Whole-body PET Imaging of T-cell Response to Glioblastoma

Glioblastoma (GBM) is a devastating fatal malignancy of the central nervous system and the most common type of histology representing up to 48% of all gliomas (1). It is uniformly difficult to treat, and the majority of patients experience disease relapse despite first-line intensive therapies such as surgical resections, radiotherapy, and chemotherapy with the DNA-alkylating agent temozolomide (2). Because GBM relapses can occur rapidly while also developing a resistance to therapeutic drugs, the median survival time is between 10 to 14 months (1, 3). Because of this poor prognosis and the associated diminished quality of life, more effective treatments are urgently needed.

Cancer immunotherapies are being increasingly investigated as a potential breakthrough for the treatment of various refractory malignancies (4). Among actively investigated immunocompetent cells, T cells play a major role in tumor eradication. Although immunotherapy in GBM is quite challenging due to tumor heterogeneity (5), capacities to subvert T-cell antitumor responses (6), and T-cell exhaustion (7), a number of preclinical and clinical attempts have been reported with successful results in the literature (8–11). There are two primary modes of immunotherapy being investigated with autologous T cells: ex vivo adoptive T-cell transfer (12, 13) and in vivo T-cell stimulating molecules. Adoptive T-cell transfer is a powerful personalized therapeutic approach with high specificity, but it is associated with higher cost and more resources per individual. T-cell targeting antibodies have been studied thoroughly and commercialized, as represented by anti–CTLA-4 antibodies and PD-1 antibodies (14, 15). OX40 (or CD134) is a promising activated T-cell surface marker, known to be a costimulatory transmembrane molecule of the tumor necrosis factor (TNF) superfamily, primarily expressed on activated effector T cells and regulatory T cells (Treg). Importantly, OX40 is not found on resting naïve T cells or the majority of resting memory T cells (16, 17). OX40 promotes proliferation and longer survival of effector and memory T-cell populations, while also suppressing the differentiation and activity of Tregs (18). Given these properties, an OX40 receptor monoclonal antibody (mAb) was tested as a treatment and showed preferable response in murine tumors (19), as well as glioma (20, 21). A third category of immune therapeutic methods is vaccination. Vaccinations prime the immune system against the tumor by introducing something highly immunogenic to the body, thus stimulating immunologic memory (2). Particularly, CpG-ODN, a Toll-like receptor (TLR) 9 agonist, is a promising stimulant, promoting maturation and activation of antigen-presenting cells and induction of Th1 T cells. We previously determined that direct injection of CpG and anti-OX40 antibody is a highly promising immunostimulatory regimen in subcutaneous murine lymphoma tumors (22). However, the effectiveness of local injection of CpG-ODN to GBM in humans has been previously reported as limited (23). On the other hand, subcutaneous vaccinations incorporating tumor cell lysates with dendritic cells were reported as an effective therapy for GBM in clinical trials (24, 25), possibly indicating that, in case of GBM, vaccination works more effectively when loaded at a remote site from the tumor than injected directly to the tumor. Indeed, multiple administrations of intradermal tumor lysate vaccine cocktail with CpG, followed by intraperitoneal administration of Fc-OX40L, was reported to exhibit a significant increase in the proliferation of brain-infiltrating CD4 and CD8 T cells followed by therapeutic effects in a murine glioma-bearing model (26).

While there has been much progress in developing immunotherapies, commercially available imaging tools capable of quantifying and monitoring response to these therapies remain limited in the clinical field. The well-established medical imaging systems CT and MRI can evaluate tumor morphology characteristics with adequate spatial resolution and can indirectly provide the information of cell density and vascularity. However, these approaches lack systemic molecular information. PET is a unique nuclear medicine modality, which enables real-time whole-body imaging of molecular targets and/or biochemical processes by administering a radiolabeled tracer. The most commonly used tracer is ¹⁸F-FDG, a glucose analogue radiolabeled with F-18. Because glucose metabolism is significantly elevated in malignant cells (27), the sensitivity of depicting tumor lesions using ¹⁸F-FDG-PET imaging is sufficient for most malignancies for the purpose of staging and monitoring therapeutic effect. The disadvantage, however, is that ¹⁸F-FDG exhibits low specificity, owing to accumulation in all hypermetabolic cells, including those found in the brain and active inflammatory cells including macrophages (28). Moreover, the half-life of F-18 is 109.7 minutes, and therefore, multiple tracer injections will be required for longitudinal observation, resulting in increased radiation dose. Recently, there have been several reports focused on engineering new tracers targeting CD8 T cells (29–31) or CAR T cells (32) using biologics labeled with longer half-life radioisotopes such as Cu-64 (12.7 hours) or Zr-89 (78.4 hours). Imaging lymphocyte trafficking with ex vivo labeled cells were reported (33), such as Th1 cells labeled with ⁶⁴Cu-PTSM (34) and γδ T cells with ⁸⁹Zr (35), but it takes more effort than in vivo labeling. Labeled antibodies are also used to track T cells in vivo. PET probes targeting checkpoint molecules such as PD-1/PD-L1/CTLA-4 have been developed as well to map the localization of checkpoint molecules expressing tumor-infiltrating lymphocytes in murine models (36) and clinical trials (37) to estimate the response for checkpoint inhibitors in various cancers. These imaging tools are aimed to predict the response to immune checkpoint inhibitors; however, emerging varieties of cancer therapy highlight the need of direct evaluation of immune cells. It is worthwhile to explore a method to depict the live T cells that act as key players to eradicate the tumor.

