ImmunoPET of Malignant and Normal B Cells with ⁸⁹Zr- and ¹²⁴I-Labeled Obinutuzumab Antibody Fragments Reveals Differential CD20 Internalization In Vivo

The CD20 antigen is expressed on the surface of B cells during all stages of B-cell development, except early pro-B cells and plasma cells. CD20 expression on B-cell malignancies such as lymphomas and leukemias has proven useful as target for antibody-based diagnosis and therapies. Rituximab was the first anti-CD20 antibody approved by the FDA (1997) for use in oncology and most lymphoma patients now receive rituximab during their treatment as monotherapy or in combination with chemotherapy (1–3). Rituximab is given to all patients with the same standard dose and scheduling established at licensing (375 mg/m² weekly × 4; ref. 4), regardless of factors known to affect pharmacokinetics and serum levels, such as CD20 expression level and density, tumor burden and localization, number of circulating B lymphocytes and antigen modulation (5). Although CD20 has long been thought to not shed, internalize or degrade upon antibody binding (6), several groups describe antigenic modulation, for example, internalization of Ag/Ab complexes (7, 8) and removal of Ag/Ab complexes from the tumor cell by monocytes, a process termed trogocytosis (9, 10). CD20 modulation might contribute to development of resistance to anti-CD20 mAb in B-cell lymphomas (11–13).

ImmunoPET, as a noninvasive molecular imaging method, combines the target specificity of antibodies with the sensitivity of positron emission tomography (PET; refs. 14, 15). Anti-CD20 immunoPET has the potential to enable personalized image-guided therapy by providing crucial information for diagnosis, prognosis, dosing, and evaluation of therapy response in lymphoma patients and could be especially informative as companion diagnostic for subsequent antibody-based therapy. Furthermore, the noninvasive detection of B cells in vivo could enable monitoring of immunotherapies, such as reconstitution after B-cell depletion, adoptive cell transfers or B-cell targeting vaccination and increase the understanding of B-cell-mediated disease states (e.g., rheumatoid arthritis, multiple sclerosis, diabetes; refs. 16, 17).

Radiolabeled rituximab has successfully been used for PET imaging of human CD20 expressing transgenic mice (18, 19), and both B-cell lymphomas and autoimmune disease in patients (20, 21). However, the long plasma half-life of full length IgGs requires delays between tracer administration and image acquisition in order to achieve sufficient clearance of the radio-antibody and therefore high target-to-background ratios. In contrast, engineered antibody fragments such as diabodies (scFv dimer) and minibodies (scFv-C_H3) possess pharmacokinetics optimized for imaging. Removal of the constant domains (C_H2, Fc) abolishes effector functions and FcRn-mediated recycling. The reduction of the overall size results in improved tissue penetration and in accelerated blood clearance with half-lives that range from 2 to 5 hours for diabodies (renal clearance) and 5 to 12 hours for minibodies (hepatic clearance) in mice (22).

In this report, we present novel PET-imaging agents based on the humanized, type II anti-CD20 mAb obinutuzumab (GA101; ref. 23). Type II anti-CD20 antibodies bind CD20 in a different orientation from type I antibodies (rituximab) and are reported to internalize less rapidly (7, 24). Cell surface retention is preferable for imaging, allowing the use of radioiodinated tracers which show greatly reduced tissue background because iodine can diffuse from the cell upon internalization and degradation of the radioiodinated antibody. As a result we hypothesized that radioiodinated obinutuzumab-based antibody fragments should be superior to rituximab-based fragments for preclinical immunoPET imaging of lymphoma mouse models and ultimately for clinical translation.

