HSV1-tk-afforded-PET, respectively (figure 1E).22 Notably, a human carcinoembryonic antigen (hCEA) reporter could also be used for the molecular imaging of adoptive T cell therapies, provided that the target tissue is not expressing any hCEA.23

Chimeric antigen receptor (CAR)-T cells. Direct labeling protocols using ⁸⁹Zr-oxine24 and ⁸⁹Zr-DFO25 have been developed for the in vivo tracking of CAR-T cells. Many host, foreign and synthetic reporter genes were also employed and provided a greater advantage for the repeated monitoring of CAR-T cells by either optical, ultrasound, nuclear imaging or hybrid. Moroz et al preclinically demonstrated the superiority of hNET/[¹⁸F]MFBG reporter/probe combination for the repeated imaging of T cells (sensitivity of<1×10⁵ T cells/cm³).26 CAR-T detection sensitivity was as high as 3000 cells with sodium iodide symporter (NIS)-based [¹⁸F]BF₄- PET (figure 1B) and revealed the inverse correlation between CAR-T tumor retention and PD-L1 expression in triple-negative breast cancer models.27 tPSMA^N9Del-based [¹⁸F]DCFPyL PET enabled the reliable detection of 2000 CAR-T cells and the visualization of CAR-T infiltration into local and metastatic lesions in a model of acute lymphoblastic leukemia.28 A novel reporter, eDHFR, derived from E. Coli allowed for successful visualization of GD2-specific CAR-T cells targeting human colorectal xenografts using [¹⁸F]TMP-afforded PET.29 A novel DOTA antibody reporter 1 (DAbR1) irreversibly binding to the lanthanoid (S)−2-(4-acrylamidobenzyl)-DOTA (AABD) was tested in xenografts and provided information on the in vivo distribution and homing of systemically administered CD19-directed CAR-T cells at the tumor using ⁸⁶Y- and ¹⁷⁷Lu-AABD.30 The first documented in-human study by Keu et al31 showed the trafficking of CD8+GRm13Z40 (interleukin-13 zetakine receptor alpha 2)-targeted CAR-T cells previously infused into the medial left frontal lobe of patients with high grade glioma, providing insights on their successful homing at the tumor by HSV1-tk /[¹⁸F]FHBG PET (figure 2A; NCT01082926).31

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Clinical application of direct and indirect molecular imaging in cell-based Immunotherapy. (A) Non-invasive reporter gene imaging of IL-13 zetakine-directed and HSV1-tk-expressing CAR-T cells in a 66-year-old patient with recurrent frontoparietal glioblastoma. [¹⁸F]FHBG PET-CT revealed increased uptake after CAR-T cells infusion 1-week post infusion. Tumor extent was assessed by T₁W MRI (top row) before and after CAR-T intratumoral infusion. Images were superimposed with [¹⁸F]FHBG PET (bottom row).31 (B) Image of an healthy volunteer showing ¹⁸F]F-AraG tracer in vivo distribution 60 min post intravenous administration of 189.07 MBq.99 The tracer exhibits hepatobiliary and renal clearance. shown are transversal PET (top) and fused PET-CT (bottom). (C) [¹⁸F]-FLT PET-CT imaging revealed T cell splenic proliferation in a patient with metastatic melanoma undergoing a durable regression post anti-CTLA-4 (tremelimumab) treatment.54 Baseline scan (left) was compared with a 3-month follow-up scan (right). (D) ⁸⁹Zr-IAB22M2C PET-CT of CD8+T cells performed in a patient with hepatocellular carcinoma on immunotherapy for 12 weeks prior to imaging.39 40 Zr-IAB22M2C-positive lesions in the PET-CT merge (left image) correspond to two liver metastases (left side), spleen (right side) and bone marrow (middle). Additional uptake was seen also in three abdominal lymph node metastases. CT alone is in right image. Images acquired 24 hours post tracer injection. (E) [¹⁸F]FLT PET-CT visualizes immune responses in four LNs (see arrows) 3 days after intranodal delivery of [¹¹¹In]/SPIO-labeled and antigen-loaded DCs and 1 hour after tracer injection.62 In fact, tracer retention was observed in the injected ln and the three draining LNs. (F) In vivo migration of ¹¹¹In]/SPIO-labeled dendritic cells from LN1 (site of injection; arrow 1/e) to other LNs (arrows 5/g) using MRI.63 Open arrow in left image (E) indicates empty ln (NO SPIO); closed black arrow in right image (G) indicates ln positive for SPIO. Images taken 2 days post intranodal injection of DCs. (G) ⁸⁹Zr-Atezolizumab PET revealed uptake in different tumor lesions (white arrows) in one representative patient 7 days postinjection.100 Pronounced uptake heterogeneity was seen within the same patient. (H) [¹⁸F]FB-IL2 uptake corresponding to activated tumor-infiltrating T cells at an adrenal gland metastasis (red arrow) in a representative melanoma patient undergoing immune checkpoint inhibitor therapy (SUV_max of 5.2).37 All patients enrolled in this study received intravenous bolus injection of~200 MBq [¹⁸F]FB-IL2 in 5 min. Transversal PET-CT image shown. All figures adapted with permission from publishers. CAR, chimeric antigen receptor; DC, dendritic cell; LNs, lymph nodes; SPIO, superparamagnetic iron oxide; PET, positron emission tomography.

