Otelixizumab is a monoclonal antibody (mAb) directed to human CD3ε, a protein forming part of the CD3/T-cell receptor (TCR) complex on T lymphocytes. This study investigated the temporal interaction between varying concentrations of otelixizumab, binding to human CD3 antigen, and expression of CD3/TCR complexes on lymphocytes in vitro, free from the confounding influence of changing lymphocyte frequencies observed in vivo. A static in vitro culture system was established in which primary human peripheral blood mononuclear cells (PBMCs) were incubated over an extended time course with titrated concentrations of otelixizumab. At each time point, free, bound, and total CD3/TCR expression on both CD4+ and CD8+ T cells and the amount of free otelixizumab antibody in the supernatant were measured. The pharmacokinetics of free otelixizumab in the culture supernatants was saturable, with a shorter apparent half-life at low concentration. Correspondingly, a rapid, otelixizumab concentration–, and time-dependent reduction in CD3/TCR expression was observed. These combined observations were consistent with the phenomenon known as target-mediated drug disposition (TMDD). A mechanistic, mathematical pharmacokinetic/pharmacodynamic (PK/PD) model was then used to characterize the free otelixizumab-CD3 expression-time relationship. CD3/TCR modulation induced by otelixizumab was found to be relatively fast compared with the re-expression rate of CD3/TCR complexes following otelixizumab removal from supernatants. In summary, the CD3/TCR receptor has been shown to have a major role in determining otelixizumab disposition. A mechanistic PK/PD model successfully captured the PK and PD in vitro data, confirming TMDD by otelixizumab.
Monoclonal antibodies (mAb) directed against human lymphocyte antigen CD3ε have been investigated clinically in a number of autoimmune diseases such as type 1 diabetes (Herold et al., 2002; Keymeulen et al., 2005, Hagopian et al., 2013), ulcerative colitis (Sandborn et al., 2010), multiple sclerosis (Weinshenker et al., 1991), and Crohn’s disease (van der Woude et al., 2010). One such anti-CD3 mAb, muromonab-CD3 (tradename Orthoclone OKT3), is approved for the treatment of steroid-resistant acute rejection of allogeneic renal, heart, and liver transplants. Despite encouraging results from clinical investigations in autoimmune diseases, especially in new-onset type 1 diabetes, there has been little progress in understanding the relationship between anti-CD3 mAb target engagement and its putative mechanisms of action. Owing to this lack of pharmacological understanding, dosage regimens appear to have been determined largely empirically, or when animal disease or in vitro models were used for extrapolation, the ability to adjust for animal-human or in vitro–in vivo differences was limited. The following pharmacology has been observed clinically with anti-CD3 mAbs: 1) Binding leads to apparent partial T-cell activation, resulting in release of proinflammatory cytokines, and a temporary disruption of normal T-cell trafficking (Chatenoud et al., 1982; Waldron Smith et al., 1997; Smith et al., 1998; -Lynch et al., 2012) and 2) internalization of the resultant mAb/CD3/T-cell receptor (TCR) complex causes loss of the complex from the T-cell surface membrane and degradation and elimination of the antibody (Reinherz et al., 1982; Press et al., 1988; Liu et al., 2000; Monjas et al., 2004), a phenomenon known as target-mediated drug disposition (TMDD) (Levy, 1994). Whereas some quantification of these clinical observations has been possible (Wiczling et al., 2010), it has been difficult separating and quantifying these individual pharmacological components at a mechanistic level owing to the close temporal relationship between them.
In an attempt to overcome the limitations of studying anti-CD3 mAb pharmacology in vivo, a static in vitro culture system was established to investigate the time course of the interaction between anti-CD3 mAb, CD3, and CD3/TCR complex, free from the confounding influence of changes in number of circulating T lymphocytes, observed as transient peripheral lymphopenia in vivo. A mechanistic nonlinear mixed effect pharmacokinetic/pharmacodynamic (PK/PD) model was then used to analyze and describe the complex pharmacological interactions in this static in vitro system. The in vitro pharmacokinetics was investigated in which primary human peripheral blood mononuclear cells (PBMCs) were used; CD3+ T lymphocytes in PBMCs is typically in the range of 50–70%. PBMCs were cultured over an extended time course with titrated concentrations of otelixizumab. At each time point, free, bound, and total CD3/TCR expression on both CD4+ and CD8+ T cells and the amount of free antibody in the supernatant were measured so that PK could be related to PD effects. To investigate the kinetics of CD3/TCR re-expression, cells were washed on day 2 to remove exogenous otelixizumab and thereby allow the rate of CD3/TCR complex re-expression to be monitored.
