We examined the control by hormones and culture conditions of the expression of pyruvate kinase L, aldolase B, and a liver-specific 5.4-kb mRNA species [Pichard, A. L. et al. (1985) Biochem. J. 226, 637-644] in three rat hepatoma cell lines, MH1C1, Fao and Faza. The expression level of these markers ranges from 2% (for pyruvate kinase L mRNA) to 10-12% (for 5.4-kb mRNA species) of the glucose-induced mRNA values found in rat liver. The mRNAs of the three liver-specific genes strongly decrease after treatment of the hepatoma cells with cyclic 8-bromo-AMP, cyclic dibutyryl-AMP or forkolin, pyruvate kinase L mRNA being the most sensitive to this inhibiting effect. In contrast, the concentration of pyruvate kinase L mRNA nuclear precursors is not modified by the cyclic AMP analogues, indicating that these agents do not act at the transcriptional level but, instead, probably destabilize the transcripts. Glucose or fructose does not modify the expression of these three marker genes in any of the studied cell lines. Insulin is inefficient in modifying concentrations of the mRNAs for pyruvate kinase L and aldolase B, alone or in the presence of carbohydrates. In contrast, it stimulates about fivefold the expression of the 5.4-kb mRNA species in the MH1C1 cell line; this stimulation is carbohydrate-independent. The hepatoma cell lines mimic, therefore, the effect of cyclic AMP on the inhibition in vivo of the expression of genes encoding glycolytic or lipogenic enzymes [Vaulont, S. et al. (1984) Biochem. Biophys. Res. Commun. 125, 135-147]. In contrast, the effect of carbohydrates [Munnich, A. et al. (1984) J. Biol. Chem. 259, 10228-10231] is undetectable. The insulin sensitivity of the liver-specific genes is conserved for the 5.4-kb mRNA species only, especially in the MH1C1 cell line, but not for the other investigated mRNAs, which seems to reflect a fundamental difference in the in vivo effect of insulin on these genes. Finally, S1 nuclease mapping of the start-site of pyruvate kinase L gene transcription shows that the normal site used in vivo is also used in the Fao and Faza lines while, in the MH1C1 line, it coexists with multiple aberrant upstream initiation sites.