First-in-Human Study of the Biodistribution and Pharmacokinetics of ⁸⁹Zr-CX-072, a Novel Immunopet Tracer Based on an Anti–PD-L1 Probody

Immune-checkpoint inhibitors yield impressive responses in patients with locally advanced and metastatic malignancies and can result in long-term survival in a subset of cancer patients. Currently, ipilimumab targeting cytotoxic T lymphocyte-associated antigen 4 (CTLA-4) and several programmed cell death protein 1 (PD-1)– and PD-1 ligand 1 (PD-L1)–targeting medicines are registered for multiple tumor types. The antitumor efficacy of immune-checkpoint inhibitors is higher when combined, as seen for nivolumab plus ipilimumab (1–3). Immune-related toxicity is induced by all immune-checkpoint inhibitors. However, the percentage of patients with side effects increases from 20% for single immune-checkpoint inhibitors to 54% when ipilimumab and nivolumab are combined (4).

These immune-related toxicities have stimulated the development of medicines that have similar pharmacologic activity but fewer side effects. CX-072, a Probody therapeutic directed against PD-L1, is a prime example of this. This recombinant monoclonal antibody prodrug is designed to be conditionally activated by proteases in the tumor microenvironment. CX-072 contains a mask with a protease-cleavable substrate at the amino-terminus of the light chain. This mask is devised to block PD-L1 binding until it is released in the tumor microenvironment. The effect of such modifications on antibody biodistribution and kinetics is unknown.

Molecular imaging is increasingly seen as an approach to facilitate drug development (5). An antibody labeled to a positron emission tomography (PET) isotope, such as zirconium-89 (⁸⁹Zr), which complements the long biological half-life of antibodies, allows the collection of data on whole-body drug distribution (6, 7). PD-L1 is expressed by tumor cells, as well as by macrophages, dendritic cells, and T cells. Clinical PET imaging with ⁸⁹Zr-atezolizumab, a zirconium-labeled PD-L1 antibody, showed uptake in tumor lesions and the spleen, lymph nodes, Waldeyer's ring, and sites of inflammation (8). Whole-body PET imaging may also provide information as a biomarker. For example, higher ⁸⁹Zr-atezolizumab tumor uptake correlated with better tumor response and overall survival in cancer patients treated with atezolizumab (8), and uptake on ⁸⁹Zr-trastuzumab PET was predictive for effect of trastuzumab–emtansine therapy in patients with breast cancer (9).

In immune-competent mice, ⁸⁹Zr-CX-072 PET showed specific drug accumulation in PD-L1–expressing human tumors with minor uptake in lymphoid tissues (10). In patients with advanced solid tumors, long-term treatment with CX-072 administered as a single agent or in combination with ipilimumab has shown durable objective responses in patients with advanced solid tumors (11, 12).

We aimed to obtain insight into CX-072 biodistribution and pharmacokinetics (PK) and to learn whether CX-072's Probody therapeutic design influences them. Therefore, we studied ⁸⁹Zr-CX-072 biodistribution in patients with locally advanced or metastatic malignancies.

This is a single-center, prospective substudy of an international multicenter phase I–II treatment study. Eligible patients had measurable advanced or metastatic solid malignancies, archival tumor tissue available, age >18 years, Eastern Cooperative Oncology Group performance status (ECOG PS) of 0–1, an anticipated life expectancy of at least 3 months, and stable hematologic and chemistry laboratory values. The main study and the substudy were approved by the medical ethical committee of the University Medical Center Groningen (UMCG). PD-L1 expression in the tumor was not required for eligibility to the substudy. All patients gave written informed consent. The study is registered in ClinicalTrials.gov (NCT03013491).

Patient screening was performed within 30 days before tracer injection and comprised a baseline computed tomography (CT), electrocardiogram (ECG), and laboratory assessment. After completing the imaging substudy part, patients received CX-072 at their assigned dose cohort 10 mg/kg every 2 weeks, with or without an initial combination with 4 cycles of ipilimumab given at 3 mg/kg at 3 weekly intervals. Treatment with CX-072 was initiated within 10 days after the last PET scan.

