Research ArticleCancer

Reporter gene imaging of targeted T cell immunotherapy in recurrent glioma

See allHide authors and affiliations

Science Translational Medicine  18 Jan 2017:
Vol. 9, Issue 373, eaag2196
DOI: 10.1126/scitranslmed.aag2196

T cells reporting for duty

Cytotoxic T cells engineered to kill tumor cells are becoming a mainstay of cancer immunotherapy. However, no matter how precisely they are engineered, once they are injected into a patient, they are no longer directly monitored or controlled by the researchers. As a result, if the treatment fails to work or causes toxicity, it is not clear whether the therapeutic cells are ineffective or whether they scattered through normal tissues and never reached the tumor. Keu et al. have designed a method to engineer these T cells with a reporter gene such that they can be tracked in people by positron emission tomography. The authors present a clinical trial demonstrating the feasibility and safety of this approach in glioma patients.


High-grade gliomas are aggressive cancers that often become rapidly fatal. Immunotherapy using CD8+ cytotoxic T lymphocytes (CTLs), engineered to express both herpes simplex virus type 1 thymidine kinase (HSV1-TK) and interleukin-13 (IL-13) zetakine chimeric antigen receptor (CAR), is a treatment strategy with considerable potential. To optimize this and related immunotherapies, it would be helpful to monitor CTL viability and trafficking to glioma cells. We show that noninvasive positron emission tomography (PET) imaging with 9-[4-[18F]fluoro-3-(hydroxymethyl)butyl]guanine ([18F]FHBG) can track HSV1-tk reporter gene expression present in CAR-engineered CTLs. [18F]FHBG imaging was safe and enabled the longitudinal imaging of T cells stably transfected with a PET reporter gene in patients. Further optimization of this imaging approach for monitoring in vivo cell trafficking should greatly benefit various cell-based therapies for cancer.


Immunotherapy holds great potential for the treatment and management of cancer patients with advanced disease (1). Through numerous divergent mechanisms, including monocyte maturation in the presence of tumor antigens (2) and the genetic engineering of immune cells (3), the body’s adaptive immune system can be primed to target malignancies normally recognized as “self.” Although the success of recent phase 3 trials has validated the principle that immunotherapy can sometimes extend cancer patient survival (4), tumor cells can escape immune surveillance and develop resistance to immunotherapy (5). Examples of resistance mechanisms include the regulation of immune checkpoints (6), resistance to cell death signaling (7), or regulation by other tumor-associated immune cells, such as regulatory T cell lymphocytes (8). Given the variable success of immunotherapy in the clinic (911), there is an urgent need to design noninvasive techniques that could give early indications of response to treatment and help predict patient outcome.

Current methods used for the assessment of tumor response to therapy primarily rely on measurements of tumor size through anatomical imaging using the response evaluation criteria in solid tumors (RECIST) (12). The limitations of anatomical imaging for response monitoring in solid tumors have been well documented (13). Assessment of immunotherapy efficacy provides particular challenges in that the influx of effector cells to the tumor microenvironment may result in “pseudoprogression” as defined by RECIST, which is not linked to long-term outcome (14). These limitations have led to the development of a specific set of immune-related response criteria (irRC) (14, 15). However, irRC measurements provide no biological information and can take 8 to 12 weeks for an initial response assessment, by which time nonresponding patients may have missed their therapeutic window and experienced unnecessary therapy-related toxicity. In addition, adaptive immune responses involving T and B lymphocytes, which have demonstrated great therapeutic potential, require monitoring of immune cell trafficking to the tumor site, lymphocyte retention, and immune cell engagement with the tumor cells. At present, techniques to noninvasively assess the in vivo viability and trafficking of these immune cells in cancer patients are not available.

We previously provided a demonstration of imaging cells in a single case report using reporter gene technology (16). In that case study, therapeutic T cells expressing a positron emission tomography (PET) reporter gene were injected into the tumor resection site of a 57-year-old male high-grade glioma patient and then visualized by molecular imaging with a PET reporter probe (16). For that study, we used genetically modified CD8+ autologous cytotoxic T lymphocytes (CTLs) stably expressing the chimeric antigen receptor (CAR) interleukin-13 (IL-13) zetakine for IL-13Rα2+ tumor targeting (1719) and the herpes simplex virus type 1 thymidine kinase gene (HSV1-tk) as a dual-purpose suicide (20, 21) and imaging reporter gene (Fig. 1) (22, 23). HSV1-tk was fused in-frame with the hygromycin phosphotransferase (hph) gene and expressed as a fusion protein. HPH expression conferred hygromycin B resistance for dominant positive selection. Through imaging of HSV1-tk reporter gene expression using 9-[4-[18F]fluoro-3-(hydroxymethyl)butyl]guanine ([18F]FHBG), which is a fluorine-18 radiolabeled analog of the anti-herpes drug penciclovir (Fig. 1, inset), we could image [18F]FHBG trapping in the brain tumor, which was thought to correspond to CTL accumulation. This corroborated the extensive preclinical assessment of this targeted strategy (24). However, this study in a single patient was limited because, at that time, we only had U.S. Food and Drug Administration (FDA) approval to perform one PET scan after the infusion of CTLs and did not have a critical baseline PET image before CTL injection. Therefore, we could not confirm that the brain-associated [18F]FHBG signal was CTL-specific (22).

Fig. 1. The HSV1-tk complex and monitoring by [18F]FHBG.

