Scaling factors for the in vitro–in vivo extrapolation (IV–IVE) of renal drug and xenobiotic glucuronidation clearance

Abstract

Aim

To determine the scaling factors required for inclusion of renal drug glucuronidation clearance in the prediction of total clearance via glucuronidation (CL_UGT).

Methods

Microsomal protein per gram of kidney (MPPGK) was determined for human ‘mixed’ kidney (n = 5) microsomes (MKM). The glucuronidation activities of deferiprone (DEF), propofol (PRO) and zidovudine (AZT) by MKM and paired cortical (KCM) and medullary (KMM) microsomes were measured, along with the UGT 1A6, 1A9 and 2B7 protein contents of each enzyme source. Unbound intrinsic clearances (CL_int,u,UGT) for PRO and morphine (MOR; 3‐ and 6‐) glucuronidation by MKM, human liver microsomes (HLM) and recombinant UGT1A9 and 2B7 were additionally determined. Data were scaled using in vitro–in vivo extrapolation (IV–IVE) approaches to assess the influence of renal CL_int,u,UGT on the prediction accuracy of the calculated CL_UGT values of PRO and MOR.

Results

MPPGK was 9.3 ± 2.0 mg g⁻¹ (mean ± SD). The respective rates of DEF (UGT1A6), PRO (UGT1A9) and AZT (UGT2B7) glucuronidation by KCM were 1.4‐, 5.2‐ and 10.5‐fold higher than those for KMM. UGT 1A6, 1A9 and 2B7 were the only enzymes expressed in kidney. Consistent with the activity data, the abundance of each of these enzymes was greater in KCM than in KMM. The abundance of UGT1A9 in MKM (61.3 pmol mg⁻¹) was 2.7 fold higher than that reported for HLM.

Conclusions

Scaled renal PRO glucuronidation CL_int,u,UGT was double that of liver. Renal CL_int,u,UGT should be accounted for in the IV–IVE of UGT1A9 and considered for UGT1A6 and 2B7 substrates.

What is Already Known about this Subject

• Multiple drugs, non‐drug xenobiotics and endobiotics are glucuronidated by human kidney.

• UDP‐glucuronosyltransferase (UGT) 1A proteins and UGT2B7 are variably expressed in the tubule components of the human nephron.

• Development of kidney specific in vitro–in vivo extrapolation (IV–IVE) approaches is hampered by the limited availability of microsomal protein per gram of kidney (MPPGK) and knowledge of the content and anatomic distribution of individual UGT enzymes.

What this Study Adds

• This study provides scaling factors to assess the relative contribution of the kidney to renal drug clearance via glucuronidation.

• It demonstrates anatomic differences in renal glucuronidation activity and provides data on the contents of UGT1A6, UGT1A9 and UGT2B7 in human kidney cortex and medulla.

• It establishes that UGT1A9 is 2.7‐fold more abundant in human kidney microsomes than hepatic microsomes, indicating the need to include renal glucuronidation unbound intrinsic clearance in IV–IVE predictions for substrates glucuronidated by UGT1A9.

Introduction

The human kidney has an array of functions that include the regulation of blood osmolarity, volume, ionic composition and pH, the production of renin, 1,25‐dihydroxyvitamin D, erythropoietin, prostaglandins and bradykinin, and the excretion of endogenous metabolites and xenobiotics. The cortex is characterized by the presence of glomeruli, tufts of capillaries and numerous convoluted epithelial structures that form the renal tubules. In contrast, the medulla consists only of tubules and renal blood vessels. A previous study from this laboratory identified that UDP‐glucuronosyltransferase (UGT) 1A proteins and UGT2B7 are variably expressed throughout the tubule components of the human nephron 1. The greatest expression of UGT1A proteins occurred in the cortical proximal convoluted tubules, while variable and lower expression of UGT2B7 was observed in both cortical and medullary tubules. Consistent with a high to low gradient of UGT1A and UGT2B7 expression from the outer cortex through to the inner medulla, we further demonstrated that the intrinsic clearance for S‐naproxen glucuronidation, which is catalyzed predominantly by UGT2B7 (with lesser contributions from UGT 1A6 and 1A9 2), was 4.5‐fold higher in kidney cortical microsomes (KCM) compared with kidney medullary microsomes (KMM) 1. Other reports have similarly demonstrated lower glucuronidation activity in renal medulla 3, 4, 5.

Multiple drugs, non‐drug xenobiotics and endobiotics, including paracetamol, carbamazepine, codeine, dapagliflozin 6, fatty acids, furosemide, gemfibrozil, ibuprofen, ketoprofen, leukotrienes, mefenamic acid, mycophenolic acid, naloxone, naproxen, prostaglandins, propofol (PRO), retigabine and valproic acid have been shown to be glucuronidated by whole human kidney microsomes [7, 8 and references therein]. Indeed, specific activities for the glucuronidation of several drugs (e.g. dapagliflozin, mycophenolic acid and PRO) by kidney exceeds that of liver 6, 9, 10, 11, and there are in vivo data suggesting that renal metabolism contributes significantly to PRO clearance 12. Despite these observations, reliable kidney‐specific in vitro–in vivo extrapolation (IV–IVE) models are not yet available and inclusion of renal glucuronidation data in IV–IVE studies has been limited 9, 13, 14, 15. In particular, the development of kidney‐specific IV–IVE models has been hampered by limited data of microsomal protein per gram of kidney (MPPGK) 13, 14 and knowledge of the abundance and anatomic distribution of individual renal UGT enzymes. Both are essential for generating the scaling factors required to convert in vitro enzyme kinetic data to an in vivo drug clearance via glucuronidation 16.

Quantitative RT‐PCR is used to determine UGT mRNA expression 17, 18 while quantification of UGT protein content has recently utilized isotopically labelled peptides and nano‐UPLC‐MS/MS analyses 19, 20, 21, 22. The latter studies have principally used human liver microsomes (HLM). By contrast, data from human kidney are limited to reports of one to two individuals, the use of pooled kidney microsomes or kidney homogenate 19, 23, 24.