Imaging regulatory T cells and gamma delta T cells

Regulatory T cells (Tregs). The efficacy of adoptive regulatory T cells (Treg) therapy (based on either polyclonal or the more potent antigen-specific Tregs), was demonstrated in numerous preclinical disease models and clinical trials.32 Nuclear-based reporter gene imaging offers a clinically compatible platform for future Tregs imaging in humans by providing the missing information on their in vivo distribution and persistence and addressing potential safety concerns. Notably, NIS-afforded SPECT-CT allowed tracking of human polyclonal Tregs to the human skin transplant in humanized mice (figure 1C). Authors further demonstrated that Tregs successful trafficking to the skin graft is regulated by Gr-1⁺ neutrophils and monocytes.33

CAR-Tregs. To improve specificity for the target antigen and therapy success, the field is quickly evolving from polyclonal to CAR-Tregs. As a result, more CARs are being developed and require in vivo testing by molecular imaging. Notably, HLA-A2 CAR-Tregs were recently engineered to constitutively express IL-10 and NIS reporter. In vitro, cells maintained their phenotype and gained an additional suppressive advantage, while NIS could be used for their non-invasive in vivo monitoring.34

Gamma delta T cells. They exert their cytotoxic activity through antigens recognition and indirectly enhance the anti-tumor activity of other immune cells by secreting multiple cytokines. Tracking with ⁸⁹Zr-oxine in a xenograft model of human triple-negative breast cancer revealed a significantly increased tumor infiltration upon pretreatment with PEGylated liposomal alendronate.35

Functional Imaging of T cells

Upon antigen recognition and engagement, T cells upregulate multiple activation biomarkers (surface, nuclear), leading to the release of anti-tumor cytokines. Those are early predictive biomarkers of immunotherapy response, providing valuable information on T cell activation and expansion (e.g., signal increase over time), and can be targeted by molecular imaging.36 Like cancer cells, activated T cells undergo metabolic reprogramming to survive the hostile tumor microenvironment (TME) and retain their antitumor activity. Those metabolic changes can also be in vivo monitored in real-time using molecular imaging.

T cell activation biomarkers. In vivo imaging of T cell-surface markers (such as the costimulatory molecules OX40 and ICOS (figure 1L), IL-2 (NCT01789827; NCT02922283 in figure 2H),37 TCR and lineage defining molecules (including CD3, CD4, and CD8) was successfully performed using direct labeling with PET radiotracers.36 Preclinical testing of ⁸⁹Zr-DFO anti-CD8 cys-diabody (figure 1I)38 and ⁸⁹Zr-Df-IAB22M2C mini-body (figure 2D; trialed in NCT04029181, NCT03107663, NCT03802123)39 revealed CD8+T cells response against cancer. A new ⁸⁹Zr-DFO-REGN5054 antibody40 for the non-invasive detection of CD8 expression is now in clinical trial (NCT05259709). Other T cell-surface markers associated with T cell activation, such as PD-1, CTLA-4, TIM-3 and LAG-3, are also immune checkpoint molecules and will be discussed in their dedicated section.

