Acute Promyelocytic Leukemia, Retinoic Acid, and Arsenic: A Tale of Dualities

Abstract

Acute promyelocytic leukemia (APL) is driven by the promyelocytic leukemia (PML)/retinoic acid receptor α (RARA) fusion oncoprotein. Over the years, it has emerged as a model system to understand how this simple (and sometimes sole) genetic alteration can transform hematopoietic progenitors through the acquisition of dominant-negative properties toward both transcriptional control by nuclear receptors and PML-mediated senescence. The fortuitous identification of two drugs, arsenic trioxide (ATO) and all-trans-retinoic acid (ATRA), that respectively bind PML and RARA to initiate PML/RARA degradation, has allowed an unprecedented dissection of the cellular and molecular mechanisms involved in patients’ cure by the ATO/ATRA combination. This analysis has unraveled the dual and complementary roles of RARA and PML in both APL initiation and cure by the ATRA/ATO combination. We discuss how some of the features unraveled by APL studies may be more broadly applicable to some other forms of leukemia. In particular, the functional synergy between drugs that promote differentiation and those that initiate apoptosis/senescence to impede self-renewal could pave the way to novel curative combinations.

Based on clinical and cytological explorations, acute promyelocytic leukemia (APL) was first identified as a specific clinical entity of rapid and unfavorable evolution: or, “the most malignant form of acute leukemia” (Hillestad 1957). APL was later associated with the presence of a specific chromosomal translocation, t(15,17), present in virtually all patients (Rowley et al. 1977), later found to drive the promyelocytic leukemia (PML)/ retinoic acid receptor α (RARA) fusion gene, paving the way to the molecular pathogenesis of this condition. Although APL can occur in children, it cannot be considered as a disease of children, neither as a cancer whose physiopathology is directly linked to developmental biology. Actually, APL can occur at every age, with an almost constant incidence (Vickers et al. 2000). Some reports have suggested that its clinical course is more severe in children (Iland et al. 2023). Some very rare familial cases were reported, but the biological reasons underlying these observations remain unknown. APL is an important model disease, deserving an article in this collection as a unique example of cure by targeted therapies.

Recent reviews have addressed APL genetics (Geoffroy and de Thé 2020). Briefly, APL and related diseases are associated with chromosomal translocations always involving a retinoic acid receptor (RAR). In the immense majority of cases, RARA is involved, but, recently, some very rare cases of atypical APLs were shown to harbor retinoic acid receptor γ (RARG) (Wu and Gao 2023) or, more rarely, retinoic acid receptor β (RARB) fusions (Osumi et al. 2018; Zhu et al. 2023). Most RARA fusions involve PML, but a variety of other variant RARA fusions were described, the most common of which is PLZF (Licht et al. 1995). Moreover, viral insertions within the RARA gene, presumably driving its overexpression, may drive an APL-like disease (Astolfi et al. 2021; Chen et al. 2022). The fact that the incidence of APL is essentially constant with age argues that it is driven by a single rate-limiting dominant genetic hit. Nevertheless, many cooperating oncogenic activations were associated with disease progression (FLT3) or relapse (WT1) (Akagi et al. 2009; Wartman et al. 2011; Madan et al. 2016; Lehmann-Che et al. 2018). Critically, the APL genome is remarkably stable when compared to most other tumors.

The cornerstones of APL therapy are arsenic trioxide (ATO) and all-trans-retinoic acid (ATRA), two targeted therapies that were discovered largely by chance and now cure over 95% of patients without DNA-damaging drugs (Estey et al. 2006; Lo-Coco et al. 2013, 2016). The physiopathology of the disease and its response to therapy have been deciphered with an unprecedented level of molecular details (Lallemand-Breitenbach et al. 2012; Dos Santos et al. 2013; de Thé et al. 2017). Here, we propose to review recent studies on the biology of APL oncogenesis, and show how the dual targeting of its PML and RARA moieties contributes to yield a leukemia cure. We also highlight how these mechanisms may be more broadly involved in leukemia biology.