The HSV1-tk complex is a genetically modified structure that was transfected into either autologous or allogeneic CTLs. This complex expresses a fusion protein consisting of a selective gene (hygromycin resistance locus) for adequate in vitro expansion of the CTL line; a PET reporter gene and safety gene, HSV1-tk; a glucocorticoid receptor (GR) resistance locus for improving CTL survival despite the high doses of steroids routinely given to high-grade glioma patients; and an IL-13 zetakine domain, which is a chimeric receptor that enables glioma cell recognition by CTLs. CTL transfection was performed by electroporation. Inside the transfected cells, HSV1-tk is transcribed and translated to produce the HSV1-TK enzyme. [18F]FHBG is a labeled analog of penciclovir and substrate for HSV1-TK. In the presence of HSV1-TK, the radiolabeled probe is phosphorylated and trapped within the cell. The magnitude of [18F]FHBG signal reflects the activity of HSV1-TK enzyme and thus HSV1-tk gene expression. pmRNA, pre-messenger RNA.

Expanding on this pilot case study, we sought to assess the ability of [18F]FHBG to longitudinally monitor CTL trafficking, survival, and proliferation in multiple patients with recurrent high-grade gliomas that were resistant to conventional therapies. Under this protocol, seven patients received several infusions of either autologous or allogeneic engineered CTLs (Fig. 1). CTL location and response were noninvasively measured by PET using a strategy outlined in Fig. 2. In addition, we evaluated the safety of the CTLs injected into the central nervous system and the safety of recombinant human IL-2 delivery in conjunction with CTL adoptive cell transfer, a strategy used to prolong the survival of engineered CTLs and potentially improve their therapeutic efficacy.

Fig. 2. Schematic demonstrating the workup and monitoring of CTL infusion after an initial treatment for high-grade glioma.

After enrollment in this study and upon tumor recurrence, an intracerebral Rickham catheter was installed for repetitive infusions of CTLs. Each patient underwent [18F]FHBG PET and magnetic resonance imaging (MRI) before and after CTL infusions. In addition, IL-2 was administered at regular intervals to further increase the survival and enhance the potency of the administered CTLs. CTL distribution was imaged by measuring changes in tumoral [18F]FHBG accumulation before and after infusions of cells.


Patient enrollment and CTL characterization

To conduct a detailed evaluation of the use of [18F]FHBG to track genetically engineered CTLs for immunotherapy, we enrolled a total of four men and three women (average age of 57 ± 8 years old with recurrent high-grade glioma) under two separate Institutional Review Board (IRB)–approved protocols (Table 1). Patient 1 in this study is the patient from our previous case report (16). Autologous (NCT00730613) (17) and allogeneic (NCT01082926) CTLs for immunotherapy were engineered to express the tumor-targeting IL-13 zetakine CAR and the wild-type HSV1-tk gene reporter. This CAR recognizes IL-13Rα2, a cell surface receptor expressed in more than 50% of glioblastoma cells but not in normal brain tissue (25, 26). The CTLs were >99% CD8. We first confirmed our previous finding (27, 28) that [18F]FHBG specifically accumulates in genetically modified CTLs expressing the HSV1-tk reporter gene rather than the naïve parental cells obtained from a human buffy coat (fig. S1). Here, [18F]FHBG accumulation at 1 hour was 12-fold higher in transfected CTLs versus nontransfected parental cells, at 7.37 ± 2.17% and 0.61 ± 0.27% [18F]FHBG uptake per 108 cells, respectively (n = 3, P = 0.006). Furthermore, 60-min incubation with [18F]FHBG had no significant effect on cell proliferation of nontransduced CTLs or those expressing HSV1-tk (taken from patient 1), up to 48 hours after the addition of the radiotracer (fig. S2; P > 0.05).

Table 1. Imaging and demographic data for all seven patients enrolled in this study.

N/A, not applicable.

View this table:


[18F]FHBG PET scans were performed in patients before and after CTL infusions to assess the spatial location of the engineered CTLs within the body. Patients 3 to 7 received IL-2 infusions to improve CTL survival. Patient vital signs and laboratory results were within normal limits throughout and after [18F]FHBG administration. Patient 1 experienced a mild, transient headache without any neurological deficits. Patient 2 developed new neurological deficits 2 days before the initiation of CTL therapy, which corresponded to vasogenic cerebral edema (diagnosed from MRI) and was not related to either [18F]FHBG or disease progression. Patient 5 had a mild rash on the shoulders with redness on the lower face at 2 hours after intravenous injection of [18F]FHBG during pre-CTL imaging. Twenty-five milligrams of oral diphenhydramine was subsequently administered to patient 5, and the rash resolved within 1 hour. None of the seven patients had any major or life-threatening events related to [18F]FHBG or CTL injections throughout the duration of their participation within this trial. Most patients died within 12 months after the last CTL infusion. Patient survival data after initial diagnosis and since the first CTL infusion are presented in table S1.


The stability and detailed pharmacokinetics of [18F]FHBG in healthy volunteers have been previously reported (29). Here, we observed similar radiotracer biodistribution (fig. S3), as assessed by standardized uptake value (SUV)mean measurements on the pre- and post-CTL injection scans using a 50-mm-diameter spherical volume of interest (VOI). In pre-CTL treatment scans, at a mean time of 137 ± 11 min after injection of [18F]FHBG (n = 6), the average (±SD) SUVmean was 0.25 ± 0.08 for left ventricular blood pool, 1.57 ± 0.69 for right hepatic dome, 0.09 ± 0.04 for lung parenchyma, and 0.36 ± 0.14 for right thigh. There was no significant difference in [18F]FHBG uptake in these tissues after immunotherapy (P > 0.05; table S2). In the normal brain, contralateral to either suspected lesions or resection sites and before CTL infusions, [18F]FHBG SUVmean was 0.04 ± 0.01 (n = 6). After CTL infusions, we observed no significant change in normal brain uptake (SUVmean = 0.04 ± 0.01; n = 6, P > 0.05). We next quantified the amount of [18F]FHBG activity in the vicinity of tumors where a breakdown of the blood-brain barrier (BBB) was observed on the basis of the postcontrast T1-weighted MRI scan. The low background from nontarget tissue (normal brain) permitted visualization of brain tissue abnormalities such as tumors (Fig. 3A), with [18F]FHBG uptake in untreated tumors and sites of tumor resection statistically higher than in the normal contralateral brain tissue (3.3- to 11.5-fold increase; P = 0.0009).

