An in vivo Look at Vertebrate Liver Architecture: Three-Dimensional Reconstructions from Medaka (Oryzias latipes)



Understanding three-dimensional (3D) hepatobiliary architecture is fundamental to elucidating structure/function relationships relevant to hepatobiliary metabolism, transport, and toxicity. To date, factual information on vertebrate liver architecture in 3 dimensions has remained limited. Applying noninvasive in vivo imaging to a living small fish animal model we elucidated, and present here, the 3D architecture of this lower vertebrate liver. Our investigations show that hepatobiliary architecture in medaka is based on a polyhedral (hexagonal) structural motif, that the intrahepatic biliary system is an interconnected network of canaliculi and bile-preductules, and that parenchymal architecture in this lower vertebrate is more related to that of the mammalian liver than previously believed. The in vivo findings presented advance our comparative 3D understanding of vertebrate liver structure/function, help clarify previous discrepancies among vertebrate liver conceptual models, and pose interesting questions regarding the “functional unit” of the vertebrate liver. Anat Rec, 2007. © 2007 Wiley-Liss, Inc.

Several conceptual models emerged in the 19th and 20th centuries to describe structure/function relationships of the vertebrate liver; lobular mammalian liver models, and a tubular liver model to describe the lower vertebrate livers of birds, fish, reptiles, and amphibians. The mammalian “classic,” “modified,” and “portal” lobule models describe morphological features encountered in two-dimensional (2D) single sectional views of the liver (e.g. histological preparations, electron micrographs) and attempt to characterize the relationship between vasculature, biliary passageways, and the hepatocellular compartment (Kiernan, 1833; Mall, 1906; Elias and Bengelsdorf, 1952; Rappaport, 1958; Fig. 1). While these models share similarities, discrepancies exist in describing hepatobiliary structure/function. For instance, Kiernan's classic lobule is a hexagonal structure with portal tracts at the hexagon corners and a central hepatic venule, whereas Mall's portal lobule places the portal tract as the central axis of the model. Rappaport's acinar model of the liver has a physiological rather than morphological basis (emphasizing afferent and efferent sinusoidal flow within the parenchyma), and attempts to describe metabolic variance along a periportal to centrolobular gradient (Jungermann, 1988). In the past 20 years “primary” and “secondary” lobule concepts have also emerged (Matsumoto et al., 1979; Saxena et al., 1999; Teutsch et al., 1999). These concepts, based on 3D reconstructions of serial sections of human (Matsumoto et al., 1979) and rodent liver (Teutsch et al., 1999) describe cone-shaped units observed at the tissue level of organization (branching of portal tracts). For instance; secondary lobules (synonymous with the “classic” lobule), comprised of six to eight primary lobules, have a central terminal hepatic vein and six portal tracts at the periphery. Although each of these models remain valuable for communicating normalcy and disease, they are incomplete in conceptually describing vertebrate hepatobiliary structure and function, particularly in 3 dimensions, and, understandably, disparities remain in applying these models across vertebrate species. As such, lobular and acinar models of the liver have never been completely accepted (Rappaport, 1958; Lamers et al., 1989). Moreover, none of these conceptual models have seen successful application to lower vertebrate hepatobiliary structure/function. Such discrepancies likely belie inadequacies in the conceptual models themselves, or rather, our lack of comparative understanding of vertebrate liver architecture.

Figure 1.

A: The “classic” and “portal” lobule, and liver “acinus” characterize mammalian hepatobiliary structure/function relationships. A “tubular” concept has been used to describe the lower vertebrate livers of birds, fish, reptiles, and amphibians. a: The tubular concept in transverse view; two rows of hepatocytes form a tubule lumen (TL) at their apical membranes, while basal membranes border sinusoidal or intercellular/intertubular space. b–d: Findings from three-dimensional (3D) reconstructions reveal “tubular” architecture is more complex than previously considered. The tubule lumen is actually an interconnected network of branching canalicular and bile preductule passageways. Findings from 3D reconstructions also suggest a common/shared structural/functional unit (indicated by brown oval in both conceptual models) among lower and higher vertebrate livers. This functional unit is comprised of (1) a “portal tract/hilus,” a single conduit containing two afferent blood supplies (hepatic portal and arterial vessels), and efferent bile duct(s); (2) a “primary efferent vascular conduit,” central vein/hepatic vein/terminal hepatic vein; and (3) an anastomosing hepatic muralium (predominantly monolayered in mammals, bilayered in lower vertebrates), perfused by a canalicular network and sinusoidal bed, that bridges these two. B: Our investigations, in conjunction with extant mammalian liver information, suggest the vertebrate hepatobiliary system to be constructed on two fundamental architectural motifs: fractal branching in portal tracts (larger bile ductules and ducts) and a polyhedral (hexagonal) motif at the canalicular level of organization. The polyhedral architectural motif (1) appears to order the arrangement of the interconnected canalicular network and individual cells, and (2) may provide maximal structural/functional relationships. The latter is suggested given hexagonal organization is the least perimeter way for space filling, meaning hepatocellular (metabolic) space is maximized, while minimal material is used for construction and maintenance of the hepatocellular membranes, and consequently in the creation of canaliculi. Because blood to bile distance is minimized (Figs. 3, 4), transport is maximized. Lastly, the presence of a hexagonal motif in both the ordered arrangement of the canalicular network, and in the organization of the hepatic lobule, suggest a hexagonal structural motif may provide structural integrity to the organ system in total through the cellular and tissue level of organization. Similarly, fractal branching of the blood and bile conduits in the portal tracts confers maximal afferent and efferent transport efficiency through these fluid compartments. Numbered segments in the hilus and bile duct regions indicate the branching order (fourth order) observed in medaka. These underlying structural principles may serve to order the system toward maximal structural integrity and metabolic and transport efficiency.

