Compelling evidence points to immune cell infiltration as a critical component of successful immunotherapy. However, there are currently no clinically available, noninvasive methods capable of evaluating immune contexture prior to or during immunotherapy. In this study, we evaluate a T-cell–specific PET agent, [18F]F-AraG, as an imaging biomarker predictive of response to checkpoint inhibitor therapy. We determined the specificity of the tracer for activated T cells in vitro and in a virally induced model of rhabdomyosarcoma. Of all immune cells tested, activated human CD8+ effector cells showed the highest accumulation of [18F]F-AraG. Isolation of lymphocytes from the rhabdomyosarcoma tumors showed that more than 80% of the intratumoral signal came from accumulation of [18F]F-AraG in immune cells, primarily CD8+ and CD4+. Longitudinal monitoring of MC38 tumor-bearing mice undergoing anti-PD-1 treatment revealed differences in signal between PD-1 and isotype antibody–treated mice early into treatment. The differences in [18F]F-AraG signal were also apparent between responders and nonresponders to anti-PD-1 therapy. Importantly, we found that the signal in the tumor-draining lymph nodes provides key information about response to anti-PD-1 therapy. Overall, [18F]F-AraG has potential to serve as a much needed immunomonitoring clinical tool for timely evaluation of immunotherapy.
These findings reveal differences in T-cell activation between responders and nonresponders early into anti-PD-1 treatment, which may impact many facets of immuno-oncology, including patient selection, management, and development of novel combinatorial approaches.
By the time they are diagnosed, most cancers have already developed mechanisms by which they evade control by the immune system (1, 2). Immunotherapy, a rapidly advancing field, aims to overcome the immunosuppressive environment in the tumors by utilizing patients’ own immune defenses. One type of immunotherapy, checkpoint inhibitors, uses mAbs against surface proteins that serve as checkpoints or regulators of the immune response. Checkpoint inhibitor therapy has led to impressive clinical successes, providing objective and durable responses in patients with advanced cancers that previously had very few treatment options. Unfortunately, immunotherapy works only in a relatively small fraction of patients with solid tumors (3). Although the reasons for immunotherapy failure are not entirely clear, it is believed that the immune activity within tumors plays a crucial role. Numerous studies have shown an association between tumor-infiltrating T cells and clinical prognosis in many solid cancers (4–7). Pathologic examination of tumor biopsies revealed three basic cancer-immune phenotypes: immune inflamed, immune excluded, and immune desert tumors (6, 8). Not surprisingly, inflamed tumors, characterized by high numbers of immune cell infiltrates in the tumor and its margin, show the best response to immunotherapy. However, even within the inflamed phenotype, there is a wide variation in response to therapy, indicating the existence of other factors, such as immune cell migration, activation, survival, proliferation, which can affect immunotherapy outcome (8, 9). Despite the vital role that the immune infiltration plays in clinical outcome, in the clinic there are currently no noninvasive immunomonitoring methods capable of evaluating immune contexture prior to or during immunotherapy in the clinic.
RECIST for immune-based therapeutics (iRECIST), currently used in the clinic for evaluation of immune response, aim to capture the response patterns unique to immunotherapeutics, but only assess changes in the tumor burden (10). The examination of biopsy specimens for the presence of immune-related biomarkers is not well suited for immunomonitoring purposes because of the variability in tissue sampling, invasiveness of biopsy procedures, and inability to inform on the complex immunologic responses in the whole body. A noninvasive, immune-specific, whole-body imaging technique has the capability to enable immunomonitoring and thus provide valuable information on the patient-specific immune status and immune response needed to achieve desired clinical outcomes.
[18F]F-arabinofuranosyl guanine (AraG) was developed by Namavari and colleagues, as a PET-imaging agent for activated T cells (11). It is a 18F-labeled analogue of arabinofuranosyl guanine (AraG), a compound that has shown remarkably selective accumulation in T cells (12, 13). Nelarabine, AraG's prodrug, has been approved by the FDA for treatment of T-cell acute lymphoblastic leukemia and T-cell lymphoblastic lymphoma. [18F]F-AraG can be phosphorylated, and trapped intracellularly, by two enzymes whose activity is upregulated in activated T cells—cytoplasmic deoxycytidine kinase (dCK) and deoxyguanosine kinase (dGK; Fig. 1; ref. 14). However, because dGK has a higher affinity for [18F]F-AraG (Supplementary Fig. S1), we expect [18F]F-AraG at tracer level to be preferentially phosphorylated by the mitochondrial kinase. Numerous studies demonstrate a critical role mitochondrial activity plays in T-cell activation and function (15–17). As a substrate for mitochondrial dGK, [18F]F-AraG seems to be an agent uniquely suited to report on T-cell activation and proliferation during immunotherapies.
