Visualization of Activated T Cells by OX40-ImmunoPET as a Strategy for Diagnosis of Acute Graft-versus-Host Disease

Allogeneic hematopoietic cell transplantation (HCT) is a well-established curative therapy for a broad range of hematologic malignancies. Unfortunately, allogeneic HCT is still associated with significant morbidity and mortality related to cancer relapse and transplant complications, namely graft-versus-host disease (GvHD). During GvHD, the interaction of donor-derived T lymphocytes with host tissues induces their activation, proliferation, and migration to target tissues, notably skin, gastrointestinal tract, and hepatobiliary system, where they mediate cytolytic attack (1). Early diagnosis of acute GvHD is essential to promptly establish appropriate treatment and prevent disease progression. At present, acute GvHD diagnosis mainly relies on clinical manifestations combined with pathologic analysis of target organs by tissue biopsies. Unfortunately, these approaches lack specificity and sensitivity and can result in iatrogenic complications. Further, these studies are only initiated after the development of overt clinical symptoms that occur later in the pathogenesis of GvHD. Diagnostic strategies allowing early and noninvasive identification of patients developing acute GvHD are therefore urgently needed for more effective intervention.

Standard radiological evaluation by ultrasound, contrast-enhanced CT or MRI is of limited utility for accurate diagnosis of acute GvHD. These anatomical imaging modalities report on gross morphologic changes that lack specificity for GvHD (2). PET imaging is a noninvasive clinical imaging modality with high sensitivity and quantitative capabilities, incentives that are ideal for the dynamic assessment of inflammatory responses. Previous reports suggested that molecular imaging by ¹⁸F-fluorodeoxyglucose ([¹⁸F]FDG) PET can visualize tissue inflammation associated with acute gastrointestinal GVHD in both murine models and in the clinic (3). Unfortunately, clinical evaluation of [¹⁸F]FDG PET for acute GvHD diagnosis in larger cohorts of patients revealed a lack of specificity and low positive predictive value (4).

The use of alternative tracers designed to detect T-cell activation more specifically has been reported. Previously we reported the ability of 2′-deoxy-2′-[18F]fluoro-9-β-D-arabino-furanosylguanine ([¹⁸F]F-AraG), a small-molecule metabolic radiotracer preferentially accumulating in activated T cells to efficiently track T-cell activation and expansion in a murine model of acute GvHD (5), an approach currently under clinical evaluation (NCT03367962). A major caveat of using small-molecule tracers is that the metabolic pathways they target may also be upregulated in other hematopoietic and nonhematopoietic tissues. ImmunoPET is a rapidly growing area of PET imaging, leveraging the high specificity and high affinity of monoclonal antibodies and antibody fragments that are radiolabeled with PET isotopes, allowing for the targeted detection of cell surface markers on specific subsets of cells (6).

We recently reported a novel immunoPET tracer developed using a murine OX40-specific monoclonal antibody (⁶⁴Cu-OX40mAb) that enables noninvasive imaging of murine OX40⁺ activated T cells (7, 8). OX40 (CD134), a member of the TNF receptor superfamily, is a T-cell costimulatory molecule whose expression is highly restricted to activated T cells (9). Previous studies have shown an increase in OX40 expression at the T-cell surface during acute GvHD in rodents (10), nonhuman primates (11), and humans (12–15). Moreover, there is evidence that OX40 plays a role in acute GvHD pathogenesis where OX40/OX40L blockade significantly attenuates disease progression in murine (16–18) and nonhuman primate (11) models of disease.

In the present work, we evaluated the utility of OX40-targeted immunoPET imaging and the sensitivity of this approach to image the kinetics and homing of activated T cells in vivo in a major MHC-mismatch murine model of acute GvHD.

BALB/cJ (H-2kd) and C57Bl/6J (H-2kb) mice were purchased from The Jackson Laboratory. Firefly luciferase (Luc)+ transgenic C57BL/6 L2G85 mice have been described previously (19) and were bred in our animal facility at Stanford University. B6.129S4-Tnfrsf4tm1Nik/J OX40 knockout mice (OX40^−/− C57BL/6) were kindly provided by Dr. Ronald Levy's laboratory at Stanford University.

