For patients with amyotrophic lateral sclerosis who take oral riluzole tablets, approximately 50% experience alanine transaminase (ALT) levels above upper limit of normal (ULN), 8% above 3× ULN, and 2% above 5× ULN. BHV-0223 is a novel 40 mg rapidly sublingually disintegrating (Zydis) formulation of riluzole, bioequivalent to conventional riluzole 50 mg oral tablets, that averts the need for swallowing tablets and mitigates first-pass hepatic metabolism, thereby potentially reducing risk of liver toxicity. DILIsym is a validated multiscale computational model that supports evaluation of liver toxicity risks. DILIsym was used to compare the hepatotoxicity potential of oral riluzole tablets (50 mg BID) versus BHV-0223 (40 mg BID) by integrating clinical data and in vitro toxicity data. In a simulated population (SimPops), ALT levels > 3× ULN were predicted in 3.9% (11/285) versus 1.4% (4/285) of individuals with oral riluzole tablets and sublingual BHV-0223, respectively. This represents a relative risk reduction of 64% associated with BHV-0223 versus conventional riluzole tablets. Mechanistic investigations revealed that oxidative stress was responsible for the predicted ALT elevations. The validity of the DILIsym representation of riluzole and assumptions is supported by its ability to predict rates of ALT elevations for riluzole oral tablets comparable with that observed in clinical data. Combining a mechanistic, quantitative representation of hepatotoxicity with interindividual variability in both susceptibility and liver exposure suggests that sublingual BHV-0223 confers diminished rates of liver toxicity compared with oral tablets of riluzole, consistent with having a lower overall dose of riluzole and bypassing first-pass liver metabolism.
Amyotrophic lateral sclerosis (ALS) is a neurodegenerative disease characterized by the death of motor neurons that leads to progressive muscle weakness and difficulties in speaking, breathing, and swallowing. The median survival time from onset to death ranges from approximately 2 to 3 years (Bensimon and Doble, 2004). The precise cause of the disease is unknown.
Riluzole is a neuroprotective drug that is thought to act by blocking glutamatergic neurotransmission in the central nervous system (Doble, 1996). Clinical trials have demonstrated that riluzole prolongs survival and time to tracheostomy in patients with ALS (Bensimon et al., 1994; Lacomblez et al., 1996). The original pivotal studies that led to the approval of riluzole for the treatment of ALS demonstrated an extended median survival time of 2–3 months and a 43% reduction in death in favor of riluzole over the trial period of 18 months. More recent studies suggest that riluzole can extend survival to 6–18 months with a longer follow up time (Brooks and Sanjak, 2004; Georgoulopoulou et al., 2013; Mitchell et al., 2006). Mandrioli et al. demonstrated that strict adherence (> 90% of days treated from time of diagnosis) to riluzole treatment is important for better survival benefit (Mandrioli et al., 2018). Challenges to riluzole tablet adherence in ALS patients include the inability to swallow, gastrointestinal (GI) complications, and drug-induced abnormalities in liver function tests. Serum alanine aminotransferase levels greater than 3 times the upper limit of normal (ULN) have been observed in 10%–15% of patients receiving riluzole oral tablets (Bensimon and Doble, 2004).
BHV-0223 is a novel 40 mg sublingually dissolving Zydis formulation of riluzole that is bioequivalent to the riluzole 50 mg oral tablet formulation. Sublingual administration of riluzole may improve adherence in patients with dysphagia (difficulty swallowing). In addition, because of its sublingual route of administration, the ability of BHV-0223 to bypass first-pass metabolism, while achieving adequate systemic drug concentrations, will diminish overall hepatic drug burden and potentially lower the risk of liver toxicity.
Quantitative systems toxicology (QST) is an approach that integrates computational and experimental methods to understand and predict the toxicity of drugs throughout their development (Bloomingdale et al., 2017). DILIsym is a QST model of drug-induced liver injury (DILI) which includes multiple hepatotoxicity mechanisms (ie, bile acid accumulation, mitochondrial dysfunction, and oxidative stress) (Battista et al., 2018; Longo et al., 2019; Shoda et al., 2014; Woodhead et al., 2019). In this study, DILIsym was used to quantitatively and mechanistically compare the liver toxicity potential of oral riluzole tablets versus BHV-0223 by integrating clinical data and in vitro toxicity data (Figure 1). Responses to conventional oral riluzole tablets and BHV-0223 were analyzed in a simulated population (SimPops) which included variability to account for potential interpatient differences in key biochemical areas related to hepatotoxicity.
