cDNA Sequencing and General Data of ESTs from HFL22w

The HFL22w cDNA library had average insert sizes of 1.0–1.5 kb. By using automatic procedures for DNA sequencing, 14,400 clones were randomly picked up and sequenced partially from one end by using T7 or SP6 primer. Of them, 743 were considered trash, defined as sequences from bacterial DNA, sequences from primer polymers, sequences containing >1% of ambiguous bases (N), or sequences shorter than 100 bp; the other 13,077 sequences were considered good ones. The rate of successful sequences was therefore 90.8% and the average read-length for good sequences is 555 bp, which, to our knowledge, is among the best in the literature. Analysis of the 13,077 ESTs of satisfactory quality revealed three groups of sequences. Group I (5819 ESTs, 44.5%) matched to known genes in the GenBank nonredundant database and were considered labels of known functional genes, among which 5666 ESTs (43.3%) matched to human genes and the other 153 ESTs (1.2%) matched to previously described genes of other species. Group II (5460 ESTs, 41.8%) exhibited no significant homology to known genes, and 18.8% (1025 ESTs) of these overlapped EST sequences in the public database (dbEST). Group III (1798 ESTs, 13.7%) were genomic sequences of unknown function, mitochondrial DNA, or repetitive sequences.

Gene Expression Profile of Active Genes in HFL22w

A catalog of genes expressed in HFL22w was established by generating a large amount of ESTs, followed by bioinformatics analysis (data available through E-mail: [email protected]). The uninformative sequences of Group III were put aside, and the remaining 11,279 ESTs of Group I and II were further analyzed and assembled into 1729 and 4768 clusters, respectively. After integration of overlapping sequences or sequences corresponding to different portions of the same gene, 5666 ESTs actually represented 1660 human genes and were summarized into 15 different functional categories in Table1. HFL22w ESTs were partitioned based upon biological roles and subcellular localization to include cell defense and homeostasis, cell division and regulation, cytokines and hormones, cytoskeleton, development, genes associated with diseases or abnormalities, gene/protein expression, hematopoiesis, liver and lipoproteins, metabolism, proteases and protease inhibitors, secretory proteins, signal transduction, transcription-related genes, and unclassified.

An expression profile of active genes in HFL22w is shown in Table2. In the list we can see several genes with certain frequency that could be expected based on the unique features and functions of HFL22w. First, in this developmental stage of human liver, cell proliferation and differentiation need a high productive level of protein synthesis as well as general metabolism, and a large amount of energy supply and protein synthesis is occurring. The genes expressed in HFL22w in the highest proportion were functionally related to the general housekeeping responsibilities of the cells, such as general metabolism, protein synthesis, and synthesis of nucleic acids and amino acids, which includes transcripts for various enzymes involved in the central reactions of metabolism, elongation factors, and ribosomal proteins, similar to EST databases previously generated from other tissues (Adams et al. 1995). Second, HFL22w is a major site of fetal hematopoiesis and immune development. ESTs associated with hematopoiesis formed the largest group of transcripts, for example, hemoglobins, globins, complement components, prothymosin-α, angiotensinogen, T-cell cyclophilin, and glycophorin A. Third, as expected, HFL22w highly expressed liver-specific genes such as serum albumin, fibrinogens, apolipoproteins, α-fetoprotein, haptoglobin, and high density lipoprotein-binding protein. In addition, genes for signal transduction, genes associated with diseases or abnormalities, and transcription-related genes were also noticeably active. Some cytokines and hormones such as insulinlike growth factor II (IGF-2), thymosin β-4, β-10, FGFR-4, lens epithelium-derived growth factor, megakaryocyte-stimulating factor, osteoclast-stimulating factor, and transforming growth factor (TGF) were also encountered in the EST data.