Tumor response was assessed every 8 weeks and after one year every 12 weeks. The objective response was defined as a complete or partial response on two consecutive tumor assessments at least 4 weeks apart, according to RECIST version 1.1 and irRECIST (13, 14).

PD-L1 expression in archival tissue biopsies, performed at the latest 27 months before the study start, was determined by PD-L1 immunohistochemistry (IHC) 22C3 pharmDx assay (Dako) and scored as the percentage of viable tumor cells showing partial or complete membrane staining of PD-L1.

Clinical-grade ⁸⁹Zr-CX-072 was developed and produced in the UMCG, as described previously (10). For quality control, ⁸⁹Zr-CX-072 met all release specifications on conjugation ratio, radioactive yield, protein purity, concentration, pH, radiochemical purity, residual solvents, sterility, and endotoxin content. Preservation of immunoreactivity after conjugation was verified by enzyme-linked immunosorbent assay (ELISA).

Cohorts of 2 to 3 patients received 37 MBq of ∼1 mg ⁸⁹Zr-CX-072 plus 0 mg, 4 mg, or 9 mg unlabeled CX-072 (∼1 mg, 5 mg, or 10 mg total tracer protein doses) until sufficient blood pool levels were reached with satisfactory visualization of tumor lesions. Sufficient unlabeled dose supplementation was determined by comparing ⁸⁹Zr-CX-072 SUV_mean in the blood pool at day 4 with other ⁸⁹Zr antibody tracers with well-known kinetics over time (9, 15, 16), thereby taking into account deposition in sink organs known for a PD-L1 antibody (8).

⁸⁹Zr-CX-072 safety was assessed through changes in laboratory test results, changes in vital signs, and summaries of adverse events (AE) before and after exposure to ⁸⁹Zr-CX-072. AE data were recorded according to NCI Common Terminology Criteria for Adverse Events v4.0.

All PET scans were performed on days 2, 4, and 7 after tracer administration and combined with a low-dose CT scan for attenuation correction and anatomic reference, on a Siemens Biograph mCT 40 or 64-slice PET/CT camera, as described previously (8). PET images were reconstructed with the parameters advised for multicenter ⁸⁹Zr-monoclonal antibody PET scan trials, according to EARL1 (17).

Quantification of tracer uptake in normal tissues and tumor lesions was performed in the ACCURATE tool by manually placing a 3D-sphere as the volume of interest (VOI) on each of the three post tracer administration PET scans for each patient (18) (RRID: SCR_020955). Tumor lesions were identified before treatment by conventional imaging techniques according to RECIST1.1. Standardized uptake value (SUV) was calculated using net injected dose, body weight, and measured radioactivity within the VOI on the PET image, corrected for decay. The tracer uptake in healthy tissue was expressed as mean standardized uptake value (SUV_mean; average uptake within a VOI), which is standard for calculating physiologic uptake in homogeneous tissues. SUV_mean is reported with ± standard deviation. In tumor lesions, uptake was calculated as maximum SUV (SUV_max; maximum uptake within a VOI) to assess target-specific uptake.

Uptake of ⁸⁹Zr-CX-072 in Waldeyer's ring, axillary, and inguinal lymph nodes was assessed visually. The uptake was compared to the background as well as liver and spleen.

Blood samples for PK analysis were obtained at 30 minutes and on days 2, 4, and 7 after tracer administration.

Magnetic beads coated with protein A were used to enrich for immunoglobulin (including intact and cleaved CX-072) in K2EDTA plasma samples. Following protein denaturation, reduction, and alkylation, the proteins were digested with trypsin, and two peptide fragments to CX-072 were monitored: 1 peptide from the CX-072 heavy chain present in both the intact and active forms of CX-072 (for quantitation of total CX-072) and 1 peptide from the CX-072 prodomain present only in the intact form of CX-072 (for quantitation of intact CX-072). Corresponding stable isotope-labeled peptides were used as internal standards. The final extract was analyzed via high-performance liquid chromatography mass spectrometry with tandem mass spectrometry detection using positive ion electrospray. The assay has a quantifiable range of 0.657 to 328 nmol/L for intact and total CX-072.

