PML-driven sumoylation of PML::RARA-bound repressors drives hematopoietic progenitor immortalization

Acute promyelocytic leukemia (APL) is driven by fusion proteins always involving one of the three retinoic acid (RA) receptors (Geoffroy et al., 2021; Zhang and Qiu, 2025), pointing to the critical role of deregulated RA signaling in initiation of the disease. In most patients, these fusions involve RARA and PML (de Thé et al., 1990a). While the pathways involved in RA and arsenic trioxide (ATO) triggered PML::RARA degradation and APL cure have been largely deciphered (de Thé et al., 2017; Dos Santos et al., 2013; Rerolle et al., 2024), why RARA and PML are the preferred fusion partners driving APL initiation remains poorly understood. RARA binds corepressors more avidly than other RARs, and repression by RA receptors plays an important role in development (Weston et al., 2003). PML::RARA disrupts both PML nuclear body formation and nuclear receptor–mediated transcriptional control (Nasr et al., 2009). Dominant-negative RARA mutants exert transforming activities in a variety of biological systems, including some breast cancers (Saitou et al., 1995; Tan et al., 2015; Tsai et al., 1992). PML::RARA is an even more potent repressor than RARA (de Thé et al., 1991). Yet, the molecular bases for PML::RARA super-repressive activity remain imperfectly understood. PML-facilitated PML::RARA homodimer formation was proposed to enhance HDAC and corepressor binding (Lin and Evans, 2000), while tethering of PML-bound repressors such as DAXX (Zhu et al., 2005) may contribute to PML::RARA-driven repression. Artificial RARA homodimers inefficiently initiate APL, even in Pml null cells, which implies that PML fusion to RARA entails an unidentified gain of function central to leukemogenesis (Licht, 2006; Sternsdorf et al., 2006; Voisset et al., 2018).

Sumoylation is a reversible posttranslational modification implicated in stress responses. Sumoylation is mediated by UBC9, the universal SUMO E2 ligase, in an enzymatic cascade similar to ubiquitination (Celen and Sahin, 2020). In the context of chromatin, sumoylation is associated with transcriptional repression (Tharuka et al., 2025). PML is a massively sumoylated protein, which serves as a scaffold to drive SUMO2 conjugation of a variety of nuclear body–associated partner proteins, particularly upon stress (Sahin et al., 2014; Tessier et al., 2022). SUMO2 conjugation was also linked to stress-induced target catabolism, as discovered by studying ATO-induced PML::RARA degradation (Jaffray et al., 2023; Jaffray et al., 2025; Lallemand-Breitenbach et al., 2001; Lallemand-Breitenbach, 2008). Within regulatory complexes, sumoylation often occurs on multiple members favoring complex stability and regulating their functions (Psakhye and Jentsch, 2012). For example, the nuclear corepressor complex, assembled by NCoR1/2, may undergo SUMO-facilitated nuclear receptor association and/or transcriptional repression (Tiefenbach et al., 2006). Here, we demonstrate that fusion of PML to RARA drives sumoylation of RARA-associated corepressors and other key partners to promote transcriptional repression and progenitor immortalization.

In primary murine hematopoietic progenitors, PML::RARA expression results in their enhanced self-renewal and differentiation block, as does overexpression of RARA, but not RARB or RARG (Fig. 1 A) (Du et al., 1999; Zhu et al., 2005). Yet, only PML::RARA ensures long-term progenitor immortalization after the fifth replating, contrasting with RARA (Fig. 1 A). Interestingly, the levels of RARA protein expression were consistently much higher than those of PML::RARA at the second passage (Fig. 1 B), suggesting that the higher RARA expression levels in immortalized progenitors compensate for its lower intrinsic repressive activity. Importantly, fusion of the POZ repression domains of PLZF onto RARB allowed efficient progenitor immortalization ex vivo (Fig. 1 A), highlighting the need of a strong basal transcriptional repression of RAR target genes in this process.

In RA absence, PML::RARA or RARA potently binds the NCoR1/2 complexes, and RARA I396 residue is required for this binding (le Maire et al., 2010). Importantly, PML::RARA^I396E or RARA^I396E failed to efficiently repress basal expression of RA-sensitive reporters and no longer immortalized primary progenitors (Fig. 1, C–E). Collectively, repression of RA receptor target genes by NCoR-bound PML::RARA or RARA is required for primary hematopoietic progenitor immortalization, but can be mimicked by fusion of a strong repressive domain onto another RA receptor.