Fig. 3. Tumor-associated CTL imaging with [18F]FHBG.

[18F]FHBG PET imaging was performed in a 66-year-old male (patient 6) with a recurrent right frontoparietal glioblastoma before (A) and 1 week after (B) CTL infusions. Allogeneic CTLs and IL-2 were injected intratumorally (red arrows) as described in Materials and Methods. Tumor recurrence was monitored by T1-weighted (T1W) MRI (top panels). [18F]FHBG PET images were fused with MR images (bottom panels), and three-dimensional (3D) volumes of interest were drawn using a 50% [18F]FHBG SUVmax threshold, outlined in yellow. (C) Voxel-wise analysis of [18F]FHBG SUV in pre- and post-CTL infusion scans.

[18F]FHBG imaging of CTLs in patients

Given the observation of some nonspecific [18F]FHBG retention in both gliomas and previous surgical sites before CTL infusion, it was essential to perform baseline and follow-up scans in the same patients after CTL injection (16). A total of 10 tumor foci were analyzed in the seven patients, with two sites each assessed for patients 1, 4, and 7. We did not obtain a pre-CTL [18F]FHBG scan from patient 1 because the use of two [18F]FHBG scans per patient was not yet approved by the FDA when the patient was enrolled in 2006. Qualitative assessment of [18F]FHBG uptake by PET imaging revealed an increase in PET signal after CTL infusions, likely corresponding with CTL cell trafficking and viability (Fig. 3, A and B, and movie S1).

To quantify [18F]FHBG radioactivity and distribution, we initially evaluated the maximum concentration (SUVmax) of [18F]FHBG on pre- and post-CTL injection scans, where we observed a large variation in response, ranging from +53% (patient 2) to −15% (patient 7, lesion B) (fig. S4). After intracerebral injection, there is the potential for CTLs to diffuse in the surgical cavity or the brain parenchyma. To optimize the monitoring of CTL trafficking to targeted areas, better imaging metrics were needed to replace single-voxel measurements such as SUVmax, particularly when the localization of CTLs, and therefore the VOI, was unknown. The first challenge was to define an accurate VOI. Given the availability of contrast-enhanced MR images taken from the same patients before and after CTL infusions, it was possible to derive VOIs from enhancing regions of the tumor. MRI-based [18F]FHBG voxel analysis, however, excluded radioactivity outside of enhancing regions (fig. S5A), which may represent CTLs that have failed to target to the tumor or potential targeting to tumor cells that were not visible on MRI. Analysis of these therapeutic cells is essential because an indication of poor CTL trafficking to tumor tissue has implications for the efficacy of the immunotherapy strategy. To circumvent the exclusion of non–tumor-associated radioactivity, we used different SUVmax thresholds to draw a growing VOI. The contouring algorithm used selected the highest [18F]FHBG SUV as the center of the VOI before generating peripheral limits based on the desired threshold value, as illustrated in fig. S5B. VOIs for all patients were analyzed using threshold values from 10 to 90% in 10% increments (fig. S6). For [18F]FHBG analysis, a 50% SUVmax threshold was selected to create VOIs that best represented [18F]FHBG distribution throughout the brain.

Having determined a suitable, nonbiased method for VOI delineation, the next challenge was to devise a method of analysis that could assess CTL trafficking (i) to a region within the brain of fixed volume, such as a tumor, and (ii) when the [18F]FHBG signal was widely dispersed after CTL infusions (for example, as a result of poor trafficking to the tumor site, CTL dilution in the ventricular system, or correct trafficking to tumor tissue not visible by MRI). The SUVmean provided a means to assess tumor-specific CTL trafficking when the VOI did not change between pre- and post-CTL scans (fig. S7; patients 2 and 6) but failed to accurately depict changes in [18F]FHBG radioactivity after CTL infusions in patients for whom the VOI was increased in the post-CTL scan. To detect potential changes in [18F]FHBG radioactivity before and after CTL infusions, regardless of the size of the VOI, we next assessed the [18F]FHBG total activity (SUVmean × VOI) for each scan.

Analysis of the total activity in patients for whom the size of the VOI did not change, such as patient 6, clearly indicated a shift in SUV histograms after CTL infusions (Fig. 3C), corresponding to a doubling of [18F]FHBG total activity from 4.1 to 8.4 in this patient. A further example is provided with patient 2, whose total activity was increased from 10.1 to 20.3 (fig. S8). Total activity measurements could also detect radioactivity disseminated over an increased volume after CTL infusions, an example of which is illustrated in Fig. 4 (A and B) (patient 7). Here, as evident from the histogram analysis (Fig. 4C), despite an absence of a large change in median [18F]FHBG radioactivity between scans, the total radioactivity increased 3.7-fold, from 1.8 in the pre-CTL scan to 6.8 after CTL infusions, in part because of a substantial increase in the VOI. A further example of changes in [18F]FHBG total activity after CTL infusions in the absence of a change in SUVmean was observed with patient 3 (fig. S9). Across all patients, in clinically confirmed sites of tumor recurrence, the [18F]FHBG total activity was significantly increased in the PET scan after CTL infusions versus the baseline scan (Fig. 5; P = 0.014, paired Wilcoxon test).

Fig. 4. [18F]FHBG PET imaging in recurrent disease and at untreated tumor sites.

