Artificial Liver Device

1. Field of the Invention

This invention relates to pluripotent parenchymal hepatocytes derived from the epiblast of pig blastocyst which have been cultured continuously for a number of years, indicating an infinitely self-renewing cell population. The stem cell characteristics of the hepatocytes indicate that the cells are unique for investigations of liver differentiation and organogenesis and are particularly desireable for use in artificial liver devices.

2. Description of the Relevant Art

Liver transplantation has become widely accepted as an effective treatment for chronic and acute liver disease. One of the major problems associated with the transplantation process, however, as been the need for an effective means for providing temporary support for patients awaiting an available donor organ. Extracorporeal devices for providing heart, kidney and lung support are well-established, however, the development of such devices which are effective for liver support has proven more elusive.

Two basic approaches have emerged as promising. Both utilize a device such as a hollow fiber cartridge containing liver-derived cells which acts as a perfusion apparatus. Primary hepatocytes attached to microcarriers (Rozga et al. 1993. Hepatology. vol. 17(2), pp. 258-265; Rozga et al. 1994. Ann. Surg. vol. 217(5), pp. 502-511) represents one approach, while transformed cells such as those derived from hepatoblastoma cell lines (Sussman et al. 1992. Hepatology. vol. 16(1), pp. 60-65) are another. These approaches have several drawbacks, however. For example, primary cells have a very limited culture capability, therefore there is a requirement for continuous cell preparation. Transformed cells, on the other hand, grow well in culture, but, since they are transformed, they exhibit functional characteristics which may be different from normal cells. Both also suffer from the lack of accessory cells such as epithelial and bile-duct epithelial cells which are important in the function of a normal liver. The need thus exists for the discovery of normal hepatocytes capable of long-term culture which could effectively mimic liver function in an extracorporal device.

The continuous secondary culture of normal parenchymal hepatocytes has previously been unsuccessful, (Williams et al. 1977. In Vitro. vol. 13, pp. 809-817; Guguen-Guillouzo et al. 1983. Exp. Cell Res. vol 143, pp. 47-54; Perraud et al. 1991. Exp. Cell Res. vol 195, pp. 59-65). In vitro hepatocyte studies have largely been limited to the use of primary cultures of hepatocytes established after collagenase perfusion of rat or mouse livers (Williams et al., supra; Berry and Friend. 1969. J. Cell Biol. vol. 43, pp. 506-520). These techniques were also amenable to the establishement of primary hepatocyte cultures in non-rodent species such as the pig (Caperna et al. 1985. J. Anim. Sci. vol. 61, pp. 1576-1586). However, the use of primary cultures is problematic since it is labor intensive and inherently variable. In addition, other epithelial cells, fibroblasts and macrophages were common "contaminants" in primary liver cell culture, and these cells can quickly overgrow or otherwise interfere not only with maintaining cultures but also with experimental manipulations and measurements (Caperna et al., supra; Furukawa et al. 1987. In Vitro. vol. 23, pp. 339-348; Langenbach et al. 1979. Cancer Res. vol 39, pp. 3509-3514). Moreover, normal liver parenchymal cell activities, such as .alpha.-fetoprotein (AFP) and albumin production, and various enzymatic functions were often lost within one to two weeks of establishment in primary culture (Guguen-Guillouzo et al., supra; Langenbach et al., supra).

An alternative to normal parenchymal hepatocyte cell culture has been the establishment of numerous human and rodent hepatocarcinoma-derived cell lines (Richardson et al. 1969. J. Cell Biol. vol. 40, pp. 236-247; Aden et al. 1970. Nature (London). vol. 282, pp. 615-616). Although several of these hepatoma cell lines were minimally deviated and expressed various proteins representative of normal hepatocyte function, they were likely to be abnormal in several aspects, particularly growth control. For example, some of these cell lines were highly tumorigenic when placed in vivo (Richardson, supra; Knowles et al. 1980. Science. vol. 209, pp. 497-499). Also, while hepatocyte growth factor (HGF) is mitogenic for primary hepatocytes, it has been reported to be cytostatic for the hepatoma cell lines HepG2, Hep3B and H35 (Higashi and Shima, 1993). Some human hepatoma cell lines also chronically produced hepatitis proteins (Aden et al., supra; Knowles et al., supra). Thus, hepatoma-derived cell cultures may confuse assessments of normal parenchymal biology in vitro, and in vivo assessments may not be possible. The ability to sustainably culture normal, functional parenchymal hepatocytes in secondary culture is therefore novel and of great value for the routine study of hepatocyte biology.

It has been hypothesized that the liver contains stem cells which support its general long term function and its regenerative ability (Sigal et al. 1992. M. J. Physiol. vol. 263, p. G139; Fausto and Meade. 1989. Lab. Invest. vol. 60, pp. 4-13). Stem cells have not been identified, although embryonic phenotypic gene expression, e.g. AFP, was found in some proliferating liver epithelia following the administration of hepatotoxic agents (Fausto et al. Purification and culture of oval cells from rat liver. In Pretlow and Pretlow, eds. Cell Separation: Methods and Selected Applications, Vol. 4. Orlando: Academic Press, 45-77; Evarts et al. 1987. Carcinogenesis. vol. 8, pp. 1737-1740). Liver "oval" cells and intrahepatic bile duct cells were also proposed as candidate liver stem cells or facultative stem cells (Sigal et al., supra; Fausto and Meade, supra). A consensus of data indicated that liver stem cells would express AFP and albumin and would be capable of differentiating into bile duct cells and parenchymal hepatocytes (Shiofiri, N. 1981. J. Embryol. Exp. Morphol. vol. 79, pp. 149-152; Evarts et al. 1989. Cancer Res. vol. 49, pp. 1541-1547; Germain et al. 1988. Cancer Res. vol. 48, pp. 4909-4918).

Thus although attempts have been made to establish normal hepatocyte cell lines which are capable of culturing over a long period of time (i.e., years) without losing characteristic activity, accomplishment of this goal has heretofore been unsuccessful.

We have established a novel cell line of normal pluripotent parenchymal hepatocytes derived from a primary culture of pig epiblast cells which represents the first example of the sustainable culture of normal parenchymal hepatocytes and which is useful in an artificial liver device.

Accordingly, it is an object of the invention to provide an artificial liver device comprising the novel hepatocyte cell line.

Other objects and advantages of the invention will become readily apparent from the following description.

The file of the patent contains at least one drawing executed in color. Copies of this patent with color drawing(s) will be provided by the Patent and Trademark Office upon request and payment of the necessary fee.

FIG. 1 shows the top view of an artificial liver device which utilizes plates as the apparatus which provides support for the culture of hepatocyte cells.

FIGS. 2 and 2A show the cross-section of the artificial liver device of FIG. 1.

FIG. 3 shows the side view of the artificial liver device of FIG. 1 having an internal semipermeable membrane compartment.

FIG. 4 is a photomicrograph of PICM-19 clone 3BT cells grown in an artificial liver device.

FIG. 5 shows the activity of 7-ethoxyresorufin-O-deethylase (P-450 1A1) in T-25 flasks and in an artificial liver device containing PICM-19 clone 3BT cells.

FIG. 6 shows urea secretion of PICM-19 clone 3BT cells in an artificial liver device.

The novel cell line was prepared as described by Talbot et al. (1993. In Vitro. vol. 29A, pp. 543-554; patent application Ser. No. 08/133,589, both herein incorporated by reference). It was derived from pig epiblast cells that had been separated from other early embryonic cells and cultured in vitro. The epiblast cell cultures were pluripotent and gave rise to several morphologically distinct cell cultures resembling neural crest cells, muscle cells, fibroblasts, endoderm, polarized epithelium, secretory epithelium and macrophages.

