Variant Forms of (cid:1) -Fetoprotein Transcripts Expressed in Human Hematopoietic Progenitors IMPLICATIONS FOR THEIR DEVELOPMENTAL POTENTIAL TOWARDS ENDODERM*

Hematopoietic stem cells have been identified as multipotent cells that give rise to all adult hematopoietic lineages. Although the hematopoietic lineage is derived from the mesodermal germ layer in the embryo, recent data suggest that bone marrow cells with an antigenic profile consistent with that of hematopoietic stem cells can also differentiate to cell types of the endodermal lineages, such as hepatocytes. However, the molecular mechanisms associated with these events are entirely unknown. For decades, (cid:1) -fetoprotein (AFP) has been used as a differentiation marker for endodermal cells, because it was thought that the transcription of AFP mRNA is tightly regulated in a developmental and tis-sue-specific process. In this report we describe two new variant forms of AFP transcripts in human hematopoietic progenitors that are not expressed in mature cells. The variant AFP (vAFP) cDNA sequences isolated from a multipotent hematopoietic cell line, K562, revealed that the vAFP differed from the authentic transcript, consisting of 15 exons, by replacing exon 1 of AFP with one or two exons located in the 5 (cid:1) -untranslated region of the AFP gene. In addition to the K562 cell line, vAFP transcripts were detected in normal bone marrow, thymus,

Stem cells, pluripotent cells capable of extensive growth without losing their multipotentiality, are becoming the centerpieces of new strategies for cell therapies, gene therapies, and tissue engineering, all contributing to a new era in medicine (1-3). The molecular machinery of self-renewal and differentiation is one of most intriguing subjects of stem cell biology. Analyses of gene expression patterns using subtracted cDNA library or cDNA microarray techniques have revealed several candidates of new or known transcripts that are expressed exclusively in stem cell populations, but not in terminally differentiated cells (4,5). These studies provide enormous information facilitating the development of approaches for understanding stem cells. However, the applied techniques have an intrinsic weakness in that they cannot detect the stem/progenitor cell-specific isoforms of transcripts having relatively long domains in common between the isoforms. Therefore, an alternative strategy is to analyze individual genes with expression that is regulated in developmentally specific ways in particular cell lineages.
Cell differentiation in the developing embryo is regulated by extrinsic inductive signals and an intrinsic programmed genetic code. Differentiation into the three germ layers (ectoderm, endoderm, and mesoderm) from primitive ectoderm (epiblast) is a crucial step during development and thought to be an irreversible process. This leads to the development of unique cell types from each germ layer, e.g. epidermal or neuronal cells from ectoderm; epithelial cells in internal organs or digestive tract from endoderm; and hematopoietic and mesenchymal cells from mesoderm. Recent studies (6,7) of cell transplantation, however, suggest that somatic stem or progenitor cells from adult tissues are able to generate cells with fates different from those normally associated with the germ layer that gave rise to the stem cells. A typical example of so-called "transdifferentiation" is that of murine bone marrow-derived ckit high Thy low Lin Ϫ Sca-1 ϩ cells giving rise to mature hepatocytes (8). This antigenic profile is used to define hematopoietic stem cells. Although reports about these examples of possible transdifferentiation in mammals are accumulating (9), it remains to be determined whether the studies observed transdifferentiation or identified more primitive progenitors than hematopoietic stem cells in bone marrow. In addition, recent reports proposed that spontaneous cell fusion between transplanted donor progenitors and recipient cells might cause misinterpretation in previous studies reported as possible transdifferentiation (10,11). In any event, biochemical analyses are necessary for understanding multipotency of the cells not only in those experimental setting but in normal physiological environment. * This work was supported by Grant RO1-DK52851 from the National Institutes of Health and grants from Incara Cell Technologies (formerly Renaissance Cell Technologies), Research Triangle Park, NC. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM  Adult liver parenchymal cells consist of hepatocytes and biliary epithelial cells. They are derived from common precursors, hepatoblasts, during embryonic development (12). The bipotent progenitors originate from multipotent endodermal cells in the foregut by inductive signals from the septum transversum surrounding the outgrowth of the endoderm (13). Although it is not known whether the bone marrow-derived cells can differentiate into hepatoblasts or can convert directly to hepatocytes, the cells possessing potential for hepatic differentiation should express some early endodermal markers before full differentiation into hepatocytes.
