INTRODUCTION

The human genome contains up to one million copies of a sequence element known as the Alu repeats, comprising ~5% of the DNA. These elements are potentially functional class III genes, transcribable by RNA polymerase III, originally derived from a 7SL gene. During the preceding 30-50 million years of primate evolution, Alu transcripts were converted into cDNA by a reverse transcription mechanism and reinserted randomly throughout the genome in retroposon fashion. An evolutionary lineage of highly conserved Alu source genes is thought to have provided the transcripts, which were converted into retroposons (1, 2). The Alu elements can be grouped into several subfamilies, each representing the progeny of one Alu source gene. The degree of sequence divergence within an Alu subfamily is a measure of the time elapsed since those Alu retroposons were introduced into the genome, with subgroups I-IV representing the oldest to the most recent retroposition events.

There has long been suggestive evidence that Alus or other polymerase III genes may function in gene regulation (3, 4, 5, 6). More recent evidence implicates specific Alu elements as influencing the expression of nearby genes (7, 8, 9, 10, 11, 12). We reported that a major class of Alu elements (subclass III-IV) has evolved to encode a composite hormone response element (HRE)¹ overlapping the internal promoter of the Alu (7). These several hexamer sequences are related to a consensus binding site, AGGTCA, recognizable by members of the nuclear receptor superfamily of ligand-activated transcription factors, including receptors for retinoic acid (RA) (RAR), thyroid hormone (T3) (TR), vitamin D (VDR), steroids, and a number of orphan receptors for which a ligand has not yet been identified (13, 14, 15, 16). Most of these nuclear receptors recognize two adjacent hexamer half-sites that can be arranged as direct repeats (DR), palindromes, or inverted palindromes. The spacing and orientation of the half-sites is a major determinant of receptor binding specificity (17, 18, 19). Receptors can bind DNA as monomers to a single half-site, as homodimers to two adjacent half-sites, or as heterodimers with the retinoid X receptors (RXRs). The RAR/TR family requires heterodimerization with RXRs for effective DNA binding and function (20, 21, 22, 23, 24).

An Alu preceding the human keratin K18 gene was shown to contain four half-sites oriented as direct repeats and spaced by 2 base pairs (DR-2) (7), consistent with the binding characteristics of RAR-RXR. DNA binding and transfection studies confirmed that these sites included two overlapping retinoic acid response elements (RARE), which were recognized by RAR-RXR heterodimers and functioned as positive activators of CAT gene expression in transient transfection assays. Those findings and previous studies (25) implicated this Alu-RARE in the regulation of the keratin K18 gene. Alu elements with encoded HREs could thus represent mobile enhancers containing recognition sites for RARs or other types of nuclear receptors. Random insertion of hundreds of thousands of these elements during primate evolution is likely to have affected the regulation of numbers of genes and may have represented an important source of genomic plasticity. To obtain evidence as to the extent to which Alu-HREs regulate neighboring genes, a literature search was performed that revealed a number of Alu elements with potential regulatory function, one being upstream of the human myeloperoxidase gene (MPO) (26). The MPO gene is expressed specifically in the myeloid lineage (27). The encoded myeloperoxidase enzyme is localized in lysosomal vesicles and catalyzes the reaction between hydrogen peroxide and chloride ions to form hypoclorous acid, a potent antibacterial agent and generator of oxidizing free radicals (28, 29). The MPO gene is transcribed in the early stages of myeloid cell differentiation, myeloblasts, and promyelocytes; transcription ceases when these cells differentiate into mature granulocytes or monocytes/macrophages (27, 30, 31, 32, 33). MPO is also expressed in acute myelocytic leukemic cells, which represent a clonal expansion of early myeloid cells blocked in their ability to differentiate.

Acute myelocytic leukemias (AML) account for 46% of all major leukemias and are subdivided into six subtypes, M1 through M6, distinguishable by cytochemical, morphological, and immunological characteristics (34). While the myeloperoxidase gene is expressed in all six subtypes of AML, the levels of expression vary considerably with the highest level seen in M3, or acute promyelocytic leukemia (APL) (35). This M3-APL subtype is associated with a chromosomal translocation t(15:17) fusing the PML gene to the RAR gene. The resultant PML-RAR fusion protein retains most of the functional domains of the parental proteins (36, 37, 38, 39) and is thought to be instrumental in leukemogenesis. APL cells respond to RA treatment by differentiating with loss of proliferation. Most clinical cases of APL can be driven into remission by treatment with all-trans RA (40).

