Introduction

Almost two decades ago, the prognosis of acute promyelocytic leukemia (APL) was critically poor due to fatal coagulation disorders at diagnosis [1, 2]. Even with conventional chemotherapy using anthracyclines, more than 70 % of APL patients showed poor prognosis [3, 4]. After introduction of all-trans retinoic acid (ATRA) in the clinical setting in combination with conventional chemotherapy, the prognosis of APL has improved dramatically, with the result that more than 85 % of patients now achieve complete remission (CR) and nearly 70 % of patients can be cured [58]. Since 1994, the marked effectiveness of As2O3 in APL patients, even in relapsed patients after combination therapy with ATRA, has been confirmed [912]. When As2O3 is utilized as a single agent, ~70 % of patients can be cured, whereas nearly 90 % of patients can be cured if As2O3 is utilized in combination with ATRA [13, 14]. Although outcomes of APL treatment with ATRA and/or As2O3 in combination with conventional chemo-drugs have improved, relapsed/refractory patients are still observed in the clinical setting and drug resistance to ATRA and As2O3 has been recognized as a critical problem.

More than 98 % of APL patients carry the t(15;17) translocation, which results in fusions of the retinoic acid receptor alpha (RARα) gene with the promyelocytic leukemia (PML) gene, PML-RARα (Fig. 1) [1517]. A very limited number of patients, showing APL phenotype without t(15;17), exhibit a variety of X-RARα fusions (Fig. 1) [1825]. Interestingly, some patients expressing X-RARα show clinical resistance to ATRA and/or As2O3. Previous reports have indicated that both ATRA [26, 27] and As2O3 [2830] have rigorously defined molecular targets, an improved understanding of their molecular mechanisms of action and resistance may be important to further improving clinical outcomes in APL treatment.

Fig. 1
figure 1

Schematic representation of PML-RARα and X-RARα fusion protein confirmed in APL. Chromosomal translocations resulting in the fusion protein are also indicated under the name of fusion protein. Long and short forms of PML-RARα with or without nuclear localizing signal (NLS) are reported [86]. ATRA and As2O3 responsiveness in the clinical setting and/or in vitro analyses is indicated in the right panel. Gray triangles indicate break points of chimeric protein. Numbers indicate the amino acid positions. A to E functional domains in RARα, DBD DNA binding domain, LBD ligand binding domain, RING really interesting new gene finger domain, B1 and B2 B-box motifs, CC coiled-coil domain, POZ/BTB pox virus and zinc finger/BR–C, ttk and bab domain, Pro proline rich domain, Zn zinc finger domain, NR nuclear reassembly, RIIA dimerization domain, BBD BCL6-binding domain, AR ankyrin repeats, + sensitive, − resistant, NR not reported

Mechanisms of action of molecular targeting drugs to APL cells

ATRA

Wild-type RARα is a nuclear hormone receptor that binds to consensus sequence DR5 (five bases spaced between two AGGTCA motifs) in target gene promoters, normally as heterodimer with retinoid X receptor (RXR) [3133]. Without ligands, ATRA and 9-cis retinoic acid, RAR-RXR heterodimer induces transcription repression throughout chromatin remodeling by recruiting transcription co-repressors, such as N-CoR/SMRT large protein complexes, that contain histone deacetylases (HDACs) [27, 3437] and histone methyltransferases [3840]. In the presence of ligand (~10−7 M), the co-repressor complexes dissociate from RAR-RXR, and transcriptional de-repression and activation are induced [3437, 41]. PML-RARα binds to DR5 of target gene promoters primarily as a homodimer, but also as a heterodimer with RXR [42, 43], and induces transcription repression by recruiting N-CoR/SMRT complexes and polycomb group repressive complex 1 and 2 (PRC1/2) [39, 40], which contain histone methyl transferases, in the absence of ligands [27] (Fig. 1). PML-RARα can be SUMOylated at K160 of the PML protein to recruit death domain-associated protein (DAXX), resulting in the transcriptional repression of target genes [44]. Even in the presence of physiological concentration of ligand (10−7 M), the co-repressor complex still binds with PML-RARα and the transcriptional repression cannot be dissolved. In the presence of pharmacological concentration of ATRA (10−6 M), transcription activation can be induced by dissociation of co-repressor complexes from PML-RARα and proteasome-dependent PML-RARα degradation [4547].

