Inhibition of human UDP-glucuronosyltransferase enzymes by lapatinib, pazopanib, regorafenib and sorafenib: Implications for hyperbilirubinemia

Accepted Manuscript Inhibition of human UDP-glucuronosyltransferase enzymes by lapatinib, pazopanib, regorafenib and sorafenib: Implications for hyperbilirubinemia John O Miners, Nuy Chau, Andrew Rowland, Kushari Burns, Ross A McKinnon, Peter I Mackenzie, Geoffrey T. Tucker, Kathleen M Knights, Ganessan Kichenadasse PII: DOI: Reference:

S0006-2952(17)30002-3 http://dx.doi.org/10.1016/j.bcp.2017.01.002 BCP 12716

To appear in:

Biochemical Pharmacology

Received Date: Accepted Date:

1 November 2016 4 January 2017

Please cite this article as: J.O. Miners, N. Chau, A. Rowland, K. Burns, R.A. McKinnon, P. I Mackenzie, G.T. Tucker, K.M. Knights, G. Kichenadasse, Inhibition of human UDP-glucuronosyltransferase enzymes by lapatinib, pazopanib, regorafenib and sorafenib: Implications for hyperbilirubinemia, Biochemical Pharmacology (2017), doi: http://dx.doi.org/10.1016/j.bcp.2017.01.002

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Inhibition of human UDP-glucuronosyltransferase enzymes by lapatinib, pazopanib, regorafenib and sorafenib: Implications for hyperbilirubinemia

John O Minersa,b,*, Nuy Chau a, Andrew Rowland a,b, Kushari Burnsa, Ross A McKinnonb, Peter I Mackenziea,b, Geoffrey T. Tuckerc, Kathleen M Knightsa and Ganessan Kichenadasseb,d

a

Department of Clinical Pharmacology, Flinders University School of Medicine, Adelaide,

Australia. b

Flinders Centre for Innovation in Cancer, Flinders University School of Medicine,

Adelaide, Australia. c

Department of Human Metabolism, Medical and Biological Sciences (Emeritus), University

of Sheffield, Sheffield, United Kingdom. d

Department of Oncology, Flinders Medical Centre, Bedford Park, Australia.

* Corresponding author. Email addresses: [email protected] [email protected] [email protected] [email protected] [email protected]

1

[email protected] [email protected] [email protected] [email protected]

2

Abstract Kinase inhibitors (KIs) are a rapidly expanding class of drugs used primarily for the treatment of cancer. Data relating to the inhibition of UDP-glucuronosyltransferase (UGT) enzymes by KIs is sparse. However, lapatinib (LAP), pazopanib (PAZ), regorafenib (REG) and sorafenib (SOR) have been implicated in the development of hyperbilirubinemia in patients. This study aimed to characterise the role of UGT1A1 inhibition in hyperbilirubinemia and assess the broader potential of these drugs to perpetrate drug-drug interactions arising from UGT enzyme inhibition. Twelve recombinant human UGTs from subfamilies 1A and 2B were screened for inhibition by LAP, PAZ, REG and SOR. IC50 values for the inhibition of all UGT1A enzymes, except UGT1A3 and UGT1A4, by the four KIs were < 10 µM. LAP, PAZ, REG and SOR inhibited UGT1A1-catalysed bilirubin glucuronidation with mean IC50 values ranging from 34 nM (REG) to 3734 nM (PAZ). Subsequent kinetic experiments confirmed that REG and SOR were very potent inhibitors of human liver microsomal β-estradiol glucuronidation, an established surrogate for bilirubin glucuronidation, with mean Ki values of 20 and 33 nM, respectively. Ki values for LAP and PAZ were approximately 1- and 2orders of magnitude higher than those for REG and SOR. REG and SOR were equipotent inhibitors of human liver microsomal UGT1A9 (mean Ki 678 nM). REG and SOR are the most potent inhibitors of a human UGT enzyme identified to date. In vitro – in vivo extrapolation indicates that inhibition of UGT1A1 contributes significantly to the hyperbilirubinemia observed in patients treated with REG and SOR, but not with LAP and PAZ. Inhibition of other UGT1A1 substrates in vivo is likely. Keywords UDP-glucuronosyltransferase, enzyme inhibition, kinase inhibitors, hyperbilirubinemia, drugendobiotic interaction, drug-drug interaction

3

Compounds Bilirubin (PubChem CID; 5280352; β-Estradiol (PubChem CID; 5757); Lapatinib (PubChem CID; 208908); Pazopanib (PubChem CID; 10113978); Propofol (PubChem CID; 4943); Regorafenib (PubChem CID; 178102514); Sorafenib (PubChem CID; 216239). Abbreviations DDI, drug-drug interaction; β-EST, β-estradiol; HLM, human liver microsomes; IV-IVE, in vitro – in vivo extrapolation; KI, kinase inhibitor; 4MU, 4-methylumbelliferone; LAP, lapatinib; NSB, non-specific binding; PAZ, pazopanib; PRO, propofol; REG, regorafenib; SOR, sorafenib; UDP-GlcUA, UDP-glucuronic acid; UGT, UDP-glucuronosyltransferase

4

1. Introduction Kinase signalling pathways regulate diverse cellular functions, including angiogenesis, apoptosis, differentiation and proliferation, and dysregulation of protein and lipid kinases is associated with a range of diseases [1,2]. In particular, mutations in the genes encoding protein and lipid kinases are linked to numerous malignancies in humans. Thus, there has been intense interest over the last two decades in the discovery and development of kinase inhibitors (KIs) for the treatment of cancer and other diseases. Thirty KIs have been approved since 2001, about two-thirds of these in the last five years [3]. Since the duration of treatment with KIs varies from weeks to years and polypharmacy is common in cancer patients (for the treatment of cancer and other co-morbidities) [4], patients receiving KIs are considered to be at high risk from drug-drug interactions (DDIs) [5]. Given the predominant role of cytochrome P450 (CYP) 3A4 and, to a lesser extent, other CYP enzymes (e.g. CYP1A2) in the metabolism of KIs, most studies investigating KIs as either victims or perpetrators of DDIs arising from enzyme inhibition have focussed on CYP3A4 inhibitors and substrates, respectively [6-9]. Further, both hepatic uptake (e.g. OATP1B1) and efflux transporters (e.g. P-glycoprotein and BCRP) are variably involved in KI disposition and DDIs may also arise from inhibition of transporter activity [10]. In addition to CYP- and transporter- mediated hepatic elimination, a number of KIs are known to be metabolised via glucuronidation [3,6,8]. Glucuronidation reactions involve UDP-glucuronosyltransferase (UGT) catalysed covalent linkage (‘conjugation’) of the substrate with glucuronic acid, which is derived from the cofactor UDP-glucuronic acid (UDP-GlcUA) [11]. UGT exists as a superfamily of enzymes [12]. The nineteen human UGT enzymes that utilise UDP-GlcUA as cofactor have been classified in the UGT 1A, 2A and 2B subfamilies. Of the hepatically expressed enzymes, UGT 1A1, 1A3, 1A4, 1A6, 1A9, 2B7 and

