Mechanism and Subcellular Site of Bilirubin Diglucuronide Formation in Rat Liver*

Two different subcellular sites and mechanisms have been proposed for the hepatic conversion of bilirubin monoglucuronide to bilirubin diglucuronide: a microsomal system requiring UDP-glucuronate and a UDP- glucuronate-independent transglucuronidation or dismutase reaction localized at the bile canalicular plasma membrane. To further define these, canalicular plasma membranes were highly purified from rat liver, and the capacity of these to form bilirubin diglucuronide was compared with that of simultaneously isolated he- patic microsomes. The canalicular liver plasma membranes were 48-1 16-fold enriched over homogenate in various canalicular marker enzyme activities; microsomal contamination was <lo% based on the NADPH-cytochrome c reductase activity. No evidence of any conversion of highly purified bilirubin IXa monoglucuronide to bilirubin diglucuronide was found with canalicular liver plasma membranes either in the absence or presence of UDP-glucuronate. In contrast, digitonin-treated microsomes isolated under similar conditions converted 31% of added bilirubin mono- glucuronide (9.4-1 7.1 PM) into bilirubin diglucuronide in 30 min, the reaction being dependent on UDP-glu-curonate. When bilirubin (12.5 MM) was added to the microsomes, 42.3% was converted to bilirubin pellet canalicular mixed membrane sedi- mm in diameter, with a particle size of 10 pm. The bile pigments and their isomers were separated by carrying out the following procedure (17). The flow rate of the mobile phase, which consisted of 5 mM heptanesulfonic acid in 0.1 M acetate buffer, pH 4.8 (solvent A), and acetonitrile (solvent B), was maintained at 2.0 ml/min for 22 min and then abruptly changed to 3.5 ml/min. At zero time, the proportion of solvent A was 75% (v/v) and that of solvent B was 25% (v/v). During the first 20 min of the run, solvent B was increased linearly from 25% (v/v) to 45% (v/v) and then, in the next 7 min, from 45% (v/v) to 80% (v/v) of the mobile phase, following which this ratio of solvents was maintained for the next 7 min, that being the time required to complete the assay. The isomers of bilirubin and its conjugates as well as the species themselves were separated by this procedure.

curonide is formed, with UDP-glucuronate as the glucuronyl donor, by a microsomal glucuronyltransferase (UDP-glucuronyltransferase, EC 2.4.1.17) (8). Second, bilirubin monoglucuronide is converted to bilirubin diglucuronide, but the enzymic process involved and the hepatocellular site of this second reaction remain uncertain. Some investigators have demonstrated that bilirubin diglucuronide is formed in the microsomes by a UDP-glucuronate-requiring transferase (9-12), whereas others have detected the formation of bilirubin diglucuronide from bilirubin monoglucuronide in hepatic plasma membranes enriched in bile canaliculi (13,14). This latter reaction requires that 2 mol of bilirubin monoglucuronide are converted to 1 mol of bilirubin diglucuronide and 1 mol of bilirubin in the absence of UDP-glucuronate. Chowdhury et al. (15) have partially purified from hepatic plasma membranes an enzyme that appears to catalyze this dismutation. Interpretation of the experimental observations has, however, been questioned since it has been reported recently that the formation of bilirubin diglucuronide from bilirubin monoglucuronide by bile canalicular enriched liver plasma membranes in uitro may result from nonenzymic dipyrrole exchange. This latter reaction forms the nonphysiological symmetrical XIIIa and IIIa isomers, which are not expected with the originally postulated dismutase reaction (16). Hence, the role of the bile canalicular plasma membrane in the formation of bilirubin diglucuronide remains controversial. In the present study, we compare the capacity of highly purified canalicular liver plasma membranes and simultaneously isolated microsomes to convert bilirubin monoglucuronide to bilirubin diglucuronide, utilizing a recently developed high performance liquid chromatographic assay that characterizes the isomeric forms of the products (17). The studies provide no evidence for an enzymic formation of bilirubin diglucuronide by canalicular liver plasma membranes and establish that UDP-glucuronate is an absolute prerequisite for bilirubin diglucuronide synthesis by hepatic microsomes. This work has previously been presented in preliminary form (18).

EXPERIMENTAL PROCEDURES
Animals-Male Wistar or Sprague-Dawley rats, weighing 200-250 g, were used in this study. The animals had free access to water, were fed Purina Rodent Chow ad libitium, and were housed in a constant temperature-humidity environment with alternated 12-h light and dark cycles. The animals were routinely killed between 7:30 and 830 a.m.