A three-step protocol was used to isolate the epiblast from trophectoderm and primitive endoderm of the blastocyst before culturing. At least two possible benefits were derived from culturing the epiblast free of the attendant heterologous cells of the blastocyst. First, the culture of the intact pig blastocyst results in the attachment and growth of the trophoblast and primitive endoderm, but under these conditions, unlike the mouse blastocyst, the pig epiblast usually does not grow but instead degenerates. In addition, it has been reported (Mummery et al. 1991. Differentiation. vol. 46, pp. 51-60) that visceral endoderm can be a potent inducer of mesodermal differentiation for mouse embryonal carcinoma (EC) cells. Thus, the prevention of interaction between the pig epiblast cells and trophectodermal or endodermal cells is considered important in maintaining the epiblast cells in culture. A second reason for culturing an isolated epiblast is that the evaluation of the culture conditions is simplified without the presence of the other blastocyst cells.

Blastocysts collected at 7 to 8 days postestrus were immunodissected to obtain the inner cell mass (ICM) and destroy trophectodermal cells. The pig ICM (PICM) was composed of an inner core currounded by multiple layers of outer cells. This arrangement is also found in the mouse blastocyst where two rounds of immunodissection have been used to isolate and culture the primitive ectoderm core of the ICM (Hogen et al. 1977. Nature. vol. 265, pp. 626-629; Pederson et al. 1977. Nature. vol. 270, pp. 435-437). The ICM was cultured for 2 to 3 days on STO feeder cells. The epiblast was then physically dissected free of associated primitive endoderm.

Immunodissection removed most or all of the trophectoderm but little or none of another layer comprising an outer layer of the ICM. This layer completely surrounded the pig epiblast immediately after the immunodissection. By mouth-pipetting with a Pasteur pipette slightly smaller in diameter than the ICM, this outer layer could be stripped away from the epiblast core except at one strong attachment point. After dissection from the epiblast and culture on STO feeder cells, it grew as colonies of polymorphic epithelial-like cells, which polarized to form domes and vesicles. Based on these observations, this outer layer surrounding the epiblast appears to be primitive endoderm in its entirety, necessitating its complete removal from the preparation.

To isolate epiblasts free of endodermal cells, immunodissected ICMs were cultured on STO feeder cells in 10% DMEM(+) for 2 to 3 days. The outer primitive endodermlike cell layer grew radially outward from the centrally located epiblast and surrounded it as a well-organized single-cell layer. As with the dissected cells discussed above, the radially spreading endoderm cells produced domes and vesicles. The final isolation of the pig epiblast was achieved by a simple physical dissection of the cultured ICM. This was possible because the associated outer layer was only loosely attached after the 2 to 3 days in culture. The epiblast consisted of a well-organized homogenous ball of 50 to 200 cells. With this combination of steps, approximately 8 of 10 blastocysts yielded an epiblast free of heterologous cells.

Primary culture of the isolated pig epiblasts on STO feeder cells produced homogeneous colonies whose cells resembled mouse embryonic stem (ES) cells (small round cells with large nuclei and nucleoli). Pig epiblast cells were approximately the same size as the cells of the D3 mouse ES cell line. Like mouse ES cells, the cells grew as tight clusters either on top of or nested in between the STO feeder cells. Also, the colonies had a rich extracellular matrix which glistened at the colony's edge under phase contrast microscopy.

After the initial attachment and spreading of the epiblast cells on STO feeder cells (24 to 36 hours), the colonies progressed in one of several ways. Typical primary colony progression was represented by two separate experiments consisting of six and eight epiblasts. Approximately 10% (1 of 14) of the colonies degenerated into senescent cells which became enlarged, irregularly shaped and nondividing. This most often occurred when the epiblast was small and consisted of only 10 to 20 cells. Fifty percent (7 of 14) of the epiblast colonies began a rapid growth pattern in which the cells became smaller and were arranged in mushrooming rosettes reminiscent of neural tube formation. In roughly another 20% (3 of 14), what seemed to be a clonal expansion of an epithelial cell type began within the colonies. These differentiated cells were flat, cuboidal and translucent, with large nuclei but small nucleoli. They rapidly produced a cell sheet which pushed the STO feeder cells and the remaining epiblast cells out of their path, creating an obvious heterologous cell-to-cell boundary at the colony's edge. The remainder (3 of 14 or about 20%) of the epiblast colonies grew out as a complex of various cell types. Portions of these colonies had cells resembling the two described above (neural rosettes and epithelial-like) along with other unique cell types. Two of these new cell types had prominent nuclei and nucleoli, but each grew in a distinct manner within the colony. One grew as an epithelial-like cell sheet and the other as tight swirled and multilayered groups of cells. These colony morphologies became increasingly distinct in their individual character if they were allowed to grow for 1 to 2 weeks or even longer before the first passage of the primary culture was attempted. The vesicle and dome-forming polarized epithelium, which was characteristic of the ICM endoderm, sometimes appeared as a component of the complex colony type. These cells may have differentiated from the epiblast. However, they also may have derived from endoderm cells that were accidentally carried over during the physical dissection of the epiblast from the cultured ICM. Only epiblast colonies composed of either the neuroectodermal-like rosettes of cells or the complex of different cells were passaged for secondary culture.

Secondary cultures of epiblast colonies resulted from subculturing the primary pig epiblast colonies. Numerous primary colonies were lost at passage 1 because of the delicate nature of epiblast cells. The epiblast colonies composed of small cells which grew as mushrooming rosettes usually gave rise to neuronlike cells upon continuous passage on STO feeder cells. Of seven such colonies cultured, five developed into mass cultures of neuronlike cells. These cell cultures were designated PICM-8, 9, 10, 11 and 20 (Table 1). The other two colonies grew very slowly after passage and soon became senescent. Several of the neuronlike epiblast colonies also gave rise to terminally differentiated pigment-producing cells if the cells had clumped during early passage. For example, one of the five colonies, PICM-9, was passaged almost entirely as grossly clumped groups of cells at passage 2. The majority of PICM-9 clumps failed to grow and instead differentiated into pigmented cells. In addition, after 2 to 3 weeks in maintenance culture, at least four other differentiated phenotypes developed from the PICM-9 subculture (Table 1). Among these were small fibroblastic cells, neuronlike cells, and one colony which formed a vesicle, the inner surface of which was covered with rapidly beating cilia. Unfortunately, these ciliated cells were not successfully passaged. Also, colonies of multinucleated contracting muscle cells developed and were cloned. The PICM-9 muscle cells were small spindle-shaped fibroblasts which grew very rapidly. They spontaneously formed

multinucleated contracting muscle cells when seeded at a low concentration on STO feeder cells.

Six epiblast colonies, which were composed of a complex of cell types, gave rise to several distinct cell cultures after passaging and cloning (Table 1). Neuronlike cells (PICM-12) and fibroblastic cells (PICM-6 and 15) were cultured. It was also possible to grow homogeneous cultures of the two cell types with prominent nuclei and nucleoli. The first of these, represented by PICM-16, were very similar to mouse ES cells in appearance. PICM-16 grew on top of the STO feeder layer as numerous individual clusters of cells after each passage. After 48 to 72 hours post passage, the PICM-16 cells in the centers of growing colonies flattened into polygonal epithelial-like cells which usually grew over the STO feeder cells. The cells at the periphery of each expanding colony retained their former ES-like appearance. Some dome formation occasionally occurred in cultures of PICM-16 which had reached confluency. Despite treatment with retinoic acid or DMSO, these epithelial-like cultures (PICM-16 and 17) have not shown any other specialized differentiation potential besides formation of polarized epithelium.

The second cell type exhibiting large nuclei and nucleoli, represented by PICM-13, 18 and 19, also grew as numerous individual clusters of cells after each passage (Table 1). These cells grew on or between the feeder cells and formed tubular, glandlike structures if maintained in STO co-culture for a week or more. The glandlike structures were primarily two-dimensional but often quite extensive and intricate in pattern. The cells assumed a pseudo-columnar orientation as the glandlike formations matured and what appeared to be secretory matter collected at the luminal side of the cells.