5

2B15 appear to be of greatest importance in drug and xenobiotic metabolism, with lesser contributions of UGT 2B4, 2B10 and 2B17 [13]. The individual UGT enzymes exhibit distinct, but overlapping, patterns of substrate and inhibitor selectivities. In particular, bilirubin is glucuronidated solely by UGT1A1 [13,14]. UGT1A9-catalysed glucuronidation contributes to the elimination of regorafenib (REG) and sorafenib (SOR) [9], and hence inhibition of UGT1A9 and other UGT enzymes by these KIs would not be unexpected. For example, dual inhibition of UGT1A1 and UGT1A9 by substrates of either enzyme has been demonstrated in this laboratory [15,16]. Consistent with this observation, inhibition of UGT1A1 by SOR in vitro has been reported [17,18]. Jaundice is a common adverse effect of both REG and SOR [19,20], and has also been reported as an adverse effect in patients treated with pazopanib (PAZ) and lapatinib (LAP) [21-23]. These observations are strongly suggestive of KI inhibition of UGT1A1-catalysed bilirubin glucuronidation since, as noted above, bilirubin is glucuronidated solely by this enzyme. Data relating to UGT enzyme inhibition by KIs are sparse and hence understanding of the propensity of these drugs to precipitate drug-endobiotic interactions and DDIs is limited. Importantly, the FDA now recommends screening of investigational new drugs for inhibition of UGTs, while the EMA recommends at least screening for inhibition of UGT 1A1 and 2B7 [24,25]. Recent studies have demonstrated that inhibition kinetic studies with human liver microsomes (HLM) and recombinant proteins as the enzyme sources together with in vitro-in vivo extrapolation (IV-IVE) approaches predict the likelihood of a DDI or drug-endobiotic interaction arising from UGT enzyme inhibition in vivo [13,26-29]. Since hyperbilirubinemia is an important adverse effect in patients treated with LAP, PAZ, REG and SOR, this study aimed to characterise the role of inhibition of UGT1A1-catalysed bilirubin glucuronidation as the cause of jaundice. More broadly, the study additionally characterised the inhibition of

6

human UGT 1A and 2B enzymes by LAP, PAZ, REG and SOR (see Figure 1 for structures), to assess the potential role of these KIs as perpetrators of DDIs. Importantly, inhibitor constants (Ki), which underpin IV-IVE, for LAP, PAZ, REG and SOR were corrected for microsomal binding and hence accurately reflect inhibition potential. Many KIs, but particularly REG and SOR, are highly lipophilic organic bases that bind extensively to the microsomal membrane [30]. Failure to account for non-specific binding (NSB) results in over-estimation of Ki, and Km in the case of a substrate, and hence underestimation of drug-endobiotic and DDI potential [13,31]. Notably, following correction for NSB, REG and SOR were identified as remarkably potent inhibitors of UGT1A1 with Ki values in the low nanomolar range making these compounds the most potent UGT inhibitors reported to date.

2. Materials and Methods 2.1. Drugs and other chemicals LAP, PAZ, REG and SOR were purchased from LC Laboratories, (Woburn, MA, USA); alamethicin (from Trichoderma viride), bilirubin, codeine, β-EST, β-estradiol-3-β-Dglucuronide, 4MU, 4MU β-D-glucuronide, PRO, and UDP-GlcUA (trisodium salt) were purchased from Sigma-Aldrich (Sydney, NSW, Australia); and codeine 6-O-β-D-glucuronide from Toronto Research Chemicals (North York, ON, Canada). Lamotrigine and lamotrigine N2-β-D-glucuronide were provided by the Wellcome Research Laboratories (Beckenham, UK). Solvents and other reagents used were of analytical reagent grade. 2.2. Human liver microsomes (HLM) and recombinant human UGTs.

7

Approval for the use of human liver tissue for in vitro drug metabolism studies was obtained from the Southern Adelaide Clinical Research Human Ethics Committee. Pooled HLM (150 donor pool; equal number of male and female donors) were purchased from Corning Gentest (Tewksbury, MA, USA). HLM were activated by pre-incubation with the pore-forming agent alamethicin (50 µg/mg microsomal protein) prior to use in incubations [32]. Human UGT 1A1, 1A3, 1A4, 1A6, 1A7, 1A8, 1A9 and 1A10 cDNAs were stably expressed in a human embryonic kidney cell line (HEK293T) [33]. Following harvesting, cells were lysed by sonication using a Vibra Cell VCX 130 Ultrasonics Processor (Sonics and Materials, Newtown, CT, USA). Lysates were centrifuged at 12,000 g for 1 min at 4°C, and the supernatant fraction was separated and stored in phosphate buffer (0.1 M, pH 7.4) at -80°C until use. SupersomesTM (Corning Gentest) expressing UGT 2B4, 2B7, 2B15 and 2B17 were used in inhibition studies given the lower expression of UGT2B proteins in HEK293T cells. 2.3. Glucuronidation inhibition by KIs 2.3.1. Inhibition of recombinant human UGT enzymes by LAP, PAZ, REG and SOR Inhibition of recombinant human UGT enzyme activities was determined in the absence and presence of LAP, PAZ, REG or SOR at four concentrations (0.01, 0.1, 1 and 10 µM). LAP, PAZ, REG and SOR stock solutions were prepared in DMSO such that the final concentration of solvent in incubations was 1.0% (v/v), which has only a minor effect on UGT activities [33]. An equivalent volume of DMSO was included in control incubations. Incubations, in a total volume of 0.1 mL, contained substrate (see below), UDP-GlcUA (5 mM), MgCl2 (4 mM) and recombinant UGT protein in phosphate buffer (0.1 M, pH 7.4), with and without LAP, PAZ, REG or SOR. The effects of each KI on UGT 1A1, 1A3, 1A6, 1A7, 1A8, 1A9, 1A10, 2B7, 2B15 and 2B17 activities were measured using the non-selective UGT substrate 4-methylumbelliferone (4MU) as the probe. Incubations were performed at a

8

4MU concentration corresponding to the apparent Km or S50 of 4MU for each enzyme (see footnote to Table 2), using protein concentrations and incubation times optimised previously in this laboratory [33]. Product (4MU-β-D-glucuronide) formation was measured by HPLC [34]. Inhibition of UGT1A4 and UGT2B4 by each KI was assessed with lamotrigine and codeine as the respective probe substrates according to established methods in the authors’ laboratory [27,28]. Concentrations of lamotrigine and codeine used in the UGT1A4 and UGT2B4 inhibition screening studies corresponded to the respective Km values for each substrate/pair. Positive control inhibitors were used in all inhibition screening experiments, as described by Pattanawongsa et al. [15]. Within- and between-day coefficients of variation for all enzyme assays, including with HLM as the enzyme source (see below), were < 5% and 10%, respectively. 2.3.2. Inhibition of UGT1A1-catalysed bilirubin glucuronidation by LAP, PAZ, REG and SOR. Bilirubin glucuronidation by recombinant UGT1A1 was measured according to Udomuksorn et al. [35]. Incubation mixtures contained UDP-GlcUA (5 mM), MgCl2 (4 mM), bilirubin (0.25 µM, which corresponds to the Km [35]) and HEK293 cell lysate expressing UGT1A1 (0.1 mg/mL) in phosphate buffer (pH 7.4, 0.1 M; 200 µL). Incubations were performed for 10 min at 37ºC, in the absence and presence of five separate added concentrations of each KI; LAP (50 – 1,500 nM), PAZ (500 – 10,000 nM), and REG and SOR (10 – 500 nM). Exposure of reactions to light and air was minimised by performing incubations in capped, amber tubes undertaken in a laboratory with dimmed lighting. The formation of bilirubin monoglucuronides and bilirubin di-glucuronide was quantified by HPLC [35]. 2.3.3. Kinetic characterisation of LAP, PAZ, REG and SOR inhibition of human liver microsomal UGT 1A1, 1A6 and 1A9 activities.

9

The kinetics and mechanisms of LAP, PAZ, REG and SOR inhibition of human liver microsomal UGT1A1 and UGT1A9 were determined using β-estradiol (β-EST) and propofol (PRO) as the respective enzyme-selective probe substrates [13]. Incubation conditions and analytical procedures used to quantify β-EST (3-position) and PRO glucuronidation were as described previously [16]. Briefly, incubations, in a total volume of 0.1 mL, contained substrate (β-EST or PRO), UDP-GlcUA (5 mM), MgCl2 (4 mM) and HLM in phosphate buffer (0.1 M, pH 7.4), with and without LAP, PAZ, REG or SOR. The HLM content of incubations was 0.25 mg/mL with β-EST and 0.5 mg/mL with PRO as the respective probe substrates. As with the inhibition screening studies, each KI was added to incubations in DMSO such that the final concentration was 1.0% v/v. Inhibition experiments were performed with four inhibitor concentrations at each of three substrate concentrations (Table 1). Binding of LAP, PAZ, REG, SOR, β-EST and PRO to HLM was accounted for in the calculation of inhibitor constants (see Results), which are expressed as unbound Ki (i.e. Ki,u). Where the fraction unbound of substrate and inhibitor (fumic) present in incubations was determined at a microsome concentration other than that used in experimental studies , fu mic values were converted to the appropriate microsome concentration (typically 0.5 mg/mL) using equation 1 [36]. Equation 1:

fu2 =

1 C2 (1− fu1 ) +1 C1 fu1

where, fu1 is the fraction unbound obtained experimentally at a specified HLM concentration (C1), and fu2 is the fumic calculated for HLM concentration C2 .