To establish the optimal anti-CD20 immunoPET tracer, we compared different sized antibody fragments (cys-diabody, 55 kDa and cys-minibody, 80 kDa) based on obinutuzumab (GA101) and Rituximab, radiolabeled with long-lived positron emitters iodine-124 (¹²⁴I, nonresidualizing, t_1/2 4.18 days) and zirconium-89 (⁸⁹Zr, residualizing, t_1/2 3.27 days). ImmunoPET tracers were evaluated in two preclinical mouse models expressing human CD20. First, in a subcutaneous B-cell lymphoma model (38C13-huCD20) and second, in transgenic mice expressing human CD20 on endogenous mature B cells. Furthermore, we utilized the differences regarding uptake and retention of ¹²⁴I- and ⁸⁹Zr-labeled GA101 antibody fragments to assess internalization of CD20/antibody complexes in vivo by comparing quantitative PET images over time, revealing differential internalization rates of healthy and malignant B cells.

The murine B-cell lymphoma line 38C13 expressing human CD20 (38C13-huCD20) was previously described (25). Both 38C13 and 38C13-huCD20 were cultured in RPMI1640 (Life Technologies) supplemented with 50 μmol/L 2-mercaptoethanol and 10% FBS.

Protocols for all animal studies were approved by the University of California Los Angeles (UCLA) Animal Research Committee. 6- to 8-week-old female SCID mice were purchased from Jackson Laboratory. Allografts were established by injecting 0.5 × 10⁶ cells in 1:1 medium:Matrigel (BD Bioscience) subcutaneously into the shoulder region and were allowed to grow for 5 to 8 days. Human CD20 transgenic mice (hCD20TM) have been described previously (16), and were backcrossed onto BALB/c backgrounds and genotypes confirmed by PCR.

Cloning, production, purification, and radiolabeling of obinutuzumab- and rituximab-based antibody fragments are described in Supplementary Data Methods.

¹²⁴I-labeled antibody fragments were generated by direct radioiodination (Pierce Pre-Coated Iodination Tubes; Thermo Scientific). For site-specific ⁸⁹Zr-radiolabeling, the chelator deferoxamine-maleimide (mal-DFO, Macrocyclics) was conjugated to the reduced C-terminal cysteine of the antibody fragments and radiolabeling was performed as described by Vosjan and colleagues (26).

Approximately 20 to 25 μg (∼1.5–2.2 MBq) of ¹²⁴I- or ⁸⁹Zr-labeled antibody fragments were injected via tail vein into tumor bearing SCID mice or huCD20TM. Thyroid and stomach uptake of radioiodine was blocked with Lugol's iodine and potassium perchlorate, respectively, as previously described (27). At 4, 8, and 24 hours postinjection, mice were anesthetized using 2% isoflurane and imaged with 10-minute acquisitions on an Inveon microPET scanner, followed by a microCT scan (Siemens). Following microPET/CT imaging, mice were sacrificed and organs and blood were collected, bilateral lymph nodes (mesenteric, inguinal, axial, and brachial) were pooled, weighed, and gamma counted. The %ID/g was calculated based on a standard containing 1% of the injected dose.

MicroPET images were reconstructed using non-attenuation filtered back-projection (FBP) or maximum a posteriori (MAP) reconstruction. MicroPET/CT overlay images are displayed as whole body maximum intensity projections (MIP). Image analysis was performed using AMIDE (28).

Tumor uptake values (%ID/g_ROI) and volumes were quantified by drawing ellipsoid regions of interest (ROI) over the entire tumor based on the microCT. The mean voxel value of the microPET scan was converted to %ID/g using the decay-corrected injected dose and empirically determined cylinder factors for ¹²⁴I and ⁸⁹Zr, respectively. Partial volume correction was based on the spheric approximation of the tumor volume (ROI) and previously determined recovery coefficient (RC) curves for ¹²⁴I and ⁸⁹Zr (29). Spleen uptake in huCD20TM was quantified by drawing small ellipsoid ROIs (∼14 mm³) within the tip of the spleen and %ID/g_ROI was calculated without partial volume correction, due to the complexity of organ shape, volume, and proximity to organs with high activity (liver, kidneys).