Isolation of Canalicular Liver P h m a Membranes-Highly purified canalicular liver plasma membranes were isolated from rat liver by a newly developed procedure involving rate zonal flotation procedure and high speed centrifugations through discontinuous sucrose gradients (19). All isolation steps were done at 0-4 "C. 100-110 g of liver tissue were routinely obtained from 10-12 normal fed rats. 10-g portions of liver were minced, washed three times in 80 ml of cold 1 mM NaHC03, and homogenized in a loose Dounce homogenizer (Type , and centrifuged at 1,500 X g for 15 min. The crude nuclear pellet was resuspended in 5.5 volumes of 56% sucrose (w/w) and stirred for 15 min to disrupt membrane aggregates. The sample was then dynamically (3,000 rpm) loaded onto a 100-ml cushion of 56% (w/w) sucrose with a variable speed Sorvall pump into a zonal rotor TZ-28 (Sorvall) and overlaid by 400 ml of 44% sucrose (w/w) and 200 ml of 36.5% sucrose (w/w), respectively. Finally, the rotor was filled to its total volume capacity (1,350 ml) with 0.25 M sucrose. The completed discontinuous sucrose gradient system was centrifuged at 20,000 rpm for 120 min. After slow deceleration to a complete stop, 70 15-ml fractions were collected from the bottom of the rotor. The fractions containing the bulk of plasma membrane fragments (the 44/36.5% sucrose interface) were combined and diluted with NaHC03 (1 mM) to 1,000 ml, and the material was sedimented at 7,500 X g for 15 min. The pellet was gently resuspended (by Vortex mixing) in 250 ml of NaHC03, and the suspension was recentrifuged at 2,700 X g for 15 min. The pellet representing a "bile canalicular enriched mixed liver plasma membrane fraction" (20) was resuspended in 25 ml of 0.25 M sucrose and tightly homogenized (Type B glass-glass Dounce homogenizer) by 50 up and down strokes to dissociate canalicular from basolateral liver plasma membranes. Canalicular liver plasma membranes were separated from the basolateral plasma membrane fragments by centrifugation of the "mixed plasma membrane" fraction through 31% (w/w) sucrose at 40,000 rpm (195,200 X g) for 3 h in a Beckman SW 41 rotor, and the vesiculated canalicular liver plasma membranes, recovered from the top of the 31% sucrose layer, were diluted in 0.25 M sucrose, sedimented at 105,000 X g for 60 min, and finally resuspended in 0.25 M sucrose.
Isolation of Microsomes-Two procedures were used for the isolation of microsomes from rat liver. As a routine source for glucuronyltransferase activity, microsomal (i.e. endoplasmic reticulum) membranes were isolated by differential centrifugation from a 20% homogenate prepared from approximately 10 g of liver in isotonic 0.25 M sucrose, 1 mM EDTA (microsomes A). Since the isolation of canalicular liver plasma membranes was started in hypotonic Na-HCO3 (see above), in some experiments both canalicular liver plasma membranes and microsomes were isolated in 1 mM NaHC03 from the same homogenate. In these experiments, the initial 1,500 X g supernatant was made isotonic (0.25 M) with sucrose and then tightly rehomogenized by 10 up and down strokes in a Type B Dounce homogenizer. Mitochondrial membranes were sedimented at 7,000 X g for 15 min, and microsomes were prepared from the corresponding postmitochondrial supernatant by centrifugation at 105,000 x g for 60 min (microsomes B). These microsomes B were either resuspended in 0.25 M sucrose or 1 mM NaHC03. Isolated membrane subfractions were stored in liquid nitrogen (-70 "C) for up to 4 weeks before use.
Analytical Methods-The purity of the canalicular liver plasma membrane fraction was determined by use of marker enzyme assays: succinate-and NADH-cytochrome c reductase for mitochondria (21), NADPH-cytochrome c reductase for microsomes (21), acid phosphatase for lysosomes (22), and galactosyltransferase for Golgi mem-branes (23). The plasma membrane markers MP-ATPase and ouabain-sensitive (Na+/K+)-ATPase activities were determined by a coupled kinetic assay as modified by Scharschmidt et al. (24). Alkaline phosphatase was assayed usingp-nitrophenyl phosphate as substrate (25). The method of Goldbarg and Rutenburg (26) was used for determination of leucylnaphthylamidase activity. 7-Glutamyl transpeptidase and alkaline phosphodiesterase I activities were measured according to Orlowski and Meister (27) and Razzell (28), respectively. The adenylate cyclase activity in the presence and absence of glucagon (10-8-10-6 M final concentration) was measured by the method of Stewart et al. (29). Protein was determined according to Lowry et al. (30) with bovine serum albumin as a standard.