2.4. Data Analysis

10

All kinetic and inhibition screening experiments were performed in quadruplicate (< 10% variance). Data points represent the mean of quadruplicate measurement ± parameter standard error of fit (SE; inhibition screening data) or mean ± standard error of the mean (SEM; bilirubin inhibition and kinetic inhibition experiments). IC50 values were calculated using equation 2, while Ki,u values for LAP, PAZ, REG and SOR inhibition of human liver microsomal β-EST and PRO glucuronidation activities were determined by comparison of goodness-of-fit parameters from fitting the expressions for competitive, non-competitive and mixed (competitive – non-competitive) inhibition of an enzyme exhibiting Michaelis-Menten kinetics [16] and competitive inhibition of an enzyme exhibiting sigmoidal kinetics [15] to experimental data using Enzfitter (version 2.0, Biosoft, Cambridge, UK). Goodness of fit of all expressions was assessed from comparison of the SEs, coefficients of determination (r2), 95% confidence intervals, and F-statistic. Inhibition kinetic data (see Results) were best described by the expression for competitive inhibition of an enzyme exhibiting sigmoidal kinetics (β-EST and PRO as the probe substrates; equation 3). Equation 2:

[1 −

[

[ +  

where v0 is the control activity and vi is the activity in the presence of the inhibitor (I). Equation 3:
=

 ×   [

  1 +   +   

where S50 is the concentration at half maximal velocity (Vmax), [I] is the inhibitor concentration, and n is the Hill coefficient.

11

DDI potential of an inhibitor may be assessed from a Ki value generated in vitro using equation 4 for drugs with negligible renal clearance [13,37]. Equation 4: AUCi = AUC

1 fm + (1− f m ) 1+ [ I ] / Ki

where AUCi/AUC is the ratio of the areas under the concentration versus time post dose curves of the victim drug with and without co-administered inhibitor, and fm is the fraction of the dose of the victim drug metabolised by the enzyme and pathway of interest. Where the victim drug is metabolised along a single metabolic pathway by a single enzyme (e.g. bilirubin), equation 5 simplifies to, Equation 5: AUCi = 1 + [I]/Ki AUC

For IV-IVE, the inhibitor concentration ([I]) after oral administration may be calculated as the maximum unbound hepatic inlet concentration. However, this requires knowledge of the absorption rate constant,(ka), the fraction absorbed into the enterocytes (Fa) and the fraction escaping first-pass gut wall metabolism (FG) of the inhibitor [13,37]. In the absence of experimental information, the values of ka and Fa.FG are generally defaulted to 0.1/min and 1.0, respectively [24]. However, as discussed in Section 4 (Discussion), this approach has limitations for KIs. Alternatively, [I] is simplistically equated to the maximum unbound concentration of the inhibitor in plasma (see Section 4). The following values of maximum plasma concentration (Cmax), reported for the usual therapeutic dose (shown in parenthesis) and fraction unbound in plasma (fu) from published clinical trial data were used to calculate the maximum unbound concentration of each KI (as Cmax x fu): LAP (1200 mg/day) – Cmax 12

1.22 mg/L, fu 0.01 [38]; PAZ (800 mg/day) - Cmax 45.1 mg/L, fu 0.005 [39]; REG (160 mg/day) - Cmax 3.90 mg/L, fu 0.005 [40]; and SOR (400 mg bd) - Cmax 7.2 mg/L (mean from 7, 21 and 28 day dosage regimens) , fu 0.005 [7,41]. Given the very high plasma protein binding (fu ≤ 0.01) of the KIs, unbound Cmax concentrations are low, ranging from 21 to 77 nM for LAP, REG and SOR, and 515 nM for PAZ.

3. Results 3.1. UGT enzyme inhibition screening studies Inhibition of a panel of recombinant human UGT enzymes (1A1, 1A3, 1A4, 1A6, 1A7, 1A8, 1A9, 1A10, 2B4, 2B7, 2B15 and 2B17) by LAP, PAZ, REG and SOR is shown in Figure 2 and derived IC50 values in Table 2. REG and SOR potently inhibited UGT1A1, with IC50 values < 100 nM. Inhibition of UGT1A1 by LAP and PAZ was marginally less potent. Potent inhibition of UGT1A9 was also observed (IC50 < 2 µM), with the same rank order as for UGT1A1 inhibition. Other enzymes inhibited by LAP, PAZ, REG and SOR with IC50 values < 10 µM included UGT 1A6, 1A7, 1A8 and 1A10. IC50 values > 10 µM were generally observed for KI inhibition of the remaining recombinant UGTs, although IC50 values for LAP inhibition UGT1A3 and UGT1A4 were 2.3 and 9.2 µM, respectively. Notably, inhibition of UGT2B subfamily enzymes was moderate to negligible (IC50 > 20 µM). 3.2. Inhibition of UGT1A1-catalysed bilirubin glucuronidation by LAP, PAZ, REG and SOR. KI concentrations present in incubations were corrected for binding to HEK293T cell lysate using the equilibrium dialysis method employed to assess KI binding to HLM [30] (see below). Values of fumic for LAP, PAZ, REG and SOR with HEK293 cell lysate as the enzyme source were 0.72, 0.92, 0.57 and 0.57, respectively. Inhibition of recombinant UGT1A113

catalysed bilirubin glucuronidation is shown in Figure 3. REG and SOR potently inhibited bilirubin glucuronidation, with respective IC50 values of 34 ± 0.5 and 48 ± 3.2 nM. Consistent with the inhibition screening studies, IC50 values for LAP and PAZ inhibition of bilirubin glucuronidation were higher; 467 ± 12 and 3734 ± 14 nM, respectively. 3.3. Kinetic characterisation of KI inhibition of human liver microsomal UGT1A1 and UGT1A9 Since human liver microsomal enzyme activities are considered to better reflect the characteristics of hepatically expressed UGTs than recombinant enzymes, kinetic studies were performed with HLM as the enzyme source. Inhibition kinetic studies were limited to UGT1A1 and UGT1A9 as these were the hepatically expressed enzymes most potently inhibited (IC50 ≤ 3 µM) by all four KIs in the screening studies. (The 3 µM cut-off is approximately an order of magnitude higher than the highest Cmax of the KIs investigated; see Section 2.4.) β-EST and PRO were used as the respective substrate probes for UGT1A1 and UGT1A9. It should be recognised that bilirubin is an unsuitable substrate for kinetic studies with HLM as the enzyme source given its instability and very high protein/membrane binding [35,42]. However, it has been established that β-EST 3-glucuronidation may be used as a surrogate for bilirubin glucuronidation [42] (discussed further in Section 4). The NSB of LAP, PAZ, REG, SOR, β-EST and PRO to HLM has recently been determined in this laboratory [15,16,30]. Values of fumic used to correct for NSB of the KIs at HLM concentrations of 0.25 mg/mL and 0.5 mg/mL (in parenthesis) were: LAP, 0.25 (0.14); PAZ, 0.73 (0.5); REG, 0.14 (0.075); and SOR, 0.14 (0.075). The fumic values used to correct for the NSB of β-EST and PRO at the relevant HLM concentration (in parenthesis) were 0.79 (0.25 mg/mL) and 0.70 (0.5 mg/mL), respectively.