PML/RARA, THE MASTER DRIVER OF “CLASSIC” APLs

The constant involvement of a RAR in the various fusion proteins demonstrates that deregulated retinoic acid signaling plays a key role in the initiation of APL. RARs are transcription factors whose activity is regulated by a group of vitamin A–derived ligands, retinoids. They control the differentiation of many tissues and exert potent effects on stem cells, including hematopoietic stem cells (Fig. 1; Cabezas-Wallscheid et al. 2017; Schönberger et al. 2022). Studies combining gene inactivation and treatment with RARA-specific retinoids have clearly demonstrated the role of RARA in myeloid differentiation (Kastner et al. 2001; Walkley et al. 2002). Expression of dominant-negative RARA defective for transcriptional activation has pro-oncogenic properties in many cellular systems, including hematopoietic progenitors or hepatocytes, as well as in specific forms of breast cancers (Tsai et al. 1992; Tsai and Collins 1993; Yanagitani et al. 2004; Khetchoumian et al. 2007; Tan et al. 2015). In several of these biological systems, mere overexpression of RARA has the same consequences (Fig. 1; Du et al. 1999). Collectively, this suggests that RARA-mediated target gene repression is important for the initiation of APL and more broadly, for transformation.

Figure 1.

Biology of acute promyelocytic leukemia (APL) oncogenesis. In physiological conditions, retinoic acid receptor α (RARA) is essential for normal hematopoiesis. However, RARA has oncogenic properties when mutated or overexpressed, and the fusion of RARA with promyelocytic leukemia (PML) drives APL through further transcription deregulation and disorganization of PML nuclear bodies (NBs).

Modulation of nuclear receptor-mediated transcription is regulated by their ligand-dependent association with protein complexes that regulate epigenetic status (including DNA and histone modifications), transcriptional initiation, and elongation. One of these key complexes is the nuclear receptor corepressor/silencing mediator of retinoic acid (N-Cor/SMRT), which avidly binds RARA and, presumably, even more avidly RARA fusions, because of their ability to self-dimerize (Fig. 2; Lin and Evans 2000; Minucci et al. 2000). PML-enforced PML–RARA/RXRA dimerization also results in a greatly extended repertoire of the oncoprotein DNA-binding sites (Kamashev et al. 2004; Martens et al. 2010). It was thus suggested that the bases for APL initiation relied on basal transcriptional repression of RARA targets, but also other nuclear receptor target genes, controlling myeloid differentiation and/or self-renewal (Fig. 1). However, the overexpression of normal RARA can contribute to myeloid malignancies ex vivo or in vivo (Du et al. 2000; Astolfi et al. 2021; Chen et al. 2022), suggesting that deregulation of RARs signaling is absolutely central, while the deregulation of de novo targets resulting from PML/RARA dimerization may be an additional contributor to malignant transformation. Recent studies have demonstrated that for some target genes, PML/RARA can be an activator (Tan et al. 2021), although whether this feature is essential for transformation is unknown. Finally, whether the deregulated expression of a small number of master RARA target genes suffices to recapitulate APL initiation remains to be clarified. Identification of these key downstream effectors would bear considerable importance for our understanding of leukemia biology, as some may be shared with other types of acute myelogenous leukemias (AMLs).

Figure 2.

Promyelocytic leukemia (PML)/retinoic acid receptor α (RARA): a double hit oncoprotein. The oncofusion protein disorganizes both PML and RARA structure/function. PML nuclear bodies (NBs) are disorganized into a microspeckled pattern and lose their function, and RARA target genes are repressed due to the recruitment of corepressors, which will block acute promyelocytic leukemia (APL) cells at the promyelocyte stage.

IS DIMERIZATION THE ONLY CONTRIBUTION OF PML TO TRANSFORMATION BY PML/RARA?

