Characterization of the Retinoid Binding Properties of the Major Fusion Products Present in Acute Promyelocytic Leukemia Cells

THE CHARACTERISTIC balanced translocation t(15q+; 17q−) of acute promyelocytic leukemia (APL) fuses a portion of the promyelocytic leukemia (PML) gene on chromosome 15 to the second coding exon of the nuclear retinoic acid receptor α (RARα) on chromosome 17. As a result of this translocation a PML/RARα chimeric gene is formed which is transcriptionally active in all cases of APL.1,2 The chimeric transcript encodes a PML/RARα fusion protein that retains both the PML “ring finger” and dimerization domains and the RARα DNA-binding and RA-binding domains.3,4 The breakpoints on chromosome 15 are clustered at three different sites in the PML locus: in intron 6 (breakpoint cluster region 1; bcr1), exon 6 (bcr2), and intron 3 (bcr3), leading to three different fusion proteins.5-8 The most frequent products are the bcr1- (also known as “long” or type “B”) and bcr3- (or “short” or type “A”) PML/RARα,4,5 which present different lengths of PML (amino acids 1-552 and 1-394, respectively). In rare APL cases, the RARα gene fuses with the PLZF (promyelocytic leukemia zinc finger) gene on chromosome 11^9,10 or with the nucleophosmin gene on 5,11 suggesting that aberrant RARα gene expression may be associated with leukemogenesis.

APL has become the first model of differentiation therapy targeted to a defined genetic defect.3,4,12 All-trans-retinoic acid (t-RA) induces disease remission of APL patients by triggering terminal differentiation of malignant cells.3,4,12 However, there is diversity of t-RA sensitivity in APL cases: t-RA induces cell differentiation and clinical remission in patients with the t(15; 17) translocation,4,12 but APL patients and cultured cells from patients with the t(11; 17) fail to differentiate in response to t-RA. In the single t(5; 17) APL case a reduced response to t-RA therapy compared to PML/RARα cases10 was found. In addition, within APLs with the t(15; 17), evaluation of patients in remission after treatment with t-RA alone or with t-RA followed by chemotherapy showed that the expression of the bcr3-PML/RARα isoform is significantly associated with a poorer prognosis compared to those with the bcr1-PML/RARα isoform.12-14 

The mechanisms underlying the heterogeneous t-RA response in APL are unclear. Retinoids activate two classes of nuclear receptor proteins of the steroid and thyroid hormone superfamily, the RARs (α, β, and γ), and the retinoid X receptors (RXRs α, β, and γ), for gene transcriptional activation.15,16 T-RA binds only to RARs, whereas 9-cis-retinoic acid (9-cis-RA) is a bifunctional ligand capable of binding to both RARs and RXRs.17-19 Whether 9-cis-RA has a different spectrum of biologic activity from t-RA is presently unknown.

With respect to retinoid binding properties, we and others6,20 have previously reported that the bcr1-PML/RARα isoform binds retinoids with the same affinity and specificity as the wild-type RARα receptor and is present predominantly in high-molecular-weight nuclear complexes.20 

In the present investigation we have characterized the structural and functional properties of the bcr3-PML/RARα and PLZF/RARα in comparison to those of the bcr1-PML/RARα product. We show that in cells from bcr3- PML/RARα APL patients and bcr3-PML/RARα or PLZF/RARα COS-1–transfected cells, these fusion proteins are present in high-molecular-weight nuclear complexes. Moreover, our results provide evidence for preferential retinoid binding to the different PML/RARα products. In fact, we found that 9-cis-RA binds with higher affinity and specificity to the bcr3-PML/RARα isoform than to the bcr1-PML/RARα or RARα. This effect was also associated with a differential activation of retinoic acid responsive elements in COS-1 cells transfected with bcr3-PML/RARα and either TRE (thyroid response element) or βRARE (RARβ response element). Taken together, these results indicate that preferential retinoid binding to different gene fusion products can be measured and suggest the possibility of specific differentiation therapy in APL patients.

