Mouse Strains Influence Clearance and Efficacy of Antibody and Antibody–Drug Conjugate Via Fc–FcγR Interaction

This article is featured in Highlights of This Issue, p. 731

Monoclonal antibodies have become a major class of therapeutics for cancer patients (1, 2). These include antibodies that modulate oncogenic signaling pathways, kill tumor cells through effector cell functions, activate antitumor immune responses, engage T cells through bispecific targeting, and deliver cytotoxic payloads in the format of antibody–drug conjugates (ADC; refs. 3, 4). There is continued interest to develop novel antibody-based therapeutics, as demonstrated by more than 1,000 active clinical studies evaluating antibodies for cancer treatment (www.clinicaltrials.gov).

Before entering clinical trials, the efficacy and toxicity of antibody drugs are studied in various in vitro and in vivo models. Most oncology drugs are tested using xenografts, which are implanted in one of the following immuno-deficient mouse strains: athymic nude, severe combined immunodeficiency (SCID), or NOD.Cg-Prkdc^scidIl2rg^tm1Wjl/SzJ (NSG; ref. 5). Compared with wild-type mice, these mice lack one or more types of immune cells, allowing the engraftment of foreign tumor cells. Among them, NSG mice are considered the most immune-compromised, because they lack mature T cells, B cells, and natural killer (NK) cells. They also have defective macrophages and dendritic cells due to impaired IL2R signaling. Therefore, NSG mice offer an excellent host for a variety of cancers, especially those of hematologic origin (6).

Despite the successful development of therapeutic antibodies, the mechanisms governing the pharmacokinetics (PK) and pharmacodynamics (PD) of them have not been well elucidated. One of the challenges is that the PK data are often generated in wild type mice or other species of animals (e.g., rat), whereas the PD studies are performed in various immune-deficient mouse strains. Therefore, a direct correlation between the PK and PD would require bridging across different strains of mice or species. Here we examine these potential differences by systematically analyzing the PK properties of antibodies and ADCs across wild-type mice and three immune-deficient mouse strains, to improve the translation between different models.

All animal studies were approved by the Seattle Genetics Institutional Animal Care and Use Committee. Balb/c, athymic nude mice, and SCID mice were purchased from Envigo. NOD.Cg-Prkdc^scidIl2rg^tm1Wjl/SzJ (NSG) mice were obtained from the Jackson Laboratory (Bar Harbor, ME). NOD/SCID/gamma^null (NOG) mice were acquired from Taconic Biosciences. Six- to 8-week-old female mice were used for these studies. For xenograft studies, 5 million cells were implanted subcutaneously and tumor growth was monitored using a digital caliper. Tumor volumes were calculated as 1/2 × length × width².

A bolus dose of antibody or ADCs was administered via intraperitoneal or intravenous injection for PK studies, whereas the hIVIg was injected intravenously. A total of six animals were injected for each arm of the PK studies. Blood samples were collected using orbital-bleeding techniques, where three of the six animals were collected with a sparse sampling schedule. Terminal cardiac puncture was used to collect blood samples from tumor-bearing animals at the end of study. Plasma was obtained by centrifuging blood samples in Eppendorf tubes containing K2EDTA (BD Bioscience).

Karpas 299 cells were obtained from Dr. Abraham Karpas of the University of Cambridge and were authenticated by IDEXX Laboratories (Columbia, MO). Human intravenous immune globulin injection (hIVIg) was purchased from Grifols Therapeutics. Recombinant hIgG1, αCD30 (clone cAC10) (7), αCD70(clone h1F6) (8), and variants of these antibodies were all produced at Seattle Genetics, Inc. Antibody sequences are included in Supplementary Information.

All antibody–drug conjugates used in this study, including αCD30-vcMMAE, αCD30G1V1-vcMMAE, hIgG1-vcMMAE, hIgG1V1-vcMMAE, and αCD70-mcMMAF were conjugated to average four drugs per antibody at Seattle Genetics, Inc. The conjugation methods were described elsewhere (9, 10).

Radioactive antibodies or ADCs were used for most pharmacokinetic studies. The solutions of antibody or ADC containing 500 mmol/L potassium phosphate (pH 8.0) and 500 mmol/L sodium chloride was added to 55 μCi N-succinimidyl propionate, [propionate-2,3-3H]-(Moravek Biochemicals, 80 Ci/mmol, 1 mCi/mL, 9:1 hexane:ethyl acetate solution) per mg of antibody or ADC. The resulting mixture was incubated at room temperature for 2 hours, before it was purified by centrifugal filtration (Amicon Ultra-15 30 kDa MWCO centrifugal filter units; Millipore). The resulting products were further filtered through sterile 0.22 μm Ultrafree-MC Centrifugal Filter Units (Millipore) and the final antibody or ADC concentration was measured. The specific activity (Ci/mg) of each product was determined by liquid scintillation counting.