As2O3

The efficacy of As2O3 on APL cells was first reported by Chen et al. in 1996 [28], who showed the dual effect of apoptosis at relatively high concentrations (0.5–2 μM/L) and partial differentiation at low concentrations (0.1–0.5 μM/L) in both ATRA-responsive and ATRA-resistant APL cells. As2O3 induces the targeting of nucleoplasmic PML-RARα with a micro speckled pattern into nuclear bodies with a normal speckled pattern prior to degradation [30, 4850]. As2O3 induces the formation of reactive oxygen species (ROS) [30], which induce multimerization of PML-RARα through intermolecular disulphide crosslinks at PML B1-domain (Fig. 2) and PML-RARα SUMOylation by ubiquitin-conjugating enzyme 9 (UBC9) [30]. A recent report indicated that As2O3 directly binds with PML at the C–C motif in the PML B2-domain, and that PML SUMOylation can be induced by enhancement of UBC9 binding at the PML RING domain [29, 30, 50]. SUMOylated PML recruits RING finger protein 4 (RNF4), which is known as a SUMO-dependent ubiquitin ligase [51], and polyubiquitylated PML-RARα can be degraded by ubiquitin–proteasome pathway [29, 49, 51].

Fig. 2
figure 2

Molecular mechanisms of action and resistance to ATRA and As2O3 in APL cells. PML-RARα are found mainly as homodimers through the C–C domain of PML, and partially as heterodimers with RXR. PML-RARα binds with target gene promoter in the absence of ligand, and recruits co-repressor complexes, such as N-CoR/SMRT complexes containing histone deacetylases (e.g. HDAC3) [3437, 41] and PRC1/2 complex containing histone methyltransferases (e.g. EZH2) [39] to repress the gene expression. Histone tail deacetylatation and/or methylation are related to transcription repression. In the presence of pharmacological concentration (1 × 106 μM) of ligand (ATRA), co-repressor complexes are dissociated from RARα, while co-activator complexes containing histone acetyltransferases (e.g. p300/CBP) are recruited, and transcription activation occurs. In the cases of PML-RARα with LBD mutations, ligand binding with LBD is interfered and co-repressor dissociation does not occur in the presence of pharmacological concentrations of ATRA. In the presence of As2O3, the formation of reactive oxygen species (ROS) is induced, and PML intermolecular disulfide crosslinks through B1 domain, that induce multimerization, and SUMOylation of PML by ubiquitin-conjugating enzyme 9 (UBC9) occur. As2O3 directly bind with PML-B2 domain and enhancing UBC9 binding and SUMOylation of PML. SUMOylated PML recruits RING finger protein 4 (RNF4), and is polyubiquitylated by RNF4, and proteasome-dependent degradation occurs. If PML-RARα has PML-B2 mutation, direct binding of As2O3 with PML is impaired, and polyubiquitylation and degradation are perturbed

Molecular mechanisms of drug resistance in APL cells

From the molecular mechanisms of ATRA and As2O3 effectiveness as indicated above, several mechanisms of drug resistance have been speculated [52]. In this section, we outline the molecular mechanisms of resistance that are thought to be significant from the clinical perspective.

RARα fusion proteins in APL

In very limited cases with APL phenotype, RARα translocations with X-genes other than PML (PLZF [18], NuMA [19], NPM [20], STAT5b [21, 53], FIP1L1 [22], PRKAR1A [23, 24], and BCOR [25]) resulting in the production of X-RARα fusion protein have been reported (Fig. 1). PML-RARα forms mainly homodimers, and it has been reported that homodimerization of PML-RARα is critical for the pathogenesis of APL [42, 43]. Sternsdorf et al. [54] indicated that forced homodimerization of RARα induces ALP-like leukemia in a mouse model, indicating that the dimerization domain of the fusion protein may be critical to the induction of leukemogenesis by X-RARα. In fact, homodimerization through specific domains (coiled-coil; PML-, NPM-, and STAT5b-, POZ/BTB; PLZF-, RIIA; PRKAR1A-, and so on) has been confirmed in all X-RARα proteins. Interestingly, in PML-, PRKAR1A- [24], and BCOR-RARα [25], heterodimerization with RXR is also important for transformation and/or RARE binding.

