Dose‐dependent acute liver injury with hypersensitivity features in humans due to a novel microsomal prostaglandin E synthase 1 inhibitor

Abstract

Aims

LY3031207, a novel microsomal prostaglandin E synthase 1 inhibitor, was evaluated in a multiple ascending dose study after nonclinical toxicology studies and a single ascending dose study demonstrated an acceptable toxicity, safety and tolerability profile.

Methods

Healthy subjects were randomized to receive LY3031207 (25, 75 and 275 mg), placebo or celecoxib (400 mg) once daily for 28 days. The safety, tolerability and pharmacokinetic and pharmacodynamic profiles of LY3031207 were evaluated.

Results

The study was terminated when two subjects experienced drug‐induced liver injury (DILI) after they had received 225 mg LY3031207 for 19 days. Liver biopsy from these subjects revealed acute liver injury with eosinophilic infiltration. Four additional DILI cases were identified after LY3031207 dosing had been stopped. All six DILI cases shared unique presentations of hepatocellular injury with hypersensitivity features and demonstrated a steep dose‐dependent trend. Prompt discontinuation of the study drug and supportive medical care resulted in full recovery. Metabolites from metabolic activation of the imidazole ring were observed in plasma and urine samples from all subjects randomized to LY3031207 dosing.

Conclusions

This study emphasized the importance of careful safety monitoring and serious adverse events management in phase I trials. Metabolic activation of the imidazole ring may be involved in the development of hepatotoxicity of LY3031207.

What is Already Known about this Subject

• Drug‐induced liver injury (DILI) with hypersensitivity response is uncommon and usually not dose dependent.

What this Study Adds

• Acute DILI with a hypersensitivity pattern occurred with high incidence and in a dose‐dependent manner in patients randomized to LY3031207.

• The pathogenesis of hepatic injury may have involved formation of reactive metabolites from the oxidation of the imidazole ring. These data suggest that certain imidazole rings be considered as a potential structure alert for hepatotoxicity.

• Careful attention to safety monitoring throughout phase I trials is recommended.

Introduction

Drug‐induced liver injury (DILI) is a serious issue for health care providers and patients; and has been a major cause of drug withdrawal and nonapproval by regulatory authorities. DILI has been well described for nonsteroidal anti‐inflammatory drugs (NSAIDs) 1, the most commonly prescribed medications for inflammatory pain. It has been hypothesized that the efficacy of these agents may be attributed to their ability to suppress inducible prostaglandin E₂ (PGE₂) synthesis 2, whereas their adverse effects may be due to their inhibition of other vital prostanoids 3, 4.

Microsomal PGE synthase (mPGES‐1) is an inducible PGE₂ synthase. In mPGES‐1 knockout mice, inflammation and painful behaviour were reduced in multiple pain models 5 without significant gastrointestinal 6 or cardiovascular pathology 7.

LY3031207, a selective mPGES‐1 inhibitor, was developed as a potential alternative to NSAIDs. After single doses in healthy subjects, LY3031207 demonstrated good tolerability up to 900 mg; and dose/concentration‐dependent inhibition of inducible PGE₂ synthesis as measured by an ex vivo human whole blood PGE2 synthesis assay 8, with a 50% maximal inhibitory concentration (IC50) of 910 ng ml^–1 (95% CI 706–1180 ng ml^–1), which translates to an 80% maximal inhibitory concentration (IC80) of 3640 ng ml^–1 (Figure S1). The pharmacokinetic and pharmacodynamic modelling based on these data suggested that, assuming at least 80% target engagement is required to be efficacious for osteoarthritis treatment 2, a LY3031207 dose around 225 mg once daily may be the therapeutic dose. In contrast to 400 mg single dose of celecoxib, LY3031207 did not inhibit thromboxane (Table S1, Figure 2a) and prostacyclin synthesis (Figure S2b) in vivo. Clinical development of LY3031207 was terminated after six cases of acute liver injury were identified in a multiple ascending dose study. Toxicological data and clinical presentations of DILI following LY3031207 administration are presented.

Methods

Nonclinical study design

The nonclinical toxicology profile of LY3031207 was established using a battery of safety pharmacology, genotoxicity and repeat‐dose studies. Repeat‐dose studies in Sprague–Dawley rats and beagle dogs were conducted according to US Food and Drug Administration Good Laboratory Practice regulations. Studies employed once‐daily oral dosing and included vehicle control groups. One‐month studies in rats evaluated 75, 200 and 1000 mg kg^–1 body weight and in dogs evaluated 50, 150 and 500 mg kg^–1. Additionally, rats were given LY3031207 for 6 months at doses of 5, 200 and 500 mg kg^–1, and dogs were given doses of 10, 30 and 100 mg kg^–1 for 3 months. LY3031207 is active against dog mPGES‐1 but not against the rat enzyme; therefore, the dog was the more relevant toxicology species.

