Overexpression of PREP-1 in F9 Teratocarcinoma Cells Leads to a Functionally Relevant Increase of PBX-2 by Preventing Its Degradation*

To bind DNA and to be retained in the nucleus, PBX proteins must form heterodimeric complexes with members of the MEINOX family. Therefore the balance between PBX and MEINOX must be an important regulatory feature. We show that overexpression of PREP-1 influences the level of PBX-2 protein maintaining the PREP-1-PBX balance. This effect has important functional consequences. F9 teratocarcinoma cells stably transfected with PREP-1 had an increased DNA binding activity to a PREP-PBX-responsive element. Because PREP-1 binds DNA efficiently only when dimerized to PBX, the increased DNA binding activity suggests that the level of PBX might also have increased. Indeed PREP-1-overexpressing cells had a higher level of PBX-2 and PBX-1b proteins. PBX-2 increase did not depend on increased mRNA level or a higher rate of translation but rather because of a protein stabilization process. Indeed, PBX-2 level drastically decreased after 3 h of cycloheximide treatment in control but not in PREP-1-overexpressing cells and the proteasome inhibitor MG132 prevented PBX-2 decay in control cells. Hence, dimerization with PREP-1 appears to decrease proteasomal degradation of PBX-2. Retinoic acid induces differentiation of F9 teratocarcinoma cells with a cascade synthesis of HOX proteins. In PREP-1-overexpressing cells, HOXb1 induction was more sustained (3 days versus 1 day) and the induced level of MEIS-1b, another TALE (three amino acid loop extension) protein involved in embryonal development, was higher. Thus an increase in PREP-1 leads to changes in the fate-determining HOXb1 and has therefore important functional consequences.

TALE (three amino acid loop extension) proteins, including a PBC and a MEINOX subfamily, are homeodomain transcription factors regulating the activity of various HOX proteins in segmentation, differentiation, and organogenesis (1)(2)(3)(4). Proteins of the PBX subfamily (Extradenticle in Drosophila, lazarus in zebrafish, CEH20 in Caenorhabditis elegans, and PBX in mammals) interact with anterior HOX, increasing DNA target selectivity. In the presence of their target DNA, HOX proteins form molecular complexes with PBX through the homeodo-mains and nearby amino acid sequences (5)(6)(7)(8)(9). Proteins of the MEINOX subfamily (homothorax in Drosophila, CEH25 in C. elegans, and MEIS and PREP in vertebrates) (4, 10 -16) form specific DNA-independent dimeric complexes with PBX, expanding the specific target selectivity and transcriptional regulatory functions (10 -16). Through the use of different interaction surfaces, PBX forms trimeric complexes with HOX and MEINOX proteins, further expanding DNA target selectivity of HOX (17)(18)(19).
Vertebrates express multiple members of the PBC and MEINOX subfamilies, at least 6 PBX (four genes), 5 MEIS (three genes), and 2 PREP (two genes) (4, 11, 20 -26). Thus a high number of possible MEINOX-PBX heterodimers may form but very little is known regarding the role of individual heterodimers. The formation of heterodimers between PBX and MEIS/PREP has multiple biological consequences that include the regulation of nuclear localization (26 -31). Neither PREP-1 nor PREP-2 contains a nuclear localization signal and thus requires PBX heterodimerization to enter the nucleus (26,28). On the other hand, PREP-1 dimerization prevents nuclear export of PBX (28,29). Thus the balance between PREP and PBX may be functionally important; therefore, its disruption may lead to important functional consequences. Because these proteins are expressed both in the embryo and in the adult, such consequences may be felt not only in embryogenesis but also in the adult life.
In man, Prep1 maps at chromosome 21q22.3 (23,32) and hence it can be present in triple copy in chromosomes of Down's syndrome patients. Indeed, PREP-1 is overexpressed in Down's syndrome fetal brains along with at least one of its target genes (33). Because PREP-1 binds DNA only as a heterodimer with PBX proteins, one would predict that increased PREP-1 could not affect the expression of other genes unless PBX proteins also were increased. Data in Drosophila and zebrafish show that homothorax and MEIS proteins stabilize PBX (30,31). If PREP-1 had the same effect, its increase would cause a generalized variation of the gene expression pattern in various tissues. Here we have overexpressed PREP-1 in cultured F9 teratocarcinoma cells. We find that the increase of PREP-1 results in increase of PBX-2 and PBX-1b proteins, that the effect on PBX-2 is post-transcriptional, and that PREP-1 prolongs its half-life, preventing its rapid degradation. The functional significance of this finding is underlined by the change in the kinetics of HOXb1 synthesis during retinoic acid-induced differentiation of F9 teratocarcinoma cells.