Materials and Methods
Uptake studies in murine immune cells
Spleens and lymph nodes of C57Bl6 mice (Jackson Lab) were harvested and primary CD4 and CD8 T cells negatively isolated using the EasySep Mouse T Cell Isolation Kit (StemCell Technologies). Cells were incubated with 10 U IL2 (PeproTech) and left untreated or activated via immobilized αCD3 (2.5 μg/mL) and soluble αCD28 (2.5 μg/mL) antibodies for 24, 48, or 72 hours (BD Biosciences). Bone marrow–derived macrophages (BMM) were generated from the bone marrow of B6 mice and left untreated or activated with 1 μg/mL lipopolysaccharide (Escherichia coli LPS; serotype O111:B4; Sigma) for 24 or 48 hours. B cells were isolated from B6 mice spleens using the EasySep Mouse B-Cell Isolation Kit (StemCell Technologies) and left untreated or stimulated with 1 μg/mL LPS for 24 or 48 hours. Bone marrow–derived dendritic cells (BMDC) were generated from the bone marrow of B6 mice. Monocytes from bone marrow were isolated with the CD14+ Selection Kit (Miltenyi Biotec) and cultured with 100 ng/mL GM-CSF (PeproTech) for 6 days. Nonadherent BMDCs were collected and then either stimulated with LPS or left untreated. Cell viability was determined by Trypan blue dye exclusion.
To determine in vitro tracer uptake, 3 × 106 cells were incubated with 1 μCi/mL [3H]F-AraG (Moravek) in in RPMI1640 supplemented with 10% heat-inactivated FBS (Hyclone), sodium pyruvate (UCSF Cell Culture Facility), nonessential amino acids (VWR Scientific), and penicillin/streptomycin (Hyclone), R10 medium, for 2 hours at 37°C. After centrifugation at 300 × g, cells were washed with PBS (GE Healthcare) and lysed with RIPA buffer (VWR Scientific). To determine retention, cells were transferred to fresh R10 medium after 2 hours of incubation with [3H]F-AraG. Cells were incubated for 1 hour at 37°C to allow for nonphosphorylated [3H]F-AraG to efflux, wash, and lyse. Ecolite Scintillation Cocktail (MP Biomedical) was added to the lysed cells, and after an hour, radioactivity counted on the Beckman LS6500 (Perkin Elmer).
FACS was performed to confirm the purity and activation status of immune cell phenotypes. Cells were stained with fluorophore-conjugated antibodies and analyzed on a FACSARIA II or III (BD Biosciences) and collected data further analyzed using FlowJo software (FlowJo). Cell viability for FACS was determined by staining primary isolated cells with an amine reactive dye (Violet Live/Dead Viability Kit; Thermo Fisher Scientific) with 405/450 nm ex/em maxima. The following antibodies were used (1:200 dilution): CD4 T cells: anti-CD3 (clone 145-2C11), anti-CD4 (GK1.5), anti-CD25 (PC61), anti-CD44 (IM7), anti-CD62L (Mel-14), anti-CD69 (H1.2F3); CD8 T cells: anti-CD3, anti-CD8 (53-6.7), anti-CD25, anti-CD44, anti-CD62L, anti-CD69; B cells: anti-CD19 (1D3), anti-CD80 (16-10A1), anti-CD86 (GL1).
The following macrophages were used: anti-CD11b (M1/70), anti-CD80, anti-CD86; DCs: anti-CD11b, anti-CD11c (HL3), anti-CD80, anti-CD86, anti-class II MHC (M5/114.15.2).
Uptake studies in human immune cells
CD4 or CD8 T cells isolated from peripheral blood mononuclear cells (PBMC; Lifeline Cell Technology) with RosetteSep Human CD4+ T-cell enrichment cocktail (StemCell Technologies) or EasySep human CD8 Positive Selection Kit II (StemCell Technologies) and cultured in R10 medium. Half of the cells were stimulated in plates coated with CD3/28 antibody beads (Miltenyi Biotec) at a 1:1 bead:cell ratio. CD8 culture medium was supplemented with 20 U/mL IL2. [3H]F-AraG uptake was measured after 48 hours of stimulation.
Th2 and memory CD4 T cells were generated according to the published method (19). CD4 T cells were isolated from whole blood (Zen-bio) that was separated using Ficoll-Paque Plus (GE Healthcare). Cells were further isolated from PBMCs with the RosetteSep Human CD4+ T-cell Enrichment Cocktail (StemCell Technologies). Half of the cells were stimulated with anti CD3/28 beads (1:1 bead:cell) for 48 hours. Th2 cells were phenotyped by flow cytometry (CD4+CCR4+CXCR3−).
