The three studies were conducted in compliance with the Declaration of Helsinki and ICH Good Clinical Practice and were registered with ClinicalTrials.gov and EudraCT. The study protocols were approved by the relevant independent Ethics Committees before starting the study. All study participants provided their written informed consent before entry into the study.

A brief description of the studies is given below. Their main characteristics are tabulated in Supplementary Section 1.1.

The primary objective of the SAD and MAD studies was to investigate the safety, tolerability, and PK of ascending single or multiple oral doses of zabedosertib. The primary objective of the FE/abs.BA study was to investigate the impact of concomitant food intake on the PK properties of zabedosertib, to confirm the data from the orientating food-effect arm in the SAD study and to determine the absolute bioavailability of zabedosertib using a ¹³C-microtracer approach (Lozac’h et al., 2018).

In addition, a panel of exploratory PD and safety biomarkers were investigated in the SAD and MAD studies to show first evidence of pharmacological activity. Of this panel, only the cytokine release determined ex vivo after stimulation is reported here.

Healthy men aged 18–50 years (inclusive) were eligible for study participation in the three reported studies (details in Supplementary Sections 1.2.1 through 1.2.3).

Both the SAD and MAD studies, which also included a single-dose part, were randomized, placebo-controlled, double-blind (sponsor-unblinded), parallel-group, and ascending-dose studies (Figures 1A,B). Participants received either zabedosertib immediate-release tablets or corresponding placebo tablets (N = 52 and N = 40 received zabedosertib in the SAD and MAD studies, respectively). ⁴ For the first dose steps in the SAD study, a liquid service formulation was used instead of tablets.

The FE/abs.BA study was an open-label, randomized, three-period, two-sequence cross-over study. Study participants (N = 10) were randomly allocated to one of two possible treatment sequences (Figure 1C): administration of zabedosertib in the fasted state in period 1 and following a standardized high-fat, high-calorie breakfast (FDA/CDER, 2002⁠; EMA, 2012a) in period 2 or vice versa. In period 3, zabedosertib was administered following a moderate-fat, moderate-calorie breakfast (details in Supplementary Section 1.3.1), as high-fat, high-calorie meals are not regarded as a typical breakfast worldwide. To investigate the absolute bioavailability of zabedosertib after oral administration, participants received an intravenous (IV), stable-isotope-labeled microtracer dose of zabedosertib (¹³C₆-zabedosertib) on top of the oral zabedosertib dose taken in the fasted state.

The study designs were chosen based on regulatory guidance and the standard procedures used in similar early phase 1 studies of the study sponsor. In single- and multiple-dose first-in-men studies, ascending-dose designs are mandatory for safety reasons. The dosages to be studied were selected to cover the intended efficacious exposure range, which was derived from rodent inflammation models (LPS models), taking into account interspecies differences such as differences in protein binding. The starting dose was selected in compliance with the relevant FDA guidance (FDA/CDER, 2005). The dose range to be studied started at an exposure where the assumed maximum concentration would not reach the half-maximum inhibitory concentration (IC₅₀) for IL-6 or TNFα inhibition (based on a non-clinical LPS model).

The studies each comprised a screening period, up to three treatment periods (Figure 1), a follow-up period, and the end-of-study visit.

Procedures included safety assessments and blood sampling for PK analyses, zabedosertib plasma protein-binding analyses (SAD and MAD studies), and exploratory PD biomarker analyses (SAD and MAD studies). ⁵ For sampling times, see Supplementary Section 1.3.2.

Primary variables for the assessment of safety and tolerability were the frequency and severity of treatment-emergent adverse events (TEAEs).

Treatment-emergent adverse events of special interest (AESIs) were (i) confirmed or suspected severe invasive bacterial infections (all studies), (ii) systemic hypersensitivity reactions (MAD study), and (iii) noninvasive infections of the skin (MAD study).

