Whole-Body Pharmacokinetics of Antibody in Mice Determined using Enzyme-Linked Immunosorbent Assay and Derivation of Tissue Interstitial Concentrations

Abstract

Here we have reported whole-body disposition of wild-type IgG and FcRn non-binding IgG in mice, determined using Enzyme-Linked Immunosorbent Assay (ELISA). The disposition data generated using ELISA are compared with previously published biodistribution data generated using radiolabelled IgG. In addition, we introduce a novel concept of ABC_IS values, which are defined as percentage ratios of tissue interstitial and plasma AUC values. These values can help in predicting tissue interstitial concentrations of monoclonal antibodies (mAbs) based on the plasma concentrations. Tissue interstitial concentrations derived from our study are also compared with previously reported values measured using microdialysis or centrifugation method. Lastly, the new set of biodistribution data generated using ELISA are used to refine the PBPK model for mAbs.

INTRODUCTION

Monoclonal antibodies (mAbs) remain a rapidly growing class of drug molecules, with 6 mAbs approved by European Union or United States FDA in 2019, and 79 mAbs in late-stage clinical evaluation.¹ Since most mAbs act in the tissue interstitial space, it is important to accurately measure/predict the concentration of these molecules in tissues, to better understand their exposure-response relationships. Tissue distribution of mAbs is typically measured in preclinical species, and predicted in humans using allometric scaling.^2–7 or physiologically-based pharmacokinetic (PBPK) models.^8,9 However, most of the published whole-body biodistribution studies of mAbs are conducted using radio/fluorescent-labelled mAbs. and there are some issues when using labelled-mAbs for quantification. First, the labels may alter the PK behaviour of mAbs based on the degree of labelling¹⁰ via changes in their physicochemical properties like charge¹¹ or hydrophobicity.^12,13 Second, even if a label does not change mAb disposition, it may not remain conjugated to the mAb throughout the duration of experiment.¹² The label may dissociate from the mAb, and the free label can cause analytical bias.¹⁴ It is also important to match half-lives of mAbs and the label to make sure the radiolabel half-life is long enough to perform the experiment, but not too long to avoid excessive radioactive exposure.¹² Third, the disposition behaviour of a label after the mAb is catabolized should be considered. Radioactive isotopes can be divided into non-residualizing (e.g. radiohalogen such as iodine-125) and residualizing (e.g. radiometal-chelates such as indium-111, zirconium-89) isotopes. When a mAb is attached to a non-residualizing isotope, the isotope typically clears rapidly from the system following degradation of the mAb. However, residualizing isotopes remain in the body for a prolonged period of time after the mAb is degraded.¹⁵ As such, whole-body PK and tissue distribution measured using labelled mAbs can differ significantly compared to unlabelled mAbs.¹⁶ In addition, it is also reported that there are certain tissue-specific properties that impart additional error when conducting tissue distribution studies using labelled mAbs. For example, mAbs labelled with iodine usually have high signals in the skin because of the highly active sodium/iodide symporters in this tissue.¹² When using zirconium labelled mAbs, it has been observed that high signals are observed in bones because of the ‘bone-seeking’ properties of the chelated zirconium catabolites.¹⁷ Tissues like liver and kidney that are involved in the elimination of proteins,¹⁸ demonstrate a high degree of residualization due to the presence of transporters or metabolic enzymes, and solid tumours often demonstrate high degree of residualization due to the absence of functional lymphatic drainage.¹⁹ Moreover, residual blood contamination is a concern since mAb concentrations in blood are much higher than in tissues.¹⁴ Direct measurement of residual blood volume in each tissue using markers such as radiolabelled albumin²⁰ or labelled red blood cells²¹ is often performed using radiometric quantification since saline perfusion is not possible. However, radiolabelled albumin exhibits extravasation into interstitium spaces of tissues even at early time points. The transudation of the albumin into interstitium spaces results in 5% increase of the measured volume after 30 minutes.²² Similarly, free label that does not bind to red blood cell causes bias to the residual blood volume measurement.²³ Thus, use of labelled markers may not help in accurately accounting for blood contamination in tissues.

To overcome these issues associated with labelled mAbs, and to assess the true disposition characteristics of mAbs, here we have generated whole-body biodistribution data for an FcRn binding and non-binding mAb (trastuzumab, non-cross-reactive) in mice using a validated Enzyme-Linked Immunosorbent Assay (ELISA). Whole body perfusion was performed to avoid residual blood contamination. While it is resource intensive to develop and validate ELISA methods to quantify mAb concentrations in tissues, the PK data generated using this method provides accurate measurement of the PK behaviour of mAbs.²⁴

Most tissue distribution studies for mAbs report total tissue concentrations. However, total tissue concentrations of mAb may not represent the “effective” concentration at the target site-of-action (typically in the interstitial space of a tissue), which is required to predict drug efficacy and toxicity.²⁵ Consequently, it is important to assess interstitial concentrations of mAbs. However, to date, only a small number of studies have directly measured tissue interstitium concentrations of mAbs using tissue centrifugation or microdialysis methods,^25,26 because quantification of mAbs in tissue interstitium is challenging and not well-established. Consequently, here we have employed our tissue distribution data and knowledge of physiological parameters for tissues to infer tissue interstitium concentrations for mAbs. Since all of our biodistribution measurements were conduct following whole body perfusion, we were able to directly measure extravascular tissue concentrations of mAbs without residual blood contamination. After accounting for interstitial volume fractions of each tissue, we were able to infer tissue interstitial concentrations of mAbs, and compare them to the published values. In addition, we have also calculated the ratio of interstitium and plasma area under the curve (AUC) values to derive interstitial antibody biodistribution coefficient (ABC_IS) values, which can be used to infer therapeutically relevant exposure of mAbs in the tissues based on the plasma exposure.²⁷

Since most mAb PBPK models have been developed using biodistribution data generated using radiolabelled mAbs, we have also updated our previously published platform PBPK model⁸ using the biodistribution data for FcRn binding and non-binding mAb analysed by ELISA. We have also evaluated the PBPK model by comparing model-generated tissue interstitium PK data with the measured values from the literature.

