Hepatic Nuclear Factor 1 (HNF1) shows a wider distribution than products of its known target genes in developing mouse

HNF1 (Hepatocyte Nuclear Factor 1; also named APF, LFB1 or HP1) has been described as a sequencespecific DNA-binding protein from rat liver that interacts with promoter elements present in many fiver-restricted genes, such as albumin, alpha- and betafibrinogen, alpha-l-antitrypsin, alpha-fetoprotein, pyruvate kinase, transthyretin and aldolase B, among others (Cereghini et al. 1988; Schorpp et al. 1988; Courtois et al. 1987, 1988; Baumhueter et al. 1988; De Simone et al. 1987; Hardon et al. 1988; Jose-Estanyol and Danan, 1988; Vaulont et al. 1989; Costa et al. 1988; Tsutsumi et al. 1989; Frain et al. 1990; Toniatti et al. 1990; Rodriguez de Cordoba et al. 1991; reviewed in Rey-Campos and Yaniv, 1991). For most of these genes, it was shown that the HNF1-binding site is essential for the transcriptional activity of the corresponding promoter (Godbout et al. 1986; De Simone et al. 1987; Courtois et al. 1987; Schorpp et al. 1988; Cereghini et al. 1988; Lichtsteiner and Schibler, 1989; Herbomel et al. 1989; Tronche et al. 1989; Toniatti et al. 1990). Titration of the factor in liver nuclear extracts, using an excess of specific oligonucleotides containing an HNF1-binding site, abolishes in vitro transcription from the albumin promoter (Cereghini et al. 1988). In addition, a highly efficient, liver-specific promoter can be obtained with only a TATA box and an HNF1 site (Tronche et al. 1989; Ryffel et al. 1989; Maire et al. 1989). Finally, spleen nuclear extracts that are inactive in transcription of the mouse albumin promoter in vitro can be rendered active by addition of purified rat HNF1 (Lichtsteiner and Schibler, 1989).

While HNF1 could be detected in liver or differentiated hepatoma cells expressing liver-specific functions, a distinct DNA-binding activity, named vAPF or vHNF1, was detected in dedifferentiated hepatomas and somatic hybrids that had lost the capacity to express liver-specific markers (Cereghini et al. 1988; Baumhueter et al. 1988). The DNA-binding properties of HNF1 and vHNF1 were shown to be indistinguishable by several criteria, thus suggesting a close relationship between the two factors at least in their DNA-binding domain (Cereghini et al. 1988). These results and the correlation between the presence of HNF1 and the differentiated state led us to postulate a key role for HNF1 during hepatocyte differentiation (Cereghini et al. 1988).

Recently, HNF1 cDNA clones from rat liver and rat hepatoma cells were independently obtained (Frain et al. 1989; Baumhueter et al. 1990; Chouard et al. 1990). Sequence analysis showed that the protein is 628 amino acids long and presents in its middle portion a clear homology to the DNA-recognition helix of homeoproteins (Affolter et al. 1990; Otting et al. 1990; Kissinger et al. 1990). Alignment of the HNF1 homeodomain with all reported homeodomain sequences (Scott et al. 1989) showed that the HNF1 homeodomain is highly divergent, lacking conserved residues inside and outside the third helix (Chouard et al. 1990). A better alignment can be obtained by the introduction of a loop of 24 amino acids between helices 2 and 3, instead of the canonical 3 amino acids in the turn (Finney, 1990; Baumhueter et al. 1990; Chouard et al. 1990; Nicosia et al. 1990). In addition, a weak homology with the POU A box of POU proteins (Herr et al. 1988) can be found upstream of the HNF1 homeodomain (Frain et al. 1989; Baumhueter et al. 1990; Chouard et al. 1990); however, both the POU-specific box and the POU homeobox are too poorly conserved to consider HNF1 as a member of the POU family of transcription factors (Chouard et al. 1990).

