Microstructural Evolution of Lipid Aggregates in Nucleating Model and Human Biles Visualized by Cryogenic Transmission Electron Microscopy

Transcription

1 Original Articles Microstructural Evolution of Lipid Aggregates in Nucleating Model and Human Biles Visualized by Cryogenic Transmission Electron Microscopy FRED M. KONIKOFF, 1 DGANIT DANINO, 2 DAPHNA WEIHS, 2 MOSHE RUBIN, 3 AND YESHAYAHU TALMON 2 Obtaining reliable information on the physical state and ultrastructure of bile is difficult because of its mixed aqueous-lipid composition and thermodynamic metastability. We have used time-lapse cryogenic transmission electron microscopy (cryo-tem) combined with video-enhanced light microscopy (VELM) to study microstructural evolution in nucleating bile. A well-characterized model bile and gallbladder biles from cholesterol and pigment gallstone patients were studied sequentially during cholesterol nucleation and precipitation. In model bile, cholesterol crystallization was preceded by the appearance of the following distinct microstructures: spheroidal micelles (3-5 nm), discoidal membrane patches ( nm) often in multiple layers (2-10), discs ( nm), and unilamellar ( nm) and larger multilamellar vesicles (MLVs). The membrane patches and discs appeared to be short-lived intermediates in a micelle-to-vesicle transition. Vesicular structures formed by growth and closure of patches as well as by budding off from vesicles with fewer bilayers. MLVs became more abundant, uniform, and concentric as a function of time. In native bile, all the above microstructures, except discoidal membrane patches, were observed. However, native MLVs were more uniform and concentric from the beginning. When cholesterol crystals appeared by light microscopy, MLVs were always detected by cryo-tem. Edges of early cholesterol crystals were lined up with micelles and MLVs in a way suggesting an active role in feeding crystal growth from these microstructures. These findings, for the first time documented by cryo-tem in human bile, provide a microstructural framework that can serve as a basis for investigation of specific factors that influence biliary cholesterol nucleation and crystal formation. (HEPATOLOGY 2000; 31: ) Abbreviations: cryo-tem, cryogenic transmission electron microscopy; VELM, video-enhanced light microscopy; PC, phosphatidylcholine; TC, sodium taurocholate; CSI, cholesterol saturation index; MLV, multilamellar vesicles. From the 1 Department of Gastroenterology, Tel Aviv Sourasky Medical Center and Minerva Center for Cholesterol Gallstones and Lipid Metabolism in the Liver, Sackler Faculty of Medicine, Tel Aviv University, Tel Aviv; the 2 Department of Chemical Engineering, Technion-Israel Institute of Technology, Haifa; and the 3 Department of Surgery B, Rabin Medical Center, Petah Tikva, Israel. Received April 28, 1999; accepted November 2, Address reprint requests to: Fred Konikoff, M.D., M.Sc., Department of Gastroenterology, Ichilov Hospital, Tel Aviv Medical Center, 6 Weizmann Street, Tel Aviv 64239, Israel. konikoff@post.tau.ac.il; fax: Copyright 2000 by the American Association for the Study of Liver Diseases /00/ $3.00/0 The hydrophobic cholesterol molecules are solubilized in bile by 2 ampiphilic substances, bile salts, and phospholipids, which act as natural detergents, forming water-soluble mixed biliary lipid aggregates. 1 It is generally accepted that bile salt-rich micelles and phospholipid-rich vesicles, in a dynamic metastable equilibrium, function as cholesterol carriers in bile. 2 The presence of additional lipid aggregates or complexes has been suggested, but remains controversial. 3-5 Moreover, despite their pathophysiological importance, the physical configuration and transformation of the lipid structures involved, especially during early cholesterol crystallization, the initial step of cholesterol gallstone formation, are poorly understood. 6 Because of the complex composition of bile, the acquisition of information regarding the physical state and ultrastructure of bile is difficult. The systematic investigation of a bile analogue, composed of the 3 major biliary lipids, cholesterol, cholate, and lecithin, has proven to be a valuable experimental system providing data relevant to native bile. 1 However, the studies have largely relied on indirect methods, such as light and neutron scattering or nuclear magnetic resonance, most of which depend on model assumptions for data interpretation or gel filtration chromatography, which involves sample processing and dilution. Conventional transmission electron microscopy is unsuitable for investigating bile because of staining or drying artifacts. 7,8 We have recently shown that cryogenic transmission electron microscopy (cryo-tem), which is a direct nonperturbing method, devoid of staining and drying artifacts, can be successfully applied to visualize lipid aggregates in model as well as native bile. 9,10 Moreover, combining cryo-tem with video-enhanced light microscopy (VELM) makes it possible to visualize biliary microstructures ranging from 1 nm to 100 microns. 