Investigation of the Influence of Protein-Losing Enteropathy on Monoclonal Antibody Pharmacokinetics in Mice

Abstract

Protein losing enteropathy (PLE), which is characterized by substantial loss of plasma proteins into the gastrointestinal (GI) tract, is a complication of a variety of GI diseases, including inflammatory bowel disease. Clinical studies have found that the clearance of monoclonal antibodies (mAb) is often increased in subjects with diseases known to cause PLE; however, direct relationships between PLE and mAb pharmacokinetics have not been demonstrated. This study employed a murine model of colitis to examine the influence of PLE on mAb pharmacokinetics. Mice were given dextran sodium sulfate (DSS, 2% w/v) supplemented tap water as drinking source for 6 days to induce colitis and PLE. Mice were then intravenously injected with 8C2, a murine IgG1 mAb. 8C2 plasma concentrations were measured up to 14 days post injection. Fecal alpha-1-antitrypsin (A1AT) clearance was measured as biomarker for PLE. DSS-treated mice developed PLE of clinically relevant severity. They also showed a transient increase in 8C2 plasma clearance and a decrease in 8C2 plasma exposure. The area under the 8C2 plasma concentration-time curve for the length of the study (AUC_0–14d) reduced from 1368 ± 255 to 594 ± 224 day μg/ml following DSS treatment (p = 0.001). A quantitative relationship between A1AT clearance and 8C2 clearance was obtained via population pharmacokinetic modeling. DSS treatment substantially increased 8C2 clearance and reduced 8C2 exposure. Increased mAb plasma clearance was highly correlated with A1AT fecal clearance, suggesting the possible utility of A1AT fecal clearance as a mechanistic biomarker to predict the pharmacokinetics of therapeutic antibodies.

INTRODUCTION

Great success has been achieved in the development of monoclonal antibodies (mAb) for the treatment of a wide range of disease conditions, including cancer and immunological disorders (1), and more than 45 mAb have been approved for therapeutic use in the US or Europe (2). The success of mAb therapeutics is due not only to their high target specificity and affinity, but also to their favorable pharmacokinetic (PK) properties. The long, 10–25-day elimination half-life of mAb therapeutics (3) allows for prolonged dosing intervals, reducing the frequency of clinical visits and improving patient compliance.

Substantial inter-individual variability (IIV) in mAb pharmacokinetics has been observed in clinical studies (4). For example, the IIV in mAb clearance ranges from 20 to 59% (percent coefficient of variation, CV%), as estimated by population pharmacokinetic modeling (4). Variability in mAb clearance may lead to variability in systemic exposure and treatment outcome, if dosing is not individualized. Although some patient factors (body weight, disease burden, presence of anti-drug antibodies) have been associated with substantial mAb PK variability, much of the IIV in mAb clearance is still unaccounted for. To date, very few studies have investigated the mechanistic determinants of mAb PK variability. It is likely that an improved understanding of mechanisms explaining IIV in mAb PK, and the development of mechanistic biomarkers that quantitatively predict mAb PK behavior, would improve the design of individualized dosing regimens, potentially allowing improved efficacy and reduced toxicity.

The gastrointestinal (GI) tract is lined with a continuous layer of epithelial cells on its luminal surface to separate the GI lumen from the internal milieu. This epithelial barrier controls the selective absorption of nutrients and the excretion of metabolic wastes and prevents massive GI loss of blood proteins such as albumin and immunoglobulins. However, under certain GI pathological conditions where the epithelial cell layer is damaged, increased permeability to macromolecules may result in a substantial loss of plasma proteins into the GI tract, which is termed protein-losing enteropathy (PLE) (5).

Many disease states have been associated with PLE, including gastric cancer (6), colon cancer (7), inflammatory bowel disease (IBD, including ulcerative colitis (8) and Crohn’s disease (5)), and systemic lupus erythematosus (9). There is published evidence showing increased gastrointestinal immunoglobulin excretion in gastric cancer (10), ulcerative colitis, and Crohn’s disease (11,12). For example, the median values of IgG intestinal clearance in 32 patients with ulcerative colitis and in 15 patients with active Crohn’s disease were found to be eightfold and fivefold greater than the median value found for healthy controls (both increases were statistically significant) (11).

The impact of GI immunoglobulin leakage on mAb PK has been suggested by several clinical studies that demonstrated that the clearance of several therapeutic mAb is higher in patients with diseases known to cause PLE compared to clearance values found for patients with diseases that are not associated with PLE. For example, the population mean clearance of infliximab is roughly 50% higher in patients with IBD than with rheumatoid arthritis or ankylosing spondylitis (13–16). The systemic exposure of trastuzumab, bevacizumab, and pertuzumab is 30–50% lower in patients with advanced gastric cancer (AGC) than in patients with other solid tumors when given the same dose (17–19).

Several clinical investigations have found mAb plasma clearance is negatively correlated with serum albumin concentration (17,20–22). Since PLE patients often develop hypoalbuminemia, and since the severity of PLE has been shown to be negatively correlated with serum albumin concentration (23), excessive GI protein loss may be able to explain the concurrence of low serum albumin concentration and high mAb plasma clearance. However, no diagnosis of PLE was attempted for the studied subjects in any of the aforementioned studies. As such, there is a lack of direct evidence for a relationship between PLE and the pharmacokinetics of mAb therapeutics.

The present study aims to evaluate the impact of PLE on the pharmacokinetics of monoclonal antibodies by analyzing the plasma PK of 8C2, a model murine IgG1 mAb, in a mouse model of PLE. Dextran sodium sulfate (DSS) administration led to a substantial increase in GI protein leakage. This was evaluated by measuring the intestinal clearance of alpha-1-antitrypsin (A1AT), a common clinical biomarker for PLE (24), using collected feces and plasma specimens. We observed increased 8C2 plasma clearance and decreased 8C2 plasma exposure in DSS-treated mice. Population PK modeling demonstrated that fecal A1AT clearance was positively correlated with 8C2 plasma clearance.

MATERIALS AND METHODS

Mice

Male Swiss Webster mice at age 6–8 weeks were purchased from Harlan Laboratories (Indianapolis, IN). Mice were housed in a temperature and humidity-controlled environment with a standard 12-h light/12-h dark cycle. All mice were allowed free access to food and drinking water. The animal study protocol was approved by the Institutional Animal Care and Use Committee (IACUC) of the State University of New York at Buffalo. All animal experiments were performed in accordance with the NIH Guide for the Care and Use of Laboratory Animals (25).