MATERIALS AND METHODS

Development of ELISA method

Chemicals and reagents

IgG (trastuzumab, HERCEPTIN®) was obtained from a local pharmacy and FcRn nonbinding IgG (i.e. IHH mutation of trastuzumab) was produced in-house.²⁸ Goat anti-human IgG-F(ab’)₂ cross-adsorbed F(ab’)₂ and goat anti-human IgG-F(ab’)₂ cross-adsorbed F(ab’)₂ conjugated with alkaline phosphatase were obtained from Bethyl Laboratories. ELISA blocking buffer was dissolved in deionized water to make blocking solution (0.05 M Tris, 0.138 M NaCl, 0.0027 M KCl, 1% BSA, pH 8.0). PNPP, RIPA buffer and 1X Halt™ protease inhibitor were obtained from Thermo Scientific™.

Sample preparation

Separate calibration curves and quality controls (QCs) were prepared for plasma and individual tissues for IgG and FcRn nonbinding IgG. IgG and FcRn nonbinding IgG stock solutions were diluted in mouse plasma and 11 tissue homogenates (heart, lung, liver, spleen, kidney, pancreas, muscle, skin, fat, bone, and brain). Concentrations of standard curves ranged from 1–250 ng/mL, and the concentrations of QCs were 1, 50, and 200 ng/mL. Tissue matrices for standard curves and QCs were weighted and then RIPA buffer with 1X protease inhibitor was used for dilution. The optimized dilution factors for normal IgG were 300 for plasma, 100 for heart, liver, spleen, and skin; 50 for lung, kidney, and bone; 30 for muscle; and 20 for pancreas, fat, and brain. The dilution factors for FcRn non-binding IgG for higher concentration samples were 100 for skin; 50 for heart, liver, lung, spleen, kidney, fat, bone; 30 for muscle; 20 for pancreas and brain, and for lower concentration samples were 5 for all tissues. Dilution factors for standards, QCs, and samples were kept the same within each matrix. After dilution, tissue samples, standards, and QCs were incubated at 4°C for 2 h. The extended incubation of spiked-in drug in tissue matrices sufficiently equilibrated the protein-protein interaction and binding in tissue matrces.²⁹ Then all samples were centrifuged at 15000 g for 15 min at 4°C.

ELISA procedure

The sandwich ELISA protocol used to quantify concentrations of IgG or FcRn nonbinding IgG included the following steps: (a) coating the 384-well plate with capture antibody, (b) blocking the nonspecific binding sites on the plate, (c) adding samples, standards, and QCs to the plate, (d) adding detection antibody to the plate, and (e) adding the substrate. Between each step, plates were washed 3 times with 1% phosphate buffer saline (PBS)-Tween (0.05% Tween-20 in 1X PBS, no pH adjustment), followed by three washes with distilled water using AquaMax2000 (Molecular Devices, Sunnyvale, CA). Briefly, Nunc® Maxicorp™ 384-well plates were coated with 30 μL/well anti-human IgG- F(ab’)₂ antibody at a concentration of 5 μg/mL in 20 mM Na₂HPO₄ (no pH adjustment) and incubated at 4°C overnight. Plates were blocked with 90 μL/well of blocking solution at room temperature for 1 h on a plate shaker. Thirty microliters per well of samples, standards, and the QCs were loaded in triplicates and incubated for 2 h at room temperature on a plate shaker. Thirty microliters per well of the 1.4 ng/μL of goat anti-human IgG-F(ab’)₂ cross-adsorbed F(ab’)₂ conjugated with alkaline phosphatase in washing buffer was used as the secondary antibody and incubated at room temperature for 1 h. Thirty microliters per well of p-nitrophenyl phosphate solution (1 mg/mL in 1X diethanolamine substrate buffer) was used as the colouring agent. The change in absorbance was measured over time (dA/dt) at 405 nm for 45 min using Filter Max F5 microplate analyzer (Molecular devices, Sunnyvale, CA). All standard curves were fitted using 4- parameter logistic equation.

Precision, accuracy, and limit of quantification

The accuracy of the assay was assessed by low, medium, and high QCs, and relative errors (%RE = [measured concentration - nominal concentration]/nominal concentration) were calculated. Intraday precision (CV%) was assessed on two occasions on the same day and interday precision on separate days. The lowest concentration on the standard curve was defined as lower limit of quantification (LLOQ) when CV% and %RE were within 25%. The highest concentration on the standard curve was defined as upper limit of quantification (ULOQ) when CV% and %RE were within 20%.

Biodistribution studies of IgG and FcRn nonbinding IgG in nude mice

Animals

All in vivo studies were approved by the Institutional Animal Care and Use Committee of the State University of New York at Buffalo (IACUC # PHC29035Y). Male athymic nude mice (~6 weeks old, 26–37 g) (Charles River, USA) bearing MDA-MB-468 tumor xenografts (197–303 mm³) were used for the biodistribution studies. Of note, the antibodies used for biodistribution do not bind to a target on the tumor cells, and hence the solid tumor behaves just like an additional tissue.

Biodistribution studies

10 mg/kg of IgG or FcRn nonbinding IgG were injected in the mice via the penile vein, and terminal samples were collected at 6, 24, 72, and 168 h. Three mice were sacrificed at each time point. Blood samples were collected in EDTA pre-coated tubes and centrifuged at 2000 g for 20 min at 4°C. Plasma was collected and stored at −20°C for further analysis. Before collecting tissue samples whole-body perfusion was performed. For the perfusion of animals, 8–10 mL of PBS was injected from the apex of the left ventricular at the rate of 10–20 mL/min. Perfusion was completed when liver was blanched to a light tan color. Heart, liver, lung, spleen, pancreas, kidney, skin, bone, muscle, fat, brain, and tumor were collected. Harvested tissue samples were blotted dry and immediately frozen in liquid nitrogen and stored at −80 °C until homogenization.