To identify partners involved in enhanced repression by PML::RARA, we expressed a DHFR-PML::RARA-BioID (thereafter indicated PML::RARA-BioID) fusion in mouse Trp53^−/− HPC-7 stem/progenitor cells, which exhibit very similar features to their parental cell (except for downregulation of basal Trp53 signaling and activation of Myc targets, as expected, Fig. S1 A), but can tolerate stable PML::RARA expression (Ferrucci et al., 1997). Comparison between DHFR-NLS-BioID and PML::RARA-BioID interactants identified 379 specific interactors (Fig. 2 A, Fig. S1 B, and Table S1). 20 of them were recently identified at high confidence (log₂ fold change >2.5) using a similar approach in different cellular systems (Katerndahl et al., 2024). A detailed comparison between PML::RARA interactors in these two systems is presented in Fig. S1 D. The greatest enrichments were observed for RARA interactors with repressive abilities, including NCoR1/2, NRIP, GPS2, HDACs, SUMO2 (Fig. 2 A), although some transcriptional activators (NCOA2/3) were also identified (Fig. 2 A). Unexpectedly, we identified mRARA/B/G, which are not considered to be direct RARA interactors. PML binders (SUMO1/2, PIAS, TET2) were also efficiently purified. In biochemical validation experiments, RA led to the expected dissociation of NCoR or GPS2, but greatly enhanced interactions with NCOA and NRIP1 (Fig. 2 B).

A significant number of these PML::RARA-interacting proteins were recently identified by SUMO2 proteomics in ATO-treated APL cells in vivo, notably NCoR1/2 and NRIP1 (Fig. 2 C and Fig. S1 C) (Tessier et al., 2022). Importantly, in transfected 293T cells, PML::RARA, but not RARA, allowed NCoR conjugation by SUMO2 (reversed by the SUMO inhibitor, TAK-981 [Lightcap et al., 2021]) (Fig. 2 D), even in ATO absence, while interestingly, this was not the case for SUMO1 (Fig. 2 E). Thus, similar to PML partners (Sahin et al., 2014), RARA-interacting proteins may also undergo SUMO2 conjugation in the presence of PML::RARA, most likely through PML-mediated tethering of UBC9.

To validate that other PML::RARA interactants may undergo basal PML::RARA-driven SUMO2 conjugation, we then expressed PML::RARA-BioID and His₁₀-SUMO2 in HPC7 cells and performed a stringent dual streptavidin plus nickel–nitrilotriacetic acid (Ni-NTA) purification to identify SUMO2-conjugated PML::RARA interactants by mass spectrometry (Fig. 3 A). NCoR1/2 were again the top hits, together with other RARA-associated master regulators including RXRA, GPS2, GATA2, IRF2BP2, NSD2, and HDAC3/5/7/9 (Fig. 3, B and C) (Andrade et al., 2016; Barysch et al., 2021; Chun et al., 2003), while other sumoylated interactors were associated with PML nuclear bodies (TET2, PIAS1, KAP1) (Fig. 3 D). To demonstrate that PML::RARA may favor basal sumoylation of RARA-associated partners in a physiological setting, we used progenitors derived from His₁₀-SUMO₃ knock-in mice expressing, or not, PML::RARA at the preleukemic stage (where PML::RARA expression is barely detectable) or after evolution to full-blown APL (where PML::RARA becomes abundantly expressed [Westervelt et al., 2003]). Only primary leukemic cells showed highly enhanced basal SUMO2 conjugation of PML::RARA-associated proteins, particularly NCoR and its associated proteins such as GPS2, as well as GATA2 or IRF2BP2 (Fig. 3 E). Basal levels and/or sumoylation of PML or PML-associated proteins were decreased, possibly reflecting the distinct differentiation status of those cells.

Among APL-associated fusion proteins, NCoR sumoylation was only observed upon PML::RARA, but not PLZF::RARA or PML::RARG expression (Fig. 3 F), in keeping with the fact that PLZF does not bind UBC9, while RARG does not efficiently recruit NCoR (Farboud et al., 2003). Moreover, PML::RARA point mutants impaired in their UBC9 recruitment ability (L73E or K160R) (Bregnard et al., 2022; Wang et al., 2018; Zhu et al., 2005) yielded a significant decrease in SUMO2 conjugation of RARA-bound NCoR or GPS2 (Fig. 3 G). These mutants are also defective for PML::RARA-driven ex vivo immortalization (Fig. 3 H) and in vivo transformation (Wang et al., 2018; Zhu et al., 2005), tightly correlating efficient NCoR1/2 sumoylation to PML::RARA-driven transformation.

Collectively, the fusion of PML to RARA allows efficient SUMO2 conjugation of multiple RARA partners, the NCoR complex, but also key other transcriptional regulators such as GATA2 and IRF2BP2, potentially modulating their transcriptional output (Chen et al., 2018; Tiefenbach et al., 2006; Wang et al., 2020) and contributing to repression-mediated immortalization.

