TRIB3 Stabilizes High TWIST1 Expression to Promote Rapid APL Progression and ATRA Resistance

Acute promyelocytic leukemia (APL), which accounts for 10%–15% of acute myeloid leukemia (AML) cases, is characterized by the t(15; 17) chromosomal translocation and is now highly curable by the combination of granulocytic differentiation induction and the PML-RARα oncoprotein–targeted agents all-trans retinoic acid (ATRA) and arsenic trioxide (ATO; refs. 1, 2). Despite the striking molecular complete remission (CR) and the very few cases of relapse associated with ATRA/ATO–based regimens, mortality events typically result from early fatal bleeding, which remains the most important challenge and the largest obstacle to curing all patients with APL (3). For instance, several studies have reported that the risk of early hemorrhagic death (HD) reaches an incidence of 10%–20% during the first month of induction (4–6). Importantly, patients with APL with a high white blood cell count face an increased risk of early HD (7). Furthermore, resistance to ATRA/ATO treatment with PML-RARα mutations still remains a therapeutic challenge for a significant proportion of patients with APL. Thus, we need to better understand the molecular mechanism of APL pathogenesis and design more effective therapeutic strategies to block the rapid progression of early death and overcome resistance.

Oncogenic transcription factors play an important role in the development of hematologic malignancy (8, 9). The dysregulation of epithelial–mesenchymal transition-inducing transcription factors (EMT-TF), including SNAI1/SNAI2, ZEB1/ZEB2, and TWIST1/TWIST2, has also been explored within the context of the aggressive invasion, chemoresistance, and poor prognosis of AML (10–12). TWIST1, a highly conserved basic helix-loop-helix (bHLH) protein, is a well-characterized EMT-TF that plays a critical role in embryonic development and cancer metastasis (13, 14). Recent studies have shown that TWIST1 is overexpressed in primary AML samples (15). Our previous study reported that high expression of TWIST1 in AML contributes to extramedullary infiltration and promotes leukemic aggressiveness (16). These findings strongly suggest that disruption of TWIST1 is involved in leukemogenesis. However, the prognosis of patients with AML with high TWIST1 expression is controversial based on several reports (15, 17). These conflicting reports likely occurred because TWIST1 expression in AML is heterogeneous (18). TWIST1 expression is higher in APL, an intriguing AML subtype in which successful clinical tumor differentiation-induction therapy has been applied and for which the clinical phenotype is inconsistent with the expression level of TWIST1 (15, 18).

In this study, we demonstrate that the EMT-TF TWIST1 is significantly highly expressed in APL and governs the survival of APL cells. We show that TRIB3 interacts with TWIST1 and stabilizes high TWIST1 expression by repressing TWIST1 ubiquitination. Our data also suggest that a peptide similar to the WR domain disturbs the TRIB3/TWIST1 interaction, impairs rapid progression during the early death of APL, and reverses resistance to ATRA therapy. These results contribute to a better understanding of the molecular mechanism of APL pathogenesis and will allow for designing more effective therapeutic strategies to block the rapid progression of early death and overcome resistance.

Leukemic cell lines were cultured in RPMI1640 medium (Invitrogen) supplemented with 10% FBS (Invitrogen), and HEK293T cells were grown in DMEM (Invitrogen) supplemented with 10% FBS. All cell lines, including NB4-R1 and PR9 cell line, were obtained from the Shanghai Institute of Hematology (Shanghai, China) as described previously (19–21). NB4-R1, a de novo ATRA-resistant cell line isolated from parental NB4 cells, was obtained from Dr. Michel Lanotte (Hospital Saint Louis, Paris, France; ref. 22). HMRP8-PML-RARα transgenic mice were generated on an FVB/NJ background using the human MRP8 expression cassette (23). The reproducible APL transplantation model can be propagated by injecting APL blasts, isolated from hMRP8-PML-RARα transgenic mice, into the tail vein of syngeneic recipients (24). All transplanted APL cells were highly purified by flow-sorting to exclude the dying cells, and then injected into the immunodeficient mice or syngeneic recipients. NB4-luc or R1-luc cells were injected into sublethally irradiated 8-week-old NOD/SCID mice through tail veins. All animal experiments were conducted in accordance with the approved guidelines provided by the Laboratory Animal Resource Center of Shanghai Jiao-Tong University School of Medicine (Shanghai, China).