Quantitative analyses of zabedosertib and [¹³C]-zabedosertib in plasma were performed using fully validated high-pressure liquid chromatography and tandem mass spectrometric detection methods; see Supplementary Section 1.4.1 for details and performance parameters. ⁶ Method validation and analyses were conducted in compliance with the pertinent regulatory guidelines (FDA/CDER/CVM, 2001⁠; FDA/CDER/CVM, 2018⁠; EMA, 2011; EMA, 2012b).

The unbound fraction (f_u) of zabedosertib in plasma was determined using equilibrium dialysis after spiking plasma samples ex vivo with [¹⁴C] zabedosertib; details are provided in Supplementary Section 1.4.2.

The impact of zabedosertib on cytokine release (e.g., TNFα) was measured after ex vivo stimulation of whole-blood cells with LPSs; details are provided in Supplementary Section 1.4.3.

Noncompartmental PK parameters for zabedosertib and [¹³C]-zabedosertib were calculated using PK software WinNonlin v5.3 or Phoenix 8.1 (Certara Companies).

Statistical analyses were conducted separately for each study using the program package SAS v9.2 (SAS Institute, Cary, NC). Within each study, data from all participants receiving placebo were pooled for safety analyses. Sample sizes were planned based on safety considerations and prior experience with similar studies. All analyses were explorative.

Dose proportionality was investigated separately for the SAD and MAD studies and within the MAD study separately for single and repeated administration. Analyses of variance were performed on log-transformed, dose-normalized PK parameters, with the dose level as an independent factor. Based on these analyses, point estimates (least squares means) and exploratory 90% confidence intervals (CIs) for dose ratios were calculated by retransforming the estimates and CIs to the original scale.

In addition, power models were applied where a linear regression of dose on log-transformed AUC and C_max data was performed (SAD study). The point estimate and 90% CI of the slope in the respective linear regression are used to characterize the dose proportionality.

The relative bioavailability of zabedosertib administered under different conditions was investigated by performing mixed models including fixed treatment effects and random subject effects on the log-transformed AUC and C_max values. For each mixed model, the point estimate and 90% CI for the ratio of “condition A/condition B” (e.g., solid form vs. liquid form), derived by inverse transformation of the estimates and 90% CIs from the model, are reported.

For the evaluation of the absolute bioavailability (FE/abs.BA study), PK parameter F was calculated as F = (AUC/D)_oral / (AUC/D)_IV based on data obtained under fasted conditions. The geometric mean as point estimate and the corresponding 90% CI were calculated for this parameter, which was assumed to be log-normally distributed.

To evaluate the impact of food intake on the PK of zabedosertib (FE/abs.BA study), log-transformed AUC and C_max were subjected to an analysis of variance, including treatment, subject, period, and sequence effects (high-fat, high-calorie meal) or subject and period effects (moderate-fat, moderate-calorie meal). Point estimates and 90% CIs for the ratio “fed/fasted” were calculated by retransforming the analysis of variance estimates to the original scale.

A population-pharmacokinetic analysis based on PK data from the SAD and MAD studies was conducted to describe the variability in the PK of zabedosertib taken as tablets under fasting conditions. The analysis was conducted using nonlinear mixed-effects modeling with NONMEM v7.3 (ICON Development Solutions, Dublin, Ireland); details are provided in Supplementary Section 1.5.