14

Inhibition kinetic plots are shown in Figure 4 and derived Ki,u values in Table 3. β-EST glucuronidation by HLM exhibits positive cooperative kinetics [15,16]. Although we have previously reported that PRO glucuronidation by HLM prepared from tissue sourced from the Flinders Liver Bank exhibits hyperbolic kinetics [43], PRO glucuronidation using the commercially sourced pool of HLM employed in this work exhibited positive cooperative kinetics (mean kinetic parameters from fitting with the Hill equation; S50 200 µM, n 2.09, Vmax 2.18 nmol/min.mg). All four KIs were competitive inhibitors of human liver microsomal β-EST and PRO glucuronidation (Figure 4). Consistent with the glucuronidation kinetics of the probe substrates (see above), inhibition of HLM-UGT1A1 and UGT1A9 were best described by equation 3 (Data analysis, Section 2.4). As observed in the inhibition screening and bilirubin inhibition experiments, REG and SOR were very potent inhibitors of human liver microsomal β-EST glucuronidation (Figure 4 and Table 3), with respective mean Ki,u values of 20 and 33 nM. Mean Ki,u values for LAP and PAZ were approximately 1- and 2orders of magnitude higher than those for REG and SOR, which again is broadly consistent with the inhibition screening data. Similar to the comparable inhibition of human liver microsomal β-EST glucuronidation by REG and SOR, these KIs were equipotent competitive inhibitors of human liver microsomal PRO glucuronidation (mean Ki,u 678 nM; Figure 4 and Table 3). However, mean Ki,u values (viz. ~ 17 µM) for LAP and PAZ inhibition were higher than expected from the inhibition screening studies. It should be noted that discordance between Ki (and Km) values generated using recombinant UGTs and HLM is not entirely unexpected [31]. 3.4. In vitro – in vivo extrapolation For a ‘victim’ drug metabolised along a single metabolic pathway by a single enzyme (e.g. bilirubin), the predicted increase in exposure in vivo (expressed as the AUC ratio) due to an

15

inhibitory DDI is given by equation 5. Maximum unbound plasma concentrations of LAP, PAZ, REG and SOR were calculated from published Cmax and fu data from clinical trials (see Data analysis, Section 2.4 and Discussion, Section 4). Based on Ki values for β-EST glucuronidation, a surrogate for bilirubin glucuronidation, REG and SOR are both predicted to increase bilirubin exposure approximately 3- fold, with only a minor effect from PAZ and no effect of LAP (Table 4). It should be noted that failure to account for the NSB of the KIs results in mean Ki values of 2270 nM, 3200 nM, 144 nM and 236 nM for LAP, PAZ, REG and SOR, respectively, resulting in minor or negligible predicted inhibition (0 – 32%) of in vivo UGT1A1 activity. No or negligible inhibition of UGT1A9 in vivo is predicted based on the Ki values for human liver microsomal PRO glucuronidation (Table 4). Importantly, however, Ki values for inhibitors of human liver microsomal UGT1A9 are overestimated by approximately an order of magnitude due to the inhibitory effects of long-chain unsaturated fatty acids released from the microsomal membrane during the course of an incubation [43-45]. Addition of bovine serum albumin (BSA, 2% w/v) to incubations sequesters the inhibitory fatty acids and reduces the Ki (and Km of substrates) by an order of magnitude, presumably giving a ‘true’ estimate of this parameter. Addition of BSA to incubations was not feasible here given the extremely high protein binding of KIs (fu typically ≤ 0.01). If Ki,u values one-tenth those shown in Table 3 are assumed for inhibition of human liver microsomal UGT1A9 by LAP, PAZ, REG and SOR, then modest inhibition (31 – 67%) of clearance via glucuronidation for a compound metabolised solely by UGT1A9 in vivo is predicted following co-administration of PAZ, REG and SOR (Table 4).

4. Discussion

16

There is growing recognition of the potential importance of drug-endobiotic interactions (especially with bilirubin as the ‘victim’) and DDIs arising from inhibition of UGT enzymes. Indeed, the FDA and EMA now recommend characterisation of UGT enzyme inhibition by new investigational drugs [24,25]. Data presented here demonstrate that, to the best of our knowledge, REG and SOR are the most potent inhibitors of any human UGT enzyme reported to date. Initial inhibition studies of UGT1A1-catalysed 4MU and bilirubin glucuronidation provided respective mean IC50 values of 45 and 33 nM for REG, and 66 and 48 nM for SOR. Kinetic studies with bilirubin are problematic given the light- and chemicalinstability of this compound along with its high protein/membrane binding and sequential glucuronidation [35,42]. Available evidence indicates, however, that β-EST 3glucuronidation serves as a surrogate for bilirubin glucuronidation, especially with respect to prediction of inhibition [42]. Mean Ki,u values for REG and SOR inhibition of human liver microsomal β-EST 3-glucuronidation were 20 and 33 nM, respectively, which are broadly consistent with the IC50 values determined with 4MU and bilirubin (given Ki is equivalent to 0.5 x IC50 for competitive inhibition). Based on these Ki,u values and plasma unbound Cmax values for REG and SOR, 3.0- and 3.4- fold increases in exposure to bilirubin are predicted (see later discussion relating to IV-IVE). Ki,u (and IC50) values for LAP and PAZ inhibition were higher, and little or no effect of LAP and PAZ on bilirubin exposure due to UGT1A1 inhibition in vivo was predicted. SOR inhibition of bilirubin glucuronidation in vitro has been investigated previously [18]. A Ki of 11.8 µM was reported for SOR inhibition of UGT1A1-catalysed bilirubin glucuronidation, which is inconsistent with inhibition of bilirubin glucuronidation in vivo. Importantly, however, there was no correction for the NSB of SOR, and experimental conditions (e.g. substrate and inhibitor concentration ranges) did not favor the detection of potent inhibition. Furthermore, as noted above, bilirubin is not a recommended substrate for

17

in vitro glucuronidation inhibition kinetic studies. The differences between the present and previous studies highlight the need for careful attention to experimental conditions in order to generate accurate kinetic constants. The extent of UGT1A1 inhibition required to precipitate hyperbilirubinemia is unclear. Early studies suggested that hepatic bilirubin glucuronidation activity is reduced to about 25 – 50% of normal in Gilbert’s syndrome [46,47], but this has not been verified using more recent, reliable experimental and analytical methods. The predicted 3-fold increase in exposure to bilirubin in vivo due to REG and SOR reported here corresponds to a two-thirds reduction in substrate clearance (assuming constant formation), and this would appear consistent with reports of hyperbilirubinemia in patients treated with these drugs. A meta-analysis identified a relative risk of 5.69 for hyperbilirubinemia in patients treated with REG (or PAZ) [22], while a very high incidence of elevated total bilirubin (~70%) was observed in advanced hepatocellular carcinoma patients treated with SOR [19]. Although disease progression and liver cirrhosis undoubtedly contributed to the hyperbilirubinemia observed in the hepatocellular patients treated with SOR, data presented here are consistent with a significant contribution of inhibition of bilirubin glucuronidation. In contrast to REG and SOR, LAP and PAZ had little or no effect on UGT1A1 activity. Apart from inhibition of UGT1A1, KIs may contribute to the development of hyperbilirubinemia by other mechanisms. For example, bilirubin is taken up into hepatocytes by OAT1B1/3, while bilirubin glucuronides are transported across the canalicular and sinusoidal membranes by MRP2 and MRP3, respectively [48]. OATP1B1/3 variably contribute to the hepatocellular uptake of KIs, including PAZ and SOR [49], and MRP2 has been linked to the efflux of SOR [10]. Further, LAP and PAZ have been reported to inhibit OATP1B1 while SOR inhibits MRP2 [10]. Thus, inhibition of bilirubin uptake and efflux and hence ‘transport-metabolism’ interplay may additionally contribute to KI-induced hyperbilirubinemia.

