⁸⁹Zr-labeled Bispecific T-cell Engager AMG 211 PET Shows AMG 211 Accumulation in CD3-rich Tissues and Clear, Heterogeneous Tumor Uptake

Immunotherapy with immune checkpoint inhibitors is currently used as part of many treatment regimens for a wide range of tumor types. Unfortunately, not all patients benefit from these drugs. This has stimulated the search for new drugs to induce an anticancer immune response, including bispecific antibodies (1).

One novel approach is the use of bispecific T-cell engager (BiTE) antibody constructs (a registered trade mark of Amgen Inc.). These consist of two single-chain variable fragment arms of which one is directed against an antigen target on the tumor cell membrane and the other often against cluster of differentiation 3 (CD3) on T-cells. Binding of both arms induces target cell–dependent T-cell activation and proliferation, leading to apoptosis of tumor cells (2). The anti-CD19/CD3 BiTE blinatumomab is approved for the treatment of patients with B-cell precursor acute lymphoblastic leukemia (3). Continuous intravenous administration is used because of its short serum half-life of 2 hours. This results from its small molecular size of approximately 55 kDa, which leads to renal filtration, and the lack of an Fc domain, which prevents salvation from lysosomal degradation (4, 5).

AMG 211 (also known as MEDI-565) is a carcinoembryonic antigen (CEA; CEACAM5)-directed BiTE. CEA, a glycosylated human oncofetal antigen, is abundantly expressed by a variety of tumors, especially adenocarcinomas of the gastrointestinal tract (6, 7). In vitro studies have shown that a low concentration of approximately 1 ng/mL of anti-CEA/CD3 AMG 211 is sufficient to activate patient-derived T-cells with subsequent lysis of patient-derived chemo-refractory CEA-positive colorectal tumor cells (8, 9).

A study in patients with advanced gastrointestinal adenocarcinomas with 0.75 μg to 7.5 mg/day AMG 211 administered intravenously over 3 hours on days 1 to 5 in 28-day cycles showed linear and dose-proportional pharmacokinetics, but no tumor responses (10). This might be related to intermittent administration and short exposure of the tumor to the drug, which has an elimination half-life of 2.2 to 6.5 hour. To achieve sustained target coverage, thereafter AMG 211 was administered and tested as a continuous intravenous infusion for 28 subsequent days in 6-week treatment cycles in a phase I study in patients with advanced gastrointestinal adenocarcinomas (11).

In bispecific antibodies, the potentially different binding affinity for the target of each of the arms might affect biodistribution. However, very limited information is available regarding whole-body distribution of bispecific antibodies and BiTE antibody constructs in patients (12, 13). Improved understanding of biodistribution of these bispecific antibody constructs might help to guide drug dosing schedules and inform potential target-related drug impact in vivo. PET with zirconium-89 (⁸⁹Zr)-labeled AMG 211 as a tracer has shown specific tracer uptake in human CEA-expressing tumor-bearing mice (14). Therefore, we performed a first-in-human feasibility study with the ⁸⁹Zr-labeled BiTE antibody construct AMG 211 and PET imaging to determine the biodistribution of ⁸⁹Zr-AMG 211 in healthy tissues and tumor lesions before and/or during AMG 211 treatment in the AMG 211 phase I study.

Patients with pathologically proven gastrointestinal adenocarcinomas were eligible for this imaging study (ClinicalTrials.gov identifier NCT02760199) if they were participating in the phase I study with AMG 211 (ClinicalTrials.gov identifier NCT02291614) at the University Medical Center Groningen (UMCG; Groningen, the Netherlands) or the Free University Medical Center (VUMC; Amsterdam, the Netherlands). Other eligibility criteria included age ≥18 years, written informed consent, and availability of ≥1 measurable lesion as assessed with CT per modified immune-related response criteria (irRC; ref. 15). For visceral lesions, this is defined as the two longest perpendicular diameters ≥10 × 10 mm, and for pathologic lymph nodes as the longest diameter perpendicular to the longest axis ≥15 mm.