Primary AML samples were obtained from the bone marrow of patients with diagnosed AML. Leukemic blasts were purified and harvested in the mononuclear layer via density gradient centrifugation. Human primary AML samples were mainly obtained from Shanghai Rui-Jin Hospital (Shanghai, China) with written informed consent from each patient and research ethics board approval in accordance with the Declaration of Helsinki.

Cells were suspended in FACS buffer containing 1% FBS and 0.1% NaN₃. The data were collected on an LSR-Fortessa X20 Flow Cytometer (BD Biosciences). Antibodies were purchased from BD Biosciences, including anti-CD11b and Annexin V/PI apoptosis flow kit.

Murine PB or bone marrow (BM) cells were cytospun onto slides and stained with Wright-Giemsa staining solution by following manufacturer's manual. The samples were evaluated under a light microscope (BX61, Olympus).

Senescence-associated b-galactosidase (SA-β-Gal) staining was detected with the Senescence Detection Kit (Abcam). The senescent cells were quantitated from at least six random fields according to the manufacturers' protocols.

To evaluate methylcellulose colony-forming unit (CFU) colony numbers in human or mouse leukemic cells, highly purified sorted cells were plated in duplicate and cultured in MethoCult medium (Stem Cell Technologies) in 12-plate dishes. On day 11, CFUs were counted from three independent experiments using the manufacturers' protocols.

The TWIST1 or TRIB3 shRNA sequence was converted from a pair of previously reported shRNA oligos (16, 25). To generate cells stably expressing TWIST1-shRNA, TRIB3-shRNA, and the negative control-shRNA (NC), the expression cassettes were transduced into leukemic cells with lentiviral vectors. siRNA oligonucleotides against TWIST1 were transfected using Lipofectamine 2000 (Invitrogen).

Total cellular RNA was extracted with RNeasy micro kit or RNeasy Mini Kit (Qiagen) by following the manufacturer's manual. For qRT-PCR, reactions were performed by using SYBR Premix Ex Taq (Applied Takara Bio Inc.) on an ABI 7500 Real-Time PCR system. The primer sequences of reference gene GAPDH were described previously (26).

The resulting cell pellets were lysed with the whole-cell lysis buffer in the boiling water for at least 10 minutes. The antibodies used included anti–β-Actin (Sigma), anti-PML-RARα (Abcam), anti-TRIB3 (Proteintech), and anti-TWIST1 (Santa Cruz Biotechnology). Protein signals were visualized using the Immobilon Western Kit (Millipore). Murine brain sections and spinal cord slices were prepared for HE staining or human CD45 (Cell Signaling Technology) IHC staining. Cells or frozen tissue samples were fixed in 4% PFA and then permeabilized in 0.2% Triton X-100. Murine brain sections and spinal cord slices were prepared for murine c-Kit (Santa Cruz Biotechnology) IHC staining. Optical sections of the cells were observed under a Leica TCS SP8 Confocal Microscope (Leica Microsystems).

Twenty days after inoculation of APL cells into syngeneic mice or NOD/SCID mice, peptidomimetics (10 mg/kg, twice) or ATRA (10 mg/kg, once) and ATO (10 mg/kg, once; Sigma-Aldrich) treatment was started by daily intraperitoneal injection for 6 to 10 successive days.

NB4 cells or APL murine blasts carrying a luciferase reporter were transplanted into mice. The luciferase substrate was injected into living animals before imaging. Then bioluminescence imaging (BLI) was performed by according to the manufacturers' protocols (Xenogen IVIS Spectrum, PerkinElmer).