The PICM-13, 18 and 19 cultures also produced a second phenotype (Table 1). After long-term maintenance culture of a month or more, adipogenic processes began in the cells. The accumulated lipid vacuoles were intensely stained by treatment with oil red O stain. The vacuole formation and oil red O staining were not seen in those cells that had undergone differentiation into the glandlike tubules.

Another cell type appeared, first as round, refractile cells that spewed forth from a central colony of the tubule-adipocyte forming precursor cells. Subsequent work has shown, however, that this type occurs independently of the tubule-adipocyte-forming cells, and the concurrence of the two cell types was probably coincidental. The round cells were only loosely attached to the feeder cells by small but obvious pseudopods. Over several days to a week in culture, approximately 50% of the cells became firmly attached to the feeder cells and adopted an amoeboid appearance and behavior. In both forms, the cells were heavily vacuolated. Grossly, the round cells were similar to maturing adipocytes, and some of the cells appeared to have eccentric nuclei as a result of vacuole coalescence. However, these cells did not show a tendency to float and were negative for staining with oil red O. Wright-stained cytopsin preparations also clearly showed the gross vacuolation, and the cell's profile was most similar to an adipocyte or a plasmacyte.

The various PICM cell cultures, with the exception of PICM-16 and PICM-19, could not be propagated under clonal conditions (i.e. from 5 to 10 cells per 1.9-cm.sup.2 well on STO feeder cells) more than once and thus appear to have a finite doubling potential. With continuous passage on STO co-culture, all of the PICM cell cultures (except PICM-16 and PICM-19) showed a decline in proliferative ability characteristic of in vitro senescence. At their senescent endpoints (approximately 40 to 50 population doublings), the various PICM cell cultures displayed cytoplasmic vacuolations, perinuclear inclusion bodies, incomplete karyokinesis, and a two- to three-fold increase in overall and nuclear size.

PICM-16 has shown no obvious senescent growth crisis or senescent morphologic changes over the course of greater than 80 passages at alternating 1:5 and 1:10 split ratios. In addition, the PICM-16 cell culture has undergone two successive single-cell clonings. Each clonal PICM-16 cell culture has maintained the same proliferation rate and normal karyotype of the parental culture.

Of the three tubule-forming-adipogenic cell cultures established, only PICM-19 has been continuously passaged. This cell culture grows very slowly, and therefore its longevity in STO co-culture remains undefined. However, PICM-19 has now been continuously passaged for over 2 years to greater than 100 cumulative population doublings, with no obvious signs of senescence. While any appropriate feeder cell population may be used successfully for culturing, STO feeder cells are preferred.

Some of the PICM cell cultures were feeder cell-dependent. After only 3 to 4 passages at 1:5 split ratios without STO feeder cells, the neuronlike cells (PICM-8, 9, 10, 11, 12 and 20) flattened, greatly enlarged, collected abundant perinuclear inclusion bodies, stopped dividing and died. The tubule-adipocyte-forming epithelial cells (PICM-13, 18 and 19) were similarly affected by passage off STO feeder cells. In contrast, the fibroblastic cell cultures (PICM-6 and 15) grew to mass cultures without feeder cells as cell strains exhibiting "normal" Hayflick limits of approximately 40 population doublings. The dome-forming epithelial-like cells (PICM-16) have been maintained without feeder cells for greater than 10 passages. However, the feeder-independent PICM-16 cultures had to be passaged at low split ratios (1:2 or 1:3) to survive and proliferate.

Added human recombinant leukemia inhibitory factor (LIF) at 10 or 20 ng/ml or purified erythroid differentiation factor (EDF) at 50 to 100 ng/ml had no apparent effect on the morphology, maintenance of morphology or growth of the primary epiblast colony when they were plated on or off STO feeder cells. In contrast, LIF at the same concentration maintained the morphology and growth potential of the D3 mouse ES cells when passaged in the absence of STO cells.

Chromosome counts were done on 10 to 20 mitotic spreads from cell cultures maintained for at least 10 passages. All the epiblast-derived cell cultures had the normal porcine complement of 38 chromosomes (35 autosomes and 2 sex chromosomes) with no visible abnormalities, additions or losses. Karyotypes were constructed of each cell type and all were found to be male (Table 1).

The inner core of the pig ICM is the epiblast because the cells derived from it were pluripotent in culture. The outer layers seem to be primitive endoderm because in culture it only gave rise to vesicle and dome-forming cells. A complete layer of cells surrounding the epiblast is incongruent with the view that the endodermal cells in vivo grow over only the side of the epiblast that faces the blastocoelic cavity (Mossman, H. W., Vertebrate fetal membranes. New Brunswick, N.J.: Rutgers University Press, 1987: 74-78; Solter and Knowles. 1975. PNAS. vol. 72, pp. 5099-5102). Inasmuch as the trophoblast cell layer surrounding the ICM is only one-cell thick, it is probable that no trophoblast cells survive the immunosurgery. This was found to be the case by Solter and Knowles, supra, in their original work with mouse blastocysts. The bulk of the endoderm multilayer can be physically stripped off the pig epiblast core except at one point of strong attachment. This arrangement suggests that the endoderm cells migrated out from the epiblast core as they surrounded it and differentiated.

The pig ICM's outer endoderm layer seemed to be comprised of at least two layers. The endoderm layer, which could be physically stripped off the epiblast core as discussed above, was the outermost layer(s) of endoderm. Left behind was a single cell layer which was closely adherent to the epiblast core. When the physically stripped epiblast core was cultured, this innermost endoderm layer also gave rise to a vesicle and dome-forming monolayer. The removal of this innermost layer of endoderm was facilitated by first culturing the intact ICM on STO feeder cells. During the 48-hour culture period, the union between the epiblast core and the innermost endoderm layer was uniformly weakened, thus enabling the physical dissociation necessary for isolating the pure epiblast.

Although significant differences in early embryonic development exist between mice and ungulates (Mossman, supra), under similar culture conditions the morphology of their embryonic stem (ES) cells should be very similar, given the following: a) the ES cells are derived from the ICM at a point before embryonic differentiation; b) they share the same functional attribute of totipotency, and structure often follows function; and c) hamster ES cells and human EC cells are very similar in morphology to mouse ES cells (Andrews et al., I. Cell lines from human germ cell tumours. In: Robertson, E. H., ed. Teratocarcinomas and embryonic stem cells: a practical approach. Oxford: IRL Press, 1987: 207-248; and Doetschman et al. 1985. J. Embryol. Exp. Morphol. vol. 87, 27-45). This "ESD-like" morphology was observed in the pig's epiblast cells during the first 48 to 72 hours in primary culture. Maintaining this morphology was difficult, however, because, as with mouse epiblasts (Hogan et al., supra), the pig epiblast cells quickly differentiated.

Mouse ES cells often suffer extensive cell death or differentiation if passaged in the absence of feeder cells (Robertson, E. Embryo-derived stem cell lines. In: Robertson, E., ed. Teratocarcinomas and embryonic stem cells: a practical approach. Oxford: IRL Press, 1987: 71-112). The cytokine LIF, also known as differentiation inhibitory activity (DIA), is the factor secreted from STO feeder cells which maintains mouse ES cells in an undifferentiated state (Rathjen et al. 1990. Cell. vol. 62, pp. 1105-1114; Smith and Hooper. 1983. Exp. Cell. Res. vol. 145, pp. 458-462; and Williams et al. 1988. Nature. vol. 336, pp. 684-687). EDF (Acrivin-A) has also been found to suppress the differentiation of mouse EC cells (van den Eijnden-van Raaij et al. 1991. A. Mech. Dev. vol. 33, pp. 157-166). Neither of these cytokines had a differentiation-inhibiting effect on the pig epiblast cultures. This was the case at both the primary colony stage and with the various subcultures established after passage. Furthermore, they did not prevent rapid senescence of the PICM cell cultures when they were removed from culture with STO cells. Thus, under the culture conditions tested, the pig epiblast cells did not respond to LIF in the same manner as mouse epiblast cells. This apparent lack of anti-differentiation or proliferation response to added recombinant LIF has also been observed in certain human EC cell lines (Pera et al. 1989. Differentiation. vol. 42, pp. 10-23; P. Andrews, personal communication, 1992) and in the culture of pig and sheep ICM-derived cells (Piedrahita et al. 1990. Theriogenology. vol. 34, pp. 879-901). Although there is no direct evidence for the absence of the LIF receptor in the pig epiblast cells, the inability to maintain these cells in an undifferentiated state may depend on up-regulating the expression of the LIF receptor. If the LIF receptor is present, its function may depend on interaction with homologous pig LIF or a synergistic interaction with other growth factors. EDF had a similar lack of effect, but this was not unexpected because it also failed to maintain the undifferentiated phenotype of the D3 mouse ES cell line when grown without STO feeder cells.