A single A/G base transition within the Alu element preceding the MPO gene was previously reported to be associated with most cases of AML (26). We noted that this base difference was within the HRE region, and the AML-associated residue resulted in an improved fit to the half-site consensus sequence. This raised the possibility that this base transition enhanced binding by a nuclear receptor or other transcription factor, altering MPO gene expression in a way that might potentiate the development of AML. To investigate this possibility, we tested the ability of several nuclear receptors to bind these elements and activate transcription in vivo. Our results indicate that the MPO-associated Alu includes a RARE as well as a T3 response element (TRE). Interestingly, a strong binding site for the general transcription factor SP1 is created by the single base difference associated with acute myelocytic leukemia.

MATERIALS AND METHODS

The oligonucleotides encoding the different response elements from the Alu sequence were annealed and inserted upstream of the CAT gene in the pBLCAT2 vector between the SalI and XbaI sites (41). One copy of each response element was present as verified by DNA sequencing. The receptor expression vectors used in this study, pECE-RAR, pECE-RXR, and pECE-TR, have been described elsewhere (Ref. 20 and references therein).

CV-1 cells (10⁴ cells/well) were plated in 96 well plates in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, 2 mM glutamine, and penicillin/streptomycin. The cells were transfected 24 h later by the calcium phosphate precipitation method. The transfection mixture contained 100 ng of reporter pBLCAT2 plasmids and the indicated receptor expression vectors. The amount of total DNA was adjusted to 200 ng with pBluescript. 24 h after DNA addition, the medium was removed, and the cells were treated with retinoic acid (1 µM) or thyroid hormone (T3) (0.1 µM) for an additional 24-h period prior to harvesting. CAT activity was assayed by a standard phase extraction method using [³H]acetyl coenzyme A and chloramphenicol and normalized against total protein concentration.

NB4 cells (42) were grown in RPMI medium supplemented with 10% heat-inactivated fetal calf serum. For transient transfections, cells were electroporated using 2 × 10⁶ cells in 400-µl and 50-µg reporter pBLCAT2 plasmids, using a BTX electroporator (capacitance, 500 microfarads; 200 V; pulse length, 20 ms).

Receptor proteins were produced by in vitro transcription-translation from the corresponding linearized pBluescript vector using rabbit reticulocyte lysate (Promega) as described previously (43, 44). To normalize for amounts of proteins, reactions containing [³⁵S]methionine were performed in parallel, and the products were analyzed by SDS-polyacrylamide gel electrophoresis. Nuclear extracts were basically prepared as described (45). The final high salt extract was dialyzed against buffer containing 20 mM Hepes, pH 7.9, 100 mM KCl, 1 mM dithiothreitol, 1 mM phenymethylsulfonyl fluoride, and 20% glycerol. Protein concentration was determined using the Bradford method (Bio-Rad). Purified human SP1 was purchased from Promega. Proteins were incubated with 1 µg of poly(dI-dC) in binding buffer (10 mM Hepes, pH 7.9, 50 mM KCl, 1 mM MgCl₂, 1 mM dithiothreitol, 10% glycerol) for 15 min at room temperature. ³²P-Labeled oligonucleotides (20,000 cpm) were then added, and incubation continued for 20 min at room temperature. A 5% non-denaturing polyacrylamide gel was used to analyze the protein-DNA complexes (44). When antibodies were used, proteins were mixed with 1 µl of non-immune serum, polyclonal RXR (kindly donated by A. Lombardo and K. Ely), or SP1 (Santa Cruz) for 30 min on ice before incubation with the DNA. In the competition experiments, the indicated amounts of nonlabeled oligonucleotides were added to the binding reaction together with the poly(dI-dC).

