Probody therapeutic CX-072 is a protease-activatable antibody that is cross-reactive with murine and human programmed death-ligand 1 (PD-L1). CX-072 can be activated in vivo by proteases present in the tumor microenvironment, thereby potentially reducing peripheral, anti–PD-L1-mediated toxicities. To study its targeting of PD-L1–expressing tissues, we radiolabeled CX-072 with the PET isotope zirconium-89 (89Zr).
89Zr-labeled CX-072, nonspecific Probody control molecule (PbCtrl) and CX-072 parental antibody (CX-075) were injected in BALB/c nude mice bearing human MDA-MB-231 tumors or C57BL/6J mice bearing syngeneic MC38 tumors. Mice underwent serial PET imaging 1, 3, and 6 days after intravenous injection (pi), followed by ex vivo biodistribution. Intratumoral 89Zr-CX-072 distribution was studied by autoradiography on tumor tissue sections, which were subsequently stained for PD-L1 by IHC. Activated CX-072 species in tissue lysates were detected by Western capillary electrophoresis.
PET imaging revealed 89Zr-CX-072 accumulation in MDA-MB-231 tumors with 2.1-fold higher tumor-to-blood ratios at 6 days pi compared with 89Zr-PbCtrl. Tumor tissue autoradiography showed high 89Zr-CX-072 uptake in high PD-L1–expressing regions. Activated CX-072 species were detected in these tumors, with 5.3-fold lower levels found in the spleen. Furthermore, 89Zr-CX-072 uptake by lymphoid tissues of immune-competent mice bearing MC38 tumors was low compared with 89Zr-CX-075, which lacks the Probody design.
89Zr-CX-072 accumulates specifically in PD-L1–expressing tumors with limited uptake in murine peripheral lymphoid tissues. Our data may enable clinical evaluation of 89Zr-CX-072 whole-body distribution as a tool to support CX-072 drug development (NCT03013491).
Combining immune checkpoint inhibitors improves survival of patients with advanced stages of several tumor types, but can elicit severe immune-related adverse events (irAE). These irAEs may be caused by immune checkpoint blockade in healthy tissues, and immune checkpoint–inhibiting antibodies with tumor-restricted activity are therefore of interest. Recently, imaging of 89Zr-atezolizumab whole-body distribution in patients with cancer showed high uptake in healthy lymphoid tissues, including spleen, lymph nodes, and Waldeyer's ring. Our preclinical imaging study in mice reveals anti–PD-L1 Probody therapeutic CX-072 is preferentially activated in tumors, followed by PD-L1–mediated uptake, whereas accumulation in spleen and other PD-L1–expressing peripheral lymphoid tissues is limited. These findings demonstrate CX-072 may reduce anti–PD-L1-mediated toxicities in healthy tissues, thereby potentially expanding its use in combination therapies. We developed and characterized clinical grade 89Zr-CX-072, which is currently studied in patients as part of a phase I/II clinical trial (NCT03013491) to support CX-072 drug development.
Immunotherapies targeting immune-regulatory checkpoints have acquired a clear role in clinical cancer care. These therapies improve survival of patients with advanced stages of several tumor types, although not all patients respond (1). Immune checkpoint inhibition can elicit a unique spectrum of immune-related adverse events (irAE) due to the role of these immune checkpoints in maintaining immunologic homeostasis, including toxicities of endocrine, hepatologic, dermatologic, cardiac, and gastro-enteric origin, that can be life-threatening (2). Combining immune checkpoint inhibitors improves response rates and overall survival for specific cancers (3–7), but these combinations often show increases in rate and severity of side effects (8–10). Immune checkpoint inhibitors with reduced peripheral, immune-related toxicities are therefore of interest.
CX-072 is a Probody therapeutic that targets the programmed death-ligand 1 (PD-L1) immune checkpoint. It is currently studied in a phase I/II clinical trial (ClinicalTrials.gov identifier NCT03013491). CX-072 potentially limits irAEs, as it is activated preferentially in the tumor microenvironment (11–13). Tumor-associated proteases can remove the masking peptide that blocks the PD-L1–binding region, yielding activated antibody with approximately 100-fold increased target affinity compared with its intact, inactivated form (Fig. 1).
Due to its design, the CX-072 tissue distribution profile is expected to diverge from other PD-L1–targeting antibodies. PET imaging is a powerful, noninvasive technique to determine in vivo antibody distribution when used with radiolabeling. It provides quantitative spatial and temporal information on tissue-targeting and target-expression. The PET isotope zirconium-89 (89Zr; t1/2 = 78.4 hours) is favorable for radiolabeling antibodies, as its physical half-life matches the time antibodies require for tumor accumulation, resulting in an optimal tumor-to-background signal (14).
Several preclinical imaging studies have reported high uptake of radiolabeled PD-L1–targeting antibodies in murine lymphoid tissues, including spleen, lymph nodes, and thymus, but also in brown adipose tissue (BAT; refs. 15, 16). PET imaging with radiolabeled anti–PD-L1 antibody 89Zr-atezolizumab in patients measured high, heterogeneous uptake in tumor lesions as well as in spleen, nonmalignant lymph nodes and Waldeyer's ring (17).