Blood samples were obtained before tracer infusion to study the presence of endogenous anti-drug antibodies (ADA) against CX-072. A validated method using solid-phase extraction with acid dissociation (SPEAD) sample pretreatment followed by a direct electrochemiluminescent assay was used to detect anti-CX-072 antibodies in serum. To maximize the detection of ADAs to CX-072, including antibodies directed against the target binding region, ADAs were extracted from serum using a 1:1 mixture of biotinylated CX-072 and biotinylated CX-075 (parental antibody of CX-072). Following SPEAD processing steps, anti-CX-072 antibodies were detected using a 1:1 mixture of sulfo-tagged CX-072 and sulfo-tagged CX-075. Sample testing was conducted in a tiered manner beginning with a screening assay, which identified presumptive positive or negative samples. This was followed by confirmatory assays. The screening sensitivity is 2.72 ng/mL.

The intactness of the ⁸⁹Zr-CX-072 was studied in serum or plasma, depending on sample availability, collected days 2, 4, and 7 with sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) as described previously (10). To detect a ±4 kDa difference between masked and unmasked antibody, ⁸⁹Zr-CX-072 was reduced to its heavy and light chains using β-mercaptoethanol. Intact ⁸⁹Zr-CX-072 was used as control and diluted to match sample radioactivity. ⁸⁹Zr-CX-072 was detected autoradiographically by exposing gels to a multipurpose phosphor plate (PerkinElmer) overnight at −20°C. Exposures were captured using a Cyclone phosphor imager. Intact tracer and masked versus unmasked antibody were quantified using ImageJ (version 1.52p, RRID: SCR_003070).

Descriptive statistics (i.e., mean and SDs), median counts, and percentages were used to provide an overview of the patient population and ⁸⁹Zr-CX-072 uptake values. Despite the early-phase nature of this study (without a priori–defined testable hypotheses and supporting sample size calculations), and the limited number of patients studied, we performed various exploratory statistical analyses making optimal use of the wealth of repeated measurements that are obtainable using PET imaging in patients with metastatic cancer. To describe the ⁸⁹Zr-CX-072 biodistribution according to administered dose and time post-tracer administration in terms of tumor uptake, we used linear mixed models. The natural logarithmic of the SUV_max was used as dependent variable (to account for its right-skewed distribution), taking within-patient and within-tumor clustering into account using nested random intercepts. The results of these models are expressed as geometric mean uptake levels. To allow post-injection time–tumor–uptake patterns to be variable between dose groups, interaction terms between days post-injection and dose were added to the models. They were tested for statistical significance using the likelihood ratio test under maximum likelihood (such tests also served to test the contribution of other variables to tumor uptake). Besides analyzing days post-tracer administration as a categorical variable, it was also analyzed continuously, choosing the best-fitting form from a linear, a log-linear, or a quadratic curve using the Akaike's Information Criterion under maximum likelihood. The ⁸⁹Zr-CX-072 biodistribution in normal tissues was assessed using similar linear mixed-effect models using SUV_mean as the dependent variable, for which a natural logarithmic (or other) transformation was not necessary.

Again, as strictly exploratory, the relation between best treatment response per patient and tumor uptake was analyzed by similar linear mixed-effect models. The relation between geometric mean tumor SUV_max per patient at day 7 post-injection and overall survival was studied using Firth's penalized maximum likelihood small-sample bias reduction method for Cox regression and assuming linearity and proportionality. All reported statistical tests are two-sided, without correction for multiple testing. Main analyses were performed using R version 3.2.1 for macOS, particularly using packages lme4 (1.1-11, RRID: SCR_015654), lmerTest (2.0-20, RRID: SCR_015656), and coxphf (1.11).

Eight patients were enrolled between March 2018 and November 2018. Patient characteristics are listed in Table 1. All patients completed the entire PET scan series, and none experienced tracer-related AEs. Two patients received a total protein dose of 1 mg, three patients 5 mg, and three patients 10 mg. Thereafter treatment consisted of CX-072 monotherapy, and one patient additionally received ipilimumab (Supplementary Table S2).