As binding of zabedosertib to IRAK4 could not be measured in vivo, a model-based approach was developed to assess the potential target occupancy of zabedosertib. This approach assumed that zabedosertib target binding increases with increasing unbound zabedosertib plasma concentration and that the maximal binding capacity is considerably lower than the maximal zabedosertib concentration, resulting in maximal target occupancy toward higher zabedosertib concentrations. Furthermore, it is assumed that zabedosertib binding to the target follows a second-order kinetic process; that is, the target binding rate is dependent on the unbound zabedosertib plasma concentration and free target [see Tummino and Copeland (2008) for further details on target binding kinetics]. When all the targets are occupied, the target binding rate becomes zero, signifying that the maximal binding capacity has been attained. Zabedosertib is assumed to dissociate from IRAK4 following first-order kinetics, which can be measured in vitro (Tummino and Copeland, 2008). The second-order binding rate constant can be estimated from the equilibrium dissociation constant Kd. As described by Winkler et al. (2021) for the IRAK4 inhibitor zimlovisertib, Kd can be derived from the IC₅₀ of the inhibition of IL6 release in an in vitro human whole-blood assay after stimulation with the toll-like-receptor-7 and -8 agonist resiquimod (R848); details are provided in Supplementary Section 1.6. IC₅₀ was estimated by fitting the concentration–IL-6 release inhibition data to a sigmoidal E_max model using R package drm (Ritz et al., 2015).

The (projected) target occupancy in healthy volunteers was estimated by linking the variability in unbound PK, as described using the population PK model, to the estimated IC₅₀ value from the in vitro human whole-blood assay and target residence time (∼5 min) using a typical model for target-binding kinetics, as described above. Information on the actual IRAK4 (target) concentration is not required under the assumption that drug concentration would be in excess of the target concentration. R package RxODE (Wang et al., 2016) was used for the estimation.

In total, 130 healthy men aged 18–50 years were included in the three studies. Their demographic characteristics were well balanced within each study (Table 1), with no notable differences between dose groups or treatment sequences (Supplementary Sections 2.1–2.3). All participants were evaluated for safety, PK, and PD; see Supplementary Sections 2.4–2.6 for participant disposition.

Zabedosertib showed good safety and tolerability. This is reflected in the fact that mainly mild TEAEs were reported and none of the TEAEs led to discontinuation of treatment. Only one adverse event, also the only serious TEAE, as described below, was assessed as severe; however, it was not related to zabedosertib due to the time elapsed since the last drug intake. Different laboratory parameters showed, in individual cases, transient changes, but there was no indication of any relationship to the dose administered and no clinically meaningful difference between zabedosertib- and placebo-treated participants. No drug-related changes were observed in vital signs or ECG parameters.

Single-dose administration of zabedosertib to healthy men led to a transient reduction in LPS-mediated cytokine release (6 h post-dose) in the ex vivo whole-blood assay (Table 7), which resolved within approximately 3 days (Figure 6). The effect was most pronounced for TNFα (6-h stimulation protocol, 0.1 ng/mL LPS; see Supplementary Section 1.4.3) and interferon gamma (24-h stimulation protocol, 100 ng/mL LPS). This was interpreted as an indication of target engagement on the enzyme IRAK4, which plays a central role in the LPS signaling due to its role in the signaling downstream of toll-like receptor 4. A dose-dependent complete or near-complete suppression of TNFα mean values was not observed.

Repeated administration of zabedosertib over 10 days led to a clear reduction in LPS-stimulated TNFα release in almost all zabedosertib-treated participants (Figure 7). The reduction seemed to increase with dose and treatment duration. After the end of treatment, the TNFα release reached the baseline level again within 3 days (Figure 6).

IC₅₀ for IL-6 release inhibition after resiquimod stimulation in the in vitro human whole-blood assay was 86 (60–122) nM (mean and 95% CI). With the assumption that this concentration would result in 50% IRAK4 occupancy by zabedosertib in healthy volunteers, the estimated steady-state target occupancy at trough after 120 mg zabedosertib BID was on average 79% (72%–84%) (Figure 8). After the same daily dose given as a single dose, that is, 240 mg, the estimated steady-state target occupancy was clearly lower [67% (58%–74%)].

The three early phase 1 studies of zabedosertib described in this publication revealed a favorable safety and tolerability and pharmacokinetic profile for zabedosertib. Single oral doses of zabedosertib of up to 480 mg and repeated oral doses up to 200 mg twice daily for 10 days (taken as tablets) were well tolerated in healthy male volunteers, without dose-limiting toxicities or severe infections.

