The Presence of Intervening Sequences in the a-Fetoprotein Gene of the Mouse*

‘lkli. ~JOURNAL OF BIOI.OCICAL Vol. 254. No. 15. Issue of August Prmfed m U.S.A The Presence Mouse* CHEMISTRY 10, pp. 7393-7399, of Intervening Sequences in the a-Fetoprotein (Received Shirley Ingram From M. Tilghman,$ the Fels Research Dimitris Institute, Kioussis, Temple Michael University School is a major constituent of fetal species which have been examined in the mouse, AFP’ is synthesized sac and liver and secreted After birth, the amount serum in all (l-4). During by the yolk into the blood and amniotic of AFP in serum decreases of u-fetoprotein synthesis in the adult Garcia Philadelphia, Ruiz, Pennsylvania March and Robert S. under a select number of pathological conditions: following liver injury (6, 7), in hepatocarcinomas (1, 2), and in testicular teratocarcinomas (8, 9). In addition, a number of congenital birth defects, including congenital nephrosis, severe Rh-he- molytic disease, and several open neural tube defects are associated with abnormally elevated concentrations of AFP in the maternal and fetal circulation and amniotic fluid (10-12). These striking and reproducible increases in the production of AFP in fetal abnormalities, in liver regenerative diseases, and in carcinomas of the liver and germ cells in humans has generated interest in the mechanism of control of the expres- sion of AFP. Recent studies by Tamaoki and his co-workers (13,14) have used cell-free translation assays to demonstrate that the rate of synthesis of a-fetoprotein in fetal and neonatal mouse liver is related to the concentration of translatable AFP mRNA. Taking this analysis one step further, Innis and Miller (15) reported the isolation of rat AFP mRNA. Complementary DNA to the AFP mRNA was used in hybridization kinetic experiments to show that the actual amount of AFP mRNA in adult liver mRNA was less than 1% of that in hepatoma cells. These studies imply that the control of AFP synthesis occurs at the level of mRNA transcription or processing, or both. Distinguishing between these possibilities will require large amounts of specific hybridization probes to the a-fetoprotein gene, as well as intimate knowledge about its structural or- ganization in the genome. Toward this end, we have con- structed chimeric plasmids containing portions of the mouse AFP mRNA sequence, and used them to identify three EcoRI fragments of mouse genomic DNA which encode in a discon- tinuous manner the single copy AFP gene. fluid. EXPERIMENTAL markedly in most strains of mice to a basal level which represents less than 0.01% of that in fetal serum (3, 5). This decline has recently been shown by Olsson et al. (5) to be inherited as an autosomal recessive trait, which they termed regulation of a- fetoprotein (ran. Reinitiation of Medicine, J. Predes for publication, occurs * This work is supported by Grants CA 23572 to S.M.T. from the National Institutes of Health and CA 12227 to the Fels Research Institute. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “aduertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. $ Recipient of an American Cancer Society Junior Faculty Award. 0 Trainee of the Medical Scientist Training Program, National Institutes of Health Grant GM 07170. ’ Abbreviations used are AFP, a-fetoprotein; dsDNA, double- stranded DNA; SSC, 0.15 M NaCl, 0.015 M sodium citrate; bp, base pairs; kb, kilobase pairs; Hepes, N-(2-hydroxyethyl)-l-piperazine-eth- anesulfonic acid, SDS, sodium dodecyl sulfate. PROCEDURES Purification of a-Fetoprotein-Amniotic fluid, pooled from 16- to Id-day mouse fetuses, was subjected to 35 to 70% ammonium sulfate precipitation. The 70% pellet was dialyzed against 50 mu Tris.HCl (pH 7.5) and electrophoresed through a nondenaturing 7% polyacryl- amide gel (16). The AFP band was identified by a comparison to fetal and adult serum protein markers, extracted and concentrated by ethanol precipitation (16). This material was used to raise antisera to AFP by bimonthly injections into a rabbit. The rabbit antisera were regularly tested for anti-AFP activity by Ouchterlony analysis (17) against purified AFP, amniotic fluid, and