Much less is known about the nonmammalian vertebrate liver. The prevailing model for these livers is the hepatic tubule, which emerged from the predominance of observations of a two hepatocyte thick parenchyma and the fact that the vast majority of studies in lower vertebrate liver over the past century have shown no clear lobular formation (Elias and Bengelsdorf, 1952; McCuskey et al., 1986; Hampton et al., 1988; Rocha et al., 1994; Hinton and Couch, 1998). The current tubular concept describes of two rows of hepatocytes (in longitudinal section), the adjacent apical membranes of which form a tubule lumen (bile passageway), into which bile is actively secreted (Figs. 1, 2). Basal membranes of hepatocytes face sinusoidal or intertubular space. Of interest; the “tubular” formation is widely held as a common hepatobiliary architecture found among all vertebrate species; it is the shared predominant phenotype in all embryonic vertebrates, to include humans, and has been used to characterize the adult phenotype for amphibians, birds, reptiles, and fishes (Elias, 1949; Arias, 1988). While the hepatic “tubule” formation predominates among lower vertebrates throughout their life span, mammalian liver undergoes transition from a tubular to laminar (muralium) architecture that is complete in man by age 5 (Arias, 1988). “Tubule” formations (denoting more than one row of hepatocytes) are also observed in liver regeneration in mammalian species after severe injury, such as after hepatectomy and from marked acute exposure to toxins/toxicants (Van Eyken et al., 1989; Vandersteenhoven et al., 1990).

Figure 2.

Noninvasive in vivo imaging of hepatobiliary structure/function in living STII medaka. A: STII medaka embryo 6 days postfertilization (dpf), left lateral view. The fluorescent cytochrome P4503A metabolite resorufin (red) in transport through the intrahepatic biliary networks (IHBPs) of the embryonic liver (L), and concentration in the gall bladder (GB). Otic vesicle (Ov), yolk sac (Y). B: Tubule-like formations, seen in transverse section, in the developing parenchyma, observed in vivo at 6 dpf. C: Concentrative transport of resorufin from hepatocyte into the tubule lumen (TL/IHBP). D1: STII medaka, 30 dpf, left lateral view, brightfield microscopy. Green algae can be seen transiting the gut lumen. D2: An in vivo confocal image of liver from the region of interest (grey square) in frame (D1), showing fluorophore transport (FITC) through IHBPs. E1–E4: In vivo confocal imaging of blood to bile transport. E1: Two rows of hepatocytes and their nuclei (HN) are seen in longitudinal section. Red blood cells (elongated ovate structures) were observed circulating through sinusoids (S/r) abutting the basal membranes of hepatocytes. E2: Single section from in vivo confocal image stack showing conservative β-Bodipy C5 Phosphocholine (green fluorescence) in transport from sinusoid (S) to canalicular space (IHBP), imaged in vivo 30 min. Post administration of the fluorophore to medaka (aqueous bath). E3: Composite of E1 and E2, localizing fluorophore concentrate to IHBPs. E4: Surface map of region of interest (ROI) of inset (white square) in frame E2; illustrating concentrative transport of the fluorophore from sinusoidal space (S) to bile space (IHBP). The ROI, showing a 17-μm span from blood to bile (across hepatocyte), was quantitatively assessed, and revealed β-Bodipy C5 Phosphocholine concentration to be ∼66 times greater in the biliary passageway than in sinusoidal space. These studies demonstrated in vivo evaluation of concentrative blood to bile transport of fluorescent probes was possible.

From the brief review above it can be understood that our comparative understanding of the vertebrate liver is increasingly important. Not only is the tubular liver structure the most ubiquitous hepatic phenotype in the world, a better comparative understanding of the vertebrate liver enhances our ability to interpret and communicate normalcy and disease across the variety of animal models employed in research. By example, small fish animal models such as medaka and zebrafish are proving increasingly invaluable to the study of vertebrate development, carcinogenesis, and for investigation of molecular mechanisms of disease and toxicity (Wittbrodt et al., 2002; Shima and Mitani, 2004; Berghmans et al., 2005; Alestrom et al., 2006).

The 3D in vivo findings we present here not only elucidate hepatobiliary architecture of the lower vertebrate liver phenotype in medaka, but may help better our comparative understanding of various conceptual models applied to vertebrate liver structure/function.


For decades, various color mutant strains of medaka (Oryzias latipes), acquired from natural and commercially available populations, have been maintained in the Laboratory of Freshwater Fish Stocks at Nagoya University, Japan. Cross-breeding from these stocks was used to produce a stable “transparent” strain of medaka (STII), homozygous recessive for all four pigments (iridiophores, leucophores, xanthophores, melanophores; Wakamatsu et al., 2001). STII medaka allow high resolution (<1 μm) noninvasive in vivo imaging of internal organs and tissues at the subcellular level. Using laser scanning confocal microscopy (LSCM) and fluorescent probes to elucidate the hepatobiliary system, we imaged liver structure and function at various stages of development in over a hundred individual living medaka. Of these, 15 medaka were used for 3D reconstructions presented here (3 medaka at 8 days postfertilization [dpf], 4 medaka at 12 dpf, 3 medaka at 30 dpf, 3 at 40 dpf, 2 medaka at 60 dpf). With widefield and confocal fluorescence microscopy, salient features of the organ system such as canaliculi, space of Disse, endothelial cells, biliary epithelial cells, red blood cells, and hepatocytes and their nuclei, were clearly resolved in vivo (Fig. 2). Confocal stacks from in vivo imaging of the hepatobiliary system were used for 3D reconstructions, and from these, architectural, morphometric, and volumetric analyses were made.

