Transposed elements more likely to be exonized in duplicated genes

Are these TE exonizations within duplicated genes part of a neo-functionalization or a non-functionalization process? Apparently, both trajectories are present within our dataset. An example for neo-functionalization is the primate-specific duplication of the gene GON4L that generated the duplicated gene YY1AP1 [38, 39]. The gene YY1AP1 is functional and also has a new role as a co-activator, YY1; it also has different levels of expression within human tissue relative to its duplicate [38, 40]. Our exonization analysis revealed that YY1AP1 contains two different Alu exonizations that do not exist within GON4L. One of the exonizations results from the Alu that was inserted prior to the gene duplication. The other resulted from an insertion of an Alu element after the duplication and subsequent exonization [41].

An example of non-functionalization and subsequent pseudogenesis is found in the duplication of the NCF1 gene. NCF1 is one of the genes responsible for the chronic granulomatous disease (CGD) and also contributes to autoimmunity [42]. In human, the gene underwent triplication, the two duplicates are well characterized [43]. Our analysis of TE exonization revealed that there was an Alu exonization within the second intron of each of the two pseudogenes. There is no evidence for this exonization within the functional gene, even though the Alu element is present within the intron of the functional gene as well. Our dataset contains both exonization of TEs that were inserted prior to gene duplication and those that were inserted after gene duplication. Two of the Alu exonizations in duplicated genes in our dataset were shown to be associated with cancerous tissues based on the tissue origin of their ESTs [28, 44]; the Alu exonization within intron 2 of the gene YY1AP1, the aforementioned duplicate of the gene GON4L [38]; and Alu exonization within intron 6 of the gene ACAD9, a paralog of ACADVL (p-value < 0.01).

TIF-IA gene underwent a hominidae-specific triplication

One of the genes from our bioinformatic analysis is transcription initiation factor IA (TIF-IA). This transcription initiation factor directs growth-dependent regulation of RNA polymerase I. It is a 75-kDa protein and levels, or activity, fluctuate in response to cell proliferation [45]. Genetic inactivation of the transcription factor TIF-IA leads to nucleolar disruption, cell cycle arrest, and p53-mediated apoptosis [46]. Therefore, TIF-IA is a key protein in adapting cellular biosynthetic activities to cell growth [47].

The gene that encodes TIF-IA is highly conserved from yeast to human and is essential for cell survival. Alignment of the human cDNA of TIF-IA against the human genome using Blat (UCSC Genome Browser [48]) revealed two imperfect copies of TIF-IA, which presumably resulted from gene duplication. The original copy of TIF-IA gene (termed locus 15), as well as the two duplicates (termed locus 28 and locus 21), are located on chromosome 16. In addition, a processed pseudo-gene of TIF-IA is located on chromosome 2, but is not transcribed according to EST data.

The two duplicates of the gene were probably generated in a sequential manner, as illustrated in Figure 1. Non-homologous recombination was probably the mechanism for the triplication, because all the copies are on the same chromosome. In detail, the original gene from locus 15 was duplicated almost completely, except for the last exon (exon 18), to locus 28. Next, a major deletion of 3,517 bp within the duplicate, which included exons 11 and 12, occurred. Then, locus 28 was duplicated to locus 21. This duplication was also incomplete, resulting loss of the last three exons (exons 16 to 18). The deletion was confirmed experimentally (data not shown).

Blat of human mRNA on the chimpanzee genome suggests that it also contains more than one copy of TIF-IA gene. The original gene is located on chromosome 16, as it is in the human genome. There are several regions of homology to TIF-IA on chromosome 16 (in addition to the WT gene), as well as a processed pseudogene on chromosome 2b. It was difficult to determine the order and pattern of duplication events in the chimpanzee genome due to incomplete sequencing and the low level of genome assembly; however, it is clear that the chimp genome contains more than one copy of the TIF-IA gene on the same chromosome.