18

Genetic polymorphism of UGT1A1 is another consideration. We have shown previously that the hazards ratio for indinavir-induced hyperbilirubinemia arising from inhibition of UGT1A1 is an order of magnitude higher in Thai HIV patients carrying Gilbert’s alleles (UGT1A1*6 and UGT1A1*28) compared to wild-type [50]. While indinavir inhibited both UGT1A1 and UGT1A1*6, it was postulated that the higher total bilirubin concentration in patients carrying the mutant UGT1A1 alleles arises from a higher baseline bilirubin concentration. There is evidence to suggest that elevations in total bilirubin concentration in patients treated with SOR may also be influenced by UGT1A1 genotype [51]. Consistent with the minor effect of PAZ on UGT1A1 activity observed here, genotyping data indicate that the hyperbilirubinemia observed in patients treated with PAZ is a benign manifestation of Gilbert’s syndrome [52,53]. Similarly, based on an analysis of the association between Class II HLA and UGT1A1*28 alleles with liver injury in patients treated with LAP, it was suggested that the observed elevation in total bilirubin concentration may be due to underlying Gilbert’s syndrome rather than LAP treatment itself (at least in HLA-negative cases) [54]. These observations are consistent with the predicted lack of inhibition of LAP and PAZ on bilirubin glucuronidation in vivo reported here. As indicated in Section 2.4 (Data analysis), the estimate of the unbound inhibitor concentration for IV-IVE is ideally defined by the hepatic input concentration of the inhibitor. Where values of ka and Fa.FG are unavailable, the FDA recommends 0.1/min and 1.0 be assumed for the respective parameters, presumably to provide a ‘worst-case’ scenario for inhibition [24]. However, these assumptions will almost certainly result in significant over-estimation of hepatic input concentration for most, if not all, KIs and therefore misrepresent the predicted inhibition of bilirubin glucuronidation by LAP, PAZ, REG and SOR. For example, KIs are highly lipophilic drugs that are generally absorbed relatively slowly (tmax typically > 2 hr). Of the KIs investigated here, a ka value appears to have been

19

published only for LAP [55]; the reported ka of 0.56/hr (or 0.01/min) is one-tenth that of the recommended assumed value (viz. 0.1/min). Similarly, values of Fa.FG are not available for most KIs. However, it appears that the net oral bioavailabilities of KIs are typically < 100%, due primarily to incomplete absorption and/or pre-systemic metabolism [3]. Fa values less than 1 probably reflect the high lipophilicities of the KIs resulting in incomplete dissolution, as highlighted previously [30]. The prediction of FG values using the ‘QGut’ model [56] is not always precise. FG values less than 1 will further decrease the predicted extent of inhibition of hepatic UGT’s, whereas their inhibition in the gut wall would have the opposite effect on the exposure of substrates given orally. Owing to the considerable uncertainties around ka, Fa and FG values, a simplified static IV-IVE approach based solely on unbound Cmax was adopted here. Further, through the use of dynamic rather than static maximum estimates of inhibitor concentration, physiologically based pharmacokinetic (PBPK) modelling is likely to provide more accurate estimates of the extent of an interaction. However, simulation software such as Simcyp does not currently accommodate endogenous compounds such as bilirubin as ‘victim’ substrates. Ideally, a systems model incorporating the turnover of bilirubin would form the basis of a comprehensive simulation. Based on the IC50 values for REG and SOR inhibition of UGT1A1-catalysed 4MU and bilirubin glucuronidation and the Ki,u values for inhibition of human liver microsomal β-EST glucuronidation, both drugs might be expected to inhibit the glucuronidation of other UGT1A1 substrates in vivo. Drugs metabolised extensively by UGT1A1 include dolutegravir, etoposide, raloxifene, raltegravir, and SN-38. Of the drugs used in the treatment of cancer, neither etoposide nor irinotecan (the pro-drug converted to SN-38) are currently coprescribed with REG or SOR. Nevertheless, the irinotecan – SOR combination has been investigated in a phase I clinical trial [17]. SOR, 400 mg bd, more than doubled the AUC and Cmax of SN-38. SOR was further shown to inhibit human liver microsomal SN-38 (a substrate

20

of both UGT1A1 and UGT1A9) with a Ki of 2.7 µM [17]. Similarly, LAP has been reported to inhibit the human liver microsomal glucuronidation of SN-38 with a Ki of 1.6 µM [57]. Neither of these studies accounted for NSB, and hence the reported Ki’s are over-estimated by at least an order of magnitude. Although inhibition of UGT1A1 is not a feature of all kinase inhibitors, several KIs in addition to LAP, PAZ, REG and SOR have been reported to inhibit UGT1A1 activity in vitro including; afatinib, crizotinib, dasatinib, erlotinib, gefitinib, imatinib and nilotinib [9, 57-59]. However, given the structural complexity of KIs (see Figure 1) and limited published Ki values for inhibition of UGT1A1, meaningful structure-effect relationships cannot be inferred at present. In addition to UGT1A1, REG and SOR potently inhibited human liver microsomal PRO glucuronidation (Ki,u ~ 0.68 µM). When the effects of long-chain unsaturated fatty acid inhibition of HLM-UGT1A9 are taken into account (see Section 3.4, Results), REG and SOR are predicted to modestly increase the AUC-ratios of UGT1A9 substrates, with lesser inhibition by PAZ. Known UGT1A9 drug substrates include edaravone, frusemide, mycophenolic acid, PRO, retigabine, and several NSAIDs (low affinity component of glucuronidation). Of further interest, UGT1A9 is the most abundantly expressed enzyme in human kidney and appears to contribute to the renal glucuronidation of numerous drugs and non-drug xenobiotics [45,60]. Thus, inhibition of UGT1A9 by KIs would be expected to result in renal accumulation of drugs and other chemicals metabolised by this enzyme. In addition to the potent inhibition of UGT1A1 and UGT1A9 by REG and SOR, the four KIs investigated here were potent inhibitors (IC50 values 0.317 to 8.5 µM) of UGT 1A7, 1A8 and 1A10, enzymes that are expressed only in the gastrointestinal tract. However, the importance of these enzymes to drug and chemical metabolism in the gastrointestinal tract is incompletely understood, and the significance of KI inhibition is therefore unclear. All four KIs were reasonably potent inhibitors of the hepatically expressed UGT1A6 (mean IC50 range

21

3.4 – 4.2 µM). However, exploratory kinetic studies of the inhibition of human liver microsomal deferiprone glucuronidation (a selective UGT1A6 substrate [13]), provided Ki,u values ranging from 9.0 µM (SOR) to > 20 µM (PAZ). The reason(s) for the discrepancies between the inhibition screening studies and the kinetic studies with UGT1A6 and HLM as the respective enzyme sources is unclear, but taken together the data indicate that inhibition of UGT1A6 in vivo is highly unlikely. In conclusion: (i) REG and SOR are extremely potent inhibitors of UGT1A1, and potent inhibitors of several other UGT1A enzymes. Indeed, REG and SOR are the most potent inhibitors of any human UGT enzyme (specifically UGT1A1) reported to date; (ii) IV-IVE predicts that inhibition of UGT1A1 is likely to contribute significantly to the jaundice observed in patients treated with REG and SOR, but not to the hyperbilirubinemia reported in patients receiving LAP and PAZ; (iii) Given the potent inhibition of UGT1A1-catalysed 4MU, bilirubin and β-EST glucuronidation observed here, REG and SOR are similarly expected to potently inhibit drug glucuronidation by UGT1A1 more widely, although any clinical relevance remains to be established; (iv) REG and SOR are potent inhibitors of UGT1A9-catalysed 4MU and PRO glucuronidation, but further studies are required to assess the scope and clinical relevance of UGT1A9 inhibition; (v) Data presented here further highlight the need to account for experimental confounders when assessing drug-endobiotic and DDI potential in vitro, especially the NSB of inhibitor (and substrate) to the enzyme source. Almost all previously published studies investigating KI inhibition of CYP and UGT enzymes have not considered the contribution of NSB, and hence inhibition potential in vivo is likely to be underestimated; and (vi) Given the potent inhibition of several UGT enzymes by the KIs investigated here, especially REG and SOR, and the possible clinical implications, further investigation of UGT enzyme inhibition by other marketed KIs is warranted.