PCR was performed with 200 ng of genomic DNA isolated from peripheral blood lymphocytes or bone marrow lymphocytes, with 0.5 µg of each primer in a 50-µl reaction volume containing 50 mM KCl, 10 mM Tris-HCl, pH 8.3, 1.5 mM MgCl₂, 200 µM nucleotides, and 2.5 units of Taq polymerase (Perkin Elmer Cetus). Nested primers (Life Technologies, Inc.) were synthesized to amplify a region extending from position +310 within the myeloperoxidase gene to position 829 bp. The first primer set was 5-CTTGGTCCTGCGCCCACAGTCCCC-3 and 5-TCCCACCTTGGGAACTGTTACCTG-3, and the second set was 5-GCTGCCCATTGGGTGGCTGTTGGA-3 and 5-AGAGGGCTGGGGCGTGGCCAGAAT-3. The cycling conditions were 94 °C for 6 min, followed by 30 cycles at 94 °C for 1 min, 55 °C for 2 min, and 72 °C for 2 min. 10 µl of the first reaction was used in a second PCR reaction with the second primer set, internal to the first set. The resultant 1129-bp PCR product was electrophoresed in agarose gels and purified. The PCR products were sequenced directly or, in some cases, ligated into a plasmid vector prior to sequencing. Each blood sample was analyzed at least twice.

RESULTS

The DNA sequences preceding the MPO gene (26, 46) include a ~300-bp Alu element situated between positions 200 and 505 (Fig. 1A). A previous report suggested that a G residue at position 463 was associated with AML more often than an A residue (Fig. 1B). That site is within a region of four potential hexamer half-sites, related to the consensus AGGTCA, located between positions 436 and 467, with the fourth half-site overlapping the B box internal promoter element of the Alu gene (Fig. 1C). All four half-sites are oriented as direct repeats with the first two separated by two base pairs (DR-2), typical of RARE; the second and third half-sites are DR-4, consistent with TRE, and the third and fourth half-sites are again DR-2 (Fig. 1C). The third and fourth half-sites fit the consensus HRE sequence (Fig. 1D), while the first and second half-sites are nonconsensus at positions four and three, respectively. The G/A base difference associated with AML is at position 5 of the first potential half-site, and the G residue creates a better fit to the HRE consensus (19).

Acute Myelocytic Leukemias Are Predominantly Homozygous for a G at Position 463

The A/G transition at position 463 in acute myelocytic leukemias has been suggested to arise by somatic mutation (26). Alternatively, this could represent allelic polymorphism in the population, with a preference for the G allele in leukemic cells. To distinguish between these possibilities, we analyzed the sequences of DNA isolated from lymphocytes from 18 AML patients and 44 control blood donors. Nested oligonucleotide primers were designed to allow PCR amplification of sequences extending from 829 to +310 of the MPO gene. Sequence analysis showed that 50% of the normal donors (22 of 44) were homozygous for the G residue at position 463, 18 were heterozygous, and 4 were homozygous for the A residue, indicating an allelic polymorphism. Examination of the DNA sequences from the AML samples revealed that 13 out of 18, or 72%, were homozygous for the G residue at 463, indicating a preference for that allele. This preference was most pronounced in the M3 and M4 subclasses (6 of 7, or 85%). The M3 subclass produces the highest levels of MPO gene expression, severalfold above that of M4 and M2, and as much as 20-fold above the levels seen in M1 and M5 (35). This suggests a correlation between the G-containing allele and AMLs with high levels of MPO gene expression. In this study, the allele with the G residue at position 463 is designated AML, while the allele with A at position 463 is designated MPO (Fig. 1B).

The composite MPO-Alu HRE contains two DR-2 elements, which are potential RAREs, and one DR-4 element, a potential TRE. To assay the ability of the Alu DR-2/DR-4 elements to function as RARE or TRE, these were inserted upstream of the CAT reporter gene in the pBLCAT2 vector. The pBLCAT2 reporter constructs containing the composite HRE, or the individual DR-2 or DR-4 elements (Fig. 1C), were transiently transfected into CV-1 cells in the presence or absence of cotransfected expression vectors encoding RAR and RXR (Fig. 2A). The AML12 element activated transcription of the CAT gene by 25-fold in the presence of both RAR and RXR expression constructs and all-trans RA. This level of activation was severalfold higher than that obtained with MPO12. The other DR-2 element, MPO34, produced a ~35-fold transcriptional activation in the presence of cotransfected receptors and RA. In the absence of cotransfected RAR/RXR, 2-3-fold lower levels of activation were observed for both AML12 and MPO34. The DR-4 element, MPO23, had little to no effect on CAT gene expression under these conditions. Curiously the activation obtained with the AML1234 element, containing all four half-sites, was significantly below the ~60-fold predicted additive effect of the AML12 and MPO34 elements, suggesting competition or interference with binding or activation when all four overlapping half-sites are present. The activation obtained with AML1234 was about twice that obtained with the MPO1234 element. These findings establish that the DR-2 elements, AML12 and MPO34, function as strong activators of CAT gene expression in CV-1 cells and provide the greatest transactivation in the presence of cotransfected RAR and RXR as well as RA.