All procedures performed on animals were approved by Stanford University's Institutional Animal Care and Use Committee and were in compliance with the guidelines of humane care of laboratory animals.

Donor CD4⁺ and CD8⁺ conventional T cells (Tcon) were isolated from splenocytes harvested from luc+ C57BL/6 mice and enriched with CD4 and CD8 MicroBeads (Miltenyi Biotec). Cell purity was consistently >95%. T-cell–depleted bone marrow (BM) cells were prepared by first crushing bones followed by T-cell depletion using CD4 and CD8 MicroBeads (Miltenyi Biotec). BALB/c recipient mice were treated with lethal total body irradiation (TBI) consisting of 880 cGy in 2 doses administered 4 hours apart. On the same day, 5 × 10⁶ BM cells from C57BL/6 mice and 1.0 × 10⁶ Tcon from luc+ or OX40^−/− C57BL/6 mice were injected intravenously. Mice were monitored daily, and body weight and GvHD score were assessed at day 4, day 7, and weekly thereafter as previously described (20).

Single-cell suspensions were prepared from cervical lymph nodes (CLN), mesenteric lymph nodes (MLN), and spleen. Extracellular staining was preceded by incubation with purified FC blocking reagent (Miltenyi Biotech) to reduce nonspecific staining. Cells were stained with the following antibodies (Biolegend): FITC anti-CD45.1 (clone A20); PE anti-OX40 (clone OX-86) or appropriate isotype control (clone RTK2071); APC anti-Thy1.1 (clone OX-7); APC/Fire750 anti-CD19 (clone 6D5) and anti-CD45.2 (clone 104); BV421 anti-CD4 (clone GK1.5); BV605 anti-CD3 (clone 17A2); and BV650 anti-CD8 (clone 53–6.7). Samples were acquired on a BD LSR II flow cytometer (BD Biosciences), and analysis was performed with FlowJo 10.5.0 software (Tree Star).

Intestinal tissues were harvested from transplanted mice (day 7 after transplant) and frozen in optimal cutting temperature embedding compound and stored at −80°C until further use. Tissue sections were cut longitudinally into 10 μm thick sections using a cryostat. Immunofluorescence staining was carried out using standard procedures using the following primary antibodies: anti-OX40 (clone OX86, BioXcell) and CD90.1-Biotin (Invitrogen). OX40 expression was detected using donkey anti-rat IgG-AF488 (Abcam) and Streptavidin-AF647 (Invitrogen), respectively. The sections were finally mounted with VECTASHIELD Antifade Mounting Medium with DAPI (Vector Laboratories) and imaged with Nikon A1R-Si-MP multiphoton confocal microscope using a Nikon CF160 Plan Apo Lambda 10x/0.45 numerical aperture objective lens.

Mice were injected with D-luciferin (10 mg/kg; intraperitoneally) and anesthetized with isoflurane. Imaging was conducted using an IVIS Spectrum imaging system (Perkin Elmer), and data were analyzed with Living Image Software 4.1 (Perkin Elmer).

Bioconjugation of murine-specific OX40mAb to DOTA-NHS chelate (Macrocyclics) was performed using previously optimized protocols (7) and is described in Supplementary Methods. DOTA-OX40mAb was radiolabeled with ⁶⁴CuCl₂ (University of Wisconsin, Madison, WI, USA) to produce ⁶⁴Cu-DOTA-OX40mAb (heron referred to as ⁶⁴Cu-OX40mAb) with final specific activity of 10 to 15 μCi/μg, radiochemical purity >99%, and labeling efficiency of 95% to 99%. Radio-ITLC and radio-HPLC using size-exclusion liquid chromatography with a Phenomenex SEC 3000 column with sodium phosphate buffer (0.1 mol/L, pH 6.8) was performed to corroborate radiochemical purity. The final formulation was prepared in PBS.