Source of Bilirubin and Bilirubin Morwglucuronide-Commercial bilirubin was obtained from British Drug House. The purity of the bilirubin was checked by high pressure liquid chromatography as outlined below.
Bilirubin monoglucuronide is not available from commercial sources. To obtain it, one has the choice of either isolating it from bile or synthesizing it from commerical bilirubin with a liver microsomal preparation. The nature of the product and the nature of the experiment will dictate to some extent which source is more suitable for a given study.
In the systems under investigation, bilirubin diglucuronide can potentially be formed from bilirubin monoglucuronide either by enzymic activity or by nonenzymic dipyrrole exchange, as outlined in Table I and Fig. 1. If enzymic glucuronidation occurs by either the UDP-glucuronate-dependent transferase or by dismutation (the reaction postulated at the canalicular level), the isomeric composition of the products will be identical to that of the substrate (in vivo, bilirubin and its products are almost completely in the IXa form; in vitro, the starting isomeric composition of the bilirubin will vary depending on its commercial source but will not be expected to change during these reactions). In contrast, if there is nonenzymic dipyrrole exchange in vitro in our reaction mixture, cleavage of bilirubin monoglucuronide will occur randomly at either side of the central methylene bridge, with random recombination of the resultant dipyrroles. The reaction products arising from this will be a mixture of bilirubin, bilirubin monoglucuronide, and bilirubin diglucuronide in the ratio 1:2:1, and the isomeric composition of the products arising from the IXa isomer will be in each case XII1a:IXa:IIIa in the ratio 121. Since bilirubin monoglucuronide has two IXa isomers (with the glucuronide conjugated to the propionic acid group attached to either C-8 or C-12 of the tetrapyrrole skeleton), there will be four species, in the 1:l:l:l equivalence, resulting from this exchange. With the progress of scrambling, the composition of the reaction mixture will approach its random equilibrium state, both in terms of species and isomeric forms. Since the contribution of dipyrrolic exchange to any reaction is revealed by changes in bilirubin isomeric composition, it was mandatory to be able to ascertain the isomeric composition of the original bilirubin monoglucuronide and of its reaction products to determine the mechanism of any conjugation observed. The high performance liquid chromatographic procedure which we use for our analyses (which is described later) provided us with this capability (17). It permitted us to detect and quantitate the isomeric forms of The structureof bilirubin IXa and the effect random dipyrrole exchange will have when the starting substrate is an equimolar mixture of the two bilirubin 1Xa monoglucuronides. A represents the part of the bilirubin molecule with an endovinyl group and B, that with an exovinyl group. G represents the glucuronide-conjugating moiety. Since the IXa endovinyl and exovinyl monoglucuronides are distinguishable, one will expect three potential bilirubin peaks, four bilirubin monoglucuronide peaks, and three bilirubin diglucuronide peaks in the chromatographic analysis. In the chromatographic runs, the isomeric outflow sequence is, for each species, XIIIa, IXa, and IIIa. bilirubin and its conjugates and, from changes. in isomeric composition, to quantitate the contribution of dipyrrole exchange to product formation under the various experimental conditions utilized.