Alignment of TIF-IA cDNA with the rhesus genome revealed the presence of the original copy of the gene on chromosome 20 and the processed pseudogene on chromosome 12, but no duplications. Similar analysis on other vertebrate genomes (C. familiaris, B. taurus, M. musculus, R. norvegicus, G. gallus, X. tropicalis, D. rerio) and non-vertebrate organisms (D. melanogaster, S. cerevisiae, S. pombe, C. elegans) revealed that each had only a single copy of TIF-IA gene. Therefore, we concluded that the duplication of TIF-IA gene occurred after the human-chimp-rhesus split but before the human-chimp split between 4 and 25 million years ago.

Alu insertion in the first intron of TIF-IA

Examination of the TIF-IA genes revealed another intriguing event: An Alu element was inserted into the first intron of the common ancestor of human, chimpanzee, and rhesus (approximately 25 million years ago). This Alu was inserted into another transposed element called L2, located in intron 1 of the TIF-IA gene. We reconstructed this scenario by examination of intron 1 of the TIF-IA gene in other mammals. The L2 element was present in all sequenced mammals except opossum. The insertion of the Alu element into the L2 transposed element led to an exonization in human. Based on alignment of ESTs to the human genome (see Additional file 2), the 3' splice site (3'ss) is donated by the L2 transposed element and two alternative 5' splice site (5'ss) are selected, one donated by the Alu sequence and the other by L2. We designated the distal 5'ss, located in the Alu sequence as 5'ss-a, and the proximal 5'ss, located in the L2 element, as 5'ss-b (Figure 2A and 2B). The overall steps that lead to the exonization of that exon were as follows: (1) During primate evolution, an Alu-Sx element was inserted into an L2 retrotransposon of the LINE family. (2) The sequence accumulated mutations that caused exonization, leading to selection of three different isoforms as demonstrated in Figure 2A. This exon is termed L2-AEx.

Transcription and translation of the wild-type and duplicate TIF-IA genes

To examine whether the duplicates in the human genome are active transcriptionally, we examined the putative promoter regions of all three loci (Figure 1). The region of 1 kb upstream of the translation start codon of locus 15 has 96.5% identity to the corresponding upstream region of locus 21. The 1-kb promoter region of locus 28 shows similarity to that of locus 15; however, this putative promoter contains a deletion of 768 bp that ends at position -117 upstream of the potential translation start codon.

The many changes that occurred between the wild-type gene and the two duplicates suggest that the duplicates are not under selective pressure to encode functional protein. Translation of the mRNA produced from the original TIF-IA gene results in a 651-aa protein with a molecular weight of 75 kDa. The duplicate at locus 28 potentially encodes a 514-aa polypeptide chain, whereas the duplicate at locus 21 potentially encodes a 39-aa polypeptide using the genuine start codon or a 106-aa polypeptide chain using the third ATG downstream of the transcription start site (TSS). Polypeptides of sizes expected from the duplicate loci were not detected by western blotting analysis with polyclonal antibodies to the wild-type TIF-IA protein (Additional file 3). Although proteins do not appear to be generated from the duplicate genes, these duplicates are active transcriptionally as indicated by ESTs that derive from these loci. In addition, we measured the relative mRNA levels based on nucleotide variation at certain positions between the loci. In P69 cells, mRNA from the wild-type gene comprises about 80% of the mRNA transcribed from the three loci and locus 28 and locus 21 constitute 3% and 17% of the mRNA, respectively (Figure 1).

Alu-exonization is exclusive to the TIF-IA duplicate in locus 21

EST analysis revealed that a large fraction of the L2-AEx-containing ESTs originated from carcinogenic tissues, such as Burkitt's lymphomas (Additional file 2). To examine the splicing patterns of exons 1 and 2 of the wild-type gene and the two duplicates, we designed a pair of primers to sequences in the flanking exons: 1 and 2 that are conserved among all three loci, meaning all three copies will be amplified simultaneously. We then performed RT-PCR analysis of human cDNA from normal tissues and from transformed cell lines. The major mRNA product from the TIF-IA genes in most of normal tissues skips the Alu exon and shows either no exonization or a negligible level of exonization (Figure 3A). In two cell lines, 293T and BJ-1, there was a substantial level of exonization (Figure 3A, lanes 9 and lane 10, respectively). These results imply that, for most of the tested cells, the exonization is a minor or non-existent event. We did not detect any exonization of L2-AEx within chimpanzee blood cells, implying that this exonization may be human specific (Figure 3A, lane 17).

