Abstract

   Acute myeloid leukemia (AML) is a blood cancer characterized by
   uncontrolled growth of myeloid cells. Overcoming AML treatment
   resistance, particularly to anthracycline-based drugs like doxorubicin
   (ADR), poses a challenge. This study investigated the role of CELF1, an
   RNA-binding protein, in ADR resistance and autophagy regulation in AML.
   CELF1 expression was elevated in multiple tumor types, including AML.
   AML cell lines exhibit varying levels of CELF1 expression, with
   drug-resistant cell lines showing higher CELF1 expression compared to
   parental cells. CELF1 knockdown reduced drug resistance, promoted cell
   death, and inhibited autophagy. Mechanistic analysis identified ATG5 as
   a potential CELF1-regulated target gene, with CELF1 knockdown reducing
   ATG5 expression and mRNA decay. These findings indicate that targeting
   CELF1 could overcome ADR resistance in AML by modulating autophagy
   through ATG5 regulation, highlighting its clinical significance as a
   therapeutic target for enhancing ADR response in AML.

   Subject terms: Cancer, Prognostic markers

Introduction

   Acute Myeloid Leukemia (AML) is a type of blood cancer characterized by
   abnormal proliferation and differentiation of myeloid cells, with high
   mortality rates primarily affecting older adults [[38]1]. The overall
   5-year survival rate is less than 40% [[39]2]. The hallmark features of
   AML include immature myeloid cell heterogeneity and rapid growth rate,
   leading to disrupted normal hematopoiesis. Despite the development of
   novel agents and improved treatment strategies, relapse and drug
   resistance remain major hurdles towards achieving optimal cure rates
   for AML patients [[40]3]. Abnormalities in key genes during the
   progression of the disease may be crucial drivers of AML development
   and drug resistance [[41]4]. Therefore, exploring novel biomarkers and
   gaining a deeper understanding of their impact on the biological
   characteristics of AML and the molecular mechanisms of clinical
   resistance could potentially enhance treatment outcomes and survival
   rates for AML patients.

   The occurrence and development of AML are largely influenced by the
   regulation of RNA-binding proteins (RBP). Research has shown that
   aberrant expression of RBPs is often associated with enhanced
   transcriptional and post-transcriptional mechanisms promoting cancer
   progression [[42]5, [43]6]. CELF1, a member of the CUGBP Elav-like
   family (CELF) of RNA-binding proteins, was initially identified as an
   RBP binding to the CUG triplet repeat sequence within muscle atrophy
   kinase mRNA. It plays a crucial role in the pathogenesis of Myotonic
   dystrophy (dystrophia myotonica, DM) [[44]7, [45]8]. Further studies
   have revealed that CELF1 binds specifically to certain RNA elements and
   regulates numerous post-transcriptional processes such as alternative
   splicing, translational control, and mRNA decay regulation, thereby
   contributing to the progression of various human diseases, including
   cancer [[46]9–[47]11]. However, research on CELF1 in AML has not been
   reported. Our study found that CELF1 is abnormally highly expressed in
   AML cells, with a more pronounced expression observed in
   Adriamycin-resistant (ADR) cells. Further exploration revealed that
   CELF1 may be involved in the regulation of autophagy in ADR resistance,
   but the key molecular mechanisms remain to be elucidated.

   The development of chemotherapy resistance in hematological
   malignancies poses a significant challenge to clinical treatment,
   representing one of the major obstacles in tumor therapy [[48]12,
   [49]13]. Research has demonstrated that autophagy plays a crucial role
   in tumorigenesis and progression [[50]14]. Specifically, autophagy can
   eliminate mutated cells, damaged cellular organelles, and genomic
   instability during early stages of tumor formation, thereby inhibiting
   tumor initiation [[51]15]. In established tumors, autophagy is enhanced
   as a response to nutrient deprivation, energy insufficiency, hypoxic
   stress, and chemotherapy-induced damage, allowing cancer cells to
   continue growth and dissemination without destruction, ultimately
   leading to the induction of acquired resistance [[52]16–[53]18]. In
   certain hematological malignancies, autophagy has been found to be a
   key mechanism that enables tumor cells to evade immune surveillance and
   killing [[54]19–[55]21]. Moreover, after exposure to chemotherapeutic
   agents, autophagy can be activated in these cancer cells, resulting in
   the development of drug resistance [[56]22–[57]24].

   This study discovered that CELF1 specifically binds to the mRNA of
   ATG5, a critical gene involved in autophagy regulation, and promotes
   its expression in ADR-resistant acute myeloid leukemia (AML) cells.
   This activation of protective autophagy enhances the AML cells’
   resistance to ADR treatment. Furthermore, we established an in vivo
   model system to validate CELF1’s regulatory role on ATG5-mediated
   autophagy as a tumor-suppressive mechanism. Therefore, targeting CELF1
   specifically or inhibiting ATG5-mediated autophagy has significant
   implications for further translational research and clinical
   application in AML treatment.