22

Ackowledgement This project was funded by a project grant from the National Health and Medical Research Council of Australia (id 1044063).

Conflict of interest The authors declare no conflicts of interest.

23

References [1] D Fabbro, 25 years of small molecular weight kinase inhibitors: Potentials and limitations, Molec Pharmacol 87 (2015) 766-755. [2] P Wu, TE Nielsen, MH Clausen, FDA-approved small molecule kinase inhibitors, Trends Pharmcol Sci 36 (2015) 422-439. [3] A Rowland, M van Dyk, AA Mangoni, JO Miners, RA McKinnon, MD Wiese, E Rowland, H Gurney, BJ Smith, MJ Sorich, Kinase inhibitor pharmacokinetics: Comprehensive summary and roadmap for addressing interindividual variability, Exp Opin Drug Metab Toxicol 31 (2017) 31-49. [4] J Lees, A Chan, Polypharmacy in elderly patients with cancer: Clinical implications and management, Lancet Oncology 12 (2011) 1249-1257. [5] J Scholler, D Leveque, Molecular pharmacokinetic determinants of anticancer kinase inhibitors in humans, Oncol Rev 5 (2011) 77-92. [6] NP Van Erp, H Gelderblom, H-K Guchelaar, Clinical pharmacokinetics of tyrosine kinase inhibitors, Cancer Treatment Rev 35 (2009) 693-706. [7] P Di Gion, F Kanefendt, ALindauer, M Scheffer, O Dorosshyenko, U Fuhr, J Wolf, U Jaedhe, Clinical pharmacokinetics of tyrosine kinase inhibitors, Clin Pharmacokinetics 50 (2011) 551-603. [8] H-J, Klumpen, CF Samer, RHJ Mathijssen, JMH Schellens, H Gurney, Moving towards dose individualization of tyrosine kinase inhibitors, Cancer Treatment Rev 37 (2011) 251260. [9] C Tlemsani, O Huillard, J Arronodeau, P Boudou-Rouquette, A Cessot, B Blanchet, A Thomas-Schoemann, R Coriat, J-P Durrand, J Giroux, J Alexandre, F Goldwasser, Effect of 24

glucuronidation on transport and tissue accumulation of tyrosine kinase inhibitors: Consequences for the clinical management of sorafenib and regorafenib, Exp Opin Drug Metab Toxicol 11 (2015) 785-792. [10] K Mandery, H Glaeser, MF Fromm, Interaction of innovative small molecule drugs used for cancer therapy with drug transporters, Br J Pharmacol 165 (2012) 345-362. [11] JO Miners JO, PI Mackenzie, Drug glucuronidation in humans, Pharmacol Ther 51 (1991) 347-369. [12] PI Mackenzie, KW Bock, B Burchell, C Guillemette, S-I Ikushiro, T Iyanagi T, JO Miners, IS Owens, DW Nebert, Nomenclature update for the mammalian UDPglycosyltransferase (UGT) gene superfamily, Pharmacogenet Genomics 15 (2005) 677-685. [13] JO Miners, PI Mackenzie, KM Knights, The prediction of drug glucuronidation parameters in humans: UDP-glucuronosyltransferase enzyme selective substrate and inhibitor probes for reaction phenotyping and in vitro – in vivo extrapolation of drug clearance and drug-drug interaction potential. Drug Metab Rev 42 (2010) 196-208 [14] PJ Bosma, J Seppen, B Goldhoorn, C Bakker, RPJ Oude Elferink, J Roy Chowdhury, N Roy Chowdhury, PLM Jansen, Bilirubin UDP-glucuronosyltransferase 1 is the only relevant bilirubin glucuronidating isoform in man, J Biol Chem 269 (1994) 17960-17964. [15] A Pattanawongsa, N Chau, A Rowland, JO Miners, Inhibition of human UDPglucuronosyltransferase enzymes by canagliflozin and dapagliflozin: Implications for drugdrug interactions, Drug Metab Dispos 43 (2015) 1468-1476. [16] JO Miners, K Bowalgaha, DJ Elliot, P Baranczewski, KM Knights, Characterization of niflumic

acid

as

a

selective

inhibitor

25

of

human

liver

microsomal

UDP-

glucuronosyltransferase 1A9: Application to the reaction phenotyping of acetaminophen glucuronidation, Drug Metab Dispos 39 (2011) 644-652. [17] K Mross, S Steinbild, F Baas, D Gmehling, M Radtke, D Voliotis, E Brendel, O. Christensen, C Unger, Results from an in vitro and clinical/pharmacological phase I study with combination irinotecan and sorafenib, Eur J Cancer 43 (2007) 55-63. [18] CJ Peer, TM Sissung, A Kim, L Jain, S Woo, ER Gardner, CT Kirkland, SM Troutman, BC English, ED Richardson, J Federspiel, D Venzon, W Dahut, E Kohn, Y Yarchoan, G Giaccone, B Widerman, WD Figg, Sorafenib is an inhibitor of UGT1A1 but is metabolized by UGT1A9: Implications of genetic variants on pharmacokinetics and hyperbilirubinemia, Clin Cancer Res 18 (2012) 2099-2107. [19] GK Abou-Alfa, D Amadori, A Santoro, A Figer, J De Greve, C Lathia, D Voloitis, Sanderson, M Moscovici, S Ricci, Safety and efficacy of sorafenib in patients with hepatocellular carcinoma (HCC) and Child-Pugh A versus B cirrhosis, Gastrointest Cancer Res 4 (2011) 40-44. [20] AA Miller, DJ Murray, K Owzar, DR Hollis, EB Kennedy, G Abou-Alfa, A Desai, J Hwang, MA Villalona-Calero, EC Dees, LD Lewis, MG Kakih, MJ Edelman, F Millard, RC Frank, RJ Hohl, MJ Ratain, Phase I and pharmacokinetic study of sorafenib in patients with hepatic or renal dysfunction: CALGB 60301, J Clin Oncol 27 (2009) 1800-1805. [21] GK Dy, AA Adjei, Understanding, recognizing, and managing toxicities of targeted anticancer therapies, CA: A Cancer Journal for Clinicians 63 (2013) 249-279. [22] R Iacovelli, A Palazzo, G Procoppio, M Santoni, P Trenta, A De Benedetto, S Mezi, E Cortesi, Incidence and relative risk of hepatic toxicity in patients treated with anti-angiogenic tyrosine kinase inhibitors for malignancy, Br J Clin Pharmacol 77 (2013) 929-938.

26

[23] O Abdel-Rahman, M Fouad, Does the use of lapatinib increase the risk of fatigue and hepatic toxicities in patients with solid tumors? A critical literature review and meta-analysis, Expert Opin Drug Safety 13 (2014) 999-1008. [24] FDA (2012). Guidance for Industry. Drug interaction studies – Study design, data analysis,

implications

for

dosing,

and

labelling

recommendations.

http://www.fda.gov/Drugs/GuidanceComplianceRegulatoryInformatiom/Guidances/default/ht m. [25]

EMA

(2012).