We performed a PET imaging study in murine models with 89Zr-labeled CX-072 to reveal its whole-body distribution. Also, we compared 89Zr-CX-072 targeting of tumor and lymphoid tissues in both an immune-compromised and an immune-competent setting. To enable clinical PET imaging of 89Zr-CX-072 distribution to tumor and lymphoid tissues in patients, we characterized and developed a good manufacturing practice (GMP)–compliant tracer.
Materials and Methods
Radiolabeling of CX-072, PbCtrl, and CX-075
CX-072, nonspecific Probody control molecule (PbCtrl), and parental antibody CX-075 (CytomX Therapeutics) were allowed to react with an 1:2 molar excess of tetrafluorphenol-N-succinyldesferal (TFP-N-sucDf; ABX GmbH) as described previously (18), with the following modification: pH was set at 4.0 to 4.5 using 1.0 mol/L ammonium acetate instead of 0.025 mol/L sulfuric acid to prevent aggregate formation. CX-072-N-sucDf, PbCtrl-N-sucDf, and CX-075-N-sucDf were purified using a Vivaspin-2 concentrator, aliquoted, and stored at −80°C. Concentration and purity were determined by a Waters size exclusion high-performance liquid chromatography system equipped with a dual-wavelength absorbance detector (280 nm vs. 430 nm), in-line radioactivity detector, and TSK-Gel SW column G3000SWXL 5 μm, 7.8 mm (Joint Analytical Systems; mobile phase: PBS 9.0 mmol/L sodium phosphate, 1.3 mmol/L potassium phosphate, 140 mmol/L sodium chloride, pH 7.2; Hospital Pharmacy UMCG; flow: 0.7 mL/min).
Immunoreactivity after conjugation to TFP-N-SucDf was assessed by indirect ELISA. Note that 96-well plates (Nunc Maxisorp) were coated with 1 μg/mL human extracellular PD-L1 domain (R&D Systems; 156-B7-100) diluted in PBS (Gibco; 0.7 mmol/L sodium phosphate, 1.5 mmol/L potassium phosphate, 154 mmol/L sodium chloride, pH 7.2) and incubated overnight at 4°C. Wells were blocked for 2 hours at room temperature (RT) with 1% BSA (Sigma-Aldrich), 0.05% Tween 20 in PBS. After blocking, plates were incubated with either unconjugated CX-072, PbCtrl, or CX-075 or their respective N-sucDf-conjugates in a concentration ranging from 0.0914 to 600 nmol/L for 60 minutes at RT. Plates were subsequently washed with 0.05% Tween 20 in PBS and incubated with horseradish peroxidase–labeled anti-human IgG antibody (Sigma-Aldrich; A0293) for 60 minutes at RT. Detection was performed with single-component TMB peroxidase substrate (BioRad), and optical density read-out was performed at 450 nm using a micro plate-reader. Immunoreactivity was analyzed by nonlinear regression Log(agonist) versus response in Graphpad Prism v7.0. and was expressed as the effective concentration needed for 50% of receptor occupation (EC50).
PD-L1–expressing human triple-negative breast cancer cell line MDA-MB-231 was a kind gift from Dr. Janet Price, MD Anderson Cancer Center (Houston, TX). Cell lines were confirmed to be negative for microbial contamination and were authenticated in January 2018 by BaseClear using short tandem repeat profiling. The murine PD-L1–positive colon cancer cell line MC38 was obtained from Dr. Walter Storkus, University of Pittsburgh (Pittsburgh, PA). MDA-MB-231 and MC38 cells were cultured in DMEM (Gibco) containing 1.0 g/L glucose and 4.5 g/L glucose, respectively, supplemented with 10% FCS. Cells were incubated at 37°C in a humidified atmosphere with 5% CO2 and were passaged no longer than 6 months. For all experiments, cells were collected in the exponential growth phase.
Flow cytometry experiments in MDA-MB-231 and MC38 cells were performed using CX-075 parental antibody, which is cross-reactive to human and murine PD-L1. Cells were trypsinized and harvested in PBS with 2% FCS and kept on ice prior to use. IgG4 isotype control antibody and CX-075 were diluted in PBS with 2% FCS to 20 μg/mL and incubated with 2 × 105 cells in 1 mL for 1 hour at 4°C. Bound primary antibody was detected using a phycoerythrin-conjugated goat anti-human IgG secondary antibody (Southern Biotech; 2040-09) diluted 1:50 in 2% FCS in PBS and analyzed on a BD Accuri C6 flow cytometer (BD Biosciences). Data analysis was performed with FlowJo v10 (Tree Star), and surface receptor expression was expressed as mean fluorescent intensity.