Clinical study design

A subject/investigator‐blind, parallel‐group, multiple‐dose, dose‐escalation study was designed to test the safety and tolerability of LY3031207 at a dose range of 25–450 mg once daily. Healthy human subjects were sequentially assigned to Cohorts I–IV (approximately 13 per group) and randomized to placebo, celecoxib or LY3031207 once daily for 28 days within each cohort. The study was conducted in a single contract research unit (CRU). Unblinded CRU pharmacy staff prepared the blinded study drug for each subject based on computer‐generated randomization codes. To maintain adequate blinding, the investigators and nursing staff at the CRU had no access to the randomization codes. In addition, celecoxib capsules were encapsulated so that the appearance of all study drugs (LY3031207, placebo, and celecoxib) were identical, and subjects within each cohort took the same number of capsules. Ascending doses of LY3031207 (25, 75, 225 and 450 mg) were to be tested sequentially. Upon review of all available safety data from the previous cohort(s), decisions to escalate the dose were made jointly by the investigator and sponsor. Study drug dosing to individual subjects and dose escalation would stop if any predefined early discontinuation criteria based on safety were met. The primary objective was to evaluate the safety and tolerability of LY3031207 as measured by the incidence rates of serious adverse events and all adverse events. The secondary outcomes of the study included plasma LY3031207 concentrations and the effects of LY3031207 on prostaglandin synthesis in vivo. Subjects signed informed consent before any procedures were performed, and the study protocol conformed to the ethical guidelines of the 1975 Declaration of Helsinki as reflected in a priori approval by Schulman Associates Institutional Review Board.

Eligible subjects included healthy males and females of nonchildbearing potential aged 18–60 years. They were normotensive with a body mass index between 18.5 and 32.0 kg/m². Subjects were excluded if they had active or recent history of cardiovascular, gastrointestinal, renal, respiratory or neoplastic illness; viral hepatitis or other active hepatic illness; or heavy alcohol or recreational drug use. Subjects were not taking concomitant medication at entry. Paracetamol was allowed for the treatment of minor adverse events as needed at a total daily did not to exceed 2000 mg.

Eligible subjects received their first (Day 1) and last two doses of the study drug at the CRU and the remaining doses as outpatients. On Days 5, 12 and 19, subjects underwent safety evaluations at the CRU outpatient clinic. Plasma LY3031207 concentrations were measured using a validated liquid chromatograph mass spectrometry method. Plasma and urine from subjects who received LY3031207 were qualitatively analysed using liquid chromatograph/high‐resolution mass spectrometry to determine preliminary metabolic profiles and evidence of reactive metabolite formation. Subjects with abnormal liver tests underwent repeat liver tests and viral hepatitis panel (A, B, C and E), autoimmune serology, and hepatobiliary ultrasound. Serum concentrations of immunoglobulins were measured using a commercially available assay.

Sample collection

At times specified in the protocol schedule, approximately 2 ml of venous blood were collected from each subject in tubes containing K₂ ethylene diamine tetra‐acetic acid (EDTA) anticoagulant. The EDTA‐blood tubes were placed on ice until centrifuged at approximately 1500–2000 × g for 15–20 minutes to produce plasma samples. These plasma samples were then stored frozen at –20°C to –70°C until analysed for LY3031207 concentrations. Pharmacokinetic plasma samples remaining after bioanalyses of LY3031207 were used for exploratory LY3031207 metabolite identification.

Urine was collected at intervals specified in the study protocol. Total amounts of urine collected were recorded before the samples were stored frozen at –20°C to –70°C until analysed.

Analysis for LY3031207 concentrations in plasma

Plasma samples were analysed for LY3031207 using a validated liquid chromatography with tandem mass spectrometric detection method (LC/MS–MS). The standard curve range was 0.50 to 1000 ng ml^–1, with a sample injection volume of 0.1 ml. Samples above the standard curve range were diluted. The interassay accuracy (% relative error) during validation ranged from –1.0 to 0.8%. The interassay precision (% relative standard deviation) during the validation ranged from 1.6 to 9.9%. Aliquots of blank human plasma from six different individuals were tested for endogenous interferences. In all cases, the LY3031207 and internal standard regions were free from significant interference (<20.0% of the response from the 0.50 mg ml^–1 lower limit of quantitation and <5.0% of internal standard response in the control zero sample). LY3031207 and the internal standard were extracted from EDTA‐treated human plasma by supported liquid extraction (Isolute SLE+ supported liquid extraction plate, 200 mg). The eluted liquid was evaporated under nitrogen, and the residue was reconstituted using an acetonitrile:water (25:75,v/v) solution and then analysed by LC/MS–MS with positive electrospray ionization. The analytes were separated using a Luna C18(2) column (5 μm, 50 × 20 mm, Phenomenex) and a mobile phase gradient programme with 30 mmol l^–1 ammonium formate in 0.1% formic acid and 0.1% formic acid in acetonitrile. Quality control (QC) samples with LY3031207 concentrations of 1.5, 40 and 800 ng ml^–1 (5000 ng ml^–1 dilution QC) were placed in the analytical run to ensure the method suitably.

