A fresh look at the somatic cell genetics of hepatoma differentiation

The last six years have seen an impressive increase in our knowledge of the factors controlling the expression of liverspecific genes. The seminal papers are referred to below, and the reader can gain an overview of the various transcription factors and their inter-relationships from Tronche and Yaniv (1992) and from Kuo et al. (1992). Despite our greatly improved knowledge, many of the classical observations in the somatic cell genetics of differentiation are at best only partly understood. This Commentary will review those observations in the light of recent findings and examine some of the remaining problems in the context of the diversity of the physiological controls that are known to influence tissue-specific gene expression.

Tissue-specific genes are distinguished from ‘house-keeping’ genes not only by their tissue-specific pattern of expression: it is to be expected from their function that tissue-specific genes should additionally be subject to elaborate regulation by extracellular influences. The products of tissue-specific genes are of course not essential for a cell, but in no way are they the luxury that conventional terminology would have us believe. Far from being desirable - in some way beneficial to the cell - their expression is generally associated with cessation of cell multiplication and, in some instances, with cell death. Tissue-specific products are provided by the cell as a service to the organism as a whole, and, in providing that service, the cell must respond to external signals conveying the needs of the organism. In addition to the classical hormones, these signals will include growth factors and cell-matrix interactions that regulate the balance between stem cell multiplication and differentiation. The reciprocal regulation of cell growth and differentiation is obvious and unavoidable for those who study myoblasts, erythroleukaemia cells or epidermal cells, and it has been accorded due attention by those who work with normal hepatocytes (Leffert et al., 1982), but it is traditionally ignored by those of us who attempt somatic cell genetics with hepatomas. The well-differentiated hepatomas are attractive material for cell biology: they grow indefinitely and can be cloned, all in the same standard media, and yet they express their liver-specific genes throughout. Understandably, this has focused attention on the fundamental mechanism maintaining tissue-type, ‘determination’, which is seen to be cell-autonomous and remarkably stable. As a result the possible interplay between the regulation of cell growth and differentiation has tended to be overlooked. Likewise, despite extensive interest in the induction of tyrosine-aminotransferase by glucocorticoids, the central consideration that hormonal regulation most probably deserves in studies of hepatoma differentiation has been underestimated, at least until recently.

There are three kinds of result commonly obtained by somatic cell geneticists studying hepatoma differentiation. First, when a hepatoma is fused to a non-liver cell, the liverspecific genes of the non-liver cell may be activated (Peterson and Weiss, 1972; Malawista and Weiss, 1974; Darlington et al., 1974; Brown and Weiss, 1975). Such a result provided the first indication of the positively acting hepatic transcription factors that have since been so well documented (Gorski et al., 1986; Cereghini et al., 1987; Licht- steiner and Schibler, 1989; Frain et al., 1989). The two remaining kinds of result, both of which involve the suspension of differentiation, can help to define the role of these transcription factors. Hybrids between hepatomas and other cells normally show ‘extinction’ of differentiation (Schneider and Weiss, 1971; Szpirer and Szpirer, 1975), and hepatomas can also ‘dedifferentiate’ spontaneously (Deschatrette and Weiss, 1974). In both instances the loss of differentiation is accompanied by the loss of hepatic transcription factors (Kuo et al., 1992), and yet in both instances differentiation can reappear without the need for tissuespecific induction. This shows that the transcription factors need not be present to maintain determination: it is perfectly stable in their absence. In extinction and dedifferentiation it is the extent rather than the nature of differentiation that is affected. The thrust of this Commentary is to suggest that extinction and dedifferentiation may have less bearing on the specification of tissue-type than is generally assumed. These phenomena might, alternatively, need to be interpreted in the context of more general regulatory processes such as hormonal and cell growth control.

A case for the involvement of hormonal regulatory mechanisms in extinction is now relatively simple to make. Cell fusion can demonstrate a factor (or factors) in fibroblasts incompatible with liver gene expression, but need that factor be responsible for the silence of liver genes in normal unfused fibroblasts (i.e. need that factor have a role in determination of the fibroblast)? The sceptic might have suggested for example that fibroblasts, as part of their differentiation, maintain an especially active phosphodiesterase. The lowering of cyclic-AMP levels that would be induced when fibroblasts were fused to hepatomas would be expected to extinguish the expression of cyclic-AMPdependent liver genes. In this scheme, cyclic-AMP should reverse extinction in the hybrids, although, needless to say, it would not be expected to induce liver gene expression in fibroblasts. A significant reversal of extinction by cyclic- AMP was indeed demonstrated as long ago as 1978 (Leichtling et al.), but it was accorded little attention, perhaps because others (and this was certainly our experience) found little or no effect of cyclic-AMP on their hybrids. However, perturbation of the cyclic-AMP signalling system is now clearly established as contributing to extinction. Killary and Fournier (1984) used hybrids between a rat hepatoma and microcells made from mouse fibroblasts to establish the presence of an extinguishing locus, tse-1 (tissue-specific extinguisher 1), on mouse chromosome 11. The locus produces profound, though not total, extinction of a number of liver-specific enzymes. It was subsequently shown that this extinction can be readily reversed by cyclic- AMP (Thayer and Fournier, 1989) and that tse-1 codes for the R1 a regulatory subunit of cyclic-AMP-dependent protein kinase (PK-A: Boshart et al., 1991; Jones et al., 1991). Retrospectively, it is perhaps unfortunate that the locus was named tse-1, which could be taken to imply a significance more fundamental to determination than one would want to associate with a subunit of PK-A. However, though we may now be forced to conclude that tse-1 is peripheral to deter - mination, it would be quite unjustifiable to dismiss it as peripheral to extinction. In their microcell hybrids, Killary and Fournier (1984) achieved the first successful dissection of extinction. The same paper makes it clear that in whole cell hybrids, containing many fibroblast chromosomes, factors additional to tse-1 are involved, but the identity of these factors and the relative roles they play remain unknown. One might have anticipated that the loss of hepatic transcription factors such as HNF-1a during extinction would be of over-riding importance, but that is not consistent with the recent experiments of Bulla et al. (1992). In these experiments, hepatomas and their hybrids were transfected with an expression vector for HNF1a, so providing them with a permanent supply of this key transcription factor. It was found that HNF1 a could neither prevent nor reverse extinction. Extinction is seen to be genuinely multifactorial, and, in the light of the involvement of tse-1, we must accept that perturbations of hormonal response mechanisms may play a significant part.