Cell extracts were prepared with lysis buffer (50 mmol/L pH 7.5 Tris, 150 mmol/L NaCl, 0.5% Triton X-100, 10% glycerol, 2 mmol/L EDTA, 1 mmol/L PMSF, 20 mmol/L, Protease Inhibitor Cocktail, Phosphatase Inhibitor Cocktail, and 2 mmol/L DTT). Supernatants were then incubated with preconjugated anti-FLAG M2 (Sigma), anti-MYC (Biotool), anti-HA (Biotool), anti-GFP (MBL), or anti-IgG (CST) beads at 4°C overnight. The beads were sequentially washed 5 times with co-IP lysis buffer. The bound proteins were eluted with 2% SDS lysis buffer and boiled at 100°C for 15 minutes. The proteins were analyzed by Western blotting according to the standard protocol.

For the in vivo ubiquitination assays, 293T cells were transiently transfected with plasmids for HA-tagged ubiquitin, Flag-PML-RARα, GFP-TWIST1, Myc-TRIB3, and other indicated constructs. Twenty-four hours after transfection, cells were treated with 10 μmol/L MG-132 or DMSO for 24 hours, and then washed with PBS and collected by centrifugation. Cells were lysed in co-IP lysis buffer. The lysate was subjected to co-IP or IP using anti-HA–conjugated agarose beads for overnight at 4°C. The samples were loaded, separated by SDS-PAGE, and immunoblotted with the indicated antibodies as described above.

Data are presented as arithmetic means ± SEM. Kaplan–Meier survival analysis, Student t tests, or χ² tests were used to calculate P values where appropriate. P < 0.05 was considered to be significant.

To evaluate the expression levels of EMT-TFs in AML, we analyzed microarray data for different cytogenetic codes from 199 AML samples from the E-MTAB-3444 database (Supplementary Fig. S1A). Detailed quantitative assessment showed that TWIST1 expression, but not TWIST2, SNAI1/2, and ZEB1/2 expression, was significantly higher in t (15;17) APL samples than in other cytogenetic types of AML, including t (8;21), inv (16), MLL-(11q23), t (9;11), and inv (3)/t (3;3; Fig. 1A; Supplementary Fig. S1B). We also examined EMT-TF mRNA expression using the RNA-seq database of The Cancer Genome Atlas (TCGA) and verified the higher quantity of TWIST1 mRNA in the M3 subtype of AML according to FAB classification (Fig. 1B; Supplementary Fig. S1C). To further confirm the aberrant high expression of TWIST1 in APL, we quantified the mRNA and protein levels of APL blasts from patients with primary AML (Fig. 1C and D; Supplementary Fig. S1D; Supplementary Table S1). Indeed, higher levels of the TWIST1 mRNA and protein were found in APL samples than in non-APL samples. Consistent with the data from human leukemic cells, higher TWIST1 expression was found in leukemic cells from PML-RARα transgenic mice than in blasts from other AML mouse models (Fig. 1E). In a series of AML cell lines, the typical APL cell line NB4 presented higher TWIST1 expression than the other cell lines (Supplementary Fig. S1E). To assess whether TWIST1 expression was correlated with the expression of the PML-RARα oncoprotein, we performed ZnSO₄-induced PML-RARα expression in PR9 cells. As expected, ZnSO₄-induced PML-RARα expression upregulated the expression of TWIST1 in a time-dependent manner; although, TWIST1 and PML-RARα were not completely colocalized (Fig. 1F). Consistent with this finding, TWIST1 was also correlated with PML-RARα expression in NB4 cells, with partial colocalization (Supplementary Fig. S1F). Thus, these results demonstrate that TWIST1, as an EMT-TF, is highly expressed in APL cells.