One such marker is ␣-fetoprotein (AFP), 1 which is a major serum protein produced primarily by visceral endoderm of the yolk sac and by hepatoblasts as well as more differentiated fetal hepatic cells (14,15). AFP is one of the earliest markers for endodermal differentiation; the transcriptional expression starts with the appearance of visceral endoderm in the early embryo and is regulated tightly by developmental and tissue-specific cues (14 -16). Therefore, in most studies that assess endodermal differentiation of human embryonic stem cells or embryonic germ cells, the expression of AFP transcripts or the products have been used as a marker of endoderm (17)(18)(19).
In this study, we have identified two variant forms of human AFP transcripts from a multipotent hematopoietic cell line, K562. The cDNA sequences revealed that the differences in the variant AFP (vAFP) mRNAs from that of the authentic transcript, which consists of 15 exons, is the presence of one or two unique exons, named exon A and exon B, replacing exon 1 of AFP. The variant forms were detected in normal CD34 ϩ Lin Ϫ hematopoietic progenitor cells but not in mature blood cells. The expression of the variant AFP transcripts suggests that hematopoietic progenitors are in a state that permits them to express certain types of transcripts that have been considered unique to the endoderm.

EXPERIMENTAL PROCEDURES
Cell Culture-The human hepatoblastoma cell line, HepG2, was maintained in Eagle's MEM supplemented with 1 mM sodium pyruvate, 2 mM L-glutamine, 50 units/ml penicillin, 50 g/ml streptomycin, 0.1 mM MEM non-essential amino acids solution, 5 g/ml insulin, and 10% fetal bovine serum. The human erythroleukemia cell line, K562 was maintained in DMEM/F12 supplemented with 2 mM Lglutamine, 50 units/ml penicillin, 50 g/ml streptomycin, 5 ϫ 10 Ϫ5 M 2-mercaptoethanol and 10% fetal bovine serum. HepG2 and K562 were obtained from the Tissue Culture Core Facility of the Cancer Center at the University of North Carolina at Chapel Hill. Normal human fetal lung fibroblasts, MRC5 (American Type Culture Collection), were cultured in the same supplemented DMEM/F12 in the presence of 10% horse serum (Sigma) and 1 ng/ml of bovine FGF-2 (Collaborative Biomedical Research).
Polymerase Chain Reaction-The primer sequences used for PCR are shown in Table I. For reverse transcriptase-PCR (RT-PCR), total RNAs were extracted from various cell lines and subjected to cDNA synthesis. Briefly, cDNAs were synthesized from 10 g of total RNAs by oligo(dT) priming and AMV reverse transcriptase (Seikagaku) in a reaction volume of 20 l at 42°C. PCR was performed in a total volume of 50 l consisting of 1 M each primer, 200 M each dNTP, 50 mM KCl, 1.5 mM MgCl 2 , 10 mM TrisHCl, pH 8.3, and 1.25 units of Amplitaq polymerase gold (PerkinElmer Life Sciences) with synthesized cDNA. Samples were heated to 94°C for 3 min followed by amplification for 30 cycles for AFP or 35 cycles for vAFP of 2 min at 94°C, 2 min 62°C, and 3 min at 72°C. After the last cycle, a final extension step was done at 72°C for 6 min. Then 5 l of each PCR reaction were analyzed by 1% agarose gel electrophoresis. For nested RT-PCR of AFP or vAFP in various human tissues, one-fourth of the cDNA synthesized from 10 g of total RNAs (CLONTECH) were used for the first PCR. One-hundredth of the volume of the first PCR product was applied for nested PCR using an internal primer set. The PCR consisted of 1 cycle of 3 min at 95°C; 30 cycles of 1.7 min at 95°C, 2 min at 62°C, 1.5 min at 72°C followed by a 4.5 min extension at 72°C. Then, 5 l of each sample were analyzed in 2% agarose gels. For RT-PCR of sorted cells from human umbilical cord blood cells, total RNA was isolated with RNeasy kit (Qiagen). cDNAs synthesized from total RNAs of sorted cells were normalized by the cell number. Nested PCR for vAFP was performed with cDNA equivalent to 48,000 cells. Single step RT-PCR for ␤-actin or CD34 was done with cDNA equivalent to that of 1000 cells. For analysis of the genomic sequence of exon A, genomic DNAs were isolated from HepG2, K562, and MRC5 with QIAamp Tissue Kit (Qiagen). Amplified PCR products, using g-ex-AS and g-ex-AA primers (Table I), were cloned into pCRII-TOPO (Invitrogen), and the DNA sequences were determined by the DNA sequencing core facility of the University of North Carolina at Chapel Hill (Dr. Laura Livingston, director).