adult mouse serum. Preparation of RNA-Yolk sac and livers, obtained from 16- to 18-day-old Swiss mouse fetuses were frozen in liquid Na. Poly(A) RNA was prepared by phenol extraction of total cellular RNA (18) followed by oligo(dT)-cellulose chromatography (19). The RNA was fractionated on an 8 to 25% sucrose gradient in 2 mM EDTA, 100 mM NaCl, 10 mM Hepes (pH 7.6), which was centrifuged at 29,000 rpm for 22 h in a Beckman SW40 rotor. Individual fractions were precipitated by the addition of 2 volumes of ethanol, resuspended in Hi0 and tested for AFP mRNA activity using a cell-free synthesizing system derived from wheat germ (20, 21). After treatment with 200 pg/ml of Downloaded from http://www.jbc.org/ at UCLA-Louise Darling Biomed. Lib. on April 25, 2017 Messenger RNA for the oncofetal protein cu-fetopro- tein was isolated from yolk sacs of 16- to H-day-old mouse embryos. This RNA was used as a template to synthesize full-length double-stranded DNA containing predominantly a-fetoprotein coding sequences. Chi- merit plasmids containing the double-stranded DNA were constructed in two ways. First, molecular linkers containing the recognition site for the restriction en- donuclease HindI were ligated to the blunted ends of the DNA, and following cleavage with HindIII, the re- combinant molecules were ligated to the single Hind111 site in the plasmid vehicle pBR322. Second, internal Pst I fragments of the double-stranded DNA were ligated to the single Pst I site of pBR322. Following transfor- mation of the host Escherichiu coli x1776, colonies con- taining those chimeric plasmids which had incorpo- rated n-fetoprotein sequences were identified by selec- tive hybridization and arrest translation procedures. Two cloned sequences derived from the 5’ and 3’ regions of the mRNA were used as hybridization probes to detect the presence of the a-fetoprotein gene within a two-dimensional fingerprint of mouse genomic DNA cleaved with EcoRI, an enzyme which does not cleave within the messenger RNA sequence. The detection of 3 EcoRI fragments demonstrates that the single cu-fe- toprotein gene must be represented discontinuously in the mouse genome, with at least two intervening se- quences. a-fetoprotein mammalian development B. Gorin,@ Gene of the


Messenger
RNA for the oncofetal protein cu-fetoprotein was isolated from yolk sacs of 16-to H-day-old mouse embryos.
This  (l-4). During development in the mouse, AFP' is synthesized by the yolk sac and liver and secreted into the blood and amniotic fluid.
After birth, the amount of AFP in serum decreases markedly in most strains of mice to a basal level which represents less than 0.01% of that in fetal serum (3,5). This decline has recently been shown by Olsson et al. (5) to be inherited as an autosomal recessive trait, which they termed regulation of afetoprotein (ran. under a select number of pathological conditions: following liver injury (6,7), in hepatocarcinomas (1,2), and in testicular teratocarcinomas (8,9). In addition, a number of congenital birth defects, including congenital nephrosis, severe Rh-hemolytic disease, and several open neural tube defects are associated with abnormally elevated concentrations of AFP in the maternal and fetal circulation and amniotic fluid (10)(11)(12). These striking and reproducible increases in the production of AFP in fetal abnormalities, in liver regenerative diseases, and in carcinomas of the liver and germ cells in humans has generated interest in the mechanism of control of the expression of AFP.
Recent studies by Tamaoki and his co-workers (13,14) have used cell-free translation assays to demonstrate that the rate of synthesis of a-fetoprotein in fetal and neonatal mouse liver is related to the concentration of translatable AFP mRNA. Taking this analysis one step further, Innis and Miller (15) reported the isolation of rat AFP mRNA. Complementary DNA to the AFP mRNA was used in hybridization kinetic experiments to show that the actual amount of AFP mRNA in adult liver mRNA was less than 1% of that in hepatoma cells. These studies imply that the control of AFP synthesis occurs at the level of mRNA transcription or processing, or both.