Data values are reported as mean ± SD unless indicated otherwise. Ex vivo biodistribution values are depicted as box-and-whiskers (min to max) graphs. For statistical analysis paired t tests and one-way ANOVA for comparing two or three groups, respectively, were performed using GraphPad Prism version 6.00.

The cys-diabody (GAcDb, 54.5 kDa) and cys-minibody (GAcMb, 83.5 kDa) were generated based on the variable domains of obinutuzumab (GA101). Rituximab-based antibody fragments (RxcDb, RxcMb) were generated similarly for comparison. Antibody fragments contain two antigen-binding sites and a C-terminal cysteine for site-specific conjugation (Fig. 1A; Supplementary Figs. S1 and S2). Purity and correct assembly of purified proteins was confirmed by SDS-PAGE analysis and size-exclusion chromatography (SEC; Figs. 1B and C; Supplementary Fig. S1). Engineered antibody fragments retained the antigen binding characteristics of the parental antibodies. GAcDb and GAcMb reached maximal binding at lower levels than rituximab and showed apparent affinities in the low nanomolar range (K_D∼4 nmol/L; Supplementary Fig. S3).

Anti-CD20 cDb and cMb were successfully radiolabeled with ¹²⁴I or ⁸⁹Zr (results are summarized in Table 1). Radiochemical purity after size exclusion spin column was >98% for all tracers. Immunoreactivity of radiolabeled antibody fragments ranged from 65% to 85% (Supplementary Fig. S4).

In vivo targeting of the antibody fragments was evaluated in SCID mice bearing subcutaneous lymphoma tumors (38C13-huCD20). Mice received 20 to 25 μg (1.5–2.2 MBq) of labeled protein intravenously and were serially imaged at 4, 8, and 24 hours postinjection (p.i.). Figure 2 displays representative time series of PET/CT scans showing specific uptake in huCD20-positive tumors and fast clearance from blood and nonspecific tissues, resulting in high-contrast immunoPET images as early as 4 hours p.i. Background activity cleared by 24 hours p.i. with only antigen-positive tumors visible in the PET scan. Specificity of tumor uptake was confirmed in mice bearing 38C13-huCD20 tumors that were blocked by co-injection of 80 μg (4 mg/kg) cold antibody fragments and in mice bearing huCD20-negative 38C13 tumors. The comparison of ¹²⁴I-GAcDb and ¹²⁴I-RxcDb (Fig. 2A and B) showed that ¹²⁴I-RxcDb cleared more quickly from the blood as indicated by decreasing activity in the heart and lower uptake in antigen-positive tumors. These observations were confirmed by ex vivo biodistribution at 24 hours p.i. (Fig. 2B; Supplementary Table S1). Uptake of ¹²⁴I-GAcDb in 38C13-huCD20 tumors was 4.6 ± 0.6 %ID/g at 24 hours p.i. and was significantly higher than in 38C13 control tumors (0.5 ± 0.1 %ID/g, P < 0.0001). ¹²⁴I-RxcDb showed 1.8 ± 0.4 %ID/g uptake in 38C13-huCD20 tumors and 0.8 ± 0.1 %ID/g in 38C13 (P < 0.0001).

ImmunoPET with ¹²⁴I-labeled GAcMb (Fig. 2C) showed higher activity in the blood (heart), likely due to the longer plasma half-life of the 80 kDa antibody fragment, resulting in higher tumor uptake compared with ¹²⁴I-GAcDb. Uptake values as determined by ex vivo biodistribution (Fig. 2D; Supplementary Table S1) were 9.2 ± 0.8 %ID/g in the 38C13-huCD20 tumors. Uptake in blocked tumors and in CD20-negative 38C13 tumors was significantly lower (3.8 ± 0.3 %ID/g, P = 0.0004 and 1.9 ± 0.3 %ID/g, P < 0.0001, respectively). ¹²⁴I-RxcMb immunoPET (Fig. 2D) also showed tracer uptake in the positive tumors (1.2 ± 0.1 %ID/g), which could be blocked by co-injection of cold RxcMb (0.6 ± 0.2 %ID/g) and was not observed in negative tumors (0.45 ± 0.2 %IF/g). 38C13-huCD20 uptake values of ¹²⁴I-RxcMb were considerably lower compared with ¹²⁴I-GAcMb.