Animal Study

Male Swiss Webster mice, aged 6–8 weeks, were given sterile-filtered tap water supplemented with 2% (w/v 2 g/100 ml) DSS (MW 36,000–50,000, colitis grade, MP Biomedicals, Solon, OH). Mice in the treatment group were allowed access to DSS water ad libitum, as their only drinking source, for 6 days. Mice in the control group were given access to tap water (as their only drinking source). All mice were administered 8C2 (20 mg/kg; average weight of mice was 29.0 g just before dosing), via intravenous penile vein injection, at the conclusion of the 6-day run-in period (i.e., at the conclusion of DSS administration in the treatment group). Blood samples were collected from the submandibular vein or from the retro-orbital plexus at 3 and 8 h, and at 1, 4, 7, 10, and 14 days after 8C2 dosing. Plasma was separated from blood by centrifuging at 1900 g for 10 min. Feces and urine were collected for 24-h periods by placing each mouse into individual metabolic cages 0, 3, 6, 9, 12 days after 8C2 dosing. Mice were sacrificed 14 days after 8C2 dosing, and tissue samples of all major organs were collected, snap-frozen in liquid nitrogen, and stored at −80°C until further analysis. Mice were monitored daily for body weight, dehydration, activity, stool consistency, and rectal bleeding. Mice experiencing pain or distress were removed from the study (as dictated by the IACUC protocol). In all, four mice in the treatment group were censored; three mice died during the study, with one dying under anesthesia during 8C2 dosing. One mouse was euthanized due to distress (significant weight loss, dehydration, and reduced activity). Data collected from surviving mice (i.e., six mice in the treatment group and five mice in the control group) were used for all subsequent analyses.

Monoclonal Antibody Production and Purification

8C2, a high-affinity murine IgG1 anti-topotecan monoclonal antibody, was used as a model antibody for the PK study. 8C2 was produced and purified as previously described (26). Briefly, 8C2-producing hybridoma cells were cultured with serum free media (Hybridoma SFM, Gibco, Grand Island, NY) supplemented with 5 μg/ml gentamicin (Gibco, Grand Island, NY) in 1-L spinner flasks in an incubator maintained at 37°C, and 5% CO₂. 8C2-containing cell culture supernatant was collected, centrifuged at 3500g for 16 min, and filtered with a sterile 0.22-μm polystyrene bottle top filter (Corning Inc., Corning, NY). Antibody was then purified from the supernatant via protein-G affinity chromatography, using a 5-ml HiTrap Protein G HP column (GE Healthcare, Uppsala, Sweden) attached to a BioLogic DuoFlow medium-pressure chromatography system (Bio-Rad, Hercules, CA). The column was first washed with 40 ml 20 mM Na₂HPO₄ (pH 7.0) before 250 ml of cell culture supernatant was run through the column. 8C2 antibody was then eluted with 40 ml 100 mM glycine buffer (pH 2.8) and fraction-collected into glass test tubes, each containing two drops of 1 M Tris-HCL buffer (pH 9.0). Purified 8C2 was buffer exchanged into phosphate-buffered saline (PBS) using a 7K MWCO Snake-Skin Dialysis Tubing (Thermo Fisher Scientific, Waltham, MA). The antibody was concentrated to about 1 mg/ml, and its exact concentration was measured by UV absorbance using NanoDrop 1000 Spectrophotometer (Thermo Fisher Scientific, Waltham, MA, 1 mg/ml IgG = 1.37 absorbance units).

ELISA for Quantification of 8C2 Plasma Concentration

Plasma concentrations of 8C2 were determined with an antigen-capture enzyme-linked immunosorbent assay (ELISA) (27). Briefly, 8C2 was captured onto the surface of microwell plates using a cationized bovine serum albumin-topotecan conjugate (cBSA-top), which was synthesized with a carbodiimide-catalyzed amide bond reaction (27). Nunc MaxiSorp 96 well plates (VWR, catalog #62409–002, Bridgeport, NJ) were coated overnight at 4°C with cBSA-top dissolved in 0.05 M MES (2 μg/ml, 250 μl/well). Plates were washed thrice with PB-Tween (0.05% Tween-20 in 0.02 M Na₂HPO₄), followed by three washes with reverse osmosis water. Plates were then blocked with 2% bovine serum albumin (US Biological, Salem, MA) in PB-Tween at room temperature for 2 h. After washing as described above, plates were incubated with standards and samples (diluted 100-fold with PBS) in triplicate (250 μl/well) for 2 h at room temperature. At the end of incubation, the plates were washed and then incubated with 250 μl of goat anti-mouse IgG (whole molecule specific)-alkaline phosphatase conjugate (Sigma, Cat #A5153, St. Louis, MO) for 1 h at room temperature (1:2000 dilution with 1% bovine serum albumin in PBS). After washing, p-nitrophenyl phosphate (Pierce, Rockford, IL) solution, 4 mg/ml in diethanolamine buffer (Pierce, Rockford, IL), was added to each well (250 μl/well). The change in absorbance at 405 nm with respect to time (dA/dt) was measured for 10 min using a plate reader (Spectra Max 340, Molecular Devices, Sunnyvale, CA), and the standard curve was obtained by fitting dA/dt vs. 8C2 standard concentrations with a 4-parameter equation. The lower limit of quantification of the assay is 5 μg/ml.

Measurement of A1AT Intestinal Clearance

A1AT intestinal clearance was calculated as follows:

$$\text{A}1\text{A}\mathrm{T}\text{ intestinal clearance}=\frac{\mathit{\text{amount of A1AT in a }}24h\mathit{\text{ feces collection}}}{\mathit{\text{A}}1\mathit{\text{AT plasma concentration}}}$$

A1AT plasma concentrations, which were treated as “steady-state” values, were measured ~ 1 h prior to the end of the respective 24-h feces collection periods. Of note, A1AT plasma concentrations were found to vary little with time in each mouse studied (average within animal CV% = 6.79%). A1AT intestinal clearance is preferred as a clinical biomarker over amount of A1AT in feces in order to account for substantial inter-individual variability of A1AT plasma concentrations in patients (24). Therefore, the clearance value was measured in our study. A1AT clearance values were further normalized by the mouse body weight measured just before 8C2 administration, and the first measurement of A1AT clearance (mouse body weight varied about 2–5% [CV%] thereafter). To measure the amount of A1AT in a 24-h feces collection, ten dry fecal pellets were randomly picked and used as a representative sample. The fecal sample and total feces collection were weighed, to determine the mass fraction of the sample relative to the collection. 1.5 ml PBS was then added to the sample. After soaking for 10 min at room temperature, the fecal pellets were homogenized through vigorous vortexing and separation with a pipette tip. The feces homogenate was centrifuged at 15,000g for 20 min under 4°C. A1AT concentration in the supernatant was measured using a commercially available ELISA kit (Immunology Consultants Laboratory, Portland, OR). The total amount of A1AT in the 24-h feces collection was then back-calculated. A1AT plasma concentration was also measured with the ELISA kit.