11 The purpose of the present study was to investigate time-dependent microstructural evolution in nucleating model and human bile during the earliest stages, which lead to cholesterol precipitation and crystallization, by combining these 2 direct imaging techniques. 261 MATERIALS AND METHODS Materials. Egg yolk phosphatidylcholine (PC), 99% pure, was purchased from Avanti Polar Lipids, Inc. (Alabaster, AL). Cholesterol and sodium taurocholate (TC), 98% pure, were purchased from Sigma Chemical Co. (St. Louis, MO) and used after recrystallization. All other chemicals and solvents were of analytical reagent grade. The glassware were acid washed and thoroughly rinsed in distilled water. Model Bile. A previously well-characterized model bile, composed of an aqueous solution of cholesterol, TC, and egg yolk PC at

2 262 KONIKOFF ET AL. HEPATOLOGY February 2000 concentrations simulating human gallbladder bile from cholesterol gallstone patients, was used for the experiments. 12 The model bile was prepared from a mixture of cholesterol (in chloroform), PC (in chloroform), and TC (in methanol) in a molar ratio of 18:37:120. After drying under N 2, the mixture was dissolved in chloroformmethanol (2:1, vol:vol) and dried again under N 2 and reduced pressure for 18 hours. The dried lipid film was resuspended in an aqueous solution (0.15 mol/l NaCl, mol/l NaN 3, ph 6-7) to a total lipid concentration of 90 g/dl. This concentrated solution was incubated at 56 C for 1 hour to obtain a clear, isotropic stock solution free of particles under light microscopy. To start the cholesterol crystallization process, 13 the concentrated solution was diluted with 0.15 mol/l NaCl at 37 C to a total lipid concentration of 10 g/dl (preparation pathway A yielding a final lipid composition of 18 mmol/l cholesterol, 37 mmol/l PC, and 120 mmol/l TC). In another set of experiments (preparation pathway B), the model bile was prepared by dissolving the dried lipid film directly to its final concentration (10 g/dl) at 56 C, forming a transparent isotropic solution. Supersaturation and nucleation were then induced by decreasing the temperature to 37 C. 14 It took 5 minutes to cool the solution to 37 C. These final model bile solutions (cholesterol saturation index [CSI] 147%, bile acid/phospholipid ratio 3.2, total volume 3 ml) were then incubated under N 2 at 37 C, and aliquots were taken for analysis at predetermined time points throughout the crystallization process. The time of dilution (or T 37 C in pathway B) was recorded as t 0. In both pathways the solutions were turbid at 37 C. Native Bile. Human gallbladder biles were obtained from gallstone patients at cholecystectomy. Informed consent was obtained according to a protocol approved by the local institutional human subjects committee. All patients had normal liver function tests, functioning gallbladders by preoperative ultrasonography, and no evidence of any hepatobiliary complications. Biliary infection was excluded by culture of bile samples. Bile samples were aspirated from the gallbladders with a large-bore needle through the right side 5 mm trochar under laparoscopic guidance at the beginning of the operative procedure before gallbladder manipulation or dissection. Fresh human bile was centrifuged (Microfuge, 5,000 g, 10 min, 25 C) to get rid of cell debris and then incubated under N 2 at 37 C. The time of aspiration was recorded as t 0. Every several hours during the next few days samples were vitrified as cryo-tem specimens and observed by light microscopy at the same time (see later). Gallstones (when available) were used to determine the stone type by measuring the relative cholesterol content. Over 70% cholesterol by weight was regarded as characteristic for a cholesterol stone and less than 40% for a pigment stone. Chemical Analysis. Biliary lipids were extracted by chloroform: methanol (2:1, vol:vol). Bile salts were determined using the 3-hydroxysteroid dehydrogenase assay. 15 Phospholipids were determined by the phosphorus assay of Bartlett. 16 Cholesterol was measured by the method of Abbel et al. 17 The CSI was calculated using the critical tables of Carey, 18 and the total lipid concentration was calculated using an average bile salt molecular weight of 445. Video-enhanced Light Microscopy. Bile samples were placed on glass slides (Clay-Adams). The specimens were observed at 25 C by an Olympus BH-2 light microscope, operated in the differential interference contrast (DIC) mode. Images from the light microscope were directly recorded on a PC computer equipped with a Cue-4 (Galai, Migdal Haemek, Israel) image analysis system and transferred to a Macintosh G3 computer (Apple, Cupertino, CA). Image processing and contrast enhancement were done using the NIH Image 1.61 and Adobe Photoshop 5.0 software packages (Adobe Systems, Inc., Mountainview, CA). In the nucleating model bile, microstructures were recorded and classified morphologically as previously described. 