Apolipoprotein(a) Gene Enhancer Resides within a LINE Element*

Apolipoprotein(a), (apo(a)), is the distinguishing protein portion of the lipoprotein(a) particle, elevated plasma levels of which are a major risk factor for cardiovascular disease. A search for enhancer elements that control the transcription of the apo(a) gene led to the identification of an upstream element that contains target binding sites for members of the Ets and Sp1 nuclear protein families. The enhancer element functions in either orientation to confer a greater than 10-fold increase in the activity of the apo(a) minimal promoter in cultured hepatocyte cells. Unexpectedly, the enhancer element is located within a LINE retrotransposon element, suggesting that LINE elements may function as mobile regulatory elements to control the expression of nearby genes.

Elevated levels of lipoprotein(a) (Lp(a)) 1 are a major risk factor for atherosclerosis, myocardial infarction, and cerebrovascular stroke (see Refs. 1 and 2 and references therein). Lp(a) is a lipoprotein consisting of low density lipoprotein to which is covalently linked a unique protein, apolipoprotein(a) (apo(a)) (3,4). Apo(a) is a large, highly polymorphic glycoprotein with sequence homology to plasminogen. Depending on individual alleles, it contains from 12 to 50 copies of a domain which resembles kringle four of plasminogen, plus a single kringle five-like domain, and an inactive protease domain (5,6). When plasminogen is activated to plasmin, it can cleave fibrin and several other protein substrates. By competing with plasminogen for fibrin binding through its apo(a) component, Lp(a) can reduce the conversion of plasminogen to the active plasmin and inhibit clot lysis (7)(8)(9). By a similar mechanism, Lp(a) can reduce plasmin-dependent activation of latent transforming growth factor-␤, resulting in smooth muscle cell activation, lipid accumulation, and inflammatory damage to the endothelium (10,11). Transgenic mice expressing the human apo(a) gene are more susceptible to the development of fatty streaktype lesions when fed a high fat diet, and they exhibit prolonged fibrinolysis (12,13).
The concentration of Lp(a) in human plasma varies about 1000-fold in the population and is inherited in a Mendelian co-dominant fashion. Roughly 97% of this variation maps to the locus of the apo(a) gene (14 -16). Hence, the regulation of apo(a) gene expression is significant in controlling pathological plasma levels of Lp(a). About one half of the large variation of apo(a) plasma concentration is due to transcription efficiency, and one half to the post-transcriptional effects of size polymorphism (16). Since isoforms of apo(a) with greater numbers of kringle domains transit through the cell less rapidly and efficiently, there is a rough inverse correlation between size of apo(a) isoforms and plasma concentration (17,18). However, even within alleles matched for isoform size, there exists a greater than 200-fold range of plasma concentration (19 -22). Transcription control is a likely cause of this variation. Numerous studies have found a large variation in apo(a) mRNA quantity in humans and primates, which corresponds with plasma concentration in most cases (23)(24)(25)(26). We have previously shown that the liver-specific expression of the apo(a) gene is partly mediated by the apo(a) proximal promoter, which is located from Ϫ98 to ϩ130 relative to its transcription start site and responds to the liver-enriched transcription factor HNF-1␣ (27). However, there are several indications that additional transcriptional regulatory elements must contribute to the high levels of apo(a) gene expression: activity of this core promoter is relatively weak in in vitro transfection assays, it does not account for the sex hormone responsiveness of the apo(a) gene, and it contains far less sequence polymorphism than could account for the inherited variation in gene expression, which is known to map to this locus. A further indication that regulatory sequences exist further 5Ј to the apo(a) gene comes from transgenic mouse experiments. Deletion of the region 5Ј to the apo(a) promoter from a yeast artificial chromosome severely compromises apo(a) expression in transgenic mice compared with a yeast artificial chromosome containing approximately 80 kb of upstream sequence. 2 The apo(a) gene is part of a cluster of genes and pseudogenes on chromosome six which share over 70% sequence identity (28,29). The apo(a) and the plasminogen genes are organized in a head to head configuration separated by about 40 kb. We have systematically tested this entire intergene region for additional apo(a) regulatory elements. In this study, we describe an apo(a) transcription control region (ACR) that is located 20 kb 5Ј to the apo(a) transcription start site. The enhancer region contains an essential element responding to the members of the Ets transcription factor family. In addition, the binding of Sp1 as well as several other transcription factors around the Ets site also contributes to maximal enhancer activity. Unexpectedly, the apo(a) enhancer region resides within a LINE element. LINE 1 (L1) elements are long interspersed repeat elements that number from 50,000 to 100,000 in the human genome. Although greater than 90% of them are truncated, full-length L1s are approximately 6 kb in length and contain a 5Ј-untranslated region with internal transcription regulatory elements, two open reading frames, and a 3Ј-untranslated region that terminates in a poly(A) tail (30 -32). This is the first report of an L1 element affecting the expression of a flanking gene. In contrast, the corresponding regions from 10 other randomly selected human L1 elements were found to have reduced enhancer activity as well as one or more nucleotide differences in the core Ets binding site sequence. Hence the L1 element upstream of the apo(a) gene may represent a small subset of these repeat elements that have the ability to alter the expression of neighboring genes.