In summary, ¹²⁴I-GAcDb and ¹²⁴I-GAcMb outperformed the respective rituximab-based tracers (¹²⁴I-RxcDb and ¹²⁴I-RxcMb) reaching higher tumor uptake and higher contrast microPET images (higher target-to-blood ratio). Furthermore, although ¹²⁴I-GAcMb reached the highest tumor uptake, the faster clearing ¹²⁴I-GAcDb reached the highest tumor-to-blood ratio at 24 hours p.i. (12.4 ± 2.5; Supplementary Table S1), resulting in better image contrast.

To investigate if a residualizing radiometal (⁸⁹Zr) may be advantageous in terms of greater tumor uptake and retention, mal-DFO was site-specifically conjugated to GAcDb and GAcMb for ⁸⁹Zr immunoPET. Both ⁸⁹Zr-GAcDb and ⁸⁹Zr-GAcMb were used for serial immunoPET imaging of the 38C13-huCD20 lymphoma model and high-contrast images showing specific tumor uptake were obtained (Fig. 3). The residualizing nature of the ⁸⁹Zr radio-metabolites caused higher nonspecific uptake, especially in organs of clearance (kidneys, liver, and spleen; Fig. 3A and B and Supplementary Table S1). Compared with the respective iodinated cys-diabody, ⁸⁹Zr-GAcDb reached only slightly higher tumor uptake (4.9 ± 0.3 %ID/g by ex vivo biodistribution), resulting in comparable tumor-to-blood ratios whereas the tumor-to-background (muscle) ratio was lower. ⁸⁹Zr-GAcDb immunoPET images confirmed primarily renal clearance of the 55 kDa protein.

⁸⁹Zr-GAcMb immunoPET showed the highest uptake of all tested tracers in 38C13-huCD20 allogeneic grafts (21.8 ± 1.2 %ID/g) and primarily hepatic clearance. Blood activities were comparable to ¹²⁴I-GAcMb, resulting in a higher tumor-to-blood ratio for ⁸⁹Zr-GAcMb (11.0 ± 1.0) compared with ¹²⁴I-GAcMb (4.0 ± 0.5; Supplementary Table S1). However, the tumor-to-muscle ratio was lower (41.1 ± 2.8 and 145.4 ±19.3, respectively) caused by higher background retention of the residualizing ⁸⁹Zr radio-metabolites.

To evaluate the in vivo targeting ability of the obinutuzumab-based PET tracers in the context of normal tissue expression of CD20, a transgenic mouse model expressing human CD20 (huCD20TM) was used (16). PET/CT images of mice injected with ¹²⁴I- and ⁸⁹Zr-labeled GAcDb and GAcMb were obtained 4, 8, and 20 hours p.i. All tracers showed rapid and specific localization to the spleen and lymph nodes, where B cells reside (Fig. 4A and B; Supplementary Fig. S5). Specificity was confirmed in antigen-negative BALB/c mice (Supplementary Table S2).