Next, we examined the exonization in 17 leukemia cell lines by RT-PCR analysis and detected a high level of exonization of both the short and long exonized form in 13 out of the 17 cell lines (76%). To specifically identify the locus from which this exonization occurred, we designed loci-specific primers and repeated the RT-PCR analysis using the pair of primers that detects all three loci and the locus-specific pairs of primers (Figure 3B). Only locus 21 showed exonization in which 5'ss-a or of 5'ss-b or both were selected and the ratio among the three isoforms varied significantly among the cell lines. Thus, there was a low level or an absence of exonization in normal human tissues and in epithelial cancer cell lines. In contrast, most leukemia cell lines exhibited a high level of exonization. The exonization was restricted to one of the duplicates (locus 21), but not to the original gene (locus 15). The cells that show exonization do not exhibit a known common characteristic, such as a specific hematopoietic lineage or mutations in a certain pathway or gene that can explain the exonization in these cells but not in the others. These results imply that there is a certain level of transcriptomic instability of the TIF-IA locus in certain cancer cells, in particular leukemic. The definition of exon/intron is abnormal for the TIF-IA locus, and perhaps other loci as well, in these cancer cells; what is defined as an intron for normal tissues is selected as an exon in leukemic cells.

The wild-type TIF-IA gene is one step away from exonization

To understand the molecular mechanisms leading to exonization in locus 21, but not in loci 15 and 28, we compared the genomic sequence of the L2-Alu among the loci (Figure 2B). Many point mutations have accumulated in the L2-Alu genomic sequence following the duplication event. One of these mutations was of particular interest, because it changed a GAG found in locus 15 and 28 to CAG in locus 21. This CAG is used as the 3'ss of the L2-AEx. GAG is a week 3'ss, whereas CAG is a strong one [49]. We, thus, hypothesized that this mutation leads to locus 21-specific exonization.

To examine this hypothesis, we cloned a mini-gene containing the human genomic DNA from exon 1 to exon 2 of locus 15 (WT gene). The mini-gene was transfected into human 293T cells, RNA was extracted, and the splicing pattern between exon 1 and 2 was examined by RT-PCR analysis using primers that are specific to the mini-gene mRNA. As expected the WT mini-gene does not exhibit Alu-exonization (Figure 4B, lane 2). However, a single point mutation at position -3 (G → C) at the putative 3'ss activated the exonization (Figure 4B, lane 3). The predominant mRNA generated following this mutation is the one that includes the L2-AExs (selection of 5'ss-b), leading to a low level of the normal mRNA. In addition, the 3'ss mutation also generated an intron retention isoform. This indicates that the L2-Alu in the wild-type gene (locus 15) is one mutation away from exonization. If such a mutation were to occur, it would terminate production of a normal TIF-IA protein in the cells almost completely, because the L2-AEx inserts a premature stop codon.

We next evaluated the effect of 5'ss strength on exonization. Strengthening of the either of the two sites did not activate the exonization (Figure 4B, lanes 4 and 5) and neither did strengthening both (Figure 4B, lane 8). When the 3'ss is functional, the selection of 5'ss-a or 5'ss-b is determined by their relative strength (Figure 4B, lanes 6, 7, and 9). Overall, these results indicate that the L2-Alu element in the WT TIF-IA gene is on the edge of exonization and that a single point mutation from G to C in position -3 of the putative 3'ss leads to exonization.

Next, we examined which nucleotides in position -3 of the 3'ss would activate exonization. We mutated the 3'ss from GAG to CAG, TAG, and AAG (Figure 4C, lanes 1, 2, and 3, respectively). Only the mutation from G to C at position -3 lead to exonization: The mutations to A and T did not. These results support our previous observation that CAG is the strongest 3'ss [49]. Also, the results show that alternative splicing is delicately balanced and is partially controlled by the strength of the 3' and 5'ss signals.