MATERIALS AND METHODS
pGL3-1.4RI and pGL3a Plasmid Construction-Plasmid pGL3-1.4RI was constructed from pGLA1 (27), with subsequent manipulation to create a unique EcoRI site and to remove unnecessary polylinker sequences. It contains human apo(a) genomic 5Ј sequences from an EcoRI site 1.4 kb upstream of the transcription start site to an MscI site 130 bp downstream of the start site, fused to the luciferase gene in pGL3basic (Promega). pGL3a is a similar luciferase reporter plasmid containing a fragment of the 5Ј-genomic sequence used above. It contains the minimal promoter of apo(a) from Ϫ98 to ϩ130 (27) and derives from pGL3-1.4RI by removal of an EcoRI-PvuII fragment and subsequent regeneration of the EcoRI site.
Bacterial Artificial Chromosome (BAC) Library Screening and Subcloning-A human genomic DNA library in the BAC plasmid (Genome Systems, Inc.) was screened with probes representing the 5Ј-flanking regions of apo(a) and plasminogen genes. One of the two positive clones (pBAC2), containing an insert of ϳ150 kb, was confirmed to have the entire intergenic region as well as portions of the coding regions of both apo(a) and plasminogen genes by Southern blot analysis with apo(a) and plasminogen cDNA probes.
EcoRI restriction fragments of pBAC2 DNA, as numbered in Fig. 1, were subcloned into pBluescript (Stratagene) either directly after digestion of pBAC2 or following PCR amplification. The pBAC2 DNA was also double-digested with SacI and SalI for the purpose of cloning fragments 8b-d into pBluescript (Stratagene), taking advantage of the fact that there is only one SalI site in the entire pBAC2 insert. Fragments cloned into pBluescript were excised with EcoRI, SacI, or KpnI, and subcloned into pGL3a. In most cases, the fragments were initially tested in only one orientation relative to the transcription direction of luciferase based on the convenience of subcloning.
Cell Culture and Transfection Assays-Human hepatoma HepG2 cells and HeLa cells were maintained in Dulbecco's modified medium supplemented with 10% fetal calf serum. 2 ϫ 10 5 cells were seeded in 12-well cell culture plates 24 h prior to transfection by a calcium phosphate co-precipitation method according to the manufacturer's protocol (Promega) as we described previously (27). Transfections were done in triplicate. Briefly, 3 g of pGL3-based luciferase expression plasmid and 1 g of the control pSV-␤-galactosidase (Promega) control plasmid, together with or without 1 and 3 g of Ets-1 expression plasmid were added to each well covered with 2 ml of culture medium. 48 h after transfection, the cells were harvested and lysed, and luciferase and ␤-galactosidase assays were performed using the Dual Light Kit (Tropix, Inc.) according to the protocol provided. To control for transfection efficiency, results are reported as the ratio of luciferase to ␤-galactosidase activity.