The specific functional expression of OX40 on T cells suggests that it could act as a biomarker of activated memory T cells working to eradicate tumors (38). We previously demonstrated activated tumor-infiltrating T cells using intratumoral vaccination to a lymphoma xenograft model using ⁶⁴Cu-DOTA-OX40 mAb (39). To expand this approach to a murine orthotopic glioma model, we treated mice with a subcutaneous cocktail vaccination of CpG, tumor lysate, and OX40 mAb, near left axillar lymph node, which is remote from the tumor. In addition, we utilized radiolabeled ⁸⁹Zr-DFO-OX40 mAb to achieve longer observation than ⁶⁴Cu-DOTA-OX40 mAb, and investigated the kinetics of whole-body immune response induced by a remote vaccination noninvasively and longitudinally by PET/CT. This is the first study to demonstrate the feasibility of visualizing and quantifying the whole-body immune response to a remotely introduced cancer vaccination for treating GBM.

The aim of the study was to explore the potential advantages of our newly developed ⁸⁹Zr-DFO-OX40 mAb PET/CT imaging approach for monitoring T-cell activation noninvasively in a glioma model using a remote immune-stimulating vaccination strategy.

All animal studies were conducted under a protocol approved by the Administrative Panel on Laboratory Animal Care (APLAC) of our institute. GL26 mouse glioma cells (gift in 2015 from Dr. Gerry Grant, Stanford University, Stanford, CA) labeled with Luc2-eGFP (GL26α8) were generated by isolation from the dissociated brain tumor using GFP as a selection marker to be implanted in multiple mice intracranially and monitored their growth noninvasively by bioluminescence of FL activities to select for the clones that had optimal in vivo growth properties. GL26α8 cells were cultured in DMEM with penicillin (100 U/mL), antibiotics and anti-mycotic (100 μg/mL), and FBS [10% (v/v)] and incubated at 37°C with 5% CO₂. Cells were used for in vivo experiments when they reached 80% confluence. Mycoplasma testing was performed every month. Orthotopic GL26α8 brain tumors were induced in 6- to 8-week-old female albino C57/BL6 mice (The Jackson Laboratory) by intracranial injection. Mice were anesthetized with 1.5% to 2.0% of isoflurane, sterilized with ethanol, and 10⁵ GL26α8 cells suspension in 3 μL PBS was injected in 3 mm depth from the surface, 3 mm right lateral to the sagittal suture, and 3 mm posterior to coronal suture. Postsurgical mice were monitored under a warm circumstance until they resumed normal behavior. Tumor engraftment was confirmed with bioluminescent imaging (BLI) with IVIS spectrum (Perkin Elmer) between 7 and 14 days after tumor inoculation. The BLI signal above 10⁵ (p/sec/cm²/sq) was defined as a baseline with at least two consecutive tumor growth confirmations.

Once tumor growth was confirmed, mice were randomly classified into vaccine group and control, and monitored every 3 to 4 days with BLI. We modified the previously reported vaccine strategy for glioma model (26). Vaccine group received a mixture of 50 μg of CpG-ODN (CpG) and 65 μg of GL26α8 tumor lysate in 100 μL PBS, and on the following day, 50 μg of anti-mouse OX40 mAb (clone: OX86, RRID: AB_1107592, Bio X Cell) in 100 μL PBS, in their left shoulders. Tumor lysate was prepared referring to the same prior study (26). The lysates were stored at −80°C freezer and thawed prior to injection. OX40 was injected 24 hours later, referring to a report that effect is well observed at that time point after in situ administration of CpG in subcutaneous lymphoma model (40). Control mice received 100 μL of PBS on both days.