For flow cytometry experiments in lymphoid tissues, spleen, lymph nodes, bone marrow, BAT, and thymus were collected from BALB/c nude and C57BL/6J mice. Single-cell suspensions were prepared using a cell strainer (Fisher Scientific). Red blood cell lysis buffer (BioLegend) was used for spleen and bone marrow samples. Single cells were analyzed for PD-L1 expression by flow cytometry, using Brilliant Violet 421 anti-mouse PD-L1 (CD274) antibody (BioLegend; 124315) and Brilliant Violet 421 Rat IgG2b isotype control antibody (BioLegend; 400639). Zombie aqua (BioLegend; 423101) was used for detection of viable cells. In splenocytes, PD-L1 expression on specific immune cell types was measured using fluorescein isothiocyanate–conjugated rat anti-mouse CD3 antibody (Thermo Scientific; 11-0032-82) diluted 1:200 in PBS with 2% FCS, allophycocyanin-eFluor780–conjugated rat anti-mouse CD335 antibody (Thermo Scientific; 47-3351-80) diluted 1:20, peridinin chlorophyll protein complex-cyanine5.5-conjugated rat anti-mouse CD11b antibody (Thermo Scientific; 45-0112-80) diluted 1:80, and allophycocyanin-conjugated rat anti-mouse CD19 antibody (Thermo Scientific; 17-0193-80) diluted 1:160. Samples were analyzed on a BD FACS Verse (BD Biosciences). BAT was excluded from measurements, as insufficient amounts of immune cells were present in these samples.
89Zr-CX-075 internalization in MDA-MB-231 and MC38 cell lines
89Zr-CX-075 was used for internalization experiments because it acts as a surrogate for potential internalization of 89Zr-CX-072 after removal of its masking peptide by tumor-associated proteases. Internalization in MDA-MB-231 and MC38 cells was assessed by adding 50 ng 89Zr-CX-075 (200 MBq/mg) to 1 × 106 cells. For control of binding without internalization, cells were incubated for 1 hour on ice. Cells were subsequently washed with ice-cold 1% human serum albumin (Sanquin) in PBS. Total PD-L1–bound 89Zr-CX-075 was determined by measuring cell-associated activity in a calibrated well-type gamma counter (LKB instruments) followed by incubation for 1 and 2 hours at 37°C, whereas controls were kept on ice. Cells were stripped of cell-surface–bound antibody using 0.05 mol/L glycine and 0.1 mol/L sodium chloride (pH 2.8) and subsequently counted for remaining cell-associated activity, ie, internalized PD-L1–bound 89Zr-CX-075, in a calibrated well-type gamma counter. Internalization was expressed as percentage of total PD-L1–bound 89Zr-CX-075, corrected for internalization in 4°C controls.
All animal experiments were performed in accordance with the Dutch code of practice “Animal experiments in cancer research” and were approved by the institutional animal care and use committee of the University of Groningen, the Netherlands.
PD-L1–expressing MDA-MB-231 human triple-negative breast cancer cells were s.c. engrafted in male BALB/cOlaHsd-Foxn1nu mice (Envigo). Tumors were allowed to grow 4 weeks, yielding tumor volumes of ±200 mm3. To study 89Zr-CX-072 target–specific tumor uptake and the potential for antigen saturation, mice received 10-μg 89Zr-CX-072, 89Zr-PbCtrl, or 89Zr-CX-075 (∼5 MBq) supplemented with 0, 40, or 240 μg of unlabeled CX-072, PbCtrl, or CX-075, respectively, via penile vein injection (n = 5–6 per group). To evaluate biodistribution in a fully immune-competent, syngeneic model, male C57BL/6JOlaHsd mice (Envigo) were implanted s.c. with MC38 murine colon adenocarcinoma cells. After 16 days of tumor growth to reach a volume of ±200 mm3, mice received 10 μg of 89Zr-CX-072, 89Zr-PbCtrl, or 89Zr-CX-075 via penile vein injection (n = 5–6 per group).
Mice underwent serial in vivo PET scans 1, 3, and 6 days post intravenous injection (pi) in a Focus 200 microPET scanner (CTI Siemens), followed by tissue collection for ex vivo biodistribution analysis. Tissues were weighed and counted in a calibrated well-type gamma counter. Tracer uptake per organ was quantified by percentage injected dose per gram tissue (%ID/g).
PET data were reconstructed using a two-dimensional ordered-subset expectation maximization reconstruction algorithm with Fourier rebinning, four iterations, and 16 subsets. Data sets were corrected for decay, random coincidences, scatter, and attenuation. PET images were analyzed using AMIDE medical image data examiner software 1.0.5, and regions of interest were drawn for tumor, spleen, and blood pool (ie, heart). Tracer uptake was quantified as mean standardized uptake value (SUVmean).