In this study we report MPPGK using five human kidneys and the results of comparative studies of UGT enzyme activity coupled with UGT protein quantification using mixed (whole) kidney microsomes (MKM), KCM, KMM and recombinant human UGT1A9 and UGT2B7. For the first time we provide data on the anatomic distribution and content of UGT 1A6, 1A9 and 2B7, identify the applicability of MKM for in vitro studies and show significant correlations between renal UGT enzyme content and activity. Additionally, unbound intrinsic glucuronidation clearances (CL_int,u,UGT) were determined for PRO and morphine (MOR) 3‐ and 6‐glucuronidation using MKM, human liver microsomes (HLM) and UGT1A9 and UGT2B7. These data were then scaled to assess the impact of renal CL_int,u,UGT on the prediction of IV–IVE for total CL via glucuronidation for PRO and MOR.

Methods

Materials

Alamethicin (from Trichoderma viride) was purchased from A.G. Scientific (San Diego, CA, USA). Zidovudine (AZT), AZT‐glucuronide, β‐estradiol (β‐E2), β‐estradiol‐3‐D‐glucuronide, MOR 3‐glucuronide (MOR(3‐G)), PRO, testosterone (TST), TST‐17‐glucuronide and trifluoperazine (TFP) from Sigma‐Aldrich (Sydney, Australia), deferiprone‐3‐O‐glucuronide and PRO glucuronide from Toronto Research Chemicals (Toronto, ON, Canada), MOR from GlaxoSmithKline (Melbourne, Australia) and MOR 6‐glucuronide (MOR(6‐G)) from Salford Ultrafine Chemicals (Manchester, UK). Deferiprone (DEF) was provided by Dr John Connelly (Apopharma Inc., Toronto, ON, Canada). For UGT protein quantification, stable isotope labelled (SIL) proteotypic peptides in aqueous solution (1 nmol 200 μl^–1) containing 5 or 20% acetonitrile were purchased from Thermo Biopolymers (Ulm, Germany; Material No. HPA97QP010, >97% purity). Ammonium bicarbonate, formic acid, acetic acid, iodoacetamide, dithiothreitol and β‐casein were purchased from Sigma‐Aldrich (St Louis, MO, USA), acetonitrile (HPLC grade) from Fisher Scientific (Pittsburg, PA, USA) and trypsin Gold mass spectrometry grade from Promega (Madison, WI, USA). Pooled HLM (150 donor pool; equal number of male and female donors) and UGT1A9 (batch 2 228 592) and UGT2B7 (batch 38 943) Supersomes™ were provided by Corning Gentest (Palo Alto, CA, USA). All other reagents and solvents used were of the highest analytical grade available.

Human kidney tissue and kidney microsomes

Kidney tissue from Caucasians undergoing radical nephrectomy for malignant disease was provided by the joint Flinders Medical Centre/Repatriation General Hospital Tissue Bank (Adelaide, South Australia). Approval for the collection and use of human kidney tissue for in vitro studies was obtained from the Southern Adelaide Clinical Human Research Ethics Committee. All tissue used was distant from the primary tumour and was reported by a specialist histopathologist to be histologically normal for the age of the donors (43–79 years), with only benign nephrosclerosis and hypertensive‐like vascular changes observed. Whole kidney (mixed cortical/medullary) tissue was available from two females (K12‐K13) and three males (K14‐K16). Paired cortical and medullary tissue (both obtained from the same kidney) was available from five males (K2, K7, K9‐K11, for details of individual subjects refer to Supporting Information Table S1).

Mixed kidney microsomes (MKM, n = 5), kidney cortical microsomes (KCM, n = 5) and kidney medullary microsomes (KMM, n = 5) were prepared by differential ultracentrifugation as detailed by Tsoutsikos et al. 25. Microsomal protein was resuspended either in Na₂HPO₄, pH 7.4 containing 20% (w/v) glycerol (for UGT activity assays) or buffer only for UGT protein quantification. MKM, KCM and KMM pools were prepared by combining equal protein amounts of microsomes from the relevant kidneys. Total protein concentration was measured 26 and all samples stored at −80°C until used.

Determination of microsomal protein g^–1 of kidney (MPPGK)

Correction factors for the loss of kidney microsomal protein during preparation of MKM (K12‐K16) were determined by measurement of NADPH cytochrome c reductase activity. Briefly, either homogenate or microsomal protein (0.1 mg ml⁻¹) was added to a solution (2.2 ml) containing cytochrome c (5 μm), potassium cyanide (300 μm) and 0.1 m phosphate buffer (pH 6.8) and preincubated (5 min, 37°C). Samples (1 ml) were transferred to two quartz cuvettes. Water (10 μl) was added to the reference cuvette and NADPH (10 mm, 10 μl) added to the sample cuvette. Absorbance was monitored at 550 nm and NADPH cytochrome c reductase activity calculated using an extinction coefficient of 27.7 cm^{– 1} mm ^{– 1} 27, 28. The recovery of microsomal protein, and hence fraction of microsomal protein loss during preparation, was determined as (equation (1)):

$$\mathrm{\text{Recovery}}\text{factor}\mathit{=}\frac{\mathrm{\text{microsomal}}\mathrm{\text{cytochrome}}\mathrm{c}\mathrm{\text{reductase}}\left(\mathrm{\text{nmol}}\mathrm{per}\mathrm{\text{tissue}}\mathrm{\text{sample}}\right)}{\text{homogenate}\text{cytochrome}\mathrm{c}\text{reductase}\left(\mathrm{\text{nmol}}\mathrm{per}\mathrm{\text{tissue}}\mathrm{\text{sample}}\right)}$$ (1)

Values of MPPGK corrected for microsomal protein loss during preparation were determined using equation (2) 29:

$$\text{MPPGK}\left(\mathrm{mg}{\mathrm{g}}^{-1}\right)=\frac{\text{specific}\text{activity}\text{homogenate}\left(\text{nmol}\mathrm{cyt}\mathrm{c}\text{reduced}{min}^{-1}\mathrm{m}{\mathrm{g}}^{-1}\right)\times \left(\mathrm{mg}\text{homogenate}\text{protein}{\mathrm{g}}^{-1}\text{kidney}\right)}{\text{specific}\text{activity}\text{microsomes}\left(\text{nmol}\mathrm{cyt}\mathrm{c}\text{reduced}{min}^{-1}\mathrm{m}{\mathrm{g}}^{-1}\right)}$$ (2)