To generate human memory CD8 T cells, CD3+ cells were isolated from whole blood with the Pan T-Cell Isolation Kit II (StemCell Technologies). Antibodies to CD3, CD4, CD8, CD45RO, and CCR7 were used to label target cells. CD8 (CD3+CD8+) and naïve CD4 (CD3+CD4+CD45RO−CCR7+) T cells were isolated by flow sorting. Sorted CD8 cells were expanded by coculture with CD4 T cells according to the published protocol (20). Half was stimulated with anti-CD3/28 beads (1:1 bead:cell) for 48 hours and uptake of [3H]F-AraG determined.
Regulatory T cells (Treg) were expanded in vitro from blood precursor CD4+/25hi/127lo natural Treg (nTreg) following the published protocol (21). Half of the expanded cells were stimulated with anti-CD3/28 beads (1:1 bead:cell) for 48 hours and used for [3H]F-AraG uptake studies.
CD8+ cells with an exhausted phenotype were generated and cultured according to the published protocol (22). Cells were stimulated with beads coated with anti CD2/3/28 (“costimulated”) or CD3/28 (“exhausted”) antibodies in the presence of 10 U/mL IL2 for 48 hours and used for a [3H]F-AraG uptake assay.
Monocytes from PBMCs were isolated with CD14 microbeads (Miltenyi Biotec). Depending on their microenvironment, macrophages can differentiate into either a proinflammatory phenotype (M1) or an alternatively activated anti-inflammatory (M2) phenotype. We generated M1 macrophages by culturing monocytes in 10 ng/mL GM-CSF. The cells were activated with 1 μg/mL LPS. The alternatively activated subtype (M2 macrophages) was generated by culturing isolated monocytes in 100 ng/mL macrophage CSF (M-CSF; BioLegend) and then activated with LPS for 48 hours to determine uptake of [3H]F-AraG. Activation was monitored by FACS analysis of surface activation markers and additionally, supernatants were analyzed for analysis of cytokine production. ELISAs of culture supernatants were performed to detect IL6, TNFα, macrophage inflammatory protein-1a (MIP-1a), and IL12 expression with kits from R&D Systems.
DCs were generated by incubating monocytes with 150 ng GM-CSF and human 50 ng/mL IL4 (Thermo Fisher Scientific). DCs were either activated with 1 μg/mL LPS for 48 hours or left unstimulated.
B cells were isolated from PBMCs with the human B-Cell Isolation Kit (StemCell Technologies) and either stimulated with or without 5 nmol/L phorbol 12-myristate-13-acetate (PMA; Sigma) for 48 hours.
Neutrophils and eosinophils were isolated from human whole blood (AllCells). Human Neutrophil and Eosinophil Enrichment Kit (StemCell Technologies) was used to purify the cells. Neutrophils were either treated with 1 μmol/L N-formyl- l-methionyl-l-leucyl-phenylalanine (fMLP; Sigma) or left untreated for 48 hours. Eosinophils were either incubated with or without 10 nmol/L C5a (PeproTech) for 48 hours.
Uptake assays were conducted in the same manner as in murine cells. FACS was performed with similar antibodies as described in uptake studies in murine immune cells, but against human proteins. Additional antibodies used for human immune cells: neutrophils: anti-CD16; eosinophils: anti-CD193, anti-CD44.
3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT; Sigma) cell survival assays were performed by first stimulating PBMCs with biotinylated anti-CD2, -3, and -28 antibody coated beads (Miltenyi Biotec) at a 1:2 bead to cell ratio for 48 hours. Cells were then incubated with either 158 μCi [18F]F-AraG, 27 ng/mL cold F-AraG, 10 U IL2, or R10 medium. The cells were then pulsed with MTT (Thermo Fisher Scientific) for 3 hours and lysates read on the SpectraMax Plate Reader (Molecular Devices).