9-cis-[10-³H]retinoic acid (24 Ci/mmol) and unlabeled retinoids used in this study were synthesized by the Department of Medicinal Chemistry, Hoffmann-La Roche, Nutley, NJ. All-trans-[³H]-retinoic acid (50.7 Ci/mmol) was purchased from DuPont/NEN (Boston, MA). COS-1 cells were obtained from American Type Culture Collection (ATCC; Rockville, MD). Rabbit anti-RARα polyclonal serum RPα(F )21 was kindly provided by Prof P. Chambon (Strasbourg, France).

Patients samples and processing.Fresh leukemic cells were obtained from peripheral blood (PB) and bone marrow (BM) aspirates of newly diagnosed, informed APL patients presenting an initial percentage of blasts that was more than 80%. Patients were classified as M3 or M3-variant according to the French-American-British (FAB) classification.22 Characterization of the PML/RARα isoform (bcr1-bcr2-bcr3) was performed by reverse transcriptase-polymerase chain reaction (RT-PCR) assay as previously described.23 Leukemic cells were isolated by Ficoll-Hypaque density gradients (Pharmacia, Uppsala, Sweden).

RNA preparation and Northern blot analysis.Total RNA was prepared by the guanidinium isothiocyanate-CsCl procedure24 from Ficoll-Hypaque–isolated APL blasts washed twice with calcium-magnesium free phosphate solution. Total RNA (10 μg/sample) was electrophoresed through a 1.2% denaturing agarose-formaldeyde gel. After electrophoresis, RNA was transferred to Nytran (Schleicher & Schuell, Hayward, CA) by capillary blotting and then cross-linked by UV irradiation. The cDNA probes were radiolabeled with [α-³²P]dCTP (3,000 Ci/mmol; DuPont-NEN) via random priming using the kit and protocols supplied by Bethesda Research Laboratories. The cDNA probes for human RARα, β, and γ^25,26 were obtained from Prof P. Chambon. The cDNA probes for human RXRα²⁷ and for mouse RXRβ and γ²⁸ were obtained from Dr R.M. Evans (The Salk Institute, San Diego, CA). The cDNA probe for chicken glyceraldehyde-3-dehydrogenase (GAPDH)29 was used as control probe to standardize the level of gene expression.

Cell culture, transfection, and retinoic acid binding assays.COS-1 cells were transiently transfected by electroporation as previously described19 with pSG5 expression vectors containing cDNAs for human RARα, bcr1-, and bcr3-PML/RARα,30 or with a pMT2 expression vector24 containing a cDNA encoding the human PLZF/RARα.9 The PLZF/RARα cDNA was reconstructed from wild-type RARα and PLZF cDNAs by PCR and conventional cloning strategies (M. Ruthardt and P.G. Pelicci, manuscript submitted). After transfection (48 to 72 hours), cells (3 to 7 × 10⁷) were removed by trypsinization and collected by centrifugation. Nuclear and cytosolic extracts were prepared from COS transfected cells and from fresh blasts from APL patients as previously described.19,31,32 Extracts were incubated with 10 nmol/L [³H]-t-RA for 18 hours at 4°C. [³H]-t-RA binding was analyzed by high-performance liquid chromatography (HPLC) performed at 4°C using a size exclusion column Superose 6 HR 10/30 (Pharmacia). The eluent was PTG buffer containing 0.4 mol/L KCl and the flow rate was 0.4 mL/min.20 

Apparent molecular weights (MW) were calculated on the basis of the elution times of a series of marker proteins used to calibrate the system, such as: blue dextran, MW 2,000,000; thyroglobulin, MW 669,000; apoferritin, MW 443,000; β-amylase MW 200,000; alcohol dehydrogenase, MW 150,000; bovine albumin, MW 66,000; ovalbumin, MW 45,000; carbonic anhydrase, MW 29,000; and lactalbumin, MW 14,200.