[18F]FHBG PET imaging was performed in a 60-year-old male (patient 7) with multifocal left hemispheric glioma. CTLs were injected into the medial left frontal lobe tumor (yellow arrows). (A) Tumor size was monitored by T1-weighted contrast-enhanced MRI (top left panels). [18F]FHBG PET images were fused with MR images (bottom left panels), and 3D volumes of interest were drawn using a 50% [18F]FHBG SUVmax threshold, outlined in red. (B) MRI and [18F]FHBG PET-MR images 1 week after CTL infusions. (C) Voxel-wise analysis of [18F]FHBG total radioactivity in pre- and post-CTL infusion scans. [18F]FHBG activity was additionally assessed in a noninjected tumor focus (red arrows) before (D) and after (E) CTL infusions, with voxel-wise analysis of this lesion performed for comparison (F).

Fig. 5. Changes in [18F]FHBG total activity in patients after CTL infusions.

The total activity (SUVmean × VOI) was measured for each patient on pre- and post-CTL [18F]FHBG scans, with the VOI obtained after a 50% SUVmax threshold. Clinically confirmed sites of tumor recurrence that received CTL infusions are indicated in blue. Noninjected tumor foci in patients with multifocal disease are indicated in gray. In patient 1, for whom no pre-CTL scan was available, the total activity values for the two tumor foci are shown in red. For reference, patient numbers are indicated next to the post-CTL [18F]FHBG values. The increase in [18F]FHBG signal after CTL infusion was statistically significant (P = 0.014, paired Wilcoxon test).

[18F]FHBG uptake in untreated tumor sites

In patients with multifocal disease, in addition to the injected tumor site, distant tumor foci not injected with CTLs were also analyzed with [18F]FHBG (patient 7; Fig. 4, D to F). No CTLs were infused locally, and therefore, no changes in [18F]FHBG uptake at the site were expected between scans. In the post-CTL scan, however, a 66% increase in [18F]FHBG total activity was measured by comparison to the baseline scan, from 1.6 to 2.7, which was further supported by minimal changes in the dispersion of voxel intensities (Fig. 4F). This is in contrast to the 5- to 15-fold changes observed at the injected disease site (Fig. 5). Small changes in [18F]FHBG uptake at untreated tumor sites after CTL infusions versus baseline scans were confirmed in an additional patient (patient 4). In this patient, [18F]FHBG total activity in the CTL-treated focus was increased 2.2-fold on the post-CTL scan, increasing from 2.9 at baseline to 6.3 at the site of tumor recurrence, whereas the focus corresponding to the primary untreated site increased just 32%, from 2.4 to 3.2 (fig. S10). The modest increase in [18F]FHBG activity in the untreated sites in these two patients may represent trafficking of CTLs to distant tumor foci in the brain.

Challenges of CTL tracking with [18F]FHBG

As a result of the complexity of implementing a robust immunotherapy treatment strategy for the treatment of high-grade gliomas in humans, a number of key experimental limitations presented themselves during the course of this study. In patient 1, FDA restrictions prevented the acquisition of a pre-CTL [18F]FHBG scan; however, we have previously shown in a post-CTL [18F]FHBG scan that CTL-associated radioactivity exquisitely corresponded to enhancing foci on MRI, which was not detectable by [18F]2-fluoro-2-deoxy-d-glucose ([18F]FDG) PET (fig. S11) (16). Therefore, we believed that CTLs had likely properly migrated to the target tumor areas, but given that high-grade glioma cells in pre-CTL scans were [18F]FHBG-positive in other patients, we could not confirm this hypothesis.

The CTL infusion techniques used in this study are also expected to play a role in the survival of CTLs in vivo and their homing to the tumor site. In patient 3 (NCT01082926), the CTLs were injected using a slow infusion technique over several hours. This delivery technique resulted in cell clumping that may have reduced cell viability; however, [18F]FHBG total activity was still increased 2.4-fold, from 3.5 to 8.4, in this patient (fig. S9), reflecting the likely survival of a substantial proportion of the CTLs. The positioning of the Rickham catheter tip next to the ventricular system may also affect optimal CTL trafficking to tumor sites. In patients 4 and 5, the Rickham catheter tip was positioned next to the ventricular system. In these patients, CTLs likely diffused into the cerebrospinal fluid, as evidenced by the diffusion of [18F]FHBG within the lateral ventricles in the post-CTL [18F]FHBG scan, which was not observed on the pre-CTL scan (fig. S10, A and B). These findings suggest that the dilution phenomenon might potentially affect the therapeutic effect by reducing the overall number of CTLs at the recurrent tumor site. On the other hand, this might be beneficial for patients whose tumors are located next to the ventricular system and are at a higher risk of spinal leptomeningeal dissemination (30, 31). Nevertheless, these explanations require further validation.

The tumor vascular supply may also affect the ability to track CAR-engineered CTLs with [18F]FHBG. In patient 5, the recurrent glioma was adherent to the dura mater, which contributed to a rich vascular supply with disrupted blood-tumor barrier. As a consequence, intravenously injected [18F]FHBG could readily diffuse from the leaky vascular supply into the tumor. In this patient, baseline [18F]FHBG total activity was substantially higher at 23.1, compared to a mean total activity of 6.2 ± 2.9 for all other patients for whom a pre-CTL scan was available (n = 7 foci). It is highly plausible that elevated baseline [18F]FHBG tumor uptake is a direct result of increased delivery and nonspecific retention of the radiotracer in the tumor. Furthermore, high baseline [18F]FHBG uptake on pre-CTL infusion scans may confound accurate measurement of CTL tracking. Total [18F]FHBG activity in patient 5 was increased by just 21% in the post-CTL infusion scan when compared to the baseline scan, at 28.1 and 23.1, respectively (fig. S12).