This study was conducted in compliance with the Declaration of Helsinki, ICH harmonized Tripartite Guideline for Good Clinical Practice (ICH-GCP) and applicable national and local regulatory requirements. This study was centrally approved by the Medical Ethical Committee of the UMCG and the Central Committee on Research Involving Human Subjects, the competent authority in the Netherlands. All patients provided written informed consent.

This two-center imaging study was performed at the UMCG and the VUMC, both university medical centers in the Netherlands. In the phase I study, patients received continuous intravenous treatment with 6,400-μg/day or 12,800-μg/day AMG 211 via a central venous access port for 28 subsequent days (“treatment period”) in 42-day cycles. The imaging study was performed before AMG 211 treatment and/or immediately after the end of the second AMG 211 treatment period of 28 days (“during AMG 211 treatment”) as is illustrated in Fig. 1.

The tracer ⁸⁹Zr-AMG 211 was produced in the UMCG under good manufacturing practice conditions, as described previously (14, 16). Briefly, AMG 211, which was produced and provided by MedImmune via collaboration with Amgen, was reacted with a 4-fold molar excess of the tetrafluorphenol-N-succinyldesferal-Fe ester (N-Suc-Df; ABX) and purified by gel filtration using PD-10 columns. The conjugate N-SucDf-AMG 211 was radiolabeled with clinical grade ⁸⁹Zr-oxalate (PerkinElmer) and again purified by gel filtration. Individual fractions were pooled on the basis of the amount of radioactivity and radiochemical purity. Quality control of intermediate and final drug product consisted of determination of conjugation ratio, aggregation, radiochemical purity, and stability. Immunoreactivity tests on the extracellular domain of CEA showed that ⁸⁹Zr-AMG 211 was still capable of specific binding to its target. In addition, a binding assay on CD3⁺ T-cells was performed to confirm binding of N-SucDf-AMG 211 to CD3.

In 8 patients, ⁸⁹Zr-AMG 211 imaging was performed before AMG 211 treatment. These patients received, via a separate intravenous line, a fixed dose of 37 MBq approximately 200-μg ⁸⁹Zr-AMG 211 alone (n = 2), or in combination with 1,800-μg (n = 4) or 4,800-μg (n = 2) cold AMG 211, administered in 3 hours. This 3-hour period was based on the MTD and infusion rate as assessed in the phase I study. Cold AMG 211 was added to guarantee sufficient tracer availability and was therefore administered before ⁸⁹Zr-AMG 211 (details in Supplementary Materials and Methods: ⁸⁹Zr-AMG 211 administration). We considered the cold AMG 211 dose to be sufficient when the circulation could be adequately visualized at each PET scan time point as used in other studies with comparable design. To mitigate AMG 211-related cytokine release syndrome, 4-mg dexamethasone was administered orally 1 hour before the cold AMG 211 infusion, and at 3 hours and 6 hours thereafter. AMG 211 treatment started 7 days after tracer injection. Moreover, in 2 patients, 200-μg ⁸⁹Zr-AMG 211 was administered over 3 hours via a separate intravenous line to study biodistribution immediately after the end of the second AMG 211 treatment period. In one of these 2 patients, PET imaging was also performed before AMG 211 treatment. After tracer infusion, patients were observed in the hospital for 24 hours to detect any side effects. The NCI Common Terminology Criteria for Adverse Events (NCI CTCAE) v4.03 were used for grading of adverse events (17).

PET/CT scans were performed from the top of the skull to mid-thigh with a 40-slice or 64-slice PET/CT camera (Biograph mCT, Siemens in the UMCG and Gemini TF or Ingenuity TF, Philips in the VUMC) initially 6, 24, and 48 hours after completion of the tracer injection. This was changed into 3, 6, and 24 hours from the second patient onwards, based on a review of data from the first patient showing rapid ⁸⁹Zr-AMG 211 clearance from the circulation [blood pool standardized uptake value (SUV)_mean 0.2 at 24 hours]. For attenuation correction and anatomic reference, a low-dose CT scan was acquired immediately before the PET scan.