Tissue homogenization

Tissue samples were weighed and RIPA buffer containing 1X Halt™ protease inhibitor was added. Based on the property of each tissue and the capability of being homogenized, different volumes of RIPA were added. Each homogenization tube contained 7 zirconium beads (3.0 mm, Benchmark, USA) for heart, liver, lung, spleen, pancreas, kidney, fat, brain, and tumor; or stainless steel beads (2.8 mm) for skin, bone, and muscle. Then, tubes were homogenized using BeadBug™ microtube homogenizer (Benchmark, USA) for 15 sec followed by 30 sec ice cool down, and repeated three times. Tissue homogenates were further diluted with RIPA buffer before analysis. Before loading the plates, tissue samples were incubated on ice for 2h for equilibrium. Then samples were centrifuged at 15000 g for 15 min at 4°C and supernatant was collected.

Analysis of pharmacokinetic data

Total tissue concentration and interstitial concentration of IgG in each tissue were back-calculated from the measured concentration in perfused tissues. Physiological values of the volumes of tissue sub-compartments were obtained from the literature, which assumed 28 g as the average weight of mice.⁸ Total tissue concentration $\left(C_i^{\mathit{\text{total}}}\right)$ was calculated using equation 1, in which residual blood content is included.

$$C_i^{\mathit{\text{total}}}=\frac{C_{\mathit{\text{plasma}}}\cdot V_i^V+C_{\mathit{\text{observed}}}\cdot \left(V_i^V+V_i^E+V_i^{BC}+V_i^{IS}+V_i^C\right)}{\left(V_i^V+V_i^E+V_i^{BC}+V_i^{IS}+V_i^C\right)}$$ (1)

Interstitium concentration $\left(C_i^{IS}\right)$ was calculated using equation 2.

$$C_i^{IS}=\frac{C_{\mathit{\text{observed}}}\cdot \left(V_i^V+V_i^E+V_i^{BC}+V_i^{IS}+V_i^C\right)}{V_i^{IS}}$$ (2)

Above, $V_i^V$, $V_i^E$, $V_i^{BC}$, $V_i^{IS}$, and $V_i^C$ are volumes of vascular, endosomal, blood cell, interstitium and cellular compartments for tissue “i”. C_observed is measured concentration via ELISA.

Area under the concentration curve (AUC_0-t) were calculated using noncompartmental analysis (NCA) with sparse sampling method in Pheonix WinNonlin version 8.1. Three sets of tissue-to-plasma AUC_0-t ratios (T/P) were calculated using measured tissue (measured T/P), total tissue (total T/P), and tissue interstitial concentrations of IgG and FcRn non-binding IgG. Antibody biodistribution coefficient (ABC) has been reported by us before,²⁷ and here we have introduced ABC_IS, which denotes tissue interstitium to plasma AUC_0-t ratio. Plasma and total tissue concentration profile of IgG and non-binding IgG were compared with published PK data generated using radiolabelled IgG by Garg et al.³⁰ and Chen et al.³¹ Total T/P were compared to ABC,²⁷ T/P reported by Garg et al.,³⁰ and T/P calculated using data reported by Chen et al.³¹ and Yip et al. data.³²

Data characterization using the refined PBPK model

The IgG and FcRn nonbinding IgG PK in plasma and tissues were simultaneously fitted to the previously published PBPK model.⁸ The rates of pinocytosis/exocytosis per unit endosomal space of vascular endothelial (CL_up), first-order degradation rate constant for FcRn unbound mAb within the endosomal space (K_deg), and vascular reflection coefficient (σ^V) for each tissue, which represents the level of resistance provided to the mAb convection by the vascular endothelial cells, were estimated by fitting the data to PBPK model. All of the model equations, physiological parameters, and other model parameters were kept the same as the published model. The model was fitted to the data using the weighted least square method in ADAPT-5 software (BMSR, CA) with the variance model as:

$$\text{Var}(\text{t})={\left({\sigma }_{\text{intercept}}+{\sigma }_{\text{slope}}\cdot \text{Y}(\text{t})\right)}^2$$

Above, Y(t) is the model output at a given time t, Var(t) is the variance associated with the output, and σ_intercept and σ_slope are the variance parameters representing a linear relationship between the standard deviation of the model output and Y(t). Final model performance was evaluated based on visual inspection, observed versus predicted plots, akaike information criterion, bayesian information criterion, and CV% of the estimated parameters.

RESULTS

Precision, accuracy, and assay range for ELISA

Representative standard curves to measure FcRn binding and non-binding IgG (trastuzumab) in plasma and all 11 tissues of mice are provided in Supplementary Figure 1. The intra- and inter-day precision and accuracy values for each calibration standards and quality control (QC) samples, for plasma and all tissues, were within ±20% of the nominal value. The total error was below 30% for all calibration standards and QCs (Supplementary Table 1 to 11). The range of antibody quantitation was 1–500 ng/mL for plasma; 1–250 ng/mL for heart, liver, lung, and spleen; 2–250 ng/mL for fat, pancreas, and skin matrices; and 4–250 ng/mL for bone, brain, kidney, and muscle.

Biodistribution of IgG and FcRn nonbinding IgG in nude mice

Pharmacokinetic profiles in plasma and tissues

Measured PK profiles of IgG and FcRn nonbinding IgG in plasma and tissues of nude mice obtained following whole-body perfusion are provided in Figure 1. In order to assess the role of residual blood in each tissue, we further calculated total tissue concentrations in each tissue by adding the residual blood content, and superimposed over the measured PK data (Figure 1). Interstitium concentrations of IgG in each tissue were also calculated, using the reported values of interstitial volume and extracellular volume for each tissue, and superimposed over the measured PK data (Figure 1). As expected, prolonged exposure of FcRn binding IgG were observed in plasma and tissues. In contrast, concentrations of FcRn non-binding IgG dropped rapidly, and were below the limit of quantification (BLQ) in both plasma and tissues at 168 hr. The clearance for IgG and FcRn non-binding IgG were 0.237 mL/h/kg and 6.01 mL/h/kg. The half-life for IgG and FcRn non-binding IgG were 206 h and 9.87 h. PK profiles in tumor without target expression showed that FcRn non-binding IgG had 20-fold shorter half-life than normal IgG (Figure 1).

Figure 1. Observed, total, and interstitial concentrations of IgG and FcRn non-binding IgG in plasma and 11 tissues.