Materials and Methods
PBMCs isolated from healthy volunteer bloods were cultured in the presence of a titration of otelixizumab over an extended time course (Fig. 1). Flow cytometry was used to measure free and bound CD3ε and TCR at the indicated time points. Simultaneously, an enzyme-linked immunosorbent assay (ELISA) was used to measure the amount of free antibody in the supernatant of these cultures. In addition, some cells were washed to remove exogenous antibody after 2 days and measurement of CD3/TCR re-expression examined at the indicated time points postwash.
Otelixizumab (also known as ChAglyCD3) is an aglycosylated nonmitogenic recombinant antibody (human γ1) directed against CD3ε chain, a protein forming part of the CD3/T-cell receptor complex (TCR) on T lymphocytes (Routledge et al., 1991; Bolt et al., 1993). Otelixizumab reference standard at 12.1 mg/ml was used for all experiments.
Isolation of PBMC.
Healthy volunteer blood were obtained from the on-site blood donation unit and appropriate informed consent was obtained from all donors. Blood was withdrawn into bags containing 1 IU/ml sodium heparin (MP Biomedicals, Santa Ana, CA) as anticoagulant. Heparinized blood was kept at room temperature and processed within 2 hours of withdrawal. PBMC separation was performed by density gradient centrifugation using 50-ml Leucosep tubes (Greiner Bio-One, Kremsmünster, Upper Austria) containing Ficoll-Paque. Blood (25 ml per tube) was centrifuged for 20 minutes at 800g with the brake off. PBMCs recovered from the interface were washed twice with Dulbecco’s phosphate-buffered saline (PBS; Gibco/Thermo Fisher Scientific, Sunnyvale, CA) and finally resuspended in complete medium for counting using the guava easyCyte flow cytometer (EMD Millipore, Billerica, MA). Complete medium comprised RPMI 1640 (Gibco) + 1 mM glutamine (Gibco), 100 IU/ml penicillin/100 μg/ml streptomycin (Gibco) and 5% heat-inactivated human AB serum (Gemini Bio-Products, West Sacramento, CA).
Time Course of PBMC Treatment with Otelixizumab.
Freshly isolated PBMC from two healthy donors (1 × 106/ml) were cultured in 24-well plates (Costar/Corning, Corning, NY) in RPMI + 5% AB serum with titrated concentrations of otelixizumab (0, 1, 3, 10, 30, 100, 300, 1000, 3000, and 10,000 ng/ml), 16 wells for each concentration. At time points 0.25, 0.5, 1, 2, 4, 8, 16 hours and 1, 2, 3, 4, 6, 8, 10, 12, and 14 days the cell suspension was removed from one well for each donor transferred to fluorescence-activated cell sorting (FACS) tubes and centrifuged 5 minutes at 1350 rpm. The supernatant was removed and stored for subsequent ELISA at –20°C and the cells were resuspended in 4 ml FACS buffer and split between two tubes for FACS staining. For CD3/TCR re-expression PBMC from each donor were cultured at 1 × 106/ml (10 ml) in RPMI + 5% AB serum with the same titrated concentrations of otelixizumab. After 48 hours cells were removed from wells and transferred to 50-ml tubes and centrifuged at 1350 rpm for 5 minutes. Cells were resuspended in 50 ml PBS and centrifuged again. After resuspension in fresh RPMI + 5% AB serum, counting cells were adjusted to 106/ml and 1 ml/well was added to 24-well plates, eight replicate wells for each population. One well of each replicate was harvested and the cells stained for FACS as for other cells at 8 hours, and 1, 2, 4, 6, 8, 10, and 12 days.
Quantification of CD3/TCR Modulation by Flow Cytometry.