Pooling scheme for metabolite identification

Plasma and urine samples were prepared as Hamilton 8 pools for area under plasma concentration vs. time curve for the time 0–24 h after dosing (AUC_0–24h) for Days 1 and 28 (separating non‐DILI and DILI subjects). Due to early discontinuation of the study, none of the subjects in the 225‐mg group completed the planned 28‐day dose; therefore, Day 12 plasma samples (single time point collection) were used for that group's multiple doses.

Method for metabolite identification

Pooled plasma samples were prepared by acetronitrile protein precipitation followed by supernatant evaporation. Dried plasma extracts were reconstituted with 10 mmol l^–1 ammonium formate/acetonitrile (70:30, v/v) before analysis. Pooled urine samples were evaporated to half original volume and injected directly onto the analytical column (Synergi Polar RP 80A, 4.6 mm inner diameter × 250 mm; Phenomenex, Torrence, CA, USA). Analytical conditions used a mobile phase gradient program with 10 mmol l^–1 ammonium formate (Mobile Phase A)/acetonitrile (Mobile Phase B) and a flow rate of 1 ml min^–1 (with 20% split flow to MS). Putative metabolite structures were identified using accurate mass liquid chromatography/mass spectrometry and liquid chromatography/tandem mass spectrometry with positive electrospray ionization (LTQ Orbitrap XL, Thermo Scientific, Waltham, MA, USA) as well as metabolite retention time comparisons. Estimates for amounts of each individual metabolite in the extracted ion chromatograms were reported as relative to the mass spectrometric peak height of LY3031207.

Statistical methods

The sample size was customary for phase I studies evaluating safety, pharmacokinetic, and/or pharmacodynamic parameters and was not based on formal hypothesis tests. Subjects who received at least one dose of study drug were included in the analysis. Retrospective analyses in the subset of subjects receiving 75 mg (n = 10) or 225 mg (n = 9) LY3031207 compared demographic characteristics of subjects with DILI with those without DILI. Categorical characteristics were compared using Fisher's exact test and continuous characteristics by analysis of variance. Immunoglobulin concentrations were log‐transformed prior to analysis, and the ratio to baseline was analysed by analysis of covariance adjusting for baseline, with fixed effect of DILI/non‐DILI outcome. Geometric mean ratios to baseline were presented for each DILI/non‐DILI outcome group with 95% confidence intervals, along with the ratio comparison of DILI to non‐DILI.

Nomenclature of targets and ligands

Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY 9, and are permanently archived in the Concise Guide to PHARMACOLOGY 2015/2016 10.

Results

Nonclinical toxicology

In the 1‐month rat study, 1000 mg kg^–1 exceeded a maximum tolerated dose; morbidity and unscheduled euthanasia in several rats was attributed to treatment‐related renal toxicity. In the 6‐month rat study, the no‐observed‐adverse‐effect level (NOAEL) was the high dose of 500 mg kg^–1. Hepatocellular hypertrophy attributed to microsomal enzyme induction, a common finding in rats administered xenobiotics 11, occurred at ≥200 and ≥50 mg kg^–1 in the 1‐ and 6‐month studies, respectively.

The dog was the more sensitive species for LY3031207‐induced toxicity. In the 1‐month dog study, maximum tolerated dose was exceeded at ≥150 mg kg^–1, with early euthanasia due to conditional decline (vomiting/reduced food consumption) and multiorgan toxicity, including gastrointestinal, renal and liver effects. Minimal or mild hepatocellular degeneration/necrosis occurred at the high dose of 500 mg kg^–1, unaccompanied by significant increases in group mean serum alanine aminotransferase (ALT), aspartate aminotransferase (AST), alkaline phosphatase (ALP) or total bilirubin values. The NOAEL was 50 mg kg^–1, and there were no notable microscopic liver findings or increases in liver enzymes. In the 3‐month dog study, the 100 mg kg^–1 dose exceeded the maximum tolerated dose due to conditional decline attributed to gastrointestinal toxicity. Liver‐related microscopic findings at nontolerated high dose levels consisted of minimal‐to‐mild centrilobular hepatocellular degeneration, characterized primarily by small disorganized hepatocytes without inflammatory infiltrates and bile stasis with occasional indistinct centrilobular canalicular bile plugs. Increased ALT and/or ALP in individual dogs were inconsistently present and mild (≤4× increase from pretest values). No microscopic evidence of hepatocellular degeneration, bile stasis, or changes in liver‐related serum biochemical parameters occurred at the NOAEL of 30 mg kg^–1 in the 3‐month study, with the exception of a mild increase in ALP (3.4× increase from pretest value) in a single female (Table S2).