The lack of response to LIF, either STO-generated or human recombinant, might also result from the inability to effectively disperse the cells of the primary pig epiblast colony. Even when done with very gentle methods, such as lipid-buffered trypsin-EDTA, the cells are easily killed. A critical requirement reported for establishing mouse ES cell lines is the early dispersion and subculturing of the primary epiblast growths (Robertson, supra). This presumably short-circuits cell-to-cell interactions which lead to differentiation induction. The initial dispersions of the pig epiblast colonies usually resulted in a significant amount of cell clumping, and, analogous to mouse ES and EC cell lines, this clumping may have stimulated differentiation in spite of the presence of LIF.

The pig epiblast cells of the primary and secondary cultures spontaneously differentiated into various cell types, which seem to be mesodermal, neuroectodermal, ectodermal and endodermal in character. As with some mouse EC cell lines, the final differentiation events of the pig epiblast cultures occurred after several weeks in maintenance culture on STO feeder cells (Nicolas et al. 1976. Cancer Res. vol. 36, pp. 4224-4231). It has long been recognized that the differentiation of various cells is induced by paracrine regulatory interactions or cell-to-cell interactions with mesenchymal cells (Balinsky, B. I., An introduction to embryology. Philadelphia: W. B. Saunders, 1970: 566-578; Drew, A. H. 1923. Br. J. Exp. Pathol. vol. 4, pp. 46-52; Gorbstein, C. 1957. Exp. Cell. Res. vol. 13, pp. 575-587). The differentiation of the pig epiblast cells may be driven by the STO feeder cells elaborating combinations of growth factors which do not predispose to prevent differentiation. For example, it has been shown that cultured fibroblasts secrete hepatocyte growth factor to the extent that it induces epithelial morphogenesis in co-cultured MDCK cells (Montesano et al. 1991. Cell. vol. 67, pp. 901-908). Also, among the known pleiotropic effects of LIF is its ability to promote the differentiation of neural crest cells and hemopoietic cells, to stimulate myoblast proliferation and to activate hepatocytes (Hilton and Gough. 1991. J. Cell. Biochem. vol. 46, pp. 21-26; Mummery, supra). Given this, it is very possible that co-culture on STO cells contributed to the differentiation observed in the pig epiblast cultures.

The pig epiblast-derived cell cultures, which were non-epithelial in morphology, had a finite life span in STO co-culture of approximately 40 cumulative population doublings (Rohme, D. 1981. PNAS. vol. 78, pp. 5009-5013). In contrast, PICM-16 showed no overt change in growth characteristics or morphology after approximately 300 population doublings in STO co-culture. Also, no gross changes in PICM-16's karyotype has occurred after two successive single cell clonings. Thus, PICM-16 is similar to mouse ES cell lines in being immortal and in retaining a normal male karyotype (Robertson and Bradley. Production of permanent cell lines from early embryos and their use in studying developmental problems. In: Rossant and Pedersen, eds. Experimental approaches to mammalian embryonic development. New York: Cambridge University Press, 1986: 475-508). The indefinite growth potential of PICM-16 may be the result of transforming genetic events, and fine band analysis must be done in an attempt to find evidence of any karyotypic abnormalities. Although the propagation of PICM-16 was not rigorously quantitated, a senescent crisis followed by resurgence of a transformed cell population, as is typical in mouse embryonic fibroblast cultures, was not observed in the PICM-16 culture.

Senescence occurred in some of the pig epiblast cultures despite the presence of 2-mercaptoethanol in the medium. 2-Mercaptoethanol has been shown to be essential to the cultivation of some cell types and, in general, is thought to be necessary for the isolation and maintenance of mouse ES cells (Robertson, supra; Tookey, J. I. 1975. PNAS. vol. 72, pp. 73-77). In addition, mouse embryonic fibroblasts have recently been shown to be protected against senescent crisis in culture if the medium contains inhibitors of serum amine oxidases or 2-mercaptoethanol (Parchment et al. 1990. PNAS. vol 87, pp. 4340-4344). PICM-16 could be an analogous event to that seen with mouse cells, but in this study, examining epiblast culture, it seems to represent the exception rather than the rule.

The differentiation of the pig epiblast cell cultures was brought about by cell aggregation combined with long-term culture. Those cell cultures (PICM-8, 9, 10, 11, 12 and 21), which start out by forming neuronal-like rosettes of cells, may be committed as neuroectodermal stem cells. As such, they would be expected to be able to form neurons and pigmented cells, which in fact, they all do, based on morphology. Also, preliminary tests indicate that PICM-8 displays a mitogenic response to LIF and EDF, both of which are known to support the differentiation and maintenance of neuronal cells (Murphy et al. 1991. PNAS. vol. 88, pp. 3498-3501; Schubert et al. 1990. Nature. vol. 344, pp. 868-870). The neuroectodermal colony, PICM-9, also gave rise to fibroblasts, contracting muscle, and ciliated epithelium after pronounced clumping of the cells and 6 weeks in maintenance culture. However, this was at passage 2, and it is likely that this degree of pluripotency is lost with continued passage of the cells. The cell cultures that differentiated into glandlike structures or adipocytelike cells (PICM-13, 18 and 19) had a distinctive morphology which is typical of parenchymal hepatocytes, thus raising the possibility that these cells represented hepatic stem cells.

The cell line, PICM-16, which is epithelial in character and most closely resembles mouse ES cells, has occasionally formed domes. Although mouse ES cell lines are not known to form domes in culture, a human EC cell line, NEC 14, apparently shares this characteristic with PICM-16 (Haseyawa et al. 1991. Differentiation. vol. 47, pp. 107-117). PICM-16 is morphologically similar to the ICM-derived pig cells isolated by other groups. Only limited information has been published on the in vitro or in vivo pluripotency of these other cell cultures (Evans and Kaufman. 1981. Nature. vol. 292, pp. 154-156; Evans et al. 1990. Theriogenology. vol. 33, pp. 125-128; Notarianni et al. 1990. J. Reprod. Fertil. Suppl. vol. 41, pp. 51-56; Piedrahita, supra).

The derivation of a round amoeboid cell type is notable. Results obtained with monoclonal antibodies to pig cell surface antigens indicate that the round cells may be of macrophage/granulocyte lineage.

The PICM-19 cells have utility as an in vitro model for fetal hepatocyte functions (e.g., hemopoiesis) and are also useful in experimentation on gene therapy and for artificial organ construction. The cell line also has significant value as a research reagent, that is, a sustainable fetal liver cell culture. Moreover, it would serve as a valuable source of novel growth factors and pharmaceuticals and for the production of viruses or vaccines (e.g., hepatitis viruses), as well as for the study of liver parasites or of parasites having a stage of development in the liver (e.g., malarial organisms), for in vitro toxicity and metabolism testing of drugs and industrial compounds, for gene therapy experimentation (since the liver is the largest vascular organ of the body), for the construction of artificial translatable livers, and for liver mutagenesis and carcinogenesis studies. In addition, the derivation of the PICM hepatocyte cultures from the pig epiblast demonstrates the possibility of deriving specialized cell types from any mammalian epiblast.