At least 98% of APL-associated fusion proteins involve PML. Either the emergence of these fusions is greatly facilitated by some topological features of the PML and RARA genes, or PML has a key role in oncogenesis. This may reflect either a central contribution of PML to the gain of functions of PML/RARA, or a critical contribution of PML allele loss to the oncogenic process. Experiments in transgenic mice demonstrated that self-dimerizing RARA inefficiently promoted the development of an APL-like disease (Sternsdorf et al. 2006). Thus, it is likely that PML provides some specific features to PML/RARA that favor transformation (Occhionorelli et al. 2011). Actually, fusion of PML to RARB or RARG may also promote transformation (Marinelli et al. 2007). Those might relate to the ability of PML to promote sumoylation (Sahin et al. 2014; Tessier et al. 2022) as posttranslational modification that was repeatedly involved in transcriptional repression (Garcia-Dominguez and Reyes 2009). In that respect, a sumoylation-defective PML/RARA mutant is impaired in its transformation abilities (Zhu et al. 2005).

Importantly, many studies have shown that PML has some features of a tumor-suppressor gene (Salomoni and Pandolfi 2002; Bernardi and Pandolfi 2007). Yet, PML may also promote the survival of some cancer cells and is almost never deleted or mutated in cancer evolution (Carracedo et al. 2012). PML is a gene that has fascinated cell biologists by its ability to nucleate subnuclear domains, PML nuclear bodies (NBs), that play a key role in stress responses and posttranslational modifications of an ever-growing list of partner proteins (Lallemand-Breitenbach and de Thé 2010, 2018). In multiple cellular systems, PML is required for senescence induction (Hsu and Kao 2018; Patra and Müller 2021). Importantly, PML/RARA disrupts the assembly of PML NBs, yielding a “microspeckled” distribution that can be used for APL diagnosis (Fig. 2; Daniel et al. 1993; Dyck et al. 1994, 1995; Koken et al. 1994). Pointing to a role of PML and NB disruption in APL pathogenesis, APL initiation is modestly accelerated in Pml absence (Rego et al. 2001; unpubl. data). Thus, PML function is altered during APL pathogenesis and this likely contributes to leukemogenesis. The actual mechanistic links between PML NB formation and tumor-suppressive properties remain debated. The first proposed mechanism was control of P53 function, most likely through the ability of NBs to recruit P53 and most of the enzymes enforcing its posttranslational modifications (Pearson et al. 2000; Matt and Hofmann 2018). Yet, PML may also drive P53-independent senescence (Mallette et al. 2004) and profoundly affects mitochondrial functions (Carracedo et al. 2012; Ito et al. 2012). Finally, sumoylation was found to exert prosenescent functions (Bischof and Dejean 2007). This may be promoted by PML NBs, which control global sumoylation (Sahin et al. 2014; Tessier et al. 2022).

Another topic of interest is the actual cell of origin of APL (Guibal et al. 2009; Wojiski et al. 2009). Studies in mouse models or in ex vivo transformation systems have suggested that PML/RARA can promote the proliferation of stem cells, not only myeloid progenitors (Welch et al. 2011). Careful analysis of preleukemic APL mice and their transition to frank APL could provide important insights into the emergence of APL, although some mice/human specificities may complicate this analysis. Altogether, the fusion of PML to RARA yields both novel properties of transcriptional deregulation and disruption of PML NBs (Fig. 2; de Thé et al. 2017). Mechanistically, some studies have suggested a direct role of PML in the acquisition of the transcriptional repression phenotype of PML/RARA (Occhionorelli et al. 2011). Conversely, PML/RARA-associated complexes were implicated in the inhibition of NB assembly (Shima et al. 2013). The PML/RARA fusion exemplifies a remarkable model in which transformation can be driven by a single dominant alteration that disrupts pathways controlled by each of its constitutive moieties.

APL THERAPIES, FROM EMPIRICISM TO CURE BY TARGETED DRUGS

The history of the successive APL therapies was reviewed elsewhere and will be only summarized here to show how they contributed to sequential changes in paradigms, illuminating APL therapy response and ultimately driving its cure. Therapeutically, APL was shown to be sensitive to high-dose anthracyclines in the early 1970s, although the latter often aggravated the bleeding diathesis and resulted in a very high incidence of early deaths (Bernard et al. 1973). Yet, up to 20% of patients could be cured by chemotherapy alone.