In vivo investigations included medaka embryos, larvae, and juveniles, from organogenesis (∼50 hours postfertilization) through 60 days. Medaka were exposed by means of aqueous bath (23°C) to the fluorescent probes Bodipy C5 Ceramide, Bodipy C5 HPC and fluorescein isothiocyanate to facilitate in vivo investigation of hepatobiliary structure/function. Each of the fluorophores saw hepatobiliary uptake and transport and excretion into the gut lumen. Exposed cohorts were typically comprised of 10 individual medaka. At various time points, individual medaka were removed from exposure cohorts, sedated with 10 μM tricaine-methane sulfonate (MS-222), mounted in aqueous medium (Embryo Rearing Medium for dechorionated embryos and hatchlings and de-ionized water for 20 dpf larvae and older) on depression well glass slides with cover slip, and imaged live using brightfield, and widefield and confocal fluorescence microscopy. Time points of study varied from 10 min to 2 hr depending on the fluorophore used and the aspect of the hepatobiliary system being studied. For instance, Bodipy C5 HPC and fluorescein isothiocyanate accumulation in the hepatic parenchyma is first observed at 10 min after exposure, with maximal saturation of the fluorophore in the hepatic parenchyma occurring at ∼30 min. Time points between 15 and 45 min post fluorophore exposure were commonly used for in vivo studies.

Imaging systems: Confocal fluorescence microscopy was performed with a Zeiss 510 Meta system and Zeiss LSM 5 Axiovision image acquisition software, an Argon and HeNe laser, Carl Zeiss C-apochromatic 40x/1.2, and C-apochromatic 10x/0.45. For wide-field fluorescence microscopy, a Zeiss Axioskopp with DAPI/TRITC/FITC filter cubes was used. Excitation/emission parameters for filter cube sets were: DAPI/UV (Ex 360–380 nm/Em All Vis >400 nm); FITC (Ex 450–490 nm/Em 515–565 nm); TRITC (Ex 528–552 nm/Em 578–632 nm). Gross anatomical imaging was performed using a Nikon SZM 1500 dissecting microscope with a Nikon DXM 1200 digital capture system (brightfield). Regarding software, image analysis and compilation was performed with EclipseNet (Nikon, USA), Adobe Photoshop (Adobe, Inc.), Amira 3D (Mercury Computer Systems, Berlin), ImageJ (V1.32j), IP Lab software (Scanalytics, Inc., version 3.55), and Zeiss Image Browser (Carl Zeiss). Fluorescent probes used were 7-benzyloxyresorufin (10–50 μM); β-Bodipy C5-HPC [BODIPY® 581/591 C5-HPC (2-(4,4-difluoro-5- (4-phenyl-1,3- butadienyl)-4-bora-3a,4a-diaza-s- indacene- 3-pentanoyl)-1-hexadecanoyl- sn-glycero-3-phosphocholine), (30 nM–10 μM)]; Bodipy FL C5-ceramide [N-(4,4-difluoro-5,7-dimethyl-4-bora-3a,4a-diaza-s-indacene-3- pentanoyl) sphingosine, (500 nM–5 μM)]; and fluorescein isothiocyanate (1 nM–50 μM). Fluorophores were acquired through Invitrogen/Molecular Probes (Carlsbad, CA). All fluorescent probes were administered to STII medaka by means of aqueous bath in concentration ranges given, at room temperature, under dark conditions.

All transmission electron microscopy (TEM) was performed at the Laboratory for Advanced Electron and Light Optical Methods (LAELOM), College of Veterinary Medicine, North Carolina State University. Individual medaka were anesthetized and fixed in 4F:1G fixative (4% formaldehyde and 1% glutaraldehyde in a monobasic phosphate buffer with a final pH of 7.2–7.4 and a final osmolality of 176 mosmol). Thin sections (Spurr resin embedded) were then examined using a FEI/Philips EM 208S Transmission Electron Microscope.


Polyhedral (Hexagonal) Architecture of the Intrahepatic Biliary System in Medaka

The 3D reconstructions from in vivo imaging revealed intrahepatic biliary passageways (IHBP) in medaka to be predominantly an interconnected network of canaliculi and bile preductules, the organization of which appeared to be based on a polyhedral (hexagonal) structural motif (Figs. 3–6). The polyhedral motif was most clearly observed in the arrangement of IHBPs, and was not evident in sinusoidal architecture. Empirical observations found IHBP branching to be distinctly angular, and statistical analysis of IHBP branching angles (in 3D) revealed clustering of angles into three main groups: 120 degrees (74.9%), 90 degrees (14.8%), and 58 degrees (10.3%). The majority of angles measured averaged 120 degrees, the interior angle of the hexagon (Fig. 4). The 3D angle measurements were made by randomly selecting intrahepatic biliary passageway bifurcation and trifurcation points in 3D reconstructions, achieved by rotating the 3D model and randomly selecting branching points. All angles of a bile segment branching point were measured, inclusive of all 3D planes (Fig. 3). Further 3D analyses revealed IHBP architecture to be based, conceptually, on a tessellating (3D) hexagonal mesh (Carle et al., 2001). This hexagonal motif was also apparent in the ratio between bile segment length and blood to bile distance (Figs. 3, 4). When bile segment (canaliculi) length was considered to correspond to one side of a hexagon, the distance (radii) from hexagon side (canaliculus) to hexagon center (sinusoid) appeared to correlate to the hexagonal ratio given in Figure 3. A hexagonal architecture has also been described, from empirical observations, in the fine structure of the biliary system in rat (Mochizuki et al., 1988; Murakami et al., 2001) and human (Yamamoto et al., 1990). Although a discussion of this topic is beyond the scope of this study, the presence of a polyhedral structural motif likely imparts structural integrity, and metabolic and transport efficiency, features provided by higher-order organizing principles such as hexagonal architecture, not uncommon in biological systems (Mainzer, 1996).

