Prediction of Human Serum Concentration–Time Profiles of Golimumab and Ustekinumab Using Pharmacokinetic Data From Common Marmosets With Assessment of Anti‐Drug Antibodies

ABSTRACT

In a previous study, retrospective predictions of human serum concentration–time profiles for three therapeutic monoclonal antibodies (mAbs) with linear pharmacokinetics were performed using allometric scaling in common marmosets and achieved favorable predictability. To generalize this method, in the present study, golimumab and ustekinumab, which exhibit linear pharmacokinetics, were repeatedly administered to marmosets. Additionally, the immunogenicity of these mAbs in common marmosets was evaluated. Golimumab and ustekinumab were administered in four repeated doses. Following the initial administration, the scaling exponents of the two mAbs were calculated, and the average scaling exponents of the five mAbs were used for the predictions. Furthermore, serum anti‐ustekinumab and anti‐golimumab antibodies were evaluated following the repeated‐dose administration. Human serum concentration–time curves of golimumab and ustekinumab were predicted using the average scaling exponents of the five mAbs. Although the predictability of the golimumab elimination rate was slightly inferior, ustekinumab showed excellent prediction performance. Low levels of anti‐ustekinumab antibodies were detected in the serum when the serum ustekinumab levels were considered to have completely diminished. Furthermore, serum anti‐golimumab antibodies were negligible during the experimental period. We identified provisional optimal scaling exponents for predicting human PK and found minimal anti‐drug antibody formation. As the cost of acquiring and maintaining cynomolgus monkeys is increasing, pharmacokinetic data of marmosets is expected to encourage the potential use of these small non‐human primates as an alternative model.

This study identified new optimal allometric scaling exponents for human PK prediction of monoclonal antibodies with linear PKs from marmoset PK data. In addition, immunogenicity in common marmosets was evaluated following the repeated dose of golimumab and ustekinumab, minimal anti‐drug antibodies were detected.

1. Introduction

Therapeutic monoclonal antibodies (mAbs) are beneficial for several diseases—including cancer, autoimmune disorders, and infectious diseases—owing to their high target specificity and lower off‐target toxicity than that of conventional small‐molecule drugs (Shepard et al. 2017). Preclinical studies on the development of therapeutic mAbs use cynomolgus monkeys (Old World monkeys) as the primary model due to their genetic similarity to humans and greater cross‐reactivity with therapeutic mAbs than other experimental animal species (Deng et al. 2011; Iwasaki et al. 2019). However, rising demand and supply shortages have significantly increased acquisition costs (Conroy 2023), underscoring the need for alternative models.

Common marmosets (Callithrix jacchus, hereafter referred to as marmosets)—also known as New World monkeys—possess several advantages as an alternative animal model. Their relatively short reproductive cycle allows systematic breeding and a stable supply. Their lower body weight (BW), approximately one‐10th the BW of a cynomolgus monkey, allows easier handling and reduces drug consumption in experiments compared with larger non‐human primates (Orsi et al. 2011; Son et al. 2020; Han et al. 2022). Although marmosets exhibit lower cross‐reactivity with mAbs than that with cynomolgus monkeys (67% vs. 82%), these estimates are based on limited data (Iwasaki et al. 2019) and thus warrant further investigation. When marmosets are used as an alternative to cynomolgus monkeys, the cost and time required for drug preparation may be reduced, particularly in the early stages of therapeutic mAb development.

Predicting human pharmacokinetic (PK) behavior is crucial during early‐stage drug development (Ovacik and Lin 2018). We attempted to retrospectively predict the human serum concentration–time curves of three therapeutic mAbs—canakinumab, adalimumab, and bevacizumab—following their intravenous or subcutaneous administration in marmosets and obtained favorable results (Tajiri et al. 2024). To generalize this prediction method, we selected golimumab, a human anti‐tumor necrosis factor‐α IgG1κ mAb (Mazumdar and Greenwald 2009), and ustekinumab, a human anti‐interleukin (IL)‐12 and IL‐23 IgG1κ mAb (Benson et al. 2011), as new model compounds and administered them repeatedly to marmosets. After the initial administration, the scaling exponents of the two mAbs were calculated, and their human serum concentration–time curves were predicted using the average scaling exponents of the five mAbs.

The data on the effects of repeated administration of mAbs on PKs and the formation of anti‐drug antibodies (ADA) in marmosets is limited. Canakinumab—a human anti‐IL‐1β IgG1κ mAb—exhibits cross‐reactivity exclusively with marmoset IL‐1β in experimental animals (Rondeau et al. 2015) and is used for treating cryopyrin‐associated periodic syndromes and diseases such as rheumatoid arthritis (RA), asthma, and psoriasis (Chakraborty et al. 2012). When canakinumab was administered subcutaneously to marmosets twice a week for 14 weeks, the maximum serum concentration (C _max) and area under the serum concentration–time curve (AUC) from 0 to 24 h after the final administration were 2.2–3.9 times higher than those after the initial administration (Canakinumab [Ilaris] New Drug Application [NDA]). However, ADA production was not reported in any of the previous studies. Conversely, when ASN004—an antibody‐drug conjugate targeting the cancer fetal antigen 5T4—was administered intravenously to marmosets twice at a 3‐week interval (0.1–1.5 mg/kg), no change was observed in the plasma concentration–time profiles between the first and second administrations. Nevertheless, ADA formation was found in the highest‐dose group (Smith et al. 2021). To the best of our knowledge, no other PK data on the repeated administration of mAbs in marmosets have been reported. Therefore, in the present study, we assessed serum concentrations and determined ADA formation following repeated administration of golimumab and ustekinumab to marmosets. We aimed to provide critical insights into the suitability of marmosets as a model for evaluating mAb PK and immunogenicity in drug development.

