An Alternatively Processed mRNA Specific for y-Glutamyl Transpeptidase in Human Tissues*

Human gamma-glutamyl transpeptidase (GGT)1 is composed of two subunits derived from a single precursor (Nash, B., and Tate, S.S. (1984) J. Biol. Chem. 259, 678-685; Finidori, J., Laperche, Y., Tsapis, R., Barouki, R., Guellaën, G., and Hanoune, J. (1984) J. Biol. Chem. 259, 4687-4690) consisting of 569 amino acids (Laperche, Y., Bulle, F., Aissani, T., Chobert, M.N., Aggerbeck, M., Hanoune, J., and Guellaën, G. (1986) Proc Natl. Acad. Sci. U.S.A. 83, 937-941). In the present study we report the cloning of an altered form of this precursor from human liver. We have isolated two clones, one 2,632 base pairs (bp) long from a fetal liver cDNA library and one 926 bp long from an adult liver cDNA library, each containing a 22-bp insertion that introduces a premature stop codon and shortens the open reading frame to 1,098 bp when compared with known human cDNA sequences specific for GGT. Sequence analysis of a human genomic GGT clone shows that this insertion of 22 bp is generated by a splicing event involving an alternative 3'-acceptor site. By polymerase chain reaction experiments we demonstrate that the alternatively spliced mRNA is present in polysomes from the microsomal fraction of a human hepatoma cell line (Hep G2) and thus could encode an altered GGT molecule of 39,300 Da (366 amino acids) encompassing most of the heavy subunit which is normally 41,500 Da (380 amino acids). The altered mRNA is detected in various human tissues including liver, kidney, brain, intestine, stomach, placenta, and mammary gland. This report is the first demonstration of an alternative primary sequence in the mRNA coding for GGT, a finding that could be related to the presence of some inactive forms of GGT detected in human tissues.

In the present study we report the cloning of an altered form of this precursor from human liver. We have isolated two clones, one 2,632 base pairs (bp) long from a fetal liver cDNA library and one 926 bp long from an adult liver cDNA library, each containing a 22-bp insertion that introduces a premature stop codon and shortens the open reading frame to 1,098 bp when compared with known human cDNA sequences specific for GGT. that anchors the enzyme in the membrane. Both subunits are synthesized from a single mRNA encoding a 64-kDa peptide (4,5). GGT is mainly distributed in tissues exhibiting absorptive and secretory processes (1). The highest activity is found in kidney and intestinal cells (l-3). In normal adult liver, the GGT activity is low and mainly located in bile ducts and canalicular membranes of the hepatocytes (6). In this organ, increases in activity are observed under various physiological and pathological conditions (7,8). Several experiments have demonstrated that these increases are often associated with structural changes in the sugar chains of the enzyme, as evidenced by a variation in the pattern of GGT isoforms in serum (9, 10). Antibodies, specific for GGT isoforms from a human primary hepatoma, have been shown to be useful for the diagnosis of some neoplastic diseases (11). In human serum (12) or in rat liver (13) such antibodies recognize reactive species which are not directly correlated with GGT activity, suggesting the presence of altered forms of GGT.
Recently we have cloned the cDNA encoding the rat kidney GGT (14), and using this cDNA we have demonstrated that several genes or pseudogenes are present in the human genome (15) on chromosome 22 (16). According to the recent cloning of GGT mRNA from human placenta (17), fetal liver (18), or the human hepatoma cell line Hep G2 (19), this multigene family codes for the same GGT precursor.
In the present report we describe the isolation and sequence of two cDNAs from human adult and fetal liver libraries. These cDNAs have a 22-bp insertion as compared with other known sequences (17)(18)(19) There is a complete homology between this GGT cDNA isolated from human adult liver and the 3'-part of the human fetal liver cDNA. The 3'-untranslated region is 9 bases longer in the adult liver cDNA than in the fetal liver cDNA.
Comparison with Other Known Human cDNA Sequences Specific for GGT-The sequences, for the two cDNAs we obtained, were compared with sequences coding for GGT isolated from human placenta (17), human fetal liver (18), and Hep G2 hepatoma cells (19) (Fig. 2). The 5'-noncoding sequences are organized differently.
They are identical from +1 to -88 except for single base pair differences at positions -23 and -84 in the Hep G2 cDNA relative to the other sequences. A region from -89 to -243 in the human fetal liver cDNA (present study) is observed in the Hep G2 cDNA from position -207 to -361. At position -139 in our clone, and is exactly homologous to the human fetal liver sequence except for nine additional bases at the 3'-end of this sequence which are shown in Fig. 2 there is a G instead of an A in the homologous Hep G2 position. A part of this sequence -174 to -243 from human fetal liver is found in the human placenta sequence at -185 to -253. However, no other homologies were detected between our human fetal liver cDNA and the other 5'-sequences so far analyzed. It should be noted that some other homologous regions exist between Hep G2 cDNA and placenta cDNA specific for GGT in the 5'-part (Fig. 2).