Mice were anesthetized using isoflurane delivered by 100% oxygen (2.0%–2.5% for induction and 1.5%–2.5% for maintenance). ⁶⁴Cu-OX40mAb (105 ± 9 μCi, 8 ± 3 μg) was administered i.v. via the tail vein. Transplanted mice and TBI controls received tracer at day 4 or day 7 after transplant. Wild-type and OX40^−/− mice were also imaged for control studies. Static (10 minutes) PET scans were acquired 24 hours after ⁶⁴Cu-OX40mAb administration using a small animal PET/CT hybrid scanner (Inveon, Siemens). CT images were acquired prior to each PET scan and within the same acquisition workflow to provide an anatomic reference for PET data and to allow for attenuation correction. PET image reconstruction and image analysis were conducted as previously described and are outlined in further details in Supplementary Methods (7).

Following the completion of the scan 24 hours after injection of tracer, ex vivo biodistribution studies were performed to measure blood- and tissue-associated radioactivity. Briefly, blood (∼100 μL) was collected via cardiac puncture, and the following tissues were collected: CLN and MLN, spleen, skin, lower part of the small intestine (above the ileocecal valve), large intestine, kidney, liver, muscle, and femur. Tissues were placed in a tube, weighed, and radioactivity measured using an automated gamma counter (Cobra II; Packard). Tissue-associated radioactivity was normalized to tissue weight and amount of radioactivity administered to each mouse, decay-corrected to the time of radiotracer injection. Data were expressed as percentage injected dose per gram of tissue (%ID/g) values.

Statistical analyses were performed using Prism 6 (GraphPad Software) and R version 3.5.1 with R studio version 1.1.453. Heatmaps were generated using Pheatmap version 1.0.12. Principal component analysis (PCA) was performed using the FactoMineR package version 1.41 and visualized using the factorextra package version 1.0.5. ROC curves were calculated and plotted using the plotROC version 2.2.1.

We first analyzed the dynamics of OX40 expression on the surface of CD4⁺ and CD8⁺ T cells recovered after adoptive transfer into allogeneic (BALB/c) or syngeneic (C57BL/6) recipients at the time of HCT. Donor-derived T cells were identified using the CD45.1⁺ and CD90.1⁺ congenic markers, and OX40 expression was assessed by flow cytometry (using a gating strategy and an isotype control outlined in Supplementary Fig. S1A and S1B, respectively). We detected a significant upregulation of OX40 on CD4 (Fig. 1A, top plots) and to a lesser extent on CD8 T cells (Fig. 1A, bottom plots) recovered from the CLN and MLN, both at day 4 and day 7 after transplantation. CD4 T cells isolated from the spleen significantly upregulated OX40 at day 4 after transplantation, whereas levels of expression at day 7 returned to baseline (Fig. 1A, top right plot). Conversely, CD8 T cells recovered from the spleen expressed higher levels of OX40 both at day 4 and day 7 compared with baseline (Fig. 1A, bottom right plot). Importantly, OX40 upregulation was uniquely observed after adoptive transfer into allogeneic recipients, whereas cells recovered upon transfer into syngeneic mice did not show any significant increase in OX40 expression (Fig. 1A, green symbols and lines). We next investigated the specificity of OX40 staining to identify activated donor-derived T cells after unsupervised clustering of splenocytes recovered at day 7 after transplantation from allogeneic recipients receiving BM alone (control group) or BM in combination with T cells (GvHD group). OX40 was selectively expressed on donor-derived activated T cells, whereas only minimal expression was detected within host-derived or BM-derived donor cells (Fig. 1B). We finally assessed donor-derived T-cell infiltration identified by CD90.1 staining (Fig. 1C, red channel) and OX40 expression (green channel) in intestinal tissue, a prime site for GvHD pathology. Immunofluorescence staining performed on small intestine samples harvested at day 7 after transplantation revealed higher numbers of CD90.1⁺ and OX40⁺ cells infiltrating the small intestine in GvHD mice in comparison with control mice (Fig. 1C). Collectively, these results indicate that OX40 is significantly upregulated at the T-cell surface during alloreactive responses in GvHD, allowing for the selective identification of activated donor-derived T cells in both lymphoid organs and the gastrointestinal tract.