Tracer integrity was studied ex vivo by SDS-PAGE in plasma obtained after sacrifice on 6 days pi. Total protein concentration was determined by the Bradford assay (20). Mini-Protean TGX precast protein gels 4%–15% (BioRad) were loaded with 40 μg of total plasma protein. As a positive control, freshly radiolabeled (intact) 89Zr-CX-072 was diluted to match activity levels of plasma samples. Gels were allowed to run for 30 to 45 minutes at 100 V. Detection was performed by exposing gels to a multipurpose phosphor plate (Perkin Elmer) overnight at −20°C; exposures were captured using a Cyclone phosphor imager.
Assessment of activated and intact CX-072 in tumor lysates
Tumor homogenates were prepared in Pierce IP lysis buffer (Thermo Scientific) with added Halt protease inhibitor cocktail kit (Thermo Scientific) using a Barocycler (Pressure Biosciences). Protein lysates in IP lysis buffer with protease inhibitor/EDTA were analyzed by the Wes system (ProteinSimple). Activated, intact CX-072 was detected using mouse anti–CX-075 primary antibody (CytomX Therapeutics) and anti-mouse secondary antibody (ProteinSimple; 042-205). Concentrations of activated CX-072 were analyzed using Compass software (ProteinSimple).
Ex vivo tissue preparation, autoradiography, and PD-L1 IHC
For ex vivo tissue analysis, formalin-fixed, paraffin-embedded (FFPE) tumor tissue blocks were prepared. For autoradiography, FFPE blocks were sliced into 4-μm tumor tissue sections and exposed to a phosphor plate overnight at −20°C. Exposures were captured using a phosphor imager (Cyclone).
For PD-L1 IHC, previously autoradiographed FFPE tumor tissue sections were deparaffinized in xylene and rehydrated. Heat-induced antigen retrieval was performed in 10 mmol/L citrate buffer (pH 6.0) for 15 minutes at 95°C to 100°C. Endogenous peroxidase was blocked by 10-minute incubation with dual endogenous enzyme-blocking reagent (Agilent Technologies). Slides were rinsed in 137 mmol/L sodium chloride, 20 mmol/L tris(2,3-dibromopropyl) phosphate, 0.1% Tween-20 (pH 7.6), and subsequently incubated with Dako serum free protein block (Agilent Technologies) for 30 minutes at RT. For human PD-L1 staining, slides were incubated with rabbit anti-human PD-L1 antibody (Abcam; ab205921) or rabbit IgG antibody control (Abcam; ab172730) diluted to 5 μg/mL in Dako serum-free protein block for 1 hour at RT. Thereafter, slides were incubated with Dako EnVision HRP system (Agilent Technologies) for 30 minutes at RT, followed by 10-minute incubation with diaminobenzidine chromogen. Hematoxylin counterstaining was applied routinely. For histologic analysis of tumors, hematoxylin/eosin staining was performed on tissue sections that were sliced subsequent to the sections used for autoradiography. Digital scans of slides were acquired by a Hamamatsu NanoZoomer 2.0-HT multi slide scanner and analyzed with NanoZoomer Digital Pathology viewer software.
Murine PD-L1 staining was performed as described previously (21). Heat-induced antigen retrieval was performed in 10 mmol/L citrate (pH 6.0) in a Lab Vision PT module (Thermo Scientific) at 95°C. Endogenous peroxidase was blocked by 10-minute incubation at RT with 3% hydrogen peroxide in PBS, followed by 15-minute avidin/biotin-blocking (Vector Labs). Slides were preincubated with 10% Dako normal rabbit serum (Agilent Technologies) for 30 minutes at RT. Goat anti-mouse PD-L1 antibody (R&D systems; AF1019) or normal goat IgG control antibody (R&D systems; AB-108-C) diluted to 0.4 μg/mL in PBS with 1% BSA was incubated overnight at 4°C, followed by 30-minute incubation at RT with rabbit anti-goat biotinylated secondary antibody (Agilent Technologies; E046601-2) diluted 1:400 in PBS with 1% BSA. For detection, slides were incubated with Vectastain Elite ABC HRP-kit (Vector Labs) for 30 minutes at RT.
Production of clinical grade 89Zr-CX-072 and stability testing
Methods for conjugation and radiolabeling were transferred to GMP environment. Analytical procedures were validated to demonstrate suitability for use in quality control testing of CX-072-N-sucDf and 89Zr-CX-072. For validation of the manufacturing process, three batches of GMP-compliant CX-072-N-sucDf intermediate product were produced and radiolabeled with 89Zr. For quality control, these batches met specifications on conjugation ratio, activity yield, purity, concentration, pH, radiochemical purity, residual solvents, sterility, and endotoxin content. Preservation of immunoreactivity after conjugation was determined by ELISA as previously described for preclinical conjugated CX-072.
CX-072-N-sucDf intermediate product was stored in sterile vials (Biopure) at −80°C. Stability of CX-072-N-sucDf batch 1 was analyzed at 0, 1, 3, 6, and 12 months after production. Quality control was performed after radiolabeling with 89Zr according to release specifications.