Among 13,077 clones, 10.8% belong to two abundant transcripts, hemoglobin γ-G and serum albumin (HSA), which had 724 and 694 copies, respectively. Other frequent transcripts were ferritin light chain, H19 gene, retinol binding protein (RBP), α 1 globin, and so on. Besides serum albumin, some other liver-specific genes were detected also, including fibrinogen-β, -γ, and -α chains; apolipoprotein-B100, -AII, and -AI; albumin; α-fetoprotein (AFP); high density lipoprotein-binding protein; and heptoglobin α 2 and β subunits, which are known to be abundant in the liver. Fifty-eight species of ribosomal proteins (total 283 ESTs; 2.2%) were sequenced in 13,077 randomly selected clones. Because mammalian ribosomes are reported to be composed of ∼70–80 distinct proteins (Wool 1986), most of the ribosomal proteins seemed to be represented, suggesting that gene/protein expression was very active in fetal liver of this developmental stage.

Table 3 shows some of the 153 ESTs of 69 different transcript species matched to nonhuman sequences. Several ESTs were found to be similar to the genes differentially regulated during development. Some of them may turn out to be involved in signal transduction during the differentiation and proliferation of the fetal liver. Further characterization would be necessary to find out the actual biological roles of these candidates. Together with the 5460 Novel ESTs (representing 4768 EST clusters, Group II), we identified 4837 EST clusters whose biological functions were not completely known that could be good candidates for full-length cDNA cloning of novel functional genes.

Identification of Tissue- and Developmental-StageSpecific Genes by the Compilation of the Expression Profiles of HFL22w and the Other Functionally Associated Tissues or Cells

Although we were profiling the active genes in HFL22w based upon 13,077 ESTs, as yet the number of ESTs collected for each expression profile obtained from the published data was only approximately 1000. It was not possible to compare the genes that appeared at low abundance. However, with those genes whose transcripts appeared at high abundance and represent typical physiological and developmental status, relatively accurate comparisons could be made and the conclusion might even be more objective. Therefore, genes listed in the tables were extracted from each of these expression profiles, detected two or more times, and the abundance of their transcripts among total ESTs compiled. Through the comparison, several gene groups associated with definite physiological and/or molecular features were identified.

We collected five other liver-associated expression profiles including human fetal liver at 19 wk (HFL19w) or 40 wk (HFL40w) of gestation, human adult liver (HAL), Itoh cells, and HepG2 cells (http://bodymap.ims.u-tokyo.ac.jp/human_1.html) and compared them with the expression profile of HFL22w established here. We extracted 773 genes whose abundance was two or more in at least one of the six expression profiles and compiled their activities (EST frequency). Only the genes whose transcripts appeared 15 or more times in the compiled expression profile are shown in Table 4. These genes were categorized into three classes according to the number of libraries in which they were detected: ubiquitous—appeared in five or six origins (filled area in the Library column, lib); common—appeared in two–four origins (hatched area in the Library column), and unique—appeared in only one origin (blank in the Library column). The functions of a gene could be assumed from the frequencies in random isolates from the different libraries shown in the compiled expression profiles. Among the 773 genes, nine (Gene Group I) appeared ubiquitously (Table 5). Some of them were likely to function in housekeeping, such as the three ribosomal proteins. The other six genes were actually tissue-specific, function-keeping genes of liver, including serum albumin, ferritin L chain, and apolipoprotein AII.

Table 4.