Bilirubin monoglucuronide was isolated from rat bile and characterized. The rat bile was extracted with chloroform containing 10 mM tetraheptylammonium chloride, and the bilirubin conjugates were separatedby thin layer chromatography on silica gel plates (Whatman LK-5, Technical Marketing Associates, Montreal, Canada). Separation of the pigments was achieved by developing the chromatograms in a solvent system of ch1oroform:methanol:water (60306, v/v) (31). The yellow pigment corresponding to bilirubin monoglucuronide was eluted from the plates with 80% methanol or 50% ethanol and concentrated under NP, and an aliquot of the product was analyzed by high performance liquid chromatography. In this system, the tetraheptylammonium chloride runs with the solvent front; it does not co-migrate with the bilirubin monoglucuronide. If the bilirubin monoglucuronide preparation was found to be contaminated with other bilirubin tetrapyrroles, it was rechromatographed on the thin layer plates until the sample contained only bilirubin monoglucuron-

~~ ~
ide. The bilirubin monoglucuronide purified in this fashion was used for our studies only if it was composed predominantly of IXa isomers. Expressed as a percentage of the total bilirubin monoglucuronide present, the average isomeric composition of seven such preparations (mean f S.D.) was: XIIIa, 3.8 2 0.1%; the IXa-C-8 isomer, 54.7 & 2.2%; the IXa-C-12 isomer, 39.8 2 0.1%; and IIIa, 1.7 2 0.1%. Bilirubin monoglucuronide was also prepared biosynthetically by use of a liver microsomal preparation solubilized with Triton X-100 in the fashion described by Jansen et al. (13) for their original dismutase studies. Its composition was quite different. The isomeric composition of the source bilirubin was: XIIIa, 6.8 +-2.6%; IXa, 78.4 2 3.6%; and IIIa, 14.8 & 1.8% (mean f S.D. of eight determinations). The bilirubin monoglucuronide preparation, extracted with ethyl acetate as originally described by Jansen et al. (13), was contaminated with varying but usually small proportions of bilirubin diglucuronide. The average isomeric composition of the bilirubin monoglucuronide produced was: XIIIa, 12.9%; the IXa-C-8 isomer, 42.5%; the IXa-C-12 isomer, 26.6%; and IIIa, 18.1%. The increase in the relative amounts of XIIIa and IIIa isomers and the decrease in the total of the IXa isomers indicated that random isomerization was occurring when the synthesis and extraction of bilirubin monoglucuronide were being carried out in this way. If scrambling had not occurred, the bilirubin monoglucuronide would have had the same isomeric composition as the parent bilirubin from which the bilirubin monoglucuronide was being prepared.
The isomeric compositions of the bilirubin monoglucuronide prepared from bile and that prepared biosynthetically by the method of Jansen et al. (13) were thus dramatically different. It appeared appropriate to utilize the preparation isolated from bile for most of our bilirubin monoglucuronide studies since this bilirubin monoglucuronide preparation has an almost exclusively IXa composition (corresponding to the in uiuo situation) and, with it, scrambling effects, when they occur, are immediately evident.
Glueuronidation of Bilirubin-The reaction mixture contained at zero time: microsomes at a concentration 1.3 mg of microsomal proteinlml (either untreated or preincubated for 60 min at 4 "C with digitonin, 0.35 mg/mg of microsomal protein (12) In initial experiments, albumin was added to the bilirubin stock solution. As demonstrated in Table 11, albumin at very low concentrations (60.02 mg/ml) had no effect on the rate of glucuronidation of bilirubin or the nature of the products formed. With increase of the albumin concentration to 0.2 or 1.0 mg/ml, although the monoglucuronidation of bilirubin was not affected, the formation of bilirubin diglucuronide was markedly decreased. Under these conditions, bilirubin monoglucuronide was the dominant product formed. When the concentration of albumin was further increased to 2.0 mg/ml, the glucuronidation of bilirubin was almost completely inhibited; bilirubin monoglucuronide was the only product detected in significant amounts. The inhibition of the glucuronidation of bilirubin by high concentrations of albumin has also been noted by others (32). Therefore, albumin was omitted from all subsequent reaction mixtures. Glwuronidatwn of Bilirubin Monoglucuronide-When digitonintreated microsomes were utilized, the reaction was carried out as

TABLE I1
Effeet of varying concentrations of albumin on tfx glucuronidatian of bilirubin Digitonin-treated rat liver microsomes were incubated with bilirubin at 37 'C under the conditions described under "Experimental Procedures" with and without the addition of defatted human serum albumin (fraction V). When canalicular liver plasma membranes were utilized as a potential source of enzyme, the reactions were carried out in the following manner. 100-500 pg of membrane protein were preincubated in 0.9 ml of a 0.1 M KHzPOl buffer, pH 6.4, containing 10 mM glucarol,4-lactone at 25 "C for 1 h. Bilirubin monoglucuronide (9.4-17.1 p~) in 0.1 M Tris-HC1 buffer, pH 7.8, was then added (final pH 6.6, that reported optimal for the dismutase reaction (13)), and the reaction mixture was further incubated at 37 "C for 10 min. The reaction was stopped by placing the reaction vessel in dry ice. The samples were stored at -20 "C overnight.