Deletion, Mutagenesis, and Diagnostic Fragments-Fragments of genomic DNA in reporter plasmids were removed using available restriction sites. Specifically, construct nos. 3 and 6 ( Fig. 2) were the result of removing a NheI fragment (one NheI site from the vector upstream of EcoRI, and the second one located at nucleotide 386 of L1, according to the L1 numbering system of Swergold (30) from either the negative or the positive orientation of the 8b fragment (construct nos. 2 and 1) in pGL3a. Construct nos. 4 and 5 are the internal deletion products of PstI or XhoI fragments, respectively, from construct no. 2. Construct nos. 7 and 8 were generated by cutting with PvuII and EcoRI either the negative or the positive orientation of the fragment 8bcontaining plasmid. Construct nos. 9 and 10 were generated by inserting a PCR subfragment of 8b into the EcoRI site in pGL3a in both positive or negative orientations. This fragment was generated with PCR primers L1-3 (gatcgaattcGCTAGCACAGCAGTCTGA) and L1-4 (gtacgaattcAGCTGCGTGCTGGGAGA) to yield the enhancer-containing region corresponding to nucleotides 386 -659 of L1 (lowercase letters denote added linker sequences).
Site-specific mutagenesis was carried out according to the mismatch PCR primer protocol (33) with oligonucleotides L1-3 and L1-4 as the outside primers. The internal primers contained the mutations described in Fig. 5A. The fragments were digested with EcoRI and subcloned into pGL3a and confirmed by sequencing prior to transfection assays. To test the enhancer containing fragment 8b with a heterologous promoter, it was ligated into the MluI and BglII sites of the pGL3-promoter vector (Promega) upstream of the SV40 early promoter.
Footprint and Gel Shift Analysis-Nuclear extracts from HepG2 cells were prepared as described previously (27). Gel mobility shift assay and DNase I footprint protection analysis protocols were described previously (34,35). The sense strand of oligonucleotide probe (containing site A; underlined) used for gel mobility shift assay is CAAAGCAGCCCG-GAAGCTCGAAC. The oligonucleotides containing mu-1, mu-2, and mu-3 used for the competition assays were the same as site A except mutated at the putative Ets site as indicated in Fig. 3A. The DNase I footprint probes were generated from construct nos. 9 and 10 (denoted ACR(ϩ) and ACR(Ϫ); plus and minus strands from the positive or negative orientations of ACR) by first end-labeling the MluI site with ␥-32 P, and the radiolabeled inserts were then released by digestion with BglII. The sequence of the Sp1 oligonucleotide used in the footprint competition assay is aattcGCTGGGGGAGGGGCGCCg.

Identification of an Apo(a) Gene Transcription Enhancer-
The genetic locus of the apo(a) and plasminogen genes is shown in Fig. 1. The apo(a) and plasminogen genes are organized in a head to head configuration with about a 40-kb intergenic region separating them (Fig. 1). This region was isolated in a BAC plasmid clone, designated pBAC2, which contains a 150-kb insert including the whole intergenic region plus parts of the apo(a) and the plasminogen genes (Fig. 1). As described under "Materials and Methods," fragments encompassing the intergene region were ligated in front of the apo(a) minimal promoter and luciferase reporter cDNA. We previously described the core promoter region that spans nucleotides Ϫ98 to ϩ130 relative to the transcription start site of the human apo(a) gene and responds to the liver-enriched transcription factor HNF-1␣ (27). Only two fragments produced activity significantly above the parent reporter plasmid containing only the core apo(a) promoter. One of these contains a previously described plasminogen enhancer located 2 kb upstream of the plasminogen transcription start site (36). It is not surprising to find that it enhances transcription in vitro from both apo(a) and plasmin-  (28,29). The genes are represented by shaded boxes with the direction of transcription indicated by arrows. The pBAC2 insert is indicated beneath. Below is a blowup of the region between the apo(a) and the plasminogen genes, with EcoRI restriction sites (E) bounding numbered fragments. Fragments 1 and 13 contain the apo(a) and plasminogen gene transcription start sites, respectively. Fragment 8 is further divided by the restriction sites indicated. The location of the LINE (L1) element is indicated with its predicted direction of transcription and "open reading frames" (shading) indicated. However, sequence analysis shows these canonical reading frames to be interrupted by numerous stop codons. Size scales are shown at top and bottom.