We next tested the ability of OX40-targeted immunoPET to capture T-cell expansion and activation during acute GvHD using ⁶⁴Cu-OX40mAb. Acute GvHD was induced using luc+ donor T cells enabling in vivo monitoring of donor T-cell expansion and accumulation using bioluminescence imaging (BLI) as reference. In vivo BLI and ⁶⁴Cu-OX40mAb injections were performed on day 4 or day 7 after transplant. PET/CT images were acquired 24 hours after tracer injection (Fig. 2, Supplementary Fig. S2). Figure 2A shows a reference atlas of a representative volume-rendered technique (VRT) PET/CT image and axial PET/CT views with the location of key clearance, lymphoid. and GvHD target tissues annotated. In agreement with previous reports (5, 19), in vivo BLI at day 4, a time point at which we could not detect any signs of clinical disease according to our scoring system, revealed donor T-cell expansion and accumulation in the spleen and in MLN of GvHD mice (Fig. 2B). PET/CT images of control mice acquired at 24 hours after tracer injection showed that ⁶⁴Cu-OX40mAb primarily accumulated in the heart and the liver, characteristic of whole antibody biodistribution, with minimal signal detected in the spleen and in the abdominal region (Fig. 2C, left plot). GvHD mice injected with tracer on day 4 exhibited a pronounced ⁶⁴Cu-OX40mAb-PET signal in the spleen, MLN, and lower abdomen (Fig. 2C) also seen in three-dimensional (3D) rotational images (Supplemental Video S1). We next performed the same analysis at day 7, a time point at which animals displayed overt signs of disease. In vivo BLI demonstrated an expansion of donor-derived T cells with highest levels of signal originating from the abdominal region (Fig. 2D), indicating a dramatic intestinal infiltration at this time point in GvHD mice. PET imaging at day 7 (Fig. 2E, Supplementary Video S2) corroborated BLI and showed high accumulation of ⁶⁴Cu-OX40mAb in the abdominal region and, to a lesser extent, in the splenic region, a finding compatible with the further migration of T cells to the intestinal tract at this later time point (19). Notably, compared with control mice, GvHD mice exhibited lower PET signals in the heart (Fig. 2C and E; Supplementary Videos S3 and S4).

To quantify the trends observed in the PET images, we performed region of interest (ROI) analysis on multiple tissues and regions (Fig. 3A; Supplementary Methods). ROI analysis of CLN was avoided due to contribution of high signal from surrounding blood vessels in the cervical region at the relevant time points, instead we focused on lymphoid compartments in the abdomen. For day 4 tracer injections, ROI quantification of PET/CT images confirmed markedly increased radiotracer uptake in GvHD versus control mice in lymphoid tissues and GvHD sites (spleen: 15.42 ± 2.29 vs. 8.04 ± 1.10%ID/g; MLN: 18.80 ± 3.41 vs. 10.82 ± 2.06%ID/g; abdominal region: 8.03 ± 2.52 vs. 3.56 ± 0.71%ID/g; P < 0.001; n = 9–10 per group, Fig. 3B). Comparison of control mice with mice exposed to TBI without receiving BM cells revealed a small but statistically significant contribution of BM-derived cells in the background signals detected in lymphoid organs, but not in the abdominal region (Supplementary Fig. S3A and S3B), further supporting the specificity of the detected ⁶⁴Cu-OX40mAb-PET signal for GvHD-mediating T cells. ROI quantification of PET/CT images obtained 24 hours after tracer injection (injection day 7, imaging day 8) revealed similar trends, with significantly higher radiotracer uptake in target tissues in GvHD mice compared with controls (spleen:12.88 ± 1.16 vs. 9.87 ± 1.26, P < 0.001; MLN: 14.62 ± 3.54 vs. 10.45 ± 1.55, P < 0.01 and abdominal region: 9.36 ± 2.85 vs. 5.33 ± 0.78, P < 0.001; n = 12 per group, Fig. 3C). Importantly, the background ⁶⁴Cu-OX40mAb-PET signals in muscle showed no significant difference between the two groups. Comparison of control mice receiving BM with TBI mice injected at day 7 confirmed the results obtained at day 4, with control mice specifically displaying higher ⁶⁴Cu-OX40mAb-PET signal than TBI mice in the secondary lymphoid organs (Supplementary Fig. S3C and S3D).