PET imaging of rhabdomyosarcoma model
The animals were cared for and used at the University of California, San Francisco (UCSF) facilities that are accredited by the American Association for Accreditation of Laboratory Animal Care (AAALAC). All animal studies were performed in accordance with UCSF Institutional Animal Care and Use Committee approved. Five- to 6-week-old Balb/c mice (n = 3; Jackson Laboratories) were injected intramuscularly with 30 μL of previously titered murine sarcoma virus-moloney leukemia virus (MSV-MuLV; virus stocks courtesy of Dr. Purnima Dubey, Wake Forest University, Winston-Salem, NC). After tumors became measurable (5 mm2; 5–7 days), overnight fasted, anesthetized mice were injected intravenously with [18F]FDG or [18F]F-AraG (100–150 μCi in 100–200 μL saline). FDG and F-AraG radiotracers were injected via tail vein 1 day apart to allow adequate tracer decay ([18F]t1/2 = 110 minutes) before the next imaging session. [18F]F-AraG –injected mice were recovered from anesthesia for 1 hour to permit tracer distribution. Static scans (10 minutes PET acquisition followed by CT scan for anatomic reference) were performed once a week for 3 weeks 1 hour after tail vein injection of the tracers using an Inveon microPET/CT scanner (Siemens Healthcare). Region-of-interest (ROI) analysis of the PET/CT data was performed using VivoQuant 3.5 software (Invicro). Partial volume correction was not performed. The percentage of injected dose per gram was calculated for each ROI.
Isolation and analysis of rhabdomyosarcoma-infiltrating lymphocytes
To determine which tumor-infiltrating lymphocytes take up the tracer in vivo, day 14 MSV-MuLV–infected tumor mice (n = 3) were injected with 1 mCi [18F]F-AraG. One mCi dose was used to ensure adequate label could be detected after cell isolation. After 1-hour uptake, period mice were euthanized, tumors excised, and single-cell suspensions were prepared as described above. CD4-, CD8-, and CD19-expressing cells were isolated from the cell suspension using positive selection antibody columns (Miltenyi Biotec). Total tumor label uptake was determined with an unfractionated cell sample. Radioactivity was counted using a gamma counter (Beckman). Percent radioactivity was calculated by comparison with the unfractionated cell sample (=100%).
PET imaging of MC38 tumor-bearing mice receiving checkpoint inhibitor therapy
MC38 were obtained from Celgene. Cells were cultured in DMEM cells with 10% FCS and antibiotics and split twice weekly. The uptake of [3H]F-AraG in MC38 cells was determined as described above (Supplementary Fig. S2). Five- to 6-week-old female B6 mice (Jackson Laboratory) were subcutaneously injected in the left shoulder with 0.5 million MC38 cells (Celgene). A week later when the tumors became palpable, mice (n = 16) were injected with [18F]F-AraG (100–150 μCi/100 μL PBS) and after 1 hour imaged using the microPET/CT scanner as described above. The mice were split into two groups (n = 8 for each group) and treated three times a week via interperitoneal injection of either 5 mg/kg of anti-PD-1 antibody (RMP1-14; Bio-X-Cell) or 5 mg/kg of Isotype control (2A3; Bio-X-Cell). Mice were periodically imaged during treatment: in week one 24 and 48 hours after the first antibody injection; in week two 24 hours after four antibody injections, and in week three 24 hours after seven antibody injections. ROI analysis of the PET/CT data was performed using VivoQuant 3.5 software. Partial volume correction was not performed. The percentage of injected dose per gram was calculated for each ROI.
Antibodies to murine PD-1 (RMP1-14), CD45,2 (30-F11), CD3 (17A2), CD4 (GK1.5), CD8 (H57-597), PD-1 (29F.1A12), TIM-3 (RMT3-23), and isotype controls (clone 2A3, rat IgG2, and hamster IgG1) were purchased from Bio-X-Cell.
For FACS analysis of tumor lymphocytes, single-cell suspensions of tumors excised at the end of the study (24 hours after the last imaging time point) were prepared by mechanical mincing, enzymatic digestion (200 U/mL collagenase type IV and 2 U/mL DNase I; Worthington Biochemical), and filtration of tumors through a 40-μm mesh. Dead cells were discriminated by staining with an amine reactive dye (Violet Live/Dead amine reactive dye; Thermo Fisher Scientific), and cell surface markers labeled with fluorophore-conjugated anti-mouse antibodies to CD45.2 (30-F11), TCRβ (H57-597), CD4 (GK1.5), CD8 (53-6.7), PD-1 (29F.1A12), TIM-3 (RMT3-23), or isotype controls. Samples were run on a FACS Aria3 flow cytometer and data analyzed with FlowJo software v10. A total of 3,000 to 30,000 events were collected per sample.