ogen core promoters, since the two promoter regions contain about 90% sequence identity. The physiological relevance of the plasminogen enhancer on apo(a) gene expression remains to be evaluated. The most dramatic effector of apo(a) transcription was a 1.8-kb SacI-KpnI fragment that is located about 20 kb upstream of the apo(a) promoter (Fig. 1, fragment 8b; Fig. 2, construct nos. 1 and 2). It coincides with one of the four liverspecific DNase I hypersensitive sites, DH III, reported by Magnaghi et al. (37). This 1.8-kb fragment is capable of enhancing apo(a) transcription in an orientation-independent manner and was therefore studied in detail.
Minimal Enhancer Region Is Defined by Deletion Analysis-To locate the enhancer elements in this region, it was subjected to deletion analysis. As shown in Fig. 2, construct no. 7, deletion of the KpnI-PvuII fragment at the 5Ј end of the 1.8-kb SacI-KpnI fragment (orientation relative to the apo(a) gene) resulted in a total loss of the enhancer activity, indicating the presence of essential positive elements within this region. By itself, this KpnI-PvuII fragment is sufficient to confer an over 10-fold activation (Fig. 2, no. 8). A further truncation (no. 6) yields a fragment without activity. We postulated that the functional enhancer domain resides within the 274-bp NheI-PvuII fragment, which gives a greater than 10-fold activation when ligated in either orientation 5Ј to the apo(a) core promoter and luciferase cDNA (Fig. 2, nos. 9 and 10). We designated this the apo(a) transcription control region (ACR). Deleting substantial amounts of the 1.8-kb SacI-KpnI fragment downstream of the ACR (construct nos. 4, 5, 8, 9, and 10) resulted in an increase of activity compared with the entire fragment, due either to potential silencer elements in these regions or to the nonspecific reduction in luciferase activity caused by increasing the size of the reporter plasmid.
Liver-enhanced transcription of the apo(a) gene is deter-mined chiefly by the core promoter region and not the ACR. We had previously shown that the apo(a) core promoter can drive transcription in hepatic HepG2 cells, but not in kidney-derived 293 cells or HeLa cells (27). When ligated to the core promoter, the ACR is able to direct robust transcription in HepG2 cells, but produces luciferase activity barely above background in 293 kidney cells or HeLa cells. In contrast, when ligated to the SV40 early promoter, it can stimulate transcription in all three cell types (data not shown). ACR Is Part of a LINE Element-DNA sequence analysis surprisingly revealed that the ACR is located in the 5Ј-untranslated region of a highly repetitive L1 element. The ACR (NheI-PvuII fragment) is located from nucleotide 386 -659 relative to the transcription start site of the consensus L1 element reported by Swergold (30). This raised the question whether the enhancer activity is specific to ACR or is a universal property of all the L1s. To this end, primers L1-3 and L1-4, designed according to consensus L1 sequences, were used to generate by PCR this 274-nucleotide region from 10 random human L1 elements. When subcloned into the pGL3a expression plasmid, each of the random L1 fragments produced from 20 to 50% of the luciferase activity shown by the ACR construct in transfected HepG2 cells (data not shown). Sequence analysis of the 10 randomly selected L1 fragments revealed several scattered differences, consistent with the estimated 5% sequence difference among human L1 elements. Only two sites (Fig. 5B, sites  A and B, shaded) were found to be consistently different between the ACR and the random L1s. The ACR is the only one that contains a CCCGGAAG in its sequence within site A, while each of the random L1s has either CG, CA, TG, or GG at this position. At site B, the sequence of the ACR is a CA, but is GG in random L1s. (Site A, but not site B, was shown to be functionally active; see below.) Thus, the ACR may represent a FIG. 2. Deletion analysis to define the enhancer region. Panel A, schematic representation of the luciferase expression plasmids. The test DNA fragment (filled bar) was ligated upstream of the apo(a) core promoter from nucleotide Ϫ98 to ϩ130 (hashed bar) and luciferase cDNA. The thin line represents regions deleted. To allow reference to the restriction sites as indicated in construct no. 1, all deletion constructs are drawn in the same, positive orientation. However, constructs were either tested with these fragments ligated in the same (positive) or opposite orientation as indicated by a ϩ or Ϫ sign in the column to the right. Panel B, luciferase activity in transfected HepG2 cells. The relative luciferase activity, normalized to ␤-galactosidase activity, is shown for each corresponding construct. The data represent the mean of three transfections plus the standard deviation. The activity of pGL3a, containing only the core promoter fragment (construct 0), was set to 100%. genetically different class of L1 element that can serve as a more efficient enhancer of, at least, the apo(a) gene when located in its proximity.