Alternative splicing is often regulated in a tissue-specific or developmental-stage-specific manner. A common explanation for differential splicing patterns is that the concentrations of splicing regulatory proteins vary depending on tissue type and developmental stage. Therefore, we examined the effect of different cellular environments on the splicing of the TIF-IA mini-gene by transfection into different cell lines; we used the mini-gene containing the G-to-C mutant at position -3 of the 3'ss. The patterns of splicing observed depended on the cell line (Figure 4D). In two cell lines, Du145 and PC3, the introns were not always excised (Figure 4D, lanes 2 and 6, respectively; see figure legend for the source of each cell line) and in U2OS, HT1080, HepG2, HeLa, and 293T cell lines, the alternatively spliced mRNA containing L2-AEx was the predominant isoform (lanes 1, 3, 4, 5, and 7, respectively). These findings show that, for this TIF-IA mini-gene, the cellular environment regulates the level of exonization and intron/exon recognition.

Alu-exonization in YY1AP, the duplicate of GON4Lb

To further support the bioinformatics analysis we demonstrated another case of a gene that underwent a duplication event, and the duplicate gene exhibits a distinct alternative splicing pattern originated from an intronic Alu element, which also exist in the ancestor gene, but is not exonized. Figure 5 demonstrates the genomic structure of the original GON4Lb gene compared to the duplicated gene, YY1AP1 [38]. An intronic Alu element is exonized between exons 13 and 14 (Fig. 5A). Aligning the Alu element in both original and duplicated genes uncovers many sequence changes between them (Fig. 5B). In contrast to the TIF-IA gene, no mutation was detected in the splice signals, thus we assume that the reason for exonization in the duplicated gene and not in the original GON4Lb gene is due to mutations in regulatory sequences such as ESRs [50]. Figure 5C examines the splicing pattern of the corresponding genomic region and revealed that the Alu element in GON4Lb gene is not exonized while the duplicated gene (YY1AP1) exhibits low levels of exonization. When the Alu-exon enters into the mature RNA of YY1AP1, it inserts a pre-mature termination codon (PTC) [51]. This strengthened our bioinformatics analysis showing a second example of exonization that is found in the duplicated gene and not in the original copy.

Dataset of TE exons in human and mouse genomes

We compiled a dataset of TE exonizations within protein coding genes [28]. The human NCBI 35 (hg17, May 2004) and the mouse NCBI33m (mm6, March 2005) assemblies were downloaded, along with their annotations, from the UCSC genome browser database [55]. Coordinates of the EST and cDNA mapping were obtained from chrN_intronEST and chrN_mrna tables, respectively. The TE mapping was obtained from chrN_rmsk tables. A TE was considered intragenic if there was no overlap with ESTs or cDNA alignments; it was considered intronic if it was found within an alignment of an EST or cDNA within an intronic region. Finally, a TE was considered exonic if it was found within an exonic part of the EST or cDNA, if it possessed canonical splice sites, and if it was an internal exon of the EST/cDNA found within the CDS or a UTR. The insertions of TEs within EST/cDNA alignments were separated into two parts: those that entered within protein coding genes relative to the list of the knownGene table in the UCSC genome browser [55] (based on proteins from SWISS-PROT, TrEMBL, TrEMBL-NEW and their corresponding mRNA from GenBank), and other insertions within cDNA/ESTs alignments that were not mapped to the known genes list and, therefore, were considered as non-protein-coding genes. Non-protein-coding genes were determined as genomic regions covered by at least two correctly spliced cDNA/ESTs (flanked by canonical splice sites) containing at least three exons that did not overlap any annotated gene based on UCSC known genes lists, versions hg17 and mm6 for human and mouse, respectively. Unspliced genes were not included in our analysis; we only considered genes with at least two introns. Internal UTR exons were considered to be internal according to the annotations of knownGenes in UCSC and the fact that they were internal within the cDNA/EST. The TE position within the gene (UTR or CDS) and the exon phase were calculated based on knownGenes table annotations of the gene start and end positions, as well as CDS start and end positions.