DILIsym version 6A was used to conduct the simulations in this article. DILIsym (http://www.dilisym.com, last accessed February 11, 2020) is a mathematical representation of DILI. DILIsym has been described previously, and many of the underlying equations have been made available in prior publications (Battista et al., 2018; Bhattacharya et al., 2012; Longo et al., 2017, 2019; Shoda et al., 2017, 2014; Woodhead et al., 2012, 2019; Yang et al., 2017). Briefly, DILIsym consists of several smaller submodels that are mathematically integrated to simulate an organism-level response. This work utilized submodels representing drug distribution, mitochondrial dysfunction and toxicity, bile acid physiology and pathophysiology, hepatocyte life cycle, and liver injury biomarkers.
SimPops are a collection of simulated individuals including parameter variability that reflects anthropometric and biochemical ranges. This study utilized an n = 285 normal healthy volunteer SimPops (Human_ROS_apop_mito_BA_v4A_1 SimPops) included in DILIsym v6A. This SimPops represents variability in parameters related to bile acid homeostasis, mitochondrial function, oxidative stress, apoptosis, and regeneration. A list of parameters varied in the human v4A-1 SimPops, as well as the sources used in the construction of the SimPops are shown in Supplementary Table S1.
SimCohorts are relatively small populations consisting of a subset of simulated individuals from existing SimPops in DILIsym. This work employed the human SimCohorts v4A-1-Multi16, which includes the baseline human as well as 15 individuals from the n = 285 human SimPops v4A-1. SimCohorts are computationally less expensive than the larger SimPops. For example, the simulation time required for simulations performed in the n = 16 SimCohorts for this study was approximately an order of magnitude lower than the time required for simulations performed in the n = 285 SimPops.
A physiologically based pharmacokinetic (PBPK) representation of riluzole was constructed within DILIsym to describe liver exposure upon conventional oral tablet and sublingual administration. The DILIsym PBPK model framework used for riluzole consists of compartments for liver, blood, muscle, gut, and other tissues. The structure of the DILIsym PBPK submodel has been discussed in detail elsewhere (Howell et al., 2012; Woodhead et al., 2012, 2014). Riluzole metabolism was represented by 1 metabolic pathway, representing the aggregate of all riluzole metabolic pathways. Riluzole metabolites were assumed not to contribute to liver toxicity (due to the lack of any in vitro hepatotoxicity data for riluzole metabolites), and thus were not tracked. The tissue distribution of riluzole was assumed to be perfusion-limited and was represented by partition coefficients. Parameters used in the PBPK submodel for riluzole are shown in Table 1. Details of the PBPK model are provided in Supplementary A.
|Riluzole blood to plasma||Dimensionless||1.1||Optimizationa|
|Riluzole gut to blood||Dimensionless||0.86||Optimizationa|
|Riluzole liver to blood||Dimensionless||2.0||Optimizationa|
|Riluzole muscle to blood||Dimensionless||0.3||Optimizationa|
|Riluzole other tissue to blood||Dimensionless||6.16||Optimizationa|
|Riluzole fraction unbound plasma||Dimensionless||0.04||96% protein-binding reported (BHV-0223 IB)|
|Riluzole molecular weight||g/mol||234.2||BHV-0223 IB|
|Riluzole renal clearance||ml/h/kg^0.75||371.9||Calculated from renal clearance reported in Le Liboux et al. (1997) (< 6 ml/min); validated with reported urinary excretion of unchanged parent|
|Riluzole gastric emptying rate, k(ge)||1/h||1.7||Optimizationa|
|Riluzole absorption rate, k(ab)||1/h||9.99||Optimizationa|
|Riluzole rate of elimination in feces||1/h||0.556||Set the first-order constant for gut absorption and fecal elimination from gut 9:1 (assuming approximately 90% absorption in humans)|
|Riluzole time for IV dose to become well-mixed in blood, k(IV)||1/h||6||Optimizationa|
|Km (Riluzole metabolite A)b||µmol/l||50||Optimizationa|
|Vmax (Riluzole metabolite A)b||nmol/h/kg^0.75||2 500 000||Optimizationa|
The PBPK representation for riluzole was based on available data for BHV-0223 and published studies of riluzole. Specifically, data on plasma riluzole exposure from a published PK study of riluzole (single 50 mg IV dose and single 100 mg oral tablet dose in healthy volunteers, Le Liboux et al., 1997) were used to optimize the PBPK model parameters (Table 1). The PBPK model was evaluated against clinical data from a completed phase 1 trial and previously published trials in healthy volunteers (Chandu et al., 2010; Le Liboux et al., 1997), including the PK study of ascending doses of riluzole (25, 50, or 100 mg dose BID).