EXPERIMENTAL PROCEDURES
Transfections-F9 (mouse embryonal carcinoma) cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and antibiotics. Cells were transfected with either pBOSPREP-1 (25) plus pcDNA3neo (Invitrogen) plasmids or with * This work was supported by grants of the Telethon Foundation, Associazione Italiana Ricerche sul Cancro (AIRC), and the Italian Ministry of University and Scientific Research (MIUR and COFIN). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. pcDNA3neo alone by electroporation (single electric pulse of 53 ms at 500 microfarads and 230 V). Cells were plated in 10-cm dishes, and stable clones were obtained by selection with 380 g/ml neomycin. To induce differentiation, trans-retinoic acid (RA) 1 from Sigma was added to a final concentration of 10 nM. The cells were collected after 0 -7 days of incubation, and nuclear extracts were prepared as described below.
Cell Extracts-Nuclear and cytoplasmic extracts were obtained as described previously (34). One million cells were washed twice with cold phosphate-buffered saline, collected with 300 l of cold buffer A (10 mM HEPES, pH 7.9, 10 mM KCl, 1.5 mM MgCl 2 , 0.5 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride) into an Eppendorf tube, left 10 min on ice, and lysed by the addition of Triton X-100 to a final concentration of 0.3%. The nuclei were collected by centrifugation. The supernatant was removed to a new tube, and 0.11 volume of buffer B was added (0.3 M HEPES, pH 7.9, 1.4 M KCl, 30 mM MgCl 2 ), incubated for 30 min at 4°C, and centrifuged. The resulting supernatant was denoted as cytoplasmic extract. Nuclear extracts were prepared by resuspending the pelleted nuclei in 60 l of buffer C (20 mM HEPES, pH 7.9, 25% glycerol (v/v), 0.42 M NaCl, 1.5 mM MgCl 2 , 0.5 mM dithiothreitol, 0.2 mM EDTA, 0.5 mM phenylmethylsulfonyl fluoride) for 30 min on ice. The extract was cleared by centrifugation.
Cycloheximide (CHX) and MG132 treatments-F9 clones were starved for 30 min in fetal bovine serum-free medium. After a 30-min incubation in complete medium, the medium was supplemented with 10 g/ml CHX or 5 M MG132 and lysed 3 h later with radioimmune precipitation buffer as described above. In other experiments, a time course was carried out by incubating cells for 0, 30, or 60 min with 5 M MG132 at 37°C. In all of the cases, at the end of incubation cells were lysed and the total or nuclear extracts were prepared as described above.
Northern Blotting and RT-PCR-Total RNA was extracted according to standard methods. For Northern blot analysis, 10 -20 g of total RNA were electrophoresed through a 1% agarose, 2.2 M formaldehyde gel and transferred to a Hybond NHϩ membrane with 20ϫ SSC. The filter was then washed in 5ϫ SSC at room temperature for 5 min and dried, and the RNA was fixed to the membrane by UV. The filter was prehybridized in formamide prehybridization/hybridization solution (5ϫ SSC, 5ϫ Denhardt solution, 50% (w/v) formamide, 1% SDS, 100 g/ml of sheared salmon sperm DNA) for 2 h at 60°C and hybridized in the same buffer containing the radiolabeled probe. The filter was washed at room temperature twice in 2ϫ SSC, 0.1% SDS; twice in 0.2ϫ SSC, 0.1% SDS; and then once in 0.2ϫ SSC, 0.1%SDS at 42°C. The human PREP-1 and mouse PBX-2 cDNAs, used as probe, were labeled with [␣-32 P]dCTP by the random primer method.
Amplification involved a first denaturation step at 97°C for 2 min followed by 25 cycles (32 cycles in the case of PREP-1), each consisting of a denaturation step at 95°C for 30 s, an annealing step at 63°C (for PBX-2 and actin) or 55°C (for PREP-1) for 30 s, and an extension step at 72°C for 1 min. To end the reaction, an extension step of 10 min at 72°C was used.