The ability of retinoids to differentiate many cell lines, including AML ones (Breitman et al. 1980, 1981), led to the first human trials of ATRA, which induced rapid in vivo differentiation of the leukemic cells and showed clearance of the disease, culminating in complete remissions (Fig. 3; Huang et al. 1988; Castaigne et al. 1990; Chomienne et al. 1990). This was the first success of “differentiation therapy” (Warrell et al. 1993; Degos et al. 1995). The success of retinoic acid treatment initiated the cloning of PML/RARA and the first physiopathological models wherein ATRA reversed the transcriptional repression of PML/RARA-silenced genes, thereby initiating differentiation of leukemic cells (de Thé et al. 1990, 1991). Differentiated APL cells (granulocytes) being very short-lived, terminal APL cell differentiation drives disease clearance. Yet, in most cases, patients rapidly relapsed, sometimes bearing on-target mutations on PML/RARA that impeded transcriptional reactivation by ATRA, formally demonstrating that ATRA is a PML/RARA-targeted therapy (Gallagher 2002). Bypassing these relapses, a combination of retinoic acid and anthracycline allowed APL eradication and could cure two-thirds of patients, so the introduction of ATRA was an actual therapeutic revolution (Fenaux et al. 1994).

Figure 3.

All-trans-retinoic acid (ATRA) and arsenic trioxide (ATO): a synergy to drive 98% cure in acute promyelocytic leukemia (APL) patients. Fewer than 5% of patients treated with single-agent ATRA are cured. Indeed, ATRA mainly induces transient complete remissions. ATO can cure up to 70% APL patients, due to its ability to target promyelocytic leukemia (PML) and trigger loss of self-renewal. Finally, the ATRA + ATO combination can lead up to 95% cures; the two molecules synergize as they target the retinoic acid receptor α (RARA) and PML moieties independently (see also Fig. 4).

Yet, the devil is in the details and it progressively appeared that several pieces did not fit in the puzzle. Not all retinoids have clinical efficacy, despite their ability to reactivate transcription and some rare variant RARA fusions were clinically insensitive to ATRA (Licht et al. 1995). Murine models clearly demonstrated that differentiation and APL regression could be uncoupled, genetically and pharmacologically (Koken et al. 1999; Nasr et al. 2008; Ablain et al. 2013). Subsequent studies in other models clearly established that differentiation of AML cells may not be the irreversible process that had once been foreseen (McKenzie et al. 2019), so that differentiation through transcriptional control was likely not the only driver of ATRA therapy.

Further understanding of APL therapy came from another clinical revolution: the exquisite sensitivity of APLs to ATO (Chen et al. 1996, 1997). Clinically, ATO may definitively cure up to 70% of APL patients as a single agent, including those that had become resistant to ATRA (Mathews et al. 2006, 2010). Thus, ATO is clinically far more efficient than ATRA. Interestingly, ATO primarily drives differentiation in vivo, although some apoptosis was also noted ex vivo (Chen et al. 1996, 1997). Critically, ATO directly targets PML/RARA, enforcing its degradation (Fig. 3; Chen et al. 1997; Zhu et al. 1997; Muller et al. 1998; Lallemand-Breitenbach et al. 2001; for review, see Zhu et al. 2002). PML/RARA degradation restores PML NB formation. Targeting of PML/RARA is enforced through its PML moiety, since ATO dramatically alters PML NBs and also degrades normal PML proteins (Zhu et al. 1997). Subsequent studies have dissected, in great molecular detail, the mechanisms of ATO binding onto PML and its consequences with respect to cell biology and biochemistry (Lallemand-Breitenbach et al. 2001, 2008; Jeanne et al. 2010; Zhang et al. 2010; Lallemand-Breitenbach and de Thé 2018; Bercier et al. 2023). Actually, ATRA also degrades PML/RARA and RARA through a molecular mechanism involving the UBR5 ubiquitin E3 ligase (Zhu et al. 1999; Kopf et al. 2000; Bruck et al. 2009; Tsai et al. 2023). Thus, both active APL drugs target each of the constitutive moieties of the PML/RARA fusion to initiate its degradation (Fig. 4). Following the PML/RARA example, many strategies (such as proteolysis-targeting chimeric drugs [PROTACs]) were implemented to initiate degradation of other driver oncogenes (Dale et al. 2021).