The activity in the spleen decreased considerably more quickly with ¹²⁴I-labeled cDb and cMb compared with ⁸⁹Zr-labeled cDb and cMb. The mean spleen uptake values (%ID/g ± SEM) determined by ex vivo biodistribution at 20 hours p.i. were 5.0 ± 0.4 for ¹²⁴I-GAcDb compared to 17.8 ± 2.4 for ⁸⁹Zr-GAcDb (P = 0.006) and 11.5 ± 1.2 for ¹²⁴I-GAcMb compared to 38.9 ± 5.7 for ⁸⁹Zr-GAcMb (P = 0.003; Fig. 4C and D and Supplementary Table S2). Because of the partial volume effect (PVE), small structures such as lymph nodes were difficult to visualize, especially with the radioiodinated antibody fragments, although ex vivo biodistribution confirmed tracer uptake in dissected lymph nodes (inguinal, axillary, mesenteric) to be 3.0 ± 1.0 %ID/g for ¹²⁴I-GAcDb and 7.1 ± 0.3 %ID/g for ¹²⁴I-GAcMb. Here, the use of residualizing ⁸⁹Zr-labeled anti-CD20 antibody fragments was advantageous and resulted in clearly distinguishable lymph nodes (Supplementary Fig. S5, microPET 2 mm section, rescaled).

Pairwise evaluation of both ¹²⁴I- and ⁸⁹Zr-labeled antibody fragments over time provides an assessment of in vivo binding and internalization of the antigen/antibody complexes (30). The ¹²⁴I-labeled antibody fragments undergo intracellular catabolism and dehalogenation upon internalization and ¹²⁴I-iodotyrosine diffuses rapidly from the cell. ⁸⁹Zr-antibody fragments show identical target cell binding but noticeable differences in retention because the residualizing ⁸⁹Zr radio-metabolites are trapped intracellularly (31).

ImmunoPET studies showed that both ¹²⁴I- and ⁸⁹Zr-anti-CD20 tracers result in high-contrast imaging, but differences were observed in uptake and retention of activity in the spleen of huCD20TM compared with the subcutaneous B-cell lymphoma model, suggesting differences in CD20 internalization or modulation between healthy B cells and B-cell lymphomas. Therefore, ROI quantification of serial microPET images and comparison of the change in radioactive signal was performed for ¹²⁴I- and ⁸⁹Zr-labeled GA101 antibody fragments in both the subcutaneously 38C13-huCD20 tumors and in huCD20TM. Peak uptake of the GAcMb appeared around 4 to 8 hours postinjection, whereas the smaller GAcDb reached peak uptake before the 4-hour time point (Fig. 5A). For both antibody fragments, the ⁸⁹Zr-labeled and ¹²⁴I-labeled versions showed similar uptake and retention in subcutaneously tumors, indicating no or very slow internalization of the radiotracers within 24 hours p.i.

Spleen uptake values in huCD20TM from ¹²⁴I-labeled GAcDb and GAcMb decreased rapidly from the spleen, at a rate faster than that in the subcutaneous lymphoma model. In contrast, activity from ⁸⁹Zr-labeled GAcDb and GAcMb was retained in the spleen, resulting in significantly different ROI values for the respective GAcDb (P = 0.0007) and GAcMb (P = 0.0003) tracers 24 hours p.i. (Fig. 5B). This observation indicates that endogenous B cells in the spleen internalize CD20/antibody complexes more rapidly than the B-cell lymphoma 38C13.

In this study, we evaluated two obinutuzumab (GA101)-based antibody fragments (cys-diabody, GAcDb and cys-minibody, GAcMb) for immunoPET of CD20 expressing B cells using two alternative radiolabels (⁸⁹Zr and ¹²⁴I). To ensure a biologically inert imaging agent, the reformatted antibody fragments lack a full Fc region reducing the interaction with complement or FcRs (32–34). Both the cDb and the cMb retained binding characteristics (type II epitope and apparent affinity K_D ∼4 nmol/L) of the parental mAb (23). Because the antibody fragments are based on a humanized mAb, translation into the clinic is facilitated. Clinical trials are of particular importance to establish pharmacokinetics and targeting in patients since clinical performance is difficult to predict based on preclinical studies in rodents.

Independent of the radiolabeling method (random iodination of tyrosine residues or site-specific conjugation of maleimide-DFO ⁸⁹Zr to the reduced cys-tag), GA101-based fragments showed specific targeting to human CD20 expressing B cells as demonstrated in a subcutaneous B-cell lymphoma model (38C13-huCD20) and in a transgenic mouse model expressing human CD20 (huCD20TM). Rapid clearance of the antibody fragments enabled high-contrast small animal PET imaging at early time points.