Clinical presentation of DILI

Twenty‐seven healthy subjects received at least one dose of 25 (n = 8), 75 (n = 10) or 225 mg (n = 9) of LY3031207 once daily; and six each received celecoxib 400 mg once daily or placebo. The baseline demographic characteristics for all treatment groups were similar (Table 1). Administration of the 225 mg dose was discontinued and dose escalation to 450 mg was abandoned according to protocol‐defined stopping criteria. Based on a clinical definition described by Aithal et al. 12, DILI was identified in two patients taking the 225 mg dose. Sixteen subjects withdrew from the study: two each in the celecoxib and placebo arms due to early trial discontinuation; three in the 75‐mg dosing cohort, with one due to family reasons and two due to early trial discontinuation; and all nine subjects in the 225‐mg LY3031207 dosing group, with two due to DILI and seven due to early trial discontinuation.

Table 1.

Demographics of patients in the multiple ascending dose study of LY3031207

	LY3031207 25 mg (N = 8)	LY3031207 75 mg (N = 10)	LY3031207 225 mg (N = 9)	Placebo (N = 6)	Celecoxib (N = 6)
Age (years), mean (SD)	37.8 (11.8)	39.0 (11.5)	45.1 (14.7)	52.7 (5.8)	44.0 (7.8)
Female sex, n (%)	1 (12.5)	2 (20.0)	4 (44.4)	2 (33.3)	0
Race, n (%)
White	5 (62.5)	6 (60.0)	3 (33.3)	2 (33.3)	4 (66.7)
Black	0	0	0	0	0
Asian	2 (25.0)	2 (20.0)	3 (33.3)	3 (50.0)	2 (33.3)
Multiple	1 (12.5)	2 (20.0)	3 (33.3)	1 (16.7)	0
Weight (kg), mean (SD)	78.8 (11.1)	72.3 (13.6)	65.8 (11.5)	71.6 (11.7)	78.0 (6.7)
BMI (kg/m 2 ), mean (SD)	24.7 (2.7)	23.8 (3.0)	23.3 (3.6)	25.1 (2.5)	23.7 (3.2)

BMI, body mass index; N, number in treatment group; n, number within category; SD, standard deviation.

On the Day 19 safety visit, two female subjects (Patients 1 and 2) who were randomized to receive 225 mg LY3031207 had AST elevations of 14‐ and 20‐fold upper limit of normal (ULN) and ALT elevations of 17‐ and 27‐fold ULN, respectively (Figure 1A). Dosing to all subjects was discontinued, and intensive hepatic monitoring was initiated. After discontinuation of study drug dosing, serum AST and ALT increases were noted in three additional subjects on 225 mg LY3031207 (Patients 3, 4, and 5). A sixth subject who completed 28 days of 75‐mg dosing had a solitary increase in ALT (Patient 6; Figure 1B). Viral and autoimmune serology was negative in all six patients.

Figure 1.

Time course of alanine aminotransferase change in six patients who developed drug‐induced liver injury after LY3031207 treatment. (A) Patients 1 and 2. (B) Patients 3 through 6. Colour‐matched arrows indicate the day of last dose. Day 1 was the day when the first dose of study drug was administered. All study days are in reference to Day 1. ALT, alanine aminotransferase

Patients 1 and 2 reported low‐grade fever, nausea, abdominal pain, and loss of appetite approximately 2 days before the aminotransferase increases were observed. On Days 22 and 21, respectively, Patients 1 and 2 were admitted to a local hospital, and N‐acetylcysteine infusions were initiated empirically. Patient 2 received N‐acetylcysteine for 7 days, but Patient 1 developed generalized urticaria and stopped N‐acetylcysteine after 4 days. Liver biopsies were performed. Histopathological examination of both biopsies revealed acute portal and lobular inflammation with numerous eosinophils and microgranulomas, Zone 3 cholestasis, and prominent lobular hepatocyte necrosis consistent with DILI (Figure 2). Both patients were discharged approximately 1 week later and monitored as outpatients until their liver test values returned to predose ranges.

Figure 2.