2. Materials and Methods

2.1. mAbs and In Vivo Studies

Golimumab (50 mg/0.5 mL prefilled syringe for subcutaneous injection) and ustekinumab (30 mg/26 mL vial for injection) were obtained from Isekyu Co. Ltd. (Nagoya, Aichi, Japan). Anti‐golimumab rabbit polyclonal antibody (pAb) and anti‐ustekinumab rabbit pAb were obtained from GenScript Biotech Corporation (Hong Kong, China).

Six male marmosets (age: 2.7–3.2 years, weight: 0.38–0.40 kg) were bred at the Yaotsu Breeding Center of CLEA Japan Inc. (Gifu, Japan). The animals were housed under controlled conditions (28 ± 5°C, 12‐h light–dark cycle) with ad libitum access to food (CMS‐1M, CLEA Japan Inc.) and water.

Of note, 12 days prior to the initial administration, approximately 300 μL of blood was collected from the marmosets' femoral vein to confirm the absence of targeted ADAs in their serum. Then, the animals were categorized into two groups (n = 3 per group) and administered 5 mg/kg (10 mL/kg) of golimumab or ustekinumab on days 0, 14, and 28, followed by a 25 mg/kg (10 mL/kg) dose on day 42. All doses were dissolved in saline and administered via the tail vein at a rate of 0.25 mL/min (for approximately 15 min).

Blood samples for golimumab and ustekinumab analyses (approximately 150 μL each) were collected without anesthesia at 20 min, 2 and 6 h, and 1, 2, 4, 7, and 10 days after the initial administration. Additional samples were collected on days 14, 16, 28, 30, 42, 44, 49, 56, and 63. On days 14, 28, and 42, blood samples were collected twice immediately before and 20 min after the second, third, and fourth administrations, respectively.

For the ADA analysis, approximately 100 μL of additional blood samples were collected on days 7, 14, 28, 42, 49, 56, 63, 92, 210, and 322 (pre‐dose of each mAb on days 14, 28, and 42). The blood samples for mAb and ADA analyses were kept in a refrigerator for 30 min and centrifuged at 1500 × g, 4°C for 15 min. The obtained serum samples were stored at −80°C until subsequent analyses. The administration of mAbs and blood sample collection were performed in compliance with the protocols (Protocol number: CLEA‐MSF2024‐1,2) approved by the Animal Care and Use Committees of Meiji Seika Pharma Co. Ltd. (Application number: C2023‐012‐0) and CLEA Japan Inc. (Application numbers: 60‐033, 040, and 065YBF).

2.2. Enzyme‐Linked Immunosorbent Assays (ELISA) of Golimumab, Ustekinumab, and Corresponding ADAs

Serum golimumab and ustekinumab levels were evaluated using the Shikari (Q‐GOL) Golimumab and Shikari (Q‐UST) Ustekinumab (Matriks Biotechnology Co., Ankara, Turkey) ELISA kits, with calibration curve ranges of 100–3000 and 3–100 ng/mL, respectively. The serum samples for golimumab and ustekinumab were diluted 5–400 and 500–10,000 times using the packaged assay buffer, and 10 and 25 μL of the diluted samples were analyzed in duplicate, respectively.

ADA formation was evaluated using Shikari (S‐ATG) Anti‐Golimumab and Shikari (S‐ATU) Anti‐Ustekinumab (Matriks Biotechnology Co.) ELISA kits according to the manufacturer's instructions. The calibration curve ranges were 30–500 for anti‐golimumab and 10–160 ng/mL for anti‐ustekinumab. Anti‐golimumab and anti‐ustekinumab rabbit pAbs were used as positive animal‐derived reference antibodies. Initially, the serum samples were diluted with the packaged assay buffer and divided into negative or positive groups based on whether the optical density values were above or below 3× the optical density of the packaged Standard Samples F (0 ng/mL, serum and stabilizer only), which was measured in triplicate and used as a cut‐off value. Samples in the positive groups were diluted with packaged confirmation reagents containing free forms of golimumab or ustekinumab, incubated for 60 min, and analyzed as described above. When the optical densities of the serum samples diluted with the confirmation reagents were reduced to ≤ 75% of those diluted with the assay buffer, the obtained serum concentrations were considered true positive. A Multiskan FC microplate reader (Thermo Fisher Scientific, MA, US) was used for analysis at 450 nm (reference: 620 nm).