Figure 4.

Mechanisms of acute promyelocytic leukemia (APL) cure by all-trans-retinoic acid (ATRA) + arsenic trioxide (ATO). ATRA or ATO induces promyelocytic leukemia (PML)/retinoic acid receptor α (RARA) degradation. Oncoprotein loss initiates passive differentiation, apoptosis, and senescence. ATRA may also actively induce APL cell differentiation, while ATO activates PML-induced senescence and apoptosis, allowing patient cure.

In principle, the degradation of a driver oncogene is the best way to inactivate it. Indeed, the dominant-negative effects of PML/RARA on retinoic acid signaling and NB assembly should disappear. Actually, in several mice models of Myc-induced leukemogenesis, extinction of oncogene expression precipitates apoptosis or senescence to drive short-lived remissions (Felsher and Bishop 1999; Jain et al. 2002). In other AML models, functional inhibition of oncogenes can drive terminal differentiation (Wang et al. 2013; Amatangelo et al. 2017; de Thé 2018). This raised an important question with respect to ATRA activity: is PML/RARA degradation sufficient for differentiation and/or APL clearance or is transcriptional reactivation of PML/RARA and RARA of any importance for clinical responses? While some of these questions remain unsettled, the use of mouse models has provided some important answers. In murine APL models, the absence of RARA did not alter the initial onset of differentiation (Kogan et al. 2000). Analyses of “uncoupled” retinoids that efficiently initiate RARA and PML/RARA-dependent transcription, but not degradation, demonstrated that these drugs are much less potent than ATRA to initiate APL clearance in vivo, despite yielding similar differentiation (Ablain et al. 2013). Genetic and pharmacological studies demonstrated that PML/RARA degradation is essential for ATRA response (Nasr et al. 2008, 2009; Ablain et al. 2013). Transcriptional activation through RARA or PML/RARA may still be of some importance. Yet, that ATO induces APL differentiation in vivo with similar kinetics and efficiency as ATRA suggests that PML/RARA loss suffices to initiate this process (Vitaliano-Prunier et al. 2014). One should stress that in APL, ATRA is almost never curative on its own. This may in fact reflect some pharmacodynamic issues, as liposome-mediated delivery of ATRA resulted in some long-term remissions (Douer et al. 2001; Tsimberidou et al. 2006), possibly reflecting a more efficient PML/RARA degradation. Yet, note that the master genes silenced by altered RARs to initiate APL self-renewal and differentiation arrest remain to be identified.

Importantly, mouse models clearly showed that normal PML is absolutely required for clinical response to either ATRA or ATO in vivo (Ablain et al. 2014; unpubl. data). Thus, therapy-induced NB reformation is an essential determinant of APL response (Rérolle and de Thé 2023). Nuclear body restoration may be the key difference between mere oncogene extinction, driving short-term remission, and long-term response or cure (Fig. 4). In the case of APL, PML/RARA loss allows PML NB reformation, which has additional growth-suppressive function independently from restored transcriptional control of retinoid targets. Indeed, PML is a master gene of senescence, at least in part through NB formation. Moreover, since ATO enforces PML NB formation independently from PML/RARA degradation, ATO-enhanced PML NB targeting is likely to be an important contributor to ATO clinical efficacy (Fig. 4). Similar to PML overexpression-driven senescence, the molecular mechanisms through which PML NB reformation upon PML/RARA destruction is required for full APL response remain imperfectly understood. PML is required for activation of P53 upon ATRA and/or ATO treatments. Moreover, in this context, PML is more important than P53 in driving responses. This could be mediated by PML global effects on sumoylation, a posttranslational modification shown to favor senescence and to be tightly linked to chemotherapy sensitivity in AMLs (Bossis et al. 2014; Tessier et al. 2022). Critically, PML mutations were discovered in therapy-resistant patients for whom PML/RARA remained normal (Lehmann-Che et al. 2014; Iaccarino et al. 2016), a situation that we recently recapitulated in murine models (unpubl. data), genetically demonstrating the critical requirement of normal PML for ATO response.