12 Cryo-Transmission Electron Microscopy. Vitrified specimens for transmission electron microscopy were prepared in a controlled environment vitrification system at 37 C and 100% relative humidity, as previously described. 19 In brief, a drop was applied onto a perforated carbon film, supported on an electron microscopy copper grid, and held by the controlled environment vitrification system tweezers. The sample was blotted and immediately (up to 5 seconds) plunged into liquid ethane at its freezing point ( 183 C). The vitrified sample was then stored under liquid nitrogen ( 196 C) and examined either in a JEOL 2000FX microscope (JEOL, Ltd., Tokyo, Japan) operated at an accelerated voltage of 100 kv or a Philips CM120 microscope (Philips Electron Optics, Eindoven, The Netherlands) operated at 120 kv. Specimens were equilibrated in the microscope at about 180 C, examined in the low-dose imaging mode to minimize electron beam radiation damage, and recorded at a nominal underfocus of 4 to 7 µm to enhance phase contrast. In the JEOL 2000FX microscope, the sample was loaded into a Gatan 626 cooling-holder and images were recorded on Kodak SO-163 film and developed in full-strength D-19 developer (Eastman Kodak Company, Rochester, NY) for maximum electron speed. In the Philips CM120 microscope, an Oxford CT-3500 cooling holder was used. Images were recorded digitally by a Gatan 791 multiscan CCD camera using the Digital Micrograph 3.1 software package (Warrendale, PA). Image processing was performed by the Adobe Photoshop 5.0 package. RESULTS Model Bile. The model bile solution was examined at sequential time intervals during a total of 14 days of incubation after dilution. Samples were inspected by VELM to determine coexisting larger microstructures and to ascertain the expected behavior of the nucleating bile model. 12 At t 0, a few small granular aggregates, at the limit of resolution of light microscopy, were seen. The aggregates became gradually more numerous with time. Figure 1 gives a sample of the structures observed by VELM in the model bile. Small aggregates, observed at 21 hours are shown in Fig. 1A. Tubes, coexisting with less well-defined aggregates, are shown in Fig. 1B at 68 hours. Plates and helices seen later on are shown in Fig. 1C and D. The same sequence of events was observed in the model bile also when it was prepared by direct dissolution to its final concentration (preparation pathway B, see Materials and Methods). In parallel with light microscopy, samples were studied by digital imaging cryo-tem at sequential time intervals of t 0, 1, 3, 6, 18, 25, 30, 42, and 90 hours from the time of dilution. Immediately after dilution (t 0, Fig. 2A), cryo- TEM revealed an abundance of spheroidal micelles, about 3 to 5 nm in diameter. In addition, moderately electron-dense discoid structures (arrows) with a diameter between nm were noted coexisting with the micelles. At the borders of the holes in the support film, 3 to 5 nm wide bilayer membrane patches, probably representing edge-on projections of the above discoid structures, were detected (arrowheads). These patches were often stacked in layers of 2 to 10. Another group of microstructures seen at t 0 were unilamellar vesicles with diameters ranging between 50 and 200 nm and occasional multilamellar vesicles with diameters up to 400 nm. Some of the membrane patches were seen to form distinct early unilamellar vesicles, and some were attached to the outer layer of discrete vesicles (inset in Fig. 2A). Large patches were noted to occupy major parts of the vesicular circumference in a way suggesting gradual growth of an additional, outer bilayer onto the vesicles. At t 1, 3, and 6 hours cryo-tem revealed micelles coexisting with an abundance of multilamellar vesicles (MLV) of various sizes, shapes, and numbers of bilayers. Most of the MLVs had concentric bilayers and a center portion devoid of

3 HEPATOLOGY Vol. 31, No. 2, 2000 KONIKOFF ET AL. 263 FIG. 1. Light micrographs of typical structures formed in the GS model system at increasing times after dilution: (A) Small granular aggregates at 21 hrs, Bar 20 µm; (B) tubes coexisting with less well-defined aggregates are the dominating structures at 68 hrs, Bar 20 µm; (C) cholesterol monohydrate crystals after 165 hrs, Bar 20 µm; (D) a rather large helix observed at 191 hrs, Bar 10 µm. bilayers. However, many contained eccentric vesicles of various sizes and shapes. Unilamellar vesicles were also seen. Free membrane patches were not observed but large patches seemed to develop into partially closed bilayers forming MLVs (see arrows in Fig. 2B). Electron-dense discs ( nm) were also occasionally seen (inset in Fig. 2B). Some MLVs were large, over 2000 nm in