Identification of the 5Ј-End of Variant AFP Transcripts-The rapid amplification of cDNA ends (RACE) method (14) was performed with modifications to identify the 5Ј-end of the vAFP transcripts. Briefly, with terminal transferase (Roche Molecular Biochemicals), a poly(dG) tail sequence was introduced into K562 cDNA derived from 10 g of total RNA. One-tenth volume of the cDNA was amplified for the first 5 cycles with 94°C for 1 min, 45°C for 1.5 min, and 72°C for 2.5 min, followed by 15 cycles with 94°C for 1 min, 60°C for 1.5 min, and 72°C for 2.5 min using SN-poly(C) primer and ex-2A, ex-3A, or ex-4A. PCR products purified with QIAquick PCR purification kit (Qiagen) were reamplified with SN primer and an internal AFP primer (ex-1A, ex-2A, or ex-3A) at 94°C for 1 min, 62°C for 1.5 min, and 72°C for 2.5 min for 30 cycles. The final PCR products were cloned into pCRII-TOPO and sequenced.
Umbilical Cord Blood (UCB) Cell Preparation and Fluorescenceactivated Cell Sorting-Human UCB was obtained from the Carolina Cord Blood Bank following institutional review board-approved informed consent. UCB was processed as described (27). Unfractionated nucleated UCB cells were prepared by deleting red blood cells with the use of Hespan (DuPont Pharma, Wilmington, CE) and hemolysis at 37°C in 0.17 M NH 4 Cl, 20 mM Tris-HCl, pH 7.2, and 200 M EDTA. Lineage-committed cells were removed from the unfractionated cells using the StemSep enrichment system (progenitor mixture, StemCell Technologies Inc, Vancouver, British Columbia). The recovered cells were termed Lin Ϫ UCB cells. Cells were purified using a FACStar Plus cell sorter (Becton Dickinson Immunocytometry Systems (BDIS), San Jose, CA). For cell sorting, the Lin Ϫ UCB cells were stained with a mixture of three phycoerythrin (PE) conjugated CD34-specific antibodies (clones Qbend10, Immu133 and Immu409, Beckman Coulter, Miami, FL). For all cell sorting, (7-aminoactinomycin D (Molecular Probes; 7AAD) was used to exclude dying cells. This was added to a final concentration of 10 g/ml during the antibody incubations. In some instances, the expression of CD38 was also monitored (clone HB7, BDIS).

Variant AFP mRNA Expressed in K562, a Human Erythroleukemia Cell
Line-The human AFP gene consists of 15 exons (Fig. 1A). The coding sequence is from exon 1 to exon 14 (20). To investigate the expression of possible variant forms of AFP mRNA in hematopoietic cells, two different portions of the AFP cDNA sequence were selected as target sequences of RT-PCR. The primer combination of ex-1S and ex-3A was used to amplify exon 1 containing the initiation methionine to exon 3, and the combination of ex-12S and ex-14A was used to amplify exon 12 to exon 14 containing the termination codon (Fig. 1B). The results of the PCR are shown in Fig. 2A. Both combinations of the primers resulted in intense amplification bands in the cDNA from HepG2, a hepatocellular carcinoma cell line (lanes 1 and 2). By contrast, cDNA from the erythroleukemia cell line, K562, showed an amplification band only for the C-terminal portion, identified by the ex-12S and ex-14A (lane 5) whereas there was no band by PCR of the primer combination of ex-1S and ex-3A (lane 4). This result suggests that K562 expresses a short form of authentic AFP transcript without the N terminus. In fact, the result of the PCR for the whole coding region of AFP using ex-1S and ex-14A primers showed that the single band of 1.8 kb (lane 3) was amplified from the HepG2 cDNA, whereas there was no detectable band in K562 (lane 6).