Human kidney microsomal UGT activities

Kidney microsomal UGT 1A1, 1A4, 1A6, 1A9, 2B7 and 2B15/17 activities were determined using β‐E2, TFP, DEF, PRO, AZT and TST as the respective enzyme selective substrates 30. For all activity studies the probe substrate concentrations were four to five times the approximate K _m or S₅₀ for each compound: β‐E2 50 μm (S₅₀ 10 μm, 31), DEF 20 mm (K _m 4 mm, 32), PRO 500 μm (K _m 125 μm, 33), AZT 5 mm (K _m 1 mm, 34) and TST 30 μm (K _m 6 μm, 35). However, due to the substrate inhibition observed with TFP glucuronidation, experiments were performed at the substrate K _m (60 μm, 36). Activity assays were performed in triplicate with MKM (n = 5), KCM (n = 5) and KMM (n = 4). Insufficient KMM was available for K11 for activity studies. HLM were used as the positive control in all activity studies. Microsomal protein concentrations, incubation components, duration of incubation, reaction termination procedure and quantification of β‐E2, DEF, PRO, AZT, TST and TFP glucuronides by HPLC were as described in previous publications from this laboratory 31, 32, 33, 34, 35, 36. Overall within‐day reproducibilities were determined for each assay by measuring glucuronide formation in 8–10 separate incubations of the same batch of pooled HLM at three substrate concentrations (low, mid, high) that spanned the ranges used to characterize enzyme kinetic parameters. Coefficients of variation ranged from 1.8% to 6.9%. The lower limits of quantification assessed as three times background noise, ranged from 0.002 to 0.005 μm, which corresponds to rates of glucuronide formation of <0.5 pmol min⁻¹ mg⁻¹.

PRO glucuronidation and MOR 3‐ and 6‐glucuronidation kinetics by HLM, MKM and recombinant UGT1A9 and UGT2B7

The kinetics of PRO (10 – 300 μm) glucuronidation and MOR (100 – 3000 μm) 3‐ and 6‐glucuronidation by pooled HLM and pooled MKM in the presence of 2% (w/v) BSA and recombinant UGT1A9 and UGT2B7 (Supersomes) were determined according to Rowland et al. 33 and Uchaipichat et al. 37, respectively. The protein concentration of MKM and HLM present in incubations was 0.5 mg ml⁻¹, while the protein concentration of UGT1A9 and UGT2B7 Supersomes in incubations was 0.25 mg ml⁻¹ and 1 mg ml⁻¹, respectively. PRO, but not MOR 38, binds to HLM plus BSA. Thus, the previously determined f _u,mic value of 0.20 was used to correct for PRO binding to incubation constituents 33.

Quantification of human kidney microsomal UGTs and recombinant (Supersome) UGT1A9 and UGT2B7

Protein from MKM, KCM and KMM was diluted to 1 mg ml⁻¹ with Na₂HPO₄ and stored at −80°C. Samples (1 μg) of recombinant UGT1A9 and UGT2B7 were added to 19 μg of rat liver microsomes (final protein concentration 20 μg per sample). Samples were analyzed using minor modifications of previously published methods 19, 20, 39. Briefly, microsomal protein (20 μg, 20 μl of 1 mg ml⁻¹ suspension) was added to a solution of ammonium bicarbonate (50 μl, 50 mm), β‐casein (0.5 μg, 10 μl of 0.05 mg ml⁻¹) and dithiothreitol (10 μl, 40 mm). Following denaturation and reduction at 65°C, samples were carbamidomethylated with iodoacetamide (10 μl, 135 mm). Trypsin (1 μg, 10 μl of 0.1 mg ml⁻¹) was then added to give a microsomal protein to trypsin ratio of 20 : 1 in each sample. Samples were then digested for 4 h at 37°C. Following the addition of acetonitrile to terminate the digestion reaction, a solution (10 μl) containing SIL peptides (1 pmol of each) was added to each sample. Residual acetonitrile was removed in a vacuum before treating with solid phase extraction (Strata‐X 33u Polymeric Reversed Phase, 10 mg ml⁻¹, Phenomenex, Torrance, CA, USA). The solid phase extraction eluate was evaporated to dryness under vacuum and reconstituted in 50 μl modified mobile phase A (water/acetonitrile/formic acid 98/2/0.1; i.e. 2% ACN instead of 1%) before analysis by nanobore UPLC‐MS/MS (selected/multiple reaction monitoring mode). Mobile phase B was 100% acetonitrile. The chromatographic and mass spectrometric instrumentation and conditions were as described previously 20. The separation conditions were 100% A at start, reducing to 58% A at 24 min, then 5% A at 24.5 min for 3 min and 100% A at 28 min for 7 min (total run time 35 min). Multiple reaction monitoring (MRM) data were processed using MultiQuant 2.0 (AB SCIEX). Peaks were smoothed prior to integration and area ratios of unlabelled/SIL peptides were determined using the sum of the two MRMs monitored.

Data analyses

All kidney microsomal UGT activities were performed in triplicate and the values averaged to obtain a single value for each individual kidney. The variation between samples was <5%. Data are reported as mean ± SD (n = 5 kidneys). Correlations between UGT protein content in MKM, KCM and KMM and the respective glucuronidation activities and correlations between the microsomal contents of UGT1A6/UGT1A9, UGT1A6/UGT2B7 and UGT1A9/UGT2B7 were assessed using the non‐parametric Spearman rank method. A value of P < 0.05 was considered statistically significant. PRO and MOR glucuronidation kinetic data were fitted with the Michaelis–Menten equation using EnzFitter (version 2.0; Biosoft, Cambridge, UK).