The PK profile of zabedosertib is characterized by increasing concentrations until 2–3 h after dosing and a plateau until approximately 6 h post-dose. For doses >60 mg, a sub-proportional increase in exposure was observed, with an absolute bioavailability of 74% at a dose of 120 mg. The nonlinearity in the pharmacokinetics could be adequately described with a one-compartmental population-pharmacokinetic model with first-order elimination, dose-dependent bioavailability, and capacity-limited binding in plasma. Explorative biomarker analysis revealed signals for target engagement or pharmacodynamic effects by reduction in the LPS-stimulated release of TNFα in an ex vivo whole-blood assay. Target occupancy at a dose of 120 mg BID was estimated to be 79%.

Across the three reported phase 1 studies, zabedosertib showed good safety and tolerability, without dose-limiting toxicities or severe infections in the healthy male volunteers. In accordance with the benign adverse event profile and favorable tolerability of the tested increasing doses of zabedosertib, vital signs, regular ECGs, and laboratory safety parameters were assessed without any signals or other noteworthy findings beyond those described above. This is in line with published data on other drugs inhibiting IRAK4, where no safety concerns in healthy volunteers or in patients were reported (Danto et al., 2019a⁠; Winkler et al., 2021). In contrast to an IRAK4-degrading molecule, no prolongations of QT-time in ECG were observed (Ackerman et al., 2023). Moreover, no signals for rhabdomyolysis were observed, which is in contrast to experiences with emavusertib (CA-4948) (a combined IRAK4/FLT3-inhibitor), for which a partial clinical (FDA)-hold was set in place due to a case of lethal rhabdomyolysis in an oncologic clinical study (treatment in ⁷ combination with azacitidine or venetoclax).

At the time when the study protocols for the studies reported here were drafted, no such safety data were available, and based on the mode of action and a potentially immune-compromising effect (Picard et al., 2007), different AESIs were defined and helped to assess the benefit-risk ratio of zabedosertib while conducting the study. In the early phase 1 program of zabedosertib with treatment durations of up to 10 days, no signals for a clinically meaningful immune suppression were found. This is supported by the fast return of TNFα levels within 72 h observed in the ex vivo LPS whole-blood assay (data available for the 200-mg BID-dose group), which suggests a quick reconstitution of the innate immune function, providing no evidence for sustained immune suppression in the study settings. The data may also point toward a quick suspension of the drug effect by zabedosertib, unlike what can be observed with molecules degrading IRAK4 or covalent binders.

The overall good safety and tolerability of zabedosertib observed in the studies reported here is in line with data collected in an up to 12-week treatment time frame ⁸ but cannot be predictive for longer-term treatment and, therefore, will have to be confirmed in a chronic treatment setting. A good prerequisite for the long-term use of zabedosertib is that there were no signals for gastrointestinal intolerability detectable across studies.

The bioavailability of zabedosertib was greater when the drug was taken in the tablet form than when it was taken in the liquid form (both taken as single doses in the fasted state). One explanation could be that the excipient Polyethylene glycol (PEG) used for the liquid formulation decreases the small intestinal transit time, which resulted in a reduced absorption from the small intestine (Basit et al., 2002⁠; Lentini et al., 2016). Another explanation could be that zabedosertib from the liquid formulation rapidly precipitates in the stomach due to its low solubility in water. The observed time to maximum plasma concentrations of approximately 5 h for the 15-mg liquid-formulation dose was also distinctively longer than the 2 h determined for the 15-mg tablet formulation in the same participants. Drug precipitates might have redissolved to a certain extent. This could explain the slow increase and late maximum of zabedosertib plasma concentrations after the administration of zabedosertib as liquid formulation.