To assess whether high TWIST1 expression plays an important role in APL progression, we introduced negative control (NC)-shRNA (shRNA) or sh-TWIST1-1/2 into two types of APL cells (NB4 cells and APL murine blasts) and found that sh-TWIST1-1 caused significant suppression of TWIST1 expression in both types of cells (Supplementary Fig. S2A). We observed that sh-TWIST1-1 decreased PML-RARα expression and induced NB4 cell apoptosis and differentiation but not senescence (Fig. 2A and B; Supplementary Fig. S2B). The efficacy and specificity of TWIST1 shRNAs were confirmed by rescue via TWIST1 overexpression (Supplementary Fig. S2C). Consistent with the findings obtained in the NB4 cell line, sh-TWIST1-1 expression induced apoptosis and differentiation in leukemic blasts of APL transgenic murine, which were generated on an FVB/NJ background using the human MRP8 expression cassette (Supplementary Fig. S2D–S2F). To evaluate the role of TWIST1 in APL progression in vivo, we inoculated TWIST1-knockdown NB4 cells into sublethally irradiated NOD/SCID mice by tail vein injection. At 3 weeks after transplantation, TWIST1 knockdown significantly suppressed APL cell invasion and prolonged the survival of recipient mice (Fig. 2C and D). Similarly, APL transgenic murine blasts expressing sh-TWIST1-1 generated fewer leukemic cells and had a much longer survival rate than the NC mice (Supplementary Fig. S2G–S2I). Interestingly, in contrast to the NC-shRNA groups, in which the mice developed a spinning in a circle syndrome or even paralysis, we did not observe significant central nervous system (CNS) infiltration of APL cells in the TWIST1-knockdown groups, indicating that TWIST1 as an EMT-TF may contribute to APL extramedullary infiltration (Fig. 2C and E; Supplementary Fig. S2G and S2J). As reported previously, the serious situation associated with APL relapse and poor prognosis largely predominated in the CNS infiltration (27–29), we showed that downregulation of TWIST1 in APL cells prevented CNS invasion in mice, although we also observed skin and eye involvement in some NC-shRNA cases of our APL mouse models (Supplementary Fig. S2K). We also examined the effect of TWIST1 depletion on blasts from patients with primay APL and confirmed the importance of TWIST1 in maintaining APL cell survival (Fig. 2F and G; Supplementary Fig. S2L). Collectively, these data suggest that TWIST1 plays an important role in APL cell survival and is critical for APL disease progression.

As mentioned previously, TWIST1 correlated with the expression of the PML-RARα oncoprotein but was not completely colocalized with PML-RARα. We then investigated whether there was evidence linking the molecular mechanisms of PML-RARα–driven APL to an indirect influence of TWIST1. Notably, the activity of TWIST1 as an EMT-TF seemed to be maintained through a TRIB3/p62–dependent interaction (30, 31). The Tribbles proteins (TRIB1, TRIB2, and TRIB3) have been shown to play critical roles in leukemogenesis via different mechanisms (32–38). TRIB3 has been reported to suppress the degradation of PML-RARα through interacting with PML-RARα and PML (39). Given the importance of TWIST1 and TRIB3 in APL pathogenesis, we assessed whether TWIST1 and TRIB3 were highly coexpressed in APL cells compared with other subtypes of AML. We analyzed the profiling data from the E-MTAB-3444 and TCGA databases and verified high TRIB3 expression in previously studied APL samples (Supplementary Fig. S3A and S3B). As expected, TWIST1 expression was significantly correlated with the expression of TRIB3 in APL patient samples (Fig. 3A). Consistent with this finding, TWIST1 and TRIB3 were also highly coexpressed in NB4 cells or ZnSO₄-induced PR9 cells (Fig. 3B; Supplementary Fig. S3C). As high expression of TWIST1 and TRIB3 in APL cells, we suspected that TWIST1 might interact with TRIB3 to regulate PML-RARα function. We found that TWIST1 coimmunoprecipitated with TRIB3 in NB4 cells and APL murine blasts (Fig. 3C; Supplementary Fig. S3D), and this finding was confirmed in HEK293T cells by coexpressing various tagged plasmids or performing glutathione S-transferase (GST) pull-down (Fig. 3D and E). In addition, using a coimmunostaining assay, we observed that TWIST1 was strongly colocalized with TRIB3 (Fig. 3F and G; Supplementary Fig. S3E). Taking into account the previously confirmed interaction of TRIB3 and PML-RARα (39), we further investigated the relationship among these three proteins. We showed the colocalization of TWIST1, TRIB3, and PML-RARα in both NB4 and plasmid-coexpressing HEK293T cells (Supplementary Fig. S3F). According to Flag-PML-RARα immunoprecipitation of APL murine blasts and plasmid-coexpressing HEK293T cells, TRIB3, but not TWIST1, bound directly to PML-RARα (Supplementary Fig. S3G and S3H). These results suggest that TWIST1 interacts with TRIB3 to indirectly modulate PML-RARα.