The functional relevance of the homeodomain homology has been confirmed by deletion analysis. The HNF1 homeobox was shown to be necessary (Frain et al. 1989), but not sufficient, for specific DNA-binding in vitro (Nicosia et al. 1990; Chouard et al. 1990). Additional sequences contained in the N-terminal half of the polypeptide are also required, and they are involved in dimerization of HNF1, which is known to occur in the presence or absence of its DNA target site (Chouard et al. 1990; Nicosia et al. 1990). Amino acids located C-terminal to the homeodomain are also functionally important since deletion of this region almost totally abolished the capacity of HNF1 to activate transcription in vitro (Nicosia et al. 1990) as well as in vivo (D. Sourdive, personal communication). The importance of the different HNF1 domains is further supported by the observation that the human and mouse HNF1 protein sequences are, respectively, 93 % and 99 % homologous to the rat counterpart (Bach et al. 1990; Kuo et al. 1990).

HNF1 mRNA is absent in dedifferentiated hepatoma cells or extinguished somatic hybrids (Baumhueter et al. 1990; Cereghini et al. 1990). Nuclear run-on experiments have shown that HNF1 is controlled at the transcriptional level in these cases (Cereghini et al.(1990). These results indicated that vHNF1 was not the result of a post-translational modification of HNF1 but a protein encoded by another gene. Several cDNA clones were recently isolated on the basis of postulated partial homology with the HNF1 homeodomain and characterized as coding for vHNF1 (Rey-Campos et al.(1990). Both HNF1 and vHNF1 were introduced into different cell types, including dedifferentiated hepatomas, by transfection with plasmids carrying the corresponding cDNAs under the control of the Rous Sarcoma Virus LTR promoter. With both factors, it resulted in the activation of a cotransfected CAT construct driven by the albumin promoter, normally inactive in these cell fines, although with variable efficiencies (Rey-Campos et al. 1991; F. Tronche, personal communication). In addition, vHNF1 appeared to be expressed along with HNF1 in differentiated hepatoma cells, although at much lower rates (Rey-Campos et al. 1991). These results showed that the mere molecular switch between HNF1 and vHNF1, previously thought to be, respectively, a positive and a negative effector of the hepatocyte-specific gene expression, was not sufficient to explain the change in activity of genes like albumin, during dedifferentiation, extinction or reversion.

Considering that HNF1 is a homeobox-containing. protein that is involved in the transcriptional control of terminal differentiation markers, it was of interest to clarify its pattern of expression during development and to compare it to that of the genes that it is known to regulate. Interesting preliminary information was obtained about the distribution of the HNF1 mRNA within subsets of adult organs by northern blot analysis (Frain et al. 1989; Chouard et al. 1990) and RNAase protection studies (Baumhueter et al. 1990; Kuo et al. 1990). Positive signals were detected in rat or mouse liver, kidney and intestine as well as in mouse stomach and, to a lesser extent, in rat spleen and thymus. Among organs that were examined, lung, brain and heart appeared negative in both rat and mouse, as well as skin, testis and thyroid in rat and spleen and ovary in mouse (Frain et al. 1989; Chouard et al. 1990; Baumhueter et al. 1990; Kuo et al. 1990).

Since strong signals were detected in various non-hepatic organs (mainly kidney and the gastrointestinal tract), we wondered whether it concerned distinct cell layers within them and wanted to complete the analysis by scoring the entire body using in situ hybridization. In addition, we wanted to ask whether HNF1 was expressed at the initial steps in liver ontogeny and whether its expression pattern would follow that of the albumin gene. We also used gel retardation assays and western blot analysis to confirm the presence of genuine HNF1 protein in kidney. Comparisons with the albumin gene expression pattern, as well as with published data about the distribution of other known HNF1 targets, indicate that this tissue-specific transcription factor depends on the cellular context for its forcing high levels of gene expression. In addition, these comparisons suggest that some HNF1-controlled gene transcription might occur in non-hepatic tissues during mouse development.