For radiolabeled materials, blood was collected into K2EDTA tubes at various time points. Plasma was isolated by centrifugation for 10 minutes at 10,000 × g. A 10 to 20 μL sample of plasma from each time point was added to 4 mL of Ecoscint-A liquid scintillation cocktail (National Diagnostics), and the total radioactivity was measured by liquid scintillation counting. The resulting disintegrations per minute values were converted to μCi, and the specific activity of the radiolabeled test articles was used to calculate the concentration of antibody or ADC remaining in the plasma at each time point.

Microtiter plates were coated overnight with a murine IgG1 monoclonal antibody with specificity to human light chain kappa (Jackson ImmunoResearch), or an anti-idiotype monoclonal antibody that binds to αCD30 antibody (Seattle Genetics, Inc.). Samples were diluted into the dynamic range of the assay with naive mouse plasma. Standards, controls, and samples were incubated on the coated and blocked plates. Bound ADCs were detected with horseradish peroxidase-conjugated goat polyclonal antibody reagent with specificity to the Fc region of human IgG (Jackson ImmunoResearch). After incubation and subsequent washing, 3,3′,5,5′ tetramethylbenzidine (ThermoFisher Scientific) was applied to the wells, and the reaction was stopped using 1N hydrogen chloride. The absorbance was measured at OD450 nm using a SPECTRAmax M2 microplate reader (Molecular Devices). Concentrations were then calculated using standards.

The PK parameters were determined using Phoenix WinNonlin (version 7.0, Certara USA, Inc.). The following PK parameters were determined: area under the plasma concentration–time curve to the last time point (AUC_last), maximum observed plasma concentration (C_max), time of maximum observed plasma concentration (T_max), terminal half-life, clearance (Cl), and calculated volume of distribution at steady-state (V_ss). AUC was calculated using the linear log trapezoidal method. Half-lives were reported for plasma concentration–time profiles that had sufficient plasma concentrations in the terminal elimination phase (at least three samples not including T_max) and an adjusted R² of ≥0.8.

IHC analysis of CD30 on formalin fixed, paraffin embedded tumor sections was performed according to the protocol described previously (11).

Mouse liver was dissociated by using Liver Dissociation Kit (Miltenyi) per manufacturer's instructions. Single-cell suspensions from spleen and bone marrow were collected and red blood cells were removed using ACK lysis buffer. Liver tissue were dissociated using GentalMacs Octo dissociator (MACS Miltenyi). Cells were stained with live/dead fixable dye (Invitrogen) to exclude dead cells from analysis. Cell surface antigens were stained with following fluorescence-labeled antibodies: CD45 (30-F11), CD11b (M1/70), Gr1 (RB6-8C5), and F4/80 (BM8). Anti-auristatin antibody was prepared at Seattle Genetics, Inc. Cells were acquired using a Becton Dickinson LSR II flow cytometry and flow cytometric data were analyzed with FlowJo software (TreeStar Software).

All analysis performed using unpaired Student t test in GraphPad Prism 7.04 software. A P value less than 0.05 was considered statistically significant.

To evaluate whether the mouse strains of xenograft models have any impact on the pharmacokinetics of antibody or ADC, we first measured the plasma concentration of radio-labeled anti-CD30 antibody, αCD30, in four strains of mice: Balb/c, athymic nude, SCID, and NSG. αCD30 is a chimeric IgG1 antibody that recognizes human leukocyte activation marker CD30 (7). Plasma concentrations at 1 hour, 4 hours, 1 day, 3 days, 7 days, and 10 days after dosing were determined using liquid scintillation counting. These measurements revealed that the plasma clearance profiles of αCD30 were comparable in the Balb/c, nude, and SCID mice (Fig. 1A). However, the antibody cleared much faster in the NSG mice, where the clearance rate reached 43.3 mL/day/kg (Fig. 1A; Table 1). Although the half-life of αCD30 was between 7.1 and 16 days in the Balb/c, nude, or SCID mice, it was only 1.4 days in the NSG mice (Table 1).

This observation prompted us to explore whether ADCs have similarly fast clearance kinetics in the NSG mice. A CD30-targeting ADC, αCD30-vcMMAE, was injected into the four strains of mice intravenously at 3 mg/kg (9, 12). In line with the observations of antibody, the ADC also cleared four times faster in the NSG mice than the other strains of mice (Fig. 1B). On average, plasma half-life of the ADC was only 1.4 days in the NSG mice, in contrast to the 6.8 days half-life in the Balb/c mice (Table 1).