2.3. Acquisition of Serum Concentration Data in Humans

Human serum concentration–time curves of golimumab following intravenous and subcutaneous injections were obtained from a previous study (Xu et al. 2010). A single dose of 100 mg golimumab was administered through a 30‐min intravenous infusion to 23 healthy male participants (mean BW: 80.0 kg) and by subcutaneous injection to 55 healthy male participants (mean BW: 78.7 kg). However, two individuals withdrew from the clinical trials, and the mean serum concentration–time curves of the remaining individuals following intravenous (n = 22) and subcutaneous (n = 54) injections were obtained. Because the BWs of the excluded individuals were unavailable, human BWs of 80.0 and 78.7 kg were used to normalize the intravenous and subcutaneous doses, respectively (1.25 and 1.27 mg/kg). For comparison, the PK parameters for patients with RA, estimated using a two‐compartment population PK model (sum of 0.3–10 mg/kg, n = 21), were obtained from a previous study (Zhou et al. 2007).

Human serum concentration–time curves for ustekinumab were obtained from the FDA document (Ustekinumab [Stelara] Drug Approval Package). Single doses of 0.09, 0.27, 0.9, and 4.5 mg/kg ustekinumab were administered via a 2‐h intravenous infusion to 4–5 patients with moderate‐to‐severe psoriasis vulgaris. After confirming that C _max and AUC were linearly related to dose, we selected the serum concentration–time curves of a 0.9 mg/kg dose as the representative dataset. Ustekinumab was administered subcutaneously at a fixed dose of 90 mg to 30 healthy participants. Because the mean BW of healthy participants was unavailable, the median BW of 81.3 kg was used for dose normalization. As the serum concentration data were presented in a graph, each concentration value was estimated using the WebPlotDigitizer software (https://apps.automeris.io/wpd/).

2.4. PK Analysis and Prediction of Human PK Parameters Using Marmoset PK Data

The mean concentration–time profiles of golimumab and ustekinumab reported in clinical studies following intravenous infusions, together with the profiles obtained after the initial intravenous administration (0–14 days) in marmosets, were fitted to a two‐compartment model. The Phoenix WinNonlin software (version 6.4; Certara USA, Princeton, NJ) was used to obtain total clearance (CL), inter‐compartmental clearance (Q), and volumes of distribution in the central (V _c) and peripheral (V _p) compartments.

The scaling exponents of adalimumab, bevacizumab, and canakinumab were obtained from Tajiri et al. (2024). In contrast, those of golimumab and ustekinumab were calculated from the observed CL, Q, V _c, and V _p values expressed on a per‐body basis in marmosets and humans. Additionally, standard BWs of 0.35 kg for marmosets and 70 kg for humans were considered, and the following equation was used:

$$\text{eCL}=\left({\log }_{10}{\text{CL}}_{\text{human}}-{\log }_{10}{\text{CL}}_{\text{marmoset}}\right)/{\log }_{10}\left({\text{BW}}_{\text{human}}/{\text{BW}}_{\text{marmoset}}\right)$$ (1)

$$\mathrm{e}Q=\left({\log }_{10}{Q}_{\text{human}}-{\log }_{10}{Q}_{\text{marmoset}}\right)/{\log }_{10}\left({\text{BW}}_{\text{human}}/{\text{BW}}_{\text{marmoset}}\right)$$ (2)

$${\text{eV}}_{\mathrm{c}}=\left({\log }_{10}{V}_{\mathrm{c},\text{human}}-{\log }_{10}{V}_{\mathrm{c},\text{marmoset}}\right)/{\log }_{10}\left({\text{BW}}_{\text{human}}/{\text{BW}}_{\text{marmoset}}\right)$$ (3)

$${\text{eV}}_{\mathrm{p}}=\left({\log }_{10}{V}_{\mathrm{p},\text{human}}-{\log }_{10}{V}_{\mathrm{p},\text{marmoset}}\right)/{\log }_{10}\left({\text{BW}}_{\text{human}}/{\text{BW}}_{\text{marmoset}}\right)$$ (4)

where eCL, eQ, eV _c, and eV _p represent the scaling exponents of CL, Q, V _c, and V _p, respectively. The average scaling exponents of the five mAbs were used to predict the CL, Q, V _c, and V _p in humans (CL_pred, Q _pred, V _{c, pred}, and V _{p, pred}, respectively) using allometric scaling.

2.5. Prediction of Human Serum Concentration–Time Profiles of Five mAbs

The human serum concentration–time profiles were predicted following intravenous infusion and subcutaneous injection of the two mAbs using the corresponding CL_pred, Q _pred, V _{c, pred}, and V _{p, pred} values, as described previously (Tajiri et al. 2024). To determine the serum concentrations following subcutaneous injection, additional PK parameters, the assumed first‐order absorption rate constant (k_a) of 0.287 day⁻¹, and the empirical bioavailability (F; F = ‒ 0.0672 × CL_pred [mL/day/kg] + 0.894) were used (Haraya et al. 2017). The human serum concentration–time profiles following intravenous infusion and subcutaneous injection of canakinumab, adalimumab, and bevacizumab were re‐predicted using the corresponding CL_pred, Q _pred, V _{c, pred}, and V _{p, pred} values (Tajiri et al. 2024).