Distinguishing between these possibilities will require large amounts of specific hybridization probes to the a-fetoprotein gene, as well as intimate knowledge about its structural organization in the genome. Toward this end, we have constructed chimeric plasmids containing portions of the mouse AFP mRNA sequence, and used them to identify three EcoRI fragments of mouse genomic DNA which encode in a discontinuous manner the single copy AFP gene. The cells were centrifuged, washed in 50 ml of 100 mM NaCl, 5 mM Tris.HCl (pH 7.5), and resuspended in 50 ml of 5 mM Tris. HCl (pH 7.5), 10 InM MgC12, 70 mM CaC12. After 20 min at room temperature, the cells were concentrated by centrifugation and resuspended in 2 ml of the same buffer. Fifty microliters of the ligated mixtures, which had been dialyzed briefly against 100 mu NaCI, 5 mM Tris.HCl (pH 7.5) were mixed with 0.2 ml of cells for 60 min at 4°C heated at 42'C for 1 min, chilled on ice for 10 min, and grown in 3 ml of LB broth supplemented with diaminopimelic acid and thymidine for 40  The filters were hybridized at 68°C in a volume of 3 ml in the presence of 0.5 to 2.0 x 10" cpm of cDNA, 4 X Denhardt's solution, 6 x SSC, 0.5% SDS, 100 pg/ml of E. coli DNA, and 50 pg/ml of poly (A) and yeast tRNA's.
After washing the filters at 50°C in 0.1 x SSC and 0.05% SDS for several h, the ftiters were autoradiographed.
Hybrid Arrest Translation-Colonies were grown in brain heart infusion media supplemented with diaminopimelic acid and thymidine to an A,, of 1.2 and amplified in the presence of 50 pg/ml of chloramphenicol.
The plasmid DNA was extracted by a modification of the procedure of Meagher et al. (28)  were run, and the DNA was transferred to nitrocellulose filters (37). The filters were pretreated and hybridized to labeled fragments of chimeric plasmids as described in detail above, using 5 to 10 x 10" cpm of DNA per filter in 20 ml of hybridization buffer. The falters were washed as above and exposed to Kodak XR-1 film in the presence of DuPont Lightning-plus intensifiers.

RESULTS
Isolation of a-Fetoprotein mRNA---In order to identify a suitable source of murine AFP mRNA, total poly(A)-containing RNA was extracted separately from 16-to M-day-old fetal livers and yolk sacs. These were subjected to sucrose gradient centrifugation, and individual fractions were tested for AFP template activity using a wheat germ in vitro protein synthesis system. The products of translation were analyzed by SDSpolyacrylamide gel electrophoresis, and AFP and albumin were identified using monospecific antibodies to each protein.
The AFP mRNA template activity migrated as an 18 S RNA species, which agrees with the size determinations of previous investigators for rat (15) and mouse (14) AFP mRNA, and with the molecular weight of the protein being 70,000 (38). The products of the 18 S mRNA translations are shown in Fig. 1. In Lane 1, two predominant bands which co-migrate with mouse albumin (upper dot) and AFP (lower dot) were synthesized by 18 S fetal liver RNA. The lower band is completely immunoprecipitable by anti-AFP sera (Fig. 1 S yolk sac mRNA were analyzed (Fig. 1, Lane 3), there was a single predominant protein band which was completely immunoprecipitable by anti-AFP sera (Lane 4) and could be competed for by authentic AFP (data not shown). The striking predominance of a-fetoprotein, the only major protein encoded by the 18 S yolk sac RNA, supports the previous observations of Wilson and Zimmerman (39), who concluded that AFP represented 40 to 60% of the protein synthesis of yolk sac, while only 20% of that in fetal liver. As a control, poly(A) RNA from adult mouse liver which should contain no AFP template activity was used to direct protein synthesis in the wheat germ system. As shown in Fig. 1, Lanes 5 and 6, no labeled protein which could be immunoprecipitated with AFP antibody was synthesized.
The relative migration of albumin and a-fetoprotein in the gel buffer system used ( Fig. 1) is contrary to their molecular weights of 68,000 and 70,000, respectively (38). By using a different buffer system (40), we could reverse the order of migration of both the in vitro products and the mature proteins, such that albumin exhibited a faster migration. We conclude that the aberrant migration observed is a result of the primary amino acid sequences of these proteins and not to any post-translational modifiiation.
Insertion of a-Fetoprotein mRNA Sequences into a Plasmid Vehicle-The 18 S poly(A) mRNA preparation from yolk sac should contain greater than 50% AFP mRNA, ensuring that at least one in two hybrid plasmids contain AFP se-  . 2) demonstrated the presence of a predominant band migrating as 2,150 base pairs of DNA. This is the size one would predict for a full-length transcript of an 18 S mRNA.