Measured tissue concentrations of IgG (black circle) and FcRn non-binding IgG (black square) are presented. After accounting for residual blood content, total tissue concentration of IgG (grey circle) and FcRn non-binding IgG (grey square) were calculated. Tissue interstitial concentration of IgG (white circle) and FcRn non-binding IgG (white square) were calculated from measured tissue concentrations. PK profiles of IgG and FcRn non-binding IgG in non-target expressing tumor, MDA-MB-468, are also presented.

After including residual blood content, concentrations increased significantly in many tissues like heart, liver, lung, spleen, kidney, pancreas, and brain. As expected, total concentration increased less in poor perfused tissue like skin, fat, and bone. The effect of residual blood on IgG concentrations was greater than FcRn non-binding IgG concentrations. Interestingly, the calculated interstitium concentrations were higher than measured tissue homogenate concentrations in many tissues such as heart, spleen, kidney, pancreas, fat, skin, muscle, and bone (Figure 1).

Comparison with other published mAb biodistribution studies

Plasma and total tissue concentrations of FcRn binding and nonbinding IgG, measured using ELISA, were compared with similar published PK data generated using radiolabelled IgG by Garg et al.³⁰ and Chen et al.³¹ (Figure 2). Both of these studies reported biodistribution of IgG in C57BL/6 control mice and FcRn-knockout mice (KO). Since the Chen et al. study only reported plasma PK and tissue-to-blood ratios at different time-points, tissue concentrations were back-calculated for comparison. All PK profiles were normalized to the dose of 10 mg/kg. Garg et al. reported plasma and tissues PK profiles generated using iodine-125 labelled IgG1, whereas Chen et al. reported plasma PK determined using ELISA and iodine-125 activity, and tissue PK determined using iodine-125 activity. In general, it was found that plasma and tissue PK of IgG in normal mice was similar between our study and two other published studies. While the data from Chen et al. was more variable, and there were subtle differences between the superimposed PK profiles, there was no obvious trend that can differentiate these profiles. However, it was found that tissue PK of FcRn non-binding IgG in wild-type mice observed by us demonstrated faster clearance compared to PK of IgG in FcRn KO mice observed by Garg et al. and Chen et al. (Figure 2). In most of the tissues, the observed PK by Garg et al. and Chen et al. were very similar, except lung where the PK by Garg et al. was similar to our study. Interestingly, plasma PK of FcRn non-binding IgG observed by us was similar to the PK of IgG in FcRn KO mice observed by Chen et al., and these two profiles demonstrated 2-fold higher clearance (6 vs. 3 mL/h/kg) compared to plasma PK of IgG in FcRn KO mice observed by Garg et al. The deviation of biodistribution profiles between studies was only found in FcRn non-binding IgG/IgG in FcRn KO mice but not in normal IgG, indicating the impact of using different analytical methods may be more pronounced when the elimination of parent molecule is more rapid.

Figure 2. Comparison of antibody biodistribution in different studies.

Total tissue concentrations of IgG (black circle) are compared with tissue concentrations of I-125 labelled IgG in wild-type mice from Garg et al. (grey circle) and Chen et al. (white circle) studies. Total tissue concentrations of FcRn non-binding IgG (black square) are compared with tissue concentrations of I-125 labelled IgG in FcRn knockout mice from Garg et al (grey square) and Chen et al. (white square) studies. The reported concentrations in brain from Garg et al. and Chen et al. were corrected with residual blood content, and thus measured tissue concentration of IgG (black circle) and FcRn non-binding IgG (black square) are presented for comparison.

Tissue to plasma AUC ratios

Table 1 shows the tissue to plasma AUC ratios (T/P) calculated using measured (measured T/P) and total (total T/P) concentrations of IgG and FcRn nonbinding IgG. The T/P of non-target expression tumor were very similar between IgG (6.34%) and FcRn non-binding IgG (6.37%), indicating the distribution of IgG in tumor was not highly dependent on FcRn related processes. Values of T/P reported by Garg et al. are also provided.³⁰ The T/P of Chen et al. were calculated based on the PK profiles.³¹ T/P were also calculated from Yip et al. study, which reported AUC_0–7 for wild-type IgG and the IgG variant (H310Q, FcRn⁻) measured using radiolabelled IgG.³² In addition, antibody biodistribution coefficient (ABC) values reported by us before are also listed in the table for comparison.²⁷ When T/P for wild-type IgG were calculated without residual blood effect, for most of the tissues this ratio was below 2%, except for higher ratios in skin (7.39%) and spleen (4.28%), and lower ratio in brain (0.197%). After accounting for the residual blood content (i.e. total tissue concentration), T/P increased by more than 3-fold in many tissues including heart, liver, lung, spleen, kidney, pancreas, and brain. Corresponding to PK profiles, T/P increased by less than 3-fold in poorly perfused tissues skin, fat, and bone.

Table 1.

Tissue to plasma AUC ratios of observed, total, and interstitial concentration in plasma and tissues

Tissue	Wild-type IgG ratio (%)						FcRn knockout or FcRn non-binding IgG ratio (%)					ABCIS (%)a
	ELISA measuredb	ELISA Totalc	I-125 Garg)d	I-125 (Chen)e	I-125 (Yip)f	ABC (Shah)g	ELISA measuredb	ELISA Totalc	I-125 (Garg)d	I-125 (Chen)e	I-125 (Yip)d	IgG	FcRn non-binding IgG	Microdialysis (Jadhav)h	Centrifuged (Eigenmann)i
Bone	2.00	4.24		11.1		7.27	3.02	4.65		10.1		10.8	16.2
Brainj	0.197	2.33	0.22	1.06	0.95	0.351	0.124	1.72	0.21	1.15	1.35	1.09	0.686
Fat	1.63	2.69		9.18		4.78	3.78	4.58		4.50		9.56	22.2
Heart	0.914	4.64	13.4	25.8	9.45	10.2	2.11	4.94	12.3	30.0	10.3	6.40	14.8
Kidney	1.70	7.02	13.0	28.7	9.39	13.7	2.01	5.97	15.7	34.4	12.7	11.3	13.4	3.77
Liver	1.17	9.40	14.0	20.9	9.45	12.1	3.13	9.37	14.3	47.4	8.75	5.88	15.7	5.39
Lung	1.72	15.7	17.1	9.62	13.1	14.9	2.18	12.8	15.7	13.4	15.4	9.15	11.6
Muscle	1.62	3.76	3.76	6.55	3.31	3.97	0.901	2.52	2.87	5.25	3.53	12.4	6.93	2.05	46.5
Pancreas	1.29	6.62		8.26		6.40	1.22	5.33		8.96		7.41	7.01
Skin	7.39	11.0	13.0	20.7	13.1	15.7	2.81	5.56	6.08	16.9	13.7	22.3	8.50	4.95	43.1
Spleen	4.28	16.0	12.0	22.0	8.90	12.8	8.34	17.2	13.8	35.5	10.7	21.4	41.7
GI			5.12		13.1	15.2			5.73		17.1