Briefly, 1 mg of anti-mouse OX40 antibody or isotype control (rat IgG1, clone: HRPN, RRID: AB_1107775, Bio X Cell) was diluted to 1 mg/mL with PBS, then buffer exchanged with pH 8.8–9.0 PBS solution. Subsequently, 10-fold excess of p-SCN-Bn-deferoxamine (DFO) was added to the antibody solution and incubated for 60 minutes at 37°C. Conjugated product showed 2.49 ± 0.12 molecules of DFO per mAb. Finally, the mixture was buffer exchanged with PBS (pH 7.4) using Vivaspin 2, 50K cutoff centrifugal concentrator. The concentration of DFO conjugated mAb was measured by Thermo Scientific NanoDrop One Microvolume UV-Vis Spectrophotometer. Na₂CO₃ (1M solution) was added to ⁸⁹Zr-oxalic acid until the pH reached 7.1–7.8, and subsequently added to DFO-OX40 mAb or DFO-isotype control solution. After 60-minute incubation at 37°C, ⁸⁹Zr-DFO-OX40 mAb was then purified by Zeba Spin Desalting Columns, 7K MWCO, 0.5 mL, 1,000 × g, 1 minute. Finally, radiochemical purity was assessed and confirmed >95% by iTLC (Instant Thin Layer Chromatography).

⁸⁹Zr-DFO-OX40 mAb or ⁸⁹Zr-DFO-rat IgG (1.5–2 MBq) was administered intravenously to orthotopic mice via tail vein, and sequential standalone PET/CT scans were acquired three times; day 1, day 2, and day 5. Time points were selected considering the physical half-life of ⁸⁹Zr and typical long circulation times of full-length antibodies (41). PET/CT studies were performed on an Inveon microPET/CT rodent model scanner, carrying a cone beam X-ray source and a flat panel detector on CT (Siemens Medical Solutions USA, Inc.). Anesthesia was induced with 2.5% isoflurane and maintained at 2%, and mice were placed in a prone position in the scanner. Ten minutes of attenuation correction of CT scan was performed just before the 15-minute static 3D PET scan. Image reconstructions were carried out on an Inveon Acquisition Workplace (Siemens Preclinical Solutions) workstation using an ordered subset expectation maximization 3D/maximum a posteriori reconstruction algorithm.

Quantitative region-of-interest (ROI) analysis of the PET images was performed on attenuation and decay corrected PET images using Inveon Research Workplace 4.2 (Siemens Preclinical Solutions). Tissue uptake values are presented as the percent injected dose per cc (%ID/cc). The spatial resolution of the Inveon PET scanner is 1.4 mm and a previous study showed that %recovery coefficient was 91% for 3 mm-size rod (42). As such, it is suggested that the tracer uptake in target of interest with <3 mm is likely to be underestimated. The maximum value in the targeted region of interest was calculated of day 1 to 5 scans to observe the tracer distribution change over time. Bilateral cervical and axillary lymph nodes plus spleen were regarded as representative lymphoid organs in this study as vaccine was injected near left axillary area and the tumor draining lymph node (TLDN) was right cervical lymph node in our mice. The maximum uptake value of these sites was collected together with representative solid organs such as the heart, the liver, the muscle, and the bilateral brain hemispheres. ROI was drawn to avoid the tracer uptake from surrounding organs referring to corresponding CT scan, when they were intensely seen. The volumetric information of whole-body lymphoid organs was incorporated for the final day 5 scan to observe the total body OX40 activity.

To accurately quantify the extent of tracer binding, we employed a method that has been commonly researched as an index of treatment response in human FDG-PET scans: total lesion glycolysis (TLG; ref. 43). TLG is a multiplication of the metabolic volume which shows pixels above a defined threshold uptake (42% of maximum uptake is the most frequently applied threshold in clinical studies; ref. 44) and mean uptake value of the volume. In our study, we defined the multiplication of the volume above 42% of maximum uptake and mean uptake value of the volume as “total lesion immune response” (TLIR; Supplementary Fig. S1). Once volume of interest (VOI) was set to a targeted lesion, the dedicated software automatically calculated the volume above 42% of maximum uptake, mean uptake in the calculated volume, and their multiplication. This analysis was performed on OsiriX MD version 8.0.1.

Immediately after the day 5 scan, mice were euthanized and blood, tumor, lymph nodes, and all major organ/tissues were collected and weighed. The radioactivity of each tissue was counted in an automated γ-counter (HIDEX), and the tissue accumulations were calculated as %ID/g (mean ± SEM). Flow cytometry was performed to the tumor, lymph nodes, and spleen in several mice immediately after biodistribution for analysis of OX40/CD4 expression.

Cells were first stained with Live/Dead aqua fixable cell stain (Molecular Probes L34957) following the manufacturer's instructions. Cells were then washed in FACS buffer (PBS and 2% BSA) and stained with a panel of antibodies including CD3-APC (hamster IgG1 k, clone:145-2C11, BD Biosciences), CD4-APC-Cy7 (rat IgG2b k, clone GK1.5:, BioLegend), CD8-PE (rat IgG2a k, clone: 53-6.7, BioLegend), and OX40-BV421 (rat IgG1 k, clone: OX-86, BioLegend). Cells were fixed in 2% paraformaldehyde and stored at 4°C prior to analysis using an LSR II flow cytometer (BD Biosciences). Compensation was performed utilizing control beads (Thermo Fisher Scientific) and data were analyzed using the FlowJo software.