Unbound intrinsic clearance (CL_int,u) values for PRO and MOR glucuronidation by MKM and HLM were calculated as V _max/K _m and subsequently scaled for microsomal protein yield (mg g⁻¹) and organ weight (g kg⁻¹ body weight) to provide a whole organ CL_int,u,UGT (ml min⁻¹ kg⁻¹). The respective microsomal protein yields (mg g⁻¹ tissue) and organ weights (g kg⁻¹ body weight) used in calculations were 9.3 mg g⁻¹ (see Results section) and 4.5 g kg⁻¹ for kidney 15, and 32 mg g⁻¹ 16 and 21.4 g kg⁻¹ 15 for liver. In the absence of a kidney‐specific model, the whole organ CL_int,u,UGT values were scaled using the well‐stirred model to give a predicted total CL via glucuronidation (CL_UGT) for both kidney and liver (equation (3)).

$$\mathrm{C}{\mathrm{L}}_{\mathrm{UGT}}=\frac{\mathrm{Q}x{f}_{\mathrm{u},\mathrm{b}}\mathrm{x}\mathrm{C}{\mathrm{L}}_{\mathrm{int},\mathrm{u},\mathrm{UGT}}}{Q+\left(f_{\mathrm{u},\mathrm{b}}\mathrm{x}\mathrm{C}{\mathrm{L}}_{\mathrm{int},\mathrm{u},\mathrm{UGT}}\right)}$$ (3)

Organ blood flows (Q) were taken as 16.4 ml min⁻¹ kg⁻¹ for kidney and 20.7 ml min⁻¹ kg⁻¹ for liver 15).The fraction of PRO unbound in blood was taken as 0.015 15, assuming a blood : plasma concentration ratio of 1.0. The fraction of a PRO dose eliminated via glucuronidation was taken as 53% 14, systemic clearance as 108 l h⁻¹ and clearance via glucuronidation as 57 l h⁻¹ 33.

Mean in vivo disposition data for MOR were calculated from values of individual studies reported in the review of Milne et al. 40: plasma clearance following intravenous administration 86.1 l h⁻¹, percent urinary recoveries of MOR 3‐ and 6‐glucuronides, 55% and 10.2%, respectively, fraction or MOR unbound in plasma, 0.75 and blood/plasma concentration ratio 1.15.

Results

Determination of MPPGK

The average weight of mixed (cortical/medullary) kidney tissue available for preparation of MKM was 0.9 ± 0.2 g. Activities of cytochrome c reductase in the mixed kidney homogenate fractions and matched microsomal fractions were 3.3 ± 0.9 (95% CI 2.5, 4.1) μmol min⁻¹ g⁻¹ and 19.1 ± 4.8 (95% CI 14.9, 23.3) μmol min⁻¹ g⁻¹, respectively. The average percentage of microsomal protein lost during preparation was 33.9 ± 11.8% (95% CI 23.5, 44.3), while the mean MPPGK corrected for loss of microsomal protein was 9.3 ± 2.0 mg g⁻¹ (95% CI 7.6, 11).

Mixed kidney microsomal UGT activities

Incubations of pooled MKM did not form the glucuronides of β‐E2 (UGT1A1), TFP (UGT1A4) or TST (UGT 2B15/2B17). In contrast, HLM (used as a positive control) catalyzed the glucuronidation of β‐E2, TFP and TST with respective mean rates of 348, 350 and 108 pmol min⁻¹ mg⁻¹. The mean rates of DEF (UGT1A6), PRO (UGT1A9) and AZT (UGT2B7) glucuronidation by pooled MKM were high, 18 999, 6309 and 1183 pmol min⁻¹ mg⁻¹, respectively. Subsequent activity studies using individual MKM from K12‐K16 and DEF, PRO and AZT established that the rates of glucuronidation of each probe substrate when expressed as either pmol glucuronide (pmol‐G) min^–1 mg⁻¹ microsomal protein or as activity normalized for UGT expression (pmol‐G min⁻¹ pmol⁻¹ UGT), were comparable with the MKM pool (Table 1).

Table 1.

Glucuronidation rates of DEF, PRO and AZT and contents of UGT 1A6, 1A9 and 2B7 using mixed kidney microsomes (MKM). Tabulated data for UGT contents, enzyme activities and normalized enzyme activities represent the mean ± SD (95% CI), n = 5 (K12‐K16). UGT enzyme activities of the MKM pool (K12‐K16) normalized for UGT content are the means of triplicate determinations

		MKM		MKM pool
UGT enzyme	UGT content (pmol‐UGT mg −1 )	UGT activity (pmol‐G † min –1 mg –1 )	Normalized UGT activity (pmol‐G † min –1 pmol‐UGT)	Normalized UGT activity (pmol‐G † min –1 pmol‐UGT )
1 A6	4.7 ± 1.4 (3.5, 5.9)	21 019 ± 7836 (14 151, 27 887)	4395 ± 639 (3835, 4955)	4008
1 A9	61.3 ± 20.1 (44.3, 78.3)	5656 ± 2191 (3736, 7576)	92 ± 6 (87, 97)	103
2B7	37.6 ± 17.4 (22.3, 52.9)	1229 ± 598 (705, 1753)	33 ± 2 (31, 35)	31

^† pmol‐G refers to formation of DEF glucuronide (UGT1A6), PRO glucuronide (UGT1A9), or AZT glucuronide (UGT2B7).

Cortical and medullary microsomal UGT activities

Similar to MKM, formation of β‐E2, TFP or TST glucuronides was not observed using pooled KCM and KMM. Rates of formation (pmol‐G min⁻¹ mg⁻¹ microsomal protein) of DEF (UGT1A6), PRO (UGT1A9) and AZT (UGT2B7) glucuronides by KCM were 1.4‐fold, 5.2‐fold and 10.5‐fold greater, respectively, than for KMM (Table 2). The average cortical : medullary (C : M) ratios of enzyme activity normalized for UGT content were 1.0, 1.1 and 1.8 for UGT 1A6, 1A9 and 2B7, respectively. Comparable enzyme activities were observed for all three substrates when using either a pool of KCM or a pool of KMM. The C : M ratios of enzyme activities for the pools of KCM and KMM were 0.9, 0.9 and 1.4 for UGT 1A6, 1A9 and 2B7, respectively (Table 2).

Table 2.