An Ets Element Is Critical for the Enhancer Activity of ACR-The sequence containing the site A, unique to the ACR, resembles the consensus binding site for the Ets family of transcription factors (38). Three mutations were created to test its importance in enhancing expression from the apo(a) core promoter. As shown in Fig. 3A, one, two, and five nucleotide substitutions increasingly disrupted the activity of the ACR. The effect of Ets-like proteins as trans-acting factors was demonstrated by co-transfection of the ACR reporter plasmid ACR(Ϫ) with an expression plasmid for the Ets-1 protein. The Ets-1 expression plasmid increased ACR-driven luciferase expression in a dose-dependent manner (Fig. 3B). However, Ets-1 protein fails to activate the ACR-driven expression to the same extent when the Ets binding site is disrupted. The residual enhancing activity produced by co-transfection with Ets-1 plasmid in this case may be the result of other weak Ets binding sites present in the ACR. Co-transfection of the Ets-1 plasmid had little effect on the activity of the core promoter region contained in pGL3a. These data strongly suggest that members of the Ets transcription factor family are capable of enhancing apo(a) gene expression through site A in the ACR.
Further confirmation of the importance of this site was provided by gel mobility shift assays. An oligonucleotide encompassing site A was used as a probe to incubate with nuclear extracts prepared from HepG2 cells (Fig. 3C). A specific DNAprotein complex was observed that was effectively competed by a 100-fold excess of unlabeled wild type oligonucleotide (lane 3) but increasingly less so by unlabeled mutant site A oligonucleotides mu-1, mu-2, and mu-3 (lanes 4 -6) and an irrelevant oligonucleotide (lane 7), consistent with the importance of an intact Ets-like factor binding site in the apo(a) enhancer.
Footprint Analysis Revealed Many Protein Protection Sites-DNase I footprint analysis of the ACR was carried out to identify other DNA binding sites that may contribute to the ACR enhancer activity. Nuclear extracts prepared from HepG2 cells protected eight regions of the 274-nucleotide ACR region from digestion with DNase I (Figs. 4 and 5). Computer analysis of the ACR sequence revealed similarity to the consensus binding sites of transcription factors in most of the protected regions: AP-4 (footprint I); Sp1 (footprints II, V, and VI); CRE-BP (III); HNF-3␤ (III, VII), Ets-1 (IV); and Oct-1 (VIII). Footprint IV co-localizes with site A, the previously identified putative Ets binding site. A hallmark of Ets binding to the major groove of DNA is a DNase I protection region in the sense strand and a hypersensitive site on the antisense strand, similar to the FIG. 3. A potential Ets binding site is important for ACR activity. Panel A, the sequences at wild type site A and three mutations of site A (mu-1, mu-2, and mu-3) are given. Luciferase activity relative to pGL3a (core promoter fragment only) is shown for the expression plasmid containing the wild type 274-bp apo(a) control region ligated in negative orientation before the core promoter fragment (ACR(Ϫ)), and the identical fragment with one of the three mutations in the Ets-1 binding target sequence. Panel B, the effects of Ets-like proteins on the ACR were tested by co-transfection of either the wild type ACR(Ϫ) plasmid or the mutated mu-3 construct with 0, 1, or 3 g of a plasmid expressing the Ets-1 protein driven by the cytomegalovirus promoter. Results are the mean of three transfections with the standard deviation. Panel C, gel shift assay was performed by incubating HepG2 nuclear extracts with a radiolabeled oligonucleotide containing wild type site A (corresponding to nucleotides 474 -496 in Fig. 5A)  pattern seen in Fig. 4 (38).