[18F]F-AraG preferentially accumulates in human-activated CD8+ T cells without affecting their viability
To demonstrate the specificity of F-AraG for activated T cells, we evaluated the uptake and accumulation of [3H]F-AraG in various resting and activated human immune cells: T cells (CD4+ and CD8+ effector and memory, regulatory, exhausted and Th2), macrophages, DCs, B cells, eosinophils, and neutrophils. The uptake of [3H]F-AraG was observed in T cells, macrophages, and DCs, whereas B cells, neutrophils, and eosinophils showed negligible accumulation (Fig. 2A). Unlike T cells where the activation led to a significantly higher uptake of [3H]F-AraG, uptake in macrophages and DCs was not dependent on their activation status (Fig. 2A). To further delineate differences in the [3H]F-AraG accumulation between T cells and macrophages and DCs, we performed efflux studies. Although human T cells retained over 80% of [3H]F-AraG, DCs accumulated less than 50% and macrophages retained slightly more than 30% (Fig. 2B). The observed behavior was consistent with the levels of nucleoside transporters and dGK in immune cells, proteins involved in the accumulation mechanism of guanine (Supplementary Fig. S3). The low but nonetheless appreciable retention of the tracer in macrophages and DCs is not expected to interfere with [18F]F-AraG's ability to evaluate T-cell activation during immunotherapies because the stimulation of macrophages and DCs did not affect the tracer uptake. Activation of all T-cell subsets led to a significantly higher [3H]F-AraG uptake compared with their naïve counterparts (Fig. 2C). However, the uptake in activated CD8 cells was significantly higher than in any other T-cell subset tested.
Because [18F]F-AraG is an analogue of nelarabine, an agent used in the treatment of T-cell malignancies, we wanted to confirm that when given in tracer amounts as an imaging agent, [18F]F-AraG will not affect T-cell viability. PBMCs exposed to [18F]F-AraG for 96 hours showed no decrease in proliferation compared with medium or IL2 control cells (Fig. 2D), indicating that [18F]F-AraG is not detrimental to T cells.
The uptake profile of [18F]F-AraG in murine immune cells is different than in human immune cells
Similar to human T cells, the activation of murine T cells led to significantly higher uptake of [3H]F-AraG (Fig. 3A). Murine DCs and BMMs did not accumulate appreciable amounts of [3H]F-AraG, whereas B cells took up the tracer regardless of their activation status. Mouse B lymphocytes lost >90% of the [3H]F-AraG following a 1-hour efflux, suggesting only a transient uptake (Fig. 3B). High expression of nucleoside transporters in combination with low levels of the trapping enzymes, dCK and dGK found in murine B cells (Supplementary Fig. S4) agree well with the observed behavior of [3H]F-AraG in murine B cells. Unlike in human T-cell subsets, the level of [3H]F-AraG uptake in murine activated CD8+ cells was similar to the uptake in activated CD4 cells (Fig. 3C). Comparison of the uptake observed in murine and human cells revealed a more favorable uptake profile in human T cells (Fig. 3D). Activated human CD8+ cells showed more than 7-fold higher [3H]F-AraG uptake compared with activated murine CD8+ cells. Compared with murine-activated CD4+ cells, human-activated CD4+ cells accumulated two times as much [3H]F-AraG.
[18F]F-AraG provides information that is distinct from FDG
Many of the small molecule PET-imaging agents that are being investigated for applications in immuno-oncology suffer from relative nonselectivity, accumulating both in immune and tumor cells. To evaluate the selectivity of [18F]F-AraG for immune cells, we compared it to 18F-fluorodeoxyglucose (FDG) in a virally induced rhabdomyosarcoma model. FDG, a widely used tracer in oncology, cannot differentiate between tumor and immune cells. Rhabdomyosarcoma, generated by the intramuscular injection of MSV-MuLV is a self-limiting tumor whose rejection, complete by 4 weeks, includes infiltration, activation, and proliferation of T cells (23). We have imaged the appearance, growth, and regression of rhabdomyosarcoma tumors using both FDG and [18F]F-AraG (Fig. 4). [18F]FDG PET that informs of glucose consumption in tumors showed a low-intensity signal in week 1 when the tumors became palpable, a high-intensity signal in week 2 when the tumors reached the peak of growth, and a decreasing signal as the tumor regressed in week 3. On the other hand, [18F]F-AraG images showed steadily increasing signal intensities starting from week 1 until reaching the highest signal intensity at the tumor rejection point, at week 3 (Fig. 4A and B). These results suggest that although FDG signal in tumors paralleled the time course of tumor growth and eventual tumor destruction by the host immune system, [18F]F-AraG signal paralleled the course of an adaptive antitumor immune response. The isolation of immune cells from the tumors confirmed abundant perfusion of rhabdomyosarcoma with T-lymphocytes bearing an activated phenotype (Supplementary Fig. S5).