RESULTS
Overexpression of PREP-1 in Mouse F9 Cells Results in PBX-2 Increase-Five F9 mouse teratocarcinoma cell clones, each transfected with either pcDNA3-neo vector (controls, C1-C5) or with both pcDNA3-neo and pBOS-PREP-1 (P1-P5), have been isolated. Immunoblotting analysis performed on nuclear extracts from different clones of control (C1-C4) and PREP-1transfected F9 cells (clones P1-P4) shows an increase in PREP-1 (Fig. 1A). To test whether overexpression of PREP-1 affected PBX protein levels, we employed different PBX antibodies. Using an antibody (PBX-1,2,3) that does not distinguish between PBX-1a, PBX-2, and PBX-3a, we found a clear increase in the intensity of the PBX band in P1-P4 clones. To identify the form of PBX increased in PREP-1-overexpressing cells, we used isoform-specific antibodies. As shown in Fig. 1A, the PBX-2 and PBX-1b bands were clearly increased in the different clones. PBX-3 antibodies gave an ambiguous result (data not shown). The reactivity toward anti-␣-actin antibodies was comparable in all of the lanes. The same result was obtained with another clone, P5 (data not shown).
We then tested the subcellular localization of PREP-1 and PBX proteins by analyzing nuclear and cytoplasmic extracts by immunoblotting. In untransfected F9 cells (and in all of the control clones, data not shown), PREP-1 was present exclusively in the nuclear extract (Fig. 1B). The cytoplasm showed two weaker bands of different sizes that may represent crossreacting components. In overexpressing P2 and P5 clones, PREP1 was found increased in the nucleus but sometimes also in the cytoplasm. The presence of PREP-1 in the cytoplasm correlated with its level of expression (compare clones P5 and P2). We have then analyzed by immunoblotting the same nuclear and cytoplasmic extracts using PBX-1,2,3 antibodies. As shown in Fig. 1B, the PBX band was present only in the nuclear extracts of PREP-1-overexpressing cells. The increase in PBX was higher in the extract of clone P2 than in clone P5, in agreement with the higher expression of PREP-1 in clone P2. The absence of PBX bands in the immunoblots of control cells in Fig. 1B is probably because of a lower efficiency of the anti-PBX antibodies with respect to anti-PREP-1 antibodies and shorter time of development of the film.
As a control, we tested the level of another nuclear protein, HMG1. As shown in Fig. 1B, HMG1 level was found unaltered in both control and PREP-1-overexpressing cells and was present uniquely in the nuclear extracts. In conclusion, the data show a net increase of some PBX proteins in PREP-1-overexpressing clones and their localization in the nucleus.
As PBX-2 was the PBX isoform mainly affected by PREP-1 overexpression (see Fig. 1A), we quantitated the level of PBX-2 by immunoblotting with anti-PBX-2 antibodies using diluted nuclear extracts of the P2 (PREP-1 overexpressing) clone and compared it with that of the undiluted extract from control clone C3. The data show that PBX-2 was increased ϳ4-fold in the nuclei of PREP-1-overexpressing clone (Fig. 1C).