In the lymphoma model, the ¹²⁴I-labeled GA101 cDb and cMb outperformed the respective rituximab-based antibody fragments, reaching higher tumor activity (4.6%ID/g vs. 1.8%ID/g for cDbs and 9.2%ID/g vs. 1.2 %ID/g for cMbs at 24 hours p.i., respectively) and higher tumor-to-blood ratios. This finding was unexpected considering that type II mAbs (obinutuzumab, GA101) show binding at approximately half the surface density of type I binders (rituximab;ref. 24).

The low specific uptake of the rituximab-based fragments is in accordance with previously published anti-CD20 minibody, scFv-Fc and diabody immunoPET imaging of the same lymphoma model (35, 36). Furthermore, rituximab is described to have a faster off-rate than many other anti-CD20 antibodies, which might influence in vivo tumor retention (37, 38).

Compared with the ¹²⁴I-labeled GAcDb and cMb, the ⁸⁹Zr-labeled antibody fragments achieved slightly higher tumor activity and retention due to the residualizing radiometal, resulting in higher tumor-to-blood ratios at 24 hours p.i.; at the same time, the tumor-to-background (muscle) ratio is lower caused by increased background activity. As expected, nonspecific uptake is most pronounced in organs of clearance (kidneys, liver, spleen), which in a clinical setting might impede detection of tumors or metastases in those organs. However, by selecting the antibody fragment according to size (diabody or minibody) the elimination pathway can be predetermined. The high uptake of ⁸⁹Zr-labeled GAcDb and GAcMb in the spleen of SCID mice is most likely caused by the abnormal structure and significantly smaller size (20–35 mg) and was not observed in BALB/c mice (39).

ImmunoPET of human CD20 transgenic mice (huCD20TM) showed specific tracer uptake in the spleen for both ¹²⁴I- and ⁸⁹Zr-labeled GAcDb and GAcMb. Lymph nodes were difficult to visualize with ¹²⁴I-labeled antibody fragments, although ex vivo biodistribution confirmed tracer uptake. Imaging and quantification of small objects such as lymph nodes (approx. 1–3 mm³) is challenging due to their size being near or below the spatial resolution of the Inveon microPET scanner (1.5–1.8 mm) and the increased PVE due to a longer mean positron range of ¹²⁴I compared with ⁸⁹Zr (3.0 and 1.1 mm, respectively; ref. 29). Another challenge for ¹²⁴I PET imaging is the simultaneous emission of a 603-keV gamma-ray with about 50% of all positrons (β⁺) that falls within the energy window of most PET scanners and causes increased background (40). ⁸⁹Zr, on the other hand, emits a prompt 909-keV gamma-ray with 100% of positrons but the disparity to the 511 keV photons prevents interference with the detection of coincident photons (41). Therefore, immunoPET using ⁸⁹Zr-GAcDb and cMb visualized even small lymph nodes with high contrast.

In summary, for B-cell lymphomas with cell surface retention of Ag/Ab complexes ¹²⁴I-labeled obinutuzumab antibody fragments outperform rituximab-based fragments and the use of ¹²⁴I is preferable to ⁸⁹Zr as it results in higher image contrast due to lower background throughout the organism. However, for immunoPET imaging of small structures like lymph nodes or metastases and of internalizing antigens, the ⁸⁹Zr-labeled antibody fragments are more suitable.