For saturation binding studies as well as competitive binding assays, bound RA was separated from free radioactivity using PD10 desalting columns (Pharmacia) as described.19,20 In saturation binding studies, incubations were performed, for 18 hours at 4°C, in the presence of increasing concentrations of [³H]-t-RA or [³H]-9-cis-retinoic acid. Binding in the presence of 200-fold molar excess of unlabeled t-RA or 9-cis-RA was defined as nonspecific binding. Specific binding resulted from total binding minus nonspecific binding. Linear least-square analysis of the Scatchard plot was performed with the aid of the computer program BDATA-EMF.33 A fixed concentration of [³H]-t-RA (10 nmol/L) and increasing concentrations of unlabeled competing ligand (ranging between 0.5 nmol/L and 1μmol/L) were used in incubations for competitive binding assays. The EC₅₀ values (retinoid concentrations that inhibit 50% of the total specific RA binding) were calculated using the nonlinear least squares regression analysis program ALLFIT.34 

Sodium dodecyl sulfate (SDS)-gel electrophoresis and Western immunoblotting.Nuclear extracts were diluted 1:1 with 2× SDS sample buffer (2% SDS, 0.125 mol/L Tris-HCl, pH 6.8, 20% glycerol, 0.02% bromophenol blue, and 10% 2-mercaptoethanol). Proteins were then fractionated by electrophoresis on an 8% SDS polyacrylamyde gel and electroblotted to nitrocellulose membrane (Hybond C Super; Amersham Life Science Inc, Arlington Heights, IL). Proteins that reacted with the anti-RARα RPα(F ) antibody21 (used at a 1:500 dilution) were detected using the ECL Western blotting detection kit (Amersham) with the reagents provided according to the manufacturer's instructions.

Transient cotransfection of COS-1 cells and transactivation assays.COS-1 cells were plated in six-well 35-mm culture dishes at a density of 2.5 × 10⁵ cells/well. Twenty-four hours after plating cells were transiently transfected using the Lipofectamine reagent (GIBCO-BRL, Gaithersburg, MD). Each well received 0.5 μg of the pSG5 expression vectors containing the coding regions for human RARα, bcr1- or bcr3-PML/RARα, 0.5 μg (βRARE)₃-tk-LUC,35 or TRE₂-tk-LUC,36 or the RARβ pr-LUC reporter,8,35 and the plasmid encoding β-galactosidase (pSV-βGal) (0.3 μg) was cotransfected to monitor transfection efficiency and for normalization of reactions. After 5 hours, transfecting medium was removed and cells were incubated in Dulbecco′s modified Eagle′s medium containing 5% heat-inactivated, charcoal stripped fetal calf serum. The next day retinoids were added for 16 hours to minimize retinoid isomerisation. Cells were washed twice with calcium-magnesium–free phosphate solution and lysate using a Reporter Lysis Buffer (Promega Corp, Madison, WI). Luciferase activity was measured using a luminometer (Berthold, Wildbad, Germany).37 Reporter gene expression was calculated in arbitrary units, relative to β-galactosidase expression. The EC₅₀ values (retinoid concentrations which induced 50% of maximal response) were calculated using the nonlinear least squares regression analysis program ALLFIT.34 

Retinoic acid binding in bcr3-PML/RARα APL patients.Previous retinoid binding studies performed in the APL cell line NB4, which express the bcr1-PML/RARα product,6 showed the presence of specific [³H]-t-RA binding corresponding to CRABP, RARs, monomeric form of bcr1-PML/RARα (110,000), and to the high-molecular-weight complexes formed by the interaction of bcr1-PML/RARα with itself or with other nuclear proteins.20,38 We have now investigated the t-RA binding properties of fresh APL blasts cells obtained from patients who expressed the bcr3 isoform of the PML/RARα transcript. Nuclear and cytosolic extracts were labeled with 10 nmol/L [³H]-t-RA in the absence or in the presence of 200-fold mol/L excess of unlabeled t-RA to determine nonspecific binding, and the t-RA binding was analyzed by HPLC size exclusion chromatography. The HPLC profile of nuclear extracts prepared from these cells showed the presence of three main peaks of specific [³H]-t-RA binding eluting at 36, 42, and 47 minutes, corresponding to molecular weights of about 500,000, 50,000, and 18,000, respectively (Fig 1). Similar HPLC profiles were consistently found in blasts isolated either from BM or from PB of bcr3-PML/RARα APL patients. The specific t-RA binding peaks corresponding to molecular weights of 50,000 and 18,000 probably represent binding to RARs and CRABP, respectively. The presence of CRABP in cytosol extracts prepared from these cells was established by the presence of a main peak of specific binding eluting at 47 minutes and corresponding to a molecular weight of 18,000. In addition, in cytosolic extracts, two small peaks of specific t-RA binding activity eluting at a retention time of 28 and 34 minutes, which probably represented contamination of cytosol with nuclear extract, could be detected.