The specificity of footprints II, V, and VI was demonstrated by a competition assay. When 100-fold unlabeled oligonucleotide that spans the footprint II region (named Sp1 oligo, since it matches the Sp1 consensus binding site, see Refs. 39 and 40) was incubated together with the ACR probe (sense strand) and nuclear extracts from HepG2 cells, footprints II, V, and VI were abolished, confirming the specific occupancy of these sites by Sp1-related proteins in HepG2 nuclei (Fig. 4, lanes 8 and 9). Thus, in addition to the Ets protein, transcription factor Sp1 is implicated in binding to the ACR. The identity of other binding proteins in this region awaits confirmation.
Mutagenesis of DNase I Protected Sites in the ACR-To assess the effects of the other footprint binding sites within the ACR, we subjected the entire ACR to further deletion and mutagenesis analysis. Mutations were created in these sites and tested in transfection assays in HepG2 cells. Fig. 5A shows each of the DNase I-protected sites boxed and numbered from I to VIII with mutations shown underneath in lowercase letters for base substitutions or as a line denoting a deletion mutation. As shown in Fig. 5B, mutations of footprints II, III-2, and VI, as well as deletion of the first 50 nucleotides, which eliminates all of footprint I and most of footprint II, do not cause a significant reduction in luciferase expression, compared with those of the wild type ACR construct. Deletion of the region including footprints VII and VIII also had no significant effect (data not shown). (It is noteworthy that the mutation in footprint VI alters site B, which was earlier noted to be one of the two sites where the sequence of the ACR differed from the 10 random L1 examples.) In contrast, destruction of the binding sites for footprints III-1, III-3, IV, and V significantly reduce the enhancer activity. The identity of the transcription factors which bind to footprint III-1 and III-3 are unknown. Footprint IV represents the target recognition site for a member of the Ets family, while footprint V is a likely target of an Sp1-related protein. DISCUSSION We report the screening of the 40 kb that lie between the 5Ј ends of the human apo(a) and plasminogen genes for cis elements other than the apo(a) core promoter that regulate the expression of the apo(a) gene in an in vitro system. We have discovered an apo(a) transcriptional control region (ACR), located midway between the apo(a) and the plasminogen genes, that is able to confer a greater than 10-fold activation when linked in either orientation to the apo(a) minimal promoter. Sequence analysis of the ACR revealed it to be part of a LINE repeat element, whose members number up to 100,000 in the human genome. Functional analysis of 10 randomly selected, corresponding regions of other human L1 elements show that they are less efficient than the ACR in enhancing the activity of the apo(a) core promoter, and that they contain sequence differences in a site corresponding to the consensus target of the Ets family of transcription factors. We also inspected 60 additional L1 sequences available in GenBank TM , and found that  (30). Boxes indicate the footprints as labeled in Fig. 4. Shown in boldface type are the nucleotide pairs at sites A and B that have consistent differences between the sequence of ACR and the 10 random genomic L1s that were shown to have reduced enhancer activity. In vitro mutations are indicated by the lowercase letters beneath the wild type sequence, or by a line indicating the deletion mutation. Numbers 1-3 indicate three different mutations that were tested within footprint III. Panel B, relative luciferase activity as described above is given for the wild type ACR fragment and the mutations indicated in panel A. they differ from the ACR in the Ets target sequence. Mutagenesis showed this Ets binding site to be indispensable for the apo(a) enhancer activity. Co-transfection of an Ets-1 expression plasmid caused a dose-dependent activation of ACR-driven expression, further implicating the role of these trans-acting proteins. DNase I footprint analysis and further site-directed mutagenesis indicate the involvement of additional factors including Sp1 in the ACR transcription enhancer activity and suggest likely interactions between several DNA binding proteins in the ACR.