To test the ability of the central DR-4 element, MPO23, to function as a TRE, CAT constructs carrying this element were transfected into CV-1 cells along with expression vectors for thyroid hormone receptor (TR) and RXR in the presence or absence of thyroid hormone (T3). Under these conditions, the MPO23 element activated transcription by ~7-fold in the presence of TR/RXR and T3, indicating the MPO23 element can function as a positive TRE (Fig. 2B).

We analyzed the binding of several nuclear hormone receptors to the MPO-Alu sites using a gel retardation assay. In vitro translated proteins were incubated with a ³²P-labeled oligonucleotide containing all four MPO half-sites in the presence or absence of in vitro synthesized RXR (Fig. 3A). TR, TR, RAR(, ,) formed strong complexes with the DNA. Binding was also found with the oncoprotein v-ErbA, a mutated form of TR, but not with the TR variant, TR2. Binding by these receptors required the presence of RXR, consistent with previous studies (18, 20, 47). The ARP-1 receptor, bound independently of RXR and TR monomer, was also found. Binding to AML1234 was also investigated, and identical patterns were observed (data not shown).

To determine which of the DR elements are recognized by these various receptors, the elements were tested as individual dimer sites in gel retardation experiments. The central DR4 element, MPO23, formed a strong complex with TR, TR, and v-ErbA, all requiring RXR (Fig. 3B). TR also formed a relatively weak monomer complex in the absence of RXR, and the mobility and intensity of this complex increased in the presence of 100 nM T3. ARP-1 bound independently of RXR, while RAR and TR2 failed to bind (Fig. 3B). Conversely, the DR2 element, MPO34, formed a complex with all three isoforms of RAR, all requiring RXR, but did not bind TR (Fig. 3C). The ARP-1 receptor bound in the presence or absence of RXR. These findings are consistent with the known binding preferences of RAR-RXR for DR-2 elements and TR-RXR complexes for DR-4 elements.

Surprisingly, the DR2 elements, AML12 and MPO12, did not form a complex with in vitro synthesized RAR-RXR (Fig. 3D) nor with the other isoforms, RAR and RAR (data not shown). This was unexpected since the AML12 element is a strong transcriptional activator of the CAT reporter gene in transfection assays and produced the highest level of transactivation in the presence of RAR-RXR expression vectors and RA (Fig. 2A). Other nuclear hormone receptors (ARP-1, TRs) were assayed and also failed to bind this potential DR-2 element (data not shown).

Because AML12, and to a lesser extent MPO12, activates transcription of a CAT reporter gene in transient transfection assays (Fig. 2A), transcription factors recognizing these elements presumably exist in CV-1 cells. To test for such factors, nuclear extracts were prepared from cell lines including CV-1, NB4, CaSki, and HeLa, treated or untreated with 1 µM RA (Fig. 4A). These extracts were incubated with ³²P-labeled AML12 and MPO12 elements, and the protein-DNA complexes were analyzed by electrophoretic mobility shift assays. Four complexes, termed I-IV, were observed with the MPO12 element, most clearly seen with the CaSki nuclear extracts. With AML12, only three of these complexes, I, II, and IV, were observed. The amount of complexes I, II, and IV formed was severalfold greater with the AML12 element than with MPO12, indicating the single base difference in the first half-site results in higher affinity binding. This single base difference also results in specific binding by complex III to the MPO12 element. Treatment with RA had no effect on the efficiency of complex formation with extracts from CV-1, CaSki, and HeLa cells. However, in the AML-M3-derived cell line, NB4 (42), RA treatment resulted in a strong inducement of complex III to MPO12, and also binding of complexes I, II, and IV to AML12 was increased.