Retinoic acid binding in bcr3-PML/RARα-COS-1 transfected cells and immunoblot analysis.To easily characterize the t-RA ligand binding properties of the bcr3-PML/RARα isoform in comparison to the previously reported RARα and bcr1-PML/RARα, COS-1 cells were transiently transfected with the cDNAs for RARα, bcr1-PML/RARα, bcr3-PML/RARα, or expression plasmid (mock). Nuclear extracts prepared from these cells were first examined by immunoblot analysis using the anti-RARα RPα(F ) antibody21 to confirm the production of intact receptors. Immunoblot analysis showed the presence of immunoreactive proteins corresponding to the approximate MW calculated from their respective cDNA sequences30,39,40 (RARα 50,000; bcr1-PML/RARα 105,000; bcr3-PML/RARα 90,000) (Fig 2A). As previously described,20 endogenous RARα was not immunodetectable in nuclear extracts from mock-transfected cells (Fig 2A). In pSG5/bcr1-PML/RARα extracts a small degree of degradation was detected, as evidenced by an immunoreactive band showing an MW of about 40,000. We next examined the binding of [³H]-t-RA to these proteins by HPLC size exclusion analysis (Fig 2B). The HPLC profile obtained from nuclear extracts prepared from pSG5/bcr3-PML/RARα transiently transfected COS-1 cells showed the presence of two main peaks of specific t-RA binding activity eluting at 31 and 37 minutes and then corresponding to an MW of more than 1,000,000 and of about 550,000, respectively. These HPLC profiles were very similar to those of nuclear extracts prepared from blasts obtained from bcr3-PML/RARα APL patients. Conversely to RARα and pSG5/bcr1-PML/RARα, a complete absence of the monomeric form of this receptor was found (Fig 2B). In fact, t-RA specific binding activity was not present in fractions corresponding to lower MW proteins. We concluded that the pSG5/bcr3-PML/RARα product is present only as macromolecular complexes that may represent pSG5/bcr3-PML/RARα homodymers and/or high-molecular-weight complexes with other nuclear proteins.

The HPLC profile of cytosolic extracts prepared from COS-1 transfected cells showed the presence of a specific t-RA binding component corresponding to an MW of 18,000 and probably representing the endogenous CRABP and similar peaks of specific binding present in nuclear extracts probably resulting from a lysis of a small percentage of nuclei during the preparation of cytosolic extracts (data not shown).

Saturation binding and Scatchard analysis of t-RA and 9-cis-RA.Previous studies20 indicated that the bcr1-PML/RARα isoform bound t-RA with similar affinity as RARα (Kd 0.13 nmol/L and 0.09 nmol/L, respectively). To characterize the binding affinities of t-RA to the bcr3-PML/RARα product we first performed saturation binding and Scatchard analysis. Nuclear extracts prepared from pSG5/bcr3-PML/RARα COS-1 transfected cells were labeled for 18 hours with increasing concentrations of [³H]-t-RA. Bound and free radiolabeled were separated with PD10 desalting columns and counted as described in Materials and Methods. A typical binding curve and Scatchard plot for [³H]-t-RA binding to the bcr3-PML/RARα is shown in Fig 3. The binding of t-RA to the bcr3-PML/RARα was specific, saturable, and exhibited high affinity similar to that described for RARα and bcr1-PML/RARα isoform.20 Scatchard analysis yielded a linear plot consisting of the presence of a single class of binding sites (r = −.91). The Kd value obtained was 0.47 ± 0.01 nmol/L. These data indicate that t-RA binds with relative lower affinity (threefold to fivefold) to bcr3-PML/RARα compared with the previously reported Kd values measured for RARα and bcr1-PML/RARα receptors.20 