Tumor-draining lymph nodes were clearly visible by both FDG and [18F]F-AraG (Fig. 4C). The signal in the lymph nodes showed similar trend to intratumoral signal but the signal intensities were not significantly different at the peak of tumor growth in week 2 (Fig. 4D). As draining lymph nodes are the primary sites of antigen-activated T cells, these results further support the selectivity of [18F]F-AraG for immune cells.
Activated CD8+ and CD4+ cells are the primary sites of [18F]F-AraG accumulation in rhabdomyosarcoma
To further demonstrate that in vivo [18F]F-AraG preferentially accumulates in T cells and not tumor cells, we have assayed the radioactivity associated with lymphocytes isolated from the tumors of mice intravenously injected with the tracer. The highest radioactivity was detected in CD8+ cells, followed closely by CD4+. Murine B cells (CD19+) took up significantly lower levels of the tracer. Overall, the combined radioactivity in tumor-infiltrating lymphocytes accounted for over 80% of the radioactivity measured in the tumors (Fig. 5A). Analysis of the isolated CD8+ and CD4+ cells showed that the overwhelming majority of the lymphocytes were activated (Fig. 5B).
[18F]F-AraG can inform on the treatment response early into the anti-PD-1 therapy
To evaluate the potential of [18F]F-AraG as an imaging biomarker predictive of response to checkpoint inhibitor therapy, we longitudinally imaged MC38-bearing mice undergoing anti-PD-1 treatment (Fig. 6A). Early in the treatment, 48 hours after a single anti-PD-1 injection, the intratumoral signal in the anti-PD-1–treated mice increased significantly over the baseline, pretreatment, scan. At the same time point, the signal in the isotype-treated mice largely remained constant (Fig. 6B and C). The difference between the 2 groups of mice continued in week 2 but was not observed at the end of study, most likely because of the large differences in tumor sizes and a mixed response to anti-PD-1 treatment (Supplementary Figs. S6 and S7). FACS analyses of tumors excised one day after the final imaging time point, correlated well with our imaging findings, revealing no differences in number of intratumoral CD8+ and CD4+ cells between treated and untreated mice (Supplementary Fig. S8).
Responders and nonresponders to anti-PD-1 therapy show differences in [18F]F-AraG signal
MC38 tumors are known to have immunosuppressive tumor microenvironment that makes the treatment less efficient and results in varying therapy responses (24). Because mice in the anti-PD-1–treated group showed variable responses to therapy, we wanted to investigate whether there were differences in [18F]F-AraG signal between responders and nonresponders within the anti-PD-1–treated group of mice (Fig. 7). Mice treated with anti-PD-1 antibody were divided into responders (n = 4) and nonresponders (n = 4) based on the size of the tumors at the end of the study. We considered responders to be mice whose tumors were well controlled or absent at the end of the study. Mice with poorly controlled or large tumors were assigned into the nonresponder group. [18F]F-AraG signal was observed in both tumor and tumor-draining lymph nodes (Fig. 7A). Only 48 hours after single PD-1 antibody injection, the intratumoral [18F]F-AraG signal in responders was significantly higher than the signal prior to starting the therapy (Fig. 7B). Although the [18F]F-AraG signal at the same time point increased in nonresponders as well, the increase was not statistically significant. Remarkably, combining the intratumoral signal with the signal observed in the draining lymph nodes resulted in an even more striking separation between responders and nonresponders (Fig. 7C).
The impressive objective and durable responses to immunotherapy have to date been achieved only in a relatively small fraction of patients with advanced solid tumors (3). The reasons for variable therapy responses are not fully understood, but compelling evidence points to robust intratumoral immune cell infiltration as a critical component of a successful immunotherapy (5, 7, 25, 26). Patients with inflamed tumors, characterized by high numbers of CD4 and CD8 cells, show the best responses to checkpoint inhibitor therapy. Variable responses even in patients with highly infiltrated tumors indicate that the immune cell infiltration at the time of diagnosis is a necessary, but not a sufficient component of a successful immunotherapy (8). A variation in therapy response in inflamed tumors suggests the existence of other factors, such as the presence of immune-suppressive cells, or impaired immune cell survival and proliferation, which can affect immunotherapy outcome (8, 9). Here, we evaluate the utility of [18F]F-AraG as a T-cell–specific agent that can address critical questions pertaining to T-cell function during immunotherapy.
A vast majority of PET-imaging agents with potential to be clinically relevant in immunomonitoring are relatively large, antibody-based molecules (27–29). However, slow clearance from circulation, long retention times in target and nontarget tissues, as well as delivery barriers in solid tumors limit their clinical utility (30). Peptide-based agents show utility in mouse models and are being developed for use in humans (31). Small molecules, such as [18F]F-AraG, which are suitable for immunomonitoring are relatively scarce (32–35). FDG has been used in a number of clinical trials that investigate various immunotherapy regimens (36). As an agent that tracks utilization of glucose in cells, a process not specific to immune cells, FDG uptake cannot accurately describe the complex nature of the immune contexture within the tumor microenvironment.