Recent studies in a similar transgenic mouse model using ⁸⁹Zr-labeled intact rituximab showed high uptake in the spleen (103 ± 12%ID/g at 24 hours p.i. from ex vivo biodistribution) but no uptake in lymph nodes (19). Possible explanations include the low injected protein dose (2–3 μg) resulting in rapid accumulation of the tracer in the spleen. The abundance of CD20 antigen throughout the body and high concentration of antigen in the spleen (antigen sink) makes imaging outside the sink challenging. In contrast to the antibody fragments presented in our study, the radiolabeled intact rituximab is biologically active and would deplete B cells if injected at higher doses. The ability of the obinutuzumab-based tracers to image small structures (lymph nodes) outside the antigen sink at low doses (15–20 μg protein/dose) will be crucial for imaging in the context of normal tissue distribution of CD20.

Muylle and colleagues investigated the impact of preloading with unlabeled rituximab on tumor targeting using ⁸⁹Zr-rituximab immunoPET in lymphoma patients and found impaired tumor targeting in the majority of patients (21). This study showed that tumor targeting is influenced by a variety of factors, for example, B-cell depletion, spleen volume, tumor heterogeneity, the site of tumor involvement, and preload of unlabeled antibody. The authors emphasize the potential of immunoPET as an inherently quantitative imaging method, for dosimetry prior to radioimmunotherapy (RIT), better selection of patients for targeted therapy, and prediction of treatment outcome.

Dosimetry studies with ⁸⁹Zr-labeled full-length antibodies in patients have suggested that nonspecific accumulation in the liver will be dose limiting, whereas for ¹²⁴I-labeled tracers bone marrow exposure from blood activity is usually the limiting factor (42–44). A dosimetry study based on biodistribution data of ⁸⁹Zr-rituximab in huCD20TM suggests liver (with predose) and spleen (without predose) as dose-limiting organs in humans (45). In both cases, the use of a smaller, fast-clearing antibody fragment like the cys-diabody and the cys-minibody could result in a reduction of the radiation exposure and absorbed dose to the patient.

In addition to confirming presence of target antigen in vivo, immunoPET has also been used to evaluate antigen modulation by pairwise comparison of residualizing and non-residualizing radiolabels conjugated to the same antibody. This was demonstrated for the slow internalization of prostate stem cell antigen (PSCA) bound by anti-PSCA minibody (¹²⁴I-A11 and ⁸⁹Zr-A11; ref. 29) and in a study by Cheal and colleagues using ¹²⁴I-cG250 and ⁸⁹Zr-cG250 to study in vivo receptor binding and internalization of carbonic anhydrase IX (30). In this study, serial ROI analysis of ¹²⁴I-GAcDb and ¹²⁴I-GAcMb immunoPET in huCD20TM showed that radioactivity decreased significantly faster from the spleen compared with the retention of activity using ⁸⁹Zr-GAcDb and ⁸⁹Zr-GAcMb. In contrast, the radioiodinated and radiometal labeled tracers demonstrated very similar retention in the subcutaneous lymphoma tumors. This suggests that CD20/GA101 complexes are indeed internalized and metabolized rapidly in huCD20 Tg B cells, compared to slower internalization in 38C13-huCD20 lymphoma cells. The results of our study are consistent with previous in vitro observations in the literature that tissue cultured B-cell lines and xenograft models show limited internalization whereas primary human B cells from healthy donors show more rapid internalization matching that seen in huCD20 Tg B cells (7, 8, 46). The differences in CD20 internalization seen in these two mouse models do not warrant general conclusions about CD20 modulation in normal and malignant cell line; however, this proof-of-principle study demonstrates the applicability of immunoPET to study the wide heterogeneity in CD20 modulation rates reported for primary lymphomas. Even though xenografts implanted into immunodeficient mice have major limitations and do not reflect the complexity of human disease, they do provide reproducible models and are widely used in the development of new antitumor therapies. ImmunoPET and the ability to assess internalization in vivo could therefore contribute to better understanding of those xenografts models and facilitate preclinical research and development of new therapies (47).