No. of
Extraction of Bile Pigments from Reaction Mixture-Bilirubin and the conjugates of bilirubin were extracted quantitatively in chloroform containing 10 mM tetraheptylammonium chloride. The mixtures were vortexed and centrifuged, and the chloroform phases were removed, pooled, and taken to dryness under N2. The pigments were then dissolved in ch1oroform:acetonitrile (53, v/v). The concentrations of bilirubin and its conjugates were determined by a recently developed high performance liquid chromatographic technique (17). All procedures were carried out under subdued light.
Analysis of Bile Pigments-The chromatographic procedure used to quantitate bilirubin and its conjugates was, briefly, as follows. The assays were carried out on a Hewlett-Packard 1084 high performance liquid chromatograph with a variable wavelength detector set at 440 nm. Separation of the bile pigments was achieved with an oven temperature at 37 "C using two Hewlett-Packard reverse phase RP-18 columns in series, each 200 mm in length and 4.6 mm in diameter, with a particle size of 10 pm. The bile pigments and their isomers were separated by carrying out the following procedure (17). The flow rate of the mobile phase, which consisted of 5 mM heptanesulfonic acid in 0.1 M acetate buffer, pH 4.8 (solvent A), and acetonitrile (solvent B), was maintained at 2.0 ml/min for 22 min and then abruptly changed to 3.5 ml/min. At zero time, the proportion of solvent A was 75% (v/v) and that of solvent B was 25% (v/v). During the first 20 min of the run, solvent B was increased linearly from 25% (v/v) to 45% (v/v) and then, in the next 7 min, from 45% (v/v) to 80% (v/v) of the mobile phase, following which this ratio of solvents was maintained for the next 7 min, that being the time required to complete the assay. The isomers of bilirubin and its conjugates as well as the species themselves were separated by this procedure.

Characteristics of Isolated Canalicular Liver Plasm Membranes
Details of the isolation procedure and a complete characterization of the isolated canalicular liver plasma membranes are published elsewhere (19). Here we report only on the purity of the membrane preparations. Intracellular marker enzyme activities were depleted in the canalicular liver plasma membranes with respect to homogenate (the relative enrichments of marker enzyme activities over homogenate were 0.1 to 0.7, except for acid phosphatase (relative enrichment = 1.1)). The relative enrichment factor for the microsomal marker NADPH-cytochrome c reductase activity was 0.4 2 0.2 (mean f S.D., n = 8). Since lysosomes and endoplasmic reticulum represent 2 and 24% of total rat liver homogenate protein (33), respectively, canalicular liver plasma membranes were more contaminated with microsomal membranes (24 X 0.4 = 9.6%) than with lysosomes (2 X 1.1 = 2.2%), despite the more favorable relative enrichment factor. Correspondingly, the contaminations with mitochondrial (succinate-cytochrome c reductase) and Golgi (galactosyltransferase) membranes were 1.6 and 0.4%, respectively. Canalicular liver plasma membranes were free of the basolateral markers (Na+/ K+)-ATPase and glucagon-stimulatable adenylate cyclase activities, but were highly enriched with respect to homogenate in the "canalicular marker" enzyme activities leucylnaphthylamidase (48-fold), y-glutamyl transpeptidase (60-fold), alkaline phosphatase (71-fold), M$+-ATPase (83-fold), and alkaline phosphodiesterase I (116-fold). These data dernon-strate the highest degree of purification so far reported for canalicular liver plasma membranes, with a total contamination with intracellular membranes below 15%.