Photomicrographs of liver biopsy. Similar histopathological changes were present in the biopsies from Patient 1 and Patient 2. (A) Porta/zone 1 (P) and centrilobular/zone 3 (C) inflammatory infiltrates composed primarily of macrophages with eosinophils were accompanied by centrilobular/zone 3 hepatocyte necrosis. (B) A higher magnification illustrates centrilobular/zone 3 inflammatory infiltrates dominated by macrophages and eosinophils, as well as cholestasis characterized by intracellular amber pigment. The inset more clearly depicts eosinophils and intracellular bile pigment. Neither fibrosis nor steatosis was evident in the biopsies. Specific stains were negative for increased iron and α‐1 inclusions (data not shown)

Of the other four patients, three developed DILI symptoms after LY3031207 dosing at 225 mg daily had been stopped early (Figure 1B), with presentation similar to that of Patients 1 and 2. Patient 3 reported generalized urticaria, and Patients 4 and 5 reported vague abdominal pain, nausea and loss of appetite. Patient 6, who had completed 28 days of 75 mg LY3031207 daily when modest increases in ALT were noted, remained asymptomatic throughout his monitoring period (Figure 1B). These four patients were monitored as outpatients until their liver test values returned to predose ranges.

The clinical presentation of all six cases was strikingly similar (Table 2): hepatocellular injury pattern in five patients (R ratio between 10 and 38) 12 and mixed hepatocellular–cholestatic in the other (R ratio of 4); and remarkable eosinophilia (6–21%, normal range 0–6%). Two patients presented with generalized urticaria. Four patients were female, and all but one were aged >50 years at study entry. Retrospective analysis showed that females and older subjects were more susceptible to DILI than males and younger subjects (P ≤ 0.05; Table S3). All six cases of DILI were judged to be related to LY3031207 dosing based on negative viral hepatitis serologies, negative serology for autoimmune hepatitis, limited alcohol consumption during the trial, and denial of the use of medications with known hepatotoxicity (including paracetamol).

Table 2.

Key information regarding LY3031207‐associated drug‐induced liver injury

ID a	Age (years)	Weight (kg)	Race	Sex	LY3031207 (mg)	Dosing duration (day)	Total dose (mg)	Urticaria	ALT Peak b	ALT c (xULN)	ALP c (xULN)	R ratio c	Total bilirubin c , d (μM)	Eosinophilia c , e (%)
1	54	61	Asian	F	225	20	4500	Yes	21	45.58	1.21	38	27	21
2	58	56	Native Hawaiian	F	225	19	4275	No	19	27.23	1.48	18	21	20
3	32	73	White	M	225	22	4950	Yes	33	5.76	0.56	10	21	10
4	57	53	Asian	F	225	12	2700	No	30	14.90	0.76	20	17	19
5	59	50	Asian	F	225	12	2700	No	18	19.16	1.17	16	14	14
6	53	88	White	M	75	28	2100	No	34	3.31	0.81	4	26	6

^aThe patient ID was assigned based on the chronological order of the DILI diagnosis.

^bDay 1 was the study day when the first dose of study drug was administered. ALT peak was the study day relative to Day 1 when the highest ALT concentrations were observed.

^cULN is defined based on the reference value of the local clinical laboratory. ALT (xULN), total bilirubin, and eosinophils are reported as a peak value; R ratio refers to ratio ALT:ALP.

^dNormal range: 0–21 μmol l^–1;

^eNormal range: 0–6%.

ALP, alkaline phosphatase; ALT, alanine aminotransferase; DILI, drug‐induced liver injury; F, female; ID, identification; M, male; ULN, upper limit of normal.

The total dose of LY3031207 received by these patients ranged from 2100 mg over 28 days (Patient 6) to 4950 mg over 22 days (Patient 3). Evaluation of the incidence of DILI by LY3031207 dose demonstrated a clear dose‐dependent increase in the likelihood of DILI with increasing dose (Table 3). Despite this trend, plasma concentrations of LY3031207 in DILI patients were comparable with subjects in the same dose group who did not develop DILI (Figure 3). In addition, the observed plasma concentrations of LY3031207 were lower than the NOAEL exposure in both dog and rat (Table S4).

Table 3.

Dose‐dependent increase in drug‐induced liver injury incidence after LY3031207 dosing

Treatment	Dose (mg)	Number of subjects	Subjects with DILI (%)	Subjects with ALT increase (%)	Subjects with eosinophilia (%)
Placebo	NA	6	0	2 (33)	0
Celecoxib	400	6	0	0	2 (33)
LY3031207	25	8	0	0	0
LY3031207	75	10	1 (10)	4 (40)	3 (30)
LY3031207	225	9	5 (56)	6 (67)	7 (78)

ALT, alanine aminotransferase; DILI, drug‐induced liver injury; NA, not applicable.

Figure 3.