Foetuses were obtained at days 10.5, 11.5, 15.5 and 16.5 of gestation from F₁ C57BL/6 ×DBA/2 female mice which had been mated to F₁ C57BL/6 ×DBA/2 males. The morning when a vaginal plug was found is day 0.

Immediately after dissection, foetuses were fixed overnight at 4°C in 4% paraformaldehyde in PBS. Dehydration through graded alcohols, paraffin embedding and preparation of 7 μm thick embryo sections was performed as described by Sassoon et al. (1988).

A 490bp PvuII fragment of rat HNF1 cDNA (nt 992 to 1482; Chouard et al. 1990) and a 550 bp PstI fragment of rat albumin cDNA clone pRSA57 (Sargent et al. 1979) were respectively subcloned in the SmaI or PstI sites of a T7/T3 promotercontaining pBS vector (Stratagene). Sense and antisense RNA probes, labeled with ³⁵S-UTP (>1000 Ci mmol^-1, Amersham) to a specific activity of 1.2 or 1.6 ×10⁹disints min^-1 μg^-1, were generated from each cDNA insert using bacteriophage T3 or T7 RNA polymerase (Pharmacia) as appropriate. Probes were hydrolyzed at pH 10 to an average fragment length of 100 bp prior to hybridization (Sassoon et al. 1988).

Before hybridization, slides were rehydrated through graded alcohols, treated with 20 μg ml^-1 proteinase K, fixed with 4 % paraformaldehyde/PBS, acetylated in 0.25% acetic anhydride in 0.1M triethanolamine, dehydrated through graded alcohols and dried, as described in Sassoon et al. (1988). Hybridizations were carried out at 52 °C for 16 h with a mixture containing 0.09 ng μl^-1 (HNF1) or 0.06 ng μl^-1 (albumin) RNA probe, 50% formamide, 10mM DTT, 0.3 M NaCl, 10mM Tris –HCl (pH7.4), 10mM NaH₂PO₄ (pH8.0), 1mM EDTA, 10% dextran sulfate, l × Denhardt’s solution (0.2% Ficoll, 0.2% polyvinylpyrrolidone, 0.2% BSA fraction V) and 1mgml^-1 yeast tRNA. After hybridization, sections were thoroughly washed at a final stringency of 50 % formamide, 2 ×SSC at 65°C, treated with 20 μgml^-1 RNAase A and 2 μgml^-1 RNAase T1 for 30min at 37°C, rinsed thoroughly, washed in 0.1 ×SSC for 15min at 37°C, dehydrated in graded alcohols containing 0.3 M ammonium acetate, air dried and covered with a Kodak NTB2 emulsion. Exposure times were 1 to 2 weeks for HNF1 probes and 3 to 7 days for albumin. Slides were counter-stained with toluidine blue. Controls included hybridization of adjacent sections with sense RNA probes, and RNAase A treatment of sections prior to hybridization with antisense probes.

Nuclear extracts were prepared as described in Cereghini et al. (1987) for rat liver and kidney and according to Schreiber et al. (1989) for H5 cells. Nuclear extracts were preincubated, during 10min at room temperature, with 10 ng of either unlabelled competitor oligonucleotide or 1 μl of a dilution of anti-HNF1 rH183 serum, 80 % (v/v) in TBST, containing or not 0.5 mg ml^-1 of peptide rHp3 as specific competitor antigen. Gel retardation assays were then performed as previously described (Cereghini et al. 1988), using 1 ng of ³²P-labelled PE56 (HNF1 site) and either 6 μg, 4.5 μg or 30 of liver, kidney or H5 cell nuclear extracts, respectively, as indicated in Fig. 3. The H5 cell signals shown in the Fig. 3 result from a 3-fold longer autoradiography exposure than for the liver and kidney ones. Resolution of high molecular weight complexes was increased by running 16 cm native polyacrylamide gels for 7h at 4°C at 12 V cm^-1.