Results

CELF1 is a potential target for ADR resistance in AML cells

   Firstly, we conducted a pan-cancer analysis of CELF1 expression using
   TCGA data, revealing significantly elevated levels of CELF1 across
   various tumors, with the most notable difference observed in AML (Fig.
   [58]1A, B). Clinical baseline data indicates a significant correlation
   between high CELF1 expression and lower classifications within the FAB
   system (M0&M1&M2), while no significant correlations were found with
   Bone Marrow Blasts proportion, Age, Cytogenetic risk, and FLT3 mutation
   (Table [59]1). Then, We selected five commonly used AML cell lines
   (HL-60, MOLM-13, U937, THP-1, and KG-1) to investigate CELF1 expression
   levels using RT-qPCR and Western Blot. The results indicate that the
   levels of CELF1 mRNA and protein expression in AML cell lines are
   significantly higher than those in HS-5. Furthermore, there are
   variations in CELF1 expression levels among the different cell lines,
   with KG-1 exhibiting the highest expression and HL-60 the lowest (Fig.
   [60]1C, D). To validate the relationship between CELF1 expression and
   ADR resistance, we conducted CCK-8 assays on KG-1 and HL-60 cells.
   Compared to HL-60 cells, KG-1 cells demonstrated significant resistance
   to ADR treatment (Fig. [61]1E). Notably, the IC[50] values of ADR in
   KG-1 cells were substantially higher than those observed in HL-60 cells
   (228.6 ± 8.3 nM vs. 84.21 ± 3.03 nM) (Fig. [62]1F). We established two
   ADR-resistant cell lines, KG-1/ADR (2558 ± 86 nM) and HL-60/ADR
   (1328 ± 41 nM), which exhibited increased IC[50] values compared to
   their parental cells. Specifically, the IC[50] values of KG-1/ADR and
   HL-60/ADR cells increased by ~10-fold (Fig. [63]1G, H). RT-qPCR and
   Western Blot analysis revealed that CELF1 expression was significantly
   upregulated in both resistant cell lines (Fig. [64]1I, J). This finding
   suggests a potential clinical relevance for targeting CELF1 to overcome
   ADR resistance in AML.

Fig. 1. CELF1 is overexpressed in in AML drug-resistant cell lines.

   [65]Fig. 1
   [66]Open in a new tab

   A, B Pan-cancer expression of CELF1 between tumor tissues and normal
   tissues from TCGA. C, D RT-qPCR and Western Blot evaluate CELF1
   expression across diverse AML cell lines. E, F The IC[50] values and
   inhibitory curve in KG-1 and HL-60 cells with ADR tested. G, H The
   IC[50] values and inhibitory curve in KG-1/ADR and HL-60/ADR cells with
   ADR tested. I, J RT-qPCR and Western Blot analysis of CELF1 expression
   in AML drug-resistant cells compared to parental cells.

Table 1.

   Correlations between CELF1 expression and clinicopathologic features of
   AML patients.
   Characteristics Low expression of CELF1 High expression of CELF1 P
   value
   n 75 75
   Gender, n (%) 0.412
    Female 36 (24%) 31 (20.7%)
    Male 39 (26%) 44 (29.3%)
   Race, n (%) 0.979
   Asian&Black or African American 7 (4.7%) 7 (4.7%)
    White 68 (45.6%) 67 (45%)
   FAB classifications, n (%) 0.022*
    M0&M1&M2 36 (24.2%) 52 (34.9%)
    M3 9 (6%) 5 (3.4%)
    M4&M5&M6&M7 30 (20.1%) 17 (11.4%)
   BM blasts(%), n (%) 0.867
    ≤20 29 (19.3%) 30 (20%)
    >20 46 (30.7%) 45 (30%)
   Age, n (%) 0.408
    ≤60 46 (30.7%) 41 (27.3%)
    >60 29 (19.3%) 34 (22.7%)
   WBC count(×10^9/L), n (%) 0.933
    ≤20 38 (25.5%) 38 (25.5%)
    > 20 36 (24.2%) 37 (24.8%)
   Cytogenetic risk, n (%) 0.487
    Favorable 15 (10.1%) 15 (10.1%)
    Intermediate/normal 44 (29.7%) 38 (25.7%)
    Poor 15 (10.1%) 21 (14.2%)
   FLT3 mutation, n (%) 0.432
    Negative 52 (35.6%) 49 (33.6%)
    Positive 20 (13.7%) 25 (17.1%)
   [67]Open in a new tab

   Clinical data were derived from The Cancer Genome Atlas (TCGA) AML
   cohort.