Biodistribution analysis using ex vivo gamma counting of tissues confirmed PET results with significantly increased tracer uptake measured in lymphoid tissues of GvHD versus control mice: spleen (days 4 and 7; P < 0.01), MLN (P < 0.01), CLN (day 4; P < 0.05, day 7; P < 0.01; Fig. 4A). Increased tracer uptake was also confirmed in GvHD target organs; small intestine (days 4 and 7; P < 0.05), colon (day 7; P < 0.05), and skin (day 7; P < 0.01; day 4; Fig. 4A, day 7; Fig. 4B). Mice with GvHD displayed significantly reduced signals from the heart as shown by ROI quantification (Fig. 3) and biodistribution analysis of heart and blood (Fig. 4) compared with control mice at day 4 and at day 7. A similar trend, reaching statistical significance only in ROI quantification at day 4, was observed for the liver (Fig. 3B). Such background signals from heart, blood, and liver are likely nonspecific, due to the presence of antibody circulating in the blood and were independent of any OX40-specific binding to circulating cells as ROI quantification and in biodistribution analysis showed similar levels of signal in WT and OX40^−/− mice (Supplementary Fig. S3E and S3F), suggesting the presence of a sink effect leading to a reduction in the concentration of circulating ⁶⁴Cu-OX40mAb tracer in GvHD mice. Correlation analysis showed a significant positive correlation between ROI quantification and biodistribution analysis, especially in spleen (R² = 0.64, P = 3.2e-08) and blood (R² = 0.61, P = 2.8e-09; Supplementary Fig. S4), further confirming the ability of OX40-immunoPET to define the localization of activated T cells in murine acute GvHD. Collectively, these results demonstrate the ability of ⁶⁴Cu-OX40mAb immunoPET to detect T-cell expansion and activation during murine acute GvHD prior to and after clinical signs of GVHD are evident.

We next analyzed the relationship between ⁶⁴Cu-OX40mAb tracer uptake and clinical signs of GvHD. As expected, ⁶⁴Cu-OX40mAb tracer uptake in spleen, MLN, and the abdomen showed only poor if any correlation with body weight and GvHD score at day 4, when mice were essentially asymptomatic (Supplementary Fig. S5), highlighting the potential for early disease detection capabilities of OX40 imaging even before overt clinical signs of GvHD appear. Conversely, ⁶⁴Cu-OX40mAb uptake in the abdominal region at day 7 negatively correlated with body weight (R² = 0.44, P = 0.0005) and positively correlated with the GvHD score (R² = 0.59, P = 0.00001; Fig. 5). Moreover, ⁶⁴Cu-OX40mAb uptake in spleen and MLN at day 7 positively correlated with the GvHD score (spleen: R² = 0.36, P = 0.0018; MLN: R² = 0.18, P = 0.042; Fig. 5). Collectively, these results show that quantification of OX40-expressing activated T cells in lymphoid organs and GvHD-target sites allows early detection of GvHD even before the appearance of clinical signs and efficiently reflects the severity of the disease.