The peptide rHp3 containing the 21 amino acids from F541 to T561 in the rat HNF1 sequence (FTSDTEASSEPGLHEPSS-PAT, Chouard et al. 1990) was chemically synthesized and covalently linked to Keyhole Limpet Haemocyanin (KLH) using glutaraldehyde (Ausubel et al. 1989). Various amounts of the partially linked KLH-rHp3 mixture were injected into a Blanc-Bouscat rabbit as follows: day 1: 400 μg with complete Freund’s adjuvant into popliteal lymph nodes; day 23: 400 μg with incomplete Freund’s adjuvant into the subscapular cavity; day 31: 200 μg in PBS intramuscularly; day 32: same intravenously; day 46 bleeding and serum (rH183) preparation. Western blot analysis was performed according to standard procedures (Harlow and Lane, 1988). Briefly, 0.6 μg and 0.45 μg of liver and kidney nuclear extracts, respectively, were separated by 10% SDS –PAGE and transferred to nitrocellulose before incubation with the first antibody and revelation by an alkaline phosphatase-conjugated anti-rabbit serum (Promega). First antibody solution consisted in a 1:250 dilution of rH183 in TBST (10 MM Tris –HCl pH 8.0; 150 MM NaCl; 0.05% Tween 20) containing or not 4 μgml^-1 of peptide rHp3 as rat HNF1-specific competitor antigen (10fold molar excess of peptidic antigen over specific antigenbinding sites, best estimated as 13 μM in undiluted serum). Molecular weight markers were E. coli beta-galactosidase (116.5 × 10³M_r), rabbit muscle phosphorylase-b (97.4 ×10³M_r), bovine serum albumin (66.2 ×10³M_r) and hen egg white ovalbumin (45 ×10³M_r), (Bio-Rad).

A segment of rat HNF1 cDNA corresponding to the 3 ’ end of the HNF1 coding region was chosen as a probe to avoid cross-hybridization with other homeoproteins, in particular with vHNF1, a protein closely related to HNF1 in its DNA-binding domain but highly divergent in the region of our probe (see Materials and methods and Rey-Campos et al. 1991). The HNF1 cDNA shows a striking conservation at the nucleotide level in mammals (Bach et al. 1990 and Kuo et al. 1990). Indeed, the sequence of our probe is 95 % homologous to the corresponding segment in mouse cDNA (Kuo et al. 1990), thus allowing us to use the rat probe in high-stringency analysis of mouse HNF1 transcripts. The specificity of our HNF1 probe was further confirmed by northern blot analysis and by competition experiments in which full-length sense RNA, synthesized in vitro from HNF1 (Chouard et al. 1990) or vHNF1 (Rey-Campos et al. 1991) cDNA clones, were added during hybridization at a 10-fold molar excess with respect to the HNF1 probe. While the HNF1 sense RNA completely abolished the signal over all the positive organs, hybridization was not affected by the vHNF1 sense RNA (data not shown).

When 10.5 days post coitum (pc) mouse embryo sections were hybridized with the HNF1 probe, a positive signal was detected over the developing liver and the visceral endoderm of the yolk sac (Fig. 1A, B). No positive signal was detected elsewhere even after 1 month’s exposure (data not shown). In contrast, at the same stage, albumin expression was evident only for the liver primordia (Fig. 1D –F), in agreement with the reported results that mouse albumin mRNA expression is 10-fold lower in yolk sac than in liver (Sellem et al. 1984).

By day 11.5 pc, the body plan of the adult mouse is already delineated (Rugh, 1990). At this embryonic stage, HNF1 labelling was visible not only in the liver cells and the yolk sac but also in the primary intestinal loops, while albumin mRNA was again only detected in the liver (data not shown).

The presence of HNF1 transcripts in tissues other than liver was even more evident for 15.5 day embryos. As shown in Fig. 2A and B, HNF1 hybridization was detected in liver as well as in gut and kidney. This is consistent with the RNAase protection data obtained with the adult rat and mouse (Baumhueter et al. 1990; Kuo et al. 1990). However, Kuo et al. (1990) reported some RNAase protection of total RNA prepared from adult mouse stomach, whereas the portions of the digestive tract that appeared to express HNF1 transcripts at day 15.5 pc were limited to distal duodenum, small and large intestine and rectum. Stomach, proximal duodenum and pancreas were negative at this stage (Fig. 2B).