If a major role of PML is to recruit UBC9 and enforce sumoylation of RARA-associated repressors, UBC9/RARA should efficiently immortalize progenitors. Indeed, UBC9/RARA fusions, but not overexpressed RARA or UBC9/RARA^I396E, allowed long-term immortalization (Fig. 4, A–C; and Fig. S2, A–C). Remarkably, UBC9/RARA fusion boosts self, NCoR, GATA2, IRF2BP2, GPS2, and RARA partner sumoylation in both 293T transfectants, transduced MLL-driven AMLs, and Lin⁻ progenitors stably expressing His₁₀-SUMO2 (Fig. 4, D and E; and Fig. S2 D), and drives potent transcriptional repression (Fig. 4 F). Similar to PML::RARA, only very low levels of UBC9/RARA suffice to confer long-term replating ability (Fig. 4 E). Actually, UBC9/RARA was a more potent repressor than PML::RARA and abolished basal progenitor differentiation (Fig. 4, B, C, and F). UBC9/RARA^I396E still exerted some repressive activities and promoted self-renewal, sharply contrasting with RARA^I396E, implying that other chromatin-associated SUMO2 targets than NCoR contribute to repression (Theurillat et al., 2020). Collectively, anchoring of UBC9 onto RAR target genes promotes their repression and drives progenitor immortalization.

If NCoR sumoylation has any significant role in PML::RARA-dependent transcriptional repression, chemical inhibitors of sumoylation should activate PML::RARA or UBC9/RARA targets. TAK-981 treatment did not alter the cellular localization of RARA fusions (Fig. S2 E), but indeed yielded reactivation of canonical RAR targets, although less efficiently than RA (Fig. 5 A and Fig. S3, A and B). As expected, this was found in PML::RARA- or UBC9/RARA-transformed primary progenitors, but not in RARA- or PLZF::RARA-immortalized ones, while TAK-981 activated Ifng gene in both settings (Fig. S3 C). Thus, repression driven by PML::RARA or UBC9/RARA fusions directly involves sumoylation.

Ex vivo treatment of PML::RARA- or UBC9/RARA-transformed primary progenitors by TAK-981 induced their differentiation (Fig. 5, B–D; and Fig. S3 D). TAK-981-initiated differentiation was also obtained in vivo, using murine APL models (Fig. 5 E), and was accompanied by target gene reactivation (Fig. 5 F). In PML::RARA- and UBC9/RARA-transformed cells, this was associated with a significant decrease in the number of clones (Fig. S3 E) and an even sharper reduction in the total number of cells (Fig. 5 G). Subsequent replating of TAK-981–treated MC I progenitors led to a progressive exhaustion of their clonogenic activities, while RARA-transformed progenitors were essentially unaffected (Fig. 5 H). Functionally, a 5-day ex vivo treatment drove some apoptosis in all transformed progenitors (Fig. S3 F). Collectively, inhibition of sumoylation in PML- or UBC9/RARA fusions reactivates RAR signaling to drive APL growth arrest and differentiation.

Transcriptional repression of RA target genes was proposed to underlie APL pathogenesis (de Thé et al., 2017), as formally demonstrated here by the PML::RARA^I396E mutant, using progenitor immortalization as a surrogate for transformation. In PLZF::RARA variant APLs, the PLZF POZ domain yields a super-repressive phenotype that entails clinical RA resistance (Licht et al., 1995). Why PML is the recurrent fusion partner and whether this relates to transcriptional repression remained poorly understood (Sternsdorf et al., 2006). We identify a novel mechanism for PML::RARA-driven hyper-repression, through PML-mediated sumoylation of PML::RARA-bound proteins, particularly the NCoR complex, whose stability, nuclear receptor binding, and repressive ability are all tightly regulated by sumoylation (Fig. 5 I) (Hua et al., 2016; Tiefenbach et al., 2006). Other key RARA-associated regulatory proteins (e.g., NCoR1/2, GPS2, GATA2, IRF2BP2, NSD2) (Fig. 3 B) also undergo PML-driven sumoylation, which may impinge on PML::RARA function (Andrade et al., 2016; Barysch et al., 2021; Chun et al., 2003). IRF2BP2 may be a RARA fusion partner in very rare APLs (Shimomura et al., 2016; Yin et al., 2015), while GATA2 forms tight complexes with RARA on DNA (Katerndahl et al., 2024; Tsuzuki et al., 2004). Chromatin sumoylation around PML::RARA binding sites may also contribute to transcriptional repression (Stielow et al., 2008). Domains of PML involved in UBC9 recruitment and sumoylation control (RING, K160) were also required for enhanced partner sumoylation, target gene regulation, TAK-981 sensitivity, progenitor immortalization (Fig. 3 H), and, critically, leukemogenesis in vivo (Wang et al., 2018; Zhu et al., 2005). NCoR complex SUMO-interacting motifs may also boost its own binding onto sumoylated PML::RARA, as in other nuclear receptors (Hua et al., 2016; Paakinaho et al., 2021), further contributing to tightly anchor the fusion protein onto this master repressive complex. That UBC9/RARA exerts much stronger transforming effects than RARA and confers phenotypic sensitivity to UBC9 inhibitors strongly argues for a direct functional role of sumoylation of RARA partners and nearby chromatin in transcriptional repression, independently from specific PML interactions such as the DAXX repressor. Finally, we also identified other PML::RARA interactors with potential importance in APL pathogenesis, such as the Ski oncogene, which binds to RARA to inhibit RA signaling (Ritter et al., 2006) and confers RA sensitivity to immortalized progenitors (Dahl et al., 1998; Melling et al., 2013).