^aAUC_0-t calculated by using calculated tissue interstitial concentrations

^bAUC_0-t in tissue/AUC_0-t in plasma*100%, calculated using measured tissue concentrations (without residual blood contamination) generated by using ELISA

^cAUC_0-t in tissue/AUC_0-t in plasma*100%, calculated using total concentrations (accounting for residual blood content) generated by using ELISA

^dValues of AUC ratios obtained from Garg et al. study

^eAUC_0-t calculated by using PK data reported from Chen et al.

^fCalculated by using AUC_0–7h reported from Yip et al.

^gABC values obtained from Shah et al.

^hValues of AUC ratios obtained from Jadhav et al.

ⁱAUC_0-t calculated by using PK data reported from Eigenmann et al.

^jThe measured concentrations from Garg et al. and Chen et al.were corrected for residual blood contents. Abuquayyas et al. reported tissue-to-plasma AUC ratios were 0.774% in WT and 0.907% in KO mice.

Since ABC values were established using biodistribution data generated using radiolabelled antibodies, from mostly non-perfused animals, it is more appropriate compare it to AUC ratios calculated after accounting for the residual blood content (total T/P). The difference between ABC values and total T/P were within two-fold for all the tissues except brain, as two values were almost identical in lung, pancreas, and muscle. However, data generated using radiolabelled antibodies in the brain were reported with residual blood correction; therefore, it is more appropriate to compare ABC values to AUC ratios calculated using measured tissue concentrations. The difference between ABC values and measured T/P were within two-fold for brain. Total T/P for normal IgG and FcRn non-binding IgG obtained by us were comparable with Garg et al. and Yip et al. The difference in total T/P values between ours and the two studies were within 2-fold in most of the tissues, except for brain, heart, kidney, and skin. However, T/P differences were within 3-fold range in heart, kidney, and skin. In fact, brain data reported by Garg. et al. and Chen et al. were with residual blood correction, and the T/P from Garg et al., but not Chen et al. was similar to the measured T/P in our study. The T/P in our study, Garg et al., and Yip et al. studies were generally lower than the ratios observed by Chen et al. We suspect that higher T/P values observed by Chen et al. may stem from the unusually higher concentration values observed by the authors at the last time point (Figure 2).

To investigate the effect of FcRn on mAbs exposure in tissues, we calculated the percentage changes in T/P for FcRn non-binding IgG compared to the normal IgG for each tissue (Figure 3). The T/P was significantly higher in liver (168%), fat (132%), heart (131%), spleen (95%), and bone (51%), and significantly lower in skin (62%). However, after accounting for residual blood, the difference became less in all tissues (within 2-fold). For example, when comparing total T/P of FcRn non-binding IgG with normal IgG, a minimal difference was observed in liver (0.319%) and heart (6.47%), spleen (7.50%), and bone (9.67%). The percentage change in T/P values for IgG in FcRn KO mice compared to WT mice were also calculated for Garg et al., Chen et al., and Yip et al. data. Our data suggested significantly higher T/P for FcRn non-binding IgG in fat (132%), while Chen et al. reported significantly lower in FcRn KO mice (102%). Garg et al. and Yip et al. did not report fat concentrations. On the other hand, measured T/P of FcRn non-binding IgG were lower than that of normal IgG in muscle (−44.4%), skin (−62.0%), and brain (−37.1%). The results agreed with Garg et al. and Chen et al., which showed the T/P of IgG in KO mice were lower than that of IgG in WT mouse in muscle, skin, and brain. Thus, the effect of FcRn on tissue distribution differs between each tissue, and the effect of FcRn may be overlooked if residual blood content is not corrected while reporting tissue exposures. Of note, although the extent of FcRn effect on T/P varied between different studies, the pattern (i.e. increase or decrease in T/P) was the same.

Figure 3. Percentage change of tissue-to-plasma area under the curve ratios (T/P) between IgG and FcRn non-binding IgG.

Area under the curve (AUC_0-t) were AUC_{0–168 h} and AUC_{0–72 h} for IgG and FcRn non-binding IgG.

PBPK model

Biodistribution data for FcRn binding and non-binding IgG generated using ELISA were simultaneously fitted using the previously published platform PBPK model.⁸ Brain data were not included in the model fitting since brain has different tissue sub-compartment and requires more dedicated model structure,³³ and the parameters for brain tissue were kept the same as the previous PBPK model. The previous PBPK model estimated four parameters which were not available from the literature, including FcRn concentration (FcRn) in the endosomal space; CL_up, K_deg, and the proportionally constant (C_LNLF) between the rate at which antibody transfers from the lymph node compartment to the plasma/blood compartment and the plasma flow of the given species. Since C_LNLF was a non-sensitive parameter, we kept it the same as the reported value. In the previous model, the vascular reflection coefficient in each tissue was set to an assigned constant based on the literature review of physiological upper limits of pore size for different vascular capillary types.³⁴ Therefore in the refined model we also estimated the vascular reflection coefficient in each tissue. Initially, we tried to estimate CL_up, K_deg, FcRn, and ${\sigma }_i^v$ simultaneously, and the determined FcRn concentration was similar to the previous PBPK model (data not shown). We thus fixed FcRn concentrations to the value obtained from published model, which allowed parameter estimations to be more precise. The refined PBPK model better characterized the PK of IgG and FcRn nonbinding IgG in plasma and tissues simultaneously. Figure 4 shows the observed and model predicted PK profiles in plasma and tissues. Simulated PK profiles using our previously published PBPK model are also reported. The estimated model parameters are provided in Table 2. The estimated CL_up, which characterized the rate of pinocytosis/exocytosis increased from 0.0366 to 1.22 L/h/L. This is the rate-limited step for the elimination of FcRn non-binding IgG in plasma and tissues, and thus the refined CL_up value allowed the model to capture the steeper slope of FcRn non-binding IgG. On the other hand, K_deg, which describe the degradation rate of unbound mAb within the endosomal space, decreased from 42.9 to 15.3 L/h, which optimized the model prediction for IgG PK. The vascular reflection coefficient $\left({\sigma }_i^v\right)$ for each tissue was estimated with good precision (CV% less than 2% for all tissues) and the optimized value was similar to the values proposed in the previous PBPK model.