Figure 3.

Examples from three-dimensional (3D) reconstructions of medaka hepatobiliary system, illustrating the types of analyses that led to findings on polyhedral (hexagonal) architecture. Grayscale background images are sections from in vivo confocal image stacks, from which 3D reconstructions were generated. In (A–D,F–H) the empty space surrounding biliary passageways (green/gold), and between sinusoids (red), is hepatocellular space (hepatocytes not rendered for visual clarity). Colored dots in A–C are given for anatomical reference. A–D: An isolated section of the intrahepatic biliary network, illustrating relationship of canaliculi (green) to sinusoids (red). A: Branching canaliculi (intrahepatic biliary networks [IHBP]) are shown between two rows of hepatocytes (confocal grayscale image). B: Same as A with sinusoids present, and C is A&B viewed directly on the “z” axis (from top). C: Example of analyses of polyhedral structural motif, which was evident in the relationship between blood to bile distance and canalicular length (see Fig. 4). Bile segment length (canaliculus, S) correlated to calculated and measured values of blood to bile distance (r), when assessed with the hexagonal ratio [r = 0.5 (S) 1.44]. D: Empirical observations (3D) of angular canalicular and bile preductule branching led to statistical analysis of branching angles, which suggested the presence of a polyhedral structural motif (see Fig. 4). E: Schematic showing modified “tubular” concept of lower vertebrate liver (note IHBPs, and the hexagonal arrangement in relation to hepatocytes/tubular profile). In medaka, we found this structural arrangement of branching canaliculi that comprises the “tubule lumen,” revealing “tubular” architecture more complex than previously considered (see Fig. 1). F: 3D reconstruction of the relationship between IHBPs and sinusoids. Between 2 sinusoids (red) can be seen a branching canalicular network (green/gold). G–I: Examples of 3D morphometric analyses, which suggest medaka parenchymal architecture to be more consistent with an anastomosing muralium. 3D measurements of hepatocellular space (between sinusoids) revealed a muralium-like architecture. G: Example dimensions given are 196.3 μm in length, 98.2 μm in height, and 24.7 μm in width. H: Same as G viewed in a horizontal longitudinal section (from the top). Sample dimensions are shown. I: Single slice from in vivo confocal stack of hepatic parenchyma. Gray arrowheads indicate sinusoids, black arrowheads hepatocyte nuclei. Hepatocellular compartment width (analyzed in vivo and in 3D) was found to be predominantly two cells thick, although occasionally up to eight cells in some areas. Average parenchymal width was 24.58 μm. Parenchymal depth varied from 21.2 μm to > 200 μm. Note that typical dimensions of 3D reconstructions were 100 μm in depth; hence, depth dimension is limited. Parenchymal metrics were derived from 127 measurements (3D) in medaka at 8, 12, 30, and 40 dpf. Space of Disse (SD), Sinusoidal Endothelial Cell (SE), Red Blood Cell (R), Sinusoid (Sn, S1–S4).

Figure 4.

Morphometric and volumetric analyses (upper left graph) of random measures (341 observations) of three-dimensional (3D) branching angles of bile segments in four individual livers (8–40 days postfertilization [dpf]) found branching angles to cluster into three primary groups: 120 degrees (∼75%), 93 degrees (∼15%), and 64 degrees (∼10%). Upper right graph: Using the hexagonal ratio in the lower portion of the plot, a polyhedral structural motif was also suggested in the relationship between blood to bile distance and canalicular length. Measured values of blood to bile distance (r) correlated to calculated values, falling between the longest (R) and shortest (r) radii in the hexagon [r = 0.5 (S) 1.44]. Where bile segment length (e.g., canaliculus) = S, blood to bile distance = r. Pearson's correlation coefficient (0.049, P = 0.6). The difference in the mean between calculated and measured r values was 0.67 μm, half the width of the canaliculus. In vivo volumetric investigations in three individual medaka at 8 dpf, 12 dpf, and 30 dpf revealed consistent ratios between intrahepatic biliary passageway volume, blood (sinusoid) volume, hepatocellular volume, and parenchymal volume.

Figure 5.