To test whether complexes I, II, III, and/or IV contain RAR or RXR, competition studies were performed. Binding of HeLa nuclear proteins to the AML12 element was competed by a 250-fold excess of unlabeled oligonucleotides containing AML12 or MPO12 sequences (Fig. 4B), but binding was not competed by the MPO34 sequence, which binds RAR-RXR heterodimers (Fig. 4B and 3C). This argues that complexes I, II, and IV do not contain RAR-RXR heterodimers. Incubation of the labeled MPO34 element with HeLa nuclear extracts produced a complex with mobility equivalent to that of in vitro synthesized RAR and RXR. This complex was not competed by an excess of unlabeled AML12 or MPO12, further indicating these elements do not bind RAR-RXR. As further evidence, we investigated the effect of several antibodies raised against RXR, RAR//, TR/, and COUP. The binding of nuclear extracts to AML12 and MPO12 was not affected by incubation with any of these antibodies (data not shown), further confirming the absence of those receptors in the complexes observed with nuclear extracts.

Examination of the AML12 sequence revealed a perfect match to a 10-base pair consensus binding site for the general transcription factor SP1 (48, 49) (Fig. 5A). The nucleotide difference between AML12 and MPO12 is within the core SP1 binding site, converting GGCGG to GGCGG. To investigate whether complexes I, II, III, and/or IV contain SP1 protein, purified SP1 was incubated with labeled AML12 and MPO12 elements. A single complex was observed that had a mobility equivalent to complex I. As observed with the nuclear extracts, SP1 binding efficiency was severalfold greater for AML12 than MPO12. These findings suggest that complex I contains the SPI transcription factor.

To gain further evidence for the presence of SP1 in complex I, we tested the effect of anti-SP1 antibody on binding by purified SP1 protein and HeLa nuclear extracts (Fig. 5B). In both cases, the SP1 antibody inhibited complex I formation and produced a supershifted complex. Complexes II and IV were also inhibited, although not as completely, suggesting these complexes may also contain an SP1-like protein. As controls, an antibody directed against RXR and non-immune serum failed to inhibit complex formation. These findings indicate that proteins in complex I, and to a lesser extent complexes II and IV, are antigenically related to SP1. Complex III, which specifically binds to MPO12, did not immunoreact with the anti-SP1 antiserum (data not shown).

As another means to assay for the presence of SP1 protein in complexes I, II, and IV, the complexes were competed with unlabeled oligonucleotides containing the AML12 sequence or the consensus SP1 binding site (Fig. 5C). The SP1 oligonucleotide specifically and completely competed with proteins in complex I as well as complexes II and IV. The SP1 consensus was in fact a more efficient competitor than the AML12 sequence. A nonspecific oligonucleotide, with no nuclear receptor or SP1 binding sequences, was unable to compete for binding of any of these complexes (not shown). As further evidence that the MPO12-specific complex III is not SP1 related, binding was not competed by an excess of cold SP1 oligonucleotide (data not shown). These findings indicate that proteins in complexes I, II, and IV, but not complex III, are SP1-like in their DNA binding preference. Complexes II and IV may be degradation products or other related proteins, since several SP1-related proteins have been identified (50).

To further compare the binding characteristics of nuclear proteins in complexes I, II, and IV with SP1 protein, mutations were introduced into the AML12 and MPO12 elements (Fig. 6A). In the mutant oligonucleotide A1m, four nucleotides within the first potential half-site and the SP1 core consensus were changed from GGCG to TTAT. This resulted in complete loss of binding by the purified SP1 protein and loss of complexes I, II, and IV from HeLa and NB4 nuclear extracts. In mutant A2m, changing four bases within the second half-site, outside the SP1 binding site, did not alter complex formation by SP1 or the nuclear extracts. Mutant A3m contained two base changes, one within the core SP1 site and one in the second half-site of the DR-2, converting the first two potential half-sites to the consensus HRE sequence; these base changes abolished binding by both SP1 and nuclear extracts. Another mutant, A4m, had one base change in the spacer between the two half-sites and within the 10-base pair SP1 consensus. This mutation severely reduced binding by both SP1 and nuclear proteins. The observation that this series of mutations similarly affected binding efficiency of SP1 and complexes I, II, and IV argues that the latter contain SP1 or proteins related to SP1.

To test whether abolition of SP1 binding activity correlates with loss of AML12 transcriptional enhancer activity, we tested the A1m mutant in transient transfection experiments using CV-1 cells (Fig. 6B). The four base changes in the core SP1 binding site in mutant A1m, which abolished SP1 binding, also negated activation of the CAT reporter gene. This finding suggests that SP1 plays a major role in the transcriptional activation conferred by the AML12 element.