3. Results

3.1. Serum Concentration–Time Curves Following Repeated Administrations of Golimumab and Ustekinumab

The serum concentration–time profiles following repeated intravenous administration of golimumab are presented in Figure 1A. After the initial administration of 5 mg/kg to the three marmosets, the individual serum concentration–time curves were highly similar, and the mean serum concentrations decreased in a biphasic manner with a terminal half‐life of 6.7 days. Following the second administration of 5 mg/kg, the serum concentration at 20 min in one animal (G1) was approximately 8‐fold lower than that observed in the other two animals. The mean ± standard deviation serum concentrations at 20 min after the first, second, and third administrations were 73.2 ± 8.7, 61.1 ± 43.6, and 67.5 ± 9.7 μg/mL, respectively, whereas the corresponding mean trough concentrations measured 14 days after each administration were 7.3 ± 1.4, 7.4 ± 3.0, and 6.8 ± 4.1 μg/mL, respectively. Although the reason for the reduced concentration 20 min after the second administration in G1 remained unclear, repeated dosing did not alter the overall serum concentration–time profiles across the three administrations. Following the fourth administration at 25 mg/kg, the individual serum concentration–time profiles remained comparable among the three marmosets. The mean serum concentration at 20 min was 375.6 ± 54.5 μg/mL, approximately 5‐fold higher than that observed after administration of 5 mg/kg. The apparent elimination half‐life estimated from the final three sampling points was 7.5 days.

FIGURE 1.

Mean and individual serum concentration–time profiles of (A) golimumab and (B) ustekinumab in common marmosets. Marmosets (n = 3 per group) were administered repeated intravenous doses of 5, 5, and 5 mg/kg on days 0, 14, and 28, followed by a high dose of 25 mg/kg on day 42. Mean serum concentrations are shown for 0–14 days, whereas individual animal data (G1–G3, U1–U3) are shown for the full study period (0–63 days).

Following the initial administration of 5 mg/kg ustekinumab to the three marmosets, the mean serum concentrations decreased in a biphasic manner, with a terminal half‐life of 7.3 days (Figure 1B). The mean ± standard deviation serum concentrations at 20 min after the first, second, and third administrations were 85.4 ± 5.7, 84.8 ± 5.2, and 92.3 ± 3.9 μg/mL, respectively, whereas the corresponding mean trough concentrations 14 days after each administration were 11.2 ± 4.7, 14.3 ± 5.7, and 11.8 ± 3.5 μg/mL, respectively. Although the 20‐min value after the third administration and the 14‐day value after the second administration increased, the absolute increased values were still small, and the effect of repeated administration was minimal. After the fourth administration at 25 mg/kg, the mean serum concentration at 20 min increased to 436.9 ± 37.8 μg/mL, approximately 5‐fold greater than that observed at 5 mg/kg. The apparent elimination half‐life estimated from the final three time points was 6.1 days.

3.2. Prediction of Human Serum Concentration–Time Profiles From the Initial Intravenous Injection Data in Marmosets

The observed CL, Q, V _c, and V _p values obtained after the initial administration in marmosets and the values in humans obtained from the literature using the two‐compartment model are summarized in Table 1. For golimumab, the observed human CL, Q, V _c, and V _p values estimated from the clinical data reported by Xu et al. (2010) were comparable to those previously reported by Zhou et al. (2007) using a two‐compartment population PK model (Supporting Information S1: Supplemental Material S2). Based on these PK parameters, the scaling exponents for golimumab and ustekinumab from marmosets to humans were calculated using Equations ((1), (2), (3), (4)). By integrating previously reported data for canakinumab, adalimumab, and bevacizumab (Tajiri et al. 2024), the average exponent values of the five therapeutic mAbs were found to be 0.73, 0.59, 0.95, and 0.97 for CL, Q, V _c, and V _p, respectively. These values were subsequently defined as the updated average scaling exponents and used for subsequent analyses (Table 2).

TABLE 1.

Observed and predicted pharmacokinetic parameters of golimumab and ustekinumab in common marmosets and humans.

Antibody	Observed in marmosets a				Observed in humans b				Predicted in humans c
	CL	Q	V c	V p	CL	Q	V c	V p	CLpred	Q pred	V c, pred	V p, pred
	(mL/day/kg)	(mL/day/kg)	(mL/kg)	(mL/kg)	(mL/day/kg)	(mL/day/kg)	(mL/kg)	(mL/kg)	(mL/day/kg)	(mL/day/kg)	(mL/kg)	(mL/kg)
Golimumab	15.7	44.0	70.3	67.8	6.27	3.99	46.9	52.4	3.75 (0.60)	5.02 (1.26)	53.9 (1.15)	57.8 (1.10)
Ustekinumab	10.8	41.3	60.4	47.7	1.83	3.88	30.3	35.8	2.62 (1.43)	4.61 (1.19)	47.3 (1.56)	40.7 (1.14)
Canakinumab d	10.5	13.0	37.9	23.2	2.51	2.03	64.1	31.2	2.52 (1.00)	1.48 (0.73)	29.0 (0.45)	19.8 (0.63)
Adalimumab d	13.1	74.4	56.8	46.8	2.80	5.76	35.1	31.9	3.13 (1.12)	8.48 (1.47)	43.6 (1.24)	39.9 (1.25)
Bevacizumab d	14.4	56.2	46.1	30.8	3.38	9.09	40.1	25.5	3.44 (1.02)	6.40 (0.70)	35.4 (0.88)	26.3 (1.03)

Note: Body weights of marmosets and humans for the allometric scaling were assumed to be 0.35 and 70 kg, respectively.

Abbreviations: CL, clearance; Q, inter‐compartmental clearance; V _c, volume of distribution in the central compartment; V _p, volume of distribution in the peripheral compartment.