Measurement of Urinary Albumin Excretion Rate

Concentrations of albumin in 24-h urine collections were measured using a commercial ELISA kit (Exocell, Philadelphia, PA). Urinary albumin excretion rates were then calculated as follows:

$$\mathit{\text{urinary albumin excretion rate}}=\mathit{\text{albumin concentration in collected urine}}\times \mathit{\text{volume of the }}24-\mathrm{h}\mathit{\text{ urine collection}}/24\text{h}$$

Volume of urine was converted from weight of urine assuming a density of 1 g/ml.

Measurement of Plasma Albumin Concentration

Mouse plasma albumin concentrations were measured using a commercial ELISA kit (Exocell, Philadelphia, PA). Plasma was obtained following the centrifugation of blood samples that were collected at 3 h, 7 days, and 14 days after 8C2 dosing.

Measurement of Plasma Murine IgG Concentration

Murine IgG concentrations in mouse plasma were measured using a commercial ELISA kit (Abcam, Cambridge, UK). Plasma was obtained following the centrifugation of blood samples that were collected 1 day prior to, and 3 h, 8 h, 7 days, and 14 days after 8C2 dosing.

Measurement of FcRn Protein Expression in Mouse Tissue

Mouse tissue samples were homogenized using a Dounce tissue grinder (Wheaton, Millville, NJ) containing 1000 μl of RIPA lysis buffer (Thermo Fisher Scientific, Waltham, MA) supplemented with protease inhibitor cocktail (Sigma, St. Louis, MO). After transfer into a microcentrifuge tube, the tissue homogenate was incubated at 4°C for 1 h with agitation. Tissue debris was pelleted by centrifuging at 14,000g for 20 min under 4°C. The supernatant was used for SDS-PAGE protein separation after being mixed with loading buffer (Pierce Lane Marker Reducing Sample Buffer, Thermo Fisher Scientific, Waltham, MA) and heated at 90°C for 5 min. Samples were loaded onto 12% Precise Tris-HEPES Gel (Thermo Fisher Scientific, Waltham, MA). Proteins were separated at a voltage of 50 V for 10 min followed by 100 V for 50 min. Proteins in the gel were transferred to an Immun-Blot PVDF membrane (Bio-Rad, Hercules, CA) at 40 V for 90 min with Tris-Glycine/20% methanol transfer buffer. The membrane was then blocked with TBST (Tris-buffered saline with 0.1% tween-20) supplemented with 5% milk for 1 h at room temperature. After blocking, the membrane was cut into halves: the section of the membrane containing protein with a size from 25 to 75 kDa was probed for FcRn with 2 μg/ml of goat anti-mouse FcRn Antibody (R&D systems, Minneapolis, MN) overnight at 4°C on a shaker followed by incubation with HRP-conjugated donkey anti-goat IgG antibody (EMD Millipore, Billerica, MA) (1:1000 dilution in TBST with 5% milk); the section of the membrane containing protein with a size larger than 75 kDa was probed for vinculin (as a loading control) with 0.05 μg/ml of rabbit anti-mouse vinculin antibody (Abcam, Cambridge, MA) overnight at 4°C on shaker followed by incubation with HRP-conjugated goat anti-rabbit IgG antibody (EMD Millipore, Billerica, MA) (1:15,000 dilution in TBST with 5% milk for liver tissue and 1:20,000 dilution for other tissue). After incubation with secondary antibodies for 1.5 h, the two halves of the membrane were washed five times with TBST and then incubated in SuperSignal West Pico Chemiluminescent Substrate (Thermo Fisher Scientific, Waltham, MA) for 5 min. Finally, the membranes were analyzed using a ChemiDOC XRS system (Bio-Rad, Hercules, CA). The density of the band for FcRn was normalized to that of the band for vinculin.

Statistical Analysis

Unpaired Student’s t tests were used to test the statistical significance of difference between the DSS treatment group and the control group for measurements made at only one time point during the study. Two-way repeated-measures ANOVA was used to test the statistical significance of difference between treatment and control groups for observations made repetitively during the study. Type I error rate was selected to be 0.05. Non-compartmental analysis of 8C2 PK data was performed by Phoenix WinNonlin 6.4 (Pharsight, Palo Alto, CA) using the linear up log down trapezoidal method.

Population Pharmacokinetic Modeling

Population PK modeling was employed to identify a quantitative relationship between A1AT intestinal clearance and 8C2 plasma clearance. Monolix software version 4.3.2 (LIXOFT, Paris, France) was used for nonlinear mixed effects modeling with the built-in Stochastic Approximation Expectation Maximization (SAEM) algorithm for parameter estimation.

A two-compartment linear mammillary model was set as the base structural model. The PK parameters of the model are linear clearance (CL), inter-compartmental distribution clearance (CL_d), volume of distribution of the central compartment (V_c), volume of distribution of the peripheral compartment (V_p).

$$V_c\times \frac{{\text{dC}}_c}{\text{dt}}={\text{CL}}_d\times C_p-\left(\text{CL}+{\text{CL}}_d\right)\times C_c;C_c(t=0)=\frac{\text{ Dose }}{V_c}$$ (1)

$$V_p\times \frac{{\text{dC}}_p}{\text{dt}}={\text{CL}}_d\times \left(C_c-C_p\right);C_p(t=0)=0$$ (2)

Inter-individual variability of PK parameters was modeled as log-normally distributed around population-estimated values:

$$P_i=P_{\text{pop}}\times e^{{\eta }_{i,P}}$$ (3)

where P_pop is the population estimate of a PK parameter denoted by P; P_i is the individual estimate of the parameter P for the ith individual, and η_i, _P represents the IIV of the ith individual on the parameter P, which is drawn from a normal distribution with mean of 0 and variance of ω².

