Imaging PD-L1 Expression with ImmunoPET

Development and Characterization of C4, aHuman RecombinantAntibody against an Extracellular Epitope on Human and Mouse PD-L1

Anti-PD-L1 antibodies were isolated from a 3 × 10¹⁰ diversity Fab phage display library.¹³ Following 3 rounds of biopanning against the biotinylated ectodomainof the human PD-L1, 27 antigen-positive Fabs of unique sequences wereidentified after screening 249 clones. Of these positive binders,clone C4 was found to cross-react with both human and mouse PD-L1proteins by ELISA, which share 73% protein sequence identity withinthe extracellular domain. C4 was further evaluated for its abilityto recognize full length PD-L1 expressed on cell surface. Using flowcytometry, we show that C4 binds to both human and mouse PD-L1-expressingcells in a dose-dependent manner, with only a slight difference inits EC₅₀ (Table 1).

We nexttested binding of C4 to recombinant, purified human andmouse PD-L1 from R&D and Sino Biological Inc. C4 antibody bindsto human PD-L1 from R&D and Sino Biological Inc. with a K_d in the low nanomolar range (4.2 ± 0.7nM and 1.5 ± 0.34 nM, respectively). C4 binds weakly to mousePD-L1 from Sino Biological Inc. (360 ± 63 nM) but does not bindto mouse PD-L1 from R&D even at high concentrations (4 μM).These findings may indicate that mouse PD-L1 may harbor features oncells required for C4 binding that are not present on recombinantantigen.

Radiolabeling and in Vitro Pharmacological Assessment of C4IgG

Desferrioxamine B (DFO) is a well-established potentchelator of zirconium-89 and has been used to append ⁸⁹Zr to antibodies for both animal and human studies.¹⁴ On this basis, we conjugated DFO to C4 by reacting commercial N-succinimidyl DFO with solvent exposed lysines. An analysisof chelate number showed that there was approximately 1 DFO per antibody(Supporting Information Figure 1). We nexttested whether DFO impacted the affinity and immunoreactivity of C4.Competition binding assays on B16F10 and PC3 cells showed virtuallyno change in the binding of C4 for mouse and human PD-L1, respectively(Table 1). The immunoreactivefraction of ⁸⁹Zr-C4 was assayed using PC3 cells, and ∼93%of the radiolabeled lot was immunoreactive for PD-L1 (Supporting Information Figure 2). Lastly, thestability of ⁸⁹Zr-C4 was measured in serum and was determinedto be >98% over 5 days in mouse serum, making it suitable for invivouse (Supporting Information Figure 3).

⁸⁹Zr-C4 Specifically Binds Mouse and Human PD-L1in Vivo

To establish whether ⁸⁹Zr-C4 can detecthuman and mouse PD-L1 expression in vivo, male nu/nu mice bearing H1975 xenografts (a PD-L1 positive model of human non-small-celllung cancer bearing oncogenic EGFR L858R/T790M) were injected withthe radiotracer via the tail vein. The biodistribution of ⁸⁹Zr-C4 was monitored over time with small animal PET/CT (Figure 1A). Analysis of theimaging data suggested that peak radiotracer accumulation was observedat 48 h after injection. Biodistribution studies at 8, 24, 48, 72,and 120 h after injection confirmed that the highest tumor uptakewas observed at 48 h (Figure 1B and Supporting Information Figure 4). Moreover, radiotracer uptake in the tumor exceeded that of bloodpool and muscle, suggesting it to be receptor mediated uptake andnot nonspecific uptake due to the enhanced permeability and retentioneffect. Lastly, bone uptake was low and did not increase over time,supportive of the stability of ⁸⁹Zr-C4 in vivo (free Zr⁴⁺ salts can have tropism to the bone). To confirm PD-L1 specificuptake in tissues, a separate mouse cohort received ⁸⁹Zr-C4or ⁸⁹Zr-C4 with 30-fold excess C4. Clear evidence of invivo blocking in the tumor tissue was observed 48 h after injectionby PET and biodistribution (Figure 1C; see Supporting Information Figure 5).