Human TWIST1 is an 18-kDa protein that contains 202 amino acids, which is much smaller than the PML-RARα fusion oncoprotein. To determine whether TWIST1 protein has a short half-life compared with PML-RARα oncoprotein, we performed the cycloheximide (CHX) chase assay with APL cells. Using the translational inhibitor CHX, we observed that APL cells accumulated endogenous TWIST1 and TRIB3 but not the corresponding mRNA after treatment with the proteasome inhibitor MG-132. The CHX chase assay indicated that TWIST1 was degraded more rapidly than PML-RARα (Fig. 3H; Supplementary Fig. S3I and S3J). In vivo ubiquitination assays also showed increased levels of ubiquitinated TWIST1 after MG132 treatment (Supplementary Fig. S3K and S3L). To check the possibility that TWIST1 is a direct transcriptional target of PML-RARα, we screened the identified putative PML-RARα oncoprotein–binding sites provided by Wang and colleagues (20), and we did not find the binding characteristic of TWIST1 as a direct transcription target of PML-RARα. These data demonstrate that APL cells degrade endogenous TWIST1 in a proteasome-dependent manner. Interestingly, we observed the half-life of TWIST1 is longer than half-life of TRIB3 (Fig. 3H). Considering that TWIST1 interacted with TRIB3 in APL cells, we hypothesized that TRIB3 might inhibit ubiquitination and degradation of TWIST1 in a TRIB3-dependent manner. TRIB3 knockdown decreased the protein level of endogenous TWIST1 and shortened the protein half-life in NB4 cells, which was reversed by MG-132 treatment (Fig. 3I; Supplementary Fig. S3M and S3N). In vivo ubiquitination assays also showed increased levels of ubiquitinated TWIST1 after TRIB3 knockdown (Fig. 3J). Together, these data indicate that TRIB3 stabilizes high TWIST1 expression in APL cells through preventing ubiquitination.

Although APL has been highly curable with ATRA-based differentiation therapy, a fraction of patients still relapse and become resistant to ATRA (40, 41). We found that ATRA treatment decreased the protein levels of TWIST1 and TRIB3 in ATRA-sensitive NB4 cells but not in ATRA-resistant NB4-R1 (R1) cells (Fig. 4A). This finding led us to consider whether TRIB3 cooperated with TWIST1 to contribute to ATRA resistance. To test this hypothesis, we introduced NC-shRNA or sh-TRIB3-1/2 into R1 cells and observed significant suppression of TRIB3 expression with sh-TRIB3-2 (Supplementary Fig. S4A). Furthermore, sh-TRIB3-2 slightly decreased TWIST1 protein levels and promoted R1 cell differentiation but not cell apoptosis (Supplementary Fig. S4B–S4D). Silencing of TRIB3 also reduced TWIST1 expression in response to ATRA and greatly reversed ATRA resistance in R1 cells (Fig. 4B and C; Supplementary Fig. S4E). In vivo ubiquitination assays revealed increased levels of ubiquitinated TWIST1 after TRIB3 knockdown in response to ATRA treatment (Fig. 4D). Moreover, TWIST1 overexpression rescued TWIST1 protein levels and impaired the differentiation induced by TRIB3 knockdown (Fig. 4E and F; Supplementary Fig. S4F–S4J). To evaluate the role of TRIB3/TWIST1 in APL ATRA resistance in vivo, we inoculated TRIB3-knockdown and/or TWIST1-overexpressing R1 cells into sublethally irradiated NOD/SCID mice by tail vein injection. At 25 days after transplantation with 6-day ATRA treatment, TRIB3 knockdown significantly reversed ATRA resistance in R1 cells and prolonged the survival of recipient mice (Fig. 4G and H). Similarly, TWIST1 overexpression rescued the suppression of R1 cells and shortened the survival of recipient mice caused by TRIB3 knockdown in vivo (Fig. 4G and H). Interestingly, compared with the NC-shRNA groups, reduced CNS infiltration of R1 cells was observed in the TRIB3-knockdown groups, and modest CNS infiltration of R1 cells was observed in the TWIST1-overexpressing group, indicating that TWIST1 may contribute to APL extramedullary infiltration (Fig. 4I). These results show that loss of TRIB3 promotes TWIST1 degradation and reverses resistance to ATRA therapy.