Figure 4. Comparison of the PBPK model-predicted vs. measured concentrations of IgG and FcRn non-binding IgG.

Comparison of the refined platform PBPK model predicted IgG (black solid line) and FcRn non-binding IgG (black dash line) with measured concentrations of IgG (circle) and FcRn non-binding IgG (square). The refined platform PBPK model better characterized the observed data than PK profiles of IgG (grey dash line) and FcRn non-binding IgG (grey dot line) simulated by previously published platform PBPK model.

Table 2.

Parameter estimated using the platform PBPK model

Parameter	Published values	Model estimated (CV%)
CLup (L/hr/L)	0.0366	1.22 (1.70)
Kdeg (1/hr)	42.9	15.3 (5.32)
${\sigma }_{\mathit{\text{heart}}}^V$	0.95	0.963 (0.614)
${\sigma }_{\mathit{\text{kidney}}}^V$	0.90	0.875 (1.31)
${\sigma }_{\mathit{\text{muscle}}}^V$	0.95	0.968 (0.215)
${\sigma }_{\mathit{\text{skin}}}^V$	0.85	0.785 (1.42)
${\sigma }_{\mathit{\text{adipose}}}^V$	0.99	0.945 (1.18)
${\sigma }_{\mathit{\text{spleen}}}^V$	0.85	0.831 (2.00)
${\sigma }_{\mathit{\text{pancreas}}}^V$	0.95	0.958 (0.198)
${\sigma }_{\mathit{\text{liver}}}^V$	0.95	0.957 (0.489)
${\sigma }_{\mathit{\text{bone}}}^V$	0.95	0.946 (0.826)
${\sigma }_{\mathit{\text{lung}}}^V$	0.93	0.911 (0.360)

CLup, the rates of pinocytosis and exocytosis per unit endosomal space; K_deg, first order degradation rate constant of FcRn unbound mAb within the endosomal space; ${\sigma }_{\mathit{\text{pancreas}}}^V$, vascular reflection coefficient of “i” tissue.

Tissue Interstitial concentrations

Interstitium concentrations for WT and FcRn non-binding IgG were calculated from measured concentration in each tissue. Interstitium to plasma AUC ratios, denoted as ABC_IS, were also calculated (Table 1). The ABC_IS values were 5- to 8-times higher than tissue to plasma AUC ratios calculated without residual blood, indicating that most mAb in a perfused tissue is distributed within the interstitium space. ABC_IS values for kidney, liver, muscle, and skin were also compared with the values obtained from Derendorf et al. and Eigenmann et al., who quantified interstitium concentrations of mAbs using microdialysis and tissue centrifugation methods. ABC_IS values calculated using our observed data were found to be in between the values calculated using the published data. In addition, interstitium PK profiles for kidney, liver, muscle, and skin, obtained from the present study, reported in the literature, and predicted using the updated PBPK model are compared in Figure 5. For muscle and skin, microdialysis experiments reported lowest interstitium concentrations and tissue centrifugation experiments reported the highest concentrations. Our calculated interstitium concentrations fell in between the concentrations measured by the two methods. For kidney, the calculated interstitium concentrations were higher than the values measured by microdialysis, while for liver, the calculated interstitium concentrations were comparable to values measured by microdialysis. The PBPK model predicted interstitium tissue concentrations were found to superimpose on the interstitium concentrations calculated from our observed data.

Figure 5. Comparison of interstitial tissue concentration of IgG obtained from different studies.

Tissue interstitial concentration of IgG (black circle) calculated from our measured tissue IgG concentrations were compared with Jadhav et al. study (black square), which directly measured interstitial concentrations of IgG by using microdialysis,²⁶ and Eigenmann et al. study (white circle), which used tissue centrifugation method.²⁵ Tissue interstitial concentrations of IgG (dashed line) predicted by the refined platform PBPK model are also presented.⁸

DISCUSSION

Investigation of mAbs concentration in tissues is important because it helps us in better understanding the target tissue penetration/distribution and off-target toxicity. Commonly used bioanalytical methods to quantify mAbs includes ELISA, direct labelling of radioisotope or fluorescent dye, and LC-MS technique. Currently, there are no studies using ELISA to quantify whole-body PK of antibodies in vivo. Knowing that each method has its advantages and disadvantages,¹⁵ an orthogonal method is needed for mAb tissue quantification. Here we have presented the first-ever whole-body biodistribution study of mAb in mice, where ELISA was used to quantify mAb concentrations in the plasma and 11 tissues. For all the tissues, intra- and inter-day precisions and accuracy for the ELISA method were within the acceptance criteria reported by FDA under the ligand binding assays guidance.³⁵ Our ELISA method was able to quantify mAb concentrations from 1–250 ng/mL. Sample pre-treatment such as tissue homogenization and dilution factors were optimized for each tissue, where the type of homogenization beads and volume of tissue sample per homogenization run were determined based on the toughness of each tissue. Different dilution factors were selected for each tissue to allow the measured concentrations to lie within the standard curve ranges, and if possible within linear range of the standard curve. To avoid any matrix effect, standard curve for each tissue was developed by using corresponding blank tissue matrix and same dilution factor as the samples.