Data were analyzed for statistical significance in GraphPad Prism v7.0 using the Mann–Whitney U test for nonparametric data followed by Bonferroni post test correction for comparison of more than two groups. Experiments were performed at least three times. P values < 0.05 were considered significant.
CX-072 characterization and binding
CX-072, a recombinant, protease-activatable antibody targeting PD-L1, is based on CX-075, a phage-derived, fully human IgG4 antibody that blocks interactions of PD-L1 with PD-1 and B7-1 molecules. In its intact form, both light chains of CX-072 are modified at their N-terminus by addition of a peptide prodomain, which serves to mask the PD-L1–binding region of the antibody. Proteolytic cleavage of the substrate within the prodomain yields the active, PD-L1–binding form of CX-072.
Binding of CX-075 parental antibody, which lacks a masking peptide, was comparable for both human (Kapp = 0.25 nmol/L) and murine (Kapp = 0.30 nmol/L) PD-L1 (Supplementary Fig. S1A), whereas CX-072 binding to murine PD-L1 was approximately 15-fold weaker than for human PD-L1. CX-072 bound to murine PD-L1 at Kapp of 152 nmol/L and bound to human PD-L1 at Kapp of 9.9 nmol/L, as determined by ELISA (Supplementary Fig. S1B).
In vivo 89Zr-CX-072 tumor and spleen uptake over time
For in vivo PET imaging studies, CX-072, PbCtrl, and CX-075 were conjugated to an average of approximately 1.2 TFP-N-sucDf chelators per antibody and thereafter radiolabeled with 500 MBq/mg 89Zr at ≥95% radiochemical purity. Immunoreactivity to PD-L1 was preserved for CX-072-N-sucDf and CX-075-N-sucDf, whereas PbCtrl-N-sucDf remained nonavid (Supplementary Fig. S1C and S1D).
To study human tumor targeting, we performed in vivo studies in human PD-L1–expressing MDA-MB-231 xenograft tumor-bearing BALB/c nude mice. In this model, we compared tumor targeting with targeting of the spleen, because this lymphoid organ is well-developed in immune-compromised mice and can be quantified with imaging. Serial PET scans on days 1, 3, and 6 pi showed tumor accumulation over time for 89Zr-CX-072 and 89Zr-CX-075, but not for 89Zr-PbCtrl (Fig. 2A). PET quantification revealed a 1.5-fold higher spleen uptake for 89Zr-CX-075 than for 89Zr-CX-072 at day 6 pi (P < 0.01). 89Zr-CX-075 spleen uptake was already higher than blood pool levels at day 1 pi and increased to a spleen-to-blood ratio of 2.6 ± 0.5 at day 6 pi (Fig. 2B), suggesting this uptake is PD-L1–mediated. Tracer activity in the blood pool decreased over time, resulting in increasing tumor-to-blood ratios for 89Zr-CX-072 and 89Zr-CX-075 from day 1 to 6 pi with highest tumor uptake observed at day 6 pi. Day 6 was selected for ex vivo biodistribution studies of subsequent mouse cohorts based on highest 89Zr-CX-072 tumor-to-blood ratio of 0.9 ± 0.2, compared with 0.4 ± 0.0 found for 89Zr-PbCtrl (P < 0.01).
Ex vivo 89Zr-CX-072 biodistribution in tumor-bearing mice
Ex vivo tracer uptake quantitation per organ revealed similar biodistribution of 89Zr-CX-072 and 89Zr-PbCtrl in healthy, nontumor tissues (Fig. 3A). Compared with 89Zr-CX-072, blood pool levels and uptake in the heart were lower for 89Zr-CX-075, whereas its uptake in liver, pancreas, stomach, ilium, bone, skin, and spleen was higher.
Next, we assessed whether tracer uptake in tumor and spleen is PD-L1–mediated through a protein dose-escalation study. A radiolabeled antibody dose of 10-μg 89Zr-CX-072, 89Zr-PbCtrl, or 89Zr-CX-075 was supplemented with 0, 40, or 240 μg of unlabeled CX-072, PbCtrl, or CX-075 to obtain total protein doses of 10, 50, and 250 μg for each tracer. We found that 89Zr-CX-072 and 89Zr-PbCtrl biodistribution in healthy tissues was not affected by increased protein dose, but 89Zr-CX-075 uptake in these healthy organs decreased with increasing total protein dose (Supplementary Fig. S2). PET quantification showed blood pool levels of all three radiolabeled molecules were unaffected by increased protein dose (Supplementary Fig. S3).
Ex vivo tracer uptake quantitation showed a protein dose-dependent tumor uptake for 89Zr-CX-072 and 89Zr-CX-075, which was not observed for 89Zr-PbCtrl (Fig. 3B). 89Zr-CX-072 tumor uptake decreased from 8.7 ± 1.0 %ID/g for the 10-μg total protein dose to 6.0 ± 1.3 %ID/g and 4.3 ± 0.7 %ID/g for the 50- and 250-μg dose groups, respectively, indicating competition of tracer with unlabeled CX-072 for PD-L1 receptor-binding. This shows 89Zr-CX-072 tumor uptake is PD-L1–driven. Similarly, 89Zr-CX-075 tumor uptake was reduced by unlabeled antibody. 89Zr-PbCtrl tumor uptake was independent of total protein dose, confirming its nonspecificity for PD-L1.