The cell line ARS-PICM-19 was deposited on Jan. 4, 1994 under the Budapest Treaty in the American Type Culture Collection, 12301 Parklawn Drive, Rockville, Md. 20852, as ATCC No. CRL 11519.

Attempts have been made to utilize rat hepatocytes for the construction of bioartificial livers (Nyberg et al. 1992. Cytotechnology. vol. 10, pp. 205-215; Li et al. 1993. In Vitro Cell Dev. Biol. vol. 29A, pp. 249-254). Since pig liver cells are physiologically more closely related to human cells than mouse or rat, the cells described herein have obvious advantages over cells of such origin as well as the obvious advantage of normal cells over tumor cells.

The novel cells (normal embryonic hepatocytes) are particularly useful for the preparation of an artificial liver device (ALD). The cells may be sphroid cultures, attached to solid supports such as glass beads or encapsulated by encapsulation agents such as alginates. A number of apparatuses are known in the art which are useful for the purpose, and the particular apparatus selected is not critical to the invention so long as the cells may be cultured in the apparatus and fluid such as media, blood or plasma may effectively circulate through the apparatus. Examples of useful apparatuses are hollow fiber cartridges, columns and plates. Various devices have been described in the literature by Kelly, J. H., U.S. Pat. No. 5,290,684, Jauregui, H. O., U.S. Pat. No. 5,043,260, Viles et al., U.S. Pat. No. 4,853,324, Ku et al. (1983. Biotechnol. Bioeng. vol. 23, pp. 79-95), Wolf et al. (1979. Int. J. of Artificial Organs. vol. 1, pp. 45-51), Ehrlich et al. (1978. In Vitro. vol. 14, pp. 443-450) and Knazek, R. H. (1974. Feder. Proc. vol. 33, pp. 1978-1981).

The novel ALD is essentially a culture vessel which contains feeder cells in addition to the embryonic hepatocytes. Adult pig hepatocytes may also be included in the cell mixture.

As discussed above, the embryonic hepatocytes display unique in vitro differentiation properties. Spontaneous differention into hepatocytes or bile duct epithelium is observed in co-culture on feeder cells. Cells from a culture having a preponderance of bile duct epithelia may be selected and cloned for use in the ALD. For example, a clone designated 3BT was derived from bile duct differentiation cells in the PICM-19 cell culture that appeared to lose growth control instead of terminally differentiating. The clone 3BT cells are high in .gamma.-glutamyl transferase (GGT) activity, which is characteristic of bile duct epithelium.

The purpose of the feeder cells is to support the growth and differentiation of the adult pig hepatocytes and the parenchymal hepatocytes. They may also conveniently be genetically engineered to perform additional functions such as production of specific growth factors or anti-serum complement factors. While any effective feeder cells may be utilized, useful feeder cells are mouse embryonic fibroblast feeder cells, such as NIH/3T3 cells, Balb/3T3 cells, CH3 10T.sup.1/2 cells or STO cells that have been mitotically inactivated by exposure to mitomycin C or radiation (Talbot et al., supra). Preferred are STO feeder cells.

Adult pig hepatocytes may be prepared by harvesting from pigs weighing about 40 kg by EGTA/collagenase perfusion of the liver followed by differential centrifugation to partially purify the hepatocytes (Caperna et al., supra). The pig hepatocytes are co-cultured with the feeder cells in a 1:1 mixture of DMEM and M199 containing 10% fetal bovine serum (10% DMEM-199) for the first week of culture. LCF medium (Hyclone Laboratories, Inc., Logan, Utah) is subsequently used to maintain the hepatocytes and to prevent the overgrowth of the culture by mesenchymal cells such as macrophages and fibroblastic cells (Caperna and Talbot. 1996. In preparation). Adult pig hepatocytes cultured under these conditions differentiate into morphologically normal, functional hepatocytes in about two weeks and may be maintained in this state for about 1 to 2 mos. Normal morphology of the cells was assessed by phase-contrast and electron microscopy. Inducible P450 enzymatic activity is high in the adult pig hepatocyte culture.

The combination of adult pig hepatocytes and embryonic parenchymal hepatocyte cells having bile duct epithelia provides the fullest spectrum of liver function possible in an in vitro device because together they represent all the parenchymal tissue present in the liver. The embryonic hepatocytes have been shown to spontaneously differentiate into liver-like cells and structures and to secrete serum proteins, i.e. the cells display in vitro ultrastructure, enzymatic and transport characteristics similar to those of parenchymal hepatocytes and bile duct epithelium. The cells contain large oval nuclei, numerous oval to elongate mitochondria with flat cristae, extensive rough and smooth endoplasmic reticulum, Golgi complexes, lipid vacuoles and glycogen granules. Biliary canaliculi with intraluminal projecting microvilli are delimited by the junctional apparatuses between adjacent cells. The cells also rapidly transported fluorescein into their biliary canaliculi from the extracellular environment. Those cells which differentiate into multicellular ductal structures have high GGT activity at their apical surfaces as shown by histochemical staining, with total GGT activity at least 19 times higher than that found in porcine hepatocytes. Metyrapone induced cytochrome P-450 content of the cells is at least one-fourth that found in porcine hepatocytes, while P-450 activity induced by 7-ethoxycoumarin is nearly equivalent to that of primary cultures of pig hepatocytes.

Cultures of fetal pig liver macrophages may also be added to the mixture. The macrophages are co-cultured on feeder cells in 10% DMEM-199 from explant cultures of liver tissue (Talbot et al. 1996. Submitted). They are easily added in pure form and may be added to the mixture if macrophage accessory function is desired. Such functions include removal of endotoxin and aged glycoproteins and production cytokines and prostaglandins. Nonparenchymal cells have also been shown to regulate nromal hepatocyte function in vitro (Billiar and Curran. 1990. J. Parenteral and Enteral Nutr. vol. 14, pp. 175S-180S; Caperna, unpublished observation).

While any effective apparatus may be utilized, for purposes of illustration, a device is described herein which utilizes a plate of two square glass or plastic sheets, about 12" on a side, separated by an about 0.5--about 1 mm space and sealed around the edge except for an intake manifold and outflow manifold for the exchange of media, blood or plasma through the internal space of the plate (FIG. 1). Adult pig hepatocytes and embryonic parenchymal hepatocytes are cultured on the internal surfaces of the plate (FIG. 2) with a total cell culture surface area of about 2 sq ft (about 1800 cm.sup.2). The plate is maintained at 37.degree. C. in a standard incubator for maintenance of the culture. Culture medium is pumped from a medium reservoir through the internal space of the plate in order tonourish the cells. Gas permeable tubing allows gas exchange between the incubator atmosphere and the medium reservoir, or a 2% CO.sub.2 /98% air mixture may be bubbled through the medium reservoir to maintain proper pH conditions. Sampling ports and meter interfaces may be incorporated into the medium loop as needed (FIG. 1). When in use for providing liver support, additional plates (depending on the severity of a patient's condition) may be used in parallel with the flow of a patient's heparin-treated plasma being distributed equally between the plates.

Plasma effluent exiting the ALD may be subjected to dialysis for protein-bound toxins using an albumin-containing dialysate (Stange et al. 1993. Artificial Organs. vol. 17, pp. 809-813). In addition, it should also be put through an anti-pig serum antibody affinity column to remove the pig hepatocyte-derived proteins from a patient's plasma. Any xenobiotics coupled to the pig hepatocyte-derived proteins would thus be removed along with the pig proteins. Plasma may then be returned to a patient after heparin antagonists had been added.