Analysis of these different responses to therapies highlights an important caveat regarding therapeutic responses in AML. In APL, the latter involves both differentiation and loss of self-renewal. In contrast to initial beliefs, the two may be somehow disentangled, as some self-renewal may persist even in terminally differentiated cells (Nasr et al. 2008; de Thé 2018; McKenzie et al. 2019). Clearly, loss of PML/RARA-mediated repression releases a differentiation program that is critical for the initial tumor debulking. A different, complementary, PML-mediated loss of self-renewal/senescence program is also required for the full therapeutic effect (Ito et al. 2008). Thus, PML NB reformation can be “passive,” following PML/RARA degradation by ATRA, or “active” following ATO-enforced NB biogenesis through targeting of normal PML (Fig. 4; Bercier et al. 2023). ATRA primarily induces differentiation through transcriptional control; ATO primarily induces loss of self-renewal through its ability to target PML (Figs. 3 and 4; Bercier et al. 2023). This explains why the combination of the two drugs is highly synergistic (Figs. 3 and 4), as first shown in animal models and subsequently validated in patients (Lallemand-Breitenbach et al. 1999; Lo-Coco et al. 2013, 2016). Mechanistically, this frontline combination induces synergistic PML/RARA degradation (because the degrons activated by ATRA or ATO binding are different) and maximal PML NB reformation through direct ATO-targeting of PML. The ATRA/ATO combination is now the gold standard, with more than 95% cures, at least in patients enrolled in clinical trials (Lehmann et al. 2017).

Overall, the APL model has highlighted a remarkable series of dualities: fusion of PML to RARA as the master gene of transformation, PML/RARA degradation through ATRA and ATO binding to its PML and RARA moieties, activation of differentiation, and loss of self-renewal. In times of immuno-oncology, whether APL eradication is purely a cell-autonomous process or whether the latter requires the immune system, as demonstrated in single-agent ATRA therapy (Westervelt et al. 2002; Padua et al. 2003; Robin et al. 2006), remains unsettled.

BROADENING THE LESSONS FROM THE APL MODEL TO OTHER AMLs?

What do 30 years of APL research provide as insights into other malignancies? At the level of pathogenesis, deregulated retinoic acid signaling has been observed in several other malignancies, notably human breast cancers and other AMLs or murine models of hepatocarcinoma (Khetchoumian et al. 2007; Tan et al. 2015). From a therapeutic point of view, AMLs that express high levels of RARA were shown to be intrinsically ATRA-sensitive (McKeown et al. 2017). This observation is in line with ex vivo experiments demonstrating that mere RARA overexpression initiates ATRA-reversible immortalization of murine hematopoietic progenitors (Du et al. 1999; Zhu et al. 2007). Mechanisms involved are currently uncharacterized, but may resemble studies wherein overexpression of thyroid hormone receptors mimicked expression of its dominant-negative mutant v-ErbA for its ability to immortalize avian erythrocytic progenitors (Bauer et al. 1998). Importantly, some clinical translations of these findings were made in AML patients unfit for chemotherapy for whom a clinical benefit was obtained from the addition of ATRA to conventional treatments (Lübbert et al. 2020; de Botton et al. 2023). Of note, several studies demonstrated that enforcing cell cycle arrest of AML cells renders them susceptible to ATRA-induced differentiation (Boutzen et al. 2016; Mugoni et al. 2019). Finally, dominant-negative RARA mutations were observed in non-APL AMLs treated by high-dose ATRA for suspicion of APL (Zhao et al. 2019), genetically demonstrating a strong RARA-dependent selective pressure of ATRA treatment. While the RARA downstream effector genes are unknown, they most likely overlap with PML/RARA-repressed genes that initiate APL. Thus, these novel ATRA/AML connections could open some important translational opportunities in a disease that remains incurable in patients unfit for chemotherapy.