The in situ hybridization shows that positive cells correspond to the epithelial layer of the lumen of intestine (Fig. 2B, E and F), while in kidney, the labelling was exclusively cortical and mostly confined to the boundary with medulla (Fig. 2B). Although the histology of the sample was not good enough for firm conclusions to be drawn, hybridization in the kidney seemed to be mostly associated with collecting tubules, while glomeruli and mesenchymal cells were negative. Attempts to obtain better conservation of the kidney morphology by cryo-sectioning failed.

A high background was systematically detected with the sense RNA probe in the pancreas, liver and spleen from 15.5 day embryos (Fig. 2C). The signal in the pancreas might indeed be due to a specific cross hybridization of our sense probe, the significance of which was not investigated. In the case of liver and spleen, the background was probably due to the usual nonspecific adsorption of RNA probes to erythrocytes, since at this stage both organs are actively hematopoietic (Rugh, 1990). However, the hybridization signal with the antisense probe was above the background, although less clearly in spleen than in liver. Still, we suggest that the spleen might contain low amounts of HNF1 mRNA at this stage (compare Fig. 2B and C). Finally, no positive signal was observed in any other tissues examined, including brain (results not shown). Similar patterns of hybridization were observed in embryos of 16.5 days (data not shown).

Thus, some minor discrepancies appear between previous data on the adult and our results on the non-hepatic embryonic tissues. Indeed, Baumhueter et al. (1990) detected a low signal in the adult rat spleen but it was not confirmed in mouse (Kuo et al. 1990); these authors also detected the HNF1 transcript in thymus, whereas no thymus expression was detectable by in situ hybridization, even at day 16.5 pc when the organ was easily recognizable (data not shown). Thus, it is likely that the HNF1 expression undergo some minor developmental and species-specific regulation.

Finally, as for earlier stages, albumin expression was detected only in the liver from 15.5 and 16.5 day embryos, even after one month of autoradiographic exposure (Fig. 2D and data not shown).

The fact that HNF1 mRNA was expressed at rather similar levels in the kidney, gut and liver of mouse embryos, raised the question of whether the protein was actually synthesized in the corresponding tissues. To test this issue, we prepared nuclear extracts from those organs in the adult. In nuclear extracts from kidney, we detected a DNA-binding activity specific to the HNF1-binding site (Fig. 3: lanes 6, 9 and 10). Among the different complexes that could still correspond to related activities distinct from the genuine HNF1, two were displaced by incubation of the extracts with an anti-HNF1 polyclonal serum (Fig. 3: lanes 6 and 7). The specificity of this supershift effect was verified by antigenic competition using the synthetic peptide against which the antibodies were raised (Fig. 3: lane 8). The efficiency of the test was verified on the liver HNF1 complex (Fig. 3: lanes 1, 2 and 3). Its specificity as regards to the HNF1-related factor called vHNF1 (Cereghini et al. 1988; Courtois et al. 1988) was tested on extracts from the H5 dedifferentiated hepatoma cell line that contains vHNF1 and not HNF1 (Cereghini et al. 1990; Fig. 3: lanes 11 –13). These results show that the HNF1 mRNA is efficiently translated in the kidney. In addition, as described elsewhere (Rey-Campos et al. 1991), the difference in the DNA-binding activities on the HNF1 site, between liver and kidney can be fully explained by a much higher level of the vHNF1 protein in the latter. Indeed, the HNF1 –DNA complexes involve dimers of the factor (Chouard et al. 1990) and vHNF1 also forms homodimers as well as heterodimers with HNF1 on DNA (Rey-Campos et al. 1991; Fig. 3). In accordance with the high levels of mRNA that we observed in the kidney, the HNF1-specific DNA-binding activity on the rat albumin proximal element (PE56, Cereghini et al. 1988) seemed very high in kidney, as compared to liver.