Figure 1

Defining the optimal time after injection to study PD-L1 expressionlevels in a xenograft model. (A) Representative coronal and transaxialPET/CT images of male nu/nu mice bearing subcutaneousH1975 tumors, a human model of NSCLC , show that peak tumor uptakeof ⁸⁹Zr-C4 occurs 48 h after injection. (B) Biodistributiondata also show peak tumor uptake of the radiotracer 48 h after injection.High uptake is also observed in PD-L1 positive tissues like the livera spleen. (C) Representative data from a blocking study acquired 48h after injection show the tumor specific uptake of ⁸⁹Zr-C4.Blocking was performed with 30-fold excess C4. Radiotracer uptakeexceeded that observed in the blood pool and muscle: (∗) P < 0.01.

⁸⁹Zr-C4 Detects PD-L1 Expression in a Patient DerivedXenograft from an Individual That Responded Durably to Anti-PD-L1Therapy

Using the conditions defined in the study of H1975tumors, we next tested whether ⁸⁹Zr-C4 can detect PD-L1expression on a PDX model of EGFR mutant (L858R) NSCLC that was isolatedat UCSF in 2014.¹⁵ The PDX was established7 months prior to the individual beginning treatment with pembrolizumabplus ipilimumab. The patient experienced a confirmed partial responsedetected by CT that lasted for 8 months (Figure 2A). Immunofluorescence of tissue sectionsshowed that the tumor tissue expresses PD-L1 (Supporting Information Figure 6). Three mice bearing bilateralsubcutaneous PDXs were injected with radiotracer and imaged seriallyout to 48 h. PET/CT showed that ⁸⁹Zr-C4 clearly resolvedthe tumors (Figure 2B). Moreover, there was no visible accumulation of heat denatured ⁸⁹Zr-C4 in the tumors from a separate mouse cohort (Figure 2B), suggesting thatthe tumor accumulation of ⁸⁹Zr-C4 is receptor mediated.Biodistribution data showed that tumor uptake of ⁸⁹Zr-C4was ∼5% ID/g and also ∼10-fold higher than the tumorassociated activity detected in mice treated with heat denatured ⁸⁹Zr-C4 (Figure 2C and Supporting Information Figure 7).Moreover, the amount of ⁸⁹Zr-C4 in the tumor was higherthan the activity observed in blood pool, muscle, and bone, underscoringthat uptake is due to specific receptor binding, and the radiotraceris stable in vivo.

Figure 2

⁸⁹Zr-C4 detects PD-L1 expression levels ina PDX derivedfrom a NSCLC patient that experienced a durable clinical responseto anti-PD-1 and anti-CTLA4 therapies. (A) Transaxial CT slices showinga soft tissue lesion in the lung prior to the initiation of pembrolizumaband ipilimumab (left), and a smaller mass 3 months after the startof therapy (right). The position of the tumor is indicated witha white arrow. This patient experienced a partial response for 8 months.The PDX was derived 7 months prior to the first CT scan. (B) Smallanimal PET/CT data showing the biodistribution of ⁸⁹Zr-C4in mice bearing bilateral PDX tumors in the flank. The tumors canbe clearly resolved, and radiotracer uptake in abdominal tissues likethe liver is observed, as expected for a large biomolecule. Mice treatedwith ⁸⁹Zr-C4 that was heat denatured (HD) for 10 min priorto injection show no evidence of radiotracer uptake in the tumor.(C) Biodistribution data showing the uptake of ⁸⁹Zr-C4in the PDX tissue 48 h after injection. The uptake is higher in thetumor compared to heat denatured ⁸⁹Zr-C4 (HD) and standardreference tissues like the blood and muscle. (D) Biodistribution dataacquired 48 h after injection in mice bearing subcutaneous H1975,PC3, A549, and the PDX tumors show the different degree of ⁸⁹Zr-C4 uptake in the tumors.

We tested specific binding of ⁸⁹Zr-C4 in PC3 andA549xenografts, two human models of prostate cancer and NSCLC that wedetermined expressed less PD-L1 compared to H1975 using a saturationbinding assay (Supporting Information Figure 8). Biodistribution studies performed 48 h after injection showedradiotracer accumulation in the PC3 and A549 xenografts to be ∼7%ID/g and 5% ID/g, respectively (Figure 2D). In each case, a blocking study was performed andconfirmed that the tumor uptake was receptor-mediated (Supporting Information Figures 9 and 10).