Previous reports have demonstrated a role for OX40 expression on T cells during murine acute GvHD induction, showing that splenocytes from OX40^−/− C57BL/6 mice (H-2K^b) had reduced GvHD induction potential upon transfer in MHC-mismatched B10.BR mice (H-2K^b) compared with their wild-type counterpart (17). In agreement, we observed similar results in our mouse model as adoptive transfer of 1 × 10⁶ T cells from OX40^−/− C57BL/6 mice (H-2K^b) into lethally irradiated BALB/c mice (H-2K^d) at time of transplantation, resulting in a significantly delayed kinetic of GvHD-related death compared with the adoptive transfer of identical numbers of cells isolated from WT C57BL/6 mice (Supplementary Fig. S6). In the same report, Blazar and colleagues showed that the administration of high (200 μg) and repeated doses of the agonistic anti-OX40mAb (clone M5) significantly increased GvHD lethality in the same murine model (17). We therefore tested the safety of the agonistic anti-OX40mAb (clone OX86) we used for the imaging study in our murine model, administered at tracer doses (described in Supplementary Methods). Administration of very low doses (15 μg—the maximal upper limit of antibody dose anticipated to ever be administered during PET imaging) of cold OX40mAb at day 4 after HCT significantly accelerated GvHD lethality compared with administration of isotype control (P < 0.0001; Fig. 6A). Conversely, we did not detect any significant impact of administration of the same dose of cold OX40mAb at day 7 after HCT compared with mice that received isotype control (Fig. 6B). Collectively, these results confirm the role of OX40 in murine acute GvHD pathogenesis and reveal that even very low doses of agonistic anti-OX40 mAb might exacerbate GvHD lethality when administered at early phases of GvHD induction.

To assess the relative contribution of signals obtained from ROI of different tissues for the detection of murine acute GvHD, we first performed an unsupervised analysis of ⁶⁴Cu-OX40mAb-PET signals detected by ROI quantification across all groups and mice (Fig. 7A). Unsupervised hierarchical clustering easily separated all GvHD mice injected at day 4 after HCT together with one mouse at day 7 after HCT (Fig. 7A). Within the second subgroup, hierarchical clustering identified a second cluster containing most of GvHD mice at day 7 with the exception of only 2 mice that clustered together with control mice (Fig. 7A). To identify the relative impact of different ROI in clustering, we performed a PCA of ROI data (Fig. 7B). The scree plot shown in Fig. 7C reveals that PC1 and PC2 combined accounted for 78.9% of the variance of the data, supporting the feasibility of this approach. PC1 alone allowed the distinction of GvHD mice injected at day 4 from all other mice, whereas a combination of PC1 and PC2 allowed the separation of GvHD mice from controls independently of the time points (Fig. 7B). Mapping of the component loadings identified abdomen, spleen, and MLN ROIs as the variables most strongly contributing to the distinction between GvHD mice and control mice (Fig. 7B). We therefore assessed the diagnostic potential of abdomen, MLN, and spleen ROI data for detection of murine acute GvHD at both day 4 and day 7 using ROC curve analysis. Muscle ROI was used as a negative control. As shown in Fig. 7D, abdomen, spleen, and MLN PET ROIs had a perfect diagnostic power for murine acute GvHD detection at day 4 (abdomen: AUC = 1; MLN: AUC = 1; spleen: AUC = 1) and an excellent diagnostic power at day 7 (abdomen: AUC = 0.96; MLN: AUC = 0.88; spleen: AUC = 0.98). As expected, the muscle ROIs had no diagnostic potential for GvHD detection either at day 4 (AUC = 0.28) or day 7 (AUC = 0.43). These results indicate that ⁶⁴Cu-OX40mAb-PET signals detected in the abdomen, MLN, and spleen ROIs have excellent diagnostic potential for detection of murine acute GvHD both before and after the appearance of clinical symptoms of disease.