To see the uptake of ⁸⁹Zr-DFO-OX40 mAb tracer reflects the exact brain tumor, brains of a few mice were compared with autoradiograph (ARG) and MRI. To obtain ARG images, mice brains were fresh frozen in optimum cutting temperature compound (O.C.T., Sakura Finetek Inc.) and sliced at 40-μm thickness using a cryostat (Microm). Tissues were exposed to an Imaging Plate (Fujifilm) and kept at −20°C in the dark for 2 weeks prior to scanning using a Typhoon phosphorimager (Amersham Biosciences). MRI T2-weighted images (echo time: 33 ms; retention time: 2,500 ms; 2 averages; 17 slices) were obtained following PET on day 2 using a 7T Varian Magnex Scientific MR system under 1.5% to 2.0% of isoflurane in anesthetized mice as reported previously (45). Fused images were created with Inveon Research Workplace 4.2.

The comparison of ⁸⁹Zr-OX40 mAb uptake between treated and vehicle, or uptake of ⁸⁹Zr-DFO-OX40 mAb versus isotype control over repeated PET scans, two-way ANOVA was applied as an analyzing method of repeated measures data. Mann–Whitney U test was used in comparison between vaccinated and control groups in terms of the spleen weight, FACS analysis, and best treatment response ratio in BLI. This method was also applied to compare best treatment response ration of BLI and TLIR in low/high–tumor burden classes. Spearman rank correlation test was applied for relationship between PET signals and biodistribution measurements or BLI signal or BLI ratio. GraphPad Prism 8.0.2 (RRID:SCR_002798, GraphPad Software) and R 3.5.1 were used. Two-sided P values under 0.05 was considered as statistically significant.

Our conceptual study design is described in Fig. 1A. In brief, we prepared orthotopic glioma mice and subsequently imaged them using immunoPET to assess the spatiotemporal dynamics of T-cell activation following administration of a cancer vaccine at a tumor remote site (left shoulder). This strategy is described in Fig. 1B. For 10 to 14 days following glioma inoculation, tumor growth was monitored using BLI. When tumor growth was confirmed with a growing signal of BLI twice within 4 days and a signal above 10⁵ (p/sec/cm²/sr) at second BLI, mice received a combined subcutaneous vaccination of CpG-ODN and glioma tumor lysates, and OX40 mAb, 5 and 6 days respectively, prior to the administration of ⁸⁹Zr-DFO-OX40 mAb immunoPET tracer. After injection of ⁸⁹Zr-DFO-OX40 mAb via tail vein, mice underwent PET/CT scans on days 1, 2, and 5. As expected, vaccinated mice exhibited marked ⁸⁹Zr-DFO-OX40 mAb uptake in multiple lymph nodes (LN) 5 days after tracer administration, which was not seen in control mice, whereas intense tracer uptake was revealed in the brain tumor in controls (Fig. 1C; Supplementary Videos S1 and S2). Vaccinated mice showed enhanced PET signal not only in the left axillary LN, which is the draining lymph node from the vaccinated site, but also in other remote areas including the right cervical LN, which is regarded as the TDLN of inoculated glioma in the ipsilateral hemisphere. This result is in agreement with previous reports that suggest antigens transferred to both superficial and deep areas of the brain drain to cervical lymph nodes (46, 47). The representative time course of ⁸⁹Zr-DFO-OX40 mAb PET signal in regions of interest from days 1 to 5 in vaccinated and control mice is shown in Fig. 1D. Bilateral lymphatic chains were exhibited in the vaccinated mice, whereas the control mouse showed a substantial tumor in the brain. We considered these organs to play a major role in immune reaction against the tumor in our model: left axillary LN as a treated site, right cervical LN as a TDLN, contralateral sites as routes of migration of activated T cells, and the spleen as a reflection of enhanced systemic T-cell activity. PET uptake in these lymph nodes and the spleen was compared between groups as well as that of the brain tumor (Fig. 1E). The highest PET signal was observed in the left axillary LN and the spleen and was significantly elevated in vaccinated (left axillary LN, P < 0.0001; spleen, P = 0.020) compared with control mice on day 5. In particular, the left axillary LN demonstrated a nearly 3-fold uptake in the vaccinated group relative to control group. Uptake in right cervical LN as a TDLN also showed higher uptake compared with control, while the brain tumor uptake was significantly higher in the control (P = 0.042). Significant splenomegaly was also seen in vaccinated mice (P = 0.019; Fig. 1F). Correlogram of ⁸⁹Zr-DFO-OX40 mAb accumulation in individual mice revealed that the LN that is closest to the vaccination site exhibited the highest OX40 mAb signals, followed by the spleen, TDLN, right axillary LN, and left cervical LN (Fig. 1G).