At the indicated time points the cells were removed, split between two tubes, and washed twice with FACS buffer (PBS + 1% FCS + sodium azide 0.1%). After blocking of Fc receptors for 5 minutes with TruStain FcX (BioLegend, Sacramento, CA) cells were surface-stained with the following murine antibodies: Tube A: CD4-PerCP-Cy5.5 (clone RPA-T4), CD8-APC (clone RPA-T8), CD3-FITC (clone SK7), and anti-Human IgG Fc-PE (clone HP6017); and Tube B: CD4-PerCP-Cy5.5 (clone RPA-T4), CD8-APC (clone RPA-T8), and TCRαβ-PE (clone IP26) (all BioLegend) for 30 minutes at room temperature protected from light. After washing twice with 3 ml FACS buffer the cells were resuspended in 300 μl Cytofix (BD Biosciences, San Jose, CA). Lymphocytes were gated on forward scatter versus side scatter and 5000 CD4 events were acquired. The mean fluorescence intensity data for phycoerythrin and fluorescein isothiocyanate fluorescence for gated CD4 and CD8 cells were converted to mean equivalent surface fluorescence (MESF) values using Quantum Simply Cellular beads (Bangs Laboratories, Fishers, IN) per the manufacturer’s instructions.
ELISA plates (96-well; Greiner Bio-One) were coated with capture antibody (anti-otelixizumab mAb clone T54.1C9.A9) at 5 μg/ml in 50 mM bicarbonate buffer overnight at 4°C, washed ×5 with PBS + 0.05%Tween 20, and blocked with Superblock (Thermo Scientific). Assay samples and otelixizumab standards (prepared in culture supernatant) were added to the prepared plates for 2 hours. After washing, 1 μg/ml detection antibody (biotinylated anti-otelixizumab mAb clone 6H12.B12.A3.G10) was added for 1 hour followed by addition of a streptavidin–horseradish peroxidase conjugate (Invitrogen, Carlsbad, CA) at 1/10,000 dilution for 1 hour. A chromogenic substrate tetramethylbenzidine (Sigma-Aldrich, St. Louis, MO) was used and color allowed to develop for 10 minutes before reaction was stopped with 0.5 M sulfuric acid. A Molecular Devices SpectroMax plate reader provided readings at 450 nm, and otelixizumab concentrations were calculated from a four-parameter fit standard curve using analytical software (Softmax Pro; Molecular Devices, Sunnyvale, CA). The lower limit of quantification for this assay was 2.5 ng/ml.
In Vitro PK/PD Model.
A TMDD model (Mager and Jusko, 2001) accounting for otelixizumab binding on both CD4+ and CD8+ lymphocytes was proposed to describe in vitro experimental data, including the extension of two binding sites (Gibiansky and Gibiansky, 2010). The proposed model is illustrated in Fig. 2.
Otelixizumab (here denoted by Cp) is administered in the central compartment, where it can bind to free CD3/TCR receptor complex on both CD4+ and CD8+ lymphocytes (FR4 and FR8, respectively) to form drug-receptor complexes (DR4 and DR8, respectively).
The model is described through the following differential equations expressed as molar concentrations:(1)(2)(3)(4)(5)with initial conditions given by Cp(0) = 0, FR4(0) = ksyn4/kdeg4 = BAS4, DR4(0) = 0, FR8(0) = ksyn8/kdeg8 = BAS8, DR8(0) = 0.
Assuming equal otelixizumab binding to CD4+ and CD8+ lymphocytes, the model can be simplified by imposing equal association and dissociation rate constants (i.e., kon4 = kon8 = kon and koff4 = koff8 = koff).
Otelixizumab molar concentration was converted into the observed concentration (ng/ml) using its molecular weight (150 kDa) and appropriate scaling (apparent volume of distribution V). Receptor dynamic measurements were expressed in MESF units assumed to be proportional to receptors molar concentrations. Otelixizumab bound to CD3 blocked the binding of all available anti-TCR and anti-CD3 detection antibodies that were tested. As a consequence, free and total receptors could be detected only where otelixizumab was not bound to CD3. This implied that total receptor measurements (MTR4 and MTR8) were proportional to free receptors molar concentration (and not to the sum of free and bound receptor concentrations):(6)(7)(8)where γFR, γDR, and γTR are conversion factors between variables and the measured quantities MFR4, MFR8, MDR4, MDR8, MTR4, and MTR8.