In the present report, we demonstrated the ability of OX40-immunoPET to efficiently visualize T-cell activation, expansion, and tissue infiltration in a murine model of acute GvHD. As our ability to successfully prevent and treat acute GvHD depends on our capacity to efficiently detect disease before clinical manifestations appear, we hypothesized that a noninvasive imaging strategy able to detect T-cell activation would be an extremely powerful approach for early diagnosis of acute GvHD. Toward this aim, the selection of the most appropriate activation marker appears to be crucial. Previous studies in animal models (10, 11) as well as in humans (12–15) have shown that expression of OX40 on T cells increases during GvHD, in particular at CD4⁺ T-cell surface (15). Importantly, OX40 outperformed other activation markers, including CD25 and CD69, for detection of alloreactive T-cell responses (14, 15, 21). After confirming OX40 upregulation at the T-cell surface during alloreactive responses in both secondary lymphoid organs and target tissues, we demonstrate that OX40-immunoPET was able to visualize T-cell activation even prior to the development of overt clinical symptoms in a murine model of GvHD. Importantly, OX40-immunoPET was able to distinguish clearly early intestinal GvHD from toxicities resulting from the conditioning regimen, which represents a major limitation to the radiologic diagnosis of GvHD (2). However, the preclinical nature of our study limits the number of confounding factors often encountered in the differential diagnosis of GvHD, notably infections. Although OX40 specificity for the T-cell compartment makes false positive originating from bacterial infections unlikely, we cannot exclude that viral infections, notably cytomegalovirus colitis, could result in positive signals in OX40-immunoPET. For interpretation of imaging results, it will therefore be critical to integrate clinical (donor/recipient serostatus) and biological (viral loads) elements for the differential diagnosis between these two entities.

The need for specific visualization of T-cell responses in vivo has motivated the development of several immunoPET agents in recent years. To date there have been a few candidates reported, targeted to lineage-defining cell surface markers expressed on T cells such as CD3 (22–25). Although these phenotypic-targeted probes can capture the dynamics of T cells, they fail to directly report on T-cell activation. Recently Pektor and colleagues reported a CD3-targeted immunoPET approach for imaging GvHD progression in a humanized mouse model (23). It is well-documented that T-cell activation mediates the downregulation of the CD3/T-cell receptor (26), a potential limitation in the context of inflammation. Another approach reported for PET imaging of GvHD progression in a humanized mouse model has been to image human class II MHC (HLA-DR) using a camelid-derived single-domain antibody (27). Although HLA-DR expression increases during T-cell activation, it is also expressed at high levels in the myeloid compartment and thus lacks specificity as an activated T-cell imaging target compared with imaging of OX40.

Our preclinical data indicate that, owing to its high sensitivity, OX40-immunoPET could detect signs of GvHD even before clinical symptoms manifest. However, given the complexity and relatively high costs of immunoPET, it is unlikely that this approach will be applied as a screening strategy for all allogeneic HCT recipients. For clinical translation of this type of imaging, it will therefore be critical to carefully establish patient selection criteria. We can imagine a scenario in which investigation by immunoPET will be triggered by clinical suspicion, for example to investigate possible gut GvHD after the appearance of skin GvHD, or after initial screening using blood biomarkers like REG3α and ST2 (28). Alternatively, immunoPET of activated T cells could also be employed as a screening strategy in patients at high risk of developing GvHD (e.g., posttransplant from MHC-mismatched and/or unrelated donors). Results of early clinical trials, like the one ongoing at our institution using the small-molecule PET tracer [¹⁸F]F-AraG to detect T-cell activation for GvHD diagnosis (NCT03367962), will lay the groundwork for defining the patient selection criteria.

A major limitation of our study comes from the exacerbation of GvHD we observed upon administration of tracer doses of agonistic anti-OX40 antibody, when given at early time points of disease (day 4). The anti-OX40 antibody significantly exacerbated GvHD similar to previous reports, although in that study repeated administration of higher doses of anti-OX40 mAbs was studied (17). These results further confirm the role of OX40 in the GvHD pathogenesis and stress the importance of the selection of appropriate clones for each application. Although agonistic anti-OX40 antibodies can be appropriate for imaging of T-cell activation in cancer immunotherapy settings (7, 8), immunoPET in immunopathologic settings targeting costimulatory molecules will probably require the use of nonagonist or even antagonist antibody clones. Although the generation of an antagonistic murine anti-OX40 clone is beyond the scope of the present report, clinical translation of OX40-immunoPET for GvHD diagnosis would probably benefit from the use of anti–human-OX40 antagonist clones such as the GBR830 clone, currently under clinical investigation for atopic dermatitis (NCT03568162; ref. 29) and recently reported to suppress xenogeneic GvHD when administered in combination with Cyclosporine A (30). Alternative approaches to minimize adverse biological effects may also include the generation of antibody fragments lacking the potency of a full-length antibody and the Fc region known to engage other immune cells, or engineered binders (31). Perturbations in the biology of immune cells by immunoPET tracers have previously been noted (32, 33) and may be further minimized by improving the specific activity of the radiolabeled probe so that significantly less mass is administered.