Alternatively the cell culture plate may be constructed with a semi-permeable membrane such as a dialysis membrane within the internal space of the cell culture plate (FIG. 3). This arrangement is advantageous in that it would prevent pig hepatocyte-derived proteins from entering the patient's plasma and it would allow for the exchange of protein-bound toxins in the patient's plasma with the pig hepatocyte-derived proteins in the plate's medium compartment which would be acting as a dialyzing fluid. The total surface are of 10 plates would be approximately 18,000 cm.sup.2 which is equivalent to a modern kidney dialysis device's surface area. Internal dialysis membranes would also serve as a preventative measure towards the escape of foreign cells from the ALD into the patient.

It is estimated that a plate as described would accommodate the culture of about one billion adult pig hepatocytes or embryonic hepatocytes. The health and differentiation status of the cultured cell within the plate may be assessed by inspection with phase-contrast microscopy. The ALD is useful as a bridging device to temporally support a patient undergoing acute liver failure until liver transplantation can be performed. Liver function assistance is also required in many shock trauma and acute infection situations. In addition, liver destruction resulting from various causes is potentially reversible, and the ALD may be used to support the patient until his or her own liver could recover function (Cattral and Levy. 1994. New Eng. J. Med. vol. 331, pp. 268-269).

The following examples are intended only to further illustrate the invention and are not intended to limit the scope of the invention as defined by the claims. PICM-19 provides an examplary representative of a number of hepatocyte cell lines independently prepared from pig blastocysts and having the properties of the novel continuous cell line described herein.

Blastocysts were collected 7 to 8 days postestrus by flushing each uterine horn with 20 ml of a modification of the Tyrodes-HEPES (TLH) mudium described by Hagen et al. (1991. J. Anim. Sci. vol. 69, pp. 1147-1150). The isolated hatched blastocysts were incubated in TSH in air at 38.degree. C. until dissection.

A double dissection procedure was carried out. First, immunodissection was performed according to Solter and Knowles, supra. Blastocysts were incubated at 37.degree. C. for 1 hour in rabbit anti-sheep splenocyte serum, prepared as described by Butler (In: Evans and Hollaender, eds. Genetic engineering of animals: an agricultural perspective. New York: Plenum Press, 1986: 175-185), and diluted 1:8 in DMEM. After washing 3 times in 10% DMEM, the blastocysts were incubated at 37.degree. C. for about 30 minutes in guinea pig serum (GIBCO, Gaithersburg, Md.) diluted 1:4 in DMEM. Lysed cells were removed by mouth pipette aspirations, and the ICMs were cultured on feeder layers of STO cells for 2 to 3 days.

A physical second dissection was done using 25-gauge hypodermic needles as dissection knives. Isolated epiblasts were co-cultured on STO feeder cells in 10% DMEM(+) ›10% DMEM with 2-mercaptoethanol (0.1 mM), nucleoside supplementation (0.01 mM guanosine, 0.03 mM adenosine, 0.01 mM cytidine, 0.01 mM uridine, and 0.01 mM thymidine), nonessential amino acids supplementation (2 times, GIBCO), and glutamine supplementation (1 mM, GIBCO) after the method of Robertson, supra)!.

Following dissection, all cells were grown on tissue culture plasticware (Nunc and Falcon). Most cell culture reagents, including fetal bovine serum (FBS), media, phosphate buffered saline (PBS), and trypsin-EDTA (0.025% trypsin, 0.43 mM EDTA) were obtained from GIBCO. STO cells, a transformed mouse fibroblast line, were obtained from the American Type Culture Collection, Rockville, Md., and were propagated in Dulbecco's modified Eagle's medium (DMEM) with high glucose, 10% FBS, and antibiotics (10% DMEM). STO feeder cells were prepared as described by Robertson, supra.

Primary epiblast colonies were passaged by two methods. One method was to physically detach the colonies by a mouth-controlled pipette and wash them through two successive drops of lipid-trypsin. The lipid-trypsin solution consisted of soybean lipid (2 mg/ml) made isotonic by the addition of 10 times Dulbecco's PBS (Ca.sup.2+ - and Mg.sup.2+ -free) and containing 0.01% wt/vol crystallized trypsin. The acidity of the lipid solution was neutralized by the addition of NaOH. After the brief washings, the colony was placed into a final small drop of lipid-trypsin (10 to 20 .mu.l) and incubated for 3 to 5 minutes at 37.degree. C. Trypsin activity was then neutralized by adding an equal volume of FBS or 10% DMEM(+) to the drop. An alternative method was to expose the primary colonies to 1.2 to 2.4 U/ml of dispase 1 in culture medium for 1 to 2 hours at 37.degree. C. With either method, the colonies were disaggregated into small clumps and single cells by repeated aspirations with a P10 Pipetman. Primary and secondary epiblast cultures were grown on STO feeder cells which were 95 to 100% confluent in either 10% DMEM(+) or 15/5 DMEM-M199 ›15/5 DMEM-M199 consisted of a 50:50 mixture of DMEM and Medium 199 (GIBCO) containing 15% FBS and 5% bovine calf serum!.

The cells derived from the preparation described in Example 1 were placed in secondary culture by growing on mitomycin C, 10 .mu.g/ml, (Sigma Chemical Company, St. Louis, Mo.), inactivated STO mouse fibroblast feeder cell layers ›CRL 1503, American Type Culture Collection (ATCC)! plated at 5.times.10.sup.4 cells/cm.sup.2. The culture medium was a 1:1 mixture of Dulbecco's modified Eagle's medium and Medium 199 with 10% FBS and antibiotics (10% DMEM-199+, GIBCO). Additional supplementation with 2-mercaptoethanol, nonessential amino acids, glutamine and nucleosides was as described in Example 1.

The PICM hepatocyte cell cultures displayed a morphology typical of primary cultured parenchymal hepatocytes (Berry and Friend, supra; Caperna et al., supra; Furukawa et al., supra). The cells were polygonal with well defined cell boundaries. Under phase-contrast microscopy they had dark, granular cytoplasm with large, distinct nuclei and nucleoli. Only mononuclear cells were observed. Previously, fetal and neonatal parenchymal hepatocytes were described as mononuclear whereas primary cultures of adult hepatocytes usually contained binuclear cells (Caperna, supra; Juan et al. 1992. Exp. Cell Res. vol. 202, pp. 495-500). Karyotype analysis of PICM-13 and PICM-19 also showed no evidence of polyploidy and confirmed the porcine origin of the cells. All the cell cultures differentiated to multi-cellular duct-like structures and into colony groupings that either accumulated and secreted lipid or developed spaces between individual cells similar to biliary canaliculi. The ductal and adipogenic differentiation phenotypes appeared to be mutually exclusive and occurred spontaneously over the course of 1 to 3 weeks in static co-culture on STO feeder-layers.

It has been found that in primary parenchymal hepatocyte cultures, small biliary canaliculi often developed within the monolayers (Furukawa, supra; Barth and Schwartz. 1982. PNAS. vol. 79, pp. 4985-4987). Likewise, small canaliculi of the same size and character developed in the monolayers of the PICM cells. Once formed, canaliculi were stable and their associated cells could retain their large polygonal shape for weeks without significant change. However, canaliculi were also observed in areas of the monolayer which eventually proceeded to duct formation and adipogenesis. In addition, individual canaliculi coalesced to form longer networks of interconnected canaliculi. Final differentiation into a mature ductal structure appeared to occur from these interconnected canaliculi. Duct formation was a terminal differentiation event since the constituent cells would no longer divide and grow if passaged, although they remained viable. Similarly, cultures which differentiated to adipogenesis grew very slowly when passaged and could not be sustainably cultured. In contrast, colonial areas that developed small canaliculi could be cloned and passaged to produce mass cultures without loss of differentiation potential. Thus, the development of canaliculi was non-terminal differentiation which could be a precursor to the terminal ductal or adipogenic differentiation programs.