Three-dimensional (3D) architecture of liver hilus, bile ductules/ducts, and canalicular network. A: In vivo image capture of the liver hilus at 30 days postfertilization (dpf). Three primary intrahepatic ducts (IHDs) conjoined a common hepatic duct (HD), which fed the cystic duct (CD), gall bladder (GB), and common bile duct (CBD). The liver hilus of medaka contained two afferent blood supplies (the hepatic portal vein [HPV] and hepatic artery [not shown]), and hepatic duct. Mucosal folds (MF) of the gut (lower left) are seen caudal to the liver. B,C,F: The 3D reconstructions aided in elucidating the relationship between the canaliculo-preductule network, bile ductules and ducts, and liver hilus. Images C (horizontal longitudinal section) and F (sagittal view) show an example of 3D reconstruction of the canaliculo-preductular network draining to a common IHD (one of three IHDs shown in B). E: A hand-traced schematic of an isolated section of the canaliculo-preductular network (diamond-shaped region of interest [ROI] indicated in C and F). One individual IHD (green arrowhead) was followed and the branching of the IHBPs mapped to a 2D schematic. Each colored line corresponds to one bile segment (∼11 μm in length). Colors are for illustrative purposes only. The 90-degree angles correlate to bile segment branching (not to scale or degree, for illustrative purposes only). The schematic is used to illustrate the interconnectedness of the canaliculo-bile preductular network. For clarity, part of the network was abbreviated by using dashed lines to illustrate feedback and interconnectedness within the IHB network. Dashed lines represent a branching network of canaliculi (as illustrated in colored lines). The canaliculo-bile preductule network occupied ∼95% of the liver body (by area). D: Schematic showing overall biliary architecture of medaka liver. Bile ductules and ducts exhibited fourth order fractal branching, and were largely found in hilar/peri-hilar region of the liver. Ductules terminated in a canaliculo-bile preductular network (illustrated in E), that followed a polyhedral (hexagonal) architectural motif.

Figure 6.

A–A2: Canaliculus/bile preductule and hepatocellular relationships. A1,A2: Three-dimensional (3D) reconstructions from confocal image in A. Four to six hepatocytes (hepatocyte nuclei shown in purple, numbered) were observed in relation to a given canaliculus (green). Circled red A indicates a bifurcating canalicular segment. A single hepatocyte was observed to contribute to one to three canaliculi. B1–B2: The 3D reconstructions of confocal image in (B), illustrating the structure of a bile preductule (BPD) junction. Red arrowheads in (B) indicate bile preductular epithelial cells (BPDECs) within the hepatic parenchyma. In 2D confocal sections BPDECs appeared to occupy the “center” of these canaliculo-bile preductule structures. BPDECs formed unique junctional complexes with surrounding hepatocytes, which created biliary passageways, termed bile preductules (BPD). The 3D reconstructions (B1,B2) elucidated the actual architecture of these canaliculo-bile preductule complexes, revealing bile preductular epithelia form multiple junctional complexes with surrounding hepatocytes, at which bile preductules are formed. In B1 and B2, the empty space surrounding BPDEC is hepatocellular space (left “empty,” or not rendered, for illustrative purposes). In vivo findings were corroborated with ultrastructural studies (C1,C2), which aided in elucidating/validating in vivo observations. Frame C is a still image from a 3D reconstruction illustrating the distribution of BPDECs (purple, indicated by red arrowhead) throughout the hepatic parenchyma. Junctional Complex/Bile Preductule (Jc/BPD), Intrahepatic Biliary Passageway (IHBP), Hepatic Nuclei (HN), Sinusoid/red blood cell (S/r), Bile Preductular Epithelial Cell (BPDE/C).

Biliary System and Hepatocytes

Linear segments of IHBPs (canaliculi) averaged 11 μm in length and 1.3 μm in diameter, and exhibited unary, binary, ternary, and quaternary branching. This finding should not be confused with fractal branching patterns (Turcotte et al., 1998); the biliary system in medaka does not appear (except for the localized area at the liver hilus) to follow a fractal branching tree motif. Rather, there is ubiquitous feedback/interconnectedness within the 3D biliary network (Fig. 5). Four to six hepatocytes were associated with each linear segment of IHBP (Fig. 6) and arranged such that a single hepatocyte could contribute to two or three bile segments (canaliculi/bile-preductules). The same hepatocellular/canalicular relationship has also been described in rat (Motta and Fumagalli, 1975), where the same side of a hepatocyte was observed to bound two or more bile canaliculi. While hepatocytes had a mean diameter of 11.3 μm, they showed varying morphology in vivo, apparently adaptive to the space they were filling (e.g., at bifurcating or trifurcating junctions). Hence, while the intrahepatic biliary network showed a more ordered structure, hepatocellular organization and morphology appeared much more adaptive to space filling and parenchymal organization.

The IHBP network, composed of canaliculi and bile preductules, occupied ∼95% of the liver corpus uniformly, with each area of livers examined (n = 15) containing approximately equal volumes of IHBPs relative to hepatocellular and vasculature volumes, regardless of how distal/proximal an observed volume of liver was from the liver hilus. Canaliculi were observed to merge with bile preductules at unique morphological sites formed by junctional complexes between bile preductular epithelial cells (BPDECs, discussed following) and hepatocytes (Fig. 6). Together, canaliculi and bile preductules comprised the IHBP network. This interconnected network of IHBPs was observed to feed three primary intrahepatic bile ducts (IHD), lined by cuboidal biliary epithelia. IHDs merged at the liver hilus into a common hepatic duct that was associated with the cystic and common bile ducts (Fig. 5). Cuboidal and large squamous biliary epithelial cells (non-BPDECs) were observed primarily localized to the hilar and peri-hilar regions of the liver.