It has been shown that 9-cis-RA binds with high affinity to each of the members of the RXR and RAR subfamilies.17,41 Scatchard analyses of the binding of [³H]-9-cis-RA to RARα performed by different groups yield Kd values of 0.24 to 0.31 nmol/L.17,41 We examined the binding of [³H]-9-cis-RA to the bcr1-PML/RARα and bcr3-PML/RARα present in nuclear extracts prepared from COS-1 cells transfected with their respective cDNAs. The [³H]-9-cis-RA binding to both bcr1-PML/RARα and bcr3-PML/RARα was specific and saturable (Fig 4). However, saturation kinetics and Scatchard analyses of the binding of [³H]-9-cis-RA to the bcr1-PML/RARα yield kd value of 0.77 nmol/L (r = −.97), whereas two kd values of 0.45 nmol/L (r = −.98) and 0.075 nmol/L (r = −.96) could be calculated for the bcr3-PML/RARα. These data indicated that 9-cis-RA binds with a lower affinity to the bcr1-PML/RARα than to the bcr3-PML/RARα. Moreover, two sites of 9-cis-RA binding with different affinities are present in the bcr3-PML/RARα fusion product. However, the analysis of the maximum binding showed that low-affinity sites were involved in the majority of this binding activity.

Competition binding.To further characterize the ligand binding properties of the bcr3-PML/RARα isoform, and the different binding affinities of t-RA and 9-cis-RA for the two major PML/RARα products, competition binding experiments were performed in COS-transfected cells using different unlabeled retinoids such as t-RA, 9-cis-RA, CH55,20,42 AM580, and TTNPB.15,16 In these studies, nuclear fractions prepared from pSG5/RARα, pSG5/bcr1-PML/RARα, and pSG5/bcr3-PML/RARα COS-1 transfected cells were incubated for 18 hours at 4°C with 10 nmol/L [³H]-t-RA in the presence of varying concentrations of the unlabeled retinoid (Table 1). According with previous observations,20 t-RA exhibited similar ability to compete for [³H]-t-RA binding to either RARα and bcr1-PML/RARα (EC₅₀ 4.9 and 4.8 nmol/L, respectively). For the bcr3-PML/RARα isoform, the calculated EC₅₀ value for competition binding by t-RA was 7.5 nmol/L. TTNPB and AM580 showed a slightly different specificity (about 1.5- to 2-fold) for competing to the three receptors, whereas CH55 had similar EC₅₀ values. Conversely, competition binding of unlabeled 9-cis-RA to RARα, bcr1-PML/RARα and bcr3-PML/RARα yield EC₅₀ values of 29.5 ± 5.5 nmol/L, 27.5 ± 4.1nmol/L, and 4.8 ± 1.3 nmol/L, respectively. These results indicated that 9-cis-RA binds with a higher specificity to the bcr3-PML/RARα isoform than to the RARα or the bcr1-PML/RARα.

Functional transactivation assay in cotransfected COS-1 cells.We next investigated whether the higher affinity and specificity of 9-cis-RA for the bcr3-PML/RARα isoform resulted in a different ability of this protein to transactivate two different retinoic acid-response elements (RARE) such as the βRARE and the TRE, compared with RARα and/or bcr1-PML/RARα. The βRARE is the direct repeat sequence (DR5) of the RARβ₂ promoter. The TRE is a palindromic sequence that also mediates thyroid hormone transactivation. In COS-1 cells, when RARα, bcr1-PML/RARα or bcr3-PML/RARα expression vectors were cotransfected with either the (βRARE)₃-tk-LUC or TRE₂-tk-LUC reporters, treatment with t-RA and 9-cis-RA induced luciferase activity in a dose-dependent manner. However, 9-cis-RA was less potent than t-RA in activating these repoter genes (Table 2). The magnitude of the (βRARE)₃-tk-LUC induction by t-RA was similar for the three receptors (EC₅₀ ranging between 0.4 to 0.9 nmol/L). 9-cis-RA was able to induce (βRARE)₃-tk-LUC about fourfold more efficiently in RARα expressing cells than in bcr1- or bcr3-PML/RARα-COS-1 transfected cells (Table 2). However, under these conditions 9-cis-RA had very similar EC₅₀ for the two PML/RARα isoforms. The different transactivation activity between t-RA and 9-cis-RA was abolished when the RARβ-pr LUC,8,35 a reporter plasmid containing the RARβ gene promoter region (−5 kb to +155) fused to the promoterless luciferase gene was used (Table 3). As recently shown,⁴³ it is possible that the mechanism of ligand-dependent transcription is modulated by a number of positive and negative factors that may not be functional when direct repeat sequences are used.