Next, we examined whether ⁸⁹Zr-DFO-OX40 mAb-PET imaging could detect immune activation after vaccination at earlier time points (Supplementary Fig. S1). To investigate this, we administered ⁸⁹Zr-DFO-OX40 mAb to orthotopic GBM mice the day after vaccination, and performed three subsequent PET scans 1 to 4 days following tracer injection. Interestingly, neither vaccinated nor control group showed enhanced PET signal in lymphoid organs until day 4 (P > 0.05 in all cases).

To confirm that PET-positive lymphoid organs was due to presence of OX40⁺ cells, flow cytometry was performed on the left axillary LN, left cervical LN, spleen, and tumor (Fig. 2A). No significant difference in CD4 expression was observed in any organs between vaccinated and control groups. However, OX40⁺ CD4⁺ T cells were significantly increased in the left axillary LN (P = 0.017), left cervical LN (P = 0.036), and the spleen (P = 0.036) of vaccinated mice (Fig. 2B), suggesting that OX40 PET signal corresponds to the number of OX40⁺ CD4⁺ T cells. No major difference in OX40 expression was seen in the tumor (P > 0.05).

To further assess the specificity of ⁸⁹Zr-DFO-OX40 mAb and its binding in lymphoid organs, immunoPET using an ⁸⁹Zr-labeled isotype control antibody (⁸⁹Zr-DFO-rat IgG1 mAb) was performed. ⁸⁹Zr-DFO-OX40 mAb showed significantly higher uptake in bilateral axillary LNs (left, P = 0.0083; right, P = 0.028), and the spleen (P = 0.026) toward day 5, compared with ⁸⁹Zr-labeled isotype control antibody (Fig. 3). This elevated uptake in LNs between ⁸⁹Zr-DFO-OX40 mAb and isotype control was not observed in control mice. In the spleen, nonspecific uptake of isotype control was observed in both vaccinated (7.2 ± 1.5 %ID/cc) and control mice (6.4 ± 0.6 %ID/cc); however, it was on average less than half and enough to be differentiated from significant ⁸⁹Zr-DFO-OX40 mAb uptake in the vaccinated mice (16.7 ± 3.2 %ID/cc). These results indicate that the ⁸⁹Zr-DFO-OX40 mAb PET signal is specifically visualizing OX40 expression within lymphoid organs. Liver uptake was significantly higher (P = 0.016) and heart showed trend of less uptake on the day 5 scan in ⁸⁹Zr-DFO-OX40 mAb PET, compared with the isotype control in the vaccinated group, possibly suggesting faster clearance of ⁸⁹Zr-DFO-OX40 mAb tracer upon vaccination. As for the tumor (right brain) and contralateral (left) brain tissue, no major difference was observed between ⁸⁹Zr-DFO-OX40 mAb and isotype PET signals in both vaccinated and control groups although in the control group higher mean ⁸⁹Zr-DFO-OX40 mAb uptake was observed relative to isotype control in the tumor. Biodistribution studies revealed similar trends; the left axillary LN exhibited significantly higher ⁸⁹Zr-DFO-OX40 mAb signal in vaccinated mice than isotype control (P < 0.0001), and the spleen also showed same tendency (no significance). Liver count was higher and blood count was lower in ⁸⁹Zr-DFO-OX40 mAb group of the vaccinated mice, suggesting ⁸⁹Zr-DFO-OX40 mAb distribution kinetics are more rapid within vaccinated animals (vs. nonvaccinated) and compared with the kinetics of the isotype control tracer. PET signal was confirmed to be well correlated with counts from biodistribution in both solid organs and lymph nodes (R² = 0.83 and 0.55, respectively).

To confirm that the location of ⁸⁹Zr-DFO-OX40 mAb uptake in the brain corresponded to inoculated glioma, coronal PET/CT was compared with MRI on day 2 and ARG on day 5. PET signal corresponded with the exact location of tumor as confirmed by MRI and ARG (Fig. 4). PET signal in the tumor on day 5 also correlated with the BLI signal of the right hemisphere in both vaccinated (R² = 0.44) and control mice (R² = 0.85). These results indicate that ⁸⁹Zr-DFO-OX40 mAb accumulated in the glioma itself as well. This may be attributed to blood brain barrier (BBB) permeability, and the extent of tracer accumulation might depend on the actual tumor volume.