For residual variability (RV), a proportional error model, constant error model, and a combined (proportional and constant) error model were tested. The combined error model was selected as the final RV model based on model performance.

$$C=C_c+\left(a+b\times C_c\right)\times e$$ (4)

where C is observed 8C2 plasma concentration; C_c is the model predicted 8C2 plasma concentration (i.e., 8C2 concentration in the central compartment). a is a coefficient for additive error, and b is a coefficient for proportional error. e is a random variable which follows a normal distribution with mean of 0 and variance of 1.

The influence of A1AT intestinal clearance on all the PK parameters with IIV modeled was tested in the covariate analysis process. Since the value of A1AT intestinal clearance is time-dependent, instantaneous A1AT clearance was interpolated by connecting values measured experimentally at adjacent time points with linear piecewise functions (as illustrated in Fig. 1b). Then the interpolated value was related to PK parameters using mathematical equations, thereby also making the PK parameters time-dependent. The forms of equations tested include linear, power, and exponential equations, all with A1AT clearance centered at its mean measured value.

$$P_{\text{pop}}=P\times \left[1+{\theta }_{\text{A}1\text{AT}}\times \left({\text{CL}}_{-}\text{A}1\text{AT}-{\text{CL}}_{-}\text{A}1{\text{AT}}_m\right)\right]$$ (5)

$$P_{\text{pop}}=P\times {\left(\frac{{\text{CL}}_{-}\text{A}1\text{AT}}{{\text{CL}}_{-}\text{A}1{\text{AT}}_m}\right)}^{{\theta }_{\text{A}1\text{AT}}}$$ (6)

$$P_{\text{pop}}=P\times \left[1+{\theta }_{\text{A}1{\text{AT}}_1}\cdot {\left(\frac{{\text{CL}}_{-}\text{A}1\text{AT}}{{\text{CL}}_{-}\text{A}1{\text{AT}}_m}\right)}^{{\theta }_{{\text{AlAT}}_2}}\right]$$ (7)

$$P_{\text{pop }}=P\times e^{{\theta }_{\text{A}1\text{AT}}\cdot \left(\text{CL}\_\text{A}1\text{AT}-\text{CL}\_\text{A}1{\text{AT}}_m\right)}$$ (8)

$$P_{\text{pop}}=P\times \left[1+{\theta }_{\text{A}1{\text{AT}}_1}\times e^{{\theta }_{\text{A}1{\text{AT}}_2}\cdot \left({\text{CL}}_{-}\text{A}1\text{AT}-\text{CL\_A}1{\text{AT}}_m\right)}\right]$$ (9)

where P_pop is the population-estimated typical value of a PK parameter; P, θ_A1AT, ${\theta }_{\text{A}1{\text{AT}}_1}$, and ${\theta }_{\text{A}1{\text{AT}}_2}$ are estimated “fixed” coefficients; CL_A1AT is the instantaneous A1AT intestinal clearance interpolated via the linear piecewise function described above; CL_A1AT_m is the mean measured A1AT intestinal clearance among all the studied animals, which is equal to 15.255 ml/day/kg.

Fig. 1.

Alpha-1-antitrypsin (A1AT) intestinal clearance increased substantially after dextran sodium sulfate (DSS) treatment. a Averaged A1AT intestinal clearance-time profile. DSS treatment increased A1AT intestinal clearance compared with the control group, although the magnitude of increase declined towards the end of experiment. Data are represented as mean ± SD. b Individual A1AT intestinal clearance-time profile. A1AT intestinal clearance was at maximum 0 or 3 days after 8C2 dosing (i.e., following the conclusion of the 6-day DSS treatment period). Large inter-individual variability among DSS-treated mice was observed

A full covariate model was built using the forward selection procedure, where covariate relationships improving the model significantly (p < 0.05, which is equal to a larger than 3.84 point reduction in the objective function value at a degree of freedom of 1) were included into the model. After re-adjusting the IIV and RV models, the significance of covariate effects in the full model was again tested by the backward elimination procedure (p < 0.001, which is equal to a larger than 10.83 point reduction in the objective function value at a degree of freedom of 1).

Model performance was evaluated based on diagnostic plots (observed vs. population-predicted concentration, observed vs. individual-predicted concentration, conditional weighted residual vs. time or predicted concentration, etc.), visual predictive check, the relative standard error (RSE) of parameter estimates, and the value of the objective function (which is equal to −2 × log-likelihood).

RESULTS

A1AT Intestinal Clearance

The clearance of A1AT eliminated from the gastrointestinal tract is widely used as a diagnostic biomarker for PLE in the clinic (28), and has been used in preclinical studies to evaluate intestinal protein leakage (29). A1AT intestinal clearance was measured 0, 3, 6, 9, and 12 days after the termination of a 6-day DSS induction period. A1AT intestinal clearance was calculated as the ratio between the amount of A1AT in a 24-h feces sample and the plasma concentration of A1AT measured within the 24-h feces collection period (refer to the Materials and Methods section for equation). The clearance value was further normalized by mouse body weight. A1AT intestinal clearance was significantly higher (two-way repeated-measures ANOVA p = 0.0193) in the DSS treatment group compared with the control group (Fig. 1a). The time-averaged A1AT intestinal clearance for DSS-treated mice was 25.8 ± 18.1 ml/day/kg, representing a 9.02 ± 7.02-fold (mean ± SD) increase compared with the mean control group value (2.58 ± 0.90 ml/day/kg, mean ± SD). Also, while A1AT intestinal clearance in the control group was fairly uniform among mice and stable across time points (CV% across all control mice and all time points = 38.9%), larger inter-individual and inter-occasion variability was observed for the DSS treatment group (CV% = 92.8%). Looking into individual mouse profiles (Fig. 1b), it was found that A1AT intestinal clearance was at maximum 0 or 3 days after 8C2 administration (when DSS treatment terminated). The peak A1AT intestinal clearance for DSS-treated mice was 46.3 ± 33.8 ml/day/kg (mean ± SD), representing an 18.3 ± 13.7-fold (mean ± SD, range 8.03- to 45.0-fold) increase relative to the mean control group value. Despite a gradual decline after reaching maximum value, A1AT intestinal clearance for DSS-treated mice were still at least 70% (2.50 ± 2.20-fold, mean ± SD) higher than the mean control group value 12 days after 8C2 dosing. The magnitude of increase in A1AT clearance after DSS treatment is comparable to clinical observations made in patients with gastric cancer (6), Crohn’s disease (30), and ulcerative colitis (11) as compared to healthy subjects. The A1AT clearance for one of the DSS-treated mice was much higher than that found for the rest of the mice (Fig. 1b). This finding, which suggests high inter-individual variability in intestinal protein loss among DSS-treated mice, is also typical of the PLE patient population (6,11,30). Given that the population PK modeling methodology is capable of handling such variability, we elected to include this mouse in our analysis (vs. treating the mouse as an outlier).