Since the myeloperoxidase gene is specifically expressed in myeloid cells, we assayed the ability of the Alu response elements to activate gene expression in the acute promyelocytic leukemia cell line, NB4 (42) (Fig. 7). When NB4 cells were transfected with the different CAT reporters in the absence of cotransfected receptors, AML12 increased CAT gene expression by 6-fold, indicating that endogenous factor(s), possibly SP1, are present and capable of activating transcription through this element. These findings are consistent with gel shift experiments (Fig. 4A) indicating apparently similar, SP1-related complexes formed on AML12 with nuclear extracts from NB4 cells as well as CV-1 and other cell types. Curiously, RA treatment of NB4 cells eliminates the transcriptional activation through AML12, in contrast to results obtained with CV-1 cells (see Fig. 2A). This finding is consistent with the observation that MPO gene expression is down-regulated in differentiating myeloid cells (28, 29, 30, 31, 32, 33) and in RA-treated NB4 cells.² The unusual and strong RA response of NB4 cells is typical of APL and is thought to be mediated through the aberrant PML-RAR fusion protein. Unlike AML12, the MPO12 element had no significant effect on CAT gene expression with or without RA treatment, indicating the A/G transition in AML12 increases transcriptional activation in myeloid cells, as in CV-1 cells. As for the other Alu response elements, MPO34 activated transcription by 6-fold in the presence of RA and in the absence of cotransfected RARs, suggesting the presence of endogenous RAR and RXR in these myeloid cells. Similarly, MPO23 activated transcription by 6-fold in the presence of T3 and in the absence of cotransfected TRs, indicating the presence of endogenous TRs. These findings demonstrate that the response elements AML12, MPO23, and MPO34 function as transcriptional enhancers in myeloid cells and thus may potentially contribute to the regulation of the MPO gene during myeloid cell development.

DISCUSSION

The MPO-associated Alu is of the evolutionarily intermediate subgroup II, which includes the majority of Alu repeats, numbering ~ 400,000 copies (1, 2, 51). The previously reported keratin K18-associated Alu (7) was of the evolutionarily recent subgroup III-IV and contained a series of three overlapping DR-2 elements (DR-2-2-2), while subgroup II Alus have a series of DR-2-4-2 elements. As shown here, this Alu sequence includes binding sites for TRs (DR-4) as well as RARs (DR-2). The first DR-2 element in the series fails to bind RAR. Instead, in the AML allele, this element represents a strong SP1 binding site, while in the MPO allele, a single base change in the core SP1 motif significantly reduces binding. Consistent with the binding studies, the AML12 site provides a strong 25-fold transcriptional enhancement to a reporter CAT gene in transient transfection assays, while the MPO12 element is severalfold less effective.

The subgroup II Alu sequences can be thought of as mobile cassettes of regulatory elements, containing RARE and TRE, in some cases overlapping an SP1 site. These sites are present in the Alu consensus sequences thought to represent the source gene sequence (1, 2, 3), implying these RARE/TRE sites existed prior to dispersal of the progeny retroposons, thus prior to the insertion of the Alu upstream of the MPO gene. Since the RARE/TRE/SP1 sites did not arise in response to selective pressures related to MPO gene regulation, these sites may not all contribute significantly to the regulation of the MPO gene. The SP1 site probably has important influence on MPO gene regulation because this sequence has improved to become a 100% match for the 10-bp SP1 consensus, as compared to 8 or 9 out of 10 matches in most Alu elements of this subfamily. Also, two alleles, with significantly different SP1 binding capability, are differentially associated with cases of acute myelocytic leukemias, suggesting this SP1 site contributes to MPO gene regulation and does so in a manner conducive to development of this leukemia.