Because the TRIB3/TWIST1 interaction is involved in the protein stability and functional maintenance of the TWIST1 protein, we further analyzed the binding interface residues in detail. We constructed mutants of TWIST1 and TRIB3 deleted for various domains (Supplementary Fig. S5A and S5B). We found that both the bHLH domain and the WR domain of TWIST1 interacted with the C terminus of TRIB3, although the bHLH domain only weakly contributes to this interaction (Fig. 5A). We have thus unveiled the critical binding domains of TWIST1 with TRIB3. As observed before, the binding of TRIB3 and TWIST1 stabilized high TWIST1 expression in APL cells (Fig. 3J; Supplementary Fig. S3K and S3L). Considering that TRIB3 interacts mostly with the WR domain of TWIST1, we hypothesized that the TRIB3-C terminus: TWIST1-WR domain interaction might stabilize high TWIST1 expression. As expected, HEK293T IP assays showed that TRIB3 was able to inhibit the ubiquitination of TWIST1, and this inhibition mainly occurred between the C-terminus of TRIB3 and the WR domain of TWIST1 (Fig. 5B; Supplementary Fig. S5C and S5D).

To directly examine the functional effect of interface residues critical for binding between TRIB3 and TWIST1, we generated four synthetic peptides containing the interface residues and various lysine sites from the TWIST1 C terminus (Fig. 5C, top). We fused these peptides to a cationic cell–penetrating peptide (42). The peptides were designed in the orientation containing K residues or the WR domain and were termed peptides 1, 2, and 3. We also designed an inactive peptide (termed peptide con) as a negative control. To determine the ability of peptidomimetics to dissociate the TRIB3/TWIST1 complex in APL cells, we purified the TRIB3/TWIST1 complex by IP using anti-TWIST1 antibodies in the presence of 1 μmol/L peptidomimetics and determined its composition by Western blotting analysis. Compared with peptide con treatment, competition with peptide 3 led to significant dissociation of the cellular TRIB3/TWIST1 complex (Fig. 5C, bottom). We then treated NB4 and R1 cells with peptidomimetics and observed that peptide 3, but not the peptide con, peptide 1, or peptide 2, induced significant differentiation and apoptosis in APL cells rather than other leukemic cells or normal hematopoietic cells (Fig. 5D, Supplementary Fig. S5E and S5F). Consistent with the significant changes in cell surface CD11b and Annexin V, we observed a significant decrease in the protein abundance of TRIB3/TWIST1 and a robust increase in intracellular caspase-3 and PARP cleavage after peptide 3 treatment for only 8 hours (Supplementary Fig. S5G). Moreover, in vivo ubiquitination assays showed increased levels of ubiquitinated TWIST1 with peptide 3 treatment (Supplementary Fig. S5H). Thus, peptide 3 is a specific peptide mimetic inhibitor of TRIB3/TWIST1 complex assembly in APL cells and induces apoptosis and differentiation of APL cells in vitro.