diameter, and occasional unilamellar vesicles had a diameter exceeding 300 nm. Some micrographs disclosed multilamellar vesicles, which seemed to be growing by budding off from other vesicles containing fewer bilayers (see inset of Fig. 2C), whereas some of the vesicles were quite elongated, almost tube-like. It is quite possible that very large MLVs were excluded from the specimen during preparation. At t 18 and 25 hours the findings were generally similar to those seen at t 1 to 6 hours, except for the impression of more abundant MLVs of smaller sizes. At t 30 hours the microstructures seen were the same as those between 6 to 25 hours. However, most vesicles were multilamellar, concentric, and relatively uniform with more regular distances (Fig. 3A) between their bilayers. Moreover, their diameter was generally smaller than at the earlier stages (from 100 to 500 nm). Yet, some very large multilamellar vesicles could still be seen. In addition, occasional plate-like structures with the classic appearance of cholesterol monohydrate crystals could be seen. At t 42 and 90 hours the findings resembled very much those seen at 30 hours. The trend of more abundant, uniform, and concentric MLVs was more obvious at these later stages. Some of the MLVs had areas of high electron density, with less distinct bilayers. Figure 3B shows, at relatively low magnification, cholesterol monohydrate plates (asterisks) coexisting with vesicular structures. When the model bile was prepared by direct dissolution to the final concentration (pathway B) only spheroidal micelles were observed at 56 C. On cooling to 37 C the solution became turbid, and the aforementioned microstructural findings occurred with minor differences in terms of frequency and abundance of appearance. The above data were highly reproducible, and the details described were seen across any given specimen and in the 3 repeat experiments performed with the model bile. Human Bile. Seven human gallbladder biles were studied. Five biles were from cholesterol and 2 from pigment gallstone patients. The chemical composition of the stones and biles and cholesterol saturation indices of the biles are presented in Table 1. Six biles (including the pigment stone biles) were supersaturated with cholesterol (CSI 1). When plotted on a ternary phase diagram of the mixed bile salt-lecithin-cholesterol system, the biliary lipid compositions of the supersaturated biles fell within the three-phase region above the micellar zone. 1 Thus, at equilibrium the

4 264 KONIKOFF ET AL. HEPATOLOGY February 2000 gallstone patients, filaments and plate-like structures were found after 3 or more hours of ex vivo incubation at 37 C, and in some of the samples, helical and tubular structures developed later (after 3 or more days). There were differences in the exact times of appearance of the various structures between the samples, but the crystallization sequences followed the expected pathways predicted from their chemical composition. 20,21 Thus, in biles No. 2, 3, 5, and 8, small plate-like crystals preceded or coincided with the appearance of helical and tubular crystals, as expected in pathway C on the equilibrium phase-diagram of corresponding model biles. 20 In the bile with the highest CSI (bile No. 6), plate-like crystals coexisted with the spherical aggregates and no intermediate crystalline structures were observed, corresponding to crystallization pathway D. In the pigment gallstone biles (No. 3 and 4), crystals could be seen only after several FIG. 2. Cryo-TEM images showing the evolution of structures in the GS bile model at increasing times after dilution. (A) At t 0 spheroidal micelles (black dots in the background) coexist with moderately electron-dense discoid structures (arrows); arrowheads point to stacks of membrane patches viewed edge-on; note 1 large vesicle (V) with membrane patches attached. Inset shows a peanut-shaped vesicle with a second membrane building on the first; (B) MLVs of various sizes and numbers of bilayers at t 1 hour. Incomplete bilayers are marked by white arrows; the inset shows a disc at t 3 hrs; (C) MLVs at t 6 hrs; inset shows budding of a vesicle. biles were expected to contain micelles, vesicles, and crystals and to encompass crystallization sequences involving filamentous, helical, and tubular as well as plate-like cholesterol crystals. 20,21 Bile No. 7 was found to be unsaturated (CSI 0.92, see Table 1), thus no de novo crystals were expected to form in that sample. By VELM, spherical aggregates a few microns in size were present in all fresh bile samples immediately after withdrawal from patients. In the supersaturated biles from cholesterol FIG. 3. (A) Multilamellar, fairly uniform vesicles observed at t 42 hrs; (B) a lower magnification image of the specimen in (A), showing cholesterol monohydrate crystals (asterisks) and multilamellar vesicles. The dark coarse network is the perforated carbon film.