In total, 118 tumor lesions were identified, of which 5 were not included in the following analyses as these were irradiated within 3–12 weeks before tracer injection. Tumor characteristics, best response, and the geometric mean of SUV_max per patient at day 7 post-injection are shown in Fig. 1.

Tumor uptake was heterogeneous within and between patients. The highest tumor lesion uptake and the highest geometric mean of SUV_max per patient were seen at the 10-mg dose. All patients showed uptake in their tumor lesions, with an overall geometric mean SUV_max of 4.73 (95% CI: 2.81–7.94; averaged over day 2–7 post-injection). Tumor uptake at 10-mg dose increased between day 2–7 post-injection, but not at other dose levels (Fig. 2). Of the 113 nonirradiated lesions defined by conventional imaging techniques, 84 were visually detected by conventional imaging and by ⁸⁹Zr-CX-072 PET. These included lymph nodes with a short axis of less than 1.5 cm but defined as malignant based on other clinical or CT features, and 5 skin lesions not visible on CT. At the same time, 28 lesions were not distinguishable from the background or did not show uptake by ⁸⁹Zr-CX-072 PET, but were detectable by conventional imaging techniques. One lesion was visible on ⁸⁹Zr-CX-072 PET that was not visible on CT imaging at baseline, but appeared malignant on the CT 6 weeks later.

The median follow-up at data cutoff on November 7, 2020, is 12 weeks (range, 4–76+). One patient experienced stable disease (SD) and two a partial response (PR). The geometric mean SUV_max in tumor lesions from patients with at least stable disease as best response was 6.84 (95% CI: 2.25–20.78; 3 patients with 52 lesions) and was 3.62 (95% CI: 1.57–8.36; 5 patients with 61 lesions) in patients with progressive disease (PD) (likelihood ratio P = 0.14; averaged over day 2–7 post-injection and adjusted for differences in administered dose between patients). The patient with anaplastic thyroid cancer who experienced a PR for 18 months had high ⁸⁹Zr-CX-072 tumor uptake with geometric mean SUV_max 11.49 (Figs. 1 and 2). Furthermore, a patient with microsatellite instable (MSI) high colorectal cancer who responded as well showed uptake in an irresectable sacral local recurrence with SUV_max 3.25 on day 7 (blood pool day 7 at 1 mg mean SUV_mean 1.65), lacking expression of PD-L1 by IHC. This tumor became resectable and was surgically removed after 15 months of treatment. At data cutoff, this patient was free of disease for 11 months. In an exploratory univariable survival analysis accounting for small sample bias (8 patients, 6 events), the hazard ratio for dying was estimated to be 1.32 (95% CI: 0.98–1.93) per unit decrease in per-patient geometric mean tumor SUV_max (likelihood ratio P = 0.070). PD-L1 tumor expression measured immunohistochemically was 90% in the patient with anaplastic thyroid tumor, in the other patients this was ≤1%.

The mean SUV_mean in the blood pool at the 10-mg protein dose on day 4 was 4.27 ± 0.45, indicating sufficient tracer levels available to reach tumor lesions. At this dose, the mean SUV_mean at day 4 in the liver was 5.05 ± 0.75 and the kidney 2.71 ± 0.64. Figure 3 shows the SUV_mean for days 2, 4, and 7 after tracer injection with 1-, 5-, and 10-mg protein dose across the normal tissues.

Remarkable was the spleen SUV_mean, which was the highest of all healthy organs. Corrected for decay at each time point, the SUV_mean of the spleen was highest at 1 mg (n = 2), lower at 5 mg (n = 3), and the lowest at 10 mg (n = 3) (test for effect of dose on uptake: likelihood ratio P = 0.017). In the liver, the average SUV_mean also lowered with increasing tracer protein doses. The spleen uptake was already high on day 2; however, in contrast to what was seen in the tumor lesions, it did not further increase the days thereafter. In both spleen and liver, uptake was stable over time. In other healthy parenchymal tissues, the PET signal decreased over the days after tracer injection.