8C2 Pharmacokinetic Study

8C2, a murine IgG1 anti-topotecan monoclonal antibody, was intravenously injected (20 mg/kg) immediately after the end of the 6-day DSS treatment period. Blood samples were collected 3 h, 8 h, and 1, 4, 7, 10, 14 days after 8C2 dosing and 8C2 plasma concentrations were measured via ELISA. Given the relatively long half-life of 8C2, a longer study period would be desired. However, consistent with prior investigations of this antibody in mice, a 14-day study period was selected to minimize the potential for a confounding influence of host anti-drug antibody development (27). The PK profiles displayed bi-exponential behavior (Fig. 2b). Although DSS-treated mice had much greater inter-individual variability in their PK profiles, the slope for the initial decline of 8C2 plasma concentration was generally steeper for DSS-treated mice than control mice, but the terminal slopes were similar for the two groups (Fig. 2b). Therefore DSS treatment appears to increase 8C2 clearance transiently during the first few days of the PK study. This corresponds well with the time-dependent change in severity of PLE, as A1AT intestinal clearance was substantially higher in DSS-treated mice at the beginning of the PK study, but it started to decline towards the control group value thereafter. An alternative explanation for the steeper initial slope is that DSS treatment increased the rate and extent of peripheral distribution of 8C2. However, no evidence was found in the literature that would support possible effects of DSS on mechanistic determinants of mAb distribution (e.g., increases in tissue fluid volume, IgG binding in tissue, increases in systemic vascular porosity). PK parameters for each individual mouse were calculated by non-compartmental analysis (NCA). Due to the transient nature of the increase in 8C2 clearance following DSS treatment, the NCA-calculated 8C2 clearance was a time-averaged value that underrepresented the impact of DSS treatment on 8C2 clearance. Also, the non-stationary PK behavior in DSS-treated mice led to an inaccurate NCA calculation of 8C2 volume of distribution at steady state (V_ss) (31). Therefore, only the area under the 8C2 plasma concentration-time curve for the length of the study (from 0 to 14 days after 8C2 dose) (AUC_0–14d) was compared between groups (1368 ± 255 day μg/ml for control group vs. 594 ± 224 day μg/ml for DSS treatment group, mean ± SD). DSS-treated mice displayed significantly lower AUC_0–14d compared with control mice (unpaired Student’s t test p = 0.001). The mean AUC_0–14d of the DSS treatment group was 57% lower than that of the control group. DSS treatment resulted in a transient increase in 8C2 plasma clearance, which was substantial enough to cause a significant reduction in 8C2 systemic exposure.

Fig. 2.

Population pharmacokinetic (PK) modeling of 8C2 pharmacokinetics. Male Swiss-Webster mice were intravenously injected with 20 mg/kg 8C2 after dextran sodium sulfate (DSS) or control treatment. Applying alpha-1-antitrypsin (A1AT) clearance as a covariate on 8C2 clearance substantially improved the observed- vs. population-predicted 8C2 concentration plot. a Base structural model. A two compartment linear mammillary model was used. 8C2 first enters the central compartment (volume of distribution is denoted by V_c) as an i.v. bolus dose, and then distributes into the peripheral compartment (volume of distribution is denoted by V_p) via the linear inter-compartmental distribution clearance (CL_d). 8C2 is eliminated from the central compartment via the linear elimination clearance (CL). b 8C2 plasma concentration-time profiles for individual mice reveal bi-exponential behavior. This plot also highlights the inter-individual variability in 8C2 PK, especially for the DSS treatment group. 8C2 concentration profile for the initial 24 h after dosing is highlighted in the inset graph. c Observed- vs. population-predicted 8C2 concentrations, as generated through application of the base structural model. The solid line is the line of identity. d Observed- vs. population-predicted 8C2 concentrations, as generated by the final model that includes the covariate effect of A1AT clearance (on 8C2 clearance). The solid line is the line of identity

Both 8C2 plasma AUC_0–14d and NCA-calculated, time-averaged clearance were found to be linearly correlated with time-averaged values of A1AT intestinal clearance (coefficient of determination R² = 0.66 and 0.94, p = 0.002 and < 0.0001, respectively) (Fig. 3a, b). Higher A1AT clearance was associated with lower AUC_0–14d and higher 8C2 clearance.

Fig. 3.

Linear correlation between biomarkers and 8C2 pharmacokinetic parameters determined by non-compartmental analysis (NCA). Data are pooled from both the treatment and control groups. Each symbol represents an individual mouse. Solid lines are the best fitted lines determined by linear regression analysis. The coefficient of determination (R²) and p value of the linear regression are also displayed. a NCA-calculated 8C2 area under the plasma concentration-time curve (AUC_0–14d) was linearly correlated with the time-averaged alpha-1-antitrypsin (A1AT) intestinal clearance observed for individual mice. b NCA-calculated 8C2 clearance was linearly correlated with time-averaged A1AT intestinal clearance. c NCA-calculated 8C2 AUC_0–14d was linearly correlated with the time-averaged plasma albumin concentration observed for individual mice. d NCA-calculated 8C2 clearance was linearly correlated with time-averaged plasma albumin concentration

Plasma Albumin Concentration

Plasma albumin concentration was measured 3 h, 7 days, and 14 days after 8C2 administration (Fig. 4). DSS-treated mice had lower plasma albumin concentrations compared with control mice (two-way repeated-measures ANOVA p = 0.0007), consistent with the common finding of hypoalbuminemia in PLE patients. Both 8C2 plasma AUC_0–14d and NCA-calculated, time-averaged clearance were linearly correlated with time-averaged plasma albumin concentration (coefficient of determination R² = 0.65 and 0.48, p = 0.003 and 0.02, respectively) (Fig. 3c, d). Lower plasma albumin concentration was associated with lower AUC_0–14d and higher 8C2 clearance. The correlations are in accordance with observations from several clinical mAb PK studies, where mAb clearance has been often found to be negatively correlated with serum albumin concentration (13,14,17,18). However, judging by the R² value, the correlation is weaker than that found between 8C2 clearance and time-averaged A1AT intestinal clearance.

Fig. 4.