^a Obtained by fitting the mean serum concentration–time curves following the initial dose (0–14 days) to a two‐compartment model.

^b Calculated from reported serum concentration–time data following intravenous infusion of 1.25 mg/kg golimumab in healthy participants (Xu et al. 2010) and 0.9 mg/kg ustekinumab in patients with psoriasis vulgaris (Ustekinumab [Stelara] Drug Approval Package).

^c Predicted using the average scaling exponent values of the five mAbs listed in Table 2. Values in parentheses were the predicted/observed ratio for each parameter.

^d The values of Observed in marmosets and Observed in human were used from Tajiri et al. 2024 for canakinumab, adalimumab, and bevacizumab.

TABLE 2.

Individual and average scaling exponents for the four pharmacokinetic parameters of five mAbs from common marmosets to humans.

	eCL	eQ	eV c	eV p
Adalimumab a	0.71	0.52	0.91	0.93
Bevacizumab a	0.73	0.66	0.97	0.96
Canakinumab a	0.73	0.65	1.10	1.06
Golimumab	0.83	0.55	0.92	0.95
Ustekinumab	0.66	0.55	0.87	0.95
<Average>	0.73	0.59	0.95	0.97

Note: eCL, eQ, eV _c, and eV _p are exponents for CL clearance;

Q, inter‐compartmental clearance; V _c, volume of distribution in the central compartment; and V _p, volume of distribution in the peripheral compartment.

^a Scaling exponent values reported in the previous report (Tajiri et al. 2024). The average exponent values of the three mAbs for CL, Q, V _c, and V _p were 0.72, 0.61, 0.99, and 0.98, respectively.

The CL_pred, Q _pred, V _c,pred, and V _p,pred values of golimumab and ustekinumab were estimated using allometric scaling (Table 1). The recalculated parameters for canakinumab, adalimumab, and bevacizumab, using newly defined scaling exponent values, are also presented in Table 1. The CL_pred of golimumab (3.75 mL/day/kg) was approximately half of the observed value (6.27 mL/day/kg), whereas the other predicted human PK parameters were similar to the observed values. Consequently, the observed serum concentrations of golimumab in humans were within 1.5‐fold of the predicted values until day 7 and within 2‐fold of the predicted values until day 20 (Figure 2A). However, the observed concentrations decreased more rapidly than the predicted concentrations, indicating a smaller CL_pred value than the observed CL value. Conversely, the observed serum concentrations of ustekinumab were within 1.5‐fold of the predicted concentrations throughout the sampling period (Figure 3A).

FIGURE 2.

Observed and predicted human serum concentration–time profiles of golimumab following (A) intravenous and (B) subcutaneous administration. (A) 100 mg (1.25 mg/kg) 0.5–1 h intravenous infusion and (B) 100 mg (1.27 mg/kg) subcutaneous injection in humans. The observed data (open and closed circles) are obtained from Xu et al. (2010) and represent the mean concentrations ± standard deviation. The predicted curves (solid lines) were generated by allometric scaling of common marmoset PK data, using the newly defined average scaling exponents for five monoclonal antibodies, as shown in Table 2. Dotted, dashed, and dashdotted lines represent 1.5‐fold, 2‐fold, and 3‐fold differences from the predicted values, respectively. The adopted absorption rate constant (k_a) and empirical bioavailability (F) values for the prediction following subcutaneous injection were derived from Haraya et al. (2017).

FIGURE 3.

Observed and predicted human serum concentration–time profiles of ustekinumab following (A) intravenous and (B) subcutaneous administration. (A) 0.9 mg/kg 2‐h intravenous infusion and (B) 90 mg (1.125 mg/kg) subcutaneous injection in humans. The observed data (open and closed squares) are derived from the FDA Drug Approval Package for ustekinumab (Stelara) and represent the mean concentrations ± or + standard deviation. The predicted curves (solid lines) were estimated by allometric scaling from common marmoset PK data, using the average scaling exponents for five monoclonal antibodies, as shown in Table 2. Dotted, dashed, and dashdotted lines represent 1.5‐fold, 2‐fold, and 3‐fold differences from the prediction, respectively. The adopted absorption rate constant (k_a) and empirical bioavailability (F) values for the prediction following subcutaneous injection were derived from Haraya et al. (2017).

For prediction of human serum concentration–time profiles following subcutaneous administration, assumed ka and F values of 0.287 day⁻¹ and 0.642 for golimumab, and 0.287 day⁻¹ and 0.718 for ustekinumab, respectively, were used as described in the Materials and Methods section. The predicted serum concentrations profile of golimumab reached a C _max of 8.4 μg/mL approximately 5 days after dosing, then gradually declined (Figure 2B). The observed C _max value were approximately 30% lower than the predicted value at 3–4 days after administration, and subsequently decreased below the 1.5‐fold difference after 4 days and the 2‐fold difference after 20 days. Conversely, the predicted serum concentration profile of ustekinumab reached a C _max of 10 μg/mL at 5–6 days after administration (Figure 3B). Although the observed C _max of ustekinumab was approximately 20% greater than the predicted value, the simulated serum concentration–time profile generally yielded the observed human PK profiles.