The AFP genes in rats, mice, and humans are organized similarly into 15 coding exons interrupted by 14 introns (20 -22). Although there has been no report about any variant form of the AFP mRNA in humans, several short forms of the AFP transcript have been characterized in rats (23,24). The rat vAFP lacks the first six or seven exons, but shares common 3Ј-sequences containing the seventh or eighth exon to the fifteenth exon, respectively (23). Detailed analysis of one variant form of rat showed that the rat vAFP used an additional exon, designated as exon V that is located in the seventh intron of the rat gene, as the first exon of the variant form. However, it turned out that human vAFP transcripts in K562 had different structures from the rat vAFP because RT-PCR of the primer combination exon 7 and ex-14A could detect the variant transcripts (data not show). Therefore, a series of 5Ј-primers from exon 2 to exon 6 were constructed (Fig. 1B) to see the difference between authentic and variant forms of AFP transcripts in K562. With each of these, ex-14A primer was used as a 3Јprimer. The results of the RT-PCR showed, intriguingly, that human variant AFP transcripts in hematopoietic cells share the entire coding exons, except for exon 1, in the authentic form that is expressed in HepG2 (Fig. 2B). Recently, it has been reported that an AFP isoform using a new exon located in intron 1 is expressed preferentially in the yolk sac and fetal liver in the mouse (25). However, no detectable amplified band in the PCR using an intron 1-specific 5Ј-primer (data not shown) suggested that the human vAFP transcripts in hematopoietic cells are novel isoforms of AFP.
Molecular Cloning of Variant AFP cDNA from K562-To determine the sequences of the 5Ј-end of the vAFP transcript in K562, the RACE method was applied as described under "Experimental Procedures." As a result, two types of variant transcripts were identified. The sequences are shown in Fig. 3A. Comparing the sequences of the variant transcripts to the GenBank TM data base, two regions of genomic sequence of the human AFP gene were identified. One type (type A) of the variant forms used an additional exon, designated as exon A, located about 5 kb upstream from exon 1 (Fig. 3B). The other type (type AB) used exon A in conjunction with another exon, designated as exon B, located about 1.6 kb upstream from exon  1, 4, and 7), ex-12S and ex-14A (lanes 2, 5, and 8), or ex-1S and ex-14S (lanes 3, 6, and 9) was used to amplify exon 1 to exon 3, exon 12 to exon 14, or exon 1 to exon 14 of AFP cDNA, respectively. B, series of 5Ј-primers from exon 1 to exon 6 (ex-1S to ex-6S) and ex-14A as the 3Ј-primer were used to detect AFP transcripts. Lanes 1, 3, 5, 7, 9, Table I. 1 (Fig. 3B). Among 19 clones analyzed in these experiments, 15 clones were type A; 3 clones were type AB; and one clone was the germ line transcript of intron 1. Although the genomic sequences containing exon A were reported originally from Watanabe et al. (26), some nucleotides in the exon A of K562 differed from the genomic sequence (Fig. 3A, underline). First, to determine whether this difference arose from artifacts associated with the anchored PCR technique, the genomic sequence of exon A in K562 cells was compared with the cloned cDNA sequence. This showed that the cDNA and genomic DNA of K562 cells were identical (data not shown). Next, to determine whether the genomic sequence of exon A of K562 is specific for the cell line, the genomic sequence of HepG2 and normal human fibroblasts, MRC5, was compared with that of K562. The results indicated no difference in the genomic sequence from all three sources, K562, HepG2, and MRC5 (data not shown). In addition, the genomic DNA sequence on the AFP locus from the data base of the human genome project (GenBank TM ; accession no. AP002825) corresponded with the sequences of K562. These results suggest that the DNA sequence of exon A cloned from K562 does not include artifacts from the PCR cloning.

Selective Expression of Variant Transcripts in K562-
The expression pattern of the variant and authentic transcripts of AFP was studied in K562 and HepG2. A sense primer of exon A for RT-PCR was designed. The combination of the exon A primer, ex-AS, and exon 14 primer, ex-14A, successfully detected the transcripts in K562, but not HepG2 (Fig. 3C). The amplified products from K562 were cloned and sequenced. As expected, the two types of variant AFP transcripts either using both exons A and B or exon A alone were followed by exon 2 to exon 14. The number of clones with a type A sequence was 7 of 12, while that of type AB clone was 5 of 12 (data not shown). This suggests that the two types of vAFP are expressed equally in K562. The far stronger signal of HepG2 by RT-PCR using ex-2S and ex-14A indicated that HepG2 expressed the authentic form only. This result demonstrated clearly that the expression pattern of variant and authentic forms of AFP rigorously discriminates between K562 and HepG2 cells.