TE exonization within duplicated genes

To analyze whether the exonization occurred within a duplicated gene or a singletons, we performed a blast search of these genes against all mRNA sequences listed as known genes (knownGene table) downloaded from UCSC human genome build hg17 [55] and searched for transcripts with at least 75% sequence similarity along 40 percent of the gene, the dataset was filtered to delete transcripts of isoforms generated by alternative splicing that belonged to the same gene and nonduplicated genes. The filtration of similar mRNA that are isoforms of the same gene was done by comparing the locus of the mRNA as indicated in UCSC genome browser; only mRNAs that mapped to different loci were considered.

Cell line maintenance

293T, HeLa, HT1080, HepG2, Du145, and U2OS cell lines were cultured in Dulbecco's Modification of Eagle Medium (DMEM), supplemented with 4.5 g/mL glucose (Biological Industries, Inc., Israel), 10% fetal calf serum (FCS), 100 U/mL penicillin, 0.1 mg/mL streptomycin, and 1 U/mL nystatin (Biological Industries, Inc.). PC3 cells were cultured in Ham's F12K medium with 2 mM L-glutamine adjusted to contain 1.5 g/L sodium bicarbonate (90%) and fetal bovine serum (10%). Cells were grown in 6-well plates or 100-mm culture dishes under standard conditions at 37°C with 5% CO₂. Cells were split at a 1:10 ratio twice weekly.

Transfection

Cells were grown to 50% confluence in 100-mm culture dishes or 6-well plates and maintained as described above. Twenty-four hours prior to transfection, cells were split, and transfection was performed using FuGENE6 (Roche) with 0.5–1 μg of plasmid DNA. Cells were harvested after 48 hr.

RNA isolation and splicing analysis

Total RNA was extracted using TRI Reagent (Sigma), followed by treatment with 2 U of RNase-free DNase (Ambion). Reverse transcription (RT) was performed on 1–2 μg total RNA in a 12.5 μL final volume containing: 100 mM DTT, 10 mM dNTPs, 100 mM oligo(dT) primer, 2 U of reverse transcriptase avian myeloblastosis virus (RT-AMV, Roche), and RT buffer. The final mixture was then incubated for 1 hr at 42°C. The spliced cDNA products derived from the expressed mini-genes were detected by PCR, using Taq polymerase (BioTools), and pEGFP-C3 specific reverse and forward primers. Primer sets for cell-line scanning purposes were designed to amplify all loci in a single PCR or to amplify every locus separately: 3 loci: 5'-CGT TAG TTC GGC CCA ATG-3', 5'-CTT CAG CAA GAC TTC TGT CAC-3'; locus 15: 5'-CTT CGT CCT CTG CAG TTA AGA AG-3', 5'-CTT CAG CAA GAC TTC TGT CAC-3'; locus 28: 5'-CGC TTC GCC CTC TGC AGT C-3', 5'-CTT CAG CAA GAC TTC TGT CAC-3' after reverse transcription (RT) with unique primer; locus 21: 5'-CTT CAC ACG TTG TTT GTC G-3', 5'-CTT CAG CAA GAC TTC TGT CAC-3'. Amplification of the chimpanzee cDNA (proliferating blood cells of a female chimpanzee from the safari in Ramat-Gan) was performed with 5'-CGT TAG TTC GGC CCA ATG-3' and 5'-GCT GGT TCT TCA ACA ACT CAA A-3'. The primers flank the Alu element in intron 1 of the TIF-IA's genes. Human GAPDH: 5'-TCG TGG AGT CCA CTG GCG TCT T-3' and 5'-TGG CAG TGA TGG CAT GGA CTG T-3'. Chimp GAPDH: 5'-TCG TGG AGT CCA CTG GCG TCT T-3' and 5'-TGG CAG TGA TGG CAT GGA CCG T-3'. GON4L: 5'-ATG AGC TGA TGG AAG AGC TG-3' and 5'-GAG GGG TGT TAA AGT TAG GAC GAG-3'. YY1AP1: 5'-CAA ATG AGC TGA TGG AAG AT-3' and 5'-GAG GGG TGT TAA AGT TAG CTT-3'. Amplification was performed for 30 cycles, consisting of denaturation for 30 seconds at 94°C, annealing for 45 seconds at 58°C, and elongation for 1.5 minutes at 72°C. The products were separated in 1.5% agarose gel; selected bands were confirmed by sequencing.