Phase-contrast microscopy of the fully differentiated duct-like structures showed phase-dense material at the lumenal surface of the structures. This arrangement indicates a secretory structure. A modified form of the duct-like differentiation structure occurred in the PICM-19 culture after about 50 population doublings. This formation was characterized by the absence of a "lumenal" space. Instead, the duct-like structures were closed with the cells assuming a distinct pseudo-columnar arrangement. Heavy deposits of secretory matter appeared between the juxtaposed cells over a period of 2 to 3 weeks in static culture.

Primary cultures of rodent parenchymal hepatocytes sometimes accumulated lipid, consistent with their lipogenic in vivo function (Richardson et al., supra; Michalopoulos et al. 1982. Cancer Res. vol. 42, pp. 4673-4682). PICM-13, 18, 19, 26, 29 and 35 produced colonies that stained intensely with oil red O indicating lipid sequestration. Although adult porcine liver has little capacity to synthesize and accumulate fatty acids, the enzymatic machinery for de novo fatty acid synthesis is present (Mersmann, H. J. In: Tumbleson, M. E., ed. Swine in Biomedical Research, Vol. 2. New York: Plenum Press, 1986: 981-995). The production of triglycerides by the PICM hepatocyte cultures may result from the embryonic nature of the cells or the in vitro culture conditions, i.e., presence/lack of specific hormonal stimuli or possibly basic medium components. The lipid appeared to be actively secreted from the cells since large globules of lipid were often present in the extracellular matrix above the cells to the extent that the cells could no longer be clearly discerned. In contrast, lipid accumulation typically seen in certain fibroblast cultures, such as mouse 3T3 adipocytes, was cell bound, and the cells were prone to detach and float off the substrate because of the buoyancy of the lipid. The firm anchorage of the PICM cells undergoing lipid accumulation probably reflects an interconnective cytoskeleton network typical of cultured epithelial cells in general and of parenchymal hepatocytes specifically (Osborn and Weber, 1982. Cell. vol. 31, pp. 303-306; Marceau et al. 1985. Can. J. Biochem. Cell Biol. vol. 63, pp. 448-457; Levy et al. 1988. Differentiation. col. 39, pp. 185-195).

The RNA from STO mouse fibroblasts (CRL 1503 ATCC), PICM-16, ST pig testis cell line (CRL 1746 ATCC), 60-day fetal pig liver, adult pig liver and from PICM-19 cells grown in static culture for various periods after passage was analyzed by Northern analysis (Fourney et al. 1988. Focus. vol. 10, pp. 5-6). Twenty mg of total RNA, prepared as described by Chomczynski and Sacchi (1987. Anal. Biochem. vol. 162, pp. 156-159), was loaded per lane and transferred to Zetaprobe membrane (Bio-Rad). The Northern blot was hybridized successively with .sup.32 P-labelled cDNA probes to pig AFP and pig albumin, followed by autoradiography. Hybridization with a human glyceraldehyde-3-phosphate dehydrogenase DNA probe (pSP6-G3PDH, Ambion) was performed as a mRNA loading control.

It was previously demonstrated that porcine parenchymal hepatocytes expressed a variety of genes specific to liver function, such as AFP and albumin (Stone, R. T. 1989. J. Anim. Sci. vol. 67, pp. 1082-1089). These two proteins in particular are diagnostic for fetal hepatocyte character since normal AFP expression was restricted to fetal livers (Stone, supra). Moreover, these proteins are expressed during various early stages of hepatic stem cell development (Sigal et al., supra). Total RNA of PICM-19 cells was assayed by Northern blot, and high levels of AFP and albumin MRNA were detected. The expression of the two genes was minimal in cultures that had recently been passaged. The expression of AFP and albumin reached a peak in PICM-19 cultures that had remained in static culture for 2 to 3 weeks and had extensively differentiated into ductal structures. A relatively high expression was maintained in culture for an additional 1 to 2 weeks although some loss of the AFP and albumin signal was seen after longer periods in culture. Total RNA preparations from fetal and adult pig liver demonstrated that, as expected, AFP expression was restricted to fetal liver while albumin expression was found in both fetal and adult livers. A pig testis epithelial cell line, an epiblast-derived epithelial cell line, PICM-16, analagous to MDCK epithelial cells, and STO feeder cells were all negative for AFP and albumin mRNA expression. The expression profile is consistent with PICM-19 cells being fetal hepatocytes (Stone, supra).

Serum-free medium ›DMEM-199 with HGF (2 ng/ml), rat tail collagen (30 mg/ml) and 1.times.MITO.sup.+ serum extender (Collaborative)! was conditioned for 3 days by 50-75% confluent monolayers of PICM-19 or STO feeder cells alone. Conditioned media samples or diluted pig serum (1:50) were mixed 1:1 with loading buffer containing SDS and 2-mercaptoethanol. Samples were loaded onto a 12% polyacrylamide gradient gel and electrophoresis performed under described conditions (Laemmli, U. K. 1970. Nature. vol. 277, pp. 680-685). Proteins were transferred to nitrocellulose (0.2.mu.), blocked with gelatin and probed with polyclonal antisera. Specific immunoreactive antigens were visualized with alkaline phosphatase conjugated secondary antibody reagents (Sigma Chemical Co., supra).

PICM-19 secreted large amounts of transferrin and albumin into the serum-free culture medium. Coincident STO feeder cells alone were compared as a control for STO cell secretory activity and for nonspecific protein residues remaining from the initial 10% fetal bovine serum (FBS) culture of the cells. The unconcentrated conditioned media were analyzed by SDS-PAGE and immunoblotting. PICM-19 cells secreted several proteins that crossreacted with an antisera to whole pig serum. The two prominent protein bands were subsequently identified as albumin and transferrin based on apparent molecular weight and crossreactivity with specific antisera to porcine albumin and transferrin. In contrast, the STO feeder cell conditioned medium showed little or no reaction with the antibodies, demonstrating that the transferrin and albumin resulted from PICM-19 secretory activity.

PICM-19 cells were plated at 1.times.10.sup.5 cells/cm.sup.2 in 4-well plates with 1.9 cm.sup.2 well surfaces (Nunc) that had been previously seeded with 5.times.10.sup.4 mitomycin-treated STO cells per cm.sup.2. Porcine TGF.beta.1 (R&D Systems) was added at the indicated concentrations in 10% DMEM-199(+) at input and every 48-72 hours with refeedings. Cells were harvested after 7 days by trypsin-EDTA treatment and counted on a hemocytometer from 4 independent wells for each treatment. Cell counts were analyzed by ANOVA. Significant differences from control wells were determined by Dunnetts "T" procedure.

DNA synthesis in adult rat parenchymnal hepatocytes was shown to be inhibited by TGF.beta. (Nakamura et al. 1985. Biochem. Biophys. Res. Commun. vol. 133, pp. 1042-1060; Carr et al. 1986. Cancer Res. vol. 46, pp. 2330-2334). Addition of porcine TGF.beta.1 inhibited the growth and viability of the PICM-19 cells. PICM-19 cells that survived TGF.beta.1 (1 ng/ml) treatment for 7 days were enlarged, flattened and elongated. Continued incubation with TGF.beta.1 (1 ng/ml) for 2 to 3 weeks resulted in the complete loss of the PICM-19 cells from the culture as judged by microscopic observation.