Diagnostic CT scans of the chest and abdomen were performed within 21 days before ⁸⁹Zr-AMG 211 injection and for response evaluation after every two AMG 211 treatment cycles.

All PET scans were reconstructed using the harmonized reconstruction algorithm recommended for multicenter ⁸⁹Zr PET scan trials (18). A single nuclear medicine physician analyzed all the PET scans for visible tracer uptake in tumor lesions and healthy tissues including lymph nodes. The total number and location of measurable tumor lesions, according to irRC, were assessed with diagnostic CT. Tumor lesions with visible tracer uptake on the ⁸⁹Zr-AMG 211 PET were considered quantifiable when the tumor size was at least 15 mm on CT to minimize potential partial volume effect. Radioactivity was quantified by manually drawing spherical volumes of interest (VOI) in healthy tissues and tumor lesions using A Medical Image Data Examiner (AMIDE) software (version 0.9.3, Stanford University, Stanford, CA; ref. 19). In healthy tissues, VOIs were drawn in the blood pool at the place of the thoracic aorta, lung, liver, spleen, kidney, intestine, brain, bone marrow, and bone cortex at the place of the femur, thigh muscle, retroperitoneum, and fat tissue. VOIs were drawn independently by two investigators, K.L. Moek and I.C. Kok, based on maximum intensity projection images of ⁸⁹Zr-AMG 211 PET or the coregistered low-dose CT if delineation was unclear on PET. ⁸⁹Zr-AMG 211 uptake was measured as SUV (formula in Supplementary Materials and Methods: calculations). We reported SUV_max (maximum voxel intensity in the VOI) for tumor lesions and SUV_mean (mean voxel intensity of all voxels in the VOI) for healthy tissues. Outliers were reassessed for accuracy. In case of a discrepancy ≥10% between the two investigators, the discrepancies were discussed and a final conclusion made. In addition, the percentage injected dose per kilogram (%ID/kg) was calculated for all VOIs (formula in Supplementary Materials and Methods: calculations). For the brain, lungs, liver, spleen, and kidneys, we used mean organ weights as reported in sudden death autopsy studies to calculate percentage of the injected dose (%ID; refs. 20, 21). We used the percentage body fat and total bodyweight to assess %ID in fat (22). The total blood volume was calculated according to Nadler's formula (23) and ⁸⁹Zr-AMG 211 serum half-life with a 1-phase decay model using GraphPad Prism software version 5.04.

To study ⁸⁹Zr pharmacokinetics, blood and urine samples were collected at each PET scan time point. In addition, ⁸⁹Zr-AMG 211 binding to immune cells was explored by counting blood fractions, and the integrity was analyzed via gel electrophoresis. More details on ⁸⁹Zr pharmacokinetics are provided in Supplementary Materials and Methods: ⁸⁹Zr pharmacokinetics.

Blood samples for soluble CEA were collected at screening and after the second AMG 211 treatment cycle. Serum soluble CEA upper limit of normal was 5 μg/L. In addition, serum antidrug antibody (ADA) levels were determined in blood samples, collected day 1 before and 7 days after tracer infusion, with an electrochemiluminescent assay for patients imaged during AMG 211 treatment. Tumor CEA expression was verified in archival tumor tissues. CEA membranous and cytoplasmic staining was scored as 3+ for strong, 2+ for moderate, 1+ for weak, and 0 for absence of any staining. A tumor was considered to express the CEA protein if at least 2+ protein expression was seen.

Statistical analyses were performed using SPSS Version 23. Unless stated otherwise, data are shown as median with interquartile range (IQR) or range in case n ≤ 3. Associations between parameters were calculated using the Spearman correlation test. P values < 0.05 were considered significant.

Nine patients were enrolled between August 2016 and May 2017. The ⁸⁹Zr-AMG 211 PET imaging study was terminated in May 2017 because of the completion of the AMG 211 phase I study. ⁸⁹Zr-AMG 211 PET imaging was performed in 7 patients before treatment with 6,400-μg/day AMG 211, in one patient during treatment with 12,800-μg/day AMG 211, and in one patient PET imaging was performed before as well as during treatment with 6,400-μg/day AMG 211. This makes the total number of PET series studied 10. Patient characteristics are shown in Table 1. CEA tumor expression was positive in all 7 patients, from whom archival tumor tissue was available.