Although the literature regarding CD20 modulation is conflicting, it seems obvious that a substantial reduction of CD20 levels on target B cells and consumption of the therapeutic antibody would impair immunotherapy and might cause resistance. Beers and colleagues reported rapid CD20 internalization of type I mAb (rituximab) in human CD20Tg B cells (7), and little modulation of CD20 from the cell surface by type II anti-CD20 mAb (tositumomab). This process appears to be an intrinsic B cell phenomenon that does not require activatory Fc gamma R engagement by effector cells as shown in γ-chain −/− (activatory Fc gamma R-negative) mice. Internalization of CD20/type I mAb complexes occurred equally rapidly (within just 2 hours) in normal human peripheral blood B cells. The same group showed in a recent study that internalization of mAbs is antibody specificity and FcγRIIb dependent (46). Although most bound mAbs bind FcγRIIb, internalization of Ag/Ab/FcR complexes occurs only for a select subset (anti-CD20 type I). However, work by Taylor and colleagues (48) demonstrated that an endocytic process, termed trogocytosis, mediated by FcγR-expressing effector cells (monocytes, macrophages, or NK cells) might also play a role in promoting CD20/mAb loss in vivo. This study showed only modest internalization of rituximab from 38C13-huCD20 cells but substantial removal of CD20 and bound type I mAbs (rituximab and ofatumumab) upon incubation with THP-1 acceptor cells. Consistent with both studies, we found no tracer internalization in subcutaneous 38C13-huCD20 lymphomas but distinct internalization in the spleen (normal B cells) of transgenic mice. Although the rational for engineering antibody fragments based on a type II anti-CD20 mAb and without the Fc region was to ensure cell surface retention, our findings suggest that a fraction of CD20 internalization is independent of the epitope binding mode and FcγRIIb binding.

Future studies should include immunoPET of more physiological relevant models, for example, syngeneic models of disseminated B-cell lymphoma in immunocompetent mice and imaging of tumors in the context of normal tissue expression in transgenic mice.

In conclusion, we presented novel anti-CD20 antibody fragments that specifically target human CD20 expressed on healthy and malignant B cells and produce high contrast immunoPET images. The combination of antibody fragments (size, plasma half-life, route of clearance) with radionuclides (residualizing vs. non-residualizing) offers optimized tracers for a variety of applications in B cell imaging. The ability to monitor CD20 internalization in vivo using noninvasive imaging could have an important impact on understanding B cell biology and support treatment design, especially when immunoPET tracers accompany subsequent antibody based therapies.

J.M. Timmerman reports receiving commercial research grants from Bristol-Myers Squibb, Kite Pharma, and Valor Biotherapeutics, and is a consultant/advisory board member for Celgene and Seattle Genetics. A.M. Wu holds ownership interest (including patents) in and is a consultant/advisory board member for ImaginAb, Inc. No potential conflicts of interest were disclosed by the other authors.

Conception and design: K.A. Zettlitz, K.K. Steward, A.M. Wu

Development of methodology: K.A. Zettlitz, J.M. Timmerman, A.M. Wu

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K.A. Zettlitz, R. Tavaré, S.M. Knowles, J.M. Timmerman

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K.A. Zettlitz, R. Tavaré, S.M. Knowles, J.M. Timmerman, A.M. Wu

Writing, review, and/or revision of the manuscript: K.A. Zettlitz, R. Tavaré, K.K. Steward, J.M. Timmerman, A.M. Wu

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R. Tavaré

Study supervision: A.M. Wu

The authors thank the Preclinical Imaging Technology Center, UCLA Crump Institute for Molecular Imaging for help with small-animal PET scans. Dr. M.J. Glennie kindly provided sequence information for the GA101 variable domains. We thank Felix B. Salazar and Reiko E. Yamada for technical assistance.

This work was supported by NIH grant CA149254, and by a generous gift from Ralph and Marjorie Crump to the Crump Institute. Small-animal imaging and flow cytometry were funded in part by the UCLA Jonsson Comprehensive Cancer Center Support Grant (NIH CA016042).

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.