Expression of Variant Forms of AFP in Normal Human
Tissues-We next examined whether the variant transcripts of AFP are expressed in normal cells. Since we anticipated the expression in normal cells would be at very low levels, a nested (two step) RT-PCR was performed. The first PCR was performed with primers, ex-A1S and ex-14A, to amplify the whole coding sequence of variant forms. Subsequently, nested PCR was performed by an internal primer combination, ex-A2S and ex-3A, which can distinguish between the presence of type A and type AB by the molecular size (Fig. 4A). The PCR results indicate that the tissue distribution of the vAFP transcripts is fairly restricted. Only primary hematopoietic organs, bone marrow (lane 1) and thymus (lane 2), and brain (lane 7) expressed the transcripts. Other tissues, including liver, lung, trachea, kidney, stomach, small intestine, colon, heart, and spleen, did not have cells expressing vAFP. These results strongly suggest that hematopoietic progenitors, but not mature blood cells, express vAFP, because the spleen is a secondary hematopoietic organ in which hemato-lymphopoiesis does not take place under normal physiological conditions. In addition, cDNA from peripheral blood cells (CLONTECH) never showed the expression of vAFP (data not shown). On the other hand, authentic AFP was detected in brain, kidney, small intestine, and thymus, but not in bone marrow. No expression in normal human fetal fibroblasts (lane 13) or small intestine (lane 4) suggests that the vAFP expression is not associated simply with cell proliferation.
Variant Transcripts of AFP Expressed in Hematopoietic Progenitors-The RT-PCR of tissue RNAs cannot define which cells express the variant forms of AFP transcripts. Bone marrow and thymus consist of a very heterogeneous cell population. Therefore, we studied purified hematopoietic progenitors from UCB by flow cytometric sorting to learn whether or not they express vAFP. Unfractionated nucleated cells in UCB contain only small numbers of CD34 ϩ cells (Fig. 5A) in which hematopoietic progenitors are highly enriched (27). The CD34 ϩ UCB cells could be enriched after removing Lin-positive cells (Fig. 5A). Both live (7AAD Ϫ ), unfractionated cells and CD34 ϩ Lin Ϫ cells were isolated by flow cytometric sorting, and the RNAs extracted were subjected to nested PCR for vAFP. A total of four different cord bloods were tested. As shown in Fig.  5B, the expression of vAFP transcripts was detected in CD34 ϩ Lin Ϫ cells, enriched hematopoietic progenitors (lanes 2,  4, 6, and 8; 4 of 4), but not in unfractionated cord blood cells Open boxes and horizontal lines indicate exons and introns, respectively. Their lengths in nucleotides are indicated by numbers above or below these elements. One type of the variant forms used one additional exon (exon A) located ϳ5-kb upstream from exon 1. The other type used exon A and another exon (exon B) located ϳ1.6 kb from exon 1. Genomic organization of the variant forms was deduced from genomic sequence of the AFP gene. C, selective expression of variant AFP transcripts in K562. RT-PCR of primer combination of exon A primer (ex-AS) and ex-14A primer were performed with cDNA from K562 (lane 1) and HepG2 (lane 2). K562, but not HepG2, expressed variant forms of AFP transcripts. The strong signal of HepG2 by RT-PCR using ex-2S and ex-14A indicates that HepG2 expresses the authentic form only. Equal amounts of cDNA were used for the reaction. Lane 3, control reaction. (lanes 1, 3, and 5; 3 of 4). Cloning and DNA sequencing experiments showed that the amplified bands corresponded to cDNA sequences of type A and type AB from K562 cells (data not shown). CD34 expression in the sorted CD34 ϩ Lin Ϫ cells was confirmed by RT-PCR (Fig. 5B).

DISCUSSION
The present study identified two variant forms of AFP expressed preferentially in hematopoietic progenitors, but not in mature blood cells or in hepatic cells. These results suggest that the AFP gene locus of chromatin in hematopoietic progenitors is open and accessible to transcription factors for the mRNA expression. In other words, chromatin-related repression of authentic AFP, which is a mechanism that will block inappropriate expression of authentic AFP in non-hepatic cells (28), is incomplete in hematopoietic progenitors and allows vAFP to be transcribed. In the case of K562, strictly speaking, very small amounts of authentic AFP transcripts could be detected when cycles of RT-PCR were increased (Fig. 4B, lane  14). This indicates that K562 express vAFP at a much higher level than authentic AFP. Conversely, vAFP could be detected in HepG2 by increasing cycles of RT-PCR (data not shown). Therefore, our data derived from the cell lines suggest that three types of patterns are present with respect to the authentic and variant transcripts of AFP. Hepatic cells (HepG2) express authentic AFP dominantly, while hematopoietic cells (K562) express vAFP dominantly. Fibroblasts (MRC5) express neither authentic AFP nor vAFP. This prompts us to ask two questions with respect to AFP expression. 1) What is the mechanism of the opening of the AFP locus in the chromatin, giving rise to whichever of the transcripts is expressed? 2) What is the transcriptional machinery associated with the different transcripts?