Median radioactivity dose administered across all patients was 35.77 MBq (IQR 34.90–36.99 MBq). Because of technical reasons, one 6-hour PET scan of one patient receiving 200-μg ⁸⁹Zr-AMG 211 + 1,800-μg cold AMG 211 was not evaluable.

With 200-μg ⁸⁹Zr-AMG 211 (n = 2), SUV_mean in the blood pool at 3 hours was 2.2, which decreased thereafter (Fig. 2A and B). The addition of 1,800-μg cold AMG 211 (n = 4) resulted in a higher blood pool SUV_mean of 4.0 (IQR 3.2–5.6) at 3 hours. The addition of 4,800-μg cold AMG 211 (n = 2) did not further increase blood pool SUV_mean at any time point. We, therefore, determined that 200-μg ⁸⁹Zr-AMG 211 + 1,800-μg cold AMG 211 was optimal for ⁸⁹Zr-AMG 211 PET imaging before AMG 211 treatment. The corresponding ⁸⁹Zr-AMG 211 serum half-life was 3.3 hours (Supplementary Table S1), indicating the optimal time points for ⁸⁹Zr-AMG 211 PET imaging period to be around 3, and 6 hours after tracer administration. Figure 2D illustrates whole-body maximum intensity projection PET images for all time points of one patient in whom imaging was performed before AMG 211 treatment using 200-μg ⁸⁹Zr-AMG 211 + 1,800-μg cold AMG 211.

Healthy tissue biodistribution in the 4 patients who received 200-μg ⁸⁹Zr-AMG 211 + 1,800-μg cold AMG 211 showed high (Fig. 2A) and prolonged (Fig. 3) tracer uptake in the CD3-rich tissues in spleen and bone marrow. Liver uptake at 3 hours showed a SUV_mean of 3.1 (IQR 2.4–3.5). AMG 211 was at that time already clearly being excreted by the kidneys. Much lower uptake at 3 hours was observed in lung, bone, muscle, abdominal cavity, brain, and body fat (Fig. 2A). In all healthy tissues analyzed, SUV_mean was highest at 3 hours and decreased over time, except for the intestines, in which the SUV_mean increased from 1.9 (IQR 1.5– 2.3) at 3 hours, to 2.5 (IQR 1.7–3.9) at 24 hours. Accumulation of ⁸⁹Zr-AMG 211 was visually observed in the colon, but not in other parts of the gastrointestinal (GI) tract known to physiologically overexpress CEA, like the stomach or esophagus (7). Healthy tissue biodistribution at 3 hours for all imaging dosing cohorts is shown in Fig. 2A. Supplementary Table S2 shows median ⁸⁹Zr-AMG 211 uptake in kidneys, liver, spleen, bone marrow, lung, and intestine across all imaging dosing cohorts per PET scan time point.

In patients receiving 200-μg ⁸⁹Zr-AMG 211 + 1,800-μg cold AMG 211, at 3 hours 26.1 %ID was present in the blood pool, 0.4 %ID in the spleen, 6.1 %ID in the liver, 32.7 %ID in the kidneys, and 3.6 %ID in the total fatty tissue. The %ID at 3 hours across all imaging dosing cohorts is shown in Supplementary Fig. S1.

A total of 61 tumor lesions ≥10 × 10 mm (median per patient: 8, range 2–14) were identified on the basis of a diagnostic CT scan (Supplementary Table S3). Of these lesions, 62% (n = 38) could be visualized on PET. In addition, visual tracer presence was observed in four presumably malignant lymph nodes <10 mm, and one lesion in the sacral bone, which was positioned outside the view of the diagnostic CT scan. Fourteen lesions were visible as “hot spots”, whereas liver (n = 27) and renal (n = 2) metastases appeared visually as “cold spots” due to the relatively high uptake in the surrounding healthy tissue. Of the 43 visible tumor lesions, 37 (86%) were PET-quantifiable (Supplementary Table S3). Two renal lesions were considered not quantifiable due to the extremely high uptake in the surrounding healthy kidney tissue, whereas four lymph nodes suspected to be malignant were not quantifiable due to the small size of these structures, which impeded quantification.