Brain tumor burden was monitored with BLI after a combined vaccination. Representative images of BLI from treatment days 0 to 8 are shown in Fig. 5A. We divided the mice into two classes according to the tumor burden on the day of vaccination: low burden with mean signal of lower than 10⁶ (p/sec/cm²/sr; vaccinated, n = 12; control, n = 9), and high burden with higher than 10⁶ (p/sec/cm²/sr; vaccinated, n = 9; control, n = 7). Treatment response was most effective in the low-burden class with one third of vaccinated mice (n = 4) demonstrating >90% decrease in tumor volume, which was not observed in the control mice (Fig. 5B). In contrast, vaccinated mice in the high-burden class exhibited <25% decrease in tumor signal. Accordingly, treatment response was found to be significant in low-burden (P = 0.015) but not in high-burden mice when compared with control mice (Fig. 5C). In addition, we investigated whether administering ⁸⁹Zr-DFO-OX40 mAb affected treatment response in any way. Importantly, there was no significant difference in BLI signal change between mice that received ⁸⁹Zr-DFO-OX40 mAb tracer dose injection (20 μg antibody) with those that received the isotype control mAb (Fig. 5D).

Because we observed that vaccinated mice showed distinct patterns of ⁸⁹Zr-DFO-OX40 mAb binding in lymphoid organs (LNs and the spleen), we compared total ⁸⁹Zr-DFO-OX40 mAb accumulation in lymphoid organs of vaccinated mice with either low (n = 4) and high tumor burden (n = 5). An anatomic depiction of LNs and other lymphoid tissues is shown in Fig. 6A. We observed enhanced PET signal in brachial and inguinal LNs in several vaccinated mice. ⁸⁹Zr-DFO-OX40 mAb PET MIP images of mice on day 5 are shown in Fig. 6B. All mice exhibited more than two visually detectable LNs and/or spleen, which indicates that the vaccine consistently works by stimulating lymphoid organs. To quantify the strength of the immune response, we introduced an index named TLIR, which was defined as a mixed index of volume and intensity of ⁸⁹Zr-DFO-OX40 mAb PET signal (Supplementary Fig. S2). TLIR in bilateral LNs of axillary, brachial, cervical, and inguinal area, and the spleen were calculated and compared between low– and high–tumor burden groups (Fig. 6C). TLIR was significantly higher in the spleen (P = 0.032) and right inguinal LN (P = 0.016) of high-burden mice, while TLIR of right cervical LN (TDLN) was higher on average for low-burden mice. TLIR ratio of TDLN/other lymphoid organs was 0.27 and 0.08 in low- and high-burden mice, respectively. The sum of TLIR and percentage decrease of BLI signal per individual mouse are shown in Fig. 6D as indicators of whole-body OX40-powered lymphatic tissue and treatment response of vaccine. Within the low–tumor burden group, two out of four mice had an almost complete response (≥99% decrease in BLI signal) along with whole-body (WB)-TLIR > 1,500 × 10⁻³ %ID. In contrast, only one mouse showed a weak treatment response (i.e., 19% decrease). Nevertheless, all mice exhibited >2,000 × 10^–3 %ID of WB-TLIR in the high–tumor burden group of mice. Taken together, these results demonstrate that when the tumor burden is low, even a modest amount of OX40⁺ activated T cells in lymphatic tissues is expected to assure a favorable treatment response after a single shot of mixed vaccine. Alternatively, if the tumor burden is already high at the time of vaccination, even a large systemic OX40-positive immune response is not sufficient to induce tumor shrinkage.

To evaluate whether LN response can predict vaccine treatment response, correlation between the uptake of ⁸⁹Zr-DFO-OX40 mAb on day 5 in each bilateral axillary LN/cervical LN/spleen and the best treatment response was calculated according to low– and high–tumor burden groups (Fig. 6A). In the low-burden class, a significant correlation was observed in left axillary LN (r = −0.94, P = 0.017), which was the closest LN from vaccination site, while the other LNs and the spleen did not show any correlation. No correlation was seen in the high-burden class at all sites. This indicates that ⁸⁹Zr-DFO-OX40 mAb accumulation in the vaccinated site can be utilized as a treatment predictor in the context of this vaccine treatment strategy.

The difficulties of treating GBM have motivated the investigation of different promising immunotherapies and new combination treatment strategies. As such, antitumor effects of therapeutics targeting various immune-cell surface markers are being assessed. OX40 is one such promising target for T-cell activating agonistic antibodies, together with 4-1BB, GITR, ICOS, CD27, and CD30 (4, 48). Several anti-OX40 agonists are already being implemented in phase I or II cancer clinical trials (49). However, there are fewer tools under investigation to directly image the expression of these immune-cell markers; thus, the evaluation of immunotherapy efficacy has been limited. We previously reported the successful development of two highly specific immune system visualization techniques. The first involved the generation of ⁶⁴Cu-DOTA-OX40 mAb, which we showed was sensitive and specific for imaging OX40⁺ activated T cells in a mouse model of syngeneic lymphoma (39) and acute graft-versus-host disease (50) using PET. The second involved the development of a PET tracer, known as ⁸⁹Zr-DFO-ICOS mAb, for detecting ICOS⁺ T cells in a murine lung cancer model (51) and CAR T cells (52). Following these promising results, we extended our pursuit of imaging T-cell activation markers to a more challenging type of tumor, GBM, and set out to quantify the characteristics of activated immune cells in the whole body after remote vaccination.