To further validate the versatility of the newly updated average scaling exponents derived from the five mAbs (Table 2), human serum concentration–time profiles for canakinumab, adalimumab, and bevacizumab from our previous study were retrospectively reanalyzed. The predicted serum concentration profiles for canakinumab following intravenous (Figure 4A) and subcutaneous (Figure 4B) administration generally yielded the observed clinical PK data (Figure 4). Although the observed concentrations after intravenous infusion slightly exceeded the 1.5‐fold prediction interval after day 40, overall predictive performance remained high. For adalimumab (Figure 4C,D) and bevacizumab (Figure 4E), the observed human serum concentrations following intravenous or subcutaneous administration were accurately reproduced using the updated scaling exponents, with nearly all observed data points remaining within the 1.5‐fold prediction interval.

FIGURE 4.

Retrospective prediction of human serum concentration–time profiles for canakinumab, adalimumab, and bevacizumab using the newly defined average scaling exponents. Human PK profiles for three mAbs reported in Tajiri et al. (2024) were re‐predicted through allometric scaling from common marmoset data using the newly defined average scaling exponents derived from five monoclonal antibodies (Table 2). (A, B) Canakinumab: 600 mg intravenous infusion (A) and 300 mg subcutaneous injection (B) in healthy participants. (C, D) Adalimumab: 1 mg/kg intravenous (C) and 1 mg/kg subcutaneous (D) injections in healthy participants. (E) Bevacizumab: 10 mg/kg intravenous infusion (τ = 1.5 h) in patients with colon cancer. Observed data (open and closed circles) are derived from the public report listed in Tajiri et al. (2024) and represent the mean concentrations ± standard deviation. In all panels, solid lines represent the predicted curves using common marmoset data, while dotted, dashed, and dashdotted lines represent 1.5‐fold, 2‐fold, and 3‐fold differences from the prediction, respectively. For subcutaneous predictions (B, D), the adopted absorption rate constant (k_a) and empirical bioavailability (F) values for the prediction following subcutaneous injection were derived from Haraya et al. (2017).

3.3. Detection of ADA Following Repeated Administration of Golimumab and Ustekinumab

Calibration curves for the quantification of anti‐golimumab and anti‐ustekinumab antibodies are shown in Figure 5A,B, respectively. Based on the established cut‐off values, the detection limits for serum anti‐golimumab and anti‐ustekinumab antibodies were considered as 0.41 and 0.01 μg/mL, respectively. To evaluate the applicability of these kits for the detection of animal‐derived ADAs, marmoset serum samples containing anti‐golimumab and anti‐ustekinumab rabbit pAbs were analyzed simultaneously. These samples exhibited concentration‐dependent response curves, and the optical signals decreased significantly upon addition of the corresponding unconjugated mAbs used as confirmation reagents.

FIGURE 5.

Optical densities in the enzyme‐linked immunosorbent assays (ELISA) for (A) anti‐golimumab and (B) anti‐ustekinumab antibodies. Closed circles represent human antibody standards in human serum proteins. Open circles represent rabbit pAb spiked into marmoset serum as positive controls to confirm the assay's applicability to marmoset samples. Open triangles represent the same rabbit pAb in marmoset serum, fortified with excess unconjugated golimumab or ustekinumab, respectively (confirmation reagent).

The method was applied to marmoset serum samples collected before and after repeated administration of golimumab, and no detectable ADA formation was observed throughout the study (Table 3). However, low levels of anti‐ustekinumab antibodies were detected in three marmosets on day 210 following repeated ustekinumab administration. These antibody levels subsequently decreased by day 322, corresponding to 112 days after the initial detection. All detected signals decreased below 20% after the addition of confirmation reagents and were considered true positive.

TABLE 3.

Serum levels of anti‐golimumab and anti‐ustekinumab antibodies following repeated administration of golimumab and ustekinumab in common marmosets.

Antibody	Animal no.	Anti‐golimumab or anti‐ustekinumab antibody (μg equivalent to human ADA/mL)
		Day −12	Day 7	Day 14	Day 28	Day 42	Day 49	Day 56	Day 63	Day 92	Day 210	Day 322
Golimumab	G1	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND
	G2	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND
	G3	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND
Ustekinumab	U1	ND	ND	ND	ND	ND	ND	ND	ND	ND	2.00	0.72
	U2	ND	ND	ND	ND	ND	ND	ND	ND	ND	2.85	1.67
	U3	ND	ND	ND	ND	ND	ND	ND	ND	ND	0.06	0.05

Note: Golimumab and ustekinumab were administered by repeated iv infusions at dose levels of 5, 5, 5, and 25 mg/kg on days 0, 14, 28, and 42 after the blood sampling for the ADA analyses, respectively. ND: Not detected (< 0.41 and < 0.01 μg/mL for golimumab and ustekinumab, respectively, depending on their cut‐off values).