The bone marrow mean SUV_mean at 10 mg was 2.54 ± 0.49 and remained stable over time. We also studied ⁸⁹Zr-CX-072 uptake in other lymphoid tissues. The ⁸⁹Zr-CX-072 uptake in the Waldeyer's ring on day 7 was present in four patients. Uptake of ⁸⁹Zr-CX-072 was present on day 7, in axillary lymph nodes in four out of seven patients, and in inguinal lymph nodes in five out of eight patients. Percentages of visual uptake in Waldeyer's ring and lymph nodes are shown in Fig. 4. Visual uptake did not differ between tracer protein doses and occurred at all doses (Supplementary Table S1).

⁸⁹Zr-CX-072 uptake was also observed at sites of inflammation in two patients: one with polyarthritis and another with unilateral rhinosinusitis (Fig. 4).

Intact and total CX-072 protein concentrations at evaluated time plots for subjects receiving a total dose of 10 mg are depicted in Fig. 5B. Notably, CX-072 concentration was below the limit of quantitation (LQ) in 32% of all PK samples and corresponded to those patients who received 1- and 5-mg doses. Following the 10-mg dose, intact and total CX-072 concentrations remained above the LQ for PK sampling duration. Measurement of ⁸⁹Zr blood radioactivity was not performed. All subjects were ADA negative at baseline.

⁸⁹Zr-CX-072 integrity in blood samples collected up to 7 days after injection in five patients showed intactness of ⁸⁹Zr-bound CX-072. CX-072 was predominantly present in its masked, unactivated form, up to day 7 (Fig. 5B,–E).

This is the first-in-human study with ⁸⁹Zr-labeled Probody therapeutic PET imaging to study CX-072 drug biodistribution. We showed clear ⁸⁹Zr-CX-072 tumor uptake in all patients, even in lesions of patients lacking IHC PD-L1 tumor expression. There was modest uptake in normal lymphoid organs without unexpected uptake in other healthy tissues.

The rationale behind the Probody therapeutic design is that the molecule is conditionally activated in the tumor. This should reduce the distribution of its activated form in normal tissues compared with current, unmodified, PD-L1 antibodies, while drug uptake in tumor lesions is not affected. In immune-competent mice, ⁸⁹Zr-CX-072 PET did indeed show drug accumulation in PD-L1 expressing human tumors similar to the parental antibody and minor uptake in lymphoid tissues (10). Our current study with ⁸⁹Zr-CX-072 PET allowed further analysis of these properties in humans. We have shown that ⁸⁹Zr-CX-072 was intact in the blood for up to 7 days and predominantly in its masked form in circulation, consistent with Probody therapeutic design and in agreement with findings from the PK analyses from the treatment studies (19). CX-072 shows a linear PK in doses that are comparable to the used tracer proteins doses in this imaging study (19), so the biodistribution of the tracer is representative for CX-072 in therapeutic dose.

All patients showed ⁸⁹Zr-CX-072 uptake into tumor lesions. At 10 mg, uptake in tumor lesions increased between days 2 and 7. In vitro studies have priorly shown ⁸⁹Zr-CX-072 bound to PD-L1 on tumor cells can internalize, followed by intracellular residualization of radioactivity (10). In the current imaging study, at all doses, the tumor-to-blood ratio increased over time and was highest in the 10-mg cohort. Normal healthy organs without target show nonspecific ⁸⁹Zr-CX-072 uptake on the PET scan, which lowers with decreasing blood pool activity. Therefore, increasing ⁸⁹Zr-CX-072 uptake in tumor lesions over time suggests that this tumor uptake is specific and that ⁸⁹Zr-CX-072 is retained in the tumor by its interaction with PD-L1.

Tumor uptake was heterogeneous within and between patients. Also, ⁸⁹Zr-CX-072 tumor uptake was seen in patients whose tumors had low PD-L1 expression by IHC on archival tissue. The same was the case in the ⁸⁹Zr-atezolizumab PET imaging study, in which a biopsy was performed immediately after the last PET scan (8). The ⁸⁹Zr-atezolizumab tumor uptake into that specific biopsied lesion was also seen when PD-L1 staining was negative.