Guideline

on

the

investigation

of

drug

interactions.

ema.europa.eu/docs/en_GB/document_library/Scientific_guideline/2012/07/WC500129606.p df. [26] JO Miners, TM Polasek, PI Mackenzie, KM Knights, The in vitro characterization of inhibitory drug-drug interactions involving UDP-glucuronosyltransferase. In, Enzyme and Transporter based drug-drug interactions, (Eds Pang KS, Rodrigues AD and Peter R), Chapter 8 (2010) 217-236, Springer (New York). [27] P Raungrut, V Uchaipichat, DJ Elliot, B Janchawee, AA Somogyi, JO Miners, In vitro – in vivo extrapolation predicts drug-drug interactions arising from inhibition of codeine glucuronidation by dextropropoxyphene, fluconazole, ketoconazole and methadone in humans, J Pharmacol Exp Ther 334 (2010) 609-618. [28] A Rowland, DJ Elliot, JA Williams, PI Mackenzie, RG Dickinson, JO Miners, In vitro characterization of lamotrigine N2-glucuronidation and the lamotrigine – valproic acid interaction. Drug Metab Dispos 34 (2006) 1304-1311. [29] V Uchaipichat, LK Winner, PI Mackenzie, DJ Elliot, JA Williams, JO Miners, Quantitative prediction of in vivo inhibitory interactions involving glucuronidated drugs from

27

in vitro data: The effect of fluconazole on zidovudine glucuronidation, Br J Clin Pharmacol 61 (2006) 427-439. [30] K Burns, PC Nair, A Rowland, PI Mackenzie, KM Knights, JO Miners, The nonspecific binding of tyrosine kinase inhibitors to human liver microsomes, Drug Metab Dispos 43 (2015) 1934-1937. [31] JO Miners, KM Knights, JB Houston, PI Mackenzie, In vitro – in vivo correlation for drugs cleared by glucuronidation: Pitfalls and promises, Biochem Pharmacol 71 (2006) 15311539. [32] S Boase, JO Miners, In vitro – in vivo correlations for drugs eliminated by glucuronidation: Investigations with the model substrate zidovudine, Br J Clin Pharmacol 54 (2002) 493-503. [33] V Uchaipichat, PI Mackenzie, X-H Guo, D Gardner-Stephen, A Galetin, JB Houston, JO Miners, Human UDP-glucuronosyltransferases: Isoform selectivity and kinetics of 4methylumbelliferone and 1-naphthol glucuronidation, effects of organic solvents, and inhibition by diclofenac and probenecid, Drug Metab Dispos 32 (2004) 413-423. [34] BC Lewis, PI Mackenzie, DJ Elliot, B Burchell, CR Bhasker, JO Miners, Amino terminal domains of human UDP-glucuronosyltransferases (UGT) 2B7 and 2B15 associated with substrate selectivity and autoactivation, Biochem Pharmacol 73 (2007) 1463-1473. [35] W Udomuksorn, DJ Elliot, BC Lewis, PI Mackenzie, K Yoovathaworn, JO Miners, Influence of mutations associated with Gilbert and Crigler-Najjar type II syndromes on glucuronidation kinetics of bilirubin and other UDP-glucuronosyltransferase 1A substrates, Pharmacogenet Genomics 17 (2007) 1017-1029.

28

[36] RP Austin, P Barton, SL Cockroft, MC Wenlock, RJ Riley, The influence of nonspecific microsomal binding on apparent intrinsic clearance, and its prediction from physicochemical properties, Drug Metab Dispos 30 (2002) 1497-1503. [37] K Ito, T Iwatsubo, S Kanamitsu, K Ueda, H Suzuki, Y Sugiyama, Prediction of pharmacokinetic alterations caused by drug-drug interactions: Metabolic interaction in the liver, Pharmacol Ther 50 (1998) 387-411. [38] HA Burris, HI Hurwitz, EC Dees, A Dowlati, KL Blackwell, B O’Neil, PK Marcom, MJ Ellis, C Overmoyer, SF Jones, JL Harris, DA Smith, KM Koch, A Stead, S Mangum, NL Spector, Phase I safety, pharmacokinetics, and clinical activity study of lapatinib (GW572016), a reversible dual inhibitor of epidermal growth factor receptor tyrosine kinases, in heavily pretreated patients with metastatic carcinomas. J Clin Oncol 23 (2005) 5305-5313. [39] HI Hurwitz, A Dowlati, S Saini, S Savage, AB Suttle, DM Gibson, JP Hodge, EM Merkle, L Pandite, Phase I trial of pazopanib in patients with advanced cancer. Clin Cancer Res 15 (2009) 4220-4227. [40] K Mross, A Frost, S Steinbild, S Hedborn, M Buchert, U Fasol, C Unger, J Kratzschmar, R Heinig, O Boix, O Christensen, A phase I dose-escalation study of regorafenib (BAY 734506), an inhibitor of oncogenic, angiogenic, and stromal kinases, in patients with advanced solid tumors, Clin Cancer Res 18 (2012) 1-10. [41] D Strumberg, JW Clark, A Awada, MJ Moore, H Richly, A Hendlisz, HW Hirte, JP Eder, H-J Lenz, B Schwartz, Safety, pharmacokinetics, and preliminary antitumor activity of sorafenib: A review of four phase I trials in patients with advanced refractory solid tumors. The Oncologist 12 (2007) 426-437.

29

[42] J Zhou, TS Tracy, RR Remmel, Correlation between bilirubin glucuronidation and estradiol-3-glucuronidation in the presence of model UDP-glucuronosyltransferase 1A1 substrates/inhibitors, Drug Metab Dispos 39 (2011) 322-329. [43] A Rowland, KM Knights, PI Mackenzie, JO Miners, “Albumin effect” and drug glucuronidation: Bovine serum albumin and fatty acid-free human serum albumin enhance the glucuronidation of UDP-glucurososyltransferase (UGT) 1A9 but not UGT1A1 and UGT1A6 activities, Drug Metab Dispos 36 (2008) 1056-1062. [44] N Manevski, PS Moreolo, J Yi-Kauhalouma, M Finel, Bovine serum albumin decreases Km values of human UDP-glucuronosyltransferases 1A9 and 2B7 and increases Vmax values of UGT1A9, Drug Metab Dispos 39 (2011) 2117-2129. [45] KM Knights, SM Spencer, JK Fallon, N Chau, PC Smith, JO Miners, Scaling factors for the in vitro – in vivo extrapolation (IV – IVE) of renal drug and xenobiotic glucuronidation clearance, Br J Clin Pharmacol 81 (2016) 1153-1164. [46] M Black, BH Billing, Hepatic bilirubin UDP-glucuronyl transferase activity in liver disease and Gilbert’s syndrome, N Engl J Med 280 (1969) 1266-1271. [47] WR Metge, CA Owen, WT Foulk, HN Hoffman, Bilirubin glucuronyl transferase activity in liver disease, J Lab Clin Med 64 (1964) 89-98. [48] D Keppler, The roles of MRP2, MRP3, OATP1B1, and OATP1B3 in conjugated hyperbilirubinemia. Drug Metab Dispos 42 (2014) 561-565. [49] EI Zimmerman, SH Hu, JL Roberts, AA Gibson, DJ Orwick, L Li, A Sparreboom, SD Baker, Contribution of OATP1B1 and OATP1B3 to the disposition of sorafenib and sorafenib glucuronide, Clin Cancer Research 19 (2013) 1458-1466.

30

[50] MA Boyd, P Sraseuebkul, K Ruxrungtham, PI Mackenzie, V Uchaipichat, M Stek, JMA Lange, P Phanuphak, DA Cooper, JO Miners, Relationship between hyperbilirubinemia and UDP-glucuronosyltransferase 1A1 polymorphism in HIV-infected Thai patients treated with indinavir, Pharmacogenet Genom 16 (2006) 321-329. [51] J Meza-Junco, QS-C Chu, O Christensen, P Rajagopalan, S Das, R Stefanyschyn, MB Sawyer, UGT1A1 polymorphism and hyperbilirubinaemia in a patient who received sorafenib, Cancer Chemother Pharmacol 65 (2009) 1-4. [52] C-F Xu, BH Reck, Z Xue, L Huang, KL Baker, M Chen, EP Chen, HE Ellens, VE Mooser, LR Cardon, CF Spraggs, L Pandite, Pazopanib-induced hyperbilirubinemia is associated with Gilbert’s syndrome UGT1A1 polymorphism, Br J Cancer 102 (2010) 13711377. [53] RJ Motzer, T Johnson, TK Choueiri, KC Deen, Z Xue, LN Pandite, C Carpenter, X-F Xu, Hyperbilirubinemia in pazopanib- or sunitinb-treated patients in COMPARZ is associated with UGT1A1 polymorphisms, Ann Oncol 24 (2013) 2927-2929. [54] CF Spraggs, LR Parham, CM Hunt, CT Dollery, Lapatinib-induced liver injury characterized by class II HLA and Gilbert’s syndrome genotypes, Clin Pharmacol Ther 91 (2012) 647-652. [55] DA Smith, KM Koch, N Arya, CJ Bowen, JM Herendeen, A Beelen, Effects of ketoconazole and carbamazepine on lapatinib pharmacokinetics in healthy subjects, Br J Clin. Pharmacol 67 (2009) 421-426. [56] J Yang, M Jamei, KR Yeo, GT Tucker, A Rostami-Hodjegan, Prediction of intestinal first-pass drug metabolism, Current Drug Metab 8 (2007) 676-684.