4. Discussion

Antibody drugs can be categorized into those exhibiting linear PK behavior driven by non‐specific IgG elimination and those exhibiting nonlinear PK characteristics driven mainly by target‐mediated drug disposition (Wang et al. 2016; Liu 2018; Ovacik and Lin 2018). For reliable PK prediction using simple allometric scaling, the target mAbs should exhibit linear PK properties within the evaluated dose range (Dong et al. 2011). In our previous study using marmosets, canakinumab, adalimumab, and bevacizumab, all of which exhibit linear PK behavior, were selected as model compounds (Tajiri et al. 2024). For golimumab and ustekinumab, both C _max and AUC reportedly increased proportionally with dose following single intravenous administration over dose ranges of 0.1–10 mg/kg and 0.09–4.5 mg/kg, respectively (Zhou et al. 2007, Ustekinumab [Stelara], FDA Drug Approval Package). Consistent with these reports, serum concentrations of golimumab and ustekinumab in marmosets following intravenous administration of 25 mg/kg were approximately 5‐fold higher than those observed after administration of 5 mg/kg (Figure 1A,B). These findings indicate that golimumab and ustekinumab exhibit linear PK behavior in humans and marmosets.

Regarding the scaling exponents from marmosets to humans, the eCL value for golimumab (0.83) was higher, whereas that for ustekinumab (0.66) was slightly lower than the mean eCL values previously obtained using three mAbs (0.71–0.73) (Table 2). Nevertheless, the average eCL value calculated from all five mAbs (0.73) was not considerably different from that of the other three mAbs (0.72). Conversely, the eQ, eV _c, and eV _p values for golimumab and ustekinumab were comparable to those of the other three mAbs. Based on these studies, the optimal exponent values for predicting CL, Q, V _c, and V _p from marmosets to humans were defined as 0.73, 0.59, 0.95, and 0.97, respectively (Table 2). When these average values were compared with those obtained from 24 therapeutic mAbs in cynomolgus monkeys (Haraya et al. 2017), eCL and eQ were slightly lower.

In contrast, V _c and V _p were almost comparable (Tajiri et al. 2024). The serum concentration–time profiles of canakinumab, adalimumab, and bevacizumab were retrospectively predicted using the newly defined average exponent values (Figure 4). The observed concentrations of canakinumab exceeded the predicted values by more than 1.5‐fold after 40 days of intravenous administration; however, the results remained consistent with those reported previously. These retrospective results using the five‐mAb average exponents were consistent with our previous findings using three‐mAb exponents, indicating that incorporating golimumab and ustekinumab data into the scaling model enhances the robustness of human PK predictions from marmoset data across various mAbs.

In this study, using the newly defined optimal exponent values, the observed CL of golimumab was approximately twice the predicted value, highlighting its high eCL (Table 1). Furthermore, the observed golimumab serum concentration decreased more rapidly than expected due to the underprediction of elimination. The observed serum concentration–time profile of golimumab after subcutaneous administration was lower than the predicted profile. The predictive accuracy of golimumab was lower than that of the other four antibodies due to several possible reasons, such as differences in the relative affinities for neonatal Fc receptors (FcRn) between humans and marmosets compared with other mAbs. FcRn is a major molecule involved in IgG recycling (Ovacik and Lin 2018) and contributes to the PK profile. Notably, the five essential amino acids of FcRn for IgG binding are the same among marmosets, cynomolgus monkeys, and humans (Tajiri et al. 2024). However, other minor differences may have affected prediction accuracy, including glycosylation and the isoelectric point, which can affect clearance (Liu 2018). Given the good predictability of golimumab from cynomolgus monkeys to humans (Haraya et al. 2017), the degree of influence of these factors on clearance may differ among marmosets, humans, and cynomolgus monkeys.

Furthermore, the predicted curve was obtained using an F‐value of 0.642 (Haraya et al. 2017), whereas the actual F‐value of golimumab was 0.511 (Xu et al. 2010). The predictions for golimumab exhibited large discrepancies in the elimination phase for IV and SC administration. The accuracy during the elimination phase did not significantly improve, even after adjusting the F value to its actual value (Supporting Information S1: S3). This suggests that F, influencing absorption processes, may not improve prediction accuracy during the elimination phase. Hence, while the discrepancy in the F value is considered one factor contributing to the inaccurate prediction of golimumab, its overall impact is likely limited.

Furthermore, the four observed PK parameters of ustekinumab in clinical settings were slightly lower than the predicted values; however, the observed human serum concentration–time profile following intravenous administration remained within the 1.5‐fold predicted interval (Figure 3A). The predicted serum concentrations after subcutaneous administration were slightly lower at C _max but mimicked the observed concentration–time curve, indicating favorable predictive performance for ustekinumab (Figure 3B). In previous studies using monkeys, although the species were not clearly specified as cynomolgus monkeys or rhesus macaques, the CL values and serum concentration–time curves of golimumab and ustekinumab in humans were predicted using single‐species allometry and an elementary Dedrick plot approach (Ling et al. 2009). The predictability of golimumab was favorable and consistent with previous findings obtained using cynomolgus monkeys (Haraya et al. 2017), in which the golimumab‐estimated eCL was similar to the optimal eCL value derived from analysis of 24 mAbs (Supporting Information S1: S1A). Conversely, in the case of ustekinumab, the predicted human CL value derived from macaque monkeys was significantly lower, and the reason for the over‐prediction in humans is unclear. Interestingly, the predictability levels for human PK of these two mAbs were inversely associated among small and large non‐human primate species.