The assessment of a potential food effect on the pharmacokinetics of zabedosertib revealed an increase in C_max by a factor of approximately 1.4 under fed conditions, that is, when zabedosertib was administered after a high-fat or moderate-fat meal compared to fasted conditions. The extent of absorption (AUC), however, was unaffected by the intake of food. Likewise, the time to maximum plasma concentrations and the relatively long half-life (∼26 h) remained unaffected by the presence or absence of concomitant food intake. The food effect on C_max may be explained by an enhanced solubility of zabedosertib, resulting in an increased absorption rate shortly after drug intake.

Following multiple-dose administration over 10 days (tablet intake 30 min after a light meal), a lack of dose proportionality, that is, a decrease in dose-normalized PK parameter values (AUC, C_max, and C_av) with increasing dose, was observed at higher dose. Similarly, no dose proportionality was observed for PK parameters of unbound zabedosertib. The time course of the zabedosertib trough concentrations indicated that steady-state conditions in plasma were reached after 2–3 days. It can therefore be assumed that the multiple-dose PK parameters determined reflect steady-state conditions.

The C_max after multiple dosing increased by a factor of 1.05 [120 mg QD] to 1.27 [60 mg BID] compared to single dosing. The mean accumulation ratio for AUC increased by a factor of 1.04 [120 mg QD] to 1.42 [60 mg BID], with lower accumulation observed at higher doses. This is less than the theoretical value of 3.65 estimated from the half-life (∼26 h) and the length of the dosing interval (12 h BID). The sub-proportional accumulation is also reflected in the linearity factor, which decreased with increasing dose, also indicating nonlinearity in PK.

The geometric mean f_u of zabedosertib determined ex vivo in pre-dose plasma samples and in plasma samples of placebo-treated participants ranged between 1.44% and 4.00%, confirming the in vitro data. The unbound fraction decreased within 24 h post-dose to some extent and reached pretreatment values at 96 h post-dose. There was a considerable variability in the f_u data, as indicated by the broad 95% prediction interval. As zabedosertib binds in plasma with high affinity to AGP, the unbound fraction is dependent on the plasma concentration of this acute-phase protein. The geometric mean AGP concentration determined in all verum-treated participants (0.612 g/L; CV: 7.08%; fasted conditions) did not significantly change after treatment with zabedosertib (e.g., at 24 h post-administration 0.585 g/L; CV: 6.55%). Nevertheless, the AGP concentration ranged among individuals from 0.41 to 0.91 g/L in the corresponding pretreatment samples, indicating that different AGP concentrations could have contributed to a certain extent to the observed variability in the unbound fraction of zabedosertib. AGP concentrations can be elevated in certain inflammatory diseases (Jaurila et al., 2020⁠; Farrag et al., 2024⁠; Agra et al., 2017). Possibly, this could lead to higher total concentrations of zabedosertib. However, given the high interindividual variability of AGP, the effect is expected to be minor.

The PK of zabedosertib after both single and multiple dosing could be adequately described by a one-compartmental model with first-order elimination, a dose-dependent bioavailability, and by assuming that zabedosertib binds to a high-affinity and low-capacity binding partner or component in the plasma with a binding affinity of 0.4 µM, that is, with an affinity like the in vitro AGP-binding affinity. The estimated first-order dissociation rate has a half-life of approximately 12 h, which is relatively long compared to in vitro estimates of plasma protein binding, with dissociation half-lives of seconds. The binding capacity of 14.9 µM is estimated to be very close to the median AGP concentration in the reported studies (15.2 µM). In addition, the individual estimates of the binding partner concentrations show a weak positive correlation (Pearson correlation coefficient of 0.28) with the observed AGP concentrations. However, it should be noted that the binding to AGP is a model assumption and does not necessarily reflect actual biology. The dose-dependent bioavailability could also partly be explained by the fact that zabedosertib is lipophilic (log P 2.4) and has a low solubility in water. However, at a dose of 120 mg, zabedosertib has a high oral bioavailability (73.9%), along with a low clearance (0.551 L/h) and long half-life (∼26 h), as demonstrated in the microtracer study (FE/abs.BA study). In addition to limited water solubility and capacity-limited protein binding, which may impact the pharmacokinetics of zabedosertib, another underlying mechanism for the observed nonlinearity in PK might be target binding to certain tissue compartments or any combination of the mechanisms. Currently, no mass balance data are available to provide further insights. Whatever the underlying mechanisms may be, our current data indicate that dose increases above 120 mg zabedosertib BID, administered as immediate-release tablets, did not result in clinically relevant higher exposure.