Glucuronidatwn of Bilirubin Microsomes A-In vitro incubations of bilirubin (12.5 p~)
with native microsomes prepared in 0.25 M sucrose and 1 mM EDTA in the presence of UDP-glucuronate but in the absence of digitonin resulted in the formation of bilirubin monoglucuronide (Fig. 2). In the substrate bilirubin, in this case, the relative proportions of the isomeric forms bilirubin XIIIa, IXa, and IIIa were 8.9,73.0, and 18.1%, respectively. Identical isomeric patterns were found in the bilirubin monoglucuronide product. Over 80% of the bilirubin substrate was monoglucuronidated in 30 min, but only a small proportion was converted to bilirubin diglucuronide (less than 4% at 30 min). In contrast, when microsomes were first treated with digitonin at 0.35 mg/mg of protein for 60 min at 4 "C and then incubated with bilirubin (12.5 PM) and UDP-glucuronate, approximately equal proportions of bilirubin monoglucuronide and bilirubin diglucuronide were formed (Fig. 3). When a bilirubin monoglucuronide preparation derived from bile (9.4-17.1 PM) rather than bilirubin was incubated with digitonin-treated microsomes, bilirubin diglucuronide was again formed and its formation was absolutely dependent upon the presence of UDP-glucuronate (Fig. 4). In neither of these reactions did the relative proportions of the isomeric forms of bilirubin (i.e. XII1a:IXa:IIIa) change during the glucuronidation reaction. Importantly there was no significant dipyrrole exchange under the conditions of the reactions (Table I11 and Figs. 2-4). In addition, the isomeric patterns detected in our preparations of bilirubin monoglucuronide were not altered in differential fashion by incubating the substrate for 30 min at 37 "C in buffers at pH 7.85, 6.78, or 6.45; furthermore, during incubation in the assay system with boiled enzyme, the isomeric distribution of a second preparation of the bilirubin monoglucuronide species did not change. These findings emphasize   the stability of the bile-derived bilirubin monoglucuronide preparation under these in vitro assay conditions. Microsomes B-To assess any potential for inactivation of dismutase activity during preparation of canalicular liver plasma membranes, microsomes were also isolated from the same homogenates and under similar conditions as canalicular liver plasma membranes, and their glucuronidation activity was assessed. As illustrated in Fig. 5, the isolation conditions did not modify the microsomal UDP-glucuronyltransferase activity. As with microsomes A (see above), >80% of the bilirubin was glucuronidated very efficiently. The conditions of the assay, however, did modify the activity. It was observed that, when these microsomes were resuspended in hypotonic NaHC03 (1 mM), only bilirubin monoglucuronide was formed; whereas, if they were suspended in isotonic sucrose, both bilirubin monoglucuronide and bilirubin diglucuronide were synthesized in their usual proportions (Fig. 5). These findings suggest that the addition of the second glucuronic acid moiety to bilirubin is somehow dependent on the integrity of certain as yet undefined structural properties of the microsomal membrane and that these properties are modified by changing the osmolality and ionic strength of the medium. Canalicular Liver Plasma Membranes-As illustrated in Fig. 6, no evidence was found with the bile-derived bilirubin IXa monoglucuronide preparation that canalicular liver plasma membranes had any capacity to glucuronidate bilirubin IXa monoglucuronide. The reactions were carried out with or without digitonin treatment of the canalicular liver plasma membranes in the presence and absence of UDPglucuronate at varying concentrations of bilirubin monoglucuronide (9.4-17.1 p M ) and at different pH levels (6.5, 6.8, and 7.8). None of these conditions resulted in the conversion of bilirubin monoglucuronide to bilirubin diglucuronide, either by enzymic reactions or by dipyrrole exchange (Fig. 6 and Table IV). In order to document whether there was any destruction of the bilirubin monoglucuronide substrate or any loss of substrate during the extraction procedure, the amount of bilirubin monoglucuronide added at zero time and that recovered at the end of the incubation were quantitated. No losses occurred; recoveries were 99.0 & 1.2% (S.D.). Bilirubin was also utilized as the substrate and no glucuronidation was observed.
The lack of reaction of the highly enriched liver canalicular membrane preparation with the highly purified bilirubin IXa monoglucuronide was, of course, puzzling. We therefore also used the partly randomized Jansen preparation of bilirubin monoglucuronide (13), which had been used in the original dismutase experiments and which contains small amounts of bilirubin diglucuronide, to determine whether with this we could detect any dismutase activity in either canalicular or basolateral membranes. At pH 6.6, further scrambling of the substrate occurred, associated with increases in both bilirubin and bilirubin diglucuronide. The reaction recorded proceeded at the same rate in the presence of either hepatic canalicular or basolateral membranes (the latter were also harvested at the time of the isolation of the canalicular membranes (19)) and in the presence of appropriate blanks. All of the changes observed could be ascribed to scrambling; no dismutase activity was observed. The addition of 1 p M ascorbic acid to the reaction mixture inhibited the randomization; in its presence, no dismutase activity was found. The results confirm those reported earlier by Sieg et al. (16). They suggest that what was originally interpreted to be a dismutase activity was really randomization. The cause for the instability of the bilirubin monoglucuronide, when prepared and extracted as originally described by Jansen et al. (13), is not clear.