The coding sequences of all the clones so far analyzed are identical except for two differences. The first difference is minor; a T instead of a C is found at position 815 in our fetal liver cDNA clone changing in alanine codon to a valine codon. The second difference, and by far more important, is a 22-bp insertion found in both our fetal and adult liver cDNAs beginning at positions 1020 and 85, respectively. The insertion modifies the reading frame and introduces a stop codon at position 1099 in the fetal liver cDNA clone and at a similar position in the adult liver cDNA clone.
The 3'-untranslated region starts at the same position for all the cDNAs. The main difference is a microheterogeneity in the length of the sequence preceding the poly(A) tail. There are three nucleotide differences between placenta cDNA specific for GGT and the other sequences. Since the insertion of 22 bp in the coding sequence necessarily alters the product of this mRNA, we further focused our work on this additional sequence.
Analysis of the 2Bbp Insertion at the Gene Level-Four genomic clones corresponding to the different subclasses of human GGT genes (15) were digested by BamHI and hybrid-ized with a labeled oligonucleotide corresponding to the 22bp insertion found in fetal and adult liver cDNAs. The result of the Southern blot is shown in Fig. 3. Only the clones corresponding to the subclass F15 and F30 hybridized with this oligonucleotide.
The faint band on clone F19 disappeared at slightly higher stringency (data not shown).
The human liver cDNA sequences were compared with the sequence of one of these genomic clones (F15) (Fig. 4). This analysis revealed that in the gene the 22-bp domain is located at the 3'-part of an intron. This sequence is bordered at each end by a 3'-acceptor consensus sequence necessary for a correct splicing (30). Therefore, the cDNAs that we characterized reflect the fact that, during maturation of the corresponding mRNAs, the internal 3'-acceptor site was used.
Detection of the 22-Base Insertion in Polysomal RNA from Hep G2 Cells and mRNA from Human Tissues-Using the oligonucleotides corresponding to the 22-bp insertion and to the control sequence as probe, we were unable to detect any GGT-specific sequence in the human liver mRNAs tested by Northern blot analysis. In order to increase the chances of detecting mRNAs containing the 22-base insertion, we used the polymerase chain reaction technique. Two oligonucleotides, designated oligo-A and oligo-B, were selected as described in Fig. 5A and used as primers for the amplification procedure.
Amplified products using RNAs from Hep G2 polysomes and supernatant as starting materials were subjected to Southern blot analysis using the GGT-specific se- A, the relative position is represented, with respect to the first ATG, of the four different oligonucleotides, A, 22, control, and B. Their respective sequences are: A, CAAGTTTGTGGATGT-GACTGAG; 22, AGCTCTGGGGTCTCGGC; control, CCTCCGA-GTTCTTCGCTGCC; B, TCAGAGATCTGGGCCCG. The mRNA from polysome preparation (s, supernatant; p, polysomes) and various tissues (aL, adult liver; fL, fetal liver; jK, fetal kidney; H, Hep G2; UK, adult kidney; B, brain; I intestine; S, stomach; P, placenta; M, mammary gland) were amplified using oligo-A and -B as described under "Experimental Procedures" and blotted. B, the same series of blots was successively hybridized with the control probe (upperpanel) or the oligo-22 (lower panel). The figure is a composite of different autoradiograms obtained from different blots hybridized simultaneously and processed under the same conditions. The size of both bands was estimated using DNA marker and from amplified cDNA of Hep G2 GGT clone (I ) or human adult liver GGT clone (2) containing the 22-bp insertion.
quences oligo-control and oligo-22, defined in Fig. 5A. If only the normal GGT mRNA is present in the polysomal fraction, we would expect only one band to be detected with a size corresponding to 81 bp when the oligo-control is used as a probe. If the mRNA with the 22-base insertion is also present in the polysomal fraction, a second band should be detected with a size corresponding to 103 bp when the oligo is used as a probe. As can be seen in Fig. 5B Fig. 5B (lower panel, lane p). Contamination of the polysomal fraction by the supernatant is highly unlikely since (i) no bands were detected in the supernatant fraction ( Fig.   5B, upper panel, lane s) and (ii) the measurement of lactate dehydrogenase in the supernatant amounted to 94% of the activity measured in the homogenized cells, whereas only 0.4% was found in the microsomes. Our results strongly support the idea that the 22-base insert-containing mRNA is translated in Hep G2 cells.
We have looked for both types of GGT mRNAs in a number of different human tissues using the same amplification procedures and probes. As evidenced by the presence of both the 81-and 103-bp band when oligo-control is used as a probe, and the 103-bp band when oligo-22 is used as a probe, both types of mRNAs are present in fetal liver, kidney, brain, intestine, stomach, placenta, mammary gland as well as in Hep G2 cells (Fig. 5B). The 103-bp band is also detected in adult liver following a longer exposure (data not shown). DISCUSSION We report here the cloning and characterization of two GGT-related cDNAs from human adult and fetal livers. These cDNAs exhibit some differences when compared with the other published cDNA sequences from human placenta (17), human fetal liver (18),or Hep G2 cells (19). The most striking difference is the presence of a 22-bp insertion in the GGT clones we have isolated. The insertion modifies the reading frame and alters the predicted translation product. Analysis of the sequence of the fetal liver cDNA reveals three large open reading frames starting with a methionine.