Previous studies suggested an obligatory role of PML::RARA as an active effector of RA-induced differentiation. Yet, one of the key PML::RARA-interacting proteins is NRIP1, an RA-dependent repressor. Accordingly, in RAR-less MEFs, PML::RARA remains a very poor activator of primary target genes (C. Esnault, unpublished data). TAK-981–induced target gene reactivation (without PML::RARA loss) demonstrates that derepression suffices to initiate the differentiation program in vivo, extending our observations that PML::RARA destruction by ATO is enough to initiate APL differentiation. PML fusion to RARA drives dimer-dependent gains of function, including relaxed DNA-binding specificity, enhanced corepressor association (Kamashev et al., 2004; Sternsdorf et al., 2006), or recruitment of PML-bound partners (Occhionorelli et al., 2011; Zhu et al., 2005). Previous studies suggested that these are insufficient to drive immortalization on their own (Sternsdorf et al., 2006; Voisset et al., 2018). The experiments reported here illustrate how critical PML/UBC9-driven posttranslational modifications of RARA partners can contribute to transformation, identifying a key gain of function explaining why PML and RARA are the preferred targets of translocations driving APL. Sumoylation was recently shown to modulate oncogenesis in other settings (Li et al., 2025; Zhang et al., 2025), but also to favor therapeutic response through nuclear receptor activation (Valima et al., 2025), possibly opening novel perspectives for sumoylation inhibitors in cancer therapy.

A mouse HPC7 Trp53^−/− stem/progenitor cell line used in this study was generated from the parental HPC7 cell line (a kind gift from Camille Lobry) by using the Alt-R CRISPR/Cas9 technology (Integrated DNA Technologies [IDT]). Briefly, HPC7 cells were electroporated with guide RNA targeting Trp53 (5′-GCG​CTG​ACC​CAC​AAC​TGC​AC-3′) and recombinant Cas9 reagent (cat# 1081061; IDT) following the recommendation of the manufacturer. Single-cell cloning was performed into 96 wells, and knockout of the Trp53 gene was confirmed by sequencing, immunofluorescence, and western blotting analysis (data not shown). Trp53^−/− HPC7 cells were retrovirally transduced with DHFR-PML::RARA-HA-BioID or DHFR-NLS-HA-BioID and selected with puromycin (1 μg/ml) for 3 days and maintained in puromycin-containing culture medium. Puromycin-resistant cells were then retrovirally transduced with His₁₀-SUMO2 (IRES-EGFP) and GFP-sorted by FACS. All HPC7 cell lines were maintained in Iscove’s modified Dulbecco’s medium plus GlutaMAX (cat# 31980030; Thermo Fisher Scientific) supplemented with 5% FBS in the presence of 100 U/ml of penicillin, 100 μg/ml of streptomycin, 0.1 mM 2-mercaptoethanol, and 100 ng/ml of soluble Kit ligand recovered from a CHO cell line stably expressing mouse c-Kit. HEK293T cells were maintained in Dulbecco’s modified Eagle’s medium plus GlutaMAX (cat# 41965062; Thermo Fisher Scientific) supplemented with 10% FBS, 100 U/ml of penicillin, and 100 μg/ml of streptomycin.

To assess transcriptional activities of RARA fusion proteins, HEK293T cells were transfected with 200 ng of indicated pSG5 RARA fusion plasmids together with luciferase reporter vectors (200 ng of pDR5-Tk-FLuc and 200 ng of pTk-RLuc for normalization) as previously described (de Thé et al., 1990b). After 24 h, transfected cells were treated with or without all-trans retinoic acid (ATRA) (1 μM) for 12 h. Measurement of firefly and Renilla luciferase activities was then monitored in a total of 36 h after transfection with the dual-luciferase assay system kit (Promega) following the recommendation of the manufacturer.