Compiled Gene Expression Profile Associated with Liver

Primary accession	lib	L1	L2	L3	L4	IC	HC	Gene definition
M15386		–	724	–	–	–	–	hemoglobin gamma-G
V00494		54	694	45	35	–	17	Serum albumin (ALB)
M11147		4	91	2	2	3	5	ferritin L chain
M32053		–	70	–	–	–	–	H19 RNA gene
X00129		–	69	–	5	–	–	retinol binding protein (RBP)
AF105974		–	68	–	–	–	–	alpha one globin (HBA1)
M16961		8	67	18	–	–	3	alpha 2-HS-glycoprotein alpha and beta chain
X01683		13	62	9	6	–	7	alpha 1-antitrypsin
J04617		–	62	–	–	11	17	elongation factor EF-1-alpha
X07868		–	55	5	–	–	–	insulin-like growth factor II
S95936		3	46	6	10	–	–	transferrin
J00129		21	41	15	8	–	–	fibrinogen beta-chain
X51473		7	38	2	5	–	–	fibrinogen gamma chain
M60854		–	37	2	–	–	4	ribosomal protein S16
U22961		–	34	–	–	–	–	mRNA clone with similarity to L-glycerol-3-phosphate:NAD  oxidoreductase and albumin gene sequences
AF042513		–	31	–	–	–	–	isolate Asn6 cytochrome b (CYTB)
AF014894		–	31	–	–	–	–	NADH dehydrogenase subunit 2
M36676		–	30	–	–	–	–	apolipoprotein B100
J04763		–	30	–	–	–	–	complement component 3 (C3)
D87666		–	30	–	–	–	–	heart mRNA for hsp90
L20955		–	29	–	–	–	2	prothymosin alpha
X00955		11	25	7	9	–	3	apolipoprotein AII
NM_000477.1		–	24	–	–	–	–	albumin (ALB)
X04225		–	24	–	–	–	–	protein HC (alpha 1-microglobulin)
NM_000559.1		–	23	–	–	–	–	hemoglobin, gamma A
X03168		–	23	–	–	–	–	S-protein
X02162		–	20	–	–	–	–	apolipoprotein AI (apo AI)
U14990		–	20	–	–	–	–	XP1PO ribosomal protein S3
M11313		6	18	3	5	–	–	alpha 2-macroglobulin
V00497		2	16	2	–	–	–	beta-globin
M64982		6	16	15	2	–	–	fibrinogen alpha chain
AF028832		–	16	–	–	–	–	Hsp90
X16064		11	15	2	2	9	9	translationally controlled tumor protein
X14420		–	10	–	–	16	–	pro-alpha 1 type 3 collagen
X00637		23	4	15	28	–	–	haptoglobin alpha 1S (Hpa 1S)
X55656		17	2	–	–	–	–	gamma-G globin
J03040		–	2	–	–	19	–	SPARC/osteonectin
M13692		–	–	4	15	–	–	alpha-1 acid glycoprotein
X13345		–	–	–	–	15	–	plasminogen activator inhibitor 1 (PAI-1)
J02775		10	–	29	13	–	2	RFLP 3′ to the apolipoprotein B gene

[i] Forty gene species matched to known human genes were listed in descending order of EST abundance of the genes in HFL22w and the order of gene definition. Only those genes whose transcripts appeared fifteen or more times are presented.

[ii] Lib, library; L1, human fetal liver aged 19 wk; L2, human fetal liver aged 22 wk; L3, human fetal liver aged 40 wk; L4, human adult liver; IC, Itoh cell; HC, HepG2 cell.

Open in new tabLink to table

On the other hand, 636 genes appeared only in one library (Table 4, blanks in Library column). Because their relatively high expression was unique to one expression profile among the listed six, they were the candidate genes whose products exerted unique functions in Itoh cells, HepG2 cells, or the liver in the different stages of development, respectively.

Eleven genes were expressed only in HFL19w and HFL22w but not in HFL40w or HAL (Table 5, Gene Group II). They were α-fetoprotein (AFP), 23-kD highly basic protein, thymosin-4, insulinoma rig-analog mRNA encoding DNA-binding protein, and seven ribosomal proteins. Genes expressed only in HAL and HFL40w but not in HFL19w or HFL22w are also listed (Table 5, Gene Group III). They, together with the genes of Gene Group II (Table 5), are developmental-stage-specific genes, which are suitable candidates for molecular probes to characterize the developmental stage of fetal liver. Further analysis of them would give impetus to the research of the molecular mechanism of liver development.