Parenchymal Architecture

Of interest, while “tubule” like formations were observed (in vivo and in single sectional views) in embryonic livers (3 dpf to 7 dpf), tubule formations were not readily apparent in 3D reconstructions (n = 15) of larval and older livers. Rather, parenchymal architecture in larval and later developmental stages of medaka appeared to be more consistent with an anastomosing, predominantly dual layered, muralium. First, 3D reconstructions revealed the “tubular” lumen to be composed, not of an extended linear biliary passageway that would create a classic “lumen,” but of a hexagonal network of branching, interconnected canaliculi and preductules, which formed the inner parenchymal framework (Figs. 3–6). In other words, tubule lumens were found to be canalicular segments with an average length between 10 and 13 μm. Second, hepatocellular muralium-like structures were apparent, which varied in height from ∼20 μm to >100 μm, and width from ∼23 μm to 83 μm, observations consistent with an anastomosing muralium (because confocal stacks were ∼100 μm in depth, the maximum measurement for plate “height” was limited to 100 μm). Third, the longest non-branching biliary segments were observed to be ∼30 μm in length, and these were few in number (∼2–3% of bile segments measured). For a tubular architecture to be present, tubules would, from these 3D investigations, be less than 30 μm in linear length and would be highly branched and interconnected (following biliary network architecture). Hence, 3D in vivo observations found medaka parenchymal structure to more resemble a predominantly two- to three-cell-thick hepatic muralium, with an average width of 26 μm and height that varied from 20 μm to > 100 μm. It follows that medaka parenchymal architecture appeared highly analogous to mammalian architecture; although predominantly two to three hepatocytes thick (and in rare instances up to seven or eight cells thick), as opposed to the predominantly one- to two-cell-thick mammalian muralium (Elias and Bengelsdorf, 1952).

Biliary Epithelia

3D investigations found BPDECs (Hinton and Pool, 1976; Hampton et al., 1988; Okihiro and Hinton, 1999) to be located throughout the parenchyma of medaka livers studied (from 8 dpf to 60 dpf). BPDECs are the putative correlary of mammalian progenitor/stem cells, given they share morphological characteristics ascribed to mammalian oval cells (bipotential progenitor cells). BPDECs, like oval cells, are phenotypically distinct from hepatocytes, characterized by a high nuclear to cytoplasm ratio with no basement membrane, and are intimately associated with the biliary system (Golding et al., 1996; Fausto and Campbell, 2003). In vivo, BPDECs were observed to form unique junctional complexes with hepatocytes, at which were created IHBPs termed bile preductules. These junctional complexes were morphologically distinct (Fig. 6). In single section view these unique morphological formations appeared as BPDECs surrounded by bile passageways on all sides (Fig. 6). 3D reconstructions from in vivo imaging revealed BPDECs to occupy the “center” of these complexes, where one or more BPDECs formed multiple junctional complexes with surrounding hepatocytes (Fig. 6). Biliary passageways (bile preductules) formed at these BPDEC/hepatocyte junctional complexes commonly showed ternary or quaternary branching. While BPDECs varied in morphology and size, showing ovate, triangular and ellipsoidal nuclei, the majority of BPDECs were commonly found to be ∼6 μm in diameter. Our in vivo findings on BPDECs in medaka are, interestingly, consistent with previous oval cell studies in rodents and humans (Farber, 1956; Golding et al., 1996; Theise et al., 1999; Fausto and Campbell, 2003; Knight et al., 2005). While we have localized BPDECs to specific locations within hepatic parenchyma, and these cells share morphological characteristics consistent with mammalian hepatic progenitor cells, BPDECs in medaka have only been partially characterized (Okihiro and Hinton, 2000). Cuboidal and squamous biliary epithelia (non-BDPEC), lining larger diameter intrahepatic biliary passageways, were observed almost exclusively near hilar and perihilar regions of the liver (Fig. 5).

Morphometric and Volumetric Information

3D reconstructions enabled highly accurate volumetric and ratiometric analyses of biliary, parenchyma, vasculature, and liver volumes (Fig. 4). We found prior ex vivo volumetric studies (Hess et al., 1973; Blouin et al., 1977; Rocha et al., 1997; Hinton et al., 2001) in vertebrate livers to correspond well with our in vivo findings.


Surprisingly, we did not observe a well-developed arborizing biliary “tree,” well characterized in mammalian studies (Ludwig et al., 1998; Masyuk et al., 2001). This “tree” describes bile ducts of the liver hilus arborizing into more highly branched, and numerous, bile ducts and ductules of diminishing diameter, as the biliary system infiltrates more distal regions of liver (relative to the liver hilus). We attribute the overall lack of an arborizing biliary tree in medaka liver to two factors: (1) mammalian studies are describing portal tract arborization, or interlobular stromal areas of the liver (meaning biliary tree arborization describes the tissue level of organization, lobule formation); (2) medaka liver is the architectural analogue of the intralobular mammalian parenchyma. The two points are discussed in detail below.

These findings raise important questions regarding current conceptual models of vertebrate liver architecture. In mammalian liver, hepatocytes have been described as irregularly shaped polygonal cells that form, predominantly, a one- to two-cell-thick wall/plate-like structure (muralium), which anastomoses throughout the liver (Elias and Bengelsdorf, 1952). Observations of dual layered hepatocytes in fish and other lower vertebrate livers, and the lack of observed lobular formations in the majority of lower vertebrate livers studied (McCuskey et al., 1986; Hampton et al., 1989; Rocha et al., 1997; Hinton et al., 2001; Akiyoshi and Inoue, 2004) led to the reasoning that lower vertebrate livers may be composed primarily of anastomosing “tubules.” In single section view (e.g. histological preparations with longitudinal orientation), hepatic tubules appear as two rows of hepatocytes, the apical membranes of which form a tubule lumen (bile passageway), and basal membranes of which border sinusoidal, intracellular, or intertubular space. As our 3D in vivo findings have shown, the true structure of the “tubular” liver is more complex than previously understood. Because one hepatocyte may contribute to one to three canaliculi, “tubule lumens” are actually composed of an interconnected hexagonally branching network of canaliculi and bile preductules. Where “tubule-like” formations were observed in 2D transverse sectional views in the embryonic liver (in vivo), “tubule” formations were very rarely encountered in larval and older fish, and were not readily apparent in 3D reconstructions. For a tubular architecture to be considered, tubules in medaka would, from our 3D in vivo investigations, average ∼ 29 μm in linear length (∼ three hepatocytes), and would be quite rare (2–3% of bile segments measured were found to span up to 29 μm). The prevailing 3D structure of the biliary system in medaka was a highly branched and interconnected network of canaliculi and bile preductules (bile segments) that averaged 11–13 μm in length (Fig. 5).