Plasma albumin concentration was lower (p = 0.0007) in the dextran sodium sulfate (DSS) treatment group compared with the control group. Plasma albumin concentration was measured 3 h, 7 days, and 14 days after 8C2 administration. Data are represented as mean ± SD

Investigation of Other Factors that May Impact Antibody Pharmacokinetics

Neonatal Fc receptor (FcRn)-mediated IgG recycling is an important mechanism protecting IgG from catabolic degradation. High endogenous IgG concentrations will compete with exogenous IgG mAb (e.g., 8C2) for FcRn binding, resulting in less efficient FcRn-mediated mAb recycling and higher mAb plasma clearance (32). Due to this known competition between endogenous IgG and exogenous IgG for FcRn binding and transport (33), the impact of DSS treatment on endogenous IgG concentrations was assessed. Total murine IgG plasma concentration was measured 1 day prior to, and 3 h, 8 h, 7 days, and 14 days after the conclusion of DSS administration. No significant differences were found for total murine IgG plasma concentrations between the DSS treatment and control groups (two-way repeated-measures ANOVA p > 0.05) (data not shown).

Nephropathy has been found to increase the systemic clearance of mAb, and urinary albumin excretion rate (UAE) has been positively correlated with mAb plasma clearance in mouse models of diabetic nephropathy (34). UAE was measured in the present study using a 24-h urine sample collected immediately after 8C2 dosing. No significant difference in UAE was found from comparisons between the DSS treatment and control groups (unpaired Student’s t test p > 0.05) (data not shown).

FcRn expression will determine the capacity of FcRn-mediated IgG recycling, and may be expected to influence IgG systemic clearance (35). The expression of FcRn in selective tissues (liver, kidney, small intestine, and large intestine) was evaluated using tissue samples harvested 14 days after 8C2 administration. No significant differences were observed in FcRn tissue expression between the DSS treatment and control groups (unpaired Student’s t test p > 0.05) (Fig. 5).

Fig. 5.

FcRn expression in the liver, kidney, small intestine, and large intestine as measured by western blot. No significant differences (p > 0.05) were detected between values determined in the dextran sodium sulfate (DSS) and control groups

Population Pharmacokinetic Modeling

Conventional regression analysis only showed a correlation between the time-averaged values of A1AT clearance and 8C2 clearance. Due to the large magnitude of change in both parameters in each DSS-treated mouse over the study duration, correlation between time-averaged values may not capture the actual relationship between the two clearance parameters. Population PK modeling is an ideal method for this situation, as it allows consideration of A1AT clearance as a time-varying covariate of 8C2 clearance.

A two-compartment mammillary model with first-order drug elimination was selected as the base structural model (Fig. 2a, refer to the Materials and Methods section for model equations and parameterization). IIV was only modeled on CL and V_c. Parameter estimates for the base model are listed in Table I. Large IIV in 8C2 CL was found.

Table I.

Population Pharmacokinetic Modeling Parameter Estimates for the Base and Final Models

Parameter	Base model		Final model
	Estimate	RSE (%)	Estimate	RSE (%)
Fixed effect parameters
CL (ml/day/kg)	12.5	32	17.2	10
θA1AT	-	-	0.0527	9
Vc (ml/kg)	104	8	103	8
CLd (ml/day/kg)	54.2	21	55.5	21
Vp (ml/kg)	94.7	16	90.3	14
IIV parametersa
IIV on CL (%)	127	45	21.3	37
IIV on Vc (%)	22.3	28	22.4	28
IIV on CLd (%)	–	–	–	–
IIV on Vp (%)	–	–	–	–
RV parameters
Constant error (μg/ml)	6.96	27	3.26	39
Proportional error (%)	9.79	28	12.7	18
Objective function value	686.81		646.65

RSE relative standard error, IIV inter-individual variability, RV residual variability, CL clearance, V_c volume of distribution for the central compartment, V_p volume of distribution for the peripheral compartment, CL_d inter-compartmental distribution clearance, θ_A1AT covariate effect of A1AT clearance on 8C2 CL

^aIIV was calculated as $\text{\%IIV}=\sqrt{e^{{\omega }^2}-1}\times 100$, where ω was estimated directly from the modeling practice. The RSE of IIV parameters was calculated from the RSE of ω based on the law of propagation of error

The quantitative relationship between PLE severity and 8C2 CL was evaluated by performing covariate analyses using A1AT clearance as covariate for 8C2 CL. As described in the Materials and Methods section, A1AT clearance for each mouse at any modeled time point was interpolated from measured values assuming the value changes linearly with time between measured time points. Making A1AT clearance a time-varying covariate for 8C2 CL allows 8C2 CL to change over the study duration. Adding the covariate relationship significantly improved the model fitting, based on significantly lowered objective function value (p < 0.001) and improved diagnostic plots. Due to the time-varying nature of A1AT clearance, a plot between η_{i, CL} and individual A1AT clearance is not feasible. However, the covariate relationship did account for a large portion of the IIV in 8C2 CL; the unexplained IIV in 8C2 CL reduced from 127 to 21.3% after addition of the covariate relationship (Table I). Also, substantial improvement in the observed vs. population predicted 8C2 concentration plot was seen, suggesting the covariate relationship improved the prediction of typical population 8C2 PK behavior (Fig. 2c, d). A linear equation with A1AT clearance centered at its mean measured value best described the relationship.

$$\begin{gathered} {\text{CL}}_{\text{pop }}=17.2\times (1+0.0527\times (\text{ CL}\_\text{A1AT}-15.255)) \\ =3.37+0.906\times C{L}_{-}\text{A}1\text{AT} \end{gathered}$$

where CL_A1AT is the A1AT intestinal clearance (ml/day/kg), and CL_pop is the typical population-estimated 8C2 clearance (ml/day/kg). Other covariate equation forms were found to be inferior to the linear equation, based on higher objective function values, and, in some cases, poor precision of parameter estimates (Table II).

Table II.