Riluzole oral tablet PBPK parameters were used as a starting point for the representation of sublingual riluzole (BHV-0223) in DILIsym. For compounds administered sublingually, a portion of the dose is absorbed from the oral mucosa and a portion is swallowed and passes through the GI tract (Bartlett and van der Voort Maarschalk, 2012; Xia et al., 2015). Because the fraction swallowed for BHV-0223 is unknown, BHV-0223 PK data (ie, plasma riluzole concentrations after a single 35 mg BHV-0223 sublingual riluzole dose) were used to estimate the portion of sublingual riluzole that is absorbed from the oral mucosa and the portion that passes through the GI tract. Simulations were conducted assuming either 0%, 25%, or 50% of the sublingual dose is absorbed via the oral mucosa, and simulated plasma riluzole concentrations after a single 35 mg sublingual dose were compared with measured concentrations after a single 35 mg BHV-0223 dose.
Riluzole was assessed in in vitro assays for the 3 main hepatotoxicity mechanisms represented in DILIsym: mitochondrial dysfunction, oxidative stress, and bile acid transporter inhibition. To assess potential mitochondrial dysfunction signals for riluzole, cellular respiration assays were conducted using a Seahorse XFe96 Flux Analyzer in HepG2 cells incubated with various concentrations of riluzole for 1 or 24 h. The potential for riluzole to induce oxidative stress was assessed by high content screening using a fluorescent probe, dihydroethidium (DHE), in HepG2 cells incubated with various concentrations of riluzole for 6 or 24 h. In these whole cell-based assays, intracellular concentrations of riluzole were determined by LC/MS/MS analysis in parallel HepG2 cultures. Inhibitory effects of riluzole for bile acid transporters were assessed experimentally using membrane vesicles overexpressing a bile acid efflux transporter (ie, BSEP, MRP3, or MRP4) and CHO cells overexpressing NTCP. Detailed experimental methods are described in Supplementary B. Mitochondrial dysfunction and oxidative stress assays were performed by Cyprotex, Inc (Macclesfield, UK). Transporter inhibition assays were performed by Solvo Biotechnology (Budaors, Hungary).
For each of the in vitro assays conducted, the results were translated into DILIsym parameters for use in the simulations. For the bile acid transporter parameters, the estimated IC50 values for riluzole were used directly as the inhibition constants in DILIsym. Mode of inhibition was assumed to be mixed inhibition with α = 5. Although competitive and noncompetitive inhibition types may result in low and high extremes of potential bile acid accumulation, respectively, mixed inhibition with α = 5 leads to a median impact on bile acid accumulation. In addition, mixed inhibitors are more common compared with pure competitive or noncompetitive inhibitors (Howell et al., 2016; Longo et al., 2019; Watkins, 2019; Woodhead et al., 2014, 2017). For riluzole-mediated mitochondrial dysfunction, the assay results comparing intracellular concentrations and oxygen consumption rate (OCR) were recapitulated in MITOsym; the resulting parameters were translated into DILIsym parameters using translation factors involving exemplar compounds, a process which has been reported elsewhere (Yang et al., 2015). For riluzole-mediated oxidative stress, the assay results were reproduced using DILIsym by mimicking in vitro conditions; appropriate parameter values for the oxidative stress effects were identified by comparing simulation results with the measured data. Further details on the translation of the experimental data into DILIsym parameters are provided in Supplementary B.