Time–concentration profile of patients with and without drug‐induced liver injury. (A) Observed LY3031207 plasma concentrations vs. time in subjects who received once daily 225 mg dose with (n = 5, red triangles, geometric means were connected with red solid line) and without drug‐induced liver injury (n = 4, blue circles, geometric means were connected with blue dashed line) on Day 1 and from predose samples through Day 12. (B) Sampling times in plots were slightly shifted to better show the comparison of drug concentrations from subjects with and without drug‐induced liver injury. The last plasma samples were obtained on Day 12 for this cohort. Conc, concentration; DILI, drug‐induced liver injury; w/o, without

Concentrations of serum immunoglobulins (Ig) were measured before and after LY3031207 treatment to further explore the potential immune‐mediated mechanism of LY3031207‐induced liver injury (Table 4). Compared with subjects who did not develop DILI, DILI patients showed increased IgE concentrations (P = 0.003). No significant changes in other immunoglobulin concentrations were noted.

Table 4.

Immunoglobulin levels in patients with and without drug‐induced liver injury

Endpoint	Patient group	n	Geometric mean [95% CI]	Ratio (non‐DILI vs. DILI) [95% CI]	P value
IgE	DILI cases	6	1.54 [1.24, 1.91]	0.64 [0.49, 0.84]	0.003
	Non‐DILI cases	12	0.99 [0.85, 1.15]
IgA	DILI cases	6	1.04 [0.90, 1.20]	0.98 [0.82, 1.17]	0.837
	Non‐DILI cases	12	1.02 [0.92, 1.13]
IgG	DILI cases	6	1.02 [0.92, 1.13]	0.99 [0.88, 1.12]	0.886
	Non‐DILI cases	12	1.01 [0.94, 1.09]
IgM	DILI cases	6	0.93 [0.79, 1.09]	1.14 [0.94, 1.38]	0.177
	Non‐DILI cases	12	1.06 [0.94, 1.18]

Subset of patients receiving LY3031207 75 and 225 mg with evaluable immunoglobulin concentrations included in analysis.

CI, confidence interval; DILI, drug‐induced liver injury; Ig, immunoglobulin; n, number meeting presence or absence of DILI.

Identification and semiquantification of LY3031207 metabolites

Examination of the clinical plasma and urine samples revealed two proposed primary metabolic pathways that involved oxidation of the imidazole ring and of the tert‐butyl functionality (Figure 4). Oxidation of the imidazole ring to the corresponding epoxide is followed by either rearrangement to Metabolite 5 (M5) or hydrolysis to M2. M3 was generated from M2 by the loss of glyoxal (C₂H₂O₂). M1 was formed from mono‐oxidation of the tert‐butyl group. Profiling of human plasma using LC/high‐resolution MS revealed the presence of LY3031207 and three metabolites (M1, M3, M5) in the cumulative extracted ion chromatograms. In all plasma pools on all days, parent drug was the predominant drug‐related component based on mass spectrometric response. M3 was the most prominent and only metabolite observed across all plasma pools. Based on liquid chromatograph/high‐resolution mass spectrometry ion intensity, the relative percentage of M3 was ≤2% of parent drug in 25‐ and 75‐mg groups and ranged from 2–10% in the 225‐mg groups. M1 and M5 represented <1% of parent drug in Day 1 plasma for 25‐, 75‐ and 225‐mg groups and Day 28 plasma for 25‐ and 75‐mg groups. In the 225‐mg dose group in Day 12 pools, M5 and M1 were observed in small amounts for unaffected subjects (≤3%) but were not observed in DILI patients. Parent drug and six metabolites (M1 to M6) were identified in pooled urine. The chromatographic profiles for each urine sample were very similar for all dose groups on all days. Parent drug was the predominant species, and metabolites M3, M5, and M1 were the most abundant metabolites across all urine pools.

Figure 4.

LY3031207 and proposed structures for metabolites in human plasma and urine (dotted line indicates formation may involve multiple steps). Plasma and urine samples were prepared as Hamilton AUC_0–24h pools for Days 1 and 28 or single point Day 12 pool for the 225‐mg group (based on history of LY3031207‐induced liver injury) 19. Metabolites were identified using accurate mass liquid chromatography/mass spectrometry and liquid chromatography/mass spectrometry/mass spectometry with positive electrospray as well as metabolite retention time comparisons. Concentration estimates for each metabolite in the extracted ion chromatograms were reported as relative to the mass spectrometric peak height of LY3031207. M1–6, metabolites M1–6

Discussion

In this small multiple ascending dose study, six cases of DILI with rapid onset (after as short as 12 days of dosing), hepatocellular injury pattern, eosinophilia, and increased serum IgE were identified. The clinical features of Patients 1 and 2 have been reported elsewhere 13. All six DILI subjects recovered fully without long‐term clinical sequelae. This outcome was attributed to intensive safety monitoring and prompt discontinuation of the study drug. Discontinuation of LY3031207 225 mg dosing and abandonment of further dose escalation prevented additional subjects from being exposed to dose‐dependent DILI risk.