All plasmid constructs are listed in Table S2. To generate the pCMV-PML::RARA-BioID2-HA plasmid, an NheI/AgeI fragment containing PML::RARA (obtained by restriction enzyme digestion of pMSCV-PML::RARA) was cloned into the pCMV-BioID2-HA vector (purchased from Addgene). To generate pMSCV-PML::RARA-BioID2-HA, an NheI/MssI fragment from the pCMV-PML::RARA-BioID2-HA plasmid was subcloned into the pMSCV-puro vector. To insert the DHFR fragment and construct pMSCV-DHFR-PML::RARA-BioID2-HA, a synthetic cDNA (GeneArt, Thermo Fisher Scientific) encompassing in-frame Escherichia coli DHFR (encoding the R12Y/G67S/Y100I destabilizing domain mutant) in frame with human PML sequence was digested with XhoI/AvrII to obtain a partial DHFR-fused PML fragment containing restriction enzyme site for vector backbone (XhoI) and PML replacement (AvrII, at 285 bp) to be cloned into pMSCV-PML::RARA-BioID2-HA. To generate NLS-BioID2 retroviral vector, a NcoI/BstBI fragment from a synthetic cDNA encoding the SV40 nuclear localization sequence (NLS) was cloned in place of the PML::RARA fragment of the pMSCV-PML::RARA-BioID2-HA plasmid, resulting in a pMSCV-NLS-BioID2-HA construct.

To obtain the pMSCV-RARB plasmid, an EcoRI/BamHI fragment from pSG-RARB (a kind gift from Albane le Maire, CNRS UMR504, INSERM 1054, Université de Montpellier, France) was subcloned into the MSCV-puro vector. For the pMSCV-Flag-HA-POZ-RARB plasmid, nested PCR was used to amplify a Flag-HA tag in frame with POZ domain from MSCV-PLZF-RARA with the following primers: Flag-HA-POZ For: 5′-GCT​GAA​TTC​GCC​ACC​ATG​GAC​TAC​AAG​GAC​GAC​GAT​GAC​AAG​CTC​GAT​GGA​GGA​TAC​CCC​TAC​GAC​GTG​CCC​GAC​TAC​GCC​GAT​CTG​ACA​AAA​ATG​GGC​ATG​ATC​CAG​CTG​CAG​AAC​CCT​AGC​CAC-3′; POZ rev: 5′-GGC​TCT​AGA​CCC​GCC​TCC​ACC​GAT​GGT​CTC​CAG​CAT​CTT​C-3′. The Flag-HA-POZ PCR product, including a glycine–serine linker, was then cloned in-frame with RARB into pMSCV-RARB.

To construct UBC9/RARA expression plasmids, UBC9 was PCR-amplified from the pSG5-Ubc9 plasmid and cloned in-frame with RARA in both pMSCV-RARA and pCMV-RARA vectors using the following primers: UBC9-MSCV-For 5′-TCT​CTC​GAG​GCC​ACC​ATG​TCG​GGG​ATC​GC-3′; UBC9-MSCV-Rev 5′-TTC​GTT​AAC​TCA​CGG​GGA​GTG​GGT​GGC-3′; CMV-UBC9-For 5′-CTG​GCT​AGC​CAC​CAT​GTC​GGG​GAT​CG-3′; CMV-UBC9-Rev 5′-ACC​GGA​TCC​CGG​GGA​GTG​GGT​GGC​C-3′.

RARA fusion point mutants were generated by using QuikChange II XL Site-Directed Mutagenesis Kit (cat# 200522; Agilent Technologies) with the following primers: PML::RARA^L73E-For 5′-GCC​CGA​AGC​TGC​TGC​CTT​GTG​AGC​ACA​CGC​TGT​GCT​CAG​GAT​GCC​TG-3′, PML::RARA^L73E-Rev 5′-CAG​GCA​TCC​TGA​GCA​CAG​CGT​GTG​CTC​ACA​AGG​CAG​CAG​CTT​CGG​GC-3′; PML::RARA^K160R-For 5′-AGG​CAC​ACC​AGT​GGT​TCC​TCC​GGC​ACG​AGG​CCC​GGC​CCC​TAG-3′, PML::RARA^K160R-Rev 5′-CTA​GGG​GCC​GGG​CCT​CGT​GCC​GGA​GGA​ACC​ACT​GGT​GTG​CCT-3′; RARA^I396E-For 5′-CCA​AGG​GGG​CTG​AGC​GGG​TGG​AAA​CGC​TGA​AGA​TGG​AGA​TCC​C-3′, RARA^I396E-Rev 5′-GGG​ATC​TCC​ATC​TTC​AGC​GTT​TCC​ACC​CGC​TCA​GCC​CCC​TTG​G-3′.