⁸⁹Zr-C4 Detects Tumor Associated PD-L1 Expressionin Immunocompetent Mice with Endogenous PD-1

To be usefulclinically, a radiotracer targeting PD-L1 must be capable of competingwith PD-1 antigen in the tumor microenvironment. To assess whether ⁸⁹Zr-C4 can detect mouse PD-L1 in an immunocompetent background,C57 BL/6J mice were inoculated in the flank with the PD-L1 positivemouse melanoma cell line B16 F10. Tumor bearing mice were treatedwith ⁸⁹Zr-C4 and imaged with PET/CT serially for 5 days(Figure 3A). As withthe H1975 study, visual inspection of the normalized PET slices showedpeak tumor uptake at 48 h after injection. Biodistribution data confirmedthis trend in the tumor associated activity (Figure 3B and Supporting Information Figure 11). Significant accumulation of ⁸⁹Zr-C4was also detected in liver and the spleen, consistent with data reportedfor other radiotracers targeting PD-L1 (∼7% ID/g and ∼6%ID/g, respectively, 48 h after injection). ⁸⁹Zr-C4 uptakewas visibly suppressed in the tumor and some normal tissues (liver,spleen) at 48 h after injection with a co-injection of 10-fold molarexcess unlabeled C4 (Figure 3C). Biodistribution data also showed tumor uptake was suppressedby cold C4, as was uptake in the liver and spleen (Figure 3D and Supporting Information Figure 12).

Figure 3

⁸⁹Zr-C4 can specifically detectmouse PD-L1 in tumorsestablished in an immunocompetent background. (A) Representative coronaland transaxial PET/CT images of male C57BL/6 mice bearing subcutaneousB16 F10 tumors, a mouse model of melanoma, show that peak tumor uptakeof ⁸⁹Zr-C4 occurs 48 h after injection. (B) Biodistributiondata also show peak tumor uptake of the radiotracer 48 h after injection.High uptake is also observed in PD-L1 positive tissues like the livera spleen. (C) Representative data from a blocking study acquired 48h after injection show the tumor specific uptake of ⁸⁹Zr-C4.Blocking was performed with 10-fold excess C4. Radiotracer uptakeexceeded that observed in the blood pool and muscle: (∗) P < 0.01.

Pharmacologically Induced Changes in PD-L1 Expression with Standardof Care Chemotherapies Can Be Quantified with ⁸⁹Zr-C4 ImmunoPET

Numerous reports have disclosed that standard of care chemotherapyor radiation therapy commonly administered prior to anti-PD1/PD-L1therapies can impact PD-L1 expression in the tumor microenvironment.^8,16,17 This observation has clear clinicalsignificance, as a standard treatment course could alter the PD-L1status of a patient’s tumor compared to an archival biopsythat might be used to qualify a patient for immunotherapy. On thisbasis, we tested whether ⁸⁹Zr-C4 PET can measure treatmentinduced changes in PD-L1 expression over time.

We first establishedon H1975 and B16 F10 cells that 48 h of exposure to paclitaxel (Taxol)and doxorubicin induced and repressed PD-L1 transcription and cellsurface expression, respectively, as expected (Supporting Information Figures 13 and 14). We next testedwhether these expression changes were sufficiently large to be quantifiedwith ⁸⁹Zr-C4 in vivo. A cohort of nu/nu mice bearing subcutaneous H1975 tumors was treated with vehicle,paclitaxel (20 mg/kg), or doxorubicin (2 mg/kg) for 3 days. At day3, ⁸⁹Zr-C4 was administered via tail vein, and treatmentwith vehicle or drug was continued for 2 days while the radiotracerwas distributed. PET/CT and biodistribution studies revealed that ⁸⁹Zr-C4 was substantially higher in the tumors of mice treatedwith paclitaxel compared to the other treatment arms, while ⁸⁹Zr-C4 was significantly lower in tumors from mice treated with doxorubicin(Figure 4A and Figure 4B). There were nosubstantial differences in the biodistribution of ⁸⁹Zr-C4in normal tissues, with the exception of increased uptake in the spleensof nu/nu mice treated with paclitaxel (Supporting Information Figure 15). An additionalcohort of C57BL/6 mice bearing subcutaneous B16 F10 tumors was treatedusing the same schema. While doxorubicin had limited effect on ⁸⁹Zr-C4 uptake in the tumor, paclitaxel increased radiotraceruptake in the tumor (Figure 4C and Figure 4D; see Supporting Information Figure 16). Region of interest analysis also showed that the increase in tumoraccumulation of ⁸⁹Zr-C4 was statistically significant (Supporting Information Figure 17).