The six cases had distinguishing features, unlike previously reported DILI with hypersensitivity. First, cases were not typical of idiosyncratic DILI, in that the incidence was quite high (56% in the 225‐mg LY3031207 group) and demonstrated a steep dose dependency. Although idiosyncratic DILI has been termed dose independent, there are conflicting data that suggest a relationship between daily doses of medications and idiosyncratic DILI 14, 15. Nevertheless, the steep dose–response curve for DILI with LY3031207 is unusual and has rarely been reported in the medical literature. Second, all six DILI cases demonstrated an IgE‐mediated hypersensitivity pattern. It was striking that >50% of subjects exposed to repeated doses of 225 mg LY3031207 seemed to have been sensitized within as few as 12 days of exposure. Elevated IgE concentration is commonly associated with type I immune hypersensitivity, which can include anaphylaxis, atopy or allergic symptoms. However, its association to DILI with hypersensitivity features has been reported in rare cases 16, and the contribution of IgE in other DILI with hypersensitivity features is unclear. Although it has been previously reported that DILI accompanied by eosinophilia may be associated with a lower risk of liver failure 17, 18, it is not entirely predictive of a benign course.

The hepatotoxicity identified in humans was in stark contrast to the nonclinical toxicology findings. Although nonclinical data identified the liver as a target organ of toxicity, the presentation of hepatotoxicity was mild, late in onset, and inconsistent across species. Hepatocellular hypertrophy in rats was attributed to microsomal enzyme induction and had no apparent effect on the health of rats. Significant liver toxicity in dogs occurred only at nontolerated doses and in the context of multiorgan toxicity; the liver changes were not considered the sole or primary cause of conditional decline. Microscopic liver changes in the dog included mild centrilobular hepatocellular degeneration/necrosis; however, the dramatic inflammatory infiltrates and prominent hepatocellular necrosis seen in human liver biopsies were not observed in dogs. Finally, no increase in liver enzymes occurred consistently or correlated reliably with the microscopic liver changes, and no eosinophilia was observed in dogs.

As with the majority of DILI cases, the mechanism for LY3031207‐associated DILI is not completely clear. The role of potential reactive metabolite formation in the pathogenesis of hepatic injury was explored. LY3031207 and a total of six metabolites were observed in plasma and urine from human subjects. Despite strong dose dependency, plasma levels of LY3031207 within each dose level were comparable in subjects with and without DILI, as were qualitatively similar metabolite profiles in urine. These data indicated that clearance of LY3031207 was similar regardless of onset of DILI; therefore, contrary to earlier speculation 13, genetic polymorphisms in the clearance pathway of LY3031207 were unlikely to be associated with the development of DILI in these subjects.

Only subtle differences in the plasma metabolite profile were observed in samples on Day 12 from subjects with and without DILI. Nonclinical data and the identification of M2 through M6 in human plasma and urine strongly suggested that the hepatic metabolism of LY3031207 involves the formation of an epoxide intermediate, a hypothetical reactive metabolite 20. Human metabolism data alone may not explain the DILI events observed because this pathway was present in both nonclinical toxicology species without significant evidence of DILI or hypersensitivity. Notably, another mPGES‐1 inhibitor that shared metabolic activation of the imidazole ring was also associated with DILI in one healthy subject with aminotransferase concentration >10× ULN in a separate multiple ascending dose study 8. Metabolic activation of different aminoimidazole compounds also involves the formation of epoxide in vitro 21, albeit via a different metabolic route. These data suggest that epoxide formation of the imidazole moiety may have played a role in the development of hepatotoxicity. Such imidazole rings may be considered as structural alert for hepatotoxicity and be avoided in the design of new chemical entities in drug discovery.

The dose‐dependent pattern seen with these cases begs the question whether pharmacological properties of LY3031207 were also involved in the observed hepatotoxicity. Although LY3031207 demonstrated pharmacology distinct from NSAIDs (Figure S1, S2a, and S2b), it demonstrated similar inhibition of inducible PGE₂ synthesis (Table S1). Depletion of hepatic PGE₂ may potentiate liver injury because PGE₂ has been hypothesized as hepato‐protective based on data from an acetaminophen‐induced mouse hepatotoxicity model 22. Interestingly, DILI was responsible for the withdrawal of five of the 17 marketed NSAIDs between 1964 and 2009 23, 24, 25, 26 . It is noteworthy that all withdrawn NSAIDs contained a phenylacetic acid moiety, which has been shown to undergo facile conjugation to the corresponding acyl glucuronide, which upon rearrangement is believed to form reactive intermediates. Of the remaining 20 NSAIDs in clinical use, hepatotoxicity risk is rare. These data suggest that DILI is unlikely to be a mechanism‐based toxicity for mPGES‐1 inhibitors. Rather, toxicity in this study was likely the result of a combination of compound specific metabolic activation, which may have been exacerbated by PGE₂ depletion. The safety profile of future mPGES‐1 inhibitors may therefore be greatly enhanced if they are devoid of the potential for intrinsic hepatotoxicity or metabolic activation.