Although immune-compromised mice were used for this model, PD-L1–mediated spleen uptake was observed for 89Zr-CX-075. Spleen uptake of 89Zr-CX-075 decreased from 25.8 ± 4.1 %ID/g at the 10-μg total protein dose to 10.8 ± 2.8 %ID/g for the 50-μg dose group and 5.3 ± 2.6 %ID/g for the 250-μg dose group. 89Zr-CX-072 did not show protein dose-dependent spleen uptake, similar to 89Zr-PbCtrl, indicating this spleen uptake is not PD-L1–driven, but part of normal, nonspecific antibody distribution.
89Zr-CX-072 and 89Zr-CX-075 showed a comparable tumor uptake of 8.7 ± 1.0 %ID/g and 8.8 ± 2.9 %ID/g, respectively, for the 10-μg total protein dose, whereas only 3.8 ± 0.2 %ID/g was observed for 10-μg 89Zr-PbCtrl. This demonstrates that CX-072's design affects biodistribution to healthy organs, but not its tumor-targeting properties.
Tumor-specific activation of CX-072 species
MDA-MB-231 tumor and spleen lysates were analyzed for presence of activated CX-072 to confirm whether CX-072 is specifically activated by proteases in the tumor microenvironment. Flow cytometry revealed PD-L1 expression by MDA-MB-231 tumor cells and splenocytes of BALB/c nude mice (Fig. 3C). IHC on ex vivo tumor and spleen tissues also confirmed presence of PD-L1. MDA-MB-231 tumor lysates had 6.9 ng/mL activated CX-072 species at the 10-μg total protein dose, 21.2 ng/mL at the 50-μg total protein dose, and highest concentration of 81.7 ng/mL was found for the 250-μg dose group (Fig. 3D). Activated CX-072 level detected in spleen lysates at the 250-μg total protein dose was 5.3-fold lower compared with tumor lysates (P < 0.05). This demonstrates that CX-072 is preferentially activated in tumor tissue and thereafter remains predominantly within the tumor microenvironment. Furthermore, circulating 89Zr-CX-072 tracer remained intact in the blood pool at 6 days pi, as confirmed by SDS-PAGE (Fig. 3E).
89Zr-CX-072 targeting of tumor and lymphoid tissues in a syngeneic mouse model
To assess 89Zr-CX-072 targeting of relevant PD-L1–expressing tissues besides tumor, tracer distribution in BALB/c nude mice was compared with fully immune-competent C57BL/6J mice bearing murine PD-L1–expressing MC38 tumors. PET imaging at 6 days pi revealed high spleen uptake for 89Zr-CX-075 in both mouse models, which was not visible for 89Zr-CX-072 and 89Zr-PbCtrl (Fig. 4A). Importantly, 89Zr-CX-072 and 89Zr-CX-075 showed comparable uptake in syngeneic MC38 tumors at 6 days pi (Supplementary Fig. S4A), whereas 3.1-fold higher spleen uptake was observed for 89Zr-CX-075 compared with 89Zr-CX-072 in C57BL/6J mice (P < 0.01).
PD-L1 expression in spleen, mesenteric lymph nodes, axillary lymph nodes, BAT, and thymus was confirmed by IHC and flow cytometry (Figs. 4B and 5A). Although limited immune cells were found in BAT, IHC revealed PD-L1 expression by adipocytes. High 89Zr-CX-075 uptake was found in these peripheral PD-L1–expressing tissues ex vivo (Fig 4C). In contrast, 89Zr-CX-072 uptake was low in these tissues and comparable with 89Zr-PbCtrl, indicating that CX-072's design limits uptake in lymphoid tissues and BAT. Notably, 89Zr-CX-075 spleen uptake was comparable in both tumor-bearing BALB/c nude and C57BL/6J mice. We found few T cells among BALB/c nude mouse splenocytes, whereas monocytes, NK cells, and B cells were abundant and showed high levels of PD-L1 expression (Fig. 5B).
Ex vivo 89Zr-CX-072 uptake was higher in MDA-MB-231 tumors compared with MC38 tumors (Supplementary Fig. S5). Flow cytometry confirmed MDA-MB-231 and MC38 cell lines both express PD-L1 in vitro, albeit at a higher level in MDA-MB-231 cells (Supplementary Fig. S5A and S5B). After 2 hours of incubation, 31.7 ±1.2% of PD-L1–bound 89Zr-CX-075 was internalized in MDA-MB-231 cells (Supplementary Fig. S5C). This finding supports the idea that 89Zr-CX-072 tracer activity can residualize in the tumor after removal of its masking peptide. Tracer internalization and 89Zr residualization may augment the PET signal, resulting in better tumor visualization. Negligible internalization was found for PD-L1–bound 89Zr-CX-075 in MC38 cells: 2.5 ±0.5% after 2 hours of incubation. Low levels of PD-L1 expression combined with limited internalization in MC38 tumors may explain why the observed 89Zr-CX-072 tumor uptake was not significantly different from 89Zr-PbCtrl (Supplementary Fig. S5D).