ADA formation following antibody administration can reduce plasma or serum concentrations of mAbs and influence clinical efficacy (Liu 2018; Ovacik and Lin 2018). The frequency of ADA formation in clinical settings differs among approved therapeutic mAbs (Vaisman‐Mentesh et al. 2020), and differences are observed for the same drug across different clinical trials. The prediction of human ADA incidence from animal studies remains difficult because of major interspecies differences in immune systems and in major histocompatibility complex genes (Brennan et al. 2010). Because golimumab and ustekinumab are human IgGs, ADA formation in non‐human primates is more common than that in humans. In some cases, ADAs were observed more frequently at low doses than at high doses. For example, intravenous administration of golimumab at 10 mg/kg to three cynomolgus monkeys induced ADA formation in all animals by day 35 (Golimumab [Simponi] NDA). Nevertheless, the continuous administration of 50 mg/kg for 25 weeks to 16 animals resulted in ADA detection in only one animal by day 84. This difference may reflect induction of immune tolerance or neutralization of circulating ADA by excess coexisting mAbs at high doses (Golimumab [Simponi] NDA).

In the present study, commercially available human ADA measurement kits were used to evaluate ADA formation in marmoset serum samples. According to the manufacturer's protocol, anti‐golimumab and anti‐ustekinumab antibodies are captured by immobilized golimumab or ustekinumab coated onto the assay wells. After washing the wells, the analytes were detected using horseradish peroxidase‐conjugated golimumab or ustekinumab. This assay design enables the detection of target‐specific ADAs, regardless of whether they originate from humans or animals. To confirm this applicability, rabbit anti‐golimumab and anti‐ustekinumab pAbs were detected simultaneously. These heterologous antibodies were quantified in a concentration‐dependent manner, and the optical densities considered decreased after the addition of unlabeled golimumab or ustekinumab as confirmation reagents. Thus, we concluded that marmoset‐derived ADAs can be detected using these assay systems.

In the repeated‐dosing study using golimumab and ustekinumab, the initial and two consecutive doses were administered at a low dose, where ADA formation was likely to be detected. The dose response was evaluated by increasing the dose 5‐fold at the fourth administration. The serum concentration–time profiles did not change with repeated administration. Furthermore, ADAs were not detected in serum samples collected up to day 92. However, when mAbs remain in the serum, ADA detection can be suppressed, resulting in false‐negative results. The assay kits used in this study do not specify a drug tolerance limit, and a previous study using the kit for Golimumab (Tawa et al. 2022) also states that ADA cannot be measured in the presence of drugs. In the present study, although the serum concentrations of golimumab and ustekinumab were not directly measured on day 92, extrapolation from the serum concentration–time profiles following the fourth administration suggest that several μg/mL of these mAbs could persist and interfere with ADAs. Therefore, samples were added on days 210 and 322, when the serum mAb concentrations were considered to have completely diminished. Low levels of anti‐ustekinumab antibody seemed to remain for a long time in the serum, whereas no anti‐golimumab antibody was detected (Table 3). The present study involves a limited number of animals, and the dosing and blood collection times differed from those in cynomolgus monkey studies. However, the levels or frequency of ADA formation against golimumab and ustekinumab in marmosets appeared comparable to, or possibly lower than, those reported in cynomolgus monkeys. Future applications of marmosets for detailed immunogenicity assessment will likely require optimization of ADA detection methods, including the determination of drug tolerance limits and the incorporation of acid dissociation steps.

5. Conclusion

To conclude, we determined the scaling exponent values of golimumab and ustekinumab and defined new optimal scaling exponents for predicting human PK profiles of first‐generation mAbs with linear PK behavior using marmoset‐derived PK data. Furthermore, evaluation of serum concentrations and corresponding ADA formation showed that ADA responses against golimumab and ustekinumab in marmosets were absent or minimal and were unlikely to influence PK behavior following repeated administration. These findings further support the applicability of marmosets as an alternative non‐human primate model to cynomolgus monkeys for preclinical PK studies. Herein, the average scaling exponents were calculated using five mAbs, including the two newly evaluated mAbs, to improve predictive robustness. Nevertheless, available PK data in marmosets remain limited. Future studies incorporating additional mAbs will therefore be necessary to further validate and refine the scaling model. Simultaneously, studies focusing on improving prediction models for nonlinear PK behavior, FcRn‐modified mAbs, and bispecific antibodies should be performed to improve translational accuracy in drug development. For the prediction of mAbs with nonlinear pharmacokinetic properties, highly sophisticated methods, including physiologically based pharmacokinetic modeling using monkey‐derived data, are currently being investigated (Glassman and Balthasar 2019; Haraya et al. 2022). Because nonlinear PK behavior is often driven by target‐mediated drug disposition, evaluation of target molecule homology between humans and marmosets will be particularly important. In cases where sufficient homology is confirmed, these advanced modeling approaches may also enable the application of marmosets to predict antibodies with nonlinear PK characteristics. Altogether, the present findings support the utility of marmosets as a promising alternative model for improving the reliability of PK prediction during early‐phase therapeutic antibody development and contribute to more efficient translation from preclinical research to clinical application.

Funding

The authors have nothing to report.

Conflicts of Interest

Nanami Furukawa, Kensuke Kagiyama, Hiroshi Ikeda, and Kazumasa Fukasawa, are employees of CLEA Japan Inc., a breeder and distributor of common marmosets. The authors declare no conflicts of interest.

Supporting information

Supporting Information S1

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.