The increase in PREP-1 and PBX proteins should be reflected in an increase of DNA binding activity to the specific DNA sequence recognized by the PBX-PREP-1 heterodimer. To characterize the DNA binding activity of control and transfected F9 cells, we used oligonucleotide b2PP2 from the HOXB2 gene (19) that carries both the PBX-HOX and the PREP-MEIS sites (see "Experimental Procedures") and that binds with high affinity all of the PREP-1-PBX dimers (19). Fig. 2 shows an EMSA comparing the binding activity of a nuclear extract of control clone C3 with that of the overexpressing clone P2. C3 displayed one very weak electrophoretically retarded band (band A), the migration of which is compatible with a complex of PREP-1 with one of the long forms of PBX (25). In the overexpressing clone P2, two bands were observed (bands A and B) and the intensity of A was increased over control extract. The migration of the weaker faster migrating band (band B) is consistent with a PREP-1 complex with one of the short forms of PBX (see below). The binding was specific in both C3 and P2 extracts as it was eliminated by incubation with a 100-fold excess of unlabeled oligonucleotide. To identify the binding components, we used specific antibodies. Anti-PREP-1 antibodies totally inhibited binding in all of the cases, showing the presence of PREP-1 in all of the binding complexes. The A band of control C3 extracts was inhibited by both PBX-1,2,3 and anti-PBX-2-specific antibodies but not inhibited by anti-PBX-3 antibodies, suggesting that the DNA-binding forms were mostly PREP-1-PBX-1a and PREP-1-PBX-2 heterodimers. In PREP-1-overexpressing P2 cells, the slower migrating A band was strongly inhibited (in addition to PREP-1 antibodies) by anti-PBX-1,2,3 and anti-PBX-2 and very weakly by anti-PBX-1 and anti-PBX-3 antibodies. Therefore, it represents mostly a PREP-1⅐PBX-2 complex. On the other hand, the B band in P2 extracts was inhibited only by PBX-1 and PREP-1 antibodies, suggesting a PREP-1-PBX-1b dimer. In conclusion, the data show in PREP-1-overexpressing cells an increase in DNA binding activity and a shift in heterodimer composition from a mostly PREP-1-PBX-1 heterodimer in control cells to mostly PREP-1-PBX-2 plus PREP-1-PBX-1b heterodimers in overexpressing cells. These results are in agreement with the immunoblotting data, which show an increase of PBX-2 and PBX-1b.

PBX-2 mRNA Is Not Increased in PREP-1-overexpressing
Cells and Is Not More Efficiently Translated-We went back to PREP-1-overexpressing cells to test whether the increase of PBX-2 was dependent on an increase of its mRNA. As shown in Fig. 3A, Northern blot analysis of the total RNA showed an abundant PREP-1 mRNA signal in the overexpressing P2 clone under conditions in which no signal was visible with RNA extracted from control clone C3. On the other hand, the signal obtained with a PBX-2 probe was very weak and was found not to be different in the two types of clones (Fig. 3B). We also used RT-PCR to check the levels of the two mRNAs. When we used PBX-2-specific primers, a band of the expected size (Fig. 3C, arrow on the right) was visible with total RNA from both control (C2, C3, and C4) and PREP-1-overexpressing clones P1, P2, and P3. The band had the same intensity in all of the clones, suggesting that the clones had comparable levels of transcript. As a control, we also amplified the endogenous actin mRNA (Fig. 3C, arrow, left side). Amplicons derived from endogenous ␤-actin mRNA also showed no difference in intensity. The amplification assay for PBX-2 mRNA was in the linear range. A control experiment showed a linear relationship between the intensity of the PCR-amplified band and the amount of input PBX-2 cDNA (Fig. 3C). Thus an increase in PBX-2 mRNA would have been detected if present.
RT-PCR with PREP-1-specific primers showed an increase of PREP-1 mRNA in overexpressing P1, P2, and P3 clones when compared with control C2, C3, and C4 clones as expected (Fig.  3D). Therefore, we conclude that the increased level of PBX-2 in PREP-1-overexpressing clones does not result from an increase of PBX-2 mRNA, suggesting a post-transcriptional mechanism for the elevation of PBX-2.
We also tested whether the difference in PBX-2 production could be attributed to a differential translation of PBX-2 mRNA in PREP-1-overexpressing cells. We pulsed cells with [ 35 S]methionine for 0 -3 h, extracted the proteins at different times, and immunoprecipitated them with specific anti-PBX-2 antibodies. SDS-PAGE and autoradiography analysis showed a time-dependent increase of the intensity of the PBX-2 band but showed no difference between control and overexpressing clones (data not shown). Therefore, we conclude that the effect of PREP-1 overexpression is neither at the level of transcription or stability of message nor at the level of its translation.

PBX-2 Has a Short Half-life, Which Is Increased in PREP-1 Overexpressing Cells and by Proteasome Inhibitors-We tested
for degradation of PBX-2 in one control (C3) clone by exposing cells for 3 h to the protein synthesis inhibitor (CHX) and testing total protein extracts by immunoblotting. CHX was used at 10 g/ml, a concentration at which it inhibited [ 35 S]methionine incorporation by 92% (data not shown). As shown in Fig. 4A, the level of both PREP-1 and PBX-2 in total extracts of C3 control clone decreased substantially after CHX addition. However, the decrease of PBX-2 was much lower in the extracts of PREP-1-overexpressing cells (clone P2). This result has been reproduced several times. Immunoblotting with anti-actin antibody is also shown.