The PK/PD model was fitted to the individual PK and PD (free, engaged, and total CD3/TCR receptors) data simultaneously using nonlinear mixed-effect statistical analysis. Model development included differences for the two donors in baseline values for CD4+ and CD8+ T lymphocytes and different otelixizumab affinities for binding, respectively, on CD4+ and CD8+ T lymphocytes. Additive, proportional, and combined residual error models were tested for each variable. The data fitting software used was NONMEM version 7.2 (Beal et al., 2009) embedded in a wrapper program Perl-Speaks Nonmem (PsN version 3.6.2) (Lindbom et al., 2004) in a Windows 7 environment. The first-order conditional estimation with interaction (FOCEI) method was used throughout the analysis, and LSODA algorithm was used to integrate differential equations (ADVAN13 subroutine with TOL = 6). A nonparametric bootstrap method (n = 1000) was used to study the uncertainty of all TMDD model parameter estimates. From the bootstrap empirical posterior distribution, relative standard errors (RSE) were obtained for the parameters.
The criteria for model selection included: 1) improved fitting of the diagnostic scatter plots (observations versus predictions, weighted residual versus predictions or time, individual weighted residual versus predictions or time), 2) convergence of the minimization, 3) residual standard errors on all the estimated parameters, 4) significant decrease in the objective function value (OFV).
Pharmacokinetics of Free Otelixizumab.
To describe pharmacological interactions in a static in vitro culture system, the pharmacokinetics of otelixizumab was investigated. Free otelixizumab concentration time-course profiles are shown in Fig. 3 from two donors at different initial drug concentrations over a 48-hour incubation period. At initial drug concentrations higher than 100 ng/ml, otelixizumab concentration levels were relatively static over 48 hours. In contrast, there was an apparent initial concentration-dependent and time-dependent reduction in free otelixizumab concentrations observed with both donors at initial concentrations lower than 100 ng/ml. At the lowest starting concentrations (3 and 1 ng/ml), free otelixizumab concentrations were below the limit of quantification for both donors. Some values were obtained by extrapolation of the standard curve at early time points, but antibody was undetectable at later time points. Collectively, these observations suggest target-mediated elimination of otelixizumab in this system.
Pharmacodynamics of Free/Bound CD3 and Total CD3 Expression.
Next, the in vitro pharmacodynamic time courses of free and bound CD3ε together with total CD3ε expression on CD4+ and CD8+ T lymphocytes in the PBMC culture following incubation with otelixizumab were determined. CD4+ T lymphocytes pharmacodynamic time-course profiles selected initial concentrations for donor 1 are shown in Fig. 4. Free CD3ε and total CD3ε expression followed similar time courses. There was a rapid decline in both free CD3ε and total CD3ε expression relative to preincubation levels. In both cases, the rate and extent of reduction appeared to increase with higher initial concentrations, the nadirs occurring as early as 0.25 hours following the start of incubation and being maintained for the whole 48-hour incubation period at initial concentrations ≥1000 ng/ml. Similar results were obtained with both CD4+ and CD8+ cells and results from both donors were comparable (data not shown).
Following washout of otelixizumab after 48 hours, there was a gradual but steady increase in free CD3ε and re-expression of total CD3ε. Based on visual inspection of the data, the rate of re-expression of total CD3ε appeared to be similar for all initial concentrations and independent of the degree of reduction of CD3ε expression induced by incubation with otelixizumab. For low initial concentrations (<3 ng/ml) expression levels returned to preincubation levels by approximately day 6 following wash out of otelixizumab. In contrast, at the highest initial concentration (10,000 ng/ml) expression had recovered to only around 80% of preincubation levels by the end of the experiment (day 14). Again, results were consistent for both CD4+ and CD8+ cells and comparable between donors (data not shown).
Characterization of the PK/PD–Time Relationship.
Mechanistic, mathematical, TMDD PK/PD models were fitted simultaneously to the PK and PD (free, engaged, and total CD3/TCR receptors) data. In terms of the model iteration and adequacy, the most robust best fit model included different baseline levels for CD3/TCR on CD4+ and CD8+ T lymphocytes (BAS4 and BAS8 respectively) for the two donors. The final model did not include any between-subject variability on model parameters, so that the resulting analysis can be considered as a naïve-pooling approach. The binding affinities for otelixizumab for CD3ε on CD4+ or CD8+ lymphocytes were found to be similar and the model was simplified by assuming equal affinity for both lymphocyte subsets (i.e., kon4 = kon8 = kon and koff4 = koff8 = koff).