In the imaging dosing cohort given 200-μg ⁸⁹Zr-AMG 211 + 1,800-μg cold AMG 211, a SUV_max of 4.0 (IQR 2.7–4.4) at 3 hours was found in tumor lesions, decreasing to 2.8 (IQR 2.0–3.3) at 24 hours. A patient-based analysis showed a slower tumor ⁸⁹Zr-AMG 211 washout than from the blood pool and from most healthy tissues, except for the spleen, bone marrow, and intestines, indicating tracer specificity (Fig. 3). Figure 4 is a heat map with log ratios for SUV across tumor lesions and healthy tissues for this imaging dosing cohort, showing that the maximum voxel intensity in tumor lesions exceeds the mean voxel intensity in healthy tissues, except for the kidneys. In the other imaging dosing cohorts, at 3 hours, a tumor lesion SUV_max of 2.9 (IQR 2.3–4.4) was found in the 200-μg ⁸⁹Zr-AMG 211 cohort, and a tumor lesion SUV_max of 3.1 (IQR 2.7–5.3) was found in the 200-μg ⁸⁹Zr-AMG 211 + 4,800-μg cold AMG 211 cohort. These findings also confirm that 200-μg ⁸⁹Zr-AMG 211 + 1,800-μg cold AMG 211 is optimal for imaging.

In all imaging dosing cohorts, ⁸⁹Zr-AMG 211 tumor uptake varied greatly within and between patients. To study this heterogeneity in tumor lesion uptake, we used the 6-hour PET scan with higher tumor-to-blood ratios than the 3-hour scan. Lesion-based analysis showed up to a 9-fold difference in ⁸⁹Zr-AMG 211 tumor lesion uptake between patients, irrespective of tumor localization (Fig. 5). Moreover, Fig. 5 illustrates representative PET/CT scans from a patient showing highly heterogeneous ⁸⁹Zr-AMG 211 uptake across lung metastases. Patient-based analysis showed a 5-fold difference in tumor lesion tracer uptake within one organ.

Analysis of relation between tumor uptake and tumor response to AMG 211 treatment was not possible, as response evaluation after the second AMG 211 treatment cycle could only be performed in 2 patients. In the other patients, treatment was stopped prematurely due to either rapid clinical progressive disease (n = 4) or adverse events (n = 1), and one patient did not start with AMG 211 treatment due to clinical deterioration caused by tumor progression.

⁸⁹Zr-AMG 211 imaging immediately after the end of the second AMG 211 treatment period was performed in 2 patients who received 28-day continuous intravenous treatment with either 6,400-μg/day or 12,800-μg/day of AMG 211 per cycle. Because of completion of the phase I treatment part of the study, no additional patients were enrolled in this imaging dosing cohort.

During AMG 211 treatment, we observed an approximately 2–3-fold higher uptake in the blood pool and an approximately 2–3-fold lower uptake in the kidneys when compared with imaging before AMG 211 treatment (Fig. 2A and C). ⁸⁹Zr-AMG 211 serum half-life exceeded 16 hours in one patient (Supplementary Table S1). Seven tumor lesions with a size ≥10 × 10 mm were detected with diagnostic CT. None of these lesions, all located outside the liver and kidneys, visually showed ⁸⁹Zr-AMG 211 uptake. No lesions were identified on PET that were not visible on diagnostic CT.

Whole-blood and urine samples for ⁸⁹Zr-AMG 211 measurements were available for 8 patients who underwent imaging before AMG 211 treatment. The SUV equivalents of ex vivo measurements of blood samples at 3, 6, 24, and 48 hours correlated well with PET-derived SUV_mean blood pool values (Spearman correlation coefficient = 0.983, P ≤ 0.01). In urine, uptake at 3 hours ranged from 13.7 in n = 1 patient receiving 200-μg ⁸⁹Zr-AMG 211 to 35.1 in n = 2 patients receiving 200-μg ⁸⁹Zr-AMG 211 + 4,800-μg AMG 211. In the 200-μg ⁸⁹Zr-AMG 211 + 1,800-μg AMG 211 cohort, the highest radioactivity (28.7 at 3 hours) was measured in urine of one diabetic patient with proteinuria.