We also identified two other gene groups through systematic analysis of the mRNA population differences between the normal cells and the tumor cells in the liver. Gene Group IV consists of the genes expressed only in the three fetal livers and the adult liver but not in the hepatoblastoma HepG2 cells (Table 5). These genes might be candidate tumor suppressor genes or genes that were inhibited during tumorigenesis. On the contrary, Gene Group V consisted of genes expressed only in the HepG2 cells but not in the normal liver in various developmental stages (data not shown). These genes might be associated with tumorigenesis of the liver. Six genes in Gene Group II (Table 5) such as α-fetoprotein (AFP); ribosomal proteins L9, L19, S3a, and L6; and insulinoma rig-analog mRNA encoding DNA-binding protein were expressed in HepG2 cells and human fetal liver in the early stage of development (age 19 and 22 wk of gestation) but not in HFL40w or HAL. Because tumor cells often express embryonic genes in abnormal ways, these six genes might represent oncogenic status in hepatoma cells.

Although Itoh cells are located in the liver, their gene expression profile was obviously different from those of hepatocytes at various developmental stages and of the hepatoma cell line HepG2. Out of 120 genes that had two or more EST copies in Itoh cells, 60 were not expressed in any of the five other liver-associated expression profiles. Genes commonly expressed with high levels in liver, such as serum albumin (ALB), fibrinogen, transferrin, apolipoprotein AI, and haptoglobin, were not detected in Itoh cells. The different expression profile of Itoh cells contributed to its different physiological function from other types of liver cells.

The compiled gene expression profile associated with hematopoiesis (data not shown) consisted of five gene expression profiles including the CD34+ hematopoietic progenitor/stem cell (Mao et al. 1998), CD4 T cell, CD8 T cell, granulocyte, and myeloblastic leukemia cell line HL60 cell (http://bodymap.ims.u-tokyo.ac.jp/human_1.html). They had 134, 38, 45, 20, and 46 genes that also expressed in HFL22w, respectively. It was obvious that the CD34+ hematopoietic progenitor/stem cell shared the most active genes with HFL22w. Among the 595 genes whose frequency was two or more in the expression profile of HFL22w, 134 (22.5%) genes were also expressed in CD34+ hematopoietic stem/progenitor cells. Some of them were hematopoietic system-specific, for example, hemoglobin γ-G (HMG), β-globin, and T-cell cyclophilin. But the similarity between the expression profile of HFL22w and granulocytes was much less. This result matched the fact that there were few differentiated granulocytes in HFL22w.

Full-Length cDNA Cloning from HFL22w

Based on the bioinformatics analysis, 110 EST clusters have been chosen initially for full-length cDNA cloning. The clone inserts were sequenced with end-sequencing, primer extension, and sequencing after partial deletion/subcloning. After assembling ESTs into contigs, we found that 74 (67.3%) of the 110 cDNA clones already contained a complete open reading frame (ORF). In the other 36 cDNA clones, an obvious but incomplete reading frame was present. In silico cloning with dbEST extension allowed us to obtain 22 (20.0%) putative entire ORFs, which were then confirmed by sequencing of material cDNA clones obtained by appropriately designed RT-PCR. For the remaining 14 (12.7%) cDNA clones that could not be extended properly with an electronic approach, rapid amplification of cDNA ends (RACE) was applied to get the 5′ or 3′ ends from appropriate tissue origins. In total, 110 cDNAs with putatively entire ORFs were obtained. Table6 shows all 110 new full-length cDNAs from HFL22w. Among these 110 full-length cDNAs, 71 contained multiple exons and 87 had a consensus polyadenylation signal near the 3′ end; the 14 polyA tails might correspond to an A-rich region of the genome when they were searched against GenBank's working draft of the human genome. It is worth pointing out that, although a polyadenylation signal was found in the majority (73/110) of cDNAs as evidence of containing the complete 3′ UTR, the integrity of the 5′ UTR needs further experimental confirmation as in reports like that of the RIKEN Genome Exploration Research Group Phase II Team and the FANTOM Consortium (Kawai et al. 2001). Among these novel genes, the majority, 76 (69.1%), encode 80–500 amino acid residues deduced from their encoding frames. According to their homology with known genes and domains, some genes might be associated with signal transduction, such as the human homolog of mouse c-Jun leucine zipper interactive protein (cDNA JZA-20), the Kluyveromyces lactis transcription initiation factor IIIB 70-kD subunit, or Bos taurus guanine nucleotide-binding protein. And some genes might be new members of certain gene families, for example, the gene for the human homolog ofSchizosaccharomyces pombe Arf GTPase-activating protein, now termed human ADP-ribosylation factor GTPase-activating protein (ARFGAP3), belonging to the ARF GAP family (Zhang et al. 2000; Liu et al. 2001). In addition, some genes are very conserved in the species' evolution because their encoded proteins exhibit similar primary structure with those derived from such organisms as Arabidopsis thaliana, Schizosaccharomyces pombe, Kluyveromyces lactis, Plasmodium chabaudi, Tetrahymena thermophila, Caenorhabditis elegans, Drosophila melanogaster (Table 6), and other mammals (Qu et al. 2001). These novel genes might be involved in critical biological processes according to their homology to known genes with established significant functions like signal transduction, metabolism, protein expression, and hematopoiesis.