For methylcellulose colony formation assay, primary murine Lin⁻ progenitors were obtained from 5-fluorouracil–treated C57BL/6NJ mice (Janvier Labs) using a lineage-positive cell depletion kit (cat# 130-090-858; Miltenyi Biotec). Briefly, bone marrow flushed-out cells were magnetically labeled with a cocktail of biotinylated antibodies against a panel of lineage antigens (CD5, CD45R [B220], CD11B, anti-Gr-1 [Ly-6G/C], 7-4, and Ter-119 antibodies) and anti-biotin microbeads. Lin⁻ progenitors were retrovirally transduced with indicated fusions and maintained in methylcellulose (cat# 3231; StemCell Technology) supplemented with a cocktail of 10 ng/ml of mIL-3, mIL-6, mGM-CSF, and 50 ng/ml of mSCF as previously described (Zhu et al., 2005), and were selected and maintained with puromycin (1 μg/ml) or neomycin (500 μg/ml). To determine the impacts of fusion proteins and TAK-981 (HY-111789; MedChemExpress) on progenitors’ immortalization, 5,000 cells per well in a 6-cell plate were cultured and subsequently replated to assess their clonogenic capacity.

To purify PML::RARA interactants or His₁₀-SUMO1/2–conjugated proteins, HPC7, Lin⁻ progenitors, and primary MLL-ENL blasts were retrovirally transduced with viruses produced by transient transfections of Plat-E packaging cells (RARA, NLS-BioID, PML::RARA-BioID, and UBC9/RARA) and selected with puromycin (1 μg/ml). GFP-positive cells (transduced with MSCV-IRES-EGFP-His10-SUMO2) were sorted by FACS. In all experiments involving the BioID constructs, cells were treated with trimethoprim (5 μM) to induce NLS and RARA fusion expression. After 24 h, NLS and RARA fusion–expressing cells were treated with biotin (50 μM) for 16 h and lysed by buffer A containing 8 M urea, 50 mM Tris, pH 8, 150 mM NaCl, 10 mM NEM (cat# E3876; Sigma-Aldrich), 15 mM imidazole, (5 U/ml) universal nuclease (cat# 88702; Thermo Fisher Scientific), and proteinase inhibitors (cat# 11 836 170 001; Roche). Protein concentration was determined using Bradford reagent (cat# B6916; Sigma-Aldrich). Equal amounts of protein lysates containing 40 μM of DUB inhibitor PR-619 (cat# 662147; Sigma-Aldrich) were incubated overnight at 4°C with Ni-NTA agarose beads (cat# L30210; Qiagen) for SUMO1/2 conjugates or streptavidin Dynabeads (cat# 65602; Thermo Fisher Scientific) for BioID proteins. The beads were washed six times with buffer A and then analyzed by western blot.

For dual streptavidin plus Ni-NTA purification experiments, His₁₀-SUMO2–conjugated proteins were first purified as described above in urea buffer A, then eluted by radioimmunoprecipitation assay (RIPA) buffer containing 250 mM imidazole, 50 mM Tris (pH 8.0), 0.15 M NaCl, 1% NP-40, 1% sodium deoxycholate, 0.1% SDS, proteinase inhibitor, and 40 μM PR-619. Biotinylated and SUMO2ylated proteins were enriched by streptavidin Dynabeads overnight at 4°C. The beads were washed three times with RIPA buffer and then analyzed by western blot.

The following antibodies were used with 1:1,000 dilution for immunodetection in precipitation samples and cell lysates: ACTIN and GAPDH (A2066, G9545; Sigma-Aldrich); RARA and PML (homemade affinity-purified rabbit and mouse); GPS2 and NCoR (PA5-76547, A301-145A; Thermo Fisher Scientific); NRIP1 (NBP3-12264; Novus); RARA, NCOA, RXRA, TUBULIN, GATA2, SUMO1, and SUMO2 (ab27574, ab10491, ab125001, ab4074, ab109241, ab32058, ab81371; Abcam); KAP1 and His tag (4124, 2365; Cell Signaling); IRF2BP2 (18847-1-AP; Proteintech).

Total RNA extraction from immortalized progenitors, APL blasts, and HCP7 cells was performed with RNeasy Plus Mini Kit (cat# 74134; Qiagen) and quantified using a NanoDrop One/One (Thermo Fisher Scientific). When required, cDNA was prepared from 1 µg of total RNA with an iScript cDNA Synthesis kit (cat# 1708891; Bio-Rad). Quantitative real-time PCR on cDNA was performed on Bio-Rad CFX thermal cycles with TaqMan gene expression assays from Life Technologies (Gapdh, 4352339E; Cyp26a1, Mm00514486_m1; Pram1, Mm02744730_g1; Ncf1, Mm00447921_m1; Hck, Mm01241463_m1; Ifnγ, Mm01168134_m1). For RNA-sequencing analysis of WT and Trp53^−/− HPC7 cells, libraries were prepared using the Illumina Stranded mRNA Prep kit and paired-end sequenced at 151 bp on an Illumina NextSeq 2000 according to the manufacturer’s instructions. Sequencing reads were pseudoaligned to the mouse mm10 reference genome and used for gene set enrichment analysis (GSEA) using Molecular Signatures Database including H, C2, and C5 gene set collections.