DILIsym v6A was used to perform simulations for comparison of oral tablet and sublingual riluzole. The clinical protocols simulated were as follows:
For both sublingual and oral riluzole tablet clinical protocols, the following simulation types were run:
|Liver:Blood Partition Coefficient (Kb)||L:B Value at 0.5 h After Dosing|
|In vitro data HepG2 (seahorse media)a||In vitro data HepG2 (HepG2 media)a||In vitro data human HC (HC media)a||In vitro data human HC (HC media)b||In silico calculations (Rodgers and Rowland)c|
|Measured or calculated||9.8||1.7||6.5||35||0.6–2||4.4 (in rats)|
|Default DILIsym value||2||7.3 (ie, simulated value with Kb = 2 in DILIsym)|
|Increased DILIsym value for sensitivity analyses||10||43 (ie, simulated value with Kb = 10 in DILIsym)|
|Riluzole Dose and Duration||DILIsym Parameter Settings||Simulated Peak ALT > 3× ULN a||Simulated Peak ALT > 5× ULN a|
|Conventional oral tablets 50 mg BID for 12 weeks||Median PK, liver Kb 2||0/285||0/285|
|High PK, liver Kb 2||0/285||0/285|
|Median PK, liver Kb 10||0/285||0/285|
|High PK, liver Kb 10||11/285||3/285|
|Sublingual 40 mg BID for 12 weeks||Median PK, liver Kb 2||0/285||0/285|
|High PK, liver Kb 2||0/285||0/285|
|Median PK, liver Kb 10||0/285||0/285|
|High PK, liver Kb 10||4/285||2/285|
Simulation results from the PBPK representation of riluzole oral tablet and clinical plasma riluzole exposure data used for optimization of the PBPK model are shown in Figure 2. The simulated plasma area under curve (AUC) and plasma Cmax values were within 1.5-fold of those observed in clinical trials (Chandu et al., 2010; Le Liboux et al., 1997). The simulations reasonably captured the plasma pharmacokinetics of riluzole, based on comparisons with training data (Figure 3A) and comparisons with data that were not used in the optimization (validation data set; Figure 3B). Simulation results from both the median PK riluzole parameterization and the “High PK” riluzole parameterization, representing individuals with increased plasma riluzole exposure, compared with clinical data, are shown in Figure 4.
PK data (ie, plasma concentrations of BHV-0023 after a single 35 mg sublingual dose) were used to estimate the portion of sublingual riluzole that is absorbed via the oral mucosa and the portion that is swallowed and passes through the GI tract. (Note that data for the 35 mg sublingual dose was used, rather than the 40 mg approved dose for BHV-0223, due to the availability of PK data for the 35 mg sublingual dose at the time of this study.) Simulated plasma concentrations after a 35 mg sublingual dose were conducted, assuming 0%, 25%, and 50% of the dose is absorbed via the oral mucosa, respectively. Simulations with 0% of the sublingual dose absorbed via the oral mucosa and 100% passed through the GI tract underestimated observed plasma concentrations following a single 35 mg sublingual dose. Simulated plasma concentrations with 25% of the dose being absorbed from the oral mucosa and 75% of the dose passing through the GI tract reasonably approximated observed plasma riluzole concentrations following a single 35 mg sublingual dose (Figure 5).
Simulated plasma concentrations and simulated liver concentrations for a single 50 mg oral tablet dose of riluzole and for a single 40 mg sublingual dose (ie, the sublingual dose corresponding to Zydis formulation) are shown in Figure 6. Notably, whereas plasma concentrations for the 2 dosing regimens (40 mg sublingual vs 50 mg oral tablet) are quite similar (ratio of 0.97 for predicted plasma AUC following 40 mg sublingual vs 50 mg oral tablet), the predicted hepatic exposure for the 40 mg sublingual dose is lower than the predicted hepatic exposure for the 50 mg oral tablet dose (ratio of 0.80 for predicted liver AUC following 40 mg sublingual vs 50 mg oral tablet).