A median of 96.07% (IQR 95.84–96.19) of ⁸⁹Zr-AMG 211 was unbound in plasma, and 2.56 % (IQR 2.07–3.07) was bound to buffy coat at 3 hours. ⁸⁹Zr-AMG 211 was intact in plasma, whereas in urine, ⁸⁹Zr-AMG 211 was mostly present in degraded form (Fig. 6).

Two patients had high serum soluble CEA levels at screening, whereas the levels in the other patients ranged between 2.4 and 42.8 μg/L. In one patient who received 200-μg ⁸⁹Zr-AMG 211, the CEA level was 217 μg/L, whereas in the other patient who received 200-μg ⁸⁹Zr-AMG 211 + 1,800-μg cold AMG 211, this level was 320 μg/L. In both patients, imaging was performed before AMG 211 treatment and showed, in comparison with patients from the same imaging dosing cohort, the highest tracer presence in the blood pool.

No induction of ADAs by tracer dose was observed in patients in whom imaging was performed before AMG 211 treatment. When imaging was performed during AMG 211 treatment, ADAs were measured in serum 1 week after tracer administration in both patients.

No ⁸⁹Zr-AMG 211-related toxicity was seen, apart from known adverse events of AMG 211 itself. Two patients, one participating in the 200-μg ⁸⁹Zr-AMG 211 imaging dosing cohort and the other in the 200-μg ⁸⁹Zr-AMG 211 + 4,800-μg cold AMG 211 imaging dosing cohort, experienced fever and/or chills. The first patient also experienced headache. All adverse events occurred within 24 hours after tracer administration and are most likely due to cytokine release. Adverse events were CTCAE grade 1, and resolved spontaneously or after administration of acetaminophen.

This is the first-in-human PET imaging study with a small BiTE antibody construct. With ⁸⁹Zr-labeled AMG 211-targeting CEA/CD3, high specific tracer accumulation was observed in CD3-rich lymphoid tissues such as the spleen and bone marrow and in tumor lesions. ⁸⁹Zr-AMG 211 was rapidly cleared from the blood pool by excretion via the kidneys, whereas uptake in tumor lesions persisted. Tumor lesions showed a clear but heterogeneous uptake within and between patients with gastrointestinal adenocarcinomas.

To date, more than one hundred bispecific antibodies have been developed, including BiTE antibody constructs, dual-affinity retargeting antibodies, and full-length antibodies (1, 24). It is well acknowledged that their development for clinical use has been more challenging for this “high hanging fruit” compared with conventional mAbs (1). The two arms differ in binding affinity for targets, which consequently might affect tissue distribution and accumulation in vivo. In human CD3-expressing transgenic immunocompetent mice bearing a murine tumor transfected with human HER2, the distribution of a HER2-CD3 full-length bispecific antibody was predominantly determined by the CD3 arm (25). This is because high affinity for CD3 reduced the systemic exposure and shifted antibody distribution away from tumors to T-cell containing tissues (25). Moreover, side effects in cynomolgus monkeys were dependent on the affinity of the CD3 part of a full-length CLL-1-CD3 bispecific antibody, with the high-affinity variant being poorly tolerated because of extensive cytokine release (26). In mice cografted with CEA-expressing tumor cells injected into the flank, and human peripheral blood mononuclear cells, fluorescence imaging with a CEA-CD3 full-length bispecific antibody showed tumor-specific accumulation mainly through CEA binding, with only minor contributions from CD3 binding (27). This antibody has a monovalent low affinity for CD3, in comparison with a higher bivalent affinity for CEA. With respect to AMG 211, binding affinity is also higher for CEA than for CD3, with an equilibrium dissociation constant of 5.5 nmol/L for CEA and 310 nmol/L for CD3 (28). Despite the lower affinity for CD3, we observed high ⁸⁹Zr-AMG 211 uptake in the spleen, and bone marrow. Because the CD3 protein complex is a defining feature of the T-cell lineage, uptake in lymphoid tissues known to be T-cell reservoirs indicate tracer specificity (29). The ⁸⁹Zr-AMG 211 accumulation we observed in the spleen and bone marrow likely represents the CD3-mediated uptake. However, this finding should be interpreted with some caution, because for some patients, uptake in spleen and bone marrow was lower than observed in the blood pool. This could indicate that to some extent, tracer uptake is nonspecific, or tissue target saturation was reached. In the GI tract, visual tracer accumulation was limited to the intestines, which may reflect tracer excretion in the gut and feces as well as CEA- and CD3-mediated tracer uptake in gut tissue. Uptake increased over time up to the 24-hour time point, indicating that more time is needed for tracer penetration into GI tissues than into organs with a rich blood supply like kidneys and liver.