Glucuronidation rates of DEF, PRO and AZT and contents of UGT 1 A6, 1 A9 and 2B7 using cortical (KCM) and medullary (KMM) microsomes. Tabulated data for UGT contents, enzyme activities and normalized enzyme activities represent the mean ± SD (95% CI), n = 5 (K2, K7, K9‐K11). Enzyme activities of the KCM and KMM pools normalized for UGT content are the means of triplicate determinations

UGT enzyme	UGT content (pmol‐UGT mg −1 )		Content ratio	UGT activity (pmol‐G † min –1 mg –1 )		Activity ratio	Normalized UGT activity (pmol‐G † min –1 pmol‐UGT)		Normalized activity ratio	Normalized UGT activity (pmol‐G † min –1 pmol‐UGT)		Normalized activity ratio (pool)
	KCM	KMM	C : M	KCM	KMM	C : M	KCM	KMM	C : M	KCM pool	KMM pool	C : M
UGT1A6	3.5 ± 2.1 (1.7, 5.3)	1.6 ± 0.7 (1, 2.2)	2.7 ± 2.2 (0.8, 4.6)	10 503 ± 7389 (4027, 16 979)	5255 ± 2465 (3095, 7415)	1.4 ± 0.4 (1.1, 1.7)	3150 ± 1697 (1663, 4637)	3939 ± 2143 (2061, 5817)	1.0 ± 0.9 (0.2, 1.8)	3506	3986	0.9
UGT1A9	40 ± 19.1 (24, 56)	7.5 ± 2.4 (5, 9)	6.2 ± 4.5 (3.2, 9.2)	5072 ± 2864 (2562, 7582)	905 ± 299 (643, 1167)	5.2 ± 4.1 (1.6, 8.8)	119 ± 25 (98, 140)	105 ± 18 (89, 121)	1.1 ± 0.2 (0.9, 1.3)	124	143	0.9
UGT2B7	24.8 ± 15.2 (12, 38)	4.4 ± 1.8 (2.9, 5.9)	7.2 ± 6.2 (1.8, 12.6)	577 ± 417 (212, 942)	70 ± 27 (47, 93)	10.5 ± 13.8 (−1.5, 22)	22 ± 4 (13, 31)	15 ± 6 (10, 20)	1.8 ± 1.4 (0.6, 3)	15	11	1.4

^† pmol‐G refers to formation of DEF glucuronide (UGT1A6), PRO glucuronide (UGT1A9), or AZT glucuronide (UGT2B7).

UGT content of human kidney microsomes and UGT 1A9 and UGT2B7 supersomes

In this study the limit of detection for all UGT proteins investigated, except UGT1A8, was <1.0 pmol mg⁻¹ protein 20. UGT 1A6, 1A9 and 2B7 were expressed in all MKM, KCM and KMM samples analyzed (Figure 1). For MKM, the rank order of mean UGT protein content was UGT1A9 (61.3 ± 20.1 pmol mg⁻¹) > UGT2B7 (37.6 ± 17.4 pmol mg⁻¹) > > UGT1A6 (4.7 ± 1.4 pmol mg⁻¹) (Table 1). The content of UGT 1A6, 1A9 and 2B7 protein was greater in KCM, with C : M ratios of 2.7, 6.2 and 7.2, respectively (Table 2). In KCM, the content of UGT1A6 varied approximately 5‐fold (range 1.2–6.1 pmol mg⁻¹), 4‐fold for UGT1A9 (range 12.8–56.2 pmol mg⁻¹) and 6‐fold for UGT2B7 (range 7.4–43.5 pmol mg⁻¹). An approximate 3‐fold variation was observed in KMM for UGT1A6 (range 0.8–2.6 pmol mg⁻¹), UGT1A9 (range 3.9–9.9 pmol mg⁻¹) and UGT2B7 (2.5–7.2 pmol mg⁻¹). By contrast, UGT 1A1, 1A3, 1A4, 1A5, 1A7, 1A8, 1A10, 2B4, 2B15 and UGT2B17 were not expressed in MKM, KCM or KMM. UGT2B10 was absent in all MKM samples (K12‐K16) and in KCM and KMM for K2, K10 and K11. The UGT1A9 (batch 2 228 592) and UGT2B7 (batch 38 943) contents of Supersomes were 553 pmol mg⁻¹ and 1593 pmol mg⁻¹, respectively.

Figure 1.

UGT protein concentration (pmol‐UGT mg⁻¹ microsomal protein) of UGT1A6, UGT1A9 and UGT2B7 of human mixed kidney microsomes (MKM, n = 5), kidney cortical microsomes (KCM, n = 5) and kidney medullary microsomes (KMM, n = 5). Kidney microsomal samples were analyzed at least in duplicate and the values were averaged to obtain a single value for each individual kidney. K2, K7, K9, K10, K11, K12, K13, K14, K15, K16

Correlations between glucuronidation activities and UGT protein contents

Rates of glucuronidation of DEF, PRO and AZT by MKM, KCM and KMM were highly correlated with the content of UGT1A6 (r _s = 0.87, P < 0.0001), UGT1A9 (r _s = 0.85, P < 0.0001) and UGT2B7 (r _s 0.94, P < 0.0001), (Figure 2), respectively. Analyses between the contents of UGT enzymes identified significant correlations between the pairs UGT1A6/UGT1A9 (r _s = 0.71, P < 0.0001), UGT1A6/UGT2B7 (r _s = 0.76, P < 0.0001) and UGT1A9/UGT2B7 (r _s = 0.89, P < 0.0001). For individual data refer to Supporting Information Table S2.

Figure 2.

Correlations between (A) deferiprone glucuronidation and UGT1A6 content, (B) propofol glucuronidation and UGT1A9 content and (C) AZT glucuronidation and UGT2B7 content of MKM , KCM and KMM . Correlation coefficients r _s were determined using the Spearman rank method.

PRO and MOR glucuronidation kinetics

Kinetic parameters for PRO glucuronidation by MKM, HLM and UGT1A9 determined in the presence of 2% BSA are shown in Table 3. PRO glucuronidation by all enzyme sources was well described by the Michaelis–Menten equation. Mean K _m values were 17, 9.6 and 9.1 μm for MKM, HLM and UGT1A9, respectively. However, the mean V _max value for PRO glucuronidation by MKM was 3.5‐fold higher than that for HLM. Hyperbolic kinetics were similarly observed for MOR 3‐ and 6‐glucuronidation by MKM, HLM and recombinant UGT2B7 supplemented with BSA. The mean K _m values for MOR 3‐ and 6‐glucuronidation by the three enzyme sources (+ BSA) ranged from 455 to 602 μm, consistent with previous reports from this laboratory 37, 38. In contrast to PRO, V _max values determined for MOR glucuronidation using HLM were higher than those measured with MKM.

Table 3.