[18F]F-AraG was found to preferentially accumulate in activated CD8+ cells (Fig. 2). Macrophages and DCs also showed uptake of the tracer, but unlike in T cells, the uptake did not depend on the activation status of these cells. In addition, the tracer was not well retained in macrophages and DCs, most likely because of the low levels of phosphorylating enzymes, dCK and dGK. Given [18F]F-AraG's strong preference for activated CD8+ cells and the lack of increase in tracer uptake in stimulated macrophages and DCs, the differences in [18F]F-AraG uptake before and after immunotherapy are expected to primarily assess the activating effect that the therapy has on T-cell effector cells.
Because [18F]F-AraG is intended as both a preclinical and a clinical tool, we investigated the uptake in murine immune cells and compared it with the accumulation of the tracer in human immune cells (Fig. 3). Activated murine CD8+ and CD4+ cells showed significantly lower [3H]F-AraG uptake than their human counterparts. In addition, although human-activated CD8+ cells accumulated significantly higher levels of the tracer than CD4 cells, murine CD8 and CD4 cells showed similar levels of uptake. As human immune cells demonstrated a more favorable uptake profile than murine immune cells, we expect [18F]F-AraG to show better characteristics when used in humans than in preclinical mouse models.
The differences between [18F]F-AraG and immune cell–nonspecific FDG were demonstrated in rhabdomyosarcoma model (Fig. 4). This tumor model is frequently used to study the role of T-cell–mediated immune response in the eradication of advanced disseminated malignancies because it involves the immune response that includes infiltration, activation, and proliferation of T cells (23, 37). Although FDG paralleled tumor growth, showing the highest intensity signal when the tumors reached the peak of growth, [18F]F-AraG paralleled the course of antitumor immune response, demonstrating steadily increasing signal that reached the highest intensity at the tumor rejection point. The differences between the two tracers were detected not only in tumors but in tumor-draining lymph nodes as well. Although the intranodal signal changed in a similar manner as intratumoral signal most likely due to metastatic involvement of the draining lymph nodes (38), the differences in signal between FDG and [18F]F-AraG were not statistically significant. Considering different immune contextures in tumors and lymph nodes, this finding further supports [18F]F-AraG's specificity for immune cells. Isolation of lymphocytes from the tumors of mice injected with the tracer confirmed that more than 80% of the intratumoral signal came from accumulation in immune cells, primarily CD8+ and CD4+ (more than 72%; Fig. 5).
To serve as an immunomonitoring agent, [18F]F-AraG must be able to inform on the response to therapy in a timely manner. Longitudinal monitoring of MC38-bearing mice undergoing anti-PD-1 treatment revealed differences in [18F]F-AraG signal between PD-1 and isotype antibody–treated mice early into treatment, 48 hours after single PD-1 antibody injection (Fig. 6). Although the signal for anti-PD-1–treated mice sharply increased compared with the baseline (pretreatment) scan, the signal in isotype-treated control mice largely remained constant. However, the differences in the intensity of the signal between the treated and untreated groups of mice were not statistically significant at any imaging time point. FACS analysis of the tumors confirmed no significant differences in number of intratumoral CD4+ and CD8+ cells in these two groups of mice (Supplementary Fig. S8). The lack of statistically significant differences in the number of CD4+ and CD8+, and thus intensity of the [18F]F-AraG signal, is not surprising as anti-PD-1 treatment led to a mixed response in MC38-bearing mice. Although administration of PD-1 antibody effectively controlled some tumors, growth of other tumors was not retarded (Supplementary Fig. S6). Remarkably, further analysis of the anti-PD-1–treated group uncovered important differences in the [18F]F-AraG signal between responders and nonresponders (Fig. 7).
First, the pretreatment intratumoral signal in the responders was higher than the baseline signal in the nonresponders, agreeing with the notion that the robust intratumoral immune infiltration leads to better therapy responses. These results suggest the ability of [18F]F-AraG to assess the inherent immunologic status of a patient, recently defined as cancer immune set point (8). To improve immunotherapy response rates, many novel therapeutic approaches aim to prime the tumors for immunotherapy by improving the immune milieu within tumor microenvironment (39). There has also been great interest in using standard local and systemic treatments, such as radiation and chemotherapy, to induce the activation of the immune system and in such a way prime the tumors for immunotherapy (40–42). A T-cell–specific agent that can inform on lesion's immune milieu could assess the most opportune treatment window after the priming therapy and allow selection of patients most likely to respond to immunotherapy.