T-RA was about threefold to fivefold more potent in activating the TRE₂-tk-LUC reporter gene in COS-1 cells expressing the bcr1- and bcr3-PML/RARα than in COS-1 cells expressing RARα. In contrast, 9-cis-RA treatment was equally potent in activating this reporter gene in RARα and bcr1-PML/RARα expressing cells. Remarkably, 9-cis-RA was about 10-fold more active in activating the TRE₂-tk-LUC in the bcr3-PML/RARα than in the RARα or in the bcr1-PML/RARα COS-1 transfected cells. These results suggested that the higher affinity and specificity of the binding of 9-cis-RA to the bcr3-PML/RARα product resulted in an increased ability of this protein to specifically transactivate the TRE₂-tk-LUC, and not the (βRARE)₃-tk-LUC, in transiently transfected COS-1 cells.

In agreement with this, preliminary results obtained from blast cells directly isolated from two bcr3-PML/RARα patients showed that 9-cis-RA was more potent than t-RA in inducing cell differentiation, as evaluated by morphology, by the nitroblue tetrazolium (NBT) dye reduction assay and by type II transglutaminase activity induction.38 In fact, the calculated EC₅₀ for these effects were estimated to be approximately 1.8 nmol/L for t-RA and 0.1 nmol/L for 9-cis-RA, respectively.

Additional differences between binding specifity and transcriptional activity were also detected for the RARα- and RAR-selective retinoids AM580 and TTNPB. AM580 and TTNPB in fact, resulted to be very effective in transactivating either TRE₂-LUC or (βRARE)₃-tk-LUC in cotransfected COS-1 cells. However, these retinoids were consistently found to be less active in transactivating these RAREs in the bcr3-PML/RARα COS-1 transfected cells (Table 2).

Retinoic acid binding properties of PLZF/RARα.We then compared the t-RA binding properties of PML/RARα isoforms with that of the PLZF/RARα, the fusion protein generated by the less frequent chromosomal rearrangement t(11; 17) present in APL and associate with refractoriness to t-RA treatment. Nuclear and cytosolic extracts were prepared from COS-1 cells transfected with a pMT2 expression vector containing a cDNA encoding the human PLZF/RARα. PLZF/RARα was able to bind t-RA and t-RA specific binding activity was found mainly associated with the nuclear fraction (Fig 5). Similarly to PML/RARα, PLZF/RARα was found present in high-molecular-weight nuclear complexes eluting with apparent MW of 950,000 and 500,000. Similar HPLC profile was obtained in nuclear extracts prepared from myeloid U937 precursor cells expressing PLZF/RARα (data not shown). Moreover, as in the case of RARα and bcr1-PML/RARα,17,20,31,41 a single site of specific high-affinity binding could be detected for PLZF/RARα by saturation kinetics and Scatchard analyses. The calculated binding affinities were 0.17 nmol/L for t-RA and 2.7 nmol/L for 9-cis-RA (data not shown).

Expression of RARs and RXRs in APL patients.Retinoid induced differentiation involves activation of multiple RA-dependent signaling pathways, including RARs and RXRs. Therefore, we have analyzed the pattern of expression of retinoid receptors in leukemic blasts expressing bcr1 or bcr3 isoforms of PML/RARα by Northern blotting analysis of RNA samples from 4 bcr1-PML/RARα and 5 bcr3-PML/RARα APL patients (Fig 6). As previously described,1,44,45 three RARα transcripts of 2.1, 3.0, and 4.8 kb were detectable in mRNAs prepared from all the APL patients. 3.1-kb RARγ and 5.6-kb RXRα transcripts were relatively highly expressed in APL patients whereas RARβ and RXRγ transcripts were undetectable. As expected, samples from bcr1- and bcr3-PML/RARα APL patients expressed the same subtypes of RAR or RXR receptor mRNAs. RXRβ transcript was about equally expressed in all the samples whereas slightly different levels of expression of RARα, RARγ, and RXRα transcripts were detectable. However, a correlation between the expression level of RARs and RXRs and the presence of a particular PML/RARα isoform was not found.