Figure 4

⁸⁹Zr-C4 can detect pharmacologically induced PD-L1 expressionchanges on the tumor cell. (A) Representative coronal and transversePET images showing the distribution of ⁸⁹Zr-C4 48 h afterinjection in a cohort of male nu/nu mice bearingsubcutaneous H1975 xenografts and treated with vehicle, paclitaxel(Taxol), or doxorubicine. The mice were treated with 20 mg/kg paclitaxelor 2 mg/kg doxorubicine for 2 days prior to radiotracer injection.(B) Representative biodistribution data in the tumor and selectednormal tissues showing that paclitaxel increases tumor PD-L1 expressionlevels, while doxorubicin suppresses it compared to vehicle. No impactwas observed on PD-L1 expressing normal tissues like liver and spleen.(∗) P < 0.01; (#) P <0.05. (C) Representative coronal and transverse PET images showingthe distribution of ⁸⁹Zr-C4 48 h after injection in a cohortof male C57BL/6 mice bearing subcutaneous B16 F10 xenografts and treatedwith vehicle, paclitaxel, or doxorubicine. The mice were treated with20 mg/kg paclitaxel or 2 mg/kg doxorubicine for 2 days prior to radiotracerinjection. (D) Representative biodistribution data in the tumor andselected normal tissues showing that paclitaxel increases tumor PD-L1expression levels, while doxorubicin had no impact compared to vehicle.PD-L1 expression was upregulated in the spleen of mice treated withdoxorubicin: (∗) P < 0.01.

Of the fourpreviously described radiolabeled anti-PD-L1 antibodies,three are neither human nor humanized, which substantially diminishesthe likelihood of near term clinical translation.¹⁸⁻²⁰ One humanizedantibody, atezolizumab, has been labeled and studied in animals.^11,12 While the authors initially labeled the antibody with indium-111for SPECT and a near-infrared dye, two imaging approaches that arefar more challenging to quantify compared to PET, they have more recentlydisclosed data showing tumors can be detected with ⁶⁴Cu-labeledantibody and PET. One aspect of the study with ⁶⁴Cu-atezolizumabperhaps deserving more careful review is the high background of theradiotracer. Indeed, the authors reported ∼10% ID/g in severaltumors known to be PD-L1 null. This value exceeds the specific accumulationof ⁸⁹Zr-C4 in several tumor models we have shown to bePD-L1 positive, including the clinically relevant NSCLC PDX. Moreover,it exceeds the threshold for the enhanced permeability and retentioneffect, assigned to be ~3–4% ID/g for biomolecules.²¹ Blood pool values of ⁶⁴Cu-atezolizumabwere also quite high at late time points after injection (∼10%ID/g at 48 h), something we did not observe with ⁸⁹Zr-C4despite both being large immunoglobulins. Lastly, all of the priorstudies utilized cell lines that overexpress PD-L1 endogenously orvia stable genetic engineering, and among the breast cancer cell lineswith endogenous overexpression, there was no evidence provided ofresponse to anti-PD-1/PD-L1 therapies which unfortunately makes itchallenging to contextualize the clinical significance of the imagingdata.

General Methods

B16 F10, H1975, A549, and PC3 cellswere acquired from ATCC and subcultured according to manufacturer’srecommendations. Paclitaxel (Taxol) and doxorubicin were purchasedfrom Sigma-Aldrich and solubilized with DMSO for in vitro studies. N-Succinimidyl-DFO was obtained from Macrocyclics (Dallas,TX) and used without further purification. Zirconium-89 was purchasedfrom 3D Imaging, LLC (Maumelle, AR). Iodine-125 was purchased fromPerkinElmer.