We clearly observed uptake in tumor lesions that persisted longer than tracer presence in the blood. These SUVs were higher than expected on the basis of preclinical data in mice-bearing CEA-expressing LS174T human colorectal adenocarcinoma xenografts (14). Moreover, in the clinical setting the CD3 arm can be studied, which is not possible in the preclinical mouse–mouse environment because AMG 211 is not cross-reactive with mouse CD3. Noninvasive whole-body PET imaging studies used to investigate the biodistribution of other drugs have shown considerable heterogeneity regarding tracer uptake in tumor lesions (30–32). We also observed striking intra- and interpatient heterogeneity in ⁸⁹Zr-AMG 211 tumor accumulation before AMG 211 treatment. This might reflect the fact that tracer accumulation is dependent on target expression as well as delivery by tumor vasculature, and tissue permeability (33). IHC target staining of multiple tumor lesions within one patient might have shed light on these differences with regards to the role of target expression. However, multiple biopsies were not part of this trial. Data on heterogeneity was lacking in the small studies, which reported tumor uptake of full-length bispecific antibodies (12, 13). In 1996, the first attempts to radiolabel bispecific antibody OC/TR F(ab’)2 (folate-binding protein-CD3) were made in a small single-photon emission CT study in patients suspected to have ovarian cancer. The tumor could be visualized in 2 of 3 patients, but the study was stopped prematurely because of unexpected severe tracer-related toxicity due to cytokine release at doses as low as 0.1 mg (12). More recently, a preliminary report described ⁸⁹Zr-labeled cergutuzumab amunaleukin (CEA-IL2v) PET in 23 patients with solid tumors, showing CEA-mediated accumulation in tumors and uptake in lymph nodes and spleen (13). Uptake in these lymphoid organs, 5 days after tracer administration, was higher than observed in our study, likely due to the relatively long half-life of full-length antibodies, enabling prolonged tracer exposure.

Our bispecific antibody construct is small (55 kDa), resulting in a short tracer half-life as determined via a 1-phase decay model. Fast serum tracer clearance was also found in PET studies with other small-sized antibody-related radiolabeled approximately 100 kDa F(ab’)₂ fragments of trastuzumab or approximately 15 kDa nanobodies developed as diagnostics patients with in breast cancer (34–36). These kinetics, therefore, require imaging assessments at earlier time points in comparison with approximately 150 kDa mAbs, for which scans are generally performed 4 to 7 days after tracer administration, thus matching the half-life of these compounds (30–32, 37). The small size of a BiTE antibody construct leads to fast renal clearance (38). For this reason, the drug was administered as continuous intravenous infusion (11). Interestingly, currently BiTE antibody constructs are being developed, which contain an Fc-domain (39, 40). This increases their size and leads to an enhanced serum half-life. In nonhuman primates, the serum half-life of various BiTE antibody constructs was extended from 6 to 44–167 hours by the addition of an Fc-domain or albumin (39). These larger BiTE antibody constructs exceed the renal filtration threshold of 60 kDa.