In further investigations, the chromosomal localization of 77 novel genes was determined, 70 of which were located by using database information of UniGene, dbSTS, dbHTGS, and Human Chromosome Databases; the location of the other 7 was determined by radiation hybrid (RH) mapping. The remaining 33 novel genes could not be mapped by either of the above methods.

DNA Sequencing

Bacteria growth and plasmid extractions of the HFL22w cDNA library (CLONTECH) were performed by a QIAprep 96 Turbo Miniprep Kit (QIAGEN). Sequencing reactions were performed on a GeneAmp PCR System 9700 thermal reactor (Perkin-Elmer) by using a BigDye Terminator Cycle Sequencing Kit (Perkin-Elmer) with T7 or SP6 primers. After removing the unincorporated dye terminators from sequencing reactions with DyeEx Spin Kits (QIAGEN), the reaction products were electrophoresed on an ABI 377-XL DNA sequencer (Perkin-Elmer–Applied Biosystems), and raw sequence data were automatically recorded.

Data Management and Bioinformatics Analysis

Sequences were edited manually by using PHRED andSequencher (version 3.0) to remove vector sequence and identify trash sequences, defined as sequences from bacterial DNA, sequences from primer polymers, sequences containing >1% of ambiguous bases (N), or sequences shorter than 100 bp. All sequence data were preserved on record tape. An in-house database for EST sequences generated from a cDNA library of HFL22w was established. The individual ESTs were searched against the GenBank nonredundant database (Release 105.0) for homology comparison by using BLASTN on theBLAST network server at the National Center for Biotechnology Information (NCBI). ESTs with an Alignment Score of the Basic Local Alignment Search (BLAST) >200 were considered to identify known genes or to have partial homology to known genes; the others were considered novel ones. Clustering of the ESTs generated in this work was performed by using PHRAP with default parameters.

Full-Length cDNA Cloning

The new sequences, considered as part of novel genes, confirmed by similarity searching against GenBank, were selected for full-length cDNA cloning. The program ORF Finder(http://www.ncbi.nlm.nih.gov/gorf/gorf.html) was applied to analyze the open reading frames. For those clones containing partial reading frames, in silico cloning and RACE were performed. In silico cloning was carried out using dbEST information, starting from the sequences obtained from the HFL22w cDNA library and then confirming these by sequencing of material cDNA clones obtained by appropriately designed RT-PCR. Sequence ambiguity existing in these contigs was clarified by further sequencing. A Smart RACE cDNA Amplification Kit (Clontech) was used to facilitate full-length cDNA cloning.

Genomic Mapping of Full-Length cDNA Clones

The chromosomal assignment of novel genes was mapped by two strategies: searching sequence databases such as Unigene, dbSTS, Human Chromosome Databases, dbHTGS at the National Center for Biotechnology Information; or radiation hybrid (RH). The Genebridge 4 RH panel (Research Genetics) was used in RH mapping.