To rule out whether the abnormal clearance was due to either the unique properties of chimeric αCD30 antibody or the specific biology of the NSG mice, the pharmacokinetic profiles of a humanized IgG1 antibody (αCD70) and an auristatin conjugate (αCD70-mcMMAF) were also assessed in the NOG mice, a highly immune-deficient mouse strain from a different vendor (8, 13, 14). Consistently, both αCD70 and αCD70-mcMMAF cleared rapidly in the NOG mice (Supplementary Fig. S1A).

We then explored whether dose level has impact on the exposure of the antibody in NSG mice. αCD30 was injected at 0.1, 1, 3, 10, or 30 mg/kg, and the plasma concentration measurement is shown in Fig. 1C. The clearance rate decreased as the dose increased, suggesting a nonlinear drug disposition pattern, which saturated between 3 and 10 mg/kg.

To address whether the reduced exposure of antibodies and ADCs in the NSG mice interferes with the efficacy assessment of antibody-based drugs, the αCD30-vcMMAE ADC was studied for its efficacy in the CD30⁺ Karpas 299 tumors subcutaneously implanted in either SCID or NSG mice (9). The tumor growth kinetics were similar between the two strains, taking approximately 17 days to reach 1,000 mm³ (Fig. 2A and B). IHC analysis confirmed that Karpas 299 cells retain comparable CD30 expression whether they are in NSG or SCID mice (Fig. 2A and B). We then treated the tumor-bearing mice with αCD30-vcMMAE ADC. In the SCID animals, a single dose of 0.3 mg/kg ADC led to growth delay; and 1 mg/kg ADC resulted in complete remission (Fig. 2C). In contrast, neither of these dosages slowed the growth of tumors implanted in the NSG mice. Moreover, 3 mg/kg ADC could only delay the tumor growth for 10 days (Fig. 2D). These data demonstrate that NSG mice did influence the efficacies of ADCs in vivo, presumably due to the reduced exposure.

Because neither αCD30 nor αCD70 antibodies binds murine homologs of CD30 or CD70, we reasoned that the rapid clearance in the NSG mice was mediated through the Fc fragment, but not the Fab domain. Therefore, we studied the pharmacokinetics of CD70 antibodies with different Fc interactions: αCD70_G1V1(E233P:L234V:L235A) lacks the binding to murine FcγRI and FcγRIV; αCD70_S239C has reduced binding to FcγRIV, but not FcγRI; and an afucosylated αCD70 has increased affinity to FcγRIV (Supplementary Table S1; refs. 15–17). Surprisingly, αCD70_G1V1 demonstrated a typical antibody pharmacokinetics profile in the NSG mice. αCD70_S239C improved the exposure comparing to the wild-type antibody. In contrast, afucosylated SEA-αCD70 did not reduce the exposure (Fig. 3A). We also compared the PK profile of a recombinant, nontargeting hIgG1-vcMMAE and hIgG1V1-vcMMAE in the NSG mice. Similar to the findings of αCD70 antibodies, ablating FcγRI and FcγRIV interaction with G1V1 mutations also rescued the exposure (Supplementary Fig. S1B). These data suggest that both FcγRI and FcγRIV played a role in the rapid antibody clearance in NSG mice.

We then explored whether excess immunoglobulin supplementation can saturate the abnormal FcγR interaction and, subsequently, increase the exposure the NSG mice. Because 1 g/kg of hIVIg was reported to increase IgG antibody clearance in mice, we selected low dose levels that are unlikely to disturb the FcRn interaction (18). When hIVIg (0, 1, 3, or 10 mg/kg, i.v.) was administered to the NSG mice 24 hours before the αCD30-vcMMAE injection (1 mg/kg, i.p.), the ADC exposure increased following the pretreatment with hIVIg in a dose-dependent manner. Moreover, pretreatment with 10 mg/kg hIVIg was sufficient to restore the half-life of ADC to 7 days in the NSG mice (Fig. 3B).

We then evaluated whether the enhanced exposure by ablating FcγR interaction or saturating with hIVIg could restore the efficacy of αCD30-vcMMAE in the Karpas 299 xenografts in the NSG mice. Remarkably, 3 mg/kg of αCD30_G1V1-vcMMAE led to complete remissions in this model. Similarly, pretreatment with hIVIg also resulted in complete remissions (Fig. 3C). Further plasma concentration measurement confirmed αCD30_G1V1-vcMMAE and hIVIg pretreatment have doubled the exposure of αCD30-vcMMAE in the NSG mice (Supplementary Fig. S1C).