Antibody Generation and Characterization

The protocolsfor biopanning, phage amplification, ELISA screening, Fab and IgGexpression and purification, as well as affinity/specificity characterizationof the isolated antibody clone C4, closely followed a previous publication.¹³ Briefly, biopanning was performed using humanPD-L1 (Sino Biological) biotinylated by the EZ-Link NHS-PEG4-biotinlabeling kit (Pierce). In the first two rounds of biopanning, PD-L1was immobilized on M280 streptavidin-coated magnetic beads (Life Technologies);in the third round, biotinylated PD-L1 was immobilized on the neutravidin-coatedmicroplate in order to avoid isolation of streptavidin magnetic beadsbinders. 10¹³ cfu phage in 1 mL of casein-PBS blockingbuffer was used in the first round, and 10¹¹ cfu phagewas used in the second and third rounds. Binding was performed for1 h at room temperature, and the beads or wells were washed with PBS,0.1% Tween-20. Concentrations of PD-L1 used for biopanning were 100nM, 20 nM, and 0.2 μg/mL, and the number of washes were 5, 10,and 25 times in rounds 1, 2, and 3, respectively. Bound phage wereeluted by 0.1 M triethylamine for 10 min followed by addition of one-halfvolume of 1 M Tris-HCl, pH 8. Eluted phage were subsequently amplifiedaccording to previously described protocols.²² After three rounds of biopanning, the Fab of selected clones wasexpressed in E. coli TG1 cells (Stratagene) to screenfor PD-L1 binders by ELISA. Unique clones were identified by DNA fingerprintingtechnique and confirmed by DNA sequencing. Fabs were reformatted intohuman IgG1 in the pTT5 vector, and the antibodies were expressed inHEK293-6E cells²³ Both the vector and cellswere obtained from National Research Council of Canada. Antibodieswere purified from the culture supernatant using protein G resin (MerckMillipore) following standard protocols.

To characterize cellsurface binding, suspension HEK6E cells were transfected with eitherhuman or mouse PD-L1 gene cloned in a CMV promoter-driven plasmid(Sino Biological, Beijing). 48 h later the cells were exposed to variousconcentration of C4 followed by a PE-labeled anti-human Fc-specificsecondary antibody (Biolegend, San Diego, CA). Mock transfection ofcells using the same plasmid lacking PD-L1 gene was used as a negativecontrol. Binding of C4 to the cells was quantified by the geometricmean fluorescence intensity (MFI), and the half maximal effectiveconcentration (EC₅₀) was determined by the sigmoidal nonlinearregression equation.

Kinetic constants for C4 antibody againsthuman and mouse PD-L1(R&D and Sino Biological Inc.) were determined using an OctetRED384 instrument (ForteBio). Six concentrations of each antigen (250nM, 125 nM, 62.5 nM, 31.25 nM, 15.625 nM, and 7.812 nM for human PD-L1and 4 μM, 2 μM, 1 μM, 500 nM, 250 nM, and 125 nMfor mouse PD-L1) were tested for binding to the C4 antibody immobilizedon anti-human IgG Fc Capture biosensors (Fortebio). All measurementswere performed at room temperature in 384-well microplates, and therunning buffer was PBS with 0.5% (w/v) bovine serum albumin (BSA)and 0.05% (v/v) Tween 20.

C4 antibody was loaded for 180 s froma solution of 300 nM, baselinewas equilibrated for 60 s, and then the antigens were associated for600 s followed by 1200 s disassociation. Between each sample, thebiosensor surfaces were regenerated three times by exposing them to10 mM glycine, pH 1.5, for 5 s followed by PBS for 5 s. Data wereanalyzed using a 1:1 interaction model on the ForteBio data analysissoftware version 8.2.