Since the presence of high amounts of the HNF1 protein in an organ where the albumin gene is silent was puzzling, we wondered whether the absence of HNF1-promoted transcription could be attributed to modification of the HNF1 protein. In an attempt to address this question, we performed western blot analysis with kidney extracts using the HNF1-specific immune serum described above. In this type of experiment, we usually observed the liver HNF1 as a broad band, illustrating a marked heterogeneity in molecular weight, comparable to silver-stained purified material (Chouard et al. 1990 and T.C., unpublished results). This pattern is probably due to the high degree of glycosylation of this protein (Lichtsteiner and Schibler 1989; T.C. unpublished results). Indeed, the HNF1 protein produced from the cDNA in a cell-free system gave rise to a sharp band which co-migrates with the fastest species of heterogenous native HNF1 (Chouard et al. 1990). In the kidney, HNF1 showed the same heterogeneity observed in liver (Fig. 4). This pattern was found in all HNF1-producing mammalian cell types analysed so far, including various cell lines transfected with an exogenous HNF1 gene (T.C., unpublished results).

Our attempts to prepare total or nuclear extracts from the intestine failed to produce any detectable HNF1 protein, whether tested by band-shift assay or western blot analysis. In control experiments, where intestine was mixed with liver prior to homogenization, in the presence of different protease inhibitors, we also failed to detect the liver HNF1 (data not shown). We concluded that, even if it was present in the intestine, the HNF1 protein could not be detected in extracts by these methods, probably because of high proteolytic activity. We were able to detect low DNA-binding activity with HNF1 specificity in cell lines originated from colon carcinoma. The band-shift pattern in Caco human cells was qualitatively comparable to that observed in kidney but we could not verify the presence of genuine HNF1 in those cells, since our anti-rat HNF1 serum did not recognize the human protein (T.C. unpublished results). Rat TRb cells, derived from a partially dedifferentiated colon carcinoma (François Martin, personal communcation) gave rise to an activity qualitatively similar to that observed in the H5 cells and not specifically altered by the anti-rat HNF1 immune serum (data not shown). Our failure to identify unambiguously an intestinal form of the HNF1 protein is probably due to technical constraints rather than to the absence of the protein in this tissue.

Molecular cloning of many mammalian transcription factors during recent years has led to a better understanding of transcriptional control mechanisms underlying the differentiation processes (for a review, see Mitchell and Tjian, 1989). For example, the case of the leucine-zipper containing C/EBP protein and the helix-loop-helix containing MyoD factor are particularly enlightening. C/EBP was first characterized by its specific DNA-binding activity in liver extracts and has now been shown to control terminal adipogenesis, also antagonizing cell growth (Umek el al. 1991). MyoD was first identified as capable of forcing undifferentiated precursor cells to enter the myogenic differentiation pathway and is now well characterized as a transcriptional factor (Weintraub et al. 1991).

Another group of tissue-specific transcription factors, isolated by biochemical approaches, happened to contain an homeodomain (Affolter et al. 1990) of new type in addition to specific conserved residues, which defined the POU-protein family (Ruvkun and Finney, 1991). One member of this family, the pituitary-specific Pit-l/GHF-1 factor, was found containing a mutated homeodomain in mouse dwarf mutants in which several cell types of the pituitary gland are affected (Li et al. 1990). Thus, the reverse genetic approach confirmed the role played by the homeobox-containing genes in transcriptional control of development, including terminal differentiation processes.