Despite the striking long-term leukemia-free survival rate after the ATRA/ATO–based regimen, the progression of APL, including early death (ED) and differentiation therapy resistance, still affects the health of a significant proportion of patients with APL (3, 7). To mimic the progressive clinical pattern of APL ED, we began to treat APL mouse models with ATRA/ATO or peptidomimetics at 20 days posttransplantation (Supplementary Fig. S6A). We noted that APL mice had very high peripheral blasts at this stage and died within 5 days if not treated, which was very similar to the clinical pattern of APL ED (Supplementary Fig. S6B). We transplanted NB4 cells into NOD/SCID mice, and at 20 days posttransplantation, we treated the engrafted mice with ATRA/ATO, peptide 3, or control peptide for 30 days. Peptide 3 significantly delayed leukemia progression, extended survival, and reduced CNS infiltration of APL cells (Fig. 6A–C; Supplementary Fig. S6C). Consistent with the results obtained in APL cell–engrafted NOD/SCID mice, APL transgenic mice treated with peptide 3 exhibited significantly delayed disease latency and death compared with the mice in the control or ATRA/ATO group (Fig. 6D and E). We also noted that peptide 3, the WR domain peptide mimetic, inhibited APL cell proliferation and impeded CNS infiltration in vivo (Supplementary Fig. S6D and S6E). We then examined whether peptide 3 interfered with TWIST1/TRIB3 interaction to reverse resistance to ATRA treatment. As expected, the use of peptide 3 in vitro greatly impaired ATRA resistance in R1 cells (Supplementary Fig. S6F). To evaluate the role of peptidomimetics in APL ATRA resistance in vivo, we inoculated R1 cells into sublethally irradiated NOD/SCID mice by tail vein injection. At 20 days after transplantation, the mice were treated with ATRA and peptidomimetics, and then monitored for response to ATRA treatment. Consistent with the use of peptide 3 in vitro, the WR domain peptidomimetic significantly impeded leukemia growth and sensitivity to ATRA and prolonged the survival of R1 cell recipient mice (Fig. 6F and G). We further confirmed the therapeutic effect of peptide 3 in preventing rapid progression of primary APL ED patient blasts in vitro. We screened 5 samples of ED from the collected APL bone marrow specimens and found that they all involved cases of high WBC and peripheral blast counts (Supplementary Fig. S6G). We treated leukemia cells with peptide 3 in vitro and confirmed that it can promote leukemia cell differentiation and apoptosis in a short time compared with ATRA/ATO or peptide con (Fig. 6H; Supplementary Fig. S6H). Therefore, the WR domain peptidomimetic can rapidly promote APL cell differentiation and apoptosis, and can prevent early death of APL and reverse induction therapy resistance (Fig. 6I).

Although the importance of EMT-TFs, including the TWIST, SNAIL, and ZEB families, has been well documented in the tumorigenesis of epithelial cancers, the role of EMT-TFs in hematologic malignancies is still unknown. In this study, we found that EMT-TF TWIST1 is highly expressed in patients with APL and is critical for leukemic cell survival. In light of a recent report indicating that the stress protein TRIB3 inhibits the degradation of PML-RARα and promotes APL progression, we found that TWIST1 and TRIB3 were highly coexpressed in APL cells compared with other subtypes of AML. We subsequently demonstrated that TRIB3 could strongly bind to the WR domains of TWIST1 to stabilize TWIST1 by inhibiting its ubiquitination.

Currently, APL is highly curative. With differentiation therapy, over 80% of patients with APL achieve long-term leukemia-free survival. However, a proportion of patients with APL suffer from early fatal bleeding and leukemic extramedullary infiltration, and the underlying mechanisms of this difference are largely unknown (27–29). On the basis of a detailed analysis and a functional screening of synthetic peptides, we discovered a peptide analogous to the TWIST1 WR domain that rapidly and specifically represses APL cell survival by disrupting the TRIB3/TWIST1 interaction. Targeting rapid TWIST1 degradation could protect against early death in APL and improved sensitivity to ATRA. Furthermore, our data also showed that TWIST1 governed CNS infiltration during progression in NB4-xenograft mice and APL transplantable mice. Cell-penetrating peptides may competitively inhibit the TWIST1/TRIB3 interaction and repress CNS progression by initiating APL cell differentiation and apoptosis. Our previous study also reported that high expression of TWIST1 in AML contributes to extramedullary infiltration and promotes leukemic aggressiveness (16). A pioneering study using the MLL-AF9 mouse model revealed that the EMT-inducer ZEB1 contributed to leukemic blast invasion and was associated with poor survival in patients with AML. These observations suggest that EMT-TFs not only play important roles in solid tumors but also promote leukemic progression and aggressive extramedullary infiltration. Thus, EMT-TFs may be novel therapeutic targets for disease progression in patients with relapsed and refractory AML.