In conclusion, the development of LY3031207 was terminated due to a rare form of dose‐dependent liver injury with hypersensitivity features. Metabolic activation of the imidazole ring may be involved in the pathogenesis of hepatotoxicity in the compound. These cases serve as a reminder of the limitation of toxicology models. Clinical investigators should approach the design and implementation of early‐phase studies with vigilance to drug toxicity and safety signals, careful monitoring of exposed subjects and prompt discontinuation of study drug, if indicated.

Competing Interests

Y.J., A.R., K.P., C.S., J.H., K.C., L.H., D.G.H., X.Y.Y., M.N., T.A.M. and W.L. are full‐time employees of and minor stockholders in Eli Lilly and Company or its subsidiaries. J.K. was a full‐time employee of Covance Clinical Research Unit. J.U. is a paid consultant for Eli Lilly and Company. This study was sponsored in full by Eli Lilly and Company or its subsidiaries.

The authors wish to thank Tamara Ball for editorial assistance. Dr Ball is a full‐time employee of inVentiv Health. Eli Lilly had contracted with inVentiv Health for assistance with this manuscript. The authors also wish to thank Dr Garret A. FitzGerald for his input on the study design, data interpretation, and manuscript review.

Contributors

Conception and design: Y.J., C.S., L.H., K.C., K.P., J.H., M.N., T.A.M.; Acquisition of data: J.K., Y.J., K.P., K.C., D.G.H., J.W.H.; Analysis and interpretation of data: Y.J., J.K., C.S., L.H., K.P., K.C., J.W.H., D.G.H., T.A.M., X.Y.Y., A.R., J.U., W.L., M.N.; Drafting: Y.J., C.S., L.H., K.C., J.W.H., D.G.H., A.R., W.L., X.Y.Y., J.U.; Critical revision: Y.J., J.K., C.S., L.H., K.P., K.C., J.W.H., D.G.H., T.A.M., X.Y.Y., A.R., J.U., W.L., M.N.; Final approval: Y.J., J.K., C.S., L.H., K.P., K.C., J.W.H., D.G.H., T.A.M., X.Y.Y., A.R., J.U., W.L., M.N.

Supporting information

Table S1 Effect of LY3031207 vs. celecoxib on serum thromboxane B concentration

Table S2 Liver‐related findings in repeat‐dose dog studies of LY3031207

Table S3 Comparison of demographic characteristics of drug‐induced liver injury cases vs. non‐drug‐induced liver injury cases

Table S4 Exposure multiple for oral administration of LY3031207

Figure S1 (A) Percent change from baseline in ex vivo whole blood PGE2 synthesis after a single dose of LY3031207 in healthy subjects. Ex vivo whole blood PGE2 synthesis after LPS stimulation is a validated assay that tests for the inhibition of inducible PGE2 synthesis after inflammatory stimulation. Bars represent mean percentage change from baseline; error bars present 90% CI. *P < 0.05 vs. placebo. (B) LY3031207 vs. inhibition of ex vivo whole blood PGE2 synthesis relationship. Symbols represent observations from each dose group, pink link represent model predicted concentration vs. % of baseline ex vivo PGE2 synthesis. Dotted lines represent IC80, drug concentration that produces 80% inhibition

Figure S2 (A) Changes from baseline in urinary 11‐dehydro‐thromboxane B2 (TXM) excretion after single doses of LY3031207 vs. celecoxib. Urinary TXM excretion has been used as a marker for in vivo thromboxane synthesis. Bars represent mean percentage change from baseline; error bars present 90% CI. * P < 0.05 vs. placebo. (B) Changes from baseline in urinary 2, 3‐dinor, 6‐keto prostaglandin F1α (PGIM) excretion after single doses of LY3031207 vs. celecoxib. Urinary PGIM excretion has been used as a marker for in vivo prostacyclin synthesis. Bars represent mean percentage change from baseline; error bars present 90% CI. * P < 0.05 vs. placebo

Jin, Y. , Regev, A. , Kam, J. , Phipps, K. , Smith, C. , Henck, J. , Campanale, K. , Hu, L. , Hall, D. G. , Yang, X. Y. , Nakano, M. , McNearney, T. A. , Uetrecht, J. , and Landschulz, W. (2018) Dose‐dependent acute liver injury with hypersensitivity features in humans due to a novel microsomal prostaglandin E synthase 1 inhibitor. Br J Clin Pharmacol, 84: 179–188. doi: 10.1111/bcp.13423.