Murine Lin⁻ cell types were analyzed using FACSCanto II (BD). Single-cell suspensions were blocked with Fc block (BD) for 15 min on ice. The following antibodies were used to stain immortalized Lin⁻ progenitor cells (1:1,000): APC-conjugated CD117 (Clone 2B8; eBioscience), APC-Cy7–conjugated anti-CD11b (Clone M1/70, BioLegend), APC-conjugated anti-FcεRI (Clone MAR-1; eBioscience), and phycoerythrin-conjugated anti-Gr1 (Clone RB6-8C5, eBioscience). Staining was performed overnight at 4°C. Cells were washed and resuspended in phosphate-buffered saline (PBS) with paraformaldehyde (PFA) 0.2% before FACS analysis.

Cytospined Lin⁻ progenitors were fixed with 4% PFA (Sigma-Aldrich) and permeabilized with 0.1% Triton X-100, and then blocked with PBS supplemented with 1% BSA and 0.05% Triton X-100. Cells were incubated with rabbit anti-human RARA antibodies (a kind gift from Scott Kogan) overnight at 4°C and then with the corresponding fluorescence secondary antibodies (Jackson ImmunoResearch or Thermo Fisher Scientific) together with 1 µg/ml of Hoechst 33342 (Thermo Fisher Scientific) for 1 h. Cells were then washed with PBS containing 0.02% Triton X-100 mounted with fluorescence mounting medium (Dako Omnis; cat# S302380) and examined with a LSM 980 confocal microscope equipped with a 63× oil objective lens (AiryScan 2 GaAsP detector; Carl Zeiss Micro-Imaging Inc.).

The His₁₀-HA-SUMO₃ knock-in strain was generated using the CRISPR/Cas9 system by the i-GONAD electroporation method with the BALB/cByJ strain. BALB/cByJ mice were purchased from Charles River. CRISPR RNA, trans-activating CRISPR RNA, single-stranded DNA, and Cas9 nuclease were purchased from IDT (CRISPR target sequence: 5′-TCA​GTT​TCC​ATT​TCT​TGA​T-3′; ssODN sequence: 5′-CAT​TTC​CCG​CCT​TCA​CAG​ACC​TAA​TAA​GAA​ATG​CAT​CAC​CAT​CAC​CAT​CAT​CAT​CAT​CAT​CAT​TAC​CCA​TAC​GAT​GTT​CCA​GAT​TAC​GCT​GAA​ACT​GAA​CCA​GTT​TCC​GTG​CAG​AAG​GTA​CCT​GCA​CCC-3′). To obtain His₁₀-HA-Sumo₃PML::RARA double mutant mice, this strain was crossed with Ctsg-PML::RARA mice on a BALB/cByJ background. For Fig. 3 E, APL blasts were collected from the bone marrow. Nonleukemic myeloid cells were enriched by depletion of lymphoid lineage using CD4/CD8, CD49b, and CD19 microbeads (cat# 130-116-480, 130-052-501, 130-121-301) and MACS column from Miltenyi Biotec.

Leukemic blasts in Fig. 5, E and F were derived from h-MRP8-PML::RARA transgenic mice and transplanted by intravenous injection into FVB mice. Serial intravenous transplantations were performed using 10⁴ GFP-sorted APL cells from bone marrow. After 2 wk, mice were intravenously treated with TAK-981 (15 mg/kg) for 3 days and bone marrow APL blasts were analyzed by May–Grünwald–Giemsa (MGG) staining and RT-qPCR. Animals were handled according to the guidelines of institutional animal care committees using protocols approved by the “Comité d’Ethique Experimentation Animal Paris-Nord N. 121” (project no. 58284). Mice were maintained in a 12-h light–dark cycle animal facility under specific pathogen-free conditions with free access to water and food (Institut de Recherche Saint Louis, Paris, France).