To choose a suitable strategy for cloning the a-fetoprotein cDNA, the 2,150 bp dsDNA was cleaved with Pst I, HindIII, and BamHI (Fig. 2). These enzymes were chosen because they each cleave the genome of the plasmid vector pBR322

Intervening
Sequences in the a-Fetoprotein Gene once, in either the ampicillin resistance gene (Pst I) or the tetracycline resistance gene (BarnHI and HindIII) (41), and thus insertion into any one of these sites allows for discrimination between colonies that contain recombinant plasmids and those that contain parental plasmids. All three enzymes cut the AFP sequence, generating unique fragments detectable by autoradiography.
The presence of clearly identified bands which correspond to a single species supports the previous estimate by translation that AFP is the only predominant species in the 18 S yolk sac mRNA. Both Hind111 and RamHI cleave the sequence once (Fig. 2, Lanes 6 and 7) and Pst I makes 3 cuts (Lane 9). In contrast, EcoRI (Lane 5) does not cleave within the AFP sequence. By performing a series of double digests such as the BarnHI-Hind111 cleavage illustrated in Fig. 2, Lane 8, it was possible to construct a restriction map of the cDNA (Fig. 3).
The orientation of the restriction map relative to the 5' and 3' ends of the a-fetoprotein mRNA, illustrated in Fig. 3, was determined using ds cDNA which had not been treated with Sl nuclease. Following cleavage of the dsDNA with Hha I, which recognizes one asymmetric site in the cDNA, the fragments were denatured by glyoxal and electrophoresed in a 1.5% agarose gel (34). The presence of a 700-bp denatured fragment demonstrates that Hha I must cleave closer to the open 3'-derived end of the dsDNA (compare Fig. 4, Lanes 2 and 3). On the other hand, denaturation of BamHI fragments, derived from a cleavage at the other end of the dsDNA, produces a 1.4-kb denatured fragment (Fig. 4, Lanes 5 and 6). The fragments representing the 5' end of the mRNA are not clearly resolved in this system, due to the difficulty in completely denaturing a snap-back sequence. two internal fragments, 520 and 900 bp in length (Fig. 3). These fragments were ligated directly to pBR322, which had been digested with Pst I and treated with bacterial alkaline phosphatase (25) to reduce the number of parental colonies. After the ligated mixture was used to transform CaCl&reated x1776 cells, hybrid colonies were identified and screened by the method of Grunstein and Hogness (26). Replica filters were hybridized separately to ["'P]cDNA's prepared from 18 S mRNA's of yolk sac, l&day fetal liver and adult liver (Fig.  5). Thus, colonies which hybridized strongly to the cDNA from yolk sac and fetal liver mRNA's, but not to adult liver cDNA's, would be expected to contain a-fetoprotein sequences and were selected for analysis. The plasmids were isolated by the procedure of Meager et al. (28) and digested with Pst I. Those containing the 520-and 900-bp (or both) Pst I fragments were thus identified.
To clone those portions of the a-fetoprotein cDNA sequence which were outside the two Pst I internal fragments, we used the method of Ulh-ich et al. (25) which required attaching lobp molecular linkers to the ends of ds cDNA. Hind111 linkers were ligated to the 2,150-kb AFP cDNA and the resulting molecules cleaved with a vast excess of Hind111 to destroy excess linkers and create ligatable ends. In this way, the dsDNA would also be cleaved internally at the Hind111 site. The dsDNA fragments were ligated to pBR322 cut by Hind111 and transformants were generated and selected by hybridization. Several clones, all derived from the 5' end of the AFP mRNA were identified.
Characterization of pBR322. AFPl and pBR322. AFP2-Two recombinant plasmids, termed pBR322. AFPl and pBR322. AFP2, were selected for further characterization. yolk sac cDNA Cleavage of pBR322. AFPl with Hind111 and electrophoresis in an acrylamide gel yielded the parental 4.3-kb fragment and a 960-bp Hind111 insert (Fig. 6, Lane 2 Lanes 3 and 4), it was possible to map the insert to the 5' end of the a-fetoprotein dsDNA sequence, as shown in Fig. 3. Likewise, pBR322. AFP2 was shown by a series of digests with Pst I (Fig. 6, Lane 5) and Pst I plus Hind111 (Lane 6) to contain the 900-bp Pst I fragment of AFP dsDNA derived from the 3' end of AFP mRNA. The two inserts overlapped each other by approximately 200 bp, within a common Pst I-Hind111 fragment (Fig. 6, arrow).