DISCUSSION
These studies confirm our earlier findings and those of others that rat liver microsomes will convert bilirubin in vitro into bilirubin diglucuronide (9-12, Table IV). This microsomal glucuronidation of bilirubin appears to occur in two steps with bilirubin monoglucuronide as an intermediate and uridine diphosphate glucuronate as the aglycon donor. The proposed canalicular bilirubin-glucuronoside glucuronosyltransferase (EC 2.4.1.95), which has been reported to convert bilirubin monoglucuronide to bilirubin diglucuronide by a dismutation in the absence of UDP-glucuronate (13,14), was not detected in our highly purified canalicular or liver plasma membrane preparation. No evidence was found for induction of randomization of the tetrapyrroles by microsomes under the conditions of the reactions, nor was bilirubin detected as a product when highly purified bilirubin IXa monoglucuronide was utilized as a substrate. These data strongly support the concept that the hepatic endoplasmic reticulum is the major subcellular site of formation of bilirubin diglucuronide (16) and demonstrate that the underlying enzymic reaction is dependent upon UDP-glucuronate.
It is difficult to compare the original work postulating the canalicular dismutase reaction with the present study since the isomeric patterns of the substrates used and the products formed were not evaluated in the earlier reports (13, 14). Thus, the possibility exists that the reported in vitro formation of bilirubin diglucuronide by bile canalicular enriched liver plasma membranes was actually due to dipyrrolic exchange (16), and our observations indicate that with the original bilirubin monoglucuronide preparation prepared by Jansen et al. (13) some element, but not canalicular membranes, promotes this kind of randomization. When highly enriched canalicular liver plasma membranes and almost pure bilirubin IXa monoglucuronide were used in the present investigation, however, no dipyrrolic exchange was observed (Table 111 and Figs. 2-4). The canalicular liver plasma membrane subfraction as isolated in this study was highly purified with respect to intracellular organelles and basolateral plasma membranes (19). The conservation of the capacity to form bilirubin diglucuronide in microsomes isolated from the same homogenates and under similar conditions as the canalicular liver plasma membranes (microsomes B; Fig. 5) indicates that the subcellular fractionation procedure did not inactivate that diglucuronidation activity, that the purified bilirubin IXa monoglucuronide molecules were freely utilizable by the conjugating mechanism, and that the purification procedure had not in some way introduced some stabilizing factor. Thus, our data strongly argue against the presence of a bilirubin-glucuronoside glucuronosyltransferase (dismutase) in bile canalicular membranes of normal rat liver. Instead, they reinforce the thesis that microsomal UDP-glucuronate-dependent glucuronyltransferase activity quantitatively accounts for the generation of the predominant product characteristically found in bile, bilirubin diglucuronide. Several recent findings now explain the earlier inability to detect significant in vitro formation of bilirubin diglucuronide by hepatic microsomes (34). First, the relative proportions of bilirubin monoglucuronide and bilirubin diglucuronide formed by microsomal UDP-glucuronosyltransferase are governed by the level of bilirubin present in the reaction mixture (9, 10). Thus, whereas at low concentrations of bilirubin (12 PM or less), bilirubin diglucuronide is the major conjugate, at higher levels of bilirubin (166 I.~M and greater), only bilirubin monoglucuronide is formed. These findings indicate that the assay conditions are crucial for the in vitro formation of bilirubin diglucuronide and that bilirubin can inhibit diglucuronide formation. Second, various alterations of the structural integrity of the microsomal membrane reveal a particular sensitivity of the microsomal UDP-glucuronate-dependent, second step bilirubin glucuronyltransferase to changes in its microenvironment (12). This sensitivity is further emphasized by the present observation that the changes produced by incubation of digitonin-treated microsomes in hypotonic (0.1 mM) NaHC03 buffer destroy the ability of the digitonin-treated membranes to add a second glucuronide in the presence of UDP-glucuronate. The exact nature of and the conditions required for the efficient coupling between the first and second step glucuronidation of bilirubin by hepatic microsomes are under further investigation. The major question which now arises is whether the glucuronyltransferase activities adding the first and second glucuronides to bilirubin are distinct and separate, even though both have now been shown to be localized to the hepatic microsomes.
In summary, the present study indicates that the hepatic endoplasmic reticulum is the major subcellular site of the enzymic formation of bilirubin diglucuronide and that the underlying enzymic reaction is dependent upon UDP-glucuronate. No enzymic conversion of a highly purified bilirubin IXa monoglucuronide to bilirubin diglucuronide was found with a highly purified canalicular enriched liver plasma membrane preparation.