Malignant gliomas are the most frequently occurring primary brain tumors and are associated with a very poor survival rate (32). Tremendous efforts have focused on the design of diagnostic and therapeutic tools for this devastating disease, among which targeted immunotherapy holds great promise (33). Unlike current conventional cancer treatments, the potential of immunotherapy resides in selectively targeting cancer cells while leaving healthy cells unharmed. One of the main immunotherapeutic methods under exploration is through CAR engineering of CTLs to become tumor-reactive and tumor-specific (34).

To fully validate immunotherapeutic strategies, it is crucial to monitor the viability, biodistribution, and trafficking of therapeutic cells to the site of the tumor. At present, there is no means to assess the fate of these cells. Imaging via MR, optical, or nuclear medicine–based techniques, however, has the potential to noninvasively monitor and track these cells in vivo. Two main approaches have been undertaken for cell tracking: direct and indirect labeling methods (35, 36). Direct labeling of cells involves incubation and retention of an imaging agent by the therapeutic cells that are injected into the subject, with the imaging agent monitored over a time course of hours to days. One example of a direct labeling method, using superparamagnetic iron oxide–labeled cells, has demonstrated great potential for the accurate assessment of image-guided cell injection into the tissue of interest (37).

Although they are relatively inexpensive and easy to perform, direct labeling methods are hampered by potential toxicity to the therapeutic cells (38). Moreover, the contrast agent becomes diluted upon cell division and is lost from the cells upon cell death, making image analysis difficult to interpret (39). To circumvent these issues, we and others (4043) have developed noninvasive indirect imaging reporter strategies to successfully track these cells. These methods enable imaging over the entire lifetime of the cell, with signal maintained after cell division, and provide information regarding cell viability (39). Most of the signal is obtained from living cells because the production of the imaging reporter is dependent on the translation of a gene into a functional protein, in a process requiring adenosine 5′-triphosphate. Here, we demonstrated in humans that therapeutic CAR T cells can be detected noninvasively after intracranial adoptive transfer, through the PET imaging of a reporter gene.

CTLs were engineered to express the tumor-targeting IL-13 zetakine CAR and the wild-type HSV1-tk reporter gene, with HSV1-TK expression monitored by [18F]FHBG PET imaging. In cell culture, [18F]FHBG uptake was 12-fold higher in HSV1-tk–expressing CTLs compared to naïve human T lymphocytes that were isolated from a human buffy coat. [18F]FHBG incubation also had no effect on CTL proliferation rate in cell culture, presumably because of the low amount of mass tracer used. The functional and cytolytic activity of CTLs obtained from patient 1 has been shown previously (18). For all other patients, insufficient tumor material was obtained at biopsy to enable primary glioblastoma or patient-derived xenografts to be established in immunocompromised mice. Therefore, we could not assess functional activity of the CTLs against each patient’s glioblastoma cells.

[18F]FHBG is a relatively poor substrate for mammalian thymidine kinases, resulting in low background accumulation in cells not expressing the HSV1-tk reporter gene, such as naïve human T lymphocytes and normal brain cells (22). Negligible brain-associated radioactivity was observed in [18F]FHBG scans both before and after CTL infusions, with [18F]FHBG not able to diffuse across an intact BBB (29), thereby providing an ideal low background signal. In glioma patients with a disrupted BBB, however, [18F]FHBG accumulation in tumors was evident before CTL infusions. Retention of [18F]FHBG in untreated tumors before CTL infusion may result from slow washout of the radiotracer from the resection cavity or be a consequence of off-target retention in the tumor cells. It was therefore vital to perform [18F]FHBG imaging both before and after CTL infusions for accurate monitoring of this CAR T cell therapy strategy.

Because the location of the CTLs within the brain cavity was unknown, better imaging metrics were needed to replace single-voxel measurements, such as SUVmax, to define both the location and quantity of the therapeutic cells. We used a contouring algorithm based on a set radioactivity threshold value relative to SUVmax to determine an accurate VOI. Using this method, we could accurately assess not only tumor-associated CTLs but also poorly targeted CTLs and CTLs that had targeted diffusely growing tumor cells undetected on contrast-enhanced MRI. For quantitation, we assessed [18F]FHBG total activity within the generated volume of interest (SUVmean × VOI) as an unbiased measure of the [18F]FHBG signal in a fixed volume region within the brain (namely, the tumor) and in the setting where the [18F]FHBG signal, and therefore CTLs, was widely dispersed after CTL infusions. This metric is comparable to total lesion glycolysis used with [18F]FDG PET, which is a more sensitive marker than SUVmax or SUVmean for the evaluation of therapeutic response and prognosis (4446).

The noninvasive imaging of genetically engineered CTLs for cancer immunotherapy held some specific challenges. Because [18F]FHBG was the first PET reporter probe to receive an investigational new drug (IND) approval from the FDA (IND #61,880), strict regulations prevented repeat [18F]FHBG scans before and after immunotherapy for the initial patient. This, coupled with the complexity of the immunotherapy strategy, markedly high cost (estimated at ~$200,000 per patient), and limited access to the patient population, restricted the number of patients who were enrolled in this study. Despite these obstacles, a significant increase in [18F]FHBG total activity representing CTL trafficking to tumor sites was observed across our patient cohort (P = 0.014, paired Wilcoxon test), highlighting the potential of this imaging strategy. It is unclear, however, whether IL-2–mediated changes in BBB permeability contributed to changes in [18F]FHBG signal in the post-CTL scan. Moreover, we could not assess accurate delivery of CTLs to the tumor and its effect on differential [18F]FHBG uptake. These variables will need to be evaluated in future studies. Although the absolute intrinsic uptake of [18F]FHBG per cell remained relatively low, the mutant HSV1-tk reporter gene, HSV1-sr39tk, may provide increased sensitivity for detection through increased trapping of [18F]FHBG in engineered cells in future clinical trials (23, 47). Finally, because of the limited number of patients enrolled on this study, it was not possible to link changes in [18F]FHBG signal before and after CTL infusions to clinical outcome. Future work will need to link CTL trafficking and viability, as measured by [18F]FHBG PET, to tumor response and patient survival. In summary, this work highlights an approach that uses a specific PET tracer to image HSV1-tk reporter gene expression in engineered CTLs to monitor CAR therapy for the treatment of high-grade gliomas.