Performance of Covariate Equation Forms

Model equation number	Objective function value
(5)	646.65
(6)	647.27
(7)	647.11a
(8)	653.55a
(9)	662.48a

^aAlso result in large relative standard error for some parameter estimates

DISCUSSION

Inter-individual variability in pharmacokinetics contributes to the uncertainty in dose-response relationships of therapeutic monoclonal antibodies. Variability in mAb PK is often studied through late phase clinical investigations, largely through the use of population-pharmacokinetic analyses that attempt to relate patient characteristics (e.g., age, weight, gender) to pharmacokinetic parameters. However, there may be unique opportunities to employ pre-clinical investigations to explore mechanistic hypotheses, and to identify mechanistic biomarkers that are predictive of pharmacokinetics in individual subjects. For example, our laboratory has recently employed a mouse model of diabetes to show a quantitative relationship between a mechanistic biomarker of diabetic nephropathy (urinary albumin excretion rate) and the systemic clearance of a model mAb (34). The present study represents another effort to employ an animal model to assess the impact of disease on mAb PK, and to explore possible mechanisms that may contribute to IIV in antibody disposition. Mice treated with DSS developed protein losing enteropathy, and were found to exhibit increased rates of plasma clearance of 8C2, a model monoclonal antibody. Fecal A1AT clearance, a biomarker for PLE, was found to be positively correlated with the plasma clearance of 8C2. This study demonstrates a direct relationship between PLE and increased plasma clearance of mAb, and demonstrates the quantitative utility of fecal A1AT clearance as a predictor of inter-individual variability in mAb pharmacokinetics.

PLE is not a primary disease; rather, it is a condition resulting from GI epithelial damage that arises through a variety of disease processes. In healthy subjects, the daily enteric loss of serum proteins only accounts for 1–2% of the serum protein pool (24). Due to the presence of “tight” intercellular junctions, the GI epithelial barrier tightly controls paracellular and transcellular transport of proteins (36,37). However, several diseases (such as IBD and GI cancer) can impair the intestinal epithelial barrier, by causing tight junction disruptions, epithelial cell apoptosis, epithelial lesions, or by increasing interstitial pressure. These effects, alone or in combination, are thought to lead to PLE (38). The most common laboratory findings of PLE are hypoproteinemia, including hypoalbuminemia and hypogammaglobulinemia, which then lead to peripheral edema (39). However, PLE may be present without apparent abnormalities in serum protein concentrations, possibly due to increased hepatic protein synthesis as a compensatory mechanism (5). The diagnosis of PLE is made by measuring the elimination rate of a specific protein into feces.

Currently, the most widely used diagnostic biomarker for PLE is the fecal clearance of A1AT (28,40). A1AT is a protease inhibitor synthesized in the liver, with a molecular weight of 57.7 kDa (41). It does not undergo proteolytic degradation in GI lumen, and it is not actively secreted or reabsorbed. Therefore, the amount of A1AT in the feces directly reflects the amount of A1AT that has “leaked” into the intestinal lumen (42), making it a reliable endogenous biomarker for loss of blood proteins via intestine.

Because of the wide spectrum of diseases that can cause PLE, its exact prevalence has not been carefully studied. Also, PLE may be overlooked clinically since hypoalbuminemia may be the result of other more common conditions such as malnutrition, impaired liver function, or proteinuria (43). Nonetheless, there is evidence demonstrating that the prevalence of PLE as a complication of primary diseases is considerable. For example, Karbach et al. studied 25 patients with Crohn’s disease, and found that A1AT intestinal clearance was higher in all patients than values found from a panel of ten healthy control subjects (30). Beeken et al. studied 30 patients with Crohn’s disease, and reported that 70% of the patients exhibited PLE (5). Among the 32 ulcerative colitis patients studied by Kapel et al., 81% had elevated intestinal protein leakage (11), and 18 of the 24 gastric cancer patients studied by Nakatani were found to exhibit substantial protein loss into the gastric lumen (6).

To our knowledge, only one mouse model has been reported as a model for the study of PLE. Bode et al. demonstrated increased albumin flux across intestinal epithelial cell monolayer in vitro and elevated fecal A1AT loss in mice with syndecan-1 (Sdc1, the predominant heparan sulfate proteoglycan expressed on intestinal epithelial cells) knockdown or knockout. Intestinal protein leakage can be further augmented in Sdc1^−/− mice by the i.v. administration of TNF-α (44). Attempts were made in our laboratory to validate the Sdc1^−/− mouse model by repeating the published protocol (44). No significant increase in A1AT fecal excretion was observed in Sdc1^−/− mice with TNF-α dosing (0.25 mg/kg i.v.) compared with wild-type littermates that did not receive TNF-α (n = 5 for both groups, unpaired Student’s t test p = 0.12). To the knowledge of the authors, no other report demonstrating PLE in this mouse model has been published to date. As indicated by Bode et al., PLE may result from a combination of genetic insufficiencies and environmental insults (29,44). The unsuccessful replication of the animal model in our laboratory may be the result of unknown environmental factors.

The DSS-induced mouse model of colitis is widely used to study human IBD (45). DSS is a water-soluble sulfated polysaccharide that is toxic to colonic epithelial cells (46). Mice treated with DSS-supplemented tap water for several days develop acute colitis that may progress to chronic colitis in Swiss-Webster (47,48) and C57BL/6 mouse (49) strains. Manifestations of DSS-treated mice include diarrhea, occult or gross intestinal bleeding, weight loss, and, in some cases, death (45). Disease activity and mortality rates are highly dependent on the concentration of DSS in drinking water and on the duration of treatment, both of which have to be optimized by individual labs. Histological features of DSS-induced colitis include colonic epithelial degeneration, neutrophil infiltration (50), and changes in the expression of tight junction proteins (51). Although PLE is a known complication of IBD in patients, and multiple pieces of evidence point to increased intestinal permeability (52,53) and hypoalbuminemia (54) in DSS-induced colitis, no direct observation of increased intestinal protein leakage has previously been reported for this mouse model. Prior to the current study, pilot studies were carried out in our laboratory to evaluate DSS dosing protocols and to assess relationships between DSS dosing and intestinal A1AT loss (data not shown).

The intestinal A1AT clearance vs. time profile of DSS-treated mice demonstrated that PLE severity changed with time during the period when 8C2 pharmacokinetics were evaluated. A1AT clearance peaked either 0 or 3 days after termination of DSS treatment, followed by a gradual decline over time. This is in line with a gradual recovery from colitis following termination of DSS dosing, during which epithelial regeneration slowly restores epithelial barrier function. The time dependent profile of A1AT clearance parallels the apparent transient increase of 8C2 plasma clearance exhibited by DSS-treated mice (i.e., where the plasma concentration data decreased rapidly with time in the first few days after dosing, followed by a slower decrease in concentration with time from day 4 through the final sample collection at day 14).