The combination of t-RA and conventional cytotoxic agents has been found to provide a more effective overall therapy for APL than conventional chemotherapy or retinoid therapy alone.3,4,12,46 Therapies including t-RA induced terminal differentiation of the leukemic promyelocytes, which are then replaced by normal hematopoiesis.3,4 However, the precise molecular mechanism of RA-induced differentiation in APL blasts has yet to be elucidated. It has been shown that the ligand binding domains of RARs and of the other members of the nuclear receptor superfamily also includes homodimerizing and heterodimerizating interfaces, ligand-dependent transcriptional activation function 2 (AF-2), and in some cases, hormone reversible transcriptional repression functions.15,43 All PML/RARα isoforms retain most of the RARα functional domains.3,4 In addition, APL blasts containing bcr1- and bcr3-PML/RARα also express transcripts encoding for RARα, RARγ, RXRα, and RXRβ. Moreover, specific retinoid binding probably corresponding to RARs and to CRABPs was detectable in these cells. These data indicate that in APL cells multiple retinoid signaling pathways are probably involved in RA-induced differentiation.

T-RA is an effective inducer of clinical remission only in patients carrying the t(15; 17) and expressing the PML/RARα product.3,4,12,47 APL patients defined as M3 by FAB criteria,22 but lacking the t(15; 17) or presenting t(11; 17), do not respond to t-RA–based treatments.10,47 Recent clinical studies indicate a different prognosis in APL patients treated with t-RA alone or with t-RA followed by chemotherapy on the basis of the fusion product present.13,14,48 In APL, the bcr3-PML/RARα isoform has been associated with the more aggressive microgranular APL variant, to CD2 expression, to higher leukocyte counts, and to poorer prognosis.12-14,49 However, in patients expressing the two major isoforms of PML/RARα treated with combined t-RA and chemotherapy no statistical differences in overall survival and complete remission duration was found.50 These findings stress the importance of a complete analysis of the molecular and clinical characteristics of APL patients to define t-RA sensitivity and response to differentiation therapy.

In this study we investigated the molecular mechanism of t-RA response in APL by analyzing functional properties such as retinoid binding and transcriptional activation of bcr1- and bcr3-PML/RARα, the two predominant fusion proteins present in APL.

In particular, we measured the binding properties of the two endogenous ligands t-RA and 9-cis-RA to bcr1- and bcr3-PML/RARα isoforms. By Scatchard analysis we showed that t-RA binds to the bcr3-PML/RARα isoform with slightly less affinity (Kd = 0.4 nmol/L) than to bcr1 PML/RARα or RARα (Kd = 0.13 nmol/L or 0.09 nmol/L, respectively). A relatively lower t-RA binding specificity for the bcr3-PML/RARα could also be measured in competitive binding analysis. Moreover, these experiments showed that 9-cis-RA is able to bind with about eightfold higher specificity to bcr3-PML/RARα than to RARα or bcr1-PML/RARα (EC₅₀ values of 4.8 nmol/L, 33.5 nmol/L, and 37.5 nmol/L, respectively). This finding was also supported by saturation kinetics studies which indicated that [³H]-9-cis-RA binds with a lower affinity to the bcr1-PML/RARα than to the bcr3-PML/RARα. Two sites of 9-cis-RA binding with different affinities were detectable in the bcr3-PML/RARα fusion product. In fact, for bcr1-PML/RARα a Kd value of 0.77 nmol/L was measured, whereas two Kd values of 0.45 nmol/L and 0.075 nmol/L could be calculated for bcr3-PML/RARα.

Both PML/RARα isoforms retain the RING domain of PML which is probably involved in protein-protein interactions and regions involved in DNA binding.49 However, the bcr3-PML/RARα product lacks the C-terminal proline/serine–rich region of PML.7 It has been shown that human RARα contains overlapping but distinct binding pockets for t-RA and 9-cis-RA in a region which corresponds to the ligand-dependent transcriptional activation function 2 (AF-2 ) domain.51 This region in RARα is predicted to form an amphipatic α-helix that is well conserved among nuclear receptors.52 Using alanine scanning mutagenesis, specific amino acids that constitute the high-affinity binding of 9-cis-RA have been defined within this region.⁵³