The validated ELISA method was used to measure in vivo biodistribution of normal IgG and FcRn non-binding IgG in nude mice. As expected, the clearance of FcRn non-binding IgG was 25 times higher than that of IgG. Since whole body saline perfusion was performed before collection of the tissues, we directly measured tissue concentrations without residual blood contamination. This was different from biodistribution studies using radioisotope for quantification, where measurement of tissue concentrations contained residual blood content and correction was further performed. Previous correction methods used directly measured blood concentrations in tissues using markers (labelled albumin, haemoglobin, red blood cell), or applied established tissue to plasma concentration ratio in specific tissue.¹⁴ Here, theoretical residual blood contents were added to the measured tissue concentrations and we “back-calculated” the total tissue concentrations, which enabled us to investigate the effect of residual blood content on tissue PK of mAbs. Total tissue exposure increased by more than 3-fold when compared to measured tissue exposure, which indicated prominent residual blood effect on mAbs tissue concentrations. This is expected since mAb concentration in plasma is much higher than that in tissues, and this is often not seen in small molecule drugs.³⁶

Plasma and total tissue concentrations of IgG and FcRn non-binding IgG generated using ELISA were compared with published PK data for IgG in wild-type and FcRn KO mice, generated using radiolabelled IgG.^30,31 Although our study used athymic nude mice and the other two studies used C57BL/6 mice, it has been reported that the PK of antibodies is comparable between these two strains of mice.³⁷ The PK profiles of normal IgG were comparable when measured by ELISA or radiometric methods. In terms of FcRn non-binding IgG/IgG in FcRn KO mice, PK profiles in plasma from our study were similar to Chen et al. study but not in tissues, and the reason needs further investigation. Both plasma and tissue PK profiles measured by us demonstrated rapid elimination when compared to IgG PK profiles reported by Garg et al. The faster elimination of FcRn non-binding IgG observed in our study may stem from several causes. First, PK of FcRn non-binding IgG in mouse could be different from the PK of normal IgG in FcRn KO mouse. An incomplete knockout may show slower drug elimination due to FcRn recycling process, resulting in different PK profiles compared to FcRn nonbinding IgG.³⁸ However, the similarity between our and Chen et al.’s plasma PK data do not support this hypothesis. Second, it is possible that iodine containing IgG catabolites or active sodium/iodide symporters in tissues such as skin could result in higher observed signals.¹² These effects will result in higher measured tissue concentrations of IgG when measured via radiolabelling method,³² especially for IgG with faster elimination, as the radiolabelled catabolites can demonstrate comparable half-life. Lastly, the immunodeficiency status of the mouse could also contribute to different biodistribution of IgG.³⁹ Our study used immunocompromised mice while Garg et al. and Chen et al. used immunocompetent mouse. We found that spleen concentrations of normal IgG in our study were higher than the ones reported by other two studies. Similarly, a previous study reported that immunodeficient strains of mice can display higher concentrations of IgG in tissues such as spleen,⁴⁰ due to lower levels of endogenous IgG and increased interaction with cells expressing Fc gamma receptors. Overall, the deviation between the PK profiles of FcRn non-binding IgG obtained using different analytical methods emphasized the effect of labeled catabolites on the PK of rapidly eliminating molecules. Therefore, we suggested using ELISA when quantifying the PK of mAbs with fast clearance or mAbs with tendency to demonstrate target mediated elimination in tissues.

To assess tissue disposition of IgG and FcRn non-binding IgG, we calculated measured T/P which used tissue concentrations without residual blood content and total T/P which used tissue concentrations after adding residual blood content. The measured T/P can help infer tissue mAb concentration without residual blood contamination. The total T/P from our study and ABC values were within 2-fold in all tissues except brain, and the measured T/P were within 2-fold of ABC values in brain.²⁷ In addition, the difference between total T/P in our study and T/P from other radiolabelled studies were within 3-fold in all tissues. While we have established T/P for normal tissues, tumor T/P remains difficult to be generalized.⁴¹ The use of immunodeficient mice bearing MDA-MB-468 tumor xenografts allowed us to calculate T/P in non-target expressed tumors. The T/P in tumor tissue was found to be 6.3 % for both IgG and FcRn non-binding IgG. Furthermore, the calculated tumor interstitium to plasma AUC ratio was around 12–24% based on reported interstitium volume (0.3 – 0.6 mL·g⁻¹),⁴² which is within two-fold from the reported tumor interstitium to plasma ratio of ~30% in target-expressing tumor at target saturating doses.⁴³ The effect of FcRn on mAb tissue distribution were significant (> 2-fold) in liver, fat, and heart, when comparing percentage changes between measured T/P for IgG and FcRn non-binding IgG. After accounting for the residual blood, the effect of FcRn became insignificant in all tissues (< 2-fold). The similarity of T/P in non-target expression tumor between IgG and FcRn non-binding IgG indicated FcRn did not contribute to tumor distribution. This is supported by the evidence that most tumor cell lines have very low to undetectable levels of FcRn.⁴⁴ The measured T/P of FcRn non-binding IgG/IgG in FcRn KO mice were more than 2-fold lower than the measured T/P of normal IgG in muscle and skin, in ours, Garg et al, and Chen et al. studies. This was supported by the evidence that skin and muscle as the major organs for IgG catabolism, and thus without FcRn protection, faster elimination of mAbs resulted in decreased tissue concentratons.^30,45

For brain, the T/P of FcRn non-binding IgG were 40% lower than the T/P of IgG. This suggested that FcRn may facilitate influx of mAbs across blood-brain barrier,^33,46 which remains to be evaluated. In fact, Yip et al.³² showed higher brain uptake at 168 h for IgG with increased FcRn binding affinity determined using I-125 labelled and In-111 labelled mAbs, indicating FcRn may play a role on IgG influx into brain. However, this increased uptake was not observed at 6 h and 24 h time point. The role of FcRn on IgG transport in brain is still controversial. Garg et al.⁴⁷ and Abuqayyas et al.⁴⁸ reported that FcRn did not contribute to the IgG exposure to brain relative to blood. On the other hand, other studies suggested that FcRn assists efflux of IgG from the brain.^49–52 This is supported by the evidence that brain expression of FcRn is co-localized with the Glut1 glucose transporter, indicating that FcRn is expressed in the capillary endothelium and thus mediates the reverse transcytosis of IgG in the brain to blood direction.⁵³ We did not find clear relation between T/P differences between IgG and FcRn non-binding IgG and FcRn expression in mouse.⁵⁴ Other approach may be able to investigate the FcRn effect on tissue distribution. For example, there were studies that used indium-111-labelled IgG and found that FcRn affect tissue distribution in skin,⁴⁵ muscle,⁴⁵ liver,^32,55 and spleen.³²