In our study, AMG 211 treatment clearly altered ⁸⁹Zr-AMG 211 biodistribution, leading to high and sustained ⁸⁹Zr-AMG 211 presence in the blood pool, which could reflect tissue target saturation. These findings support the continuous intravenous infusion approach to deliver uninterrupted therapeutic pressure by maintaining AMG 211 exposure of the tumor (2, 10). In addition, the absence of tumor lesion visualization might be indicative of tumor target saturation. Although an approximately 10% to 25% reduced uptake in tumor lesions after treatment has been shown via serial PET imaging for two membrane receptors targeting antibody tracers, clear evidence of tumor saturation was not found in these studies (30, 32). Also, other factors like perfusion and anatomical location could be responsible for lack of tumor visualization we observed in patients imaged during AMG 211 treatment. We observed ADAs in both patients in whom imaging during AMG 211 treatment was performed. Previously, in a phase I study, ADAs were present in 48% of patients who received AMG 211 treatment on days 1 to 5 in 28-day cycles (10), despite the fact that BiTE antibody constructs are thought to be less immunogenic due to the lack of an Fc domain in comparison with full-length antibodies (41). AMG 211 comprises a humanized CEA arm and a deimmunized CD3 arm, therefore mouse residues remain, which may be one cause for ADA generation in the absence of an Fc domain. The presence of ADAs might have altered ⁸⁹Zr-AMG 211 pharmacokinetics and could have led to reduced ⁸⁹Zr-AMG 211 availability in the blood pool by triggering an additional clearance pathway through immune complex formation and subsequent degradation through phagocytic cells in the liver and spleen (42).

In this study, we demonstrated that imaging with ⁸⁹Zr-AMG 211 is very informative regarding CEA/CD3 BiTE antibody construct, whole-body biodistribution, and tumor targeting. We showed CD3-specific tracer accumulation in lymphoid organs and clear tumor uptake that was highly heterogeneous, both within and between patients. This approach can support rational trial design for such innovative antibody targeting strategies.

E.G.E. de Vries is a consultant/advisory board member for Sanofi, Daiichi Sankyo, NSABP, Pfizer, and Merck. No potential conflicts of interest were disclosed by the other authors.

Conception and design: K.L. Moek, F.V. Suurs, J.A. Gietema, M.N. Lub-de Hooge, R.S.N. Fehrmann, E.G.E. de Vries

Development of methodology: K.L. Moek, S.J.H. Waaijer, C.P. Schröder, M.N. Lub-de Hooge

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K.L. Moek, S.J.H. Waaijer, I.C. Kok, A.H. Brouwers, C.W. Menke-van der Houven van Oordt, C.P. Schröder, A. Jorritsma-Smit, M.N. Lub-de Hooge, R.S.N. Fehrmann, D.J.A. de Groot, E.G.E. de Vries

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K.L. Moek, S.J.H. Waaijer, I.C. Kok, A.H. Brouwers, C.W. Menke-van der Houven van Oordt, C.P. Schröder, S.V.K. Mahesh, R.S.N. Fehrmann, D.J.A. de Groot, E.G.E. de Vries

Writing, review, and/or revision of the manuscript: K.L. Moek, S.J.H. Waaijer, I.C. Kok, F.V. Suurs, A.H. Brouwers, C.W. Menke-van der Houven van Oordt, T.T. Wind, J.A. Gietema, C.P. Schröder, A. Jorritsma-Smit, M.N. Lub-de Hooge, R.S.N. Fehrmann, D.J.A. de Groot, E.G.E. de Vries

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K.L. Moek, T.T. Wind, M.N. Lub-de Hooge, D.J.A. de Groot

Study supervision: R.S.N. Fehrmann, D.J.A. de Groot, E.G.E. de Vries

We thank the patients for participating in the study. We thank Sabine Stienen from Amgen Research Munich GmbH, and Kam Cheung from Amgen Thousand Oaks for their advice on trial design and interpretation of data. We thank Anouk Funke for her assistance in figure design. We thank Linda Pot for the labeling procedures, and Johan Wiegers and Cemile Karga for their assistance with PET data transfer. Research support from Amgen was made available to the institution. The study drug was supplied by Amgen.

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.