Derived kinetic constants for morphine (MOR) and propofol (PRO) glucuronidation in the presence of 2% (w/v) BSA, scaled CL_int,u,UGT g^–1 of tissue, and predicted in vivo glucuronidation clearances (CL_UGT) for kidney and liver. The protein sources were pooled mixed human kidney microsomes (MKM), pooled human liver microsomes (HLM), and recombinant human UGT1A9 and UGT2B7

		V max * (pmol min −1 mg −1 protein)	CL int,u,UGT † (μl min −1 mg −1 protein)	CL int,u,UGT (ml min −1 g −1 tissue)		Predicted CL UGT (ml min −1 kg −1)
Substrate/Enzyme	K m * (μm )			Kidney	Liver	Kidney	Liver
PRO/MKM	17 ± 0.7	3070 ± 44	181	1.68	–	0.11	–
PRO/HLM	9.6 ± 0.5	880 ± 12.9	91.7	–	2.93	–	0.9
PRO/UGT1 A9	9.1 ± 0.03	997 ± 2.1	110	0.114	0.15	0.008	0.05
MOR(3‐G)/MKM	492 ± 2.1	1756 ± 3.4	3.6	0.03	–	0.1	–
MOR(6‐G)/MKM	455 ± 10	300 ± 2.0	0.66	0.006	–	0.02	–
MOR(3‐G)/HLM	577 ± 10	2544 ± 16	4.4	–	0.14	–	1.9
MOR(6‐G)/HLM	533 ± 22	570 ± 9.1	1.1	–	0.04	–	0.48
MOR(3‐G)/UGT2B7	602 ± 0.4	835 ± 0.23	1.39	0.0003	0.002	0.0009	0.03
MOR(6‐G)/UGT2B7	574 ± 5.8	183 ± 0.7	0.32	0.00007	0.0005	0.0002	0.007

^* Parameter ± standard error of parameter fit.

^† Unbound intrinsic clearance calculated as V _max/K _m.

3‐G = morphine 3‐glucuronidation.

6‐G = morphine 6‐glucuronidation.

Predicted PRO and MOR CL_UGT by kidney and liver

MKM to HLM CL_int,u,UGT (μl mg⁻¹ min⁻¹) ratios were 2.0 for PRO, and 0.8 and 0.6 for MOR(3‐G) and MOR(6‐G), respectively (Table 3). Predicted PRO glucuronidation CL_UGT by kidney (0.11 ml min⁻¹ kg⁻¹) represented 12% of that in liver while scaled total renal MOR glucuronidation CL_UGT was 5% of that in liver (Table 3). Extrapolated renal/hepatic PRO CL_UGT based on quantification of recombinant UGT1A9 protein was 7% and 6% of the values obtained with MKM and HLM, respectively. Similarly for recombinant UGT2B7 scaled renal/hepatic MOR glucuronidation CL_UGT was 0.9% and 1.6% of CL_UGT obtained with MKM and HLM, respectively.

Discussion

The glucuronidation of numerous drugs and other xenobiotics by human kidney has been demonstrated in multiple studies (see Introduction). However, estimates of renal drug metabolic clearance have used either microsomal protein per gram of liver (45 mg g⁻¹, 13), the yield from rat kidney (6 mg g⁻¹, 9, 41) or a single report of 12.8 mg g⁻¹ 14. The latter study, which used 5 g of human kidney tissue (unspecified region), reported a 3.8‐fold variation in MPPGK (range 4.7–18.2 mg g⁻¹) and a 38% loss of microsomal protein during preparation. Using approximately 1 g of mixed cortical/medullary tissue from five subjects, we determined a mean MPPGK of 9.3 mg g⁻¹ (range 6–12 mg g⁻¹). Loss of microsomal protein during preparation was ~34%, comparable with that of Al‐Jahdari et al. 14, but less than the 47% reported for human liver 29.

The MPPGK determined here is approximately 30% of the value for human liver (32 mg g⁻¹) 16 and therefore estimation of renal drug clearance using hepatic microsomal yield data is clearly inappropriate. A lower microsome yield from kidney in comparison with liver would be anticipated because of the diverse morphology of renal tubule cells that includes well‐developed endoplasmic reticulum (ER) in the cortical proximal tubule cells, but sparse ER in tubule cells of the inner zone of the medulla 42. Additionally, differences in the quantitative distribution of UGTs throughout the human nephron mirror the cortical/medullary gradation of ER 1, 43. It is unknown if MPPGK changes with ageing, but there were insufficient samples in this study to investigate relationships between MPPGK and the age and gender of the donors. (For individual data refer to Supporting Information Table S2).

The mean rates of DEF, PRO and AZT glucuronidation observed for MKM and paired KCM/KMM were reasonably close in value, 18 999, 6309 and 1183 vs. 15 758, 5977 and 647 pmol‐G min⁻¹ mg⁻¹. The difference in AZT glucuronidation accords with the impact of alternative gene splicing on the activity of UGT2B7 in kidney tissue 44. Additionally, rates of DEF, PRO and AZT glucuronidation normalized for UGT expression for pooled kidney microsomes are comparable with the rates obtained with the individual kidney microsomes comprising the pool, irrespective of whether the enzyme source was KCM or KMM. Thus, the use of MKM is an appropriate experimental tool for kidney‐specific IV–IVE models. Previous studies that have reported the use of kidney microsomes for in vitro studies have not specified the anatomic region and hence extrapolated kinetic data may not reflect whole kidney metabolic clearance.

Consistent with other studies reporting lower UGT activity in the medulla than cortex, we found 1.4‐, 5‐ and 11‐fold differences between cortex and medulla in the rates of glucuronidation of DEF, PRO and AZT, respectively. The lower activity in medulla equates with the approximate 3‐, 6‐ and 7‐fold difference between KCM and KMM in the expression of UGT 1A6, 1A9 and 2B7, respectively. As the region of kidney used for preparation of microsomes impacts on glucuronidation capacity 1, use of KCM will result in an over‐estimation and KMM in an under‐estimation of CL_int via glucuronidation, potentially skewing the role of the kidney in the glucuronidation clearance of drugs and xenobiotics. However, normalizing enzyme activity for UGT expression negates the difference in glucuronidation capacity between KCM and KMM. In addition to anatomic differences in renal glucuronidation, cortical blood flow (4.7 ml g⁻¹ tissue min^–1) exceeds that of the medulla (2.3 ml g⁻¹ tissue min^–1) 45. As noted previously, in the absence of kidney‐specific IV–IVE models use of perfusion‐limited kinetics and assumptions of the well‐stirred model may bias IV–IVE predictions 15.