Study design

The primary research objective was to test the feasibility of [18F]FHBG gene reporter imaging for the monitoring of therapeutic T cell distribution and viability in glioma patients. Patients from 18 to 70 years old with an average clinical status (a Karnofsky performance status of >60) and evidence of disease progression 12 weeks after the end of radiotherapy were included in this study. Exclusion criteria were a survival expectation of less than 4 weeks; any organ dysfunction; a large tumor requiring decompression surgery; a tumor located in the basal ganglia, thalamus, or brainstem; a patient treated for severe infection or recovering from major surgery; a history of IL-2 intolerance; and a history of ganciclovir and/or MRI contrast allergy or intolerance. The authors were not blinded to the results, and no randomization was performed.

Patient recruitment

Seven patients (four men and three women; mean age ± SD, 57 ± 8 years old) were prospectively enrolled from April 2006 to October 2013 in two separate clinical trials to evaluate the safety and feasibility of intratumoral immunotherapy with CTLs (Table 1). Patients 1 and 2 were treated with autologous CTLs (NCT00730613) (17), whereas patients 3 to 7 received allogeneic CTLs (NCT01082926). Patients included in this study were diagnosed with recurrent glioblastoma overexpressing the IL-13Rα2 receptor, thus allowing the recognition of malignant glioma cells by the genetically modified CTLs (17, 19, 48). It was previously shown that patient prognosis and survival are inversely proportional to the abundance of tumor-associated IL-13Rα2 (25). Nevertheless, the presence of these specific targets helped to define patients who would likely benefit and respond to immunotherapy. No patient received additional radiotherapy or chemotherapy throughout this study. All patients were initially treated and followed up at the City of Hope, Duarte, CA. The IRB and the local ethics committee approved all study protocols, and informed consent was obtained from each patient.

CTL engineering

The T cell manufacturing process was described previously in detail (17). For patients 1 and 2, T cells were expanded from clones that had been selected from genetically modified T cells isolated from the patient’s peripheral blood mononuclear cells (PBMCs). A plasmid DNA construct encoding the IL-13 zetakine CAR and the wild-type HSV1-tk genes under the transcriptional control of a modified human elongation factor-1α promoter and the cytomegalovirus immediate/early promoter was transfected into the isolated cells using electroporation in a cell production facility at the City of Hope, Duarte, CA. Hygromycin-resistant CTLs were cloned in limiting dilution, then expanded using the rapid expansion method (REM) to numbers in excess of 109, and subsequently cryopreserved (17). After the diagnosis of tumor relapse, the cryopreserved cells were thawed, REM-expanded, and formulated for intracranial infusion in 2 ml of preservative-free normal saline (PFNS). For patients 3 to 7, the GRm13Z40-2 CTL line was derived from a healthy volunteer donor apheresis unit. The GRm13Z40-2 CTL line is an ex vivo expanded allogeneic genetically modified oligoclonal CD8+ CTL line that expresses the IL-13 zetakine CAR and contains a biallelic deletion in the GR to confer steroid resistance. PBMCs were subjected to electroporation to introduce plasmid IL-13 zetakine CAR DNA, followed by activation with monoclonal antibody muromonab CD3 (30 ng/ml; trade name Orthoclone OKT3, marketed by Janssen-Cilag), ex vivo REM expansion (17), and selection in the presence of cytotoxic concentrations of hygromycin (0.2 mg/ml). Hygromycin-selected IL-13 zetakine CTLs were then subjected to transfection using the chimeric Ad5/35 adenovirus encoding a zinc finger nuclease (ZFN) pair that targets the human GR, which caused permanent disruption of the reading frame. Adenovirus-transfected IL-13 zetakine CTLs were subjected to selection in dexamethasone. Surviving T cells underwent large-scale REM expansion followed by cryopreservation to generate an allogeneic IL-13 zetakine CTL bank. After the diagnosis of recurrent glioblastoma, the cryopreserved cells were thawed and formulated for intracranial infusion in 2 ml of PFNS.

Infusion methods

Patients 1 and 2 underwent craniotomy for tumor resection and placement of Rickham reservoir and catheter for CTL injection, as previously detailed (17). These patients were treated with a series of 12 infusions of autologous CTLs (each for 10 min) on days 1, 3, and 5 over a period of 5 weeks, with a break on week 3. If the initial dose of 1 × 107 CTLs was tolerated, subsequent infusions were increased to 5 × 107 and then 1 × 108 autologous CTLs per injection. Patients 3 to 7 had unresectable recurrent glioblastoma. These patients underwent stereotactic tumor biopsy and insertion of Rickham reservoir/catheter for intratumoral CTL infusion. These patients were treated with a series of four infusions of 1 × 108 allogeneic CTLs along with escalating doses of recombinant human IL-2 to prolong the survival of CTLs (49). Allogeneic CTLs were infused on days 1 and 3 for two consecutive weeks; IL-2 was infused on days 2, 3, 4, and 5 of the first week (2500 IU/day) and on days 1, 2, 3, 4, and 5 of the second week (5000 IU/day) (table S3). In patient 3, CTLs and IL-2 were administered directly into the tumor cavity by convection-enhanced delivery, which is a drug delivery strategy designed to circumvent the central nervous system defense system and improve drug distribution (50). It allows for delivery of high concentrations of therapeutic agents directly into brain tumors and surrounding parenchyma with minimal systemic toxicity. Because of cell clumping associated with slow infusion, the CTL delivery technique was switched to a “manual push” technique for 10 min in patients 4 to 7.