The final model included combined residual error model for total CD3/TCR receptor measurements and free otelixizumab concentration, and an additive residual error model was adopted for free and engaged CD3/TCR receptors. All parameters were estimated with reasonable precision (<40%). Goodness-of fit diagnostic plots, including comparisons of observed and predicted values, as well as residual analysis, suggested that although the model tended to underpredict the lowest free and total CD3/TCR observations, it provided a reasonable approximation for the data (Supplemental Data). The proposed model captured the PK/PD time-course profiles reasonably well for both donors and all initial concentrations of otelixizumab (Figs. 3 and 4).
The parameter estimates for the model are reported in Table 1 together with relative standard errors (RSE) based on 1000 bootstrapped datasets. The binding constants were estimated as kon = 51.5/nM per day and koff = 4.64 per day, suggesting high affinity of otelixizumab to human CD3ε. Correspondingly, the derived equilibrium dissociation constant, KD (koff/kon) was 90 pM.
Estimated internalization rates (kint4 = 1.26/day and kint8 = 1.29/day) were five times higher than degradation rates (kdeg4 = 0.273/day and kdeg8 = 0.275/day).
We established a static in vitro culture system to investigate and characterize otelixizumab pharmacology using human cells. We were able to study the dose range and time course with sufficient resolution to observe and quantify the concentration and time-dependent PK behavior consistent with target-mediated disposition of the antibody. A mechanistic, mathematical model (TMDD model) used to describe this phenomena was fitted to the PK/PD–time data. The model adequately described all the data and was sufficient to explain the nonlinear disposition of otelixizumab in this culture system. The rate and degree of change of free, bound, and total CD3ε expression was found to be determined by the concentration of otelixizumab. A moderate increase was observed in total CD3/TCR receptor profiles at the end of the experiment. Although data described in this paper does not allow any conclusion regarding the cause of this modest increase, it is possible that a small degree of cell death (approximately 10% was observed during the 14-day culture) is a contributing factor. This was, however, not considered a relevant feature to implement in the final model as it was judged to be relatively minor compared with the degree of CD3/TCR decreases achieved rapidly following otelixizumab engagement. Although the use of a constant synthesis rate for CD3/TCR receptors does not account for possible changes in cell culture compositions attributable to cell death, the proposed model captured the concentration-dependent kinetic profiles for free, bound, and total CD3ε reasonably well. At saturating concentrations, the engagement of target CD3ε molecules was found to be rapid, leading to full occupancy of CD3ε within minutes. At lower concentrations, binding of antibody gradually achieved equilibrium. The estimated parameters appeared reasonable and consistent with an antibody of this type. According to the model, the affinity constant for otelixizumab binding to CD3ε is KD = 90 pM. This finding is in line with observations made in a previous study (Wiczling et al., 2010), in which the estimated concentration of otelixizumab producing 50% reduction in free CD3/TCR sites was in the range of 14–16 ng/ml (= 93.8–107.2 pM). However, Wiczling et al. estimated a nonlinear PK model (Michaelis-Menten elimination) with a Michaelis-Menten constant (Km = 0.968 μg/ml) that is not in agreement with our in vitro results. In fact, a Michaelis-Menten PK model can be considered still an approximation of a full TMDD model in which the nonlinear elimination Km is related to the binding and internalization process (Km = (koff + kint)/kon) (Ma, 2012). In our specific case, since the estimated dissociation constant is faster than the internalization rate, the net result of the binding-internalization process is driven by the affinity estimate (Km ∼ KD). Wiczling et al. suggested that binding to CD3/TCR complexes did not affect otelixizumab pharmacokinetics, and they considered the short terminal half-life (t1/2 of approximately 0.5 days) to be related to the absence of FcRn recycling. The data presented here suggest that the process of CD3/TCR internalization is stimulated as a consequence of antibody binding, with an estimated internalization rate of drug-receptor complexes significantly faster than the free receptor degradation (kint = 1.26–1.29 day−1 > kdeg = 0.27–0.28 day−1), resulting in a mean residence time of the free receptor of approximately five times higher than the one of the drug-receptor complexes. This phenomenon is a common mechanism when natural ligand binds to its corresponding receptor and has been described for antigen binding to the T-cell receptor (Huang et al., 1999), and the B-cell receptor (Caballero et al., 2006), and for ligand binding to the granulocyte-macrophage–colony-stimulating factor receptor (Vainshtein et al., 2015). Following removal of antibody from the culture, the rate of re-expression of CD3ε was observed to be slower than otelixizumab-induced downregulation. Model estimates for internalization and degradation rate constants for CD3/TCR receptors reflect these findings. At saturating concentrations, complete modulation was achieved within about 30 minutes, whereas full CD3/TCR re-expression to pretreatment levels required about 8 days.