Because 3D investigations reveal medaka hepatobiliary architecture, representative of the lower vertebrate tubular phenotype, to more closely resemble a dual-layered anastomosing muralium, akin to the single-cell-thick muralium described in mammals, important questions are raised. If hepatic tubules comprise a two-hepatocyte-thick muralium, how are they integrated into a muralium-like structure? Although tubule formations may theoretically comprise a dual-layered muralium, this would necessitate an intertubular space, which raises another question; if tubules are present, are tubules joined at the intertubular space? If so, does this not describe a muralium structure? These important questions remain to be answered, given our findings from in vivo 3D reconstructions suggest a dual-layered plate-like muralium predominates in larval and later life stages of medaka, whereas tubule-like formations were observed in the embryonic liver.

The tubular concept of the lower vertebrate liver was derived largely from 2D observations of histological and electron micrograph (EM) preparations. In histological sections, the fine structure of the biliary system is difficult to discern (canaliculi average 1–2 μm in diameter), and when resolved, we now know, due to a hexagonal architectural formation, would show a random order of appearance in a 2D section. For instance, using histological sections Rocha et al. (1994) described biliary passageways in trout as appearing randomly dispersed throughout the liver; this would be an accurate 2D observation, or how the 3D hexagonal architecture of the canalicular network we have described would appear in a 2D histological section. Consequently, from a 2D perspective, a dual-layered muralium may appear, or be interpreted as, a tubule-like formation. Hence, given the interconnected 3D biliary architecture we have elucidated in this study, it may be that a “tubular” formation and dual layered plate-like muralium are perhaps, one and the same, and that discrepancies in understanding muralium architecture in lower vertebrates have arisen from varied 2D viewpoints (and thereby varied interpretation), and a lack of 3D studies in vertebrate liver structure at the canalicular level of organization.

If fish and mammals share similar muralium architecture, what can explain the absence of “portal tracts” and lobule formation in lower vertebrate livers? Addressing these questions involves consideration of organ system ontogeny and the nature of the “functional unit” of the vertebrate liver. Although this is beyond the scope of this article, a brief discussion is warranted. First, while portal tracts/triads are well described in mammalian liver and integral to current mammalian liver conceptual models (lobular, acinar), they are not often observed in lower vertebrates. In lower vertebrates, anatomical structures that may be perhaps the evolutionary, or even functional, precursors to portal tracts have been described as venous biliary arteriolar tracts (VBAT), venous arteriolar tracts (VAT), venous biliary tracts (VBT), biliary tracts (BT), and arteriolar tracts (AT) (Rocha et al., 1995; Hampton, 1988). These morphological features bear some anatomical resemblance to the portal tract, although no semblance of lobule formation in lower vertebrates has been found. Second, based on our 3D reconstructions, it can be hypothesized that the hepatobiliary systems of both medaka and mammals share a common fundamental functional unit: (1) a “portal tract/hilus,” a single conduit containing two afferent blood supplies (hepatic portal and arterial vessels), and efferent bile duct(s); (2) a “primary efferent vascular conduit,” central vein/hepatic vein/terminal hepatic vein; and (3) anastomosing hepatic plates/cords, perfused by a canalicular network and sinusoidal bed, that bridges these two. A similar conceptual functional unit, the hepatic microcirculatory subunit (HMS), was proposed by Ekataksin et al. (1997). This wedge-shaped unit is composed of “base” (the portal tract/hilus), “apex” (the central vein/terminal hepatic vein/hepatic vein), and a continuous system (muralium) of anastomosing hepatic plates (laminae hepatic) connecting the base and apex, perfused by sinusoidal bed (labyrinthus hepatic). Hence, it can be considered that, in the lower vertebrate liver of medaka; the liver hilus is the correlary of the portal tract, and hepatic vein the corollary of the central vein/terminal hepatic vein. In between these is the laminae hepatic, the fine structure of which, up to this study, has remained a mystery in medaka and other lower vertebrates.