Two alleles of the MPO gene differ at one position within the first half-site in the upstream Alu-HRE. Most AML patients are homozygous for the AML allele, which has a G residue at position five of the first half-site. This single base transition creates the strong SP1 binding consensus within AML12, which is correlated with a 25-fold transcriptional enhancement of a CAT reporter gene in transient transfection assays. The other allele has an A residue at position 5, which negates the SP1 binding site, and this element (MPO12) confers a much weaker transcriptional advantage to a reporter CAT gene. Since SP1 is a positive transcription factor, this suggests that the AML allele enhances MPO gene expression, and this expression somehow potentiates the leukemic phenotype. The MPO gene is normally transcribed in early myeloid cells (myeloblasts and promyelocytes). MPO gene expression ceases when these cells differentiate into monocytes or granulocytes, although the MPO enzyme remains stored in cytoplasmic vesicles (27, 30, 31, 32, 33). AML cells represent clonal expansions of early myelocytes, which have lost the ability to differentiate in response to normal cellular signals, and thus continue to express the MPO gene. Clearly, the SP1 site in the AML allele could promote MPO gene expression, but it is unclear why enhanced MPO gene expression should be linked to the leukemic state. The MPO enzyme, when released by granulocytes or monocytes, catalyzes the reaction of chloride and hydrogen peroxide to yield hypochlorous acid, a strong oxidant. In the presence of superoxide, released by macrophages, hypochlorous acid generates hydroxl radicals (29), which react with most biological molecules, creating secondary radicals of variable reactivity. There is evidence that production of oxygen radicals by monocytes inhibits the natural killer cell immune response (52). If so, higher expression of the MPO enzyme by leukemic cells carrying the AML allele might enable those cells to preferentially escape immune surveillance, which could explain why fewer AML patients carry the MPO allele. Another possibility is that free radicals produced by the MPO pathway result in DNA damage leading to the leukemia; myeloperoxidase has been linked to inflammation-associated cancers through DNA damage (53, 54) and the production of carcinogens (55, 56).

With the exception of the myeloid leukemic cell lines NB4 and HL60, it would be difficult to obtain isolates of early myeloid stages representing the in vivo stages of these leukemic cells. Therefore, it is difficult to show directly whether the potent RARE represented by M34 influences MPO gene expression in vivo in these early myeloid stages. There are, however, several lines of evidence indicating that RAR plays an important role in myeloid cell differentiation and in the acquisition of acute myelocytic leukemia. First, retinoic acid induces NB4 (M3 class) and HL60 (M2 class) cells to differentiate. A retinoic acid-resistant variant of HL60 cells was found to have a point mutation in the ligand binding domain of the RAR gene (57). Introduction of wild type RAR allowed this mutant cell line to differentiate in response to RA (58). Second, a chromosomal breakpoint in the RAR gene causes M3-AML or APL, presumably mediated by the PML-RAR fusion protein. Treatment of APL cells with RA induces rapid differentiation and loss of cellular proliferation (40). As further evidence for the involvement of RAR in the acquisition of AML, a second translocation t(11;17), which causes AML, also interrupts the RAR gene, fusing it to the Kruppel-related gene, PLZF (59). Thus, disruption of RAR function is linked to the inability to differentiate, resulting in maintenance of the undifferentiated myeloid state in which the MPO gene is expressed.

The AML12 element is not an RAR recognition site but does enhance transcription of the CAT gene most effectively in CV-1 cells in the presence of cotransfected RAR-RXR and RA, apparently by indirect means. In contrast, in NB4 cells, RA treatment results in loss of transcriptional activation by AML12, coincident with loss of MPO gene expression³ and loss of cellular proliferation. These findings suggest that AML12 enhances transcription of MPO, in a manner influenced by RA and its receptors, and abrogated by the PML-RAR induced differentiation process.

AML12 does not appear to represent a binding site for PML-RAR. Complexes I, II, and IV, which form on AML12, are observed not only in nuclear extracts of NB4 cells but also in extracts from CV-1, HeLa, and CaSki cells, which lack the PML-RAR fusion protein (Fig. 4A). As further evidence, antisera against RAR, which recognizes the PML-RAR fusion protein, failed to supershift or inhibit binding of complexes I, II, III, or IV to either the AML12 or MPO12 elements (data not shown).

An understanding of the complex, overlapping nuclear receptor/SP1 binding sites in the Alu elements will be important for determining which of the highly abundant Alu elements is contributing to the regulation of nearby polymerase II genes. Presumably, most Alu elements will have inserted too distant from genes to have an effect, while other Alu inserts may have had a negative effect, bringing a gene under control of nuclear receptors or SP1 in a way that was deleterious to the organism, and individuals carrying such Alu inserts would have been deleted from the gene pool. Conversely, some Alu insertions may have benefited the organism by bringing particular genes under control of nuclear receptors or SP1, and individuals carrying those insertions would have been retained in the population with a selective advantage. The question then is: which of the hundreds of thousands of Alu inserts are contributing to the regulation of nearby genes, and which are without significant effect?