This study is the first to quantify the UGT protein content of MKM, KCM and KMM and confirms that UGT 1A6, 1A9 and 2B7 are the predominant forms expressed in human kidney 19, 24. With the exception of UGT 1A6 and 1A9, no quantifiable protein was detected for any of the other UGT1A enzymes investigated. These data support studies reporting either UGT mRNA or UGT protein expression for UGT 1A6 and 1A9 in human kidney 17, 18, 46. The 13‐fold greater abundance of UGT1A9 compared with UGT1A6 supports a greater role for UGT1A9 in the renal metabolism of xeno‐ and endo‐biotics 8. In contrast, UGT 1A1, 1A3, 1A8 and 1A10 were not detected, which accords with the lack of mRNA expression in kidney 17, 46. UGT 1A4, 1A5 and 1A7 mRNA expression has been reported 18, 46 but we and Fallon et al. 19 found no evidence of the corresponding UGT proteins. Absence of UGT 1A1 and 1A4 protein in renal microsomes is supported by our observation of lack of β‐E2 and TFP glucuronidation.

Similar to UGT 1A6 and 1A9, UGT2B7 protein was detected in all MKM, KCM and KMM samples analyzed. On average, UGT2B7 protein expression was approximately half that of UGT1A9, consistent with the lesser contribution of UGT2B7 to overall renal glucuronidation. Although mRNA expression has been reported for UGT2B4 and UGT2B15/17 46 the corresponding UGT proteins were not detected in this and other studies 19, 24 highlighting the fact that mRNA expression is not predictive at the protein level 47. Additionally, we did not observe glucuronidation of the UGT2B15/17 ‘probe’ substrate TST by kidney microsomes.

Significant correlations were observed here between UGT 1A6, 1A9 and 2B7 enzyme activity and UGT protein content and between UGT1A6/UGT1A9, UGT1A6/UGT2B7 and UGT1A9/UGT2B7 content. Previous studies have reported a significant correlation between UGT1A6/UGT1A9 expression using HLM 21, 48. Correlations between UGT 1A6, 1A9 and 2B7 protein content may be explained by common regulatory factors including PXR, CAR, HNF‐1α and HNF‐4α 49, 50, 51. Additionally, UGT1A9 and UGT2B7 terminate the biological activity of arachidonic acid metabolites and hence are integral to the regulation of renal haemodynamics 7, 8.

Based on experience with the IV–IVE of hepatic drug glucuronidation clearance 30, the results presented here provide an approach for the assessment of the relative contribution of renal and hepatic drug and chemical clearance via glucuronidation from in vitro data. As with the IV–IVE of hepatically eliminated drugs the crucial first step is the accurate prediction of in vitro CL_int (V _max/K _m) with MKM as the enzyme source (or maximal clearance where sigmoidal kinetics are observed). Activation of MKM with alamethicin is necessary to remove the ‘latency’ associated with microsomal UGTs 34 and, since the predominant UGT enzymes in MKM are UGT 1A9 and 2B7, incubations were performed in the presence of 2% (w/v) BSA to overcome the inhibition of these enzymes by inhibitory long‐chain unsaturated acids released from membranes during the course of an incubation 15, 33, 52, 53, 54. However, substrate binding to HLM and BSA must be accounted for in the calculation of K _m (and hence CL_int,u,UGT). The microsomal CL_int,u,UGT may then be scaled to the whole kidney CL_int,u,UGT using the MPPGK reported here (i.e. 9.3 mg g⁻¹), which in turn may be scaled to renal glucuronidation clearance. Although the equation for the well‐stirred model was used here (and previously by others 15), its relevance to the prediction of renal drug glucuronidation clearance in vivo is yet to be established, especially considering blood flow in the cortex and medulla differ 45.

Following this approach, the extrapolated renal glucuronidation CL_int,u,UGT of the UGT1A9 substrate PRO and the UGT2B7 substrate MOR were determined here, along with the predicted renal and hepatic glucuronidation clearances for these drugs. The sum of the extrapolated renal and hepatic PRO glucuronidation clearances was 1.01 ml min⁻¹ kg⁻¹ or 4.2 l h⁻¹ for a 70 kg person. PRO is cleared in vivo by both UGT and cytochrome P450 mediated oxidation, predominately UGT1A9 33 and CYP2B6 14. As indicated in Methods the estimated renal clearance via glucuronidation in vivo is 57 l h⁻¹ 33. Based on the combined renal and hepatic in vitro kinetic data reported here the predicted CL_UGT compared with that observed in vivo is substantially under‐estimated. Of the total PRO glucuronidation clearance predicted from the in vitro data, renal glucuronidation accounted for 11% (i.e. 0.11 vs. 1.01 ml min⁻¹ kg⁻¹). A significant contribution of renal glucuronidation to PRO elimination accords with the observation that extra‐hepatic clearance may account for up to 27% of total clearance in patients undergoing liver transplantation, although the separate renal metabolic clearances via glucuronidation and oxidation were not determined 55. Gill et al. 15 recently reported a comparison of the glucuronidation of a number of compounds, including PRO, by HLM and HKM. In contrast to our data (kidney : liver ratio 0.6), these authors noted a higher kidney : liver ratio (1.63) for extrapolated PRO CL_int,u,UGT. However, as noted by Gill et al. 15, the effect of BSA on in vitro CL_int,u,UGT with HLM as the enzyme source was considerably lower than the approximate 10‐fold increase expected for UGT1A9 substrates and lower than the 11.2‐fold increased they observed for HKM. Had the BSA effect been similar to previous reports, recalculating the data reported by Gill et al. 15 for the contribution of the kidney to PRO glucuronidation clearance provides a value of 11%, identical to this study. Predicted CL_UGT observed here under‐estimated the known clearance of PRO by glucuronidation (viz. 57 l h⁻¹) by 93%. Similarly the in vitro glucuronidation data generated by Gill et al. 15 under‐estimated known PRO CL_UGT by approximately 75%. A number of factors may have contributed to the differences in predicted CL_UGT between the two studies. In the study of Gill et al. 15 in vitro CL_int,u,UGT was estimated using the substrate depletion approach compared with the measurement of glucuronide formation here, the anatomical region used for preparation of kidney microsomes was not specified and a value for MPPGK of 12.8 mg g⁻¹ was employed. Nevertheless, it is apparent that, despite the determination of MPPGK and knowledge of renal UGT abundance, current IV–IVE approaches under‐predict PRO glucuronidation clearance in vivo.