One of the outstanding questions in human biliary architecture has been; does each duct of Hering correspond to one canaliculus? Or do canaliculi form a confluence before entering the canals of Hering (Saxena et al., 1999)? Our 3D studies in medaka show the canalicular network to form a confluence before entering the hilar intrahepatic ducts, the anatomical corollary of the canals of Hering. Given the findings presented here and that: bile canalicular diameter is conserved in vertebrate livers (1–2 mm); early observations by Elias and Bengelsdorf (1952) suggest a polygonal hepatocellular formation; that a “chicken-wire” pattern of bile canaliculi has been described in human liver (Ekataksin et al., 1995); hexagonal architecture has also been described in the fine structure of the biliary system in rat (Murakami et al., 2001) and human (Yamamoto et al., 1990); the tubular structure (dual layered parenchyma) appears to be conserved among all embryonic vertebrates; erosion cast studies by Murakami et al. (2001) show a hexagonal canalicular network feeding intrahepatic bile ducts (rat); bipotential progenitor-like cells in both medaka and mammals are closely associated with the biliary system, it is not unlikely that all vertebrate livers share the same fundamental functional unit. Hence, it is interesting to consider the medaka hepatobiliary system, representative of lower vertebrate architecture, as a single functional unit (akin to the HMS), while mammalian hepatobiliary systems can be considered to be composed of multiple functional units organized into hepatic lobules (typically 1–2 mm; Fig. 1). In mammals, we hypothesize that portal tracts arborize within the liver to form the classic intrahepatic biliary/vascular trees, giving rise to lobulation (multiple functional units). Hence, as conceptual models go, the primary architectural differences between medaka and mammalian hepatobiliary systems appears to arise at the tissue/organ level of organization (multiple lobules in mammals vs. single lobule in medaka). These differences can be attributable to the metabolic and structural demands/needs between lower and higher vertebrate hepatobiliary systems (in the context of organ system ontogeny and functional capacity), where mammals see a comparatively higher metabolic demand, typically higher body temperature, higher load bearing on the liver resulting from gravity, and relatively larger mammalian liver mass. It follows that higher vertebrates likely show “lobulation” of the liver in support of greater organ mass and metabolic demand; the iteration of a single hepatic functional unit (e.g. HMS). In more massive mammalian livers the formation of hexagonal lobules, organized by stromal tissue (portal tracts/triads), would impart structural integrity to the organ (due to the physical properties of hexagonal packing; Weaire and Phelan, 1994; Hales, 2001), much needed in an organ of such mass (∼3–5% of body weight in mammals), and one highly perfused with a liquid medium (the liver of mammals can store up 30% of total blood volume). It follows, from an ontological viewpoint, that appearance of VBATs, VATs, VBTs, BTs, and ATs in larger piscine species may be evidence of the emergence of “lobulation” among lower vertebrates such as teleosts. Such a consideration begs more detailed investigation of larger fish species, and reptilian and amphibian livers. Lastly, the hexagonal structural motif appears to be conserved among all vertebrates; found not only in the organization of the canalicular network (intralobular parenchyma), but also in the formation of the classic lobule (arborization of portal tracts).

Reaching a conceptual model of the “functional unit” of the liver has long been sought, and the spatial architecture of the lower vertebrate liver, particularly the biliary system, has long been in question. Differences in extant conceptual models have arisen from how lower and higher vertebrate livers have been viewed at various levels of biological organization (cellular, tissue, organ), in which anatomical plane, and whether in two or three dimensions. Given various ways of viewing the liver and the prior technological constraints (lack of 3D and in vivo tools), understandably, discrepancies have appeared to exist between lower vertebrate and mammalian hepatobiliary conceptual models. Such discrepancies can be illustrated in a study by Akiyoshi and Inoue (2004), who investigated two hundred different teleost species (histological preparations) and described varying liver architecture among them (muralium as cord-like [one cell], tubular [two cells], and solid [>two cells]). Indeed, we observed all three structural forms in our in vivo and 3D investigations, each comprising the hepatic parenchyma, although the muralium of medaka was found to be predominantly two cells thick. Similar observations have been made in developing mammalian liver by McCuskey et al. (2003). Hence, it appears the limits of spatial observation/understanding permitted by single sectional views of the liver have led to these, while accurate descriptions from a 2D viewpoint, discrepancies in interpretation, and thereby discrepancies in a comparative understanding of the 3D architecture of vertebrate liver. Such discrepancies can be understood considering the difficult task of extrapolating 2D information to 3D architectural models. Particularly when 2D observations in single sectional views of the liver have accurately characterized the 3D structure we have elucidated here. It can be said that in one sense, prior observations of lower vertebrate liver architecture that described a tubular conceptual model were correct, in so far as they were communicating the 2D morphology encountered in single sectional views of the liver.


Collectively, these findings elucidate the 3D architecture of the medaka hepatobiliary system and improve our comparative understanding of vertebrate liver structure and function. These findings also present interesting questions regarding the “functional unit” of the vertebrate liver; where the hepatobiliary system in medaka can be, as a conceptual model, considered a single functional unit, akin to the individual lobule. We have also addressed prior discrepancies among conceptual models of vertebrate hepatobiliary system architecture, and the findings presented should help with integration of prior observations of vertebrate liver structure/function into a more cohesive conceptual framework. In summary, in vivo and 3D findings in medaka show: (1) parenchymal architecture is predominantly a two-cell-thick muralium, although tubule-like formations were observed in embryonic livers; (2) the hepatic muralium is organized through a polyhedral (hexagonal) structural motif, revealed in biliary architecture; (3) the intrahepatic biliary system is an interconnected network of canaliculi and bile preductules; (4) the canaliculo-preductular network occupies the majority of the liver corpus (∼95%) uniformly, with equidiameter IHBPs (1–2 μm) observed throughout the liver; (5) larger bile ductules and ducts were observed predominantly at the liver hilus, consequently an arborizing biliary tree was largely absent, seen only in the rudimentary branching of intrahepatic ducts from the hepatic duct; and (6) the livers of these small fish are replete with BPDECs, the putative mammalian correlative of bipotential progenitor/stem cells.


Thanks to Dr. David Miller, Laboratory of Pharmacology and Chemistry, National Institute of Environmental Health Sciences, Research Triangle Park, for providing access to their laser scanning confocal microscopy facility, and special thanks to the Duke University Integrated Toxicology Program. This publication was made possible by a grant from the National Center for Research Resources, a component of the National Institutes of Health. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of NCRR or NIH.