A) STO mouse embryonic fibroblasts (ATCC CRL 1503), used as feeder cells, were grown as monolayers in standard T-flasks in DMEM with high glucose and 10% fetal bovine serum (10% DMEM). STO cells were treated with mitomycin-C to block cell division. Mitomycin-C-treated STO feeder cells were seeded on the inner surface of the ALD (previously coated with Collagen Type I at 20-60 .mu.g/ml) at 60,000 cells per cm.sup.2. One day following establishment of feeder cell monolayers in the ALD, PICM-19 clone 3BT cells were seeded onto each ALD surface (100,000-400,000 cells per cm.sup.2). A 1:1 mixture of low glucose DMEM and Medium 199 supplemented with nucleosides, .beta.-mercaptoethanol, non-essential amino acids and 10% fetal bovine serum (10% DMEM-199.sup.+ ; Talbot et al., 1994, supra) was recirculated through the ALD with a peristaltic pump to establish and maintain the monolayer. Alternatively, serum-free medium, DMEM-199.sup.+ supplemented with dexamethasone (1 .mu.M), DMSO (1%) and ITS.sup.+ (Collaborative Research, BD, Inc.), was used to maintain the cultures after establishment in the ALD. Cells were cultured either at pH 7.0 (to promote and maintain the mature hepatocyte differentiated state) or at pH 7.8, an environment that enhanced bile duct formation. Control of pH was accomplished by manipulating atmospheric CO.sub.2.

B) After seeding with feeder cells as described in A), a mixture of freshly isolated adult pig hepatocytes and PICM-19 clone 3BT cells (2:1) were seeded onto each ALD inner surface (100,000-400,000 cells/cm.sup.2). A 1:1 mixture of low glucose DMEM and Medium 199 supplemented with nucleosides, mercaptoethanol, non-essential amino acids and 10% fetal bovine serum (10% DMEM-199.sup.+ was circulated through the ALD with a peristaltic pump to establish and maintain the monolayer. Alternatively, serum-free medium, DMEM-199 supplemented with dexamethasone (1 .mu.M), DMSO (1%) and ITS.sup.+ were used to maintain the cultures after establishment in the ALD. Cells were cultured either at pH 7.0 or at pH 7.8, as above.

C) After seeding with feeder cells as described in A), a mixture of hepatocytes, PICM-19 cells and liver macrophages (10:5:1) to closely mimic liver cellular composition were added (Talbot et al., 1996, In press.) A 1:1 mixture of low glucose DMEM and Medium 199 supplemented with nucleosides, .beta.-mercaptoethanol, non-essential amino acids and 10% fetal bovine serum (10% DMEM-199.sup.+, as described above) were recirculated through the ALD with a peristaltic pump to establish and maintain the monolayer. Alternatively serum-free medium, DMEM-199.sup.+ was utilized as described in A) and B).

By growing the cultures on the clear flat perfusion vessel, it was possible to continually monitor cell morphology and differentiation by direct light microscopy. Typical parenchymal chord morphology was observed in PICM-19 clone 3BT cells grown in the ALD for two weeks with a recirculating flow rate of 3 ml/min (FIG. 4).

Equivalent numbers of 3BT cells were plated on STO cells in T-25 flasks and in the ALD. After three weeks of growth, 3-methylcholanthrene (3MC) was added to the device and and 2 T-25 flasks in serum-containing media. An additional 2 flasks of 3BT cells were incubated without 3MC. Two days later, all culture vessels were washed three times with PBS. Serum-free medium containing 7-ethoxyresorufin (8 .mu.M) and dicumerol (10 .mu.M) was added to the culture vessels and at specified times, samples of media were withdrawn, centrifuged and the resorufin concentration was determined by fluorescence spectroscopy. The recirculating flow rate in the ALD was 3 ml/min; initial volume was 40 ml.

After two weeks of growth, serum-containing medium was replaced with serum-free medium and aliquots of media were removed over a 48-hr period for the determination of urea concentration. At 24 hr, NH.sub.4 Cl (5 mM) was added to the medium. The recircultaing flow rate in the ALD was 3 ml/min; initial volume was 40 ml.

Induction of 7-ethoxyresorufin-O-deethylase activity with 3-methylcholanthrene was achieved in PICM-19 clone 3BT cells at activity levels equivalent to that seen in static cultures (FIG. 5). The synthesis and secretion of urea into media indicated the presence of highly active urea cycle activity (FIG. 6).

In summary, the parenchymal hepatocytes are not easily sustained in culture (Guguen-Guillouzo et al., supra; Perraud et al., supra; Tong et al. 1992. Exp. Cell. Res. vol 200, pp. 326-332), since the PICM hepatocyte cells have been shown to be strictly feeder-dependent and senesced if plated at split ratios higher than 1:5 off feeder cells. The PICM hepatocyte cells grew on top of or intermingled with the feeder-layer cells, and all the cultures grew slowly. PICM-19 had a doubling time of approximately 72 hours by standard culture enumeration and by counting the cells of individual colonies. PICM-19 has been in continuous culture for over 2 years, representing over 100 population doublings, assuming a 100% plating efficiency at each passage. Over the course of this period, single-cell cloning of PICM-19 was performed at least six times with the establishment of mass cultures of the subclones. The differentiation potential of the cells did not change over the passage history of the PICM-19 culture. The subclones of PICM-19 also retained the ability to form canaliculi and ductal structures and to accumulate cellular and extracellular lipid. The morphology of the undifferentiated PICM-19 cells remained the same with no signs of senescence during the continuous culture of subclonings. Therefore the PICM-19 cell culture is an infinitely self-renewing cell population, i.e., a cell line.

The PICM hepatocyte cell cultures were all derived by co-culture on STO feeder cells. STO feeder cells were shown to provide an environment suitable for the isolation and maintenance of hepatocarcinoma cells (Aden et al. 1979. Nature (London). vol. 282. pp. 615-616), mouse embryonic stem cells (Robertson, supra) and mouse primordial germ cells (Matsui et al. 1992. Cell. vol. 70, pp. 841-847; Resnick et al. 1992. Nature. vol. 359, pp. 550-551). The STO feeder cells were shown to produce a variety of growth factors (Robertson, supra; Matsui, supra; Schmitt et al., supra; Montesano et al.(67), supra; Montesano et al. 1991. Cell. vol. 66, pp. 697-711; Hilton et al. 1991. J. Cell Physiol. vol. 146, pp. 207-215). Some of these were mediators of hepatocyte growth and function such as LIF and HGF (Montesano et al. (67), 1991, supra; Montesano (66), 1991, supra; Hilton et al., supra). Hepatocytes were shown to express receptors for both LIF and HGF (Hilton, supra; Prat et al. 1991. Int. J. Cancer. vol. 49, pp. 323-328). LIF was reported to induce acute phase response proteins in hepatocytes (Baumann and Wong. 1989. J. Immunol. vol. 143, pp. 1163-1168), and HGF has been widely described as a potent hepatocyte mitogen (Zarnegar and Michalopoulos. 1989. Cancer Res. vol. 49, pp. 3314-3320; Strain et al. 1991. J. Clin. Invest. vol. 87, pp. 1853-1857; Nakamura et al. 1984. Biochem. Biophys. Res. Commun. vol. 122, pp. 1450-1459). Both LIF and HGF stimulated (P>0.05) replication of PICM-19 cells when assayed in a 5% FBS and reduced STO feeder cell environment. Thus, two growth factors directly related to hepatocyte growth and function are present in the STO feeder layer and may support the development of hepatocytes from the primary epiblast culture.

The PICM-19 cell culture is the only known example of a sustainable culture of pluripotent parenchymal hepatocytes. Their fetal phenotype and their ability to differentiate into ductal structures, adipogenic cells, and monolayers of larger polygonal cells indicate that the PICM cells represent early hepatocyte progenitors (Sigal et al, supra; Fausto and Meade, supra). Their establishement in secondary or infinite culture may not reflect some charcteristic of vitality which is peculiar to their epiblast source. The STO co-culture also supported the cultivation of hepatocytes from the 26-day pig fetal liver. These cell cultures formed monolayers of polygonal cells with interconnective canaliculi that could differentiate further to produce extracellular lipid or multicellular ductal structures. However, preliminary results indicate that the fetal liver-derived hepatocyte cultures are finite and senesce after variable degrees in in vitro passage. Thus, epiblast culture enables a derivation of fetal hepatocytes prior to differentiation into committed progenitor cells.

All references cited hereinabove are herein incorporated by reference.