Similar to PRO, the predicted renal glucuronidation clearance of the UGT2B7 substrate MOR was also low (5%) compared with the predicted hepatic clearance (0.12 vs. 2.38 ml min⁻¹ kg⁻¹), which is consistent with the lower expression of UGT2B7 in kidney observed here and by Margaillan et al. 24. The ratio MOR (3‐G) : MOR (6‐G) was 5 : 1 and 4 : 1 for MKM and HLM, respectively. Similar to PRO, extrapolation of the combined hepatic and renal CL_UGT values for MOR 3‐ and 6‐glucuronidation (10.1 l h⁻¹) under‐predicted the observed mean in vivo glucuronidation clearance (approximately 56.2 l h⁻¹ 40) by 82%. We have observed previously that IV–IVE approaches similarly under‐predict the known in vivo glucuronidation clearance of the UGT2B7 substrate AZT 34, 54, 56.

The contents of UGT 1A6, 1A9 and 2B7 reported here also potentially allow the prediction of renal/hepatic drug glucuronidation clearance in vivo from kinetic data generated using recombinant human UGTs as the enzyme source. We determined in vitro CL_int,u,UGT values for PRO and MOR glucuronidation by batches of Supersomes expressing UGT1 A9 and UGT2B7, respectively. The specific contents of these UGTs were determined (see Results) and, as anticipated with over‐expression, the UGT 1A9 and 2B7 contents of Supersomes (pmol‐UGT mg⁻¹) were 10‐fold and 43‐fold greater than the respective pmol‐UGT mg⁻¹ of MKM. However, this was not reflected in proportional increases in glucuronidation activities. Compared with the observed total PRO glucuronidation clearance of 57 l h⁻¹ the use of recombinant UGT1A9 resulted in a substantial under prediction (0.24 l h⁻¹) although the ratio of the predicted renal to hepatic glucuronidation clearance by UGT1A9 (0.16) was similar to MKM : HLM (0.12). Total MOR glucuronidation clearance (0.17 l h⁻¹) using recombinant UGT2B7 was <1% of the predicted in vivo clearance. However, the ratios of MOR(3‐G) to MOR(6‐G) were similar for all protein sources MKM 6 : 1; HLM 4 : 1 and UGT2B7 5 : 1 (c.f. 5.4 in vivo) The poor prediction accuracy based on recombinant enzymes may be accounted for by the high levels of inactive protein when UGTs are expressed in insect cells 57.

In conclusion, the cortex has a 5‐fold greater abundance of total UGT proteins and hence will have a greater glucuronidation capacity for UGT 1A6, 1A9 and 2B7 substrates than the medulla. This is particularly relevant for UGT1A9 substrates as it is 2.7‐fold more abundant in human kidney (61.3 pmol mg⁻¹ microsomal protein) than in human liver (approximately 23 pmol mg⁻¹, 20, 48, 58. Our data support the observation that predicted renal glucuronidation CL_int,u,UGT is 2‐fold higher in kidney than liver for UGT1A9 substrates 15 and hence CL_int,u,UGT should be included in IV–IVE approaches for substrates glucuronidated by UGT1A9. The renal expression of UGT1A6 and UGT2B7 is approximately half that of liver (9.7 and 80.7 pmol mg⁻¹, respectively 20, 58). Using renal UGT expression data from this study and liver expression data from Fallon et al., 20 the kidney : liver UGT expression ratios for UGT1A6 and UGT2B7 are 0.48 and 0.46. The latter accords well with the kidney : liver scaled CL_int,u,UGT ratios of 0.15–0.33 for glucuronidated drugs primarily cleared by UGT2B7 15 and indicates a greater role normally for UGT2B7 in the hepatic metabolism of xenobiotics. However, in situations impacting on hepatic metabolism (e.g. drug–drug interactions, poor metabolizer phenotype, hepatic transporter defects, hepatic disease) or in the determination of renal metabolite formation or toxicity evaluations, renal glucuronidation by UGT2B7 and indeed UGT1A6 may be an important consideration. Further, we show that the use of recombinant UGT 1A9 and 2B7 proteins is appropriate for determining the contribution and fractional metabolism of substrates but recombinant UGT 1A9 and 2B7 substantially under‐predict CL_UGT. Clearly further work is still required on the development of relative activity factors, intersystem extrapolation factors and kidney specific IV–IVE models applicable for predicting renal metabolic clearance in vivo.

Competing Interests

All authors have completed the Unified Competing Interest Form at http://www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare no support from any organization for the submitted work, no financial relationships with any organizations that might have an interest in the submitted work in the previous 3 years and no other relationships or activities that could appear to have influenced the submitted work.

KMK had partial grant support from the Faculty of Health Sciences and Flinders Medical Centre Research Foundation, Flinders University; PCS had partial support from the National Institutes of Health (NIH); Instrumentation grant S10, RR024595.

We thank Virginia Papangelis, Flinders Medical Centre/Repatriation General Hospital Tissue Bank Manager for provision of renal tissue and Dr Charles Crespi and Dr Christopher Patten, Corning Gentest for the provision of UGT expressing Supersomes.

Contributors

Participated in research design: KMK, JOM.

Conducted experiments: KMK, SMS, JKF, NC.

Performed data analysis: KMK, SMS, JKF, NC, PCS, JOM.

Wrote or contributed to writing of manuscript: KMK, JKF, JOM.

Final approval of manuscript: KMK, SMS, JKF, NC, PCS, JOM.

Supporting information

Table S1 Human kidney donor details and drug history. Table S2 UGT protein concentration (pmol‐UGT mg⁻¹ microsomal protein) of UGT1A6, UGT1A9 and UGT2B7 of human mixed kidney microsomes (MKM, n = 5), kidney cortical microsomes (KCM, n = 5) and kidney medullary microsomes (KMM, n = 5) and microsomal protein per gram of kidney (mg g⁻¹, MPPGK) of MKM. Data are tabulated as the means of duplicate (UGT 1A6, 1A9 and 2B7) or triplicate (MPPGK) determinations

Supporting info item

Knights, K. M. , Spencer, S. M. , Fallon, J. K. , Chau, N. , Smith, P. C. , and Miners, J. O. (2016) Scaling factors for the in vitro–in vivo extrapolation (IV–IVE) of renal drug and xenobiotic glucuronidation clearance. Br J Clin Pharmacol, 81: 1153–1164. doi: 10.1111/bcp.12889.