Clinical assessment

Before each [18F]FHBG scan, we obtained a physical exam, a 12-lead electrocardiogram (ECG), vital signs (blood pressure, heart rate, respiratory rate, body temperature, and blood oxygenation), blood sampling (for complete blood chemistries, complete metabolic panel, international normalized ratio, and activated partial thromboplastin time), and urine (for urine analysis and pregnancy test, if needed). These values were used as a baseline. The patients’ routine medications and medical history were also recorded. Vital signs were taken at 5, 10, 15, 60, and 120 min after [18F]FHBG injection. ECGs were monitored every 15 min up to 120 min after the tracer injection and were analyzed for PQ/QRS and QT changes.

After the PET scan was completed, another set of vital signs and ECG were collected. Within 24 hours of the scan, the participants returned to the clinic, and vital signs and ECG were again collected, as well as blood and urine samples. Seven days later, a similar assessment was performed in addition to a neuropsychological test (mini mental state examination) that was also performed at baseline. Any adverse events noted by the members of the research team or the participants during the whole procedure were recorded for up to a week during follow-up.

Imaging studies

[18F]FHBG was synthesized as previously described (22). The mean [18F]FHBG activity injected was 259 ± 7 MBq (7.0 ± 0.2 mCi), with no significant differences between pre- and post-CTL infusion PET scans (P > 0.05). The maximum amount of unlabeled product was less than 2 μg, and the specific activity was greater than 37 Gbq/μmol (22).

The tracer was administered as a bolus injection for less than 5 s via a hand or an arm vein. About 2 hours after intravenous injection of [18F]FHBG, PET images were obtained from the skull vertex to the mid-thighs. Scans for patients 1 and 2 were performed with a standalone PET scanner (Siemens ECAT EXACT HR+ PET scanner, CTI PET Systems Inc.). [18F]FHBG PET scans, at 7 min per bed position, were performed for both patients 1 and 2, but no baseline scan was done for patient 1 (16). In addition, delayed brain PET images were obtained for patient 1. After the emission scan, the patient underwent a transmission scan for 3 min per bed position, using three rotating 68Ge rod sources. PET images were reconstructed with the ordered subset expectation maximization algorithm with attenuation correction on a 128 × 128 matrix size. The images for patients 3 to 5 were obtained with a PET/computed tomography (CT) scanner (Siemens Biograph TruePoint 64-slice). A topogram scan [120 kilovolt peak (kVp); 35 mA⋅s] was acquired, followed by a low-dose noncontrast CT (helical mode; rotation, 0.5 s, slice thickness, 5.0 mm; 120 kVp; 100 mA) for attenuation correction and anatomic localization. PET images were obtained in 3D mode for 7 min per bed position, using an iterative reconstruction algorithm (two iterations and 24 subsets) on a 168 × 168 matrix size and a Gaussian filter cutoff of 5.0 mm. Images for patients 6 and 7 were obtained on another PET/CT scanner (Siemens Biograph mCT 64-slice). All image protocols were similar to those on the Biograph TruePoint, except for the matrix size, which was larger at 200 × 200. All PET images were decay-corrected to the time of injection. Each patient was asked to void their bladder before each scan to reduce their total radiation exposure. All [18F]FHBG PET images were obtained at the University of California, Los Angeles, CA. All PET scanners were fully calibrated and tested for quantitative accuracy; these procedures were performed according to the manufacturer’s recommendations.

In addition, brain [18F]FDG and MRI were performed before and after CTL infusions, using standard oncological protocols. These images were obtained at the City of Hope, Duarte, CA, and were scheduled within a few days of the [18F]FHBG imaging.

Image analysis

All data analyses were performed at Stanford University (CA), University College London (London, U.K.), and Hôpital de la Cité-de-la-Santé de Laval (Quebec, Canada). [18F]FHBG images were co-registered to MR images on MIM Encore version 6.4 (MIM Software) and on Siemens Inveon version 4.2 (Research Workplace). Automatic image registration was applied initially, and manual positioning was performed as necessary. [18F]FHBG images were analyzed in all axial planes and with different thresholds to detect subtle abnormal uptake. SUVs were measured in different areas using a 3D 50-mm-diameter spherical VOI for healthy tissue. Values were obtained for the left ventricle of the heart (referred to as blood pool); the left and right lung parenchyma; the right hepatic dome (referred to as liver); the left and right quadriceps musculature (referred to as thigh); the bone marrow; the left and right frontal lobes, parietal lobes, temporal lobes, and occipital lobes; and the cerebellum (table S2). Activity within all brain areas except the site of surgical resection or tumor recurrence was averaged to represent the background brain uptake. An irregular 3D VOI was drawn over the abnormal uptake area in the brain (tumor recurrence and/or CTL foci). We used the native volume growth algorithm from MIM Encore version 6.4 (MIM Software) and applied different thresholds for delineation. A 10% threshold of SUVmax was initially drawn, and 10% increments were subsequently applied until 90% of SUVmax was obtained. MR and [18F]FHBG PET images were always compared side-by-side to help in visual adjustment of boundaries. In addition, MR and [18F]FHBG PET images were fused together to optimally visualize the uptake area. To do this, rigid body registration with rotations was initially applied, followed by deformable registration. In most cases, the delineation on [18F]FHBG PET images followed the enhancing boundaries visualized on T1 postgadolinium sequence on MR images. For treatment response, we always considered the background activity as part of baseline normal brain uptake. Therefore, we subtracted background activity from the activity measured in each VOI.