In the mitochondrial respiration assay, riluzole decreased the OCR in a concentration-dependent manner after 1 and 24 h incubation. These data suggest that riluzole is a mitochondrial electron transport chain (ETC) inhibitor. To define the DILIsym parameters for riluzole-mediated mitochondrial ETC inhibition, the 1 h in vitro data were simulated within MITOsym and subsequently translated into DILIsym values as described in the Materials and Methods section. The optimized ETC inhibition parameter value listed in Table 4 reasonably recapitulated riluzole effects on the OCR as shown in Figure 7.
|Mechanism||DILIsym Parameter Name||Unit||Value b|
|Mitochondrial dysfunction||Coefficient for ETC inhibition||µM||382|
|Oxidative stress||RNS/ROS production rate constant||ml/nmol/h||6 × 10−4|
|Bile acid transporter inhibition||BSEP inhibition constanta||µM||200|
|NTCP inhibition constanta||µM||No inhibition|
|Inhibition constant for basolateral effluxa||µM||125|
Riluzole increased reactive nitrogen and oxygen species (RNS/ROS) in a concentration-dependent manner after 24 h incubation, but not following 6 h incubation (Figure 8; 6 h data not shown). These data suggest that riluzole can elicit oxidative stress. DILIsym parameters for riluzole-induced production of RNS/ROS were optimized to recapitulate intracellular concentrations versus cellular RNS/ROS data by simulating in vitro -like conditions within DILIsym (Figure 8, Table 4).
Experimental data indicated that riluzole inhibited BSEP and MRP4; riluzole had no effect on MRP3 or NTCP. Inhibition constants for riluzole are presented in Table 4.
Clinical dosage regimens of riluzole oral tablet (50 mg BID) and sublingual riluzole (40 mg BID) were simulated for 12 weeks in the v4A_1 SimPops. In the SimPops simulations, no ALT elevations > 3× ULN were predicted for either dosing protocol (oral tablet or sublingual) with Median PK and high or default liver exposure assumptions (Table 3). In the simulations with High PK and high liver exposure (liver Kb 10), the predicted incidence of ALT elevations was higher for oral tablet dosing (11 of 285 individuals) versus sublingual dosing (4 of 285 individuals).
Mechanistic Investigation Simulations were conducted to evaluate the contributions of each hepatotoxicity mechanism (ie, riluzole-mediated mitochondrial ETC inhibition, riluzole-mediated bile acid transport inhibition, and riluzole-mediated ROS) to the simulated ALT elevations, as described in the Materials and Methods section. To investigate the importance of each mechanism to predicted hepatotoxicity, the oral tablet dosing protocol with the High PK scenario and high liver Kb (ie, 10) was simulated in SimCohorts (v4A-1-Multi16) with sequential omission of 1 potential mechanistic contributor at a time. With removal of the ROS mechanism, no ALT elevations were predicted, indicating the oxidative stress is required for simulated ALT elevations (Table 5). In contrast, there was no impact on the incidence of ALT elevations with removal of either mitochondrial ETC inhibition or removal of bile acid transport inhibition, indicating that these 2 mechanisms are not required for simulated ALT elevations (Table 5). These results demonstrate that the primary driver of ALT elevations in the DILIsym riluzole simulations is oxidative stress.
|Riluzole Dose and Duration||DILIsym Parameter Settings||Mechanisms||Simulated Peak ALT > 3× ULN a||Simulated Peak ALT > 5× ULN a|
|Conventional oral tablets 50 mg BID for 12 weeks||High PK, liver Kb 10||All||3/16||1/16|
|No mitochondrial toxicity||3/16||1/16|
|No bile acid transport inhibition||3/16||1/16|
Riluzole is a medication that has been developed to treat ALS. BHV-0223 is a novel 40 mg sublingually dissolving Zydis formulation of riluzole that is bioequivalent to the riluzole 50 mg oral tablet formulation. Because of its sublingual route of administration, BHV-0223 first-pass metabolism with BHV-0223 is mitigated, achieving adequate drug concentrations with reduced hepatic exposure. A sublingual formulation of riluzole may be particularly beneficial to ALS patients, because these patients often have difficulty swallowing.
DILIsym is a QST model of DILI that can be applied to predict hepatotoxicity based on in vitro mechanistic data and in vivo clinical data and can provide insight into the underlying mechanisms responsible for DILI. DILIsym analyses performed in this study support the advantage of the sublingual administration of riluzole.