In summary, this study demonstrates the utility of OX40 as a sensitive imaging biomarker for the early and specific visualization of activated T cells in GvHD. Integrated with tissue biopsies and endoscopic evaluation, we anticipate that this whole-body imaging approach of T-cell activation by immunoPET could significantly improve upon current GvHD diagnosis and provide earlier diagnosis where interventions may be more effective for improved clinical care.

I.S. Alam reports grants from Ben & Catherine Ivy Foundation, grants from The Canary Foundation, grants from National Cancer Institute (R01 1 CA201719-02), and grants from Parker Institute for Cancer Immunotherapy during the conduct of the study. R. Negrin reports grants from National institutes of Health (R01 CA23158201) and grants from Parker Institute for Cancer Immunotherapy (P01 CA49605) during the conduct of the study. S.S. Gambhir reports grants from National Cancer Institute (R01 1 CA201719-02) and grants from Parker Institute for Cancer Immunotherapy (P01 CA49605) during the conduct of the study. No potential conflicts of interest were disclosed by the other authors.

I.S. Alam: Conceptualization, data curation, software, formal analysis, validation, investigation, methodology, writing-original draft, project administration, writing-review and editing. F. Simonetta: Conceptualization, data curation, software, formal analysis, funding acquisition, validation, investigation, visualization, methodology, writing-original draft, project administration, writing-review and editing. L. Scheller: Data curation, methodology, conducted experiments. A.T. Mayer: Conducted experiments. S. Murty: Conducted experiments. O. Vermesh: Conducted experiments. T.W. Nobashi: Conducted experiments. J.K. Lohmeyer: Conducted experiments. T. Hirai: Conducted experiments. J. Baker: Generated the Luc+ transgenic C57bl/6 L2g85 mouse strain. K.H. Lau: Data curation, methodology. R. Negrin: Conceptualization, resources, supervision, funding acquisition, writing-review and editing. S.S. Gambhir: Conceptualization, resources, supervision, funding acquisition, writing-review and editing.

The authors would like to acknowledge the Stanford Center for Innovation in In-Vivo Imaging (SCI³) and, in particular, Drs. Timothy Doyle and Frezghi Habte for supporting the preclinical imaging experiments. We also thank the Stanford shared FACS facility for their support. In addition, we are extremely grateful to the following: Dr. Idit Sagiv-Barfi for supporting in vivo studies, Drs. Corinne Beinat and Michelle James for their helpful advice on histology, and Dr. Martin Schneider for supporting microscopy.

This work was supported in part by funding from the Ben & Catherine Ivy Foundation (S.S. Gambhir), the Canary Foundation (S.S. Gambhir), NCI R01 1 CA201719-02 (S.S. Gambhir), R01 CA23158201 (R. Negrin), P01 CA49605, the Parker Institute for Cancer Immunotherapy (S.S. Gambhir and R. Negrin), the Geneva University Hospitals Fellowship to F. Simonetta, the Swiss Cancer League (BIL KLS 3806-02-2016 to F. Simonetta), the Fondation de Bienfaisance Valeria Rossi di Montelera (Eugenio Litta Fellowship to F. Simonetta), and the American Society for Blood and Marrow Transplantation (New Investigator Award 2018 to F. Simonetta).

We dedicate this paper to the loving memory of the late Professor Sanjiv Sam Gambhir. His extraordinary impact, vision, and humanity live on and will continue to guide us for years to come.

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.