5 HEPATOLOGY Vol. 31, No. 2, 2000 KONIKOFF ET AL. 265 TABLE 1. Stone Content and Bile Composition of Model and Native Biles Studied Stone Content Bile Composition Bile No. Type Ch (%) Ca (%) Ch PL BS TLC CSI (mm) (%) (mm) (%) (mm) (%) (g/dl) (%) 1 Model Cholesterol Pigment Pigment Cholesterol Cholesterol Cholesterol Cholesterol days of ex vivo incubation. In bile No. 7, no crystals appeared throughout the entire follow-up. Two native biles were studied with our new digital imaging cryo-tem system, whereas 5 were studied with the older system. Electron micrographs of typical microstructures are presented in Fig. 4. Small spheroidal micelles, unilamellar vesicles and concentric multilamellar vesicles were observed in the bile samples on the day of withdrawal (Fig. 4A) except in bile No. 7, where only micelles were seen. Free bilayer membrane patches were not seen in native biles, although occasional electron-dense discs were observed. The micelles were of the same size (3-5 nm) and shape (spheroidal) as in the model bile. The vesicles were spherical or peanut shaped, with a maximal diameter ranging between 70 to 300 nm. The multilamellar vesicles seen in the fresh human biles were usually concentric, resembling those seen at later stages in the model bile. Their presence by cryo-tem corresponded to the appearance of spherical aggregates in light micrographs. At the stage when plate-like cholesterol monohydrate crystals appeared by light microscopy, with or without intermediate crystalline structures, multilamellar vesicles were always detected by cryo-tem. These coexisted sometimes with unilamellar vesicles and were noted after the initial appearance of unilamellar vesicles and micelles, irrespective whether cholesterol crystals were seen or not. On several occasions cholesterol crystals were captured by cryo-tem. Sometimes MLVs were seen to adhere to the crystal edges, aligning with the crystal surface (Fig. 4B, arrows). We saw this effect even when the vesicles and crystals were not close to a hole edge, when the possible effect FIG. 4. Typical microstructures observed by cryo-tem in human bile 3 hours after collection. (A) MLVs on a background of spheroidal micelles; (B) MLVs adhering to a single cholesterol monohydrate crystal (arrowheads point the borders of the crystals); note how vesicle membranes coat the edges of the crystalline plate (arrows); spheroidal micelles aggregate at the crystalline borders, e.g., at the down-pointing arrowhead. The dark coarse network seen in both micrographs is the perforated carbon support film.

6 266 KONIKOFF ET AL. HEPATOLOGY February 2000 of mechanical squeezing of the liposome onto the crystal is not at play. Although we have imaged many liposomes that are pushed against the hole edge in this system, we never saw the opening of liposomes into free bilayers in a manner seen in Fig. 4B. Thus it seems that the coating of the crystal with bilayers emanating from liposomes is indeed a form of interaction between the liposomes and crystals, which may play a role in crystal growth. Because cholesterol crystals are composed of hydrogen, carbon, and oxygen atoms only, the contrast between them and the vitrified water matrix is very low. Apparently, we do observe them by cryo-tem only when they reach the thickness that gives sufficient contrast. By that time they are many micrometers long, as can be seen in Fig. 3B. The use of digital imaging facilitates the search for cholesterol crystals in the specimen while keeping the electron dose to a minimum. Quite often micelles pile-up at crystal edges (as they do at holes in the support film) during thinning of the specimen to its final dimension. 19,22 This effect can be clearly seen around the cholesterol monohydrate crystal in Fig. 4B (arrowheads). DISCUSSION In the present study, we have employed the combination of VELM and cryo-tem to study sequentially the microstructural evolution of biliary lipid aggregates in model as well as native human gallbladder bile. Our findings underscore the value (and analytical strength) of these techniques in obtaining direct microstructural information and monitoring dynamic processes in bile. These findings enable one to delineate a sequence of microstructural developments, which occur during the earliest stages of cholesterol nucleation leading to crystal formation and precipitation. We have documented several different microstructures appearing in nucleating bile. Micelles were the predominant aggregates present initially in all biles studied and were identical in the model and native biles. Their size (3-5 nm) is consistent with data obtained previously by indirect methods. 1,23 Their shape was spheroidal, confirming prediction by others, as well as our previous cryo-tem observations. No thread-like micelles were seen, as expected in a ternary system containing cholesterol. 