Since those chimeric proteins all hold RARα DNA binding domain (DBD) and ligand binding domain (LBD), ATRA responsiveness is speculated in all cases. However, ATRA resistance has been confirmed clinically in cases showing PLZF-RARα [18, 34, 41] and STAT5b-RARα [21, 53, 55] fusions. One explanation for ATRA resistance is that the N-CoR/SMRT-corepressor complex interacts with PLZF, even in the presence of pharmacological concentration of ATRA, such that transcriptional de-repression cannot occur at RARα target gene promoters [34, 41]. The molecular mechanisms of ATRA resistance in STAT5b-RARα-expressing cells has not been fully explicated. Wild-type Stat5b is localized in cytoplasm, but STAT5b-RARα aberrantly localizes in nucleus [21]. STAT5b is a component of the janus kinase (JAK)-STAT signaling pathway, and phosphorylation of STAT5b by JAK causes homodimerization and translocation into the nucleus, where it acts as a transcription factor [56]. Aberrant transcription regulation of STAT5b target genes in addition to RARα target genes by STAT5b-RARα may be related to ATRA resistance.

On the other hand, As2O3 resistance in clinical setting was observed in patients expressing PLZF- [57, 58], STAT5b- [55], and BCoR-RARα [25]. The As2O3-binding C–C motif is confirmed in PML-B2 domain, and As2O3 binding is critical for the multimerization followed by PML-RARα degradation [29, 30, 42]. Lack of As2O3 binding sites in X-RARα protein may be one explanation of loss of As2O3 responsiveness. However, no direct effect of As2O3 on RARα has been reported.

Mechanisms of resistance to ATRA

A number of mechanisms have been proposed to explain ATRA resistance in APL patients expressing PML-RARα, such as amino acid substitution in RARα LBD domain by genetic mutations, increased catabolism of ATRA, presence of cytoplasmic retinoic acid binding protein (CRABP), and abnormal ATRA delivery to the cell nucleus. Only genetic mutations on the RARα LBD domain in PML-RARα have been confirmed as an ATRA-resistant mechanism, from both clinical observations and in vitro molecular analyses [5966]. Genetic mutations (missense, nonsense, and deletions) on RARα LBD domain have been confirmed in ATRA-resistant patients and APL cell lines, which grow despite pharmacological concentrations of ATRA, as summarized in Fig. 3. These mutations accumulate in the three subregions (zones I, II, and III in Fig. 3) of the LBD domain [66]. Gallagher et al. [66] reported that PML-RARα LBD mutation was confirmed 18 of 45 (40 %) relapse patients treated with ATRA/chemotherapy. In vitro analyses using ATRA-resistant NB4 cells (NB4-R1, -R2 [67], -R4 [60], and -RA [61]) and mutated-PML-RARα expressing Cos-1 cells [65] indicated that ATRA binding affinity with mutated PML-RARα was generally lower than that with PML-RARα without mutations, due to conformational changes in LBD. Furthermore, ligand-dependent N-CoR/SMRT co-repressor release and co-activator recruitment (e.g. ACTR histone acetyltransferase), which are critical for the transcriptional activation of genes with RARE sites and morphological cell differentiation, was impaired under the therapeutic dose of ATRA [60, 65, 67].

Fig. 3
figure 3

Genetic mutations resulting in amino acid substitution in PML-RARα LBD confirmed in clinically ATRA-resistant patients and ATRA-resistant cell lines. Mutations are confirmed in 3 cluster regions (zones I to III) in RARα-LBD [66]. Red letters indicate amino acids substituted in specific patients and/or cells. Amino acid substitutions and deletions in ATRA-resistant patients are indicated in blue letters. Substitution in ATRA-resistant cell lines indicated in black. Names of cell lines are indicated in brackets. The position of the mutation is described with reference to normal amino acid sequence of RARα1 [31]

To overcome ATRA resistance, a number of therapeutics has been tested in vitro and in vivo. Several clinical reports indicated that As2O3 rescue most of relapsed/refractory patients treated with ATRA/chemotherapy [912, 68]. Am80, a synthetic retinoid that shows higher binding affinity with PML-RARα than ATRA, is utilized in the clinical setting [6971]. Am80 is approximately 10 times more potent than ATRA as an in vitro inducer of differentiation in NB-4 and HL60 cells, and is chemically more stable than ATRA [72, 73]. Histone deacetylase (HDAC) inhibitors [74], such as sodium butyrate (NaF), valproic acid (VPA), and trichostatin A (TSA), have been utilized with ATRA and are expected to transcriptionally activate PML-RARα target genes to inhibit co-repressors complexes that contain HDACs [7577]. Another approach to overcoming the resistance uses other molecular targeting therapeutics, such as gemtuzumab ozogamicin (GO), an anti-CD33 monoclonal antibody linked with calicheamicins [78, 79].