Our hypothesis of functional differences between the two isoforms is also supported by the fact that different protein complexes are probably formed by the two PML/RARα fusion products. At variance to RARα,20,31 both PML/RARα isoforms were found in high molecular nuclear complexes by HPLC size exclusion analysis of nuclear extracts prepared from bcr1-PML/RARα and bcr3-PML/RARα transiently transfected COS-1 cells. However, the bcr3-PML/RARα product was present only in protein-protein complexes as shown by the absence of any RA-specific binding corresponding to the monomeric form of this protein either in COS-1 transfected cells as well as in APL patients nuclear extracts. In similar experimental conditions, RA-specific binding activity corresponding to the monomeric form of the bcr1-PML/RARα product was always measurable in NB4 and in COS-1 transfected cells.20,38 In addition, identical HPLC profiles were obtained if nuclear extracts from mock, RARα, bcr1-, and bcr3-PML/RARα COS-1 transfected cells were labeled with [³H]-9cis-RA (data not shown).

It has been shown that both PML/RARα and PLZF/RARα fusion proteins contain identical portions of RARα and retain the ability to act as RA-dependent transcription factors, to bind to RA-response elements, and to antagonize the function of the wild-type RARα.5-9,54 Therefore, it appears that they both have the potential to directly influence the RARα-dependent endogenous pathway that control terminal myeloid differentiation. The introduction of PLZF/RARα cDNA into COS-1 cells by transient transfection resulted in high levels of nuclear t-RA specific binding activity. The affinity of t-RA binding to PLZF/RARα was similar to that of PML/RARα. However, APL patients with the t(11; 17) expressing the PLZF/RARα are refractory to t-RA–induced differentiation9,10,54 and the expression of PLZF/RARα in hematopoietic precursor cells blocks terminal differentiation (M. Ruthardt and P.G. Pelicci, manuscript submitted). Therefore, it seems that despite PML/RARα and PLZF/RARα having similar retinoid-binding properties they have different effects in vivo on the RA-signaling pathway. The biochemical consequences of PML or PLZF sequences on the in vivo activity of the two RARα fusion proteins might differ and underlie their different biologic activities. However, the presence of both PML/RARα and PLZF/RARα in similar high-molecular-weight nuclear complexes suggests the involvement of these complexes in the block of promyelocytic differentiation.

Previous reports indicate only minor differences between the two PML/RARα isoforms in the activation of various RAREs in a variety of target cells.5-8 In this study we extended these previous findings, most of which had been obtained using a single t-RA concentration, by analyzing the t-RA and 9-cis-RA dose-dependent activation of different RAREs in RARα, bcr1-PML/RARα and bcr3-PML/RARα COS-1 transfected cells. We found that bcr1-PML/RARα and bcr3-PML/RARα overexpressed in COS-1 cells had similar alterations in activating (βRARE)₃-tk-LUC and RARβ pr-LUC when compared with wild-type RARα. However, the higher affinity and specificity of 9-cis-RA binding to the bcr3-PML/RARα product resulted in an increased ability (about 10-fold) of this fusion product to specifically transactivate the TRE₂-tk-LUC. It should be noted that the TRE is usually present in few RA target genes and is less responsive to RA than the DR-5 motif, thus far the strongest and the most abundant class of RARE present in the promoter of the RARβ and in the promoters of several RA-responsive genes.43,55 In agreement with these results, we found the 9-cis-RA to be more potent than t-RA in the induction of differentiation of blast cells from two bcr3-PML/RARα APL patients in culture (data not shown).

In conclusion, our results indicate that distinct functional properties are associated with the two major PML/RARα isoforms. This in turn may provide a rationale for targeting differential retinoid therapy at the specific molecular lesion of APL patients.

The manuscript by M. Ruthardt and P.G. Pelicci listed as submitted is now in press (Ruthardt M, Testa U, Nervi C, Ferrucci PF, Grignani F, Puccetti E, Grignani F, Peschle C, Pelicci PG: Mol Cell Biol, 1997).

We thank Dr. P. Chambon for providing RARα antibody and RARs cDNA and Dr R.M. Evans for providing RXRs cDNA. The RARβ pr-luc was kindly provided by Dr A. Dejean.