Targets related to T cell activation. Granzyme B is a predictive biomarker of successful checkpoint immunotherapy and can be imaged by ⁶⁸Ga-NOTA-GZP PET (figure 1H; NCT04169321).41 Similarly, a preclinically tested ⁸⁹Zr-labeled monoclonal antibody targeting IFN-gamma may provide insights on immune T cell activation and function.42

Nuclear Factor of Activated T cells (NFAT)-mediated T cell activation. Ponomarev et al described a TCR-dependent NFAT-HSV1-tk-GFP inducible PET-FLI dual-reporter system under the control of an artificial cis-acting NFAT-specific enhancer (figure 1D).43 44 On TCR-MHC complex engagement, the NFAT-HSV1-tk-GFP starts reporting on NFAT-mediated transcription of IL-2 and other cytokines, thereby providing information on the in vivo activation status of T cells. The combination of NFAT-inducible and constitutive reporters allowed to simultaneously visualize the trafficking, proliferation and activation of donor T cells and T cell precursors during T cell development after allogeneic hematopoietic stem cell transplantation in a mouse model of graft-versus-host disease.45

Metabolic targets. [¹⁸F]FDG imaging is routinely used to assess the changes in glucose metabolism and monitor the response in patients undergoing immunotherapy, including checkpoint inhibition.36 Other ¹⁸F-based examples are in overview from van der Veen et al.46

Imaging the Tumor Microenvironment (TME)

As the TME plays a critical role in immune surveillance and immune evasion, its major regulators can be considered promising immunotherapeutic targets. Immunosuppressive conditions include the upregulation of several immune checkpoints negatively regulating T cell activity (e.g., PD-1/PD-L1, CTLA-4, TIM-3 and LAG-3). Targeting of checkpoints by so-called ‘immune checkpoint inhibitors’ led to the Food and Drug Administration (FDA) approval for clinical use of (1) PD-1 inhibitors (e.g., pembrolizumab, nivolumab), (2) PD-L1 inhibitors (e.g., atezolizumab, avelumab, durvalumab) and (3) CTLA-4 inhibitor (ipilimumab). All immune checkpoints are amenable to molecular imaging and a wide range of agents has already been developed to non-invasively interrogate their heterogeneous expression and dynamics in preclinical models and clinical trials.36 46

Programmed cell death protein 1 (PD-1/CD279). A preclinical evaluation of ⁸⁹Zr-labeled FDA-approved Pembrolizumab (Keytruda; humanized IgG4) suggested its use for the selection of patients benefitting from anti-PD-1 therapy.47 Suitability of ⁸⁹Zr-labeled FDA-approved anti-PD-1 antibody Nivolumab as a diagnostic and disease monitoring tool is being investigated in a clinical trial (2015-004760-11).48 Accurate identification of early responses and resistance to PD-1 blockade was assessed by ⁸⁹Zr-labeled Pembrolizumab PET (NCT02760225).49

Programmed cell death protein ligand 1 (PD-L1/CD274). Antibody-based therapies targeting PD-L1 aim to monitor its in vivo expression as a predictive biomarker for checkpoint inhibitor therapy success. Examples are the newly developed nanobodies against PD-L150 and BMS-986192 tracer51 for SPECT and PET imaging, respectively. The latter tracer is now in two clinical trials (NCT03843515, NCT3520634).

Cytotoxic T lymphocyte-associated protein 4 (CTLA-4/CD152). Imaging CTLA-4 as a predictive biomarker of anti-CTLA-4 therapy is a useful tool. To reliably monitor CTLA-4 in vivo expression, the FDA-approved CTLA-4 antibody Ipilimumab was radiolabeled with multiple PET tracers and validated in xenograft52 and humanized mouse models.53 In a clinical trial, investigators are studying if high uptake of ⁸⁹Zr-Ipilinumab correlates to treatment response (NCT033313323). As CTLA-4 blockade prevents engagement to B7 costimulatory molecules and promotes T cell activation and proliferation, [¹⁸F]FLT PET can detect the associated changes in DNA synthesis during anti-CTLA-4 therapy in secondary lymphoid organs (e.g., spleen) of patients with metastatic melanoma (figure 2C; NCT00471887).54

Lymphocyte activation gene 3 protein (LAG-3). The in vivo expression of LAG-3 can be assessed using ^99mTc labeled nanobodies55 and ⁸⁹Zr labeled REGN3767 fully human anti-LAG-3 antibody.56 Clinical studies with ⁸⁹Zr-DFO-REGN3767 are ongoing in patients with diffuse large B-cell lymphoma (NCT04566978) and advanced solid tumors (NCT04706715).