Molecular mechanisms of resistance to As2O3

Even for relapsed/refractory patients following treatment with ATRA/chemotherapy, As2O3 therapy is highly effective, with a complete remission rate of more than 80 % [8082]. Although the CR rate is high even in relapsed patients, resistance to As2O3 treatment has been recognized as a clinically critical problem. Information on As2O3 resistance remains limited compared with that on ATRA resistance.

Recently, we reported two cases showing clinical As2O3 resistance after treatment with ATRA/chemotherapy, which exhibited missense mutations leading to substitution of amino acids in the PML-B2 domain in PML-RARα [50, 68, 83]. One patient with the M3 variant, expressing PML-RARα short form without nuclear localizing signal (NLS) [84], showed ATRA and As2O3 resistance at his terminal stage. Significant clonal expansion of PML-RARα mutant leading to A216V (PML-B2 domain mutation) and G391E (RARα-LBD mutation) was confirmed in leukemia cells harvested at the terminal stage (Fig. 4a, b). In vitro analysis using wild-type and mutant PML-RARα (PR-B/L-mut)-expressing HeLa and HL60 cells indicated that PML-RARα (short form) localized in cytoplasm as micro speckled pattern without As2O3, and as a macro granular pattern after adding As2O3 (Fig. 4c; PML-RARα). In contrast, PR-B/L-mut localized in cytoplasm with diffuse pattern without As2O3, and no change was confirmed in the presence of As2O3(Fig. 4c; PR-B/L-mut). Another case carried an L218P mutation, also in the PML-B2 domain (PR-B2-mut2), in PML-RARα long form with NLS. PML-RARα long form localized in nucleus, while PR-B2-mut2 was diffusely localized in the nucleus. No change was confirmed with or without As2O3 (Fig. 4c; PR-B2-mut2). Further in vitro analysis using PML-RARα overexpressed HeLa cells indicated that SUMOylation of PR-B/L-mut and PR-B2-mut2 after As2O3 treatment was strictly impaired. Recent reports have indicated that direct As2O3 binding to PML-B2 domain is critical for the serial reaction including SUMOylation, multimerization, and degradation [29, 30]. Jeanne et al. conclude that dicysteine C212/C213 in PML-B2 domain may be the direct As2O3 binding motif. From these results, genetic mutations identified in As2O3-resistant patients resulting in A216V and L218P may contribute to As2O3 resistance through impairment of direct As2O3 binding to PML-RARα due to conformational changes in As2O3 binding sites. Further accumulation of patients for genetic analyses is required for confirming the clinical significance of PML-B2 domain mutations in As2O3 resistance.

Fig. 4
figure 4

Genetic mutations resulting in amino acid substitution in PML-B2 domain confirmed in clinically As2O3 resistant APL patients. a Schematic representation of PML-RARα chimeric protein with B2-domain mutation. One patient held PML-B2 mutation (A216V) and RARα-LBD mutation (G391E) on short form PML-RARα (PR-B/L-mut), and another patient held PML-B2 mutation (L218P) on long form PML-RARα [68]. b As2O3 direct binding dicysteine motif (C212/C213) [29, 30] and mutated positions in As2O3-resistant patients (C216 and L218) occur quite close to each other. c Flag-tagged PML-RARα short form, PR-B/L-mut, and PR-B2-mut2 were over expressed in HeLa cells with or without As2O3. Over expressed PML were detected by immunofluorescence staining using anti-Flag antibody. When using PML-RARα short form without As2O3, PML body was confirmed in the microspeckled pattern in cytoplasm. After incubation with As2O3, PML bodies showed macro granular patterns. When using PR-B/L-mut or PR-B2-mut2, the PML body showed diffuse pattern in cytoplasm or nucleus. No difference was seen with/without As2O3

Conclusion

Although the overall survival of APL has been significantly prolonged since the introduction of ATRA and As2O3, relapse/refractory disease due to ATRA and/or As2O3 resistance remains a serious clinical problem. Additional genetic mutations in PML-RARα and another gene, such as FLT3-ITD or TP53 [66, 85], may contribute to disease progression and drug resistance in APL. Detailed genomic analyses using clinical samples harvested repeatedly from patients may help for predicting prognosis, selecting effective targeting drugs, understanding molecular backgrounds, and designing sophisticated new therapeutic strategies.