Specifically, PBPK representations of riluzole oral tablet and sublingual riluzole (BHV-0223) developed in DILIsym predicted similar plasma concentrations following a single 50 mg oral tablet dose of riluzole compared with a single 40 mg sublingual dose of riluzole. While plasma concentrations for the 2 dosing regimens were similar, the predicted hepatic exposure for the 40 mg sublingual dose was lower than the predicted hepatic exposure for the 50 mg oral tablet dose, (Figure 6), consistent with a subsequent study demonstrating that 40 mg BHV-0223 administered sublingually is bioequivalent to a 50 mg dose of a conventional riluzole oral tablet.
To compare the liver toxicity potential of the oral tablet and sublingual formulations of riluzole, DILIsym simulations were performed for both sublingual (40 mg BID sublingual riluzole) and oral tablet (50 mg BID PO riluzole) clinical protocols in a simulated population (SimPops) that includes variability in parameters relevant to hepatotoxicity mechanisms. Simulations were performed with Median PK and with a “High PK” parameterization (Figure 5) to represent individuals with high plasma riluzole exposure. In addition, given uncertainty in the liver to blood partition coefficient (liver Kb) value for riluzole, simulations were performed with either a liver Kb value of 2 (default value) or an increased liver Kb value of 10.
No ALT elevations > 3× ULN were predicted in the SimPops for either dosing protocol (oral tablet or sublingual) with Median PK and high or default liver exposure assumptions (Table 3). For simulations with High PK and high liver exposure (liver Kb value of 10), the predicted incidence of ALT elevations was higher for oral tablet dosing (11 of 285 individuals) versus sublingual dosing (4 of 285 individuals). This represents a relative risk reduction of 64% associated with sublingual administration of riluzole versus conventional riluzole tablets.
The low incidence of ALT elevations (ie, 11/285 or 4% of simulated individuals) in the SimPops simulations with High PK and high liver exposure for the riluzole oral tablet dosing protocol (Table 3) approximates the low incidence of ALT elevations reported for ALS patients treated with riluzole. Serum ALT elevations > 3× ULN have been observed in 10%–15% of patients with ALS taking riluzole (Bensimon and Doble, 2004). The slight underprediction of the incidence of ALT elevations may be due to the use of a normal healthy volunteer SimPops which does not capture any potential hepatic dysfunction in ALS patients prior to treatment. A high incidence of mild liver dysfunction in ALS patients has been reported (based on the evaluation of liver function tests in 37 ALS patients) (Nakano et al., 1987).
Mechanistic investigation simulations were performed to assess the contributions of each hepatotoxicity mechanism (ie, riluzole-mediated mitochondrial ETC inhibition, riluzole-mediated bile acid transport inhibition, and riluzole-mediated oxidative stress). The oral tablet dosing protocol with High PK and liver Kb value of 10 was simulated in an n = 16 SimCohorts with sequential removal of 1 potential hepatotoxicity mechanism at a time. With omission of the ROS mechanism, no ALT elevations were predicted, demonstrating that riluzole-mediated oxidative stress is required for simulated ALT elevations (Table 5). In contrast, removal of either riluzole-mediated mitochondrial ETC inhibition or removal of riluzole-mediated bile acid transport inhibition did not impact the incidence of ALT elevations (Table 5). These results indicate that oxidative stress is the primary driver of ALT elevations in the DILIsym riluzole simulations.
As noted above, 1 limitation of this study is that the normal healthy volunteer SimPops utilized for this study does not capture any disease characteristics of ALT which may contribute to susceptibility to liver injury. An additional limitation of this study is that, as described above, this study utilized a relatively simple approach (ie, median PK vs High PK parameterization) to assess the potential impact of pharmacokinetic variation. However, whereas the full range of potential riluzole exposure levels was not simulated, the approach utilized for this study accomplished the objective of assessing the sensitivity of the hepatotoxic response to different exposure levels.
In conclusion, using DILIsym to integrate clinical and mechanistic in vitro toxicity data for riluzole allowed for a quantitative comparison of the liver toxicity potential of the oral tablet and sublingual formulations of riluzole. DILIsym analyses suggest that sublingual BHV-0223 has reduced hepatic exposure and, consequently, less risk of liver toxicity compared with riluzole oral tablets.
The authors would like to acknowledge the members of the DILI-sim Initiative for their support in the development of the DILIsym model. More information on the DILI-sim Initiative can be found at www.dilisym.com.