Tumor tracer uptake showed target-mediated, specific kinetics. Therefore, this discrepancy may be explained by heterogeneous PD-L1 expression within one tumor lesion, where tissue analysis is limited when taking only a small biopsy (20). Also, heterogeneity for PD-L1 expression between archival tissue material and current metastases may play a role. Moreover, PD-L1 expression by tumor cells is dynamic, causing temporal differences (20). Therefore, molecular imaging with a tracer that internalizes, including ⁸⁹Zr-CX-072, may provide an accumulated image of target expression over time. In contrast, IHC shows target expression at a given time point and could miss this dynamic expression. Finally, as molecular imaging shows target expression in the total lesion and in the whole body, it overcomes biases on sampling and heterogeneity between and within tumor lesions.

⁸⁹Zr-CX-072 tumor uptake was highest for the 10-mg dose, although the high PD-L1 expression in one patient may have played a role in this. Exploratory analyses suggest a positive association between uptake and patient outcome, corroborating our previous findings in the ⁸⁹Zr-atezolizumab PET imaging study. However, the number of patients is too low to draw firm conclusions. Highest SUV_max occurred in a patient with anaplastic thyroid cancer with tumor PD-L1 expression of 90%, who experienced a partial response. Recent studies of patients with anaplastic thyroid cancer suggest that molecular-based personalized therapies, including PD-L1 antibody treatment, show improvements in survival (21, 22).

At the 10-mg ⁸⁹Zr-CX-072 dose, we had the opportunity to compare the results of PET imaging with 10 mg of the ⁸⁹Zr-labeled PD-L1 antibody atezolizumab (8). At that dose, ⁸⁹Zr-CX-072 uptake in the bone marrow and lymph nodes was indeed lower than ⁸⁹Zr-atezolizumab. Remarkably, after initial significant ⁸⁹Zr-CX-072 uptake in the spleen on day 2, no further increase occurred through day 7. This ⁸⁹Zr-CX-072 spleen uptake was similar to ⁸⁹Zr-atezolizumab at 1 hour after injection. However, ⁸⁹Zr-atezolizumab spleen uptake increased subsequently. PD-L1 is expressed in the spleen by immune cells but also by littoral cells that line the splenic sinuses of the red pulp. High perfusion of the spleen is the primary filter for the blood (23, 24). Therefore, the rapid high uptake of ⁸⁹Zr-CX-072, already on day 2 post-injection, may well be explained by specific binding of ⁸⁹Zr-CX-072 to PD-L1 expressing littoral cells.

The affinity of CX-072 to PD-L1 (K_D = 9.9 nmol/L) is lower compared with atezolizumab (K_D = ∼0.30 nmol/L) when masked and similar when activated (K_D = 0.25 nmol/L; ref. 10). So, even in its masked form, CX-072 still has measurable, although low, affinity for PD-L1, which could result in PD-L1 binding without protease cleavage of the linker and mask.

Interestingly, spleen uptake decreased when protein dose supplementation of CX-072 as part of the tracer increased. This phenomenon was not studied with ⁸⁹Zr-atezolizumab PET imaging, as only a 10-mg dose was applied, but it was also reported for the labeled PD-L1 antibody ⁸⁹Zr-durvalumab (25). Therefore, the rapid high uptake, already on day 2 post-injection, likely by binding of ⁸⁹Zr-CX-072 to PD-L1 expressing littoral cells, is rapidly partly saturated by higher protein doses. The same saturation did not occur in other healthy tissues.

For other lymphoid tissues, for example, bone marrow ⁸⁹Zr-CX-072 quantitative uptake was approximately half as much compared with the ⁸⁹Zr-atezolizumab PET imaging study. Uptake of ⁸⁹Zr-CX-072 in nonmalignant lymph nodes and of ⁸⁹Zr-atezolizumab was scored the same way. Visual uptake of lymph nodes in the ⁸⁹Zr-CX-072 study was ∼20% lower for axillary and inguinal stations compared with ⁸⁹Zr-atezolizumab PET imaging study. Uptake in other healthy tissues was comparable with uptake seen with other antibodies, namely, low uptake in the brain, lung, cortical bone, muscle, a subcutaneous tissue, and higher uptake in liver, kidney, and intestine, reflecting antibody metabolism and elimination (9, 15, 16).