The mouse has proved an excellent model system for studying tissue-specific and developmentally regulated transcriptional control of authentic AFP in vivo as well as in vitro (14 -16). Extensive studies have established that the transcriptional control of the AFP gene is mediated by five cis-acting regulatory domains, including the AFP promoter, three distinct enhancer elements, and one repressor region located between the AFP promoter and the upstream enhancers (reviewed in Ref. 29). There are a number of transcription factors binding to the promotor region to control the expression of the authentic AFP form (reviewed in Ref. 29). It has been proposed that GATA factors, especially GATA4, could be a master gene for endodermal differentiation (30) and for the expression of fetoprotein transcription factor, which activates AFP expression in early liver development (31). On the other hand, other GATA factors are key regulators for hematopoiesis (32). Interestingly, several possible GATA family binding sites were identified in the exon A genomic sequence as well as other transcription factor binding sites associated with hematopoiesis such as AML-1 (33) and MZF-1 (34). Because exon A was identified as only one 5Ј-end of two variant forms by our RACE method in all independent clones, exon A is possibly the first exon of vAFP in K562. Thus, it is noteworthy to see whether these transcription factors expressed in hematopoietic progenitors are involved in the transcription of vAFP. However, the transcriptional start site of the variant forms of AFP in hematopoietic cells remains to be determined.
Although extensive research has been performed to elucidate AFP gene regulation in various species (15,35), there has been no report about human variant AFP transcripts nor identification of exon A and exon B 5Ј-upstream of exon 1 in any species. Variant AFP transcripts identified in rodents were expressed preferentially in hepatic or yolk sac cells (24,25). These suggest that the vAFP transcripts in the human hematopoietic cells are different from ones identified in rat or in mouse not only in their molecular structures, but also in their expression pattern. Interestingly, all variant forms of AFP reported in the previous studies and this study use unique first exons. Because the different usage of the first exon suggests that the vAFP expression is dependent on a unique promotor (24,25), further studies are required to elucidate the hematopoietic cell-specific promotor of vAFP.
At present, apart from the hematopoietic organs, the brain is the only tissue where vAFP expression is observed. It has been shown that cultured neurospheres, which are considered as neural stem cells, isolated from the anterior portion of the lateral wall of the lateral ventricles in adult mice were able to differentiate to endodermal cells (36). Therefore, it is important to determine the identity of vAFP ϩ cells in the brain and isolate them from brain tissue in order to ask whether they have a capability to form neurospheres in vitro. The expression of vAFP in non-endodermal cells may represent a unique immature state, which has the developmental potential of endodermal cells.
The function of the vAFP is unclear because the amount of mRNA was extremely low, and no protein products were ever found in assays using immunohistochemistry with anti-AFP antibodies on hematopoietic progenitors. 2 In addition, there is no long open reading frame starts in exon A or B to connect with exon 2 of authentic AFP if only ATG is considered as an initiation codon. In that case, a possible initiation codon at exon 3 would be used for the translation so that the translated product from vAFP transcripts would be a truncated form. However, because there is TTG and CTG in exon B to connect the open reading frame of exon 2, it might be possible for one of them to work as an alternative initiation codon (37,38). In any event, further studies are required to elucidate the possible functions of vAFP transcripts.
As a final point, this study raises a warning about the use of RT-PCR techniques for the detection of AFP mRNA. AFP is used commonly as a molecular marker for endodermal differentiation (17)(18)(19) or certain carcinomas such as hepatomas or germ cell tumors (15,35). Various RT-PCR techniques have been developed for the detection of circulating carcinoma cells in peripheral blood (39). Nevertheless, none of them, to the best of our knowledge, used exon 1-specific primers to detect the authentic form of AFP. Thus, the use of not only the exon 1-specific primer but the exon A-specific primer for RT-PCR to analyze AFP mRNA expression will be necessary to consider which types of transcripts are expressed in the cells examined.
In conclusion, this is the first study in which variant forms of an endodermal marker, AFP, have been found, with a selective pattern of expression in human hematopoietic progenitors. Our results suggest that the expression of vAFP transcripts may reflect the ability of progenitors for hematopoietic cells to differentiate to endodermal cells and provides a unique opportunity for studying the mechanism.