As further evidence that the two cloned DNA's contained AFP sequences, the hybrid ,arrest procedure of Paterson et al. (30) was used. Total pBR322. AFPl DNA was hybridized to poly(A)-containing yolk sac RNA under conditions favoring RNA-DNA hybrids (30), and the mixture was translated in a wheat germ extract. The translation products from the control RNA are shown in Fig. 7 Intervening Sequences in the a-Fetoprotein Gene but not the 6.0 kb fragment. That the two probes recognized nonidentical fragments argues that the number of EcoRI fragments which hybridize to AFP sequences cannot be explained on the basis of multiple AFP gene copies. Rather, the differential hybridization pattern supports the conclusion that there is a single copy of the gene, represented in at least three EcoRI fragments. The absence of any EcoRI cleavage site in the entire AFP cDNA sequence (Fig. 2), further suggests that the two detectable EcoRI sites in the genomic sequence are contained within intervening sequences in the gene. The Southern blots also establish the orientation of the three fragments with respect to the AFP mRNA, with the 6.0-kb fragment representing the 5' end of the mRNA, the 3.0-kb fragment representing a portion of the mRNA which includes the region overlapped by the two probes, and the 5.2-kb fragment representing the 3' end of the mRNA. DISCUSSION We have constructed several chimeric plasmids which contain specific regions of the a-fetoprotein mRNA sequence. The plasmids were detected initially by differential hybridization of the cloned sequences, derived from yolk sac RNA, to fetal and adult liver RNAs. Second, restriction endonuclease mapping of the enriched cDNA, containing predominantly AFP cDNA, permitted rapid confirmation of the identity of the cloned sequences by detection of specific cleavage sites. Finally, the specific arrest in the translation of a-fetoprotein mRNA following hybridization of the plasmid DNA to total yolk sac RNA confirmed the identity of pBR322. AFPl and pBR322. AFP2.
These plasmids, when labeled and used as hybridization probes to visualize the a-fetoprotein gene within the complexity of total mouse genome DNA, were able to detect three separate EcoRI fragments, containing AFP mRNA sequences. The absence of an EcoRI site in the entire mRNA sequence, and the total size of the three fragments, 14.2 kb, argue that the a-fetoprotein gene must be represented discontinuously in the genome, with at least two intervening sequences at the two EcoRI sites. Thus, the AFP gene appears to be encoded in a manner analogous to the rabbit (42) and mouse (31, 42) /?-globin genes and the chick ovalbumin gene (43)(44)(45)(46)(47). These unique genes, along with tRNA genes from yeast (48,49), ribosomal genes in Drosophila (50)(51)(52), adenovirus 2 (53)(54)(55)(56), and SV40 genes (57,58) have been shown in the last year to be interrupted at one or more positions in the coding sequence by noncoding DNA, which is transcribed into RNA (59) and removed during RNA processing (60,61).
Two other important facts can be drawn from Fig. 8. First, the a-fetoprotein gene must be represented only once in the genome, as the two probes recognize different EcoRI fragments. Second, the order of the fragments, relative to the 5' and 3' ends of the mRNA is established as 6.0, 3.0, and 5.2 kb.
The hybridization probes used in Fig. 8 are missing approximately 150 bp of DNA at the 5' and 3' ends of the cDNA sequences. Therefore, we cannot know whether the entire AFP gene is contained within the three EcoRI fragments. In addition, this analysis would not detect EcoRI fragments which contain intervening sequence DNA alone, or less than 50 bp of coding sequence. To counteract these uncertainties, and to define in greater detail the organization of the AFP gene, we have recently cloned the three EcoRI fragments" and intend to use them to select for DNA fragments derived from a mouse embryonic DNA library, by the methods recently described by Maniatis et al. (62). In this way, an accurate map ,' M. B. Gorin and S. M. Tilghman, manuscript in preparation. of the entire region containing and surrounding the mouse AFP gene can be drawn.