TWIST1 contains two highly conserved and functionally different domains: the bHLH domain for DNA binding and the WR domain for heterodimer formation (43). Recently, the WR domain was also reported to exhibit transactivation activity and to interact with RUNX2, SOX9, p53, RELA, and p62 (31, 44–47). Notably, through a p62-dependent interaction, the WR domain is necessary for the proteolytic activity of TWIST1 (31). TRIB3 inhibited autophagic substrate clearance by interacting with p62 and led to UPS-dependent accumulation of EMT-TFs, such as TWIST1, ultimately promoting tumor growth and metastasis (30). These results strongly implicated that TRIB3 was highly implicated in TWIST1 degradation. Here, we provided clear evidence that the binding of TRIB3 to the WR domain inhibited TWIST1 ubiquitination and stabilized high TWIST1 expression to promote APL progression and ATRA resistance. Intriguingly, despite the absence of lysine sites in the WR domain, the TWIST1 WR domain was required for TWIST1 ubiquitylation, which appears to be caused by recruitment of potential ubiquitin E3 ligases.

APL is commonly driven by the t (15;17) chromosomal translocation, which yields the PML–RARα fusion oncoprotein as a transcriptional repressor. ATRA- and/or ATO-triggered PML-RARα proteolysis are required for the elimination of APL cells. Recent studies have shown that PML-RARα has a half-life of over 8 hours and requires approximately 12 hours to be partially degraded in response to ATRA and/or ATO (39, 48, 49). If this is the case, a significant number of patients with APL who experience rapid progression and a high-risk state, including high white blood cell (WBC) counts, fatal bleeding, and severe infection, would not rapidly repress APL cell survival via ATRA- and/or ATO-initiated PML-RARα degradation. Similarly, the proposal that ATRA- and/or ATO-triggered PML-RARα degradation in vitro and in vivo exerts any effect has been controversial. Therefore, we hypothesized that PML-RARα induces a gain-of-function transcriptional activation to upregulate modulators of APL pathogenesis. We provided clear evidence that high TWIST1 expression promotes APL progression and that TWIST1 proteolysis initiated by a novel peptide might help reshape the therapeutic design for rapid APL eradication. Although a previous study indicated that TRIB3 suppresses the degradation of PML-RARα by interacting with PML-RARα and PML, it is difficult to determine how TRIB3 instability might protect the half-life of PML-RARα. Our results strongly suggest that the binding of TRIB3 to TWIST1 inhibits TWIST1 ubiquitination and stabilizes high TWIST1 expression to promote APL progression and ATRA resistance. This characteristic is a very effective target for preventing APL early death and ATRA resistance.

No potential conflicts of interest were disclosed.

Conception and design: W. Zhang, J. Xu

Development of methodology: J. Lin, W. Zhang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. Lin, W. Zhang, L.-T. Niu

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. Lin, W. Zhang, L.-T. Niu

Writing, review, and/or revision of the manuscript: J. Lin, W. Zhang, J. Zhu, J. Xu

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): W. Zhang, Y.-M. Zhu, X.-Q. Weng, Y. Sheng, J. Zhu

Study supervision: J. Xu

We thank Dr. Chun-Lin Shen for her assistance with NOD/SCID in vivo bioluminescence imaging. This study was founded by the National Natural Science Foundation of China (81800099, 81400106, 81430002, 81770206), the Shanghai Rising-Star Program (17QA1402200, 19QA1407800), the Shanghai Excellent Youth Medical Talents Training Program (2018YQ09), and National Science and Technology Major Project (2018ZX09101001). W. Zhang gratefully acknowledges to be supported by Global Scholar-in-Training Award (GSITA) from AACR.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.