Bioconjugation Chemistry

C4 (208μL at a concentrationof 9.6 mg/mL) was dispersed in 200 μL of 0.1 M sodium bicarbonatebuffer (pH 9.0). The final reaction mixture was adjusted to a totalvolume of 0.5 mL by adding a sufficient amount of 0.1 M sodium bicarbonatebuffer. p-Isothiocyanatobenzyldesferrioxamine(Df-Bz-NCS) was dissolved in DMSO at a concentration of 20 mM. Df-Bz-NCSsolution was added to the antibody solution to give a 3.9 molar excessof the chelator over the molar amount of mAb. The Df-Bz-NCS was addeddropwise and mixed rigorously during the addition. The concentrationof DMSO was kept below 2% v/v to avoid any precipitation. After 30min at 37 °C, the reaction mixture was purified with a PD-10column using a gentisic acid mobile phase (5 mg/mL of gentisic acidin 0.25 M sodium acetate, pH 5.4–5.6). The pH of the elutedC4-DFO solution was adjusted to 7.0 and stored at −20 °Cuntil time of use.

Determination of Relative Binding Affinityof DFO-C4

The binding of fixed concentration of a radioligand ¹²⁵I-C4 (0.4 μM) was also measured at equilibrium onPC3 and B16F10cells in the presence of an incrementally increasing series of concentrationsof a DFO-C4 and naked C4 (a range of 10 μM to 1 pM). The cocktailwas incubated at room temperature for 1 h. Following incubation, thecells were washed twice with ice cold PBS and retained for analysis(unbound fraction). The cells were lysed with 1 mL of 1 M NaOH andcollected (cell associated fraction). The unbound and cell associatedfractions were counted in a γ counter and expressed as a percentageof the total activity added per number of cells. The data were plotted,and the IC₅₀ was determined using PRISM software.

Radiochemistry

A solution of ⁸⁹Zr-oxalicacid (5 mCi; 10 μL) was neutralized with 2 M Na₂CO₃ (5 μL). After 3 min, 0.30 mL of 0.5 M HEPES (pH 7.1–7.3)and 0.5 mg of DFO-C4 (pH = 7) were added into the reaction vial. Afterincubation for 120 min at 37 °C, the reaction progress was monitoredby iTLC using a 20 mM citric acid (pH 4.9–5.1) mobile phase.The decay corrected radiochemical yield was consistently >98.5%.

Iodination with iodine-125 was done in precoated iodination tubes(Pierce). 100 μg of C4 was dispersed in 100 μL of PBSsolution and added to the precoated iodination tubes. In a separate1.5 mL Eppendorf tube, a solution of 1 μL of HCl (0.2 M), 2.5μL of phosphate buffer (0.5 M, pH = 8), and 10 μL of potassiumiodide solution (1 mg/mL) was prepared. 1 mCi of iodine-125 was addedinto the Eppendorf and then added to the iodination tubes. After 15min of reaction the solution was halted and purified via PD10 columnusing PBS as the mobile phase. The purity was assessed via iTLC andwas consistently >98% pure.

Determination of ChelateNumber

The number of DFO attachedto the Ab was measured with a radiometric isotopic dilution assays.From a stock solution, aliquots of ⁸⁹Zr oxalate (10 μL,20 μCi, pH 7.7–7.9) were added to 12 solutions containing1:2 serial dilutions of nonradioactive ZrCl₄(aq) (100 μLfractions; 1000–0.5 pmol, pH 7.7). The mixture was vortexedfor 30 s before adding 5 μL aliquots of C4 (5 mg/mL, 2.5 μgof mAb, sterile PBS). The reactions were incubated at room temperaturefor >2 h before quenching with DTPA (20 μL, 50 mM, pH 7.0).Control experiments confirmed that ⁸⁹Zr complexation toDFO-conjugated proteins was complete in <2 h. The extent of complexationwas assessed by iTLC and counting the activity at the baseline andsolvent front. The fraction of ⁸⁹Zr-radiolabeled protein(Ab) was plotted versus the amount of nonradioactive ZrCl₄ added. The number of chelates was calculated by measuring the concentrationof ZrCl₄ at which only 50% of the protein was labeled,multiplying by a factor of 2, and then dividing by the moles of proteinpresent in the reaction. Isotopic dilution assays revealed an averageof 1.08 ± 0.3 accessible chelates per protein molecule for C4.