Only two of these are surrounded by a consensus sequence usually observed for correct initiation of translation (31). We focused our attention on the first one since it corresponds to the ATG described for the heavy subunit of GGT. In this reading frame, the 22-bp insertion introduces a stop codon at position 1099. The putative truncated protein of 39,287 Da would consist mainly of the GGT heavy subunit differing only in the last 26 amino acids of the carboxyl-terminal portion. This protein would be devoid of any GGT activity since the catalytic activity is associated with the light subunit (32). The second open reading frame, out of phase with the first one, starts at position 1045 and would encode a protein of 24,108 Da (225 amino acids). This protein would encompass the complete sequence of the small GGT subunit since the NH,-terminal part of the light chain of GGT (determined from human kidney (33)) corresponds to amino acid 37 on this protein (see Fig. 1). Nevertheless, although the initiator ATG is in a favorable context for translation (31) we have no evidence that this second open reading frame is utilized in uiuo.
The translation of the first reading frame is highly likely due to the following lines of evidence. First, we demonstrate that this mRNA is found in microsomal polysomes. Second, a doublet has been already described for the heavy subunit of human kidney GGT following immunoprecipitation (34). These bands correspond to a glycosylated form of the protein with a M, of 53,000 and 50,000, respectively. The light species has been attributed to proteolytic degradation of the normal GGT heavy subunit (53,000). According to the present work, one can hypothesize that the core proteins are generated from the normal mRNA (M, 41,239) and from the insert containing mRNA (M, 39,287). Of particular relevance to our study is the fact that immunohistochemistry and histoenzymology of human GGT in liver (13) and in serum (12) do not exactly correlate. It has been proposed that either an inhibitor decreases the GGT activity (12) or that an altered form of GGT exists in human liver or in serum.
Altered proteins have already been observed in other sys-terns such as in the case of human apolipoprotein B (35). A premature stop codon generates a protein of 250 kDa instead of the native form of 512 kDa. For GGT, a high molecular weight antigen with no activity has been observed in the rat (36). It has a molecular mass between 85 and 95 kDa and could not result from a truncated protein if one excludes an aggregation process. Such an aggregation has been observed for the purified small subunit of the rat kidney GGT (37).
We have no explanation for the fact that we have preferentially cloned cDNAs containing an extra 22 base pairs in the coding region. Although it has been shown in several cases that nonsense mutations correlate with a decrease in the steady state level of mRNA (38,39), the aberrant GGT mRNA might be as abundant as the normal one. This means that the detected levels of mRNA might not be reflective of true protein (enzymatic) levels.
On the basis of sequencing results of a genomic clone, mRNAs containing the extra 22 bases could result from the use of an alternative 3'-acceptor site during splicing (Fig. 4) as already described for other systems (40,41). In our case the on/off splicing regulation mechanism is not tissue-specific as demonstrated by polymerase chain reaction amplification. Both forms of mRNA are detected in the hepatoma cell line Hep G2, kidney, liver, brain, mammary gland, intestine, and stomach.
The 5'-noncoding region of our human fetal cDNA clone is unusually long (744 bp). In fact, most of the leader sequences on vertebrate mRNAs fall in the range of 20-100 nucleotides (31) if one excepts proto-oncogenes.
Long leader sequences are not incompatible with efficient translation, provided that any upstream open reading frame initiated on an ATG in a favorable context for initiation does not overlap with the main open reading frame. In the 5'-untranslated region of the fetal liver GGT mRNA, two ATG (-709, -316) are susceptible to open two short reading frames (18 and 16 amino acids) which terminate at positions -655 and -268, respectively, thus without any sparing effect on GGT mRNA translation.
The 5'-noncoding region characterized in the present study differs from those of the human placenta (17) and Hep G2 cells (19). In this region, homologous sequences are detected among the different cDNAs but at different positions with respect to the ATG. In humans we have characterized a multigene family for GGT (15), and it is possible that the different mRNAs are encoded by different genes. Nevertheless, such an organization in the 5'-noncoding region has already been observed in other systems where only one gene is active (42). Since it is not known whether these clones are full-length, there is insufficient information to conclude whether there might be a unique promoter or multiple promoters. A similar observation has been made for the unique rat GGT gene encoding GGT mRNAs varying in their 5'untranslated region (43). Concerning the 3'-noncoding region, there are only differences in length between the polyadenylation signal AATAAA and the poly(A) tail. Such a microheterogeneity has been described for other mRNAs (44) and has not yet been linked to any regulatory process.
Our results demonstrate a possible complex regulation of GGT genes in humans. Until now, only one modification in the processing of GGT has been described for Hep G2 cells (45) in which the precursor is not cleaved into two subunits. This is not due to an alteration in the primary sequence (19) but rather in a modification of the processing factors. We are now exploring the possibility that the human multigene family specific for GGT encodes altered primary polypeptide structures under different physiopathological conditions.