The severity of PLE in patients is associated with substantial inter-individual variation. For example, Florent et al. demonstrated a 15 ± 9-fold (mean ± SD, range 1 to 27) increase in intestinal A1AT clearance in 21 patients with IBD compared with the average value of normal subjects (40). Karbach et al. demonstrated a 17 ± 18-fold (mean ± SD, range 3 to 75) increase in intestinal A1AT clearance in 25 patients with Crohn’s disease compared with the average value of healthy subjects (30). Kapel et al. reported a median of fivefold (range 0 to 24) increase in 32 ulcerative colitis patients compared with 20 healthy controls (11). Nakatani reported a 7 ± 8-fold (mean ± SD, range 1 to 26) increase among 24 patients with gastric cancer compared with ten controls (6). In our study, the maximum measured A1AT clearance in each DSS mouse was 8 to 45-fold (18 ± 14, mean ± SD) that of the mean control group value. And the time-averaged individual A1AT clearance was 3 to 22-fold (9 ± 7, mean ± SD) that of the average control group value. The magnitude and variability of the rise in A1AT clearance following DSS treatment resemble the increases in A1AT clearance observed in patients with IBD or gastric cancer (compared with A1AT clearance in healthy subjects). Although there is likely some variability in the volume of DSS-supplemented drinking water consumed daily among mice, it has been reported that this variability is not correlated with severity of disease (55). Therefore, no measure was taken to ensure the uniformity of daily DSS water volume consumed.

Population modeling of 8C2 PK data revealed a correlation between A1AT intestinal clearance and 8C2 clearance. Including A1AT clearance as a covariate of 8C2 CL significantly improved model fit, and reduced the unexplained IIV in 8C2 CL. The time-varying nature of A1AT clearance was accounted for through linear interpolation of A1AT clearance between observations. Of note, the nonstationary nature of 8C2 pharmacokinetics complicated the development and interpretation of the fittings of the base structural PK model. When the base model, which assumes stationary pharmacokinetics, was fitted to the data, most of the IIV in 8C2 PK was attributed to the volume of the peripheral compartment (V_p), with large IIV in V_p (291%) and small IIV in CL (23.1%). Since the effect of DSS is mainly restricted to the intestines, there is no mechanistic reason why DSS treatment would impact the extent of 8C2 peripheral distribution. We hypothesized that the large IIV estimated in V_p was resultant from the lack of structure in the base model to represent a transient effect of DSS on 8C2 clearance. To improve fitting performance beyond the results of the base model, A1AT clearance was employed as a time-varying covariate of 8C2 CL. The addition of this covariate relationship greatly reduced the estimated IIV in V_p, supporting our speculation. We also explored A1AT covariate effects on CL_d or V_p. IIV on CL_d was not well estimated (negligible value and poor estimation precision). The addition of IIV on V_p did not improve beyond the final model. Adding A1AT clearance as a covariate of V_p did not improve model performance. Additionally, the estimated relationship between A1AT clearance and V_p is fairly flat, and it explained little of the IIV on V_p (IIV on V_p reduced from 291 to 251% after adding the covariate relationship). To experimentally assess DSS treatment impact on the extent of 8C2 peripheral distribution, future studies could be done by starting the 8C2 PK study at different time points after DSS dosing and to assess changes in 8C2 PK with respect to time after DSS dosing. Based on the model-estimated covariate relationship, an increase in A1AT clearance of 1 ml/day/kg will result in a 0.906 ml/day/kg increase in population-estimated 8C2 CL. The difference in magnitude of increase may relate to differences in the intestinal permeability of the two proteins, perhaps due to the difference in their molecular weight (57.7 kDa for A1AT vs. 150 kDa for 8C2).

Increased intestinal loss of therapeutic mAb has been associated with reduced treatment efficacy. In a clinical study conducted in 30 patients with severe ulcerative colitis that were just starting treatment with infliximab, an anti-TNF-α antibody, it was found that patients who did not respond to infliximab therapy had significantly higher infliximab concentration in feces (56). Also, low serum infliximab concentration has been identified to be predictive of poor treatment outcome (57,58). Although it is not clear if increased fecal loss of infliximab is reflective of disease severity which could also affect treatment response, it is speculated that low infliximab exposure as a consequence of high fecal loss may be linked to treatment failure. The current preclinical study demonstrates a profound impact of severe PLE on the systemic clearance and exposure of a model mAb, providing support to this hypothesis.

Plasma albumin concentration was significantly lower in DSS-treated mice compared with control mice, consistent with the hypoalbuminemia commonly observed in PLE patients. A reduction in total murine IgG plasma concentration after DSS treatment was not observed, which is also consistent with published results for the DSS model (59). Colonic infiltration of IgG-producing plasma cells is a histological hallmark of IBD patients, and this was also found in DSS-treated mice (60). Since the 8C2 PK study has shown a transient increase in murine IgG clearance after DSS treatment, it is possible that the lack of significant decrease in plasma murine IgG concentration is due to increased production of IgG by plasma cells infiltrating into the inflamed colonic mucosa.

FcRn is responsible for protecting both IgG and albumin from lysosomal degradation, thereby extending their serum half-life (61). Due to the half-life extension effect of FcRn on both IgG and albumin, others have hypothesized that the clinically observed negative correlation between serum albumin concentration and the clearance of several therapeutic mAb is caused by IIV in FcRn expression or function (20). Although the negative correlation was observed in our study, no significant differences in FcRn protein expression were found between the DSS-treated and control groups. FcRn expression is regulated by proinflammatory cytokines. TNF-α and IL-1β have been shown to upregulate (62), and IFN-γ to downregulate (63) FcRn protein expression and function in vitro. It has been reported that DSS treatment increases the colonic secretion of TNF-α, IL-1β (53,55), and IFN-γ (47). The lack of change in FcRn expression after DSS treatment may be due to the opposing effects of these cytokines on FcRn expression. Also, since FcRn expression was only measured on tissue samples collected 14 days after termination of DSS induction, it is possible that potential alterations in FcRn expression were resolved as animals recovered from DSS treatment. The expression of proinflammatory cytokines was reported not to be stationary during the progression and resolution of DSS-induced colitis (47,53); therefore, it is possible that FcRn expression decreased during the DSS induction and post-induction phases, but returned to baseline values prior to sample collection. This possibility will be explored in future investigations.

CONCLUSION

This study demonstrates that DSS-treated mice developed PLE, and that DSS-treated mice exhibit a transient increase in 8C2 mAb plasma clearance, and lower 8C2 plasma exposure. A1AT intestinal clearance, a biomarker for PLE, was positively correlated with 8C2 clearance, as found using either linear regression or population pharmacokinetic modeling. Clinical studies, perhaps including population PK studies assessing A1AT intestinal clearance as a covariate for mAb PK, are warranted to further validate our findings and, possibly, to allow improved prediction of mAb pharmacokinetics in PLE patients.