To examine the mechanism of degradation of PBX-2, we employed MG132 (5 M), an inhibitor of the proteasome. As shown in Fig. 4B, in control C3 cells the level of PBX-2 increased in the extract after the addition of MG132 for 3 h. We also compared the effect of MG132 on the level of PBX-2 in nuclear and cytoplasmic extracts of C3 control cells. The data show (Fig. 4C) that in the presence of MG132 there is a timedependent accumulation of PBX-2 and PREP-1 in the cytoplasm. We conclude that the level of PBX-2 is controlled by degradation, most probably in the proteasome because it is inhibited by MG132. It seems reasonable to conclude that when PREP-1 is overexpressed, the level of PBX-2 increases because its decay decreases.

Overexpression of PREP-1 Affects the Kinetic of Synthesis of HOXb1 and the Levels of MEIS-1b in Retinoic Acid-treated F9
Cells-F9 teratocarcinoma cells undergo differentiation toward different lineages when treated with RA. In this process, a sequential induction of HOX and PBX gene products occurs (35,36). The effect on PBX proteins induction was already noted to be post-transcriptional (36). In addition, in the process of induction, MEIS-1a and MEIS-1b also are induced (19). Because PREP-1 was shown to participate in this process by forming ternary complexes with HOXb1 and PBX (19), we tested whether the overexpression of PREP-1 in F9 cells affected the induction of HOXb1, PBX-1, PBX-2, MEIS-1a, and MEIS-1b. As shown in Fig. 5A, cells were treated with 10 nM RA for 0 -6 days and samples were removed daily and tested by immunoblotting. No HOXb1 was evident at day 0 in either control (clone C3) or PREP-1-overexpressing (clone P2) cells. In control cells, HOXb1 was transiently induced after 1 day and already disappeared at day 2. However, in PREP-1-overexpressing cells, the induction of HOXb1 was much more sustained and was still evident after day 3, disappearing thereafter (Fig. 5A). When the same blot was rechallenged with anti MEIS-1 antibodies, a very weak band of MEIS-1b was induced at day one in control cells, which increased at day 2 and decayed at days 4 -6. In PREP-1-overexpressing cells, the overall kinetics was not changed but the overall level was noticeably increased. We also tested the effect of RA on PBX proteins induction. As shown in Fig. 5B with Pbx1,2,3 antibodies (that recognize all of the long forms of PBX-1, PBX-2, and PBX-3), we observed in control cells a strong time-dependent increase of a band corresponding to PBX-1a and a very weak induction of a lower molecular weight band (PBX-2) as expected (19,34). In PREP-1-overexpressing cells, the induction of the PBX-1a band was unaffected. The level of the PBX-2 band was much higher and was not affected by the RA treatment.
On the basis of these experiments, we conclude that the overexpression of PREP-1 can have functional consequences as it affects the level of proteins functionally involved in the differentiation of F9 teratocarcinoma cells. DISCUSSION PBX proteins have an essential role in development as demonstrated by gene inactivation studies. In mice, the disruption of the Pbx-1 gene leads to a complex phenotype showing severe hypoplasia or aplasia of multiple organs, deficient cell growth and differentiation, widespread patterning defects of the axial and appendicular skeleton, and intra-uterine death at embryonic days 15 and 16 (37)(38)(39). In zebrafish, Pbx-4 (lazarus) is required globally for segmental patterning in the hind brain and in the anterior trunk (40). The combined deficiency of Pbx-2 and pbx4 prevents hind brain segmentation and causes r2-r6 rhombomeres to acquire a ground state r1 identity (41). Clearly, it is important to elucidate the mechanisms that control the level of PBX proteins.
Ectopic expression of human PREP-1 in Drosophila induces nuclear localization of extradenticle and leads to a homeotic  (19). ocular and limb phenotype (29). In zebrafish, the elimination of Prep-1 function has very severe consequences on development. 2 In mouse, the insertional inactivation of PREP-1 is embryonic lethal. 3 In this paper, we have identified a novel important function of PREP-1, the stabilization of PBX-2.