Now we can look more closely at spontaneous dedifferentiation. Variant hepatomas lacking all or most of their liver specific traits were first detected as clones of cells with altered morphology (Deschatrette and Weiss, 1974), and they have also been isolated using special selective regimes (Choo and Cotton, 1977). Weiss and her colleagues found that their variants showed widely ranging stability: some redifferentiated spontaneously on continued growth, whilst others were much more stable. The most stable clones were screened with glucose-free medium (which selects for gluconeogenesis) and were found either to be completely stable, or else to revert with a very low frequency, e.g. 10^-8 (Deschatrette et al., 1980; Moore and Weiss, 1982). Similar results from this laboratory agreed entirely with these findings, but additional tests using medium supplemented with glucocorticoids and cyclic-AMP-raising agents revealed a substantial hormonal component to dedifferentiation. In some cases, variants that had otherwise seemed totally stable were induced to redifferentiate en masse: dedifferentiation in these variants was largely or wholly due to some alteration in hormonal responsiveness (Goss, 1984a). Others of our variants were found to have a different or more complex basis, since they showed no immediate response to hormones. Nevertheless, revertants could be selected from these cells in hormone-supplemented media with frequencies of at least 10^-6 (Goss, 1984b). The nature of the changes underlying dedifferentiation in variants such as these still remains a mystery. Recently, it has been suggested (Mendel et al., 1991; Kuo et al., 1992) that a dedifferentiated variant commonly used in molecular genetic experiments, Faofl-C2, is a ‘differentiation mutant’, a view that might be taken to imply the mutational loss of some key regulatory element. However, this suggestion, which rests on a consideration of the rarity of dedifferentiation and of subsequent reversion, should perhaps be regarded with some caution. It would appear difficult to argue that Faofl-C2 arose by a rare event, given that this variant is one of those that were initially detected by a simple visual search for hepatoma clones with altered morphology (Deschatrette and Weiss, 1974). There might also be cause to doubt the significance of the frequency cited for the reversion of Faofl-C2 (10^-9), since this refers to the recovery of gluconeogenesis as detected by selective medium in the absence of hormones (Deschatrette et al., 1980). In fact, it has been reported that Faofl-C2 can be induced to regain albumin synthesis by growth for two days as cell aggregates in suspension. The frequency of this reversion is greater than 1 in 10, and the subsequent recovery of gluconeogenesis (still measured in the absence of hormones) is then apparently raised in some instances to 10^-6 (Deschatrette, 1980). This pattern of reversion is clearly difficult to accommodate in a model dependent on spontaneous back-mutation. A further complication that applies to Faofl-C2, and to certain related variants, is that these cells have gained some factor capable of suppressing differentiation (hybrids between differentiated and dedifferentiated cells show extinction: Deschatrette and Weiss, 1975; Deschatrette et al., 1979). The identity of this suppressor is unknown. It had seemed that a likely contender was the DNA-binding protein HNF10 (also called vAPF or vHNF1: Cereghini et al., 1987). HNF10 resembles HNF1a and also recognises the same DNA motif. Furthermore, in dedifferentiation (and in extinction) when HNF1a falls, HNF10 rises. However, transfection experiments have shown that HNF10 has no significant repressor activity; indeed it may serve as a transcriptional activator (Mendel et al., 1991; Rey-Compos et al., 1991).

In summary, dedifferentiation is poorly understood. It is generally reversible, and so determination may not be involved. There is as yet no convincing evidence that the variants are ‘differentiation mutants’: the lability of dedifferentiation is probably better suited by explanations based on gene dosage or epigenetic effects. In some cases, as in extinction, hormonal response mechanisms are implicated, and there is also evidence for an unidentified repressor of differentiation.