To explore what type of cells may contribute to the abnormal Fc interaction in NSG mice, we focused on the myeloid cell compartment because the NSG mice lack T cells, B cells, and NK cells (6). Flow cytometry analysis revealed a high percentage of bone marrow granulocytic myeloid cells (CD11b⁺Gr1^high) in the immune-deficient SCID and NSG mice, compared with the wild-type Balb/c or nude mice (Fig. 4A). Strikingly, the number of CD11b⁺F4/80⁺ myeloid cells in the bone marrow was 20 times more in the NSG mice when compared with the other strains of mice (Fig. 4B). The high frequency of CD11b⁺F4/80⁺ myeloid cells was also observed in the spleen and liver of NSG mice (Fig. 4C and D). The high abundance of CD11b⁺F4/80⁺ myeloid cells suggest that these cells may mediate antibody clearance in the NSG mice. Indeed, a subpopulation of the CD11b⁺F4/80⁺ cells was found to bind MMAE-conjugated antibody by flow cytometry (Fig. 4E). These data suggest that the myeloid cells in these organs may contribute to the antibody distribution and clearance in the NSG mice.

Preclinical PK/PD modeling is widely used in industry to enable the selection and optimization of first-in human doses and regimens, including predicting human efficacious doses (19). However, translating the biological effects between species remain challenging (20). Here we report a unique, rapid clearance phenomenon of therapeutic antibodies in the widely used NSG and NOG mice, where the abnormal clearance reduced antitumor activity of ADC. Importantly, this phenomenon also occurs for several antibodies (chimeric αCD30, humanized αCD70, or nontargeted humanized IgG1). Recently, Sharma and colleagues have reported the anomalous biodistribution of trastuzumab and cetuximab in the NSG mice using in vivo imaging technology (21). These complementary findings uncover a previously underappreciated role of mouse strains on the preclinical evaluation of therapeutic antibodies. For example, a previous study reported that the advantage of stable drug linker was minimal in a tumor model grown in NSG mice in contrast to the improvement in SCID xenograft model (22). A systematic evaluation of how antibodies, including other subclasses of IgG antibodies, behave across multiple mouse strains may enhance the translation of preclinical PK/PD modeling into human studies.

Multiple mechanisms may influence the pharmacokinetics of therapeutic antibodies and ADCs (23). First, binding to target antigen expressed on certain tissues may cause target-mediated drug disposition. Considering none of the antibodies used in this study are cross-reactive to mouse antigens, the contribution of target-engagement was not examined here. Second, binding with neonatal Fc receptor (FcRn) governs the recycling and half-life of IgG antibodies in circulation (24). Our study indicates a third mechanism may influence the pharmacokinetics of IgG antibodies through interaction with FcγR on myeloid cells. Interestingly, several studies have demonstrated that Fc–FcγR interaction with tumor-associated macrophages and other antigen-presenting cells may influence the antitumor efficacy of antibody and ADCs in animal models (25–27). However, the contribution of FcγR binding on antibody PK has not been well understood, and it may vary between species (28). Nevertheless, future clinical studies comparing the pharmacokinetics of therapeutic antibodies with altered FcγR and FcRn binding capacity will help elucidate the specific contributions of these receptors in human.

Although researchers tend to choose the most immune-competent mouse strains for preclinical studies, highly immune-deficient strains such as NSG are often required to model human disease. For example, in NOD/SCID mice only a rare fraction of human melanoma cells can establish tumor, in contrast to NSG mice where 25% of melanoma cells can establish tumors (29). Similarly, development of patient-derived xenograft models, particularly those of hematologic origin, has had more success using NSG mice as host (30). More recently, humanized mice derived from NSG mice are adopted to study infectious disease, cancer, graft versus host disease (31). Our study shows that the study of therapeutic antibodies may have been penalized in NSG mouse models, due to the anomalous PK profile. However, one can circumvent the caveat by pretreating mice with hIVIg or using FcγR-mutated antibodies. This study also raises concern on combination studies involving two antibodies in NSG mice, where potential pharmacokinetics changes may confound the results. Nevertheless, with the recent development of transgenic strains containing human FcRn or FcγR genes, NSG mice incorporating these transgenes may serve as better animal models for preclinical evaluation of therapeutic antibodies (32, 33).

J. Neale is a director at and has ownership interest (including stock, patents, etc.) in Magenta Therapeutics. Dennis R. Benjamin has ownership interest (including stock, patents, etc.) in Seattle Genetics, Inc. No potential conflicts of interest were disclosed by the other authors.

Conception and design: F. Li, V.F.-S. Shih

Development of methodology: V.F.-S. Shih

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.L. Ulrich, V.F.-S. Shih, J.H. Cochran, J.H. Hunter, L. Westendorf

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): F. Li, V.F.-S. Shih, J.H. Cochran, J.H. Hunter, J. Neale, D.R. Benjamin

Writing, review, and/or revision of the manuscript: F. Li, V.F.-S. Shih, J.H. Cochran, J.H. Hunter, L. Westendorf, J. Neale, D.R. Benjamin

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

Study supervision: F. Li

We appreciate technical help from many colleagues at Seattle Genetics, especially from Devra Olson, Ivan Stone, Cassie Baker Lee, Serena Wo, and Christine O'Day.

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.