Currently, data on the potency of several IRAK4 inhibitors that are in clinical development in humans have been published. The IRAK4 inhibitor zimlovisertib, for example, showed positive proof of concept in a phase 2b trial in patients with active rheumatoid arthritis (Danto et al., 2019b). This drug has one of the highest potencies with an IC₅₀ of 0.2 nM (Danto et al., 2019b), whereas zabedosertib has an IC₅₀ of 8 nM; the half-life of zimlovisertib after single dosing was shorter than that of zabedosertib, but it could be increased to 38.8 ± 21.1 h with the development of a modified-release formulation (Danto et al., 2019a). For zabedosertib, this would not be required as its half-life is ∼26 h, and BID dosing was needed to maximize exposure. Due to the BID dosing, long half-life, and the plateau-like concentration–time profile of zabedosertib, sufficiently high exposure over 24 h was achieved.

As demonstrated by Winkler et al. (2021), in vitro IL-6 inhibition correlates with IRAK4 target occupancy. Based on this assumption and the reported pharmacokinetics of 120 mg BID zabedosertib, the calculated target occupancy is ∼80%. These observations, together with the positive proof of concept for zimlovisertib, support the assumption that zabedosertib can be dosed sufficiently high to achieve a clinical effect. This is furthermore supported by the observed distinct PD effects in humans in a proof-of-mechanism study, which demonstrated a clear suppression of systemic inflammation and suppression of local inflammation following a systemic LPS challenge and a local (akin) imiquimod challenge (Jodl et al., 2024).

In accordance with sponsor-internal standards in force at the time of the conduct of the phase 1 studies, only men were included although the envisaged indications of the drug include conditions that affect both men and women. Healthy women who would not benefit from participation in the studies were excluded from study participation for safety reasons because no reproductive toxicity studies had been conducted when the three studies were conducted and to control PK variability at this early stage of development. Furthermore, the vast majority of study participants were White/Caucasian although the protocols had no restrictions with regard to race and ethnicity. An intrinsic limitation of the three studies is the overall relatively small sample size of the individual dose groups and the exclusion of elderly study participants. Another limitation is that the phase 1 program in healthy volunteers covered only a treatment length of up to 10 days. Therefore, conclusions with regard to long-term safety and tolerability and predictions for a longer treatment time frame are limited.

Yet another limitation of the studies might be that the measured systemic exposures may not represent the desired exposures at the site of action (target tissue) of the respective target indication. Furthermore, an ex vivo whole-blood assay was performed in this phase 1 program to reflect the pharmacodynamic effects of zabedosertib, but from this assay, no dose decision could be derived for phase 2 due to a certain “floor effect” in cytokine suppression. Therefore, IL-6 correlation with IRAK4 target occupancy was used as substitute, and a separate proof-of-mechanism study was performed (Jodl et al., 2024).

The above-described exposure limitations define, from a PK perspective, 120 mg BID as the highest feasible dose for future studies with zabedosertib. The very good safety and tolerability of this dose over up to 10 days and the high projected target coverage, which is expected to lead to a clinical effect in the relevant target indications, suggest advancing zabedosertib further in its development.

Based on the projected target occupancy and the favorable pharmacokinetic and safety profile, as well as distinct pharmacodynamic effects in a proof-of-mechanism study, zabedosertib 120 mg twice daily was selected for further clinical development in patient studies.