10 The predominance of micelles in this study might partly be related to the method of preparation of the model bile. However, observations from the native biles and the 2 preparation pathways in this study, as well as other model bile systems 1,24,25 support the universality of spheroidal micelles in bile. The second structures seen initially were discoidal membrane patches. These have not been previously described in bile. Although slightly larger ( nm vs. 50 nm), they do resemble the primordial vesicles reported by Gantz et al. recently, also using cryo-tem, 24 but are in fact flat membranes before they close to form vesicles. In our study, additional lamellar bilayer structures were noted concomitantly with the patches. The size and timing of these bilayer structures suggest that they are edge-on projections of the above discoid membrane patches. Whether these have any relevance to the lamellar structures observed by others is unclear. 3,4,26 However, their transient nature preceding the abundance of vesicles indicates that they are short-lived ( 1 hours), intermediate structures in the transformation from micelles to vesicles. Previous studies of micelle-to-vesicle transitions were performed primarily by indirect methods, such as quasi-elastic 27 and small-angle neutron scattering. 28,29 These have shown increasing size and elongation of micelles, with the formation of rod-like structures as intermediates in the process. Previous studies on micelle-to-vesicle transition in our laboratory have shown similar patches as transient structures formed during egg PC membrane reconstitution from PC/sodium cholate mixed spheroidal micelles. 30 In the present study, we have observed the discoid bilayer structures and some MLVs in addition to micelles and unilamellar vesicles seen in an earlier study using the same model bile system. 10 This seeming discrepancy is explained by the high resolution digital imaging system, which enables not only better resolution, but also increased versatility in scanning the specimen. Moreover, smaller exposure doses, and thus less electron beam damage to the samples, enable higher quality images. The free membrane patches were, however, not observed in the native biles examined in this study. In our study, supersaturation was induced by diluting a highly concentrated mixed lipid solution (pathway A), or cooling a mixed micellar solution (pathway B), whereas Gantz and co-workers diluted a concentrated mixed micellar solution. 24 All these manipulations enable determination of an initial time point (t 0), but do not precisely reflect the natural situation. Cholesterol and phospholipids are secreted into bile as vesicles, whereas bile salts are secreted separately. 2 Bile salt micelles are then believed to solubilize vesicles during bile flow in the biliary tree, forming mixed micelles and increasingly supersaturated vesicles. These enter the gallbladder where bile is concentrated and starts to nucleate and crystallize. Hence, a micelle-to-vesicle transition may not occur at all in the native situation, within the gallbladder, explaining the lack of these intermediate structures in native biles. Another possibility is, of course, that the free membrane patches were missed in native bile because of sampling error. This, however, is unlikely in view of the negative findings in all 7 human biles examined, and the reproducibility in the model bile. Uni-, oligo-, and multilamellar vesicles were the additional structures detected before and during the process of cholesterol nucleation and crystallization. These were seen in the model as well as in all native biles, including noncholesterol gallstone biles. Despite the accepted notion that unilamellar vesicle aggregation and fusion precede the nucleation process, these phenomena were not noted in the present study. The reason for this is unclear. The available data on these processes are mainly indirect. Although the lack of vesicle aggregation and fusion in the present study, like in that of Gantz et al., 24 does not exclude their existence, it does not support their presumed extensive presence and pivotal role in the nucleation process in bile. The findings of our study are in close general agreement with the observations of Gantz et al. in a similar model system prepared by dilution. 24 They used a less concentrated lipid solution and followed the process at room temperature (22 C) up to 480 hours. Thus, the time schedules cannot be compared. They did observe 2 distinct populations of unilamellar vesicles (sphere/ellipsoid and cylinder/arachoid), as opposed to a continuum of shapes and sizes in our study. This might be explained by the lower temperature, which decreases the rate of events, enabling more prolonged demonstration of primordial vesicles (up to 4 days) and their transformation products (spheroidal vesicles).