Although procedures, equipment, and reconstruction protocol were uniform, results comparing ⁸⁹Zr-CX-072 with ⁸⁹Zr-atezolizumab should be interpreted cautiously, as the current study was not a head-to-head comparison of the two antibodies. However, a comparison of the kinetics of both tracers shows ⁸⁹Zr-CX-072 has similar kinetics of nonspecific uptake in normal tissues, similar kinetics of specific uptake in tumor lesions, and less specific uptake in organs with the target outside the tumor microenvironment.

This is the first report of imaging in humans using a ⁸⁹Zr-labeled Probody therapeutic that is conditionally activated in the tumor microenvironment. The findings of accumulation in the tumor, and modest uptake in normal lymphoid organs, support tumor-associated protease cleavage of the mask with subsequent target engagement in the tumor and decreased target engagement outside the tumor. Taking together, our findings suggest that already significant insight into the whole-body distribution of a new specialized designed medicine can be reached with a small number of patient subjects.

M.N. Lub-de Hooge reports conflict of interest related to CytomX-UMCG. A.H. Brouwers reports grants from CytomX Therapeutics, Inc. and nonfinancial support from CytomX Therapeutics, Inc. during the conduct of the study. M.T.L. Nguyen is a full-time employee of CytomX Therapeutics, Inc. H. Lu is a full-time employee of CytomX Therapeutics, Inc. J.A. Gietema reports grants from Roche, AbbVie, and Siemens during the conduct of the study. D.J.A. de Groot reports grants from CytomX Therapeutics, Inc. during the conduct of the study. O. Vasiljeva is a full-time employee of CytomX Therapeutics, Inc. E.G.E. de Vries reports grants from CytomX Therapeutics, Inc. during the conduct of the study, as well as other support from Daiichi Sankyo, NSABP, and Crescendo Biologics, and grants from Amgen, Bayer, Crescendo Biologics, G1 Therapeutics, Genentech, Regeneron, Roche, Servier, and Synthon outside the submitted work. No disclosures were reported by the other authors.

L. Kist de Ruijter: Writing–original draft, writing–review and editing, acquisition of data (acquired and managed patients, etc.), analysis and interpretation of data. J.S. Hooiveld-Noeken: Writing–review and editing, acquisition of data (acquired and managed patients, etc.). D. Giesen: Writing–review and editing, acquisition of data. M.N. Lub-de Hooge: Conceptualization, writing–review and editing, acquisition of data. I.C. Kok: Writing–review and editing, acquisition of data (acquired and managed patients, etc.). A.H. Brouwers: Methodology, writing–review and editing, image interpretation. S.G. Elias: Data curation, formal analysis, visualization, writing–review and editing, analysis and interpretation of data (statistical analysis, biostatistics, computational analysis). M.T.L. Nguyen: Methodology, writing–review and editing. H. Lu: Data curation, methodology, writing–review and editing. J.A. Gietema: Writing–review and editing, acquisition of data (acquired and managed patients, etc.). M. Jalving: Writing–review and editing, acquisition of data (acquired and managed patients, etc.). D.J.A. de Groot: Conceptualization, formal analysis, writing–review and editing, acquisition of data (acquired and managed patients, etc.). O. Vasiljeva: Conceptualization, data curation, writing–review and editing. E.G.E. de Vries: Conceptualization, data curation, writing–original draft, writing–review and editing.

We thank Dr. R. Boellaard for support PET analyses. S. Davur, E. Ureno, and S. Viswanathan (CytomX Therapeutics, Inc.) assisted with producing and characterizing clinical-grade ⁸⁹Zr-CX-072. We thank the Clinical Development team of CytomX Therapeutics, Inc. (especially V. Huels, M. Will, and S. Yalamanchili) for the support of substudy design and execution. A research grant of CytomX Therapeutics, Inc. supported this study to the UMCG, Groningen, the Netherlands. The study drug CX-072 and control molecules were supplied by CytomX Therapeutics, Inc. (PROBODY is a U.S. registered trademark of CytomX Therapeutics, Inc.).

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.