31

[57] N Zhang, Y Liu, H Jeong, Drug-drug interaction potentials of tyrosine kinase inhibitors via inhibition of UDP-glucuronosyltransferases, Scientific Reports 5 (2015) 17778. [58] L Ai, L Zhu, L Yang, G Ge, Y Cao, Y Liu, Z Fang, Y Zhang, Selectivity for inhibition of nilotinib on the catalytic activity of human UDP-glucuronosyltransferases, Xenobiotica 44 (2014) 320-325. [59] Y. Liu, J. Ramirez, MJ Ratain, Inhibition of paracetamol glucuronidation by tyrosine kinase inhibitors, Br J Clin Pharmacol 71 (2011) 917-920. [60] KM Knights, JO Miners, Renal UDP-glucuronosyltransferases and the glucuronidation of xenobiotics and endogenous compounds, Drug Metab Rev 42 (2010) 60-70.

32

FIGURE LEGENDS Figure 1. Structures of lapatinib, pazopanib, regorafenib and sorafenib. Figure 2. Inhibition of recombinant human UGT enzymes by lapatinib, pazopanib, regorafenib and sorafenib. Bars represent the mean ± standard error of the parameter fit from quadruplicate experiments. Figure 3. Inhibition of UGT1A1-catalysed bilirubin glucuronidation by lapatinib (panel A), pazopanib (panel B), regorafenib (panel C) and sorafenib (panel D). Points represent the mean ± SEM of quadruplicate measurements. Figure 4. Dixon plots for lapatinib (LAP), pazopanib (PAZ), regorafenib (REG) and sorafenib (SOR) inhibition of human liver microsomal βestradiol (β-EST; panels A, C, E and G, respectively), and propofol (PRO; panels B, D, F and H, respectively) glucuronidation. Points represent the mean ± SEM of quadruplicate measurements.

33

Table 1. Probe substrate (β-estradiol, β-EST; propofol, PRO) and lapatinib (LAP), pazopanib (PAZ), regorafenib (REG) and sorafenib (SOR) concentrations used in inhibition kinetic experiments with human liver microsomes as the enzyme source a,b

a

Probe substrate

Probe substrate concentrations (µ M)

LAP concentrations (µ M)

PAZ concentrations (µM)

REG concentrations (µ M)

SOR concentrations (µ M)

β-EST

3, 6, 15 (2.37, 4.74,11.85)

PRO

50, 100, 200 (35, 70, 140)

0, 1, 2, 3, 4 (0, 0.25, 0.50, 0.75, 1.0) 0, 10, 20, 30, 40 (0, 1.43, 2.86, 4.29, 8.58)

0, 0.50, 1.0, 1.5, 2.0 (0, 0.365, 0.730, 1.10, 1.46) 0, 5, 10, 15, 20 (0, 2.88, 5.75, 11.5, 17.25)

0, 0.05, 0.10, 0.20. 0.30 (0, 0.007, 0.014, 0.028, 0.042) 0, 5, 10, 20, 40 (0, 0.375, 0.75, 1.5, 3)

0, 0.15, 0.30, 0.45, 0.60 (0, 0.021, 0.042, 0.063, 0.084) 0, 5, 10, 20, 40 (0, 0.375, 0.75, 1.5, 3)

Concentrations are shown as the concentrations of compounds added to incubations of HLM, along with concentrations corrected for non-

specific binding (i.e. fumic) in parenthesis. b

See Results for fu mic values of each probe substrate and protein kinase inhibitor.

34

Table 2. IC50 values (µM) for lapatinib (LAP), pazopanib PAZ), regorafenib (REG) and sorafenib (SOR) inhibition of recombinant human UGT enzymesa,b 1A1 0.536 ± 0.069

1A3 2.3 ± 0.36

1A4 9.2 ± 0.33

1A6 3.4 ± 0.33

1A7 1.1 ± 0.04

1A8 1.5 ± 0.06

1A9 1.1 ± 0.05

1A10 8.2 ± 0.01

2B4 NI

2B7 NI

2B15 NI

2B17 NI

PAZ

1.1 ± 0.01

35.5 ± 4.1

11.3 ± 0.26

3.5 ± 0.46

0.805 ± 0.030

0.187 ± 0.002

2.0 ± 0.04

7.5 ± 0.10

NI

19.8 ± 2.7

44.2 ± 14.5

NI

REG

0.045 ± 0.009

19.3 ± 2.6

16.0 ± 0.2

3.9 ± 0.47

0.383 ± 0.013

5.2 ± 0.1

0.261 ± 0.017

8.5 ± 0.01

NI

65.5 ± 21.5

NI

39.1 ± 1.0

SOR

0.066 ± 0.001

27.3 ± 3.0

16.1 ± 0.2

4.2 ± 0.20

0.317 ± 0.029

1.59 ± 0.11

0.306 ± 0.001

7.6 ± 0.12

54.4 ± 6.2

52.2 ± 1.3

58.7 ± 3.9

21.5 ± 0.15

LAP

a

Data presented as mean ± standard error of the parameter fit (from quadruplicate experiments)

b

Concentrations of 4-methylumbelliferone (4MU) employed in the inhibition screening studies (Section 2.3.1) corresponded to the approximate Km or S50 for each recombinant UGT enzyme: 100 µM (UGT1A1); 1100 µM (UGT1A3); 100 µM (UGT1A6); 15 µM (UGT1A7); 750 µM (UGT1A8); 10 µM (UGT1A9); 30 µM (UGT1A10); 300 µM (UGT2B7); 250 µM (UGT2B15); and 4000 µM (UGT2B17). Inhibition of UGT1A4 and UGT2B4 was assessed using lamotrigine (1500 µM) and codeine (2000 µM) as the respective probe substrates. NI – Negligible inhibition; estimated IC50 > 100 µM

35

Table 3. Ki,u values (µM) for the inhibition of human liver microsomal UGT1A1 and UGT1A9 by lapatinib (LAP), pazopanib (PAZ), regorafenib (REG) and sorafenib (SOR)a. Enzyme

LAP

PAZ

REG

SOR

HLM-UGT1A1

0.567 ± 0.034

2.34 ± 0.14

0.020 ± 0.002

0.033 ± 0.004

HLM-UGT1A9

16.8 ± 3.6

16.7 ± 1.6

0.679 ± 0.049

0.678 ± 0.021

a

Data shown as mean ± SEM of quadruplicate experiments.

36

Table 4. Predicted AUC-ratio for a victim drug glucuronidated by human hepatic UGT1A1 and UGT1A9 resulting from inhibition by lapatinib (LAP), pazopanib (PAZ), regorafenib (REG) or sorafenib (SOR).

HLM enzyme inhibited

a

Predicted AUC ratio resulting from inhibition by: LAP

PAZ

REG

SOR

UGT1A1

1.04

1.24

3.00

3.35

UGT1A9

1.00 (1.01)a

1.00 (1.31)

1.06 (1.59)

1.07 (1.67)

Data in parenthesis show predicted AUC-ratio assuming the experimental Ki,u values for

inhibition of UGT1A9 are overestimated by an order of magnitude (see Results and Discussion).

37

38

39

40

41

Graphical abstract

42