   AML acute myelogenous leukemia, FAB French-American-British
   classification, WBC White Blood Cell.

CELF1 knockdown promotes drug sensitivity and reduces autophagy in AML cells

   In the treatment of acute myeloid leukemia (AML), the development of
   drug resistance remains a major cause of chemotherapy failure and
   disease relapse. Recent studies suggest that autophagy, a critical
   regulatory mechanism of cellular metabolic homeostasis, plays a pivotal
   role in AML drug resistance. Therefore, We performed Western blot to
   assess the expression of three key autophagy markers (p62, Beclin-1,
   LC3-II/LC3-I) to evaluate changes in autophagy levels in drug-resistant
   cells and their parental counterparts. The results revealed increased
   LC3-II/LC3-I ratios, decreased p62 protein levels, and elevated Beclin1
   protein levels in HL-60/ADR and KG-1/ADR cells compared to HL-60 and
   KG-1 cells, indicating significantly enhanced autophagic activity in
   the resistant cells (Supplementary Fig. [68]S1A). Treatment with the
   autophagy inhibitor Bafilomycin A1 (BafA1) markedly suppressed
   autophagic activation in HL-60/ADR and KG-1/ADR cells (Supplementary
   Fig. [69]S1B). Tandem GFP-mRFP-LC3 fluorescence analysis showed that
   BafA1 treatment decreased the number of yellow and red puncta in
   HL-60/ADR and KG-1/ADR cells, confirming its role in reversing
   autophagy activation in drug-resistant cells (Supplementary Fig.
   [70]S1C). Furthermore, CCK-8 assays demonstrated that BafA1 reduced the
   IC[50] values of ADR in both resistant cell lines (Supplementary Fig.
   [71]S1D).

   To validate the role of CELF1 in AML drug resistance, we used siRNA to
   knockdown CELF1 expression in KG-1/ADR and HL-60/ADR cells, with the
   silencing efficiency demonstrated in Supplementary Fig. [72]S2A, B.
   CCK-8 assay results indicated that downregulation of CELF1
   significantly diminished the drug-resistant phenotype (Fig. [73]2A, B).
   Furthermore, flow cytometry results showed that CELF1 knockdown
   significantly increased ADR-induced cell apoptosis (Fig. [74]2C, D).
   Results of Western blot revealed a significant upregulation of p62
   expression and a marked decrease in the ratio of LC3-II/LC3-I and
   Beclin-1 expression after CELF1 knockdown (Fig. [75]2E), indicating
   that CELF1 knockdown can inhibit autophagy levels. Transmission
   electron microscopy images confirmed these findings, showing a
   significant reduction in the number of autophagosomes in CELF1
   knockdown KG-1/ADR and HL-60/ADR cells compared to the control group
   (Fig. [76]2F, G). Additionally, we conducted tandem mRFP-GFP-LC3
   fluorescence analysis. The results indicated that knockdown of CELF1
   resulted in a reduction in the number of yellow and red puncta in
   HL-60/ADR and KG-1/ADR cells, effectively block autophagic flux (Fig.
   [77]2H, I). The above results indicate that CELF1 knockdown promotes
   AML cell apoptosis and inhibits autophagy.

Fig. 2. CELF1 knockdown promotes cell apoptosis and inhibits autophagy.

   [78]Fig. 2
   [79]Open in a new tab

   A, B CCK8 assess IC[50] values and inhibition curves after CELF1
   knockdown in KG-1/ADR and HL-60/ADR cells. C, D Flow cytometry analysis
   of apoptosis in cells following CELF1 knockdown. E Western blot
   analysis of autophagy markers in cells after CELF1 knockdown. F, G
   Autophagic organelles visualized by transmission electron microscopy in
   cells after CELF1 knockdown. H, I Tandem mRFP-GFP-LC3 fluorescence
   assay performed in CELF1 knockdown cells to monitor autophagic flux.