The current work demonstrates that ⁸⁹Zr-DFO-OX40 mAb PET specifically delineates stimulated lymphoid organs following administration of a cancer vaccine applied in an orthotopic GBM model. After receiving combined cancer vaccine in the remote site, multiple LNs and spleen, particularly the LN proximal to the vaccination site, exhibited robust ⁸⁹Zr-DFO-OX40 mAb PET signals, reflecting an increased number of CD4⁺ OX40⁺ T cells. When GBM burden was low, a single pair of mixed vaccines led to a significant decrease in the tumor signal in BLI while requiring less boost of immune stimulation indicated by TLIR in whole-body lymphoid organs. Of note, the tumor shrinkage was correlated with the ⁸⁹Zr-DFO-OX40 mAb PET signals at LN nearest to the vaccinated site.

Consequently, immunoPET with ⁸⁹Zr-DFO-OX40 mAb was quite successful in visualizing stimulated LNs and the spleen, with a combined vaccination. Five days after tracer injection, all vaccinated mice showed vivid ⁸⁹Zr-DFO-OX40 mAb uptake in the vaccinated LN with a high contrast to blood pool. Accumulation within other LNs and the spleen varied among individual mice, but overall, every animal exhibited multiple hot regions of marked tracer uptake within lymphoid organs (i.e., lymph nodes and/or spleen). This directly shows that the immune activation triggered by the remote vaccination not only occurred at the vaccinated site but also systemically. Calzascia and colleagues demonstrated that the anatomic location in which antigen-specific lymphocytes are activated may “imprint” their ability to traffic to their ultimate destination (53); specifically, naïve T cells activated in the cervical lymph nodes demonstrate a “central nervous system homing” phenotype that may be influenced by T-cell expression of the α₄β₁ integrin (54). In this study, we found that the vaccination induced a trend toward higher ⁸⁹Zr-DFO-OX40 mAb uptake in tumor-draining LN, following by the nearest LN from vaccinated site and the spleen. In addition, we observed that a single vaccination worked efficiently in mice with low-burden GBM with less total volume of immune activity compared with large tumors by calculating WB-TLIR. Conversely, when a tumor was large [>10⁶ (p/sec/cm²/sr)], little tumor shrinkage was induced nevertheless even stronger immune reaction quantified as TLIR was introduced, especially at the spleen and remote LNs from the brain such as inguinal LN, as shown in ⁸⁹Zr-DFO-OX40 mAb PET. TDLN was the only lymphoid organ that demonstrated higher ⁸⁹Zr-DFO-OX40 mAb uptake in the low–tumor burden group, compared with high–tumor burden group. The reason of the higher uptake in TDLN in the low–tumor burden group is potentially explained as follows. Low tumor burden showed a significant response to immunotherapy, which was in line with previous clinical studies. That is, the low baseline tumor volume was associated with better overall survival in immune checkpoint blockade users with other cancers such as melanoma (55) and non–small cell cancer (56). Moreover, PD-1 inhibitor pembrolizumab brought survival benefit to recurrent patients with GBM when the tumor burden was reduced with surgical resection (57). These clinical studies suggest that the smaller tumor volume is related to better response to immunotherapy. On the other hand, recent evidence points to the need for the recruitment of newly primed and peripherally (e.g., in TDLN) expanded effector T cells to ensure efficacy of immunotherapy (58). Clinical efficacy and durability of antitumor immunity were reported a relation with elevated frequencies of central-memory or early-effector T cells with the ability to home to lymph nodes (59). These reports imply the accumulation of primed cytotoxic OX40⁺ T cells in TDLN is a factor for good therapy response. Taken altogether, we consider low–tumor burden group showed higher uptake in TDLN because both factors were associated with responders. However, this result is of limited value because there was a large overlap in TDLN uptake between low– and high–tumor burden groups without any significance. We should validate if the responders truly show high ⁸⁹Zr-DFO-OX40 mAb uptake in TDLN in a larger number of cohorts in further studies. We also briefly showed the correlation of higher ⁸⁹Zr-DFO-OX40 mAb uptake in the LN nearest from the vaccinated site and the better treatment response when the tumor was small. We assume this correlation implies that the PET signals at the vaccinated LN reflect the combined effect of T-cell priming and migration to the tumor sites. Conversely, in nonresponders, the vaccination stimulated the lymphoid organs but supposedly failed in priming and clonal expansion of T cells sufficiently. Considering above, vaccination may be preferred at the tumor-draining LN to facilitate the priming of T cells for the best therapeutic response, although we need to verify this in future studies. However, it is meaningful to confirm that the vaccination at a remote LN was effective, as the tumor-draining LN is not always accessible in the clinical practice, especially in cases of GBM. All in all, our data indicate that the strongest immune response in the whole body occurs at the nearest LN from the vaccinated site, which may help to predict the real treatment response. These phenomena might be a potential reference for future perspective of how immunotherapies help contribute to cancer therapies.