Immunoreactive Fraction

Five different dilutions ofcells ranging from 0.5 to 10 million cells/mL were prepared in PBSsolution containing BSA. 10 μCi of ⁸⁹Zr-C4 was incubatedfor 1 h at room temperature in each cellular fraction. Following theincubation, the cell suspensions were washed twice with ice-cold PBSand centrifuged. The radioactivity associated with cell pellet (cell-boundactivity) was determined by counting the tubes in a γ counterwith corresponding standards and blanks (for nonspecific binding).The inverse of cell-bound radioactivity (total over bound) was plottedas a function of the inverse of cell concentration. The data werefitted to a linear fit according to a least-squares linear regressionmethods. The Y-intercept of the regression line representsthe inverse of immunoreactive fraction.

Serum Stability Determination

In vitro stability ofDFO-mAb was assessed in serum for up to 3 days. For the serum stabilitystudies, 50 μCi ⁸⁹Zr- DFO-mAb was added to 500 μLof 100% fetal bovine serum and the mixture was incubated at 37 °C.Aliquots were reserved at different time points: 1 h, 4 h, 24 h, 75h, and 5 days and assessed by iTLC. Activity within the serum sampleswas resolved by iTLC, and each experiment was carried out in triplicate.

Saturation Binding Assay

The number of PD-L1 receptorsper cells was determined by a saturation binding experiment performedin H1975, A549, and PC3 cells using ¹²⁵I-C4. The amountof radioligand added was increased while maintaining a constant numberof cells/receptors. The three saturation binding assays were donewith a range of nine different concentrations between 0.65 and 120nM of the radioligand. The nonspecific binding was determined at threedifferent concentrations (0.65, 10, and 120 nM) by co-incubation ofseparate treatment arms with a 1000-fold excess over the K_D of unlabeled C4. Cells were incubated at 4 °C for1 h to achieve equilibrium. Following incubation, the cells were washedtwice with ice cold PBS and retained for analysis (unbound fraction).The cells were lysed with 1 mL of 1 M NaOH and collected (bound fraction).The unbound and cell associated fractions were counted in a γcounter and expressed as a percentage of the total activity addedper number of cells. Experiments were performed in triplicate. Thespecific binding was obtained by subtracting the nonspecific bindingfrom total. The specific binding was plotted against the concentrationof the radioligand. A Rosenthal plot was used to determine the B_max with PRISM software.

Cellular UptakeAssays

Cells were seeded at a densityof 4 × 10⁵ cells per well in 12-well plates and grownat 37 °C, 5% CO₂ for 24 h. Cells were treated withvehicle or the indicated therapy for 24, 48, or 72 h. Doxorubucinwas used at doses of 100 nM and 1 μM. Paclitaxel was used at200 nM, 2 μM, and 20 μM. On the day of the experiment,cells were subjected to a PBS wash followed by incubation for 1 hat 37 °C, 5% CO₂ in PBS with ¹²⁵I-C4 (0.5μCi). Following incubation, the cells were washed twice withice cold PBS and retained for analysis (externalized fraction). Thecells were lysed with 1 mL of 1 M NaOH and collected (cell associatedfraction). The externalized and cell associated fractions were countedin a γ counter and expressed as a percentage of the total activityadded per equal relative number of cells. Experiments were performedin triplicate. The cell associated activity was expressed as a % oftotal activity to which the cells were exposed. This value was furthernormalized to cell number to correct for treatment induced changesin cell viability (determined by treating and counting separate wellscontaining cells).

Small Animal PET/CT

Three- to five-week-oldmale nu/nuimmunocompromised mice and immunocompetent C57BL/6 mice were purchasedfrom Charles River. Nu/Nu mice were inoculated with 1 × 10⁶ H1975 or A549 cells subcutaneously into one flank in a 1:1mixture (v/v) of PBS and Matrigel (Corning). Tumors were palpablewithin 8–14 days with H1975 and 14–21 days with A549after injection. Three- to five-week-old male C57BL/6 mice (CharlesRiver) were inoculated with 1 × 10⁶ B16F10 subcutaneouslyinto one flank in the same mixture of Matrigel and PBS. Tumors werepalpable within 3–5 days after injection. Tumor bearing mice(n = 5 per treatment arm) received between 50 and300 μCi of solution in 100 μL saline solution volume intravenouslyusing a custom mouse tail vein catheter with a 28-gauge needle anda 100–150 mm long polyethylene microtubing. ∼300 μCiwas injected for the mice for imaging and ∼50 μCi forthe mice for biodistribution. The mice were imaged on a dedicatedsmall animal PET/CT scanner (Inveon, Siemens Healthcare, Malvern,PA). Mice were imaged at multiple time points after injection outto 5 days. Animals were scanned for 20 min for PET, and the CT acquisitionwas 10 min.