T cell immunoglobulin and mucin domain-containing protein 3 (TIM-3). TIM-3 in vivo expression was monitored in a metastatic melanoma mouse model using the ⁶⁴Cu-NOTA radiolabeled rat anti-mouse TIM-3-specific monoclonal antibody (RMT3-23) by PET.57 A clinical trial investigated the in vivo distribution of ¹⁸F-CFA for the reliable monitoring of immune activation post anti-TIM-3 and anti-PD-1 treatment (NCT03409419). However, no results are available to date.

T cell immunoreceptor with Ig and ITIM domains (TIGIT). Two antibody-based TIGIT-specific ⁶⁴Cu- and ⁸⁹Zr-labeled tracers were developed for the preclinical quantification of TIGIT expression on TILs.58 These probes could be used to select which patients are candidates for anti-TIGIT immunotherapy.

Imaging other immune cell types

Natural killer (NK) cells. The lack of monitoring and diagnostic platforms to assess of NK-based immunotherapies efficacy has constituted a major drawback to the development of novel therapeutic strategies. NK cells can be imaged in real-time by MRI.59 As some iron-oxide nanoparticles have already been approved by the FDA (e.g., ferumoxide, ferumoxytol and ferucarbotran), this approach can be directly translated into the clinic for the short-term monitoring of engineered NK cells. However, nuclear-based imaging remains the most promising methodology for their clinical translation due to the increasing number of FDA-approved probes.59 Notably, Sato et al demonstrated that ⁸⁹Zr-oxine labeling of adoptively transferred NK cells from rhesus macaque, here used as a clinically relevant preclinical model, does not impact neither cell phenotype nor viability or function of therapeutic cells, thereby providing a clinically translatable platform for their in vivo tracking in humans.60

CAR-NK cells. They provide the unique opportunity to engineer off-the-shelf allogeneic products readily available for clinical use. Direct labeling is a readily applicable method to monitor their in vivo fate in patients,59 whereas the challenges posed by the intrinsic properties of NK cells (e.g., poor transduction with retroviral and lentiviral methods) have delayed the development of an indirect and clinically compatible long-term monitoring tool.

Dendritic cells (DCs). In vivo tracking of DCs is typically achieved by direct labeling. Aarntzen et al proved that pretreatment of melanoma patients with an extra dose of ¹⁹F-labeled DCs does not improve their relocation to the lymph nodes, a prerequisite for them to trigger an effective immune response.61 The same author used [¹⁸F]FLT to visualize T and B cell proliferation as a readout of early immune responses after administration of DC vaccine therapy in melanoma patients with lymph nodal metastases (figure 2E; NCT00243594, NCT00243529).62 Accuracy of intranodal and internodal delivery and in vivo distribution of DCs can be assessed by superparamagnetic iron oxide nanoparticles (SPIONs) MRI (figure 2F).63 A comprehensive list of MRI-based DCs imaging is in overview from Bulte and Shakeri-Zadeh.64 Alternatively, DCs can be imaged using reporter genes, including NIS65 and ferritin.66

Monocytes and macrophages. As highly phagocytic cells, they can efficiently take up SPIONs67 and ¹⁹F-PFC68 and be monitored by MRI. Tumor-associated macrophages preferentially incorporate the FDA-approved iron oxide nanoparticle compound ferumoxitol and can be preclinically imaged by MRI.67 CAR-macrophages (CAR-M)69 were pioneered in Gill’s laboratory at UPenn and entered the very first in-human phase 1 multicenter clinical trial for the treatment of metastatic HER2-positive solid tumors (NCT04660929). With more of CAR-M being developed, we expect to see an increase of preclinical and clinical testing employing molecular imaging.