Second, the signal in the responders detected 48 hours after only one injection of the PD-1 antibody was significantly higher than the baseline signal. Although at the same time point, the signal in the nonresponders also increased, the increase was not statistically significant. This indicates the importance of a pretreatment scan, as it allows for a subject to serve as its own control. Future studies will be crucial to determining whether the pretreatment scan is necessary for the accurate assessment of therapy response.
Third, the change in signal in the tumor draining lymph nodes holds key information about response to anti-PD-1 therapy. Combining the intratumoral [18F]F-AraG signal with the signal observed in the draining lymph nodes resulted in an even more striking separation between responders and nonresponders. Our findings are in strong agreement with a recent study that determined a critical role of tumor-draining lymph nodes in PD-1/PD-L1 therapeutic efficacy (43). Frasen and colleagues showed abolition of anti-PD-1/PD-L1 therapy response in the absence of immune activation in the tumor-draining lymph nodes and subsequent trafficking of CD8+ cells to the tumor. These results highlight the importance and suggest a great value in evaluating immune response not only in tumors but also in the primary sites of antigen presentation and tumor metastasis—tumor-draining lymph nodes. Tumors and lymph nodes represent two distinct environments, both in terms of nutrient and oxygen availability—factors that can greatly impact T-cell survival, expansion, and effector function (44). A closer look into differences between tumor and lymph nodes using [18F]F-AraG has a great potential to reveal metabolic constraints facing T cells in a hostile tumor environment. By addressing questions regarding interplay between immune activity and metabolism, [18F]F-AraG could also play an important role in the development of new combinatorial therapeutic strategies that offer metabolic support to traditional immunotherapeutics (45, 46). One can envision that a technique capable of capturing complex immunologic status and responses not only within tumors but in the whole body could provide an “in vivo immunoscore,” a measure of patient's immune status predictive of clinical outcome.
Comprehensive evaluation in human and murine immune cells as well as in preclinical models suggest the great utility of [18F]F-AraG to serve as a much needed immunomonitoring clinical tool for timely evaluation of immunotherapy. Potential limitations of applying this imaging approach in the clinic remain to be addressed in future studies. The largely unknown heterogeneity of immune response kinetics in patients with cancer and in different lesions within the same patient with cancer may preclude our ability to choose a general optimal time to image treatment efficacy. In addition, the adequacy of [18F]F-AraG sensitivity in relation to the clinically significant changes in immune contexture is yet to be determined. Besides investigating these issues in clinical trials, future studies will also focus on combinatorial immunotherapies and other preclinical models.
Disclosure of Potential Conflicts of Interest
J. Levi is a R&D Director at and has ownership interest in CellSight Technologies. T. Lam is a lab manager at CellSight Technologies and also has ownership interest (including stock, patents, etc.) in CellSight Technologies. S.R. Goth has ownership interest (including stock, patents, etc.) in CellSight Technologies. S. Yaghoubi has ownership interest (including stock, patents, etc.) in CellSight Technologies. K.F. Schmidt has ownership interest (including stock, patents, etc.) in Celgene Corporation. D. Jennings is a Senior Scientist at Celgene Corporation and has ownership interest (including stock, patents, etc.) in Celgene Corporation. No potential conflicts of interest were disclosed by the other authors.
Conception and design: J. Levi, S.R. Goth, S. Yaghoubi, G. Ren, R. Khattri, K.F. Schmidt, D. Jennings, H. VanBrocklin
Development of methodology: T. Lam, S.R. Goth, S. Yaghoubi, G. Ren, J.E. Blecha
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T. Lam, S.R. Goth, S. Yaghoubi, J. Bates, G. Ren, T.L. Huynh, K.F. Schmidt
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. Levi, T. Lam, S.R. Goth, S. Yaghoubi, J. Bates, G. Ren, R. Khattri, K.F. Schmidt
Writing, review, and/or revision of the manuscript: J. Levi, T. Lam, S.R. Goth, S. Yaghoubi, J. Bates, G. Ren, K.F. Schmidt, H. VanBrocklin
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Levi, T. Lam, S. Jivan, R. Khattri, H. VanBrocklin
Study supervision: J. Levi, T. Lam, S. Yaghoubi, K.F. Schmidt, D. Jennings
This work was supported in part by the NIH grants NCI SBIR N44C031 014-51 (to S. Yaghoubi), HHSN261201100119C (to S. Yaghoubi), R44CA221624-01 (to S. Goth and J. Levi), and by Celgene.
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.