7 HEPATOLOGY Vol. 31, No. 2, 2000 KONIKOFF ET AL. 267 In the native biles, MLVs were quite uniform and concentric and had even bilayer distances. In contrast, the model bile MLVs were initially quite irregular, eccentric, and seemed to go through a process of maturation before reaching the final regular forms seen in the native biles. This suggests that the processes observed in the native biles ex vivo represent post factum evolution occurring in a system that has relaxed thermodynamically by cholesterol crystallization and desaturation. Hence, earlier stages of cholesterol nucleation and crystallization, which occur within the gallbladder initially, may not be accessible for ex vivo investigation at all. This, of course, is a general limitation of studying gallstone pathogenesis, which has to be kept in mind. It stresses the importance of complementary studies in model systems, especially those of pathophysiological composition, as done in the present study. It may be added that, because the data of human biles in this study do not essentially differ from those seen in the model system, one can infer that the effects of biliary proteins on the lipid aggregates that form throughout the nucleation process are relatively minor. This does not, however, exclude an effect, because the time sequence and rates may be remarkably altered without affecting the morphology of the lipid aggregates involved. Because of their large size, cholesterol crystals are usually not caught on the cryo-tem grids. However, in the present study, we often saw cholesterol crystals in vitrified specimens of both the model systems and human bile samples using our new digital imaging system, which facilitates the search for specific microstructures at low electron exposure imaging conditions. An interaction of the crystals with both MLVs and micelles was noted. These findings suggest an active interaction, possibly feeding the growth of the crystal by supersaturated MLVs and micelles. Smaller or earlier crystal forms were not seen in the present study by cryo-tem. This may reflect a limitation of the methodology because cholesterol molecules are not very electron dense, as opposed to phospholipids because they lack the phosphate groups the latter have. Because micelles, discoid structures and small unilamellar vesicles were all seen at the earliest time point observed (t 0), it is impossible to categorically determine which came first. However, because vesicles became increasingly abundant as a function of time, it might be extrapolated that they form after the initial presence of micelles and possibly also after the transient membrane patches. Moreover, because free bilayer patches disappeared while those attached to vesicle outer layers became more abundant as a function of time, it strongly suggests that the patches develop into closed bilayer structures and MLVs. Based on the findings of the present study and the earlier reasoning, a sequence of microstructural events can be suggested to occur in supersaturated nucleating bile: Unilamellar vesicles form within a background of spheroidal micelles and develop into oligo- and MLVs in stages, both via adhering membrane patches and budding from other vesicles. MLVs are initially large, irregular, and eccentric but become gradually smaller, concentric, and uniform with increasing numbers of bilayers having regular distances. Cholesterol crystals start to appear after the process of MLV formation. At the growing crystal edges, vesicles as well as micelles rearrange in terms of shape, alignment, and density, suggesting feeding of crystal growth from these lipid aggregates. In this study, no fusion of unilamellar vesicles was demonstrated, and the earliest crystal nucleus or its origin could not be detected. In model bile, discoidal bilayers, or membrane patches, are the first structures that form within the background of spheroidal micelles. These are short-lived intermediates in a micelle-to-vesicle transition. The patches tend to stack upon each other and adhere to vesicles when present. Whether this kind of micelle-to-vesicle transition occurs in vivo within the gallbladder is unclear. Subsequent lipid aggregate transitions are, however, identical in model and native biles. We believe that the details of lipid aggregate transitions and the nucleation process may be further elucidated by experiments using this methodology, selectively focusing on the critical time points before detection of the earliest cholesterol crystals. Acknowledgment: The authors wish to thank H. Laufer, J. Schmidt, and B. Shdemati for excellent technical help and assistance. We thank Dr. A. Kaplun for his involvement in the earlier stages of the work. The cryo work presented was performed at the Cryo-TEM Hannah and George Krumholz Laboratory for Advanced Microscopy. The project was partially supported by the United States Israel Binational Science Foundation, and by a Center of Excellence grant from the Israel Science Foundation, funded by the Israel Academy of Sciences and Humanities. REFERENCES 1. Cabral DJ, Small DM: Physical chemistry of bile. In: Schultz SG, Forte JG, Rauner BB, eds. Handbook of Physiology - The Gastrointestinal System III, Section 6. Baltimore, MD: American Physiological Society, Waverly Press, 1989: Carey MC, LaMont JT. Cholesterol gallstone formation. 