In the same spirit, we have previously characterized a liver-specific transcription factor, APF, as a nuclear protein that interacts with critical promoter sequences from the rat albumin gene as well as from several other liver-restricted genes (Cereghini et al. 1988; Tronche et al. 1989; Herbomel et al. 1989). APF was later purified, its corresponding cDNA was isolated and sequenced, and shown to be identical to the independently cloned factors LFB1 and HNF1 (Chouard et al. 1990; Frain et al. 1989; Baumhueter et al. 1990). Several structural features of HNF1 led us and others to classify this protein as the prototype of a novel subclass of transcription factors distantly related to homeoproteins (Frain et al. 1989; Finney, 1990; Baumhueter et al. 1990; Chouard et al. 1990). Preliminary RNAase protection data on tissue distribution of the HNF1 mRNA in adult organs revealed unexpectedly strong signals in non-hepatic tissues (Baumhueter et al. 1990; Kuo et al. 1990). Nevertheless, an overview of the entire body, as well as the information about the HNF1 expression during the early steps of organogenesis were still lacking. In addition, quantitative data, coming from organ extractions, appeared to us as misleading in the absence of histological information. Therefore, it seemed necessary to perform an in situ hybridization study of the HNF1 expression pattern during mouse development.

The present report shows that the HNF1 mRNA is already transcribed by day 10 pc in post-gastrulation embryos, in cells that have begun to differentiate into more-commited cell types, as in the liver primordium. At day 15.5 pc, the HNF1 transcription is detected not only in liver but also in kidney and gut. These results are consistent with RNAase protection data in adult rat and mouse, although some discrepancies appear in non-hepatic tissues such as stomach, spleen and thymus suggesting that the HNF1 gene transcription might undergo some weak tissue-, developmental- and species-specific variations (see the result section). In addition, the intensity of the in situ hybridization signals suggests that the level of HNF1 RNA in the positive cells from gut and kidney at days 15.5 or 16.5 is at least similar if not higher than that of the liver cells.

The detection of the HNF1 mRNA in non-hepatic tissues did not prove that the HNF1 protein was actually synthesized in such organs. Kidney nuclear extracts were shown to contain high amounts of true HNF1 protein in addition to related DNA-binding activities due to the presence of vHNF1 (see the result section and Rey-Campos et al. 1991). The kidney HNF1 protein was indistinguishable from the liver one by immunodetection and as active in DNA-binding. In contrast, we failed to detect HNF1 in nuclear extracts from intestine. However, we believe it to be due to high protease activity, rather than an hypothetical tissue-specific post-transcriptional regulation. This is further suggested by the detection of HNF1-like DNA-binding activity in some colon carcinoma cells (see the result section).

Since these results strongly suggested that high levels of authentic HNF1 transcription factor might be active ‘in tissues other than liver during mouse development, we wondered whether this was consistent with the expression pattern of genes that it is known to control. In the case of albumin, in agreement with published results, we observed high levels of its transcript in the liver as soon as it begins to bud from primitive gut; in contrast, its expression in tissues expressing high levels of HNF1 mRNA, like visceral endoderm of the yolk sac at day 10.5 or 11.5 pc, was very low; in the 15.5 day pc gut or kidney, the albumin gene transcription was undetectable even after one month exposure while its expression in liver was obvious after only 1 – 2 days of autoradiography. The situation is not different in the adult rat, where albumin remains barely detectable in kidney (Nahon et al. 1988; Poliard et al. 1988). Similar results have been reported for other HNF1-dependent genes such as rat alpha-fetoprotein (Poliard et al. 1988; Nahon el al. 1988), mouse alpha-1-antitrypsin (Ruther et al. 1987; Koopman et al. 1989), rat pyruvate kinase (Guder and Ross, 1984), or rat alpha- and betafibrinogen (Baumhueter et al. 1990). One exception is aldolase B for which relatively high levels were observed in kidney although still lower than in liver (Ruppert et al. 1990). Thus, the general rule for known HNF1 target genes remains a marked specificity of expression in hepatocytes.