Experiments were performed at the Pitié-Salpêtrière (P3S) Proteomics Core Facility of Sorbonne University. Ni-NTA– or streptavidin-purified proteins were gel-purified and trypsin-digested overnight and processed essentially as described previously (Hospital et al., 2018). Mass spectrometry data acquisition was performed on a NanoElute II UHPLC system coupled to a TIMS-TOF-HT mass spectrometer (Bruker Daltonics), with a nano-electrospray ion source (Bruker CaptiveSpray source). Peptides were injected into a reversed-phase C18 nanocolumn (Aurora 3, 25 cm long, 75 μm internal diameter, 1.7 μm, 120 Å pores, IonOpticks). The mobile phases used were A (0.1% (vol/vol) formic acid in Milli-Q grade water) and B (0.1% (vol/vol) formic acid in acetonitrile). The separation was carried out at 220 nl/min in a 30-min acetonitrile gradient (5–16%/22–25% B). All samples were acquired in a DIA-PASEF acquisition mode, with an m/z range 100–1,700 and an ion mobility (IM) range of 0.7–1.3 1/K₀. Isolation windows were fixed at 35 Da, with a cycle time estimate of 1.17 s. Mass and IM calibration were done linearly using three ions from the Agilent ESI LC/MS tuning mix (m/z 622.0290, IM 0.9917 1/K₀; m/z 922.0098, IM 1.1984 1/K₀; m/z 1,221.9906, IM 1.3934 1/K₀). Mass spectrometry data were analyzed by software Data-Independent Acquisition by Neural Networks, version 1.8.1. Data filtering, statistical, and differential analyses were all carried out with PROSTAR software, version 1.30.7. Data were refined by excluding missing lines and contaminants from the dataset, followed by log₂ transformation of LFQ intensities. Normalization was carried out by the method of quantile centering, within each condition. Significant differences were evaluated by the Limma test, using an adaptive Benjamini–Hochberg correction. Proteins were deemed significant if P values were consistent with a false discovery rate (FDR) cutoff of 5%. GSEA was performed using Gene Ontology (C5.all) containing 16,107 gene sets.

Dot plots shown in Fig. 2 A and Fig. 3 B were generated by ProHits-viz. GraphPad Prism software was used to perform two-way ANOVA and an unpaired two-tailed t test and to determine the P value and look for significant changes in the data generated from the fusion-expressing cells and TAK-981 treatment in vivo or ex vivo. All data are expressed as the mean ± SD. For all graphs, *P = 0.01–0.05, **P = 0.001–0.01, and ***P < 0.001.

Table S1 shows a list of 379 significant PML::RARA-associated proteins identified in HPC7 cell lines. Table S2 shows a list of plasmid constructs used in the study. Fig. S1 shows data on PML::RARA-interacting proteins and a detail comparison of PML::RARA interactants identified in HPC7 BioID and Lin⁻ progenitor TurboID. Fig. S2 shows data on exploration of UBC9/RARA and its I396E mutant and impact of TAK-981 in cellular localization of RARA fusion proteins. Fig. S3 shows data on TAK-981 response in primary progenitors.

The mass spectrometry proteomics data (Fig 2 and Fig 3) have been deposited to the ProteomeXchange consortium via the PRIDE partner repository (Perez-Riverol et al., 2025) with the dataset identifier PXD078941. All data are available upon reasonable request from the corresponding authors.

We warmly thank T. Ley (Washington University in St. Louis, St. Louis, MO, USA) for his generous gift of CatG-PML::RARA^+/+ C57BL/6 mice (Westervelt et al., 2003), Camille Lobry (INSERM, Paris, France) for HPC7 cells, and S. Kogan (University of California, San Francisco, San Francisco, CA, USA) for anti-hRARA antibody. We warmly thank T. Wandless (Stanford Medicine) for insightful discussions on the selection of destabilizing domains and the use of trimethoprim, the P3S Proteomics Core Facility of Sorbonne University, and the Orion technological core (IMACHE-IBiSA) of Center for Interdisciplinary Research in Biology, as well as the technical platforms of Institut de Recherche Saint-Louis for expert analyses and advice.

Work in the laboratory is supported by the European Research Council (PML-Therapy #785917, RARA-AML #101140666), as well as Institut National du Cancer (PLBIO #2022-105) and Equipe labellisée par la Ligue Nationale contre le Cancer. The animal facility was supported in part by the French National Research Agency, through the France 2030 program, ANR-23-IAHU-0005, Paris St. Louis Leukemia Institute.

Author contributions: Hsin Chieh Wu: conceptualization, data curation, formal analysis, investigation, methodology, project administration, supervision, validation, visualization, and writing—original draft, review, and editing. Emmanuel Laplantine: data curation, formal analysis, investigation, methodology, project administration, validation, visualization, and writing—review and editing. Cécile Esnault: data curation and investigation. Michiko Niwa-Kawakita: investigation, methodology, resources, and writing—review and editing. Yi Zhang: formal analysis. Viviane De Almeida Bastos: data curation, formal analysis, and investigation. Marie-Claude Geoffroy: conceptualization, data curation, formal analysis, investigation, methodology, resources, supervision, validation, visualization, and writing—original draft, review, and editing. Hugues de Thé: conceptualization, formal analysis, funding acquisition, investigation, project administration, resources, supervision, validation, and writing—original draft, review, and editing.