Interstitial concentrations of IgG are important to understand target engagement at the site-of-action and prediction of drug efficacy and safety. To facilitate the prediction of these concentrations we have proposed the concept of ABC_IS values. ABC_IS values were more than five times higher than measured T/P, indicating that most of the extravascular antibodies reside in interstitium space. The calculated ABC_IS values and interstitium concentrations of certain tissues were further compared with measured values from published studies that directly analysed interstitium concentrations. It was found that interstitium concentrations measured by Jadhav et al. using microdialysis for muscle, skin, and kidney were lower than calculated values. This difference may stem from the relatively high in vitro recovery of microdialysis probe, which would result in underprediction of tissue concentrations.²⁶ The reported in vitro recovery in Jadhav et al. study were 19% at 0.1 μg/mL and around 9% at 1 μg/mL or higher. This is higher than the reported values, for example in vitro recovery is typically below 5% for 10 kDa protein across a commercial 100 kDa molecular weight cutoff membrane (MWCO) perfused at a flow rate of 1 μL/min,⁵⁶ or 3% for IgG across 1000 kDa MWCO at the same flow rate.⁵⁷ Another factor that can influence the interstitium concentrations using microdialysis is tissue hydration. The entry of fluid by ultrafiltration across the highly porous dialysis membrane may result in fluid accumulation into tissue and dilution of mAb concentration in intertitial space. The effect of dilution will be more prominent for the relatively low concentration of mAbs within tissue space, and consequently can result in underestimation of interstitial concentrations.⁵⁸ On the other hand, the tissue centrifugation method by Eigenmann et al. showed higher interstitium concentration in muscle and skin. This could result from contamination by plasma content, as the reported residual plasma fractions assessed by Eigenmann et al. were 0.065 and 0.056 for skin and muscle, respectively.²⁵ The ABC_IS values for muscle, skin, and kidney calculated from our study were within the range of the values reported by the published studies that directly measured interstitial mAb concentrations. In addition, our ABC_IS value for liver was similar to the value obtained from the microdialysis study. As such, the ABC_IS values derived from our studies may help predict drug exposure at the site-of-action based on plasma exposure of mAb. In addition, as shown for ABC values, ABC_IS values may be translatable to other animal species. Of note, the molecular size, isoelectric point (pI), and different composition of interstitium (e.g. collagen, glycosaminoglycans, hyaluronan etc.) in each tissue, along with the diseased status of tissue (e.g. inflammation),⁴² can affect macromolecule distribution in interstitial space. Furthermore, if an antibody exhibits target mediated drug disposition (TMDD), the interstitium concentrations predicted by ABC_IS will be higher or lower than the actual interstitium concentrations, depending on whether the target expressed in the tissue demonstrates high or low internalization rate, respectively. As such, while the proposed ABC_IS value can help in inferring interstitium exposure in healthy and non-target expressing tissues for normal IgG, its application for different protein molecules or target expressing or diseased tissue need to be further investigated. Nonetheless, one can easily use ABC_IS-predicted interstitium concentrations to calculate receptor occupancy at the site-of-action using simple binding equations. Besides, going forward we plan to compare our model predicted tissue interstitium concentrations with directly-measured antibody interstitium concentrations obtained using different methods (e.g. microdialysis, tissue centrifugation, open flow microperfusion,⁵⁹ prenodal lymph isolation, wick implantation, capillary ultrafiltration, etc.)⁴² to further validate the ABC_IS values.

The low (i.e. <16%) T/P ratios obtained from measured or total tissue concentrations indicates limited antibody concentrations in tissue homogenates. In contrast, the ABC_IS values were found to be higher than total T/P in most tissues, indicating the concentrations in the biophase were actually higher than the expected concentrations based on ABC values. As such, plasma concentrations that are required to get efficacious concentrations at the site-of-action could be lower than expected based on ABC values.

The biodistribution data generated in this manuscript were further used to refine the PBPK model for mAbs. The innovative features of our biodistribution data include: (a) use of ELISA to quantify IgG concentrations in plasma and tissues; (b) performing whole-body perfusion of animals before collection of tissues; and (c) use of normal IgG and FcRn non-binding IgG data to understand the role of FcRn in the disposition of IgG. While our previously published mAb PBPK model well captured previously published data of IgG PK in wild-type and FcRn KO mice, our observed data for FcRn non-binding IgG in wild-type mice were overpredicted. It should be noted that previous PBPK models have been established using radiolabelled antibodies, and contamination of free label could affect the accuracy of these biodistribution results, especially for molecules with rapid elimination (e.g. FcRn non-binding IgG). Assuming FcRn non-binding IgG analyzed by ELISA better represented the accurate disposition of the fast-eliminated molecules, here we fitted the PBPK model to our new disposition data. The refined PBPK model with newly estimated CLup, Kdeg, and ${\mathrm{\sigma }}_{\mathrm{i}}^{\mathrm{v}}$ parameters was able to characterize plasma and tissue data from our studies better, and may provide a more refined model for establishing PBPK/PD relationship. Moreover, the model predicted IgG concentration in tissue interstitium were comparable to our calculated and all the literature reported interstitium concentrations, which were directly measured by different bioanalytical methods. This indicates that the updated PBPK model was able to predict antibody exposure in biophase reasonably well. However, it is important to note that the refinement of the PBPK platform model presented here is based on a single species and needs to be validated using other animal species.

In summary, we describe whole-body disposition of IgG and FcRn non-binding IgG in mice determined using ELISA. We have demonstrated how the disposition data using ELISA differs from previously published biodistribution data using radiolabelled IgG. Additionally, we have presented the concept of ABC_IS values, which more accurately infer the concentrations of antibody at the site-of-action compared to simple ABC values. We have also presented a comparison of tissue interstitium concentrations derived by us with previously reported values measured directly using microdialysis and centrifugation methods. Lastly, the new set of biodistribution data measured using ELISA were used to refine the PBPK model for mAbs.