These findings are in agreement with data obtained from mouse models using surrogate anti-CD3 antibody fragments (Mehta et al., 2010). These observations are supportive of the notion that CD3ε reappearance requires new synthesis and assembly of complexes, known to be a slow process that takes days, as opposed to re-expression of transiently internalized existing complexes, which has been described as a relatively rapid process (minutes to hour) (Menné et al., 2002). It should be noted that the exact fate of otelixizumab was not investigated in this study. Several mechanisms are possible, with some studies demonstrating endocytosis of antibody receptor complexes (Liu et al., 2000; Monjas et al., 2004; Kuhn et al., 2011), whereas others suggest shedding of the complex from the cell surface (Reinherz et al., 1982; Press et al., 1988).
In contrast, despite considerable investigation, the degree, duration of action of otelixizumab binding to CD3ε, CD3 expression, and its relationship to unbound antibody has only been partially elucidated from clinical data alone. For practical reasons, it has been difficult to study a sufficiently wide dose range with an adequate blood sampling schedule following treatment of otelixizumab in patients to determine whether the dynamics of antibody binding can be separated from the dynamics of the observed transient lymphopenia. Thus, attempts to quantify these relationships clinically have been largely empirical and simple (Wiczling et al., 2010). Whereas the published models describe the data adequately, and may be useful for limited interpolation or extrapolation, they lack the mechanistic detail to provide insight about the underlying pharmacological behavior of otelixizumab. Understanding the in vivo interaction of otelixizumab to the CD3 target is challenging since observed data may not be sufficient to estimate relevant system parameters. In this case, the in vitro model can provide relevant information for the receptor synthesis (ksyn), degradation/internalization rate (kdeg and kint), and binding affinity (KD) that can help to better inform any in vivo model in which other phenomena (mAb distribution, disposition, and systemic clearance attributable to nonspecific target binding, together with target expression, lymphocytes trafficking, etc.) would increase the complexity of the model structure, including identifiability issues for relevant system parameters. In general, the development of mechanistic models from in vitro experiments offers a valuable approach to addressing the complexity of in vivo systems, for example, by using the in vitro estimates as priors for the in vivo models.
Preclinical investigation of these phenomena in vivo is limited by the almost complete lack of crossreactivity of otelixizumab with animal CD3 and lack of suitable animal surrogate antibody. In conclusion, we have demonstrated using an in vitro cell culture system and a mechanistic model–based approach to data analysis, that we can successfully study and quantify the complex antibody-receptor complex PK/PD interactions that were hypothesized for otelixizumab. This will probably be applicable to other membrane-bound targets, particularly for immune cell targets in which antibodies or other protein therapeutics may cause altered cell trafficking. The parameters estimated from such analysis may allow investigators to compare and contrast other data, and the models can be used to design dosage regimens and optimize sample schedules for future clinical trials.
Participated in research design: van Maurik, MacDonald, Page.
Conducted experiments: Page.
Performed data analysis: Mezzalana, De Nicolao.
Wrote or contributed to the writing of the manuscript: van Maurik, MacDonald, Page, Mezzalana, Zamuner, De Nicolao.
- enzyme-linked immunosorbent assay
- fluorescence-activated cell sorting
- binding affinity
- monoclonal antibody
- mean equivalent surface fluorescence
- peripheral blood mononuclear cells
- phosphate-buffered saline
- peridinin-chlorophyll-cyanine 5.5
- side scatter
- T-cell receptor
- target-mediated drug disposition
- Copyright © 2015 by The American Society for Pharmacology and Experimental Therapeutics