Plasmid construct

A genomic DNA from 293T cell line (Gentra) corresponding to exon 1 through exon 2 of the TIF-IA gene (from locus 15) was PCR amplified and cloned into pEGFP-C3 vector (Clontech) between XhoI and BamHI sites under the control of the human cytomegalovirus (CMV) immediate early promoter, giving a ~1.9 kb insert. F: 5'-AAA AAA ACT CGA G GC TGA TTG GCT GAA GGT TG-3'; R: 5'-AAA AGG ATC C CA GCA ATA GTT GTA TTC TGA CCT AAC C-3'.

Site-directed mutagenesis

Specific overlapping oligonucleotide primers that contained the desired mutation were used to insert the mutation using PfuTurbo DNA polymerase (Stratagene). After PCR amplification, the reaction was digested with DpnI restriction enzyme (New England Biolabs) for 1 hr at 37°C; 1–3 μL of the reaction were transformed into E. coli XL-1 strain and colonies were picked for mini-prep extraction (Qiagen) and sequenced. L2-AEx-3'ss (-3)G->C: 5'-GAA GAC ATC TGT CAT TTC AGC TTC CAC TTG AAT G-3' and 5'-CAT TCA AGT GGA AGC TG A AAT GAC AGA TGT CTT C-3'. L2-AEx-3'ss (-3)G->T: 5'-GAA GAC ATC TGT CAT TTT AGC TTC CAC TTG AAT G-3' and 5'-CAT TCA AGT GGA AGC TA A AAT GAC AGA TGT CTT C-3'. L2-AEx-3'ss (-3)G->A: 5'-GAA GAC ATC TGT CAT TTA AGC TTC CAC TTG AAT G-3' and 5'-CAT TCA AGT GGA AGC TT A AAT GAC AGA TGT CTT C-3'. L2-AEx-5'ss-a (5)C->G: 5'-CAT GGT GGC AGG TGC G TG TAA TCC CAG CTA C-3' and 5'-GTA GCT GGG ATT ACA C GC ACC TGC CAC CAT G-3'. L2-AEx-5'ss-b (-1)A->G: 5'-GTA TTT CCA ATA GAG TGA ACG GTA AGT GAA ATG AAA AAC AGC-3' and 5'-GCT GTT TTT CAT TTC ACT TAC C GT TCA CTC TAT TGG AAA TAC-3'.

Western blotting

Lysis buffer (50 mM Tris, pH 7.5; 1% NP40; 150 mM NaCl; 0.1% SDS; 0.5% deoxycholic acid; protease inhibitor cocktail and phosphatase inhibitor cocktail I and II; Sigma) was used for protein extraction. Then lysates were centrifuged for 30 minutes at 14,000 rpm at 4°C. Total protein concentrations were measured using BioRad Protein Assay (BioRad). Proteins were separated in 10% SDS-polyacrylamide gel electrophoresis (SDS-PAGE) and then electroblotted onto a Protran membrane (Schleicher and Schuell). The membranes were probed with polyclonal anti-TIF-IA antibody (kindly provided by Ingrid Grummt) at 1/10000 dilution followed by rabbit secondary antibody. Immunoblots were visualized by enhanced chemiluminescence (Lumi-Light Western Blotting Substrate; Roche) and exposure to x-ray film.

Transposed elements analysis

RepeatMasker^© software version 3.1.0 [56] was used for the detection of transposed elements.

Analysis of the relative mRNA levels

The PCR product from RT-PCR of P69 cells was excised and purified following electrophoresis on 1.5% agarose gel (Promega, Madison, WI, USA). Direct sequencing was performed using the ABI PRISM (Applied Biosystems, Foster-City, CA, USA). The variation percentage from direct sequencing was calculated for the reverse primer; the presented percentages represent an average of three separated positions (31, 63 and 105) along exon 2 of the mRNAs. The nucleotides were quantified by the Discovery Studio (DS) Gene 1.5 program (Accelrys Inc., San Diego, CA, USA).