Different strategies were used for evaluating therapy response. We used SUVmax, SUVmean, 3D volumes encompassing the tumor, individual voxel histogram analysis, and a parameter that takes into consideration the [18F]FHBG concentration and the VOI. This parameter, referred to as total activity, corresponds to the SUVmean multiplied by the VOI. To calculate the percentage of signal change between two scans, we used a formula that accounted for the background activity in the brain:Embedded ImageMovie rendering was performed with Amira version 5.2.2 (Visage Imaging Inc.).

Assessment of [18F]FHBG toxicity to transduced CTLs

The patient-derived CTLs GRm13Z40 (HD040-IL13z-ZFN) (control cell ID #cJ05105) and WCB-031-4C11-01 [UPN-031-IL13-zeta_DIIHyTK-pMG (pj00429#7)Cl.4C11] (cell ID #cj02331) were obtained from the T cell therapeutics laboratory at the City of Hope, Duarte, CA, as frozen vials. The CTLs were recovered in RPMI (Roswell Park Memorial Institute) medium with l-glutamine and Hepes. Human IL-2 (50 U/ml) was added as a supplement. Cells (15,000) for each cell type were plated in a 96-well tissue culture dish in RPMI medium. For the cells that needed to be exposed to the tracer, 7.4 kBq of [18F]FHBG was added to each well. One hour after exposure to the tracer, cells were washed with phosphate-buffered saline twice and either lysed for counting accumulated radioactivity or plated in medium for an additional 24 and 48 hours. Viability assessment was performed using the Presto Blue assay as per the manufacturer’s instructions (Thermo Fisher). Presto Blue uses the reducing environment of living cells to quantify viability. In brief, the desired volume of Presto Blue was added to the cells to a final concentration of 10%. The accumulated fluorescence was quantified using a microplate reader (excitation, 540 nm; emission, 600 nm) 30 min and 1 hour after the addition of Presto Blue.

Statistical analysis

Statistical analysis was performed in Microsoft Excel 2010 version 14.0.6 and GraphPad Prism Software version 5.0. Values were expressed as means ± SD. To compare uptake in healthy organs before and after CTL infusions and when comparing the effect of [18F]FHBG incubation on CTL proliferation rate, we used paired Student’s t test. For comparison of [18F]FHBG total activity measurement in lesions before and after CTL infusions for all patients, a paired Wilcoxon test was used. Probability values less than 0.05 were considered as statistically significant.


Fig. S1. [18F]FHBG uptake in parental and genetically modified CTLs.

Fig. S2. Assessment of FHBG toxicity to transduced CTLs.

Fig. S3. [18F]FHBG whole body biodistribution.

Fig. S4. SUVmax corresponding to tumor tissue in patients before and after CTL infusions.

Fig. S5. [18F]FHBG VOI delineation.

Fig. S6. Detailed threshold analysis to define the localization of [18F]FHBG radioactivity.

Fig. S7. SUVmean corresponding to tumor tissue in patients before and after CTL infusions.

Fig. S8. [18F]FHBG PET imaging in patient 2.

Fig. S9. [18F]FHBG PET imaging in patient 3.

Fig. S10. [18F]FHBG total activity in recurrent disease and the site of surgical resection.

Fig. S11. Post-CTL infusion [18F]FHBG PET and [18F]FDG PET imaging.

Fig. S12. [18F]FHBG PET images of a patient with a glioblastoma that originated from the dura mater.

Table S1. Patient survival data after diagnosis and initial CTL infusion.

Table S2. [18F]FHBG uptake in organs before and after CTL infusions.

Table S3. Injection protocol for patients 3 to 7.

Movie S1. Tumor-associated [18F]FHBG activity in the same patient before and after CTL administration.


  1. Acknowledgments: We thank G. Gowrishankar, X. Yan, A. Hoehne, and R. Reeves for technical assistance. Funding: This work was supported in part by National Cancer Institute (NCI) In vivo Cellular and Molecular Imaging Center P50CA114747 (S.S.G.), NCI RO1 CA082214 (S.S.G.), NCI RO1 CA135486 (S.S.G.), the Ben and Catherine Ivy Foundation (S.S.G.), NCI R01 CA155769 (B.B.), NCI R21 NS081594 (B.B.), NCI R21 CA189223 (B.B.), and Sangamo BioSciences. T.H.W. was supported in part by a Sir Henry Dale Fellowship jointly funded by the Wellcome Trust and the Royal Society (grant number 107610/Z/15/Z). Author contributions: The project was initially conceptualized and the entire project was supervised by S.S.G. The therapeutic trial leads were M.C.J., B.B., and C.B. The imaging study was designed by S.S.G. and S.Y. Image/data acquisition was performed by K.V.K., S.Y., A.K., R.M., J.W., J.R.W., S.F., C.B., M.A.-A., J.C., W.T., T.H.W., J.R., M.C.J., B.B., and S.S.G. K.V.K., T.H.W., S.Y., F.H., J.R., B.B., and S.S.G. participated in the design and/or interpretation of the reported experiments or results. T.H.W. and K.V.K. wrote the manuscript, which was then further refined by S.S.G. All other authors reviewed and edited the intermediate and final versions of the manuscript. Competing interests: T.H.W. is a non-equity holding consultant for CellSight Inc. C.B. and M.C.J. are inventors on patent application WO 02/088334 held by the City of Hope National Medical Center, which covers “Chimeric immunoreceptor useful in treating human cancers.” S.Y. and S.S.G. are cofounders and equity holders of CellSight Inc., which develops strategies to image cell therapies in humans. M.C.J. is on the scientific advisory board and an equity holder of Juno Therapeutics, which develops novel cancer immunotherapies. Data and materials availability: Various reporter gene constructs and cell products are available from S.S.G. under a material transfer agreement with Stanford University and/or the City of Hope National Medical Center.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article