The co-registration between PET and CT images wasobtained using the rigid transformation matrix from the manufacturer-providedscanner calibration procedure since the geometry between PET and CTremained constant for each of PET/CT scans using the combined PET/CTscanner. During the imaging procedure, animals were anesthetized withgas isoflurane at 2% concentration mixed with medical grade oxygen.The photon attenuation correction was performed for PET reconstructionusing the co-registered CT-based attenuation map to ensure the quantitativeaccuracy of the reconstructed PET data.

Biodistribution Studies

Biodistribution studies wereconducted to evaluate the uptake of ⁸⁹Zr-C4 in mice bearingsubcutaneous tumors. At the indicted time after radiotracer injection,animals were euthanized by CO₂(g) asphyxiation, and 14tissues (including the tumor) were removed, weighed, and counted ona γ-counter for accumulation of ⁸⁹Zr radioactivity.The mass of ⁸⁹Zr-C4 formulation injected into each animalwas measured and used to determine the total number of counts perminute by comparison to a standard syringe of known activity and mass.The data were background- and decay-corrected, and the tissue uptakewas expressed in units of percentage injected dose per gram of drytissue (% ID/g).

Immunofluorescence

PDX tumors wereharvested and snap-frozenin 2-methylbutane and stored at −80 °C. Serial sections(10 μm thickness) were collected using a cryostat (Leica CM3050).The slices were mounted on coated glass slides (Superfrost Ultra Plus,Thermo Scientific) and stored at −80 °C until assayed.Tumors sections were fixed with paraformaldehyde (4% v/v), blockedwith bovine serum albumin (5% m/v) in Tween (0.5% m/v), and then incubatedfor 1 h at temperature room in a humidified chamber with a rabbitmonoclonal primary antibody against PDL1 (E1L3N XP, rabbit mAb no.13684, 1:500) or EGFR L858R (43B2, rabbit mAb no. 3197, cell signaling,1/400). Then the sections were rinsed three times and incubated for30 min at room temperature with a secondary anti-rabbit antibody (AlexaFluor 488, 1:1000). Sections were mounted with ProLong gold antifadereagent containing 4′,6-diamidino-2-phenylindole (DAPI, Invitrogen).Slides were imaged using an AxioObserver Z1 microscope (Zeiss). Theslides were counterstained with hematoxylin and eosin to show cellularmorphology.

A mosaic acquisition with a magnification of 20×was performed covering entirely the tumors section using an AxioObserverZ1 microscope (Zeiss). The fusion of mosaic images was generated byAxioVision 4.6. A sequential acquisition of separate wavelength channelswas used to avoid fluorescence crosstalk. Exposure times were setso that pixel brightness was never saturated and kept constant duringthe entire image acquisition and between different experiments.

Animal Therapy Studies

Tumor bearing mice were injectedintraperitoneally with doxorubicin (Sigma, 2 mg/kg body weight; equivalentto a dose of 6.5 mg/m² in patients). All mice were treatedonce per day for 6 days (4 days before the injection and 2 days followingthe radiotracer injection). Paclitaxel was suspended in HPMT solution(0.5% w/v hydroxypropylmethylcellulose dissolved in water plus0.2% v/v Tween 80). Tumor bearing mice were treated once daily viaoral gavage with paclitaxel (25 mg kg^–1 d^–1) for 6 days (4 days before the injection and 2 days following theradiotracer injection).

Statistical Analysis

Data were analyzedusing the unpaired,two-tailed Student’s t-test. Differences atthe 95% confidence level (P < 0.05) were consideredto be statistically significant.