Several limitations are included in this study. We utilized only the GL26 glioma cells that were labeled with the Luc2-eGFP reporter to avoid the issue of immunogenicity. This subline does not respond to innate immune system that may affect the interpretation of acquired data. Another limitation of this study was that the remote vaccination site was fixed at the subcutaneous contralateral shoulder. If the vaccination was tried in other parts of the body, ⁸⁹Zr-DFO-OX40 mAb PET may have shown a different distribution of enhanced OX40-positive lymphoid organs. Moreover, we did not test different doses or optimize the number of shots of the vaccination protocol or conduct a follow-up study to assess overall survival, as it takes several days to observe animals with PET while the tumor is rather progressive when the initial vaccination was not effective. Consequently, the treatment effect was limited to the “low” tumor burden group, and it was not explored if the higher tumor burden group responds to more intensive treatment regimens. Furthermore, direct evaluation of intratumoral lymphocytes in between responders and nonresponders were unachievable due to the rapid shrinkage of tumor in responders; therefore, we could not corroborate the response resulting from the migrated cytotoxic T cells. Finally, we included OX40 mAb in the vaccination regimen but did not verify in this study if our OX40 imaging approach could also be employed to monitor other types of vaccination; however, OX40-targeted immunoPET will be implemented in other vaccination regimens without OX40 mAb for other diseases, considering our previous study (39) showing the clear ⁶⁴Cu-DOTA-OX40 mAb uptake in the lymphoma in in situ CpG treatment group, followed by tumor shrinkage. Validation of OX40 PET imaging is warranted in future studies. The aim of this study was focused on evaluating the utility of OX40 PET imaging to visualize how the systemic lymphoid organs respond following vaccination and how that response predicts and contributes to the therapeutic effect in a murine glioma model. Further analysis will be required to confirm the optimal strategy in the context of vaccination therapy.

As OX40 imaging was capable of monitoring lymphoma model without OX40 mAb vaccination in our previous study (39), this imaging approach is expected to be applied to various clinical occasions including the monitoring of other types of vaccination to malignancy, but also other benign diseases associated with immune rejections such as GVHD. Dosimetry studies will be performed when translated into clinics. One of major caveats upon translation to clinics is a potential unknown biological effect of ⁸⁹Zr-DFO-OX40 mAb to the patient immune system. To avoid any immunogenicity, engineering of OX40 antibody may be needed in the future. The rising clinical demand of immunotherapies highlights the importance of an accurate imaging strategy to noninvasively evaluate the distribution of stimulated immune-related organs longitudinally. The successful observation of stimulated lymphoid organs in this murine glioma model further supports the value of ⁸⁹Zr-DFO-OX40 mAb PET and encourages future human study.

S.S. Gambhir reports grants from NIH and grants from Ben and Catherine Ivy Foundation during the conduct of the study. No disclosures were reported by the other authors.

T.W. Nobashi: Conceptualization, data curation, formal analysis, investigation, visualization, methodology, writing–original draft, writing–review and editing. A.T. Mayer: Conceptualization, data curation, supervision, investigation, visualization, methodology, writing–review and editing. Z. Xiao: Data curation, formal analysis, methodology. C.T. Chan: Supervision, methodology, writing–review and editing. A.M. Chaney: Resources, data curation, software, visualization, methodology. M.L. James: Resources, supervision, methodology, writing–review and editing. S.S. Gambhir: Resources, supervision, funding acquisition, project administration.

S.S. Gambhir reports grants from National Cancer Institute (R01 CA201719-05). The research was supported in part by the Ben & Catherine Ivy Foundation (to S.S. Gambhir). T.W. Nobashi was supported by Wagner-Torizuka Fellowship 2017-2019 sponsored by Nihon Medi-Physics Co., Ltd. We presented a portion of this study and received 2nd place of Young Investigator Award in the Center for Molecular Imaging Innovation & Translation at SNMMI 2020. The authors would like to thank the Stanford Center for Innovation in In-Vivo Imaging (SCi3) and the Stanford shared FACS facility for their support. In addition, we are grateful to Drs. Israt S. Alam and Gayatri Gowrishankar for supporting in vivo imaging studies. We also would like to dedicate this work to the memories of Dr. Sanjiv Sam Gambhir (1962-2020) and the valuable mentorship he provided.

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.