As PREP-1 monomers do not bind DNA (25), it was surprising to observe an increase in DNA binding activity of the nuclei of PREP-1-overexpressing cells. PREP-1 can dimerize with all of the PBX isoforms (25), and different DNA-binding heterodimers are found in different cells (14,19,25). Uninduced F9 cells contain mainly PREP-1-PBX-1 heterodimers (Fig. 2). In these cells, overexpression of PREP-1 results in increased PBX proteins, mainly PBX-2 and PBX-1b in five of the five clones analyzed (Fig. 1). In control cells, PREP-1 is present only in the nuclear extracts, whereas in the transfected clones with highest expression it was found also in the cytoplasm (Fig. 1B). On the other hand, PBX-2 was very low in control F9 cells but was increased in PREP-1-transfected cells and was mostly localized in the nucleus. We interpret the presence of PREP-1 in the cytoplasm as an indication that this protein is produced in excess of PBX.
The increase of PBX-2 in PREP-1-overexpressing clones is not due to a transcriptional or translational effect. In fact, the mRNA level (as judged by Northern and RT-PCR analysis) (Fig. 3) was found to be identical in both control and PREP-1-overexpressing clones. Also, immunoprecipitation of pulse-labeled cells did not reveal any difference between control and PREP-1-overexpressing clones (data not shown) and hence no differential rate of translation. The increase of PBX-2 appears to be due to a stabilization effect. Indeed, in control F9 cells treated with the protein synthesis inhibitor CHX, PBX-2 decayed rapidly. This decay was drastically reduced in PREP-1overexpressing cell clones. In agreement with this observation, treatment with MG132 led to an increase of PBX-2 in both nucleus and cytoplasm. MG132 sensitivity strongly suggests that PBX-2 degradation occurs via the proteasome. PBX-2 accumulation in the cytoplasm of MG132-treated cells (Fig. 4C) suggests that under normal conditions the excess monomeric PBX-2 exits the nucleus, enters the cytoplasm, and is targeted to degradation in the proteasome (Fig. 6).
The increase of PBX-2 half-life in the presence of excess PREP-1 indicates that the synthesis of PREP-1 is another important factor in determining the nuclear localization of PBX proteins. Nuclear localization of PBX is known to be the result of dimerization with PREP-1-MEIS (27,28) and is controlled by the phosphorylation of relevant residues of PBX (45). We now show the importance of the rate of synthesis of PREP-1 and of the balance between PREP-1 and PBX in determining the nuclear localization of PBX. Proteasome degradation of these proteins presumably occurs when one of them is in excess over the other, and we speculate that monomeric proteins are targeted to proteasome. The artificial increase in PREP-1 leads to a greater entry/residence of PBX-2 in the nucleus with decreased proteasomal degradation (Fig. 6). It is interesting to note that PBX-2 contains an N-terminal glycine-rich region (Fig. 6). A similar N-terminal region is present in Sp1 and represents the target for its proteasomal degradation (46). A similar sequence is also present in the middle of NF-kB p105 where it represents the processing signal for the generation of the p50 subunit (47).
We have shown that PREP-1 overexpression has a functional relevance in cell differentiation. The well described process of retinoic acid-induced differentiation of F9 teratocarcinoma cells into various cell lineages including neurons passes through the cascade synthesis of HOX proteins, in particular HOXb1. HOXb1 is an important gene that confers identity to rhombomere r4 in the hind brain and hence is essential in the development of specific nerves, neural crest cells, and branchial arches (48,49). Induction by RA also induces PBX and MEIS proteins, which together with PREP-1 and HOXb1 give raise to various trimeric complex that are essential, for example, for HOXb2 expression (19). The discovery that PREP-1 overexpression induces changes in the levels and in the kinetics of expression of such essential genes would be expected to have an important function also in hind brain development in vivo. Hence we feel that the discovery that PREP-1 stabilizes PBX-2 and perhaps other PBX proteins as well is a functionally important finding.
PREP-1 stabilization of PBX may also have an important role in diseases in which the balance between PREP-1 and PBX-2 is changed. For example, because PREP-1 maps to chromosome 21q22.3 (23,32), the presence of a third copy of the gene (in chromosome 21 trisomy) with its consequent increased expression (33) might cause an imbalance in the PREP-1/ PBX-2 ratio in certain cells, increasing the level of PBX and hence the amount and nature of DNA-binding complexes. This, in turn, might change the genetic program of the cells.