PML has also rich connections with human pathogenesis. It may be involved in other chromosomal translocations (Kurahashi et al. 2011) and is tightly linked to many viral infections, through its ability to act as a first-line innate response (Scherer and Stamminger 2016). PML was also tightly linked to the natural history of many epithelial cancers, massively induced at the initial stages of transformation (possibly as a reflection of interferon [IFN] induction), but lost at an advanced stage, perhaps reflecting bypass of a senescence checkpoint (Koken et al. 1995; Gurrieri et al. 2004). This is actually very reminiscent of the activation of DNA damage response upon oncogenic activation (Bartkova et al. 2005; Gorgoulis et al. 2005). However, in P53-mutant tumors, PML was proposed to have prosurvival functions, possibly linked to the effects of PML on mitochondrial fitness (Carracedo et al. 2012; Ito et al. 2012). In some settings, PML expression in tumors is tightly correlated to survival. The absolute requirement of PML for APL response to ATRA or ATO begged the question of its role in the pathogenesis or therapeutic response of other hematological malignancies. In AMLs associated with mutation of the NPM1 chaperone, mutant NPM1 impedes PML NB formation (El Hajj et al. 2015; Martelli et al. 2015). Oxidative stress induced by Actinomycin D restored PML NB formation and the latter was required for AML cell lines to respond through senescence induction (Gionfriddo et al. 2021; Wu et al. 2021). PML expression levels are primarily tuned by IFN signaling (Stadler et al. 1995). Myeloproliferative neoplasms (MPNs) driven by an activating mutation in Jak2 are often treated with IFN, which induces a slow clearance of transformed cells (Kiladjian et al. 2008). PML NB biogenesis is dependent on PML protein expression, PML assembly upon ATO treatment or ROS exposure, and PML degradation by a number of well-identified proteolytic pathways (Gamell et al. 2014; Niwa-Kawakita et al. 2017; Lallemand-Breitenbach and de Thé 2018). Thus, maximal NB formation can be enforced by a combination of IFN and ATO (Quignon et al. 1998). Should PML NBs be one of the downstream effectors of IFN activity in MPN, its activity should be enhanced by ATO. A dramatic synergy was indeed observed in murine models of the disease for MPN clearance and leukemia-initiating cell eradication. Importantly, the latter was dependent on Pml presence (Dagher et al. 2021). How general is the PML dependency for response to cancer therapy awaits further studies. Given the role of PML to control P53 signaling and the importance of the latter in chemotherapy response, it would be particularly interesting to explore Pml dependency on chemotherapy response in murine cancer models. Similarly, a structure–function analysis of PML role in therapy response could orient as to which of the many proposed functions of PML (sumoylation control, mitochondrial fitness, antioxidant, P53 control, etc.) is actually responsible for these effects.

CONCLUDING REMARKS: TOWARD A BIOLOGY OF CURE?

APL is currently one of the only examples of cure by targeted therapies that has now become the clinical gold standard. As such, it has allowed an unprecedented mechanistic exploration of the molecular requirements for disease eradication. It was long thought that dissection of APL pathogenesis and therapy response would only be relevant to APL. Yet, recent studies have shown that deregulated RARA signaling (and ATRA sensitivity) is more broadly implicated in AML than the sole APL setting (McKeown et al. 2017). Similarly, PML contributes to therapy response for several other leukemia/therapy pairs (Rérolle and de Thé 2023). In fact, studies of ovarian cancer response to platinum-based chemotherapy have also suggested a positive impact of PML expression, at least in part through mitochondrial control (Gentric et al. 2019). More broadly, PML is a key regulator of sumoylation (Tessier et al. 2022), a posttranslational modification essential to multiple forms of stress responses, including senescence induction (Bischof and Dejean 2007). Mechanistically, how PML arbitrates post-therapy cell fate remains to be determined.

Beyond the targeting of retinoid signaling by ATRA and PML/sumoylation by ATO, APL studies could enlighten and broaden the dual contributions of differentiation and senescence in cancer therapy (Fig. 4). In APL, differentiation is insufficient for cure per se, but most likely contributes to cure through both transient debulking and some reduction in self-renewal. Retinoids are not the only drugs that may initiate leukemia differentiation (Wang et al. 2013; de Thé 2018). Similarly, senescence and cell cycle arrest may be triggered by many drugs. The synergism between differentiation and senescence may constitute a model for future curative associations in AMLs, and perhaps more broadly in cancers, showing how their dual effects discovered in APL could cooperate in other conditions.