Imaging Cellular Immunotherapies: a Future Perspective

With the constant expansion of the cellular immunotherapy arsenal, imaging the in vivo fate of cellular immunotherapies is paramount for the reliable assessment and prediction of therapeutic responses. As reporter genes have multiple advantages over the direct labeling approach, they could be fully integrated into clinical practice. The clinical studies involving HSV1-tk (figure 2A and table 2) have set the stage for expanding the portfolio of clinically used reporters for cellular immunotherapies imaging. However, some impediments have slowed down the process, including the regulatory concerns of using viral delivery systems to introduce the reporter into the desired therapeutic cells. Since commercially available CAR-T products are now generated through viral delivery (e.g., the lentiviral-based Kymriah and the retroviral-based Yescarta), the skepticism of using them for the delivery of reporter is no longer justified. In addition, vectors could be rendered even safer by using self-inactivating ‘third generation’ versions currently tested in gene therapy.70 To avoid the viral-mediated random positioning of reporter genes into the host genome, safer ways based on gene editing (e.g., ZFPs, TALENs and CRISPR-Cas) are available.71 The delivery at the same time of both the CAR and the reporter (or reporters) is desired, therefore the construction of GMP-compatible cassettes for the delivery of large payloads is needed. Pairing two different reporters (including an inducible one) would allow the simultaneous in vivo assessment of distribution and targeting of cellular immunotherapies with the monitoring of their functional status. However, it would also require the development of clinical imaging protocols specifically tailored to this set up. While they provide good image contrast, xenogenic (e.g., HSV1-tk) and synthetic reporters can elicit a host immune response. Although this may not represent an issue for preclinical studies, particularly when heavily immunocompromised mouse models are used, it is indeed a barrier for their full integration into clinical practice. Host reporters are a valid alternative as they are recognized as ‘self’ by the immune system. To this class belong two promising candidates for clinical translation, namely the human sodium iodide symporter (hNIS) and the human norepinephrine transporter (hNET) with their respective clinically compatible radioactive PET probes, [¹⁸F]BF₄^- and [¹⁸F]MFBG.14 The limited endogenous expression, the favorable target-to-background ratio, exquisite detection sensitivity and faster tracer clearance, make NIS a superior reporter for the imaging of cellular immunotherapies. Since NIS tracers do not rely on cyclotron accessibility (e.g., ^99mTcO₄^-, ^xyxI^-) and can be produced by automated synthesis at high molar activities (e.g., ¹⁸F-BF₄^-), their production is greatly simplified and cost effective, making them broadly available for imaging. Moreover, the short half-life of [¹⁸F]BF₄^- greatly facilitates the management of patients in a clinical setting.

As host reporters are endogenously expressed in the human body and are responsible for unfavorable contrast, reporters within this class (e.g., hNIS, hNET, hPSMA) need to be strategically selected depending on the expected distribution of the cell-based immunotherapy and location of the tumor. The human somatostatin receptor subtype 2 (hSSTr2) can be endogenously expressed in several immune cell types, including T cells, and may, therefore, not be suited for the imaging of cellular immunotherapies. With the FDA approval of more compatible probes, we do expect a significant increase in usage of reporters in immunotherapy-related clinical trials, but their full integration into clinical practice is still premature. Direct imaging has, therefore, become the fastest approach for clinical translation. Despite the evident limitations intrinsic to the methodology, the technological advancement of the total-body PET (with its >40 times higher sensitivity than conventional PET) can be used to extend the tracking time of directly labeled cellular immunotherapies and immune-related biomarkers. Many antibodies targeting immune cell therapeutics or immune biomarkers have been successfully conjugated to imaging probes. Some of them are currently being tested in clinical trials, others will soon reach the same stage. However, the applicability of these probes for the in vivo monitoring of cellular immunotherapies and biomarkers will be affected by the pharmacodynamics and pharmacokinetics of the therapeutic compound.

Imaging cellular immunotherapies is already playing a central role in guiding the FDA approval of the next generation of off-the-shelf cell-based immunotherapies and in reliably evaluate response in treated patients. Tumors can respond in a heterogeneous manner to cellular immunotherapy.72 Biomarkers and checkpoints expression can be heterogeneous too.73 While biopsies alone may not be able to catch with the same level of precision the complexity of the immune response, the latter can be reliably captured by molecular imaging between patients or within the same patient and, sometimes, within the same tumor lesion. Therefore, the clinical application of molecular imaging is not exclusively limited to the in vivo monitoring of therapeutic immune cells and their activation status, but also of immune cell biomarkers as a readout of heterogeneous response to immunotherapy, contributing to the identification of patients most likely to benefit from these therapies and thus avoiding unnecessary side effects and reducing the costs associated with ineffective treatment. The increased number of clinical trials of cell-based immunotherapies employing molecular imaging and the constantly expanding clinically approved portfolio of imaging probes will provide justification toward the consolidation of this tool into clinical practice in the years to come. Moreover, radiomics in becoming increasingly important by offering a large array of prospective biomarkers for the interpretation and even prediction of immunotherapy treatment and unfavorable effects. This predictive value increases when combined to other ‘omics’ and/or artificial intelligence.