1. Physicalchemistry of bile and biliary lipid secretion. Prog Liver Dis 1992;10: Somjen GJ, Marikovsky Y, Wachtel E, Harvey RBC, Rosenberg R, Strasberg SM, Gilat T. Phospholipid lamellae are cholesterol carriers in human bile. Biochim Biophys Acta 1990;1042: Corradini SG, Arancia G, Calcabrini A, Guardia PD, Baiocchi L, Nistri A, Giacomelli L, et al. Lamellar bodies coexist with vesicles and micelles in human gallbladder bile. Ursodeoxycholic acid prevents cholesterol crystal nucleation by increasing biliary lamellae. J Hepatol 1995;22: Cohen DE, Kaler EW, Carey MC. Cholesterol Carriers in Human Bile Are Lamellae Involved? HEPATOLOGY 1993;18: Konikoff F. New mechanisms in cholesterol gallstone disease. Isr J Med Sci 1994;30: Talmon Y. Staining and drying-induced artifacts in electron microscopy of surfactant dispersions. J Colloid Interface Sci 1983;93: Kilpatrick PK, Miller WG, Talmon Y. Staining and drying-induced artifacts in electron microscopy of surfactant dispersions. II Change in phase behavior produced by variation in ph modifiers, stain, and concentration. J Colloid Interface Sci 1985;107: Kaplun A, Talmon Y, Konikoff FM, Rubin M, Eitan A, Tadmor M, Lichtenberg D. Direct visualization of lipid aggregates in native human bile by light- and cryo-transmission electron microscopy. FEBS Lett 1994;340: Kaplun A, Konikoff FM, Eitan A, Rubin M, Vilan A, Lichtenberg D, Gilat T, et al. Imaging supramolecular aggregates in bile models and human bile. Microsc Res Technique 1997;39: Konikoff FM, Kaplun A, Gilat T. Imaging and monitoring cholesterol crystallization in bile. Scanning Microsc 1997;11: Konikoff FM, Laufer H, Messer G, Gilat T. Monitoring cholesterol crystallization from lithogenic model bile by time-lapse density gradient ultracentrifugation. J Hepatol 1997;26: Konikoff FM, Chung DS, Donovan JM, Small DM, Carey MC. Filamentous, helical and tubular microstructures during cholesterol crystallization from bile: evidence that biliary cholesterol does not nucleate classic monohydrate plates. J Clin Invest 1992;90: Holzbach RT, Kibe A, Thiel E, Howell JH, Marsh M, Hermann RE. Biliary proteins. Unique inhibitors of cholesterol crystal nucleation in human gallbladder bile. J Clin Invest 1984;73:35-45.

8 268 KONIKOFF ET AL. HEPATOLOGY February Turley SD, Dietschy JM. Re-evaluation of the 3-alfa-hydroxysteroid dehydrogenase assay for total bile acids in bile. J Lipid Res 1978;19: Bartlett GR. Phosphorous assay in column chromatography. J Biol Chem 1959;234: Abell LL, Levy BB, Brodie BB, Kendall FE. A simplified method for the estimation of total cholesterol in serum and demonstration of its specificity. J Biol Chem 1952;195: Carey MC. Critical tables for calculating the cholesterol saturation of native bile. J Lipid Res 1978;19: Bellare JR, Davis HT, Scriven LE, Talmon Y. Controlled environment vitrification system: an improved sample preparation technique. J Electron Microsc Tech 1988;10: Wang DQH, Carey MC. Complete mapping of crystallization pathways during cholesterol precipitation from model bile: influence of physicalchemical variables of pathophysiologic relevance and identification of a stable liquid crystalline state in cold, dilute and hydrophilic bile salt-containing systems. J Lipid Res 1996;37: Wang DQH, Carey MC. Characterization of crystallization pathways during cholesterol precipitation from human gallbladder biles: identical pathways to corresponding model biles with three predominating sequences. J Lipid Res 1996;37: Talmon Y: Cryogenic temperature transmission electron microscopy in the study of surfactant systems. In: Binks BP, ed. Modern characterization methods of surfactant systems. New York: Marcel Dekker, 1999; Mazer NA, Benedek GB, Carey MC. Quasielastic light-scattering studies of aqueous biliary lipid systems. Mixed micelle formation in bile salt-lecithin solutions. Biochemistry 1980;19: Gantz DL, Wang DQH, Carey MC, Small DM. Cryoelectron microscopy of a nucleating model bile in vitreous ice: formation of primordial vesicles. Biophys J 1999;76: Fudim-Levin E, Bor A, Kaplun A, Talmon Y, Lichtenberg D. Cholesterol precipitation from cholesterol-supersaturated bile models. Biochim Biophys Acta 1995;1259: Tao S, Tazuma S, Kajiyama G. Apolipoprotein-A-I stabilizes phospholipid lamellae and thus prolongs nucleation time in model bile systems an ultrastructural study. Biochim Biophys Acta 1993;1166: Schurtenberger P, Mazer N, Waldvogel S, Kanzig W. Preparation of monodisperse vesicles with variable size by dilution of mixed micellar solutions of bile salt and phosphatidylcholine. Biochim Biophys Acta 1984;775: Hjelm RP Jr, Thiyagarajan P, Alkan H. A small-angle neutron scattering study of the effects of dilution on particle morphology in mixtures of glycocholate and lecithin. J Appl Cryst 1988;21: Long MA, Kaler EW, Lee SP. Structural characterization of the micellevesicle transition in lecithin-bile salt solutions. Biophys J 1994;67: Kaplun, A. Microstructures in phospholipd surfactant systems Technion-Israel Institute of Technology, Haifa. (PhDThesis).