In this context, our results show that large amounts of HNF1 protein, as demonstrated in the kidney, are not sufficient to force the high expression levels of its known target genes. Thus, even though target gene expression in hepatocytes is strongly dependent on the occupation of HNF1 sites located in their regulatory regions (Schorpp et al. 1988; Cereghini et al. 1988; Monaci el al. 1988), their proper regulation must involve some other factors. Such co-factors might first act positively, either at the level of chromatin organization of the various loci or at the level of interaction with the transcriptional apparatus itself. Indeed, recent results indicated a cell-type dependence of HNF1 transcriptional activity on co-transfected albumin or beta-fibrinogen promoter derivatives (F. Tronche, personal communication; Kuo et al. 1990). Moreover, HNF1 was not able to turn on the endogenous albumin gene of several cell types when transiently expressed (F. Tronche and G. Giriffo, personal communication). Second, other factors might act negatively on the expression of specific HNF1 target genes in the appropriate tissues. For example, a ‘silencer’ element was characterized as part of the –10.5 kb albumin enhancer (Herbst et al. 1990 and references therein) and a so-called ‘repressor’ segment was recently localized between the promoter and the first enhancer of the alpha-fetoprotein gene and shown to be able to switch off its expression specifically after birth (Vacher and Tilghman, 1990).

Nevertheless, several genes known to contain functional HNF1-binding site(s), are expressed in non-hepatic organs which contain HNF1, at levels frequently above the background observed in other tissues. Indeed, it is not surprising to find a coexpression in liver and gut of many genes since hepatocytes and epithelial cells of the intestine are derived from the embryonic foregut. Interestingly, this is the case for alphafetoprotein, for which mutations in an HNF1-binding site had a dramatic effect on transcription in gut or Caco-2 cells, even more than in fiver (Tyner et al. 1990; Vacher and Tilghman, 1990). In contrast, it is puzzling to observe co-expression in these endoderm-derived tissues and in kidney which is derived from mesoderm. In situ hybridization data available for several HNF1 target genes reveal low expression in kidney tubular structures, the same structures that are suggested to express the HNF1 mRNA by in situ hybridization. Albumin and alphafetoprotein were found in rat foetuses in all developing tubular cells, while post-natally they were mostly associated with the distal convoluted tubules, the collecting ducts and the loops of Henle (Poliard et al. 1988). For mouse alpha-l-antitrypsin, while no hybridization was observed in the metanephros of 14.5 days pc or later embryos, adult mice showed a positive signal probably associated with the distal convoluted tubules or the loops of Henle (Koopman et al. 1989).

Actually, the presence of high levels of HNF1 in developing non-hepatic tissues suggests the existence of not-yet-identified target genes that would be controlled by this factor. Gene disruption or inactivation of HNF1 in different cell lines should help identify such genes. Obvious candidates for such genes are those coexpressed in kidney and liver or intestine: this is the case of different glycosidases and peptidases which are expressed in both the brush borders of the small intestine and kidney proximal convoluted tubuli (Semenza, 1986); similarly, the gluconeogenic enzymes, phosphoenolpyruvatecarboxykinase (PEPCK), fructose-1,6-biphosphatase and glucose-6-phosphatase, as well as the positive regulatory factor alf (implicated in an albino lethal deletion), are expressed in liver and in the kidney proximal tubules (Guder and Ross, 1984; Ruppert et al. 1990). Practically, the mechanisms by which derivatives of different embryonic layers such as endoderm (liver and gut) and mesoderm (kidney) follow common expression patterns remain to be elucidated. Analysing the determinants of the regulation of regulators such as HNF1 and its relatives might enlighten these questions, in terms of both development and evolution of the regulatory processes.

We are grateful to B. David-Watine for the gift of nuclear extracts, to Franck Bourgade for advice in rabbit immune serum preparation, to J. Rey-Campos and S. Cereghini for making available the vHNF1 cDNA clone, to J. Gaillard, G. Lyons and M.-O. Ott for valuable discussions, to J. Ham, Curtis Pfarr and M. Weiss for critical reading of the manuscript and to J. Ars for typing the text. This work was supported by grants from INSERM, ARC, LNFCC and the FRMF. M. B. was supported by a fellowship from the ARC.