Similar to our observations in BALB/c nude tumor-bearing mice, ex vivo tracer uptake quantitation in C57BL/6J tumor-bearing mice revealed comparable biodistribution in healthy organs for 89Zr-CX-072 and 89Zr-PbCtrl (Supplementary Fig. S4B). 89Zr-CX-075 uptake was higher in liver, ilium, and brain, whereas blood pool levels were lower compared with 89Zr-CX-072. 89Zr-CX-075 is potentially affected by target-mediated drug disposition, whereas 89Zr-CX-072 is not, presumably due to protection by its masking peptide. In addition, residual activity measured in tumor-bearing mice at 1, 3, and 6 days pi suggests a slightly faster metabolism for 89Zr-CX-075 compared with 89Zr-CX-072 (Supplementary Fig. S6).
89Zr-CX-072 tumor uptake on autoradiography correlates to PD-L1–expressing tissue
Finally, we studied ex vivo macroscopic tracer distribution in FFPE tumor tissue slices using autoradiography to further confirm PD-L1–specific tumor targeting. Autoradiography revealed a heterogeneous distribution pattern for 89Zr-CX-072 and 89Zr-CX-075, but not for 89Zr-PbCtrl (Fig. 6). PD-L1 staining was observed in both viable and necrotic tumor tissue and correlated to regions showing high uptake of 89Zr-CX-072 on autoradiography. Tumor tissue regions showing low 89Zr-CX-072 uptake on autoradiography had corresponding low levels of PD-L1 expression. In contrast, 89Zr-PbCtrl distributed to nontumor areas, whereas PD-L1 expression was present in viable tumor tissue, indicating that the observed uptake is not PD-L1–specific. 89Zr-CX-075 also distributed to tumor tissue regions expressing high levels of PD-L1.
Clinical grade 89Zr-CX-072 for patient studies
To enable PET imaging in patients, we optimized and validated a robust, GMP-compliant manufacturing process for clinical grade 89Zr-CX-072. Three validation batches of clinical grade CX-072-N-sucDf intermediate product were produced and subsequently radiolabeled with 89Zr, which met prior set quality specifications. CX-072-N-sucDf intermediate product demonstrated stability up to 12 months. Therefore, CX-072-N-sucDf shelf-life is currently set at 12 months and may be extended if the product remains within specifications at future time points. An investigational medicinal product dossier for 89Zr-CX-072 was compiled, submitted, and approved by the competent authority as part of clinical trial application. A clinical study evaluating 89Zr-CX-072 biodistribution in patients is currently ongoing at our center as part of the CX-072 phase I/II study.
This study demonstrates that 89Zr-CX-072 is activated in tumor tissue, followed by PD-L1–mediated tumor uptake, whereas PD-L1–mediated accumulation in healthy, peripheral PD-L1–expressing tissues is limited. Importantly, we found comparable tumor uptake for 89Zr-CX-072 and its parental antibody 89Zr-CX-075 in both human xenograft and syngeneic tumor-bearing mice, indicating that CX-072's required activation by proteases does not hamper tumor uptake.
Low uptake of 89Zr-CX-072 was found in PD-L1–expressing lymphoid tissues and BAT, similar to 89Zr-labeled nonspecific Probody control molecule, which shows this uptake is not mediated by PD-L1. This finding is remarkably different from lymphoid tissue-targeting properties of other antibodies whose PD-L1–binding regions are not masked. PD-L1 is abundantly expressed by subsets of immune cells present in healthy lymphoid tissues of mice and humans (15, 22). As a consequence, PD-L1–targeting antibodies generally show high uptake in the spleen, which potentially affects their biodistribution and pharmacokinetic profile. Several preclinical imaging studies using radiolabeled PD-L1–targeting antibodies in tumor-bearing mice have reported high tracer uptake in the spleen (15, 16, 21, 23–29). PET imaging of PD-L1 in patients with cancer clearly demonstrated high 89Zr-atezolizumab uptake in the spleen among other lymphoid tissues (17).
Most preclinical PD-L1–imaging studies have used radiolabeled antibodies specific for either murine or human PD-L1 (15, 16, 21, 25, 27, 28, 30). 89Zr-CX-072 human/murine cross-reactivity enabled us not only to evaluate its human tumor targeting, but also to acquire data on its uptake in murine lymphoid tissues. Peripheral PD-L1 expression in healthy organs, including lymphoid tissues, strongly affected biodistribution of 89Zr-CX-075 parental antibody, but not biodistribution of 89Zr-CX-072 in tumor-bearing BALB/c nude mice. However, lymphoid tissues in such immune-compromised mice are underdeveloped, leading to absence or anomalous frequencies of specific immune cell types. To overcome this drawback, we used syngeneic MC38 tumor-bearing immune-competent C57BL/6J mice to study CX-072's biodistribution in greater detail.