Previous data from our laboratory have shown that the apo(a) minimal promoter is mediated by a liver-enriched transcription factor HNF-1␣ and can confer liver specific expression (27). In many cases, both promoters and more distant enhancer elements control the tissue specific expression of genes. In the case of apo(a), the ACR does not appear to be liver-specific, suggesting that the apo(a) proximal promoter is sufficient to determine tissue specificity of its expression. (Liver represents the sole organ of significant apo(a) production; much reduced levels of its mRNA have also been detected in brain and in testes of rhesus monkeys (41), but upon liver transplant, human subjects entirely convert their isoform of circulating apo(a) from recipient to donor phenotype (42).) The ACR may function only by increasing the expression level of the rather weak, but tissue-specific apo(a) promoter. One example of an interplay between liver-specific and nonspecific elements of an opposite type is found in the apolipoprotein E gene, where tissue specificity is determined by a hepatic control region that is located 19 kb from the start of the apoE gene (43). A number of other genes are now known to be controlled by remote transcription elements located thousands of nucleotides from transcription start sites (43)(44)(45)(46). The function of the ACR must ultimately be evaluated in its normal genomic context in vivo. We have begun to engineer transgenic mice with mutagenized and normal regions of the human apo(a) locus in yeast artificial chromosomes. Human subjects can also be screened for polymorphisms or mutations in this region which can be shown to effect apo(a) expression.
Based on functional criteria and on comparison to the sequence of other L1 elements, the most notable feature of the ACR is the presence of an Ets family target site. Evidence for its functional importance derives from footprint, gel shift, and mutagenesis data, and the effect of co-transfection with an Ets-1 expression plasmid. Ets is an oncogene that was first discovered in avian erythroleukemia virus to encode a transcription factor that is capable of activating erythrocyte-specific gene expression. It is now known that there is a family of Ets-like transcription factors with a variety of tissue distributions and target genes. For example, the thrombopoietin gene, which is predominantly expressed in liver, contains a functional Ets-binding site that is regulated by a ubiquitously expressed Ets-related transcription factor, E4TF1, in hepatocytes (47). Further studies will be needed to clarify which member(s) of the Ets family are responsible for binding to the ACR in hepatocytes. Ets proteins have been reported to act synergistically with other transcription factors, including Sp1, in the regulation of many viral and cellular genes (48,49). Although an intact Ets binding site is essential for the enhancer activity of the ACR, our data indicate that binding sites for Sp1 as well as other transcription factors also contribute to the optimal function of this apo(a) gene enhancer.
The fact that the ACR resides within the 5Ј-untranslated region of a LINE sequence may have significant implications for the functions of L1 elements, as well as for the regulation of the apo(a) gene. The nearly 100,000 L1 long interspersed repeats comprise an estimated 5% of the human genome.
Roughly 10% of these represent full-length, 6-kb elements, while only an estimated 30 -60 are active elements that are capable of transposition via reverse transcription of an RNA intermediate (50). The LINE sequence within the apo(a) locus appears to be full-length. However, it does not meet the criteria of a mobile L1 element, as it contains numerous stop codons in both consensus open reading frames necessary for such activity. Nor is it a member of the Ta subclass of L1 elements to which most of the mobile elements belong (50,51). Although it is unable to code for active mobility proteins, our data indicate that this particular L1 element is capable of interacting with DNA binding proteins and influencing the expression of an adjacent gene.
It has been proposed that methylation of CpG dinucleotides in the 5Ј region of L1 sequences interferes with their expression (52). Methylation at such sites may interfere with the interaction of DNA-binding proteins such as Ets (53). Southern blotting with methylation-sensitive and insensitive restriction enzymes and a probe 5Ј to the apo(a) gene L1 element indicate that the CpG dinucleotide in the Ets binding site (site A) is unmethylated in human liver cells (data not shown). This is consistent with an in vivo interaction of an Ets family protein with this site.
There have been several reported cases in which de novo insertions of L1 sequences in either the germline or somatic cells have interrupted genes and caused human disease (54 -58). To our knowledge, we describe the first example of a LINE element that acts as a cellular enhancer to stimulate the expression of a nearby gene. In a related example, a subclass of Alu DNA repeats has been found to function as an estrogen receptor-dependent enhancer (59). This raises the possibility that L1 elements may serve as potentially mobile gene control elements that could influence the evolution of gene expression in mammals, as well as help to control the expression of the potentially atherogenic apo(a) gene.