It is relevant in this context to consider the only well-characterized genetic mutation that disrupts liver differentiation. This recessive condition, which results from deletions in the region of the albino locus on chromosome 7, is described in detail by Gluecksohn-Waelsch (1979). Affected mice die shortly after birth with morphological abnormalities of liver, kidney and thymus. Their livers produce minimal or significantly reduced amounts of a wide range of liver-specific proteins, including tyrosine aminotransferase, glucose-6-phosphatase, and some of the plasma proteins. Other liver-specific traits are expressed at normal levels and show normal patterns of hormonal induction, indicating that the hormonal response mechanisms are intact. The chromosomal deletions do not include the structural genes for the affected traits. One test of this was to fuse lethal-albino hepatocytes to a well-differentiated rat hepatoma. Hybrids were produced that expressed normal mouse tyrosine aminotransferase (TAT): the lethal-albino deletion had in no way compromised the TAT structural gene (Cori et al., 1981). It appeared that the deletion involved a regulatory locus, which has since been named hsdr-1, for hepatocyte-specific developmental regulation-1. Further characterisation of mutant liver revealed a marked reduction of the levels of mRNA for the hepatic transcription factors HNF1a and HNF4 (Gonzalez et al., 1990; Tonjes et al., 1992). Most recently, however, it has been shown that lethal-albino deletions all involve the structural gene for fumarylacetoacetate hydrolase (FAH), an enzyme of tyrosine metabolism, and that the phenotype of mutant hepatocytes reverts to normal when cultured in vitro in conditions that avoid the intracellular accumulation of tyrosine metabolites (Klebig et al., 1992; Ruppert et al., 1992). Furthermore, the mutant phenotype can be mimicked in normal hepatocytes by supplementing the medium with an excess of homogentisate, a non-toxic precursor of fumaryl- acetoacetate (Ruppert et al., 1992). These results can be expected to contribute to our understanding of inborn FAH- deficiency in man, but it is disappointing that the mutation does not define a true regulatory gene. The connection between fumarylacetoacetate accumulation and the mutant phenotype is obscure. One suggestion of Ruppert et al., (1992) is that the oncogene c-fos is involved (c -fos is overexpressed in mutant cells, perhaps in response to alkylation damage caused by fumarylacetoacetate). Whatever the explanation, here is a timely reminder that even when a proven genetic mutation is shown to affect both liver differentiation and the levels of liver-specific transcription factors, it need not necessarily follow that the fault is fundamental to the mechanism of differentiation.

Having seen that considerable caution is needed in interpreting the disruption of a process as complex as differentiation, let us nevertheless return to our problems in somatic cell genetics. It will be clear from the preceding sections that, though it may make a significant contribution, hormonal regulation accounts for only part of extinction and dedifferentiation. Can we assume that the balance of these phenomena has to do with determinative mechanisms? This is the point at which I would suggest that we should first assess the contributions of other general regulatory mechanisms, for instance those controlling cell growth. This is where the hard data run out, but let us construct an argument. Fusion to a fibroblast will add to a hepatoma the growth regulatory mechanisms typical of a fibroblast, and the very process of isolating hybrids is likely to select to some extent for the activity of those mechanisms. Might not this alteration to the hepatoma suspend differentiation, just as the introduction and expression of the RSV oncogene does in melanoblasts and chondroblasts? Like extinction in hybrid cells, RSV-induced extinction does not affect the underlying determination of the cells. This is clear from the reversibility of the effect in experiments using virus temperature-sensitive for transformation (Boettiger et al., 1977; Pacifici et al., 1977). Similar results have been obtained in the case of foetal rat hepatocytes transformed with temperature-sensitive SV40 (Schlegel-Haueter et al., 1980). Fournier and his colleagues have identified a second tissue-specific extinguisher locus (Chin and Fournier, 1989): it will be interesting to see if this plays any role in the regulation of cell growth. Dedifferentiation can be viewed in the same light, especially since some dedifferentiated variants are known to have gained the capacity to suppress differentiation (see above). Might the gained func - tion of these dedifferentiated cells represent the unbalanced activity of an oncogene? This view would help explain the apparently high frequency with which dedifferentiated variants can be obtained (Choo and Cotton, 1977; Goss, 1984b), since they might then possess some growth advantage. Cur - rent work in this laboratory on hybrids between a differentiated and a dedifferentiated hepatoma is revealing a close correlation between dedifferentiation and an increased degree of cellular transformation. Such a reappraisal of extinction and dedifferentiation is perhaps long overdue, considering the findings of Peehl and Stanbridge (1982) and of Harris (1985). When hybrid cells are transplanted into animals and so provided with a natural hormonal and nutritional background, they either grow as dedifferentiated tumours or, if they contain active tumour suppressor genes, they stop growing and differentiate, a clear demonstration of the reciprocal regulation of differentiation and tumori- genicity. Perhaps we should take more care to remember that hepatomas are tumours, and that tumours show a tendency to ‘progress’, that is to become dedifferentiated and faster growing. It would seem possible that extinction and dedifferentiation could ultimately prove to have more bearing on tumorigenesis and on the progression of tumours than on the fundamental mechanism of tissue specification. In that case, a better understanding of extinction and dedifferentiation will become all the more desirable.