Interestingly, PET scans and ex vivo biodistribution revealed high uptake of 89Zr-CX-075 parental antibody in spleen and lymph nodes in both BALB/c nude and C57BL/6J mice. Few T cells were present in BALB/c nude mice, but B cells, NK cells, and macrophages were abundant and have been shown to be fully functional (31). We found that PD-L1 is highly expressed by monocytes of BALB/c nude mice, in line with previous research (32). B cells and NK cells also demonstrated PD-L1 expression, but only limited amounts of PD-L1 were present on T cells. Nevertheless, BALB/c nude mice demonstrated levels of PD-L1 expression comparable with immune-competent mice. This shows that even though BALB/c nude mice lack a fully functional immune system, they were well-suited for studying 89Zr-CX-072 targeting of lymphoid tissues, and potentially for other PD-L1–targeting antibodies.
89Zr-CX-072 tumor-specific uptake depends on high protease activity within the tumor microenvironment and limited protease levels in healthy organs, including PD-L1–expressing lymphoid tissues (11, 12). Proteases that can activate CX-072 are generally upregulated during tumorigenesis. They include urokinase-type plasminogen activator (uPA), matriptase (MT-SP1), and selected matrix metalloproteinases, which are all associated with many types and stages of cancer (33–37). As an indirect read-out of protease activity in the tumor microenvironment, we measured high concentrations of activated CX-072 in tumor lysates. This indicates that sufficient levels of proteases were present in tumor tissue to activate CX-072.
Our conclusions are limited by the fact that CX-072, in its intact form, binds with lower affinity to murine PD-L1 compared with human PD-L1. This finding suggests that the masking peptide has greater ability to block CX-072 binding to murine PD-L1. However, once 89Zr-CX-072 is activated, the tracer has similar binding affinity for both murine and human PD-L1. Uptake in PD-L1–expressing lymphoid tissues may be higher in the human setting than we found in mice if large amounts of inactivated 89Zr-CX-072 are taken up by these tissues. Clinical 89Zr-CX-072 PET imaging will reveal whether the tumor-specific activation we observed in mouse models is translatable to patients. To this end, we validated the production process of a GMP-compliant 89Zr-CX-072 tracer for clinical PET imaging as a tool to support CX-072 drug development.
In conclusion, we found that 89Zr-CX-072 accumulates specifically in PD-L1–expressing tumor xenografts with limited uptake in murine peripheral lymphoid tissues and BAT, thus supporting the hypothesis that CX-072 may reduce anti–PD-L1-mediated toxicities in these healthy tissues.
Disclosure of Potential Conflicts of Interest
A research grant to E. G. E. de Vries was obtained from CytomX Therapeutics and made available to the institution. I. Popova, B. Howng, M. Nguyen and O. Vasiljeva are employees of CytomX Therapeutics, which developed and owns the intellectual property rights pertaining to CX-072.
Data and Materials Availability
Data that support the findings of this study are available from E.G.E. de Vries and O. Vasiljeva upon request.
Conception and design: D. Giesen, L.N. Broer, M.N. Lub-de Hooge, O. Vasiljeva, E.G.E. de Vries, M. Pool
Development of methodology: D. Giesen, L.N. Broer, M.N. Lub-de Hooge, B. Howng, O. Vasiljeva, M. Pool
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D. Giesen, L.N. Broer, B. Howng, M. Nguyen, O. Vasiljeva, E.G.E. de Vries, M. Pool
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D. Giesen, L.N. Broer, I. Popova, B. Howng, M. Nguyen, O. Vasiljeva, E.G.E. de Vries, M. Pool
Writing, review, and/or revision of the manuscript: D. Giesen, L.N. Broer, M.N. Lub-de Hooge, I. Popova, B. Howng, M. Nguyen, O. Vasiljeva, E.G.E. de Vries, M. Pool
Administrative, technical, or material support (ie, reporting or organizing data, constructing databases): D. Giesen, L.N. Broer, M. Nguyen, O. Vasiljeva, E.G.E. de Vries, M. Pool
Study supervision: M.N. Lub-de Hooge, O. Vasiljeva, E.G.E. de Vries, M. Pool
We thank Linda Pot-de Jong (University Medical Center Groningen) for assistance in the validation and production processes of clinical grade 89Zr-CX-072; Shanti Davur, Eric Ureno, and Sridhar Viswanathan (CytomX Therapeutics, Inc.) for assistance with production and characterization of 89Zr-CX-072 and 89Zr-PbCtrl; and Gerwin Sandker (Radboud University Medical Center) for helping with IHC on murine PD-L1. Research support from CytomX Therapeutics, Inc. was made available to the institution. The study drug 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.