CELF1 enhances AML cell sensitivity to ADR through inhibition of autophagy

   Previous studies have indicated that autophagy is a potential mechanism
   of resistance to various cytotoxic drugs in a range of cancers,
   including hematologic malignancies. Therefore, we further investigated
   whether CELF1 promotes ADR resistance in AML cells by inhibiting
   cellular autophagy. We conducted rescue experiments by treating
   KG-1/ADR and HL-60/ADR cells with the autophagy inducer rapamycin
   (Rapa). CCK-8 assay revealed that cells treated with Rapa exhibited
   higher ADR IC[50] values compared to the si-NC group, indicating
   increased ADR resistance. This suggests that enhancing autophagy in AML
   cells indeed contributes to ADR resistance. Our previous findings
   demonstrated that depletion of CELF1 significantly reduced ADR
   resistance, and adding Rapa on this basis partially restored ADR
   resistance (Fig. [80]3A, B). Similarly, flow cytometry results showed
   that Rapa alone could inhibit cell apoptosis, while adding Rapa to
   CELF1 knockdown cells partially suppressed the apoptosis induced by
   CELF1 downregulation (Fig. [81]3C, D). This suggests that CELF1 likely
   promotes AML cell resistance to ADR through regulating cellular
   autophagy. Subsequently, we used Western blot analysis to examine the
   expression of autophagy markers. The results showed that compared to
   the si-NC group, p62 expression was significantly reduced in the Rapa
   group, while the ratio of LC3-II/LC3-I and Beclin-1 expression markedly
   increased, indicating that Rapa significantly enhanced cellular
   autophagy levels. In CELF1-knocked down cells treated with Rapa, p62
   expression increased, and the ratio of LC3-II/LC3-I and Beclin-1
   expression decreased significantly compared to the group with only
   CELF1 knockdown (Fig. [82]3E). Additionally, transmission electron
   microscopy results demonstrated an increase in autophagosome formation
   with Rapa treatment, with more autophagosomes observed in CELF1
   knockdown cells treated with Rapa compared to the group with only CELF1
   downregulation(Fig. [83]3F, G). The results of the tandem mRFP-GFP-LC3
   fluorescence assay showed that treatment with Rapa alone significantly
   increased autophagic flux in HL-60/ADR and KG-1/ADR cells. However,
   knockdown of CELF1 expression in these cells partially reversed the
   Rapa-induced enhancement of autophagic flux(Fig. [84]3H, I). These
   suggest that the addition of Rapa can restore cellular autophagy and
   ADR drug sensitivity in CELF1 knockdown cells.

Fig. 3. ADR resistance and autophagy induction following CELF1 knockdown and
Rapa treatment.

   [85]Fig. 3
   [86]Open in a new tab

   A, B CCK8 assess IC[50] values and inhibition curves after CELF1
   knockdown and Rapa treatment in AML cells. C, D Flow cytometry analysis
   of apoptosis in cells following CELF1 knockdown and Rapa treatment. E
   Western blot analysis of autophagy markers in cells after CELF1
   knockdown and Rapa treatment. F, G Autophagic organelles visualized by
   transmission electron microscopy in cells after CELF1 knockdown and
   Rapa treatment. H, I Tandem mRFP-GFP-LC3 fluorescence assay performed
   in cells after CELF1 knockdown and Rapa treatment to monitor autophagic
   flux.

ATG5 is a potential downstream target of CELF1

   In order to identify downstream target genes regulated by CELF1 in AML
   cells, we performed RNA-seq high-throughput sequencing on KG-1/ADR
   cells before and after CELF1 knockdown. The transcriptome analysis
   encompassed 33,094 genes, of which 987 showed statistically significant
   changes in expression following CELF1 knockdown, with 490 genes
   upregulated and 497 genes downregulated (|log2FC|≥ 1, P < 0.05). The
   heatmap displayed illustrates the top 100 upregulated and downregulated
   genes (Fig. [87]4A). In the GO enrichment analysis, several GO terms
   related to DNA replication and cell cycle processes were enriched,
   including DNA replication initiation, DNA strand elongation,
   double-strand break repair, and cell cycle phase transition, which were
   the most enriched biological processes. This finding was consistent
   with our cell proliferation data (Fig. [88]4B). In the KEGG pathway
   enrichment analysis (Fig. [89]4C), the most significant pathways,
   listed in ascending order of p-value, included DNA replication, Cell
   cycle, Pathways in cancer, MAPK signaling pathway, and PI3K-Akt
   signaling pathway. To identify CELF1 target genes, we applied more
   stringent criteria to the differentially expressed genes from
   sequencing, using a P < 0.01 and a fold change of 3, presented in a
   volcano plot (Fig. [90]4D). After excluding non-coding genes from the
   153 downregulated genes, we obtained 146 protein-coding genes.
   Intersection analysis with the Human Autophagy Database(HADb) revealed
   ATG5 and CDKN1A as potential target genes of CELF1 (Fig. [91]4E).
   Furthermore, analysis of the correlation between CELF1 and ATG5/CDKN1A
   expression in the TCGA database showed a significant positive
   correlation between CELF1 and ATG5 (R = 0.481, P < 0.001), while no
   significant correlation was found between CELF1 and CDKN1A (R = 0.083,
   P = 0.311) (Fig. [92]4F). Subsequent RT-qPCR and Western blot analyses
   demonstrated that knocking down CELF1 significantly reduced ATG5
   expression, while the expression of CDKN1A did not show significant
   changes (Fig. [93]4G, H), indicating that ATG5 is an important target
   gene regulated by CELF1 in controlling cellular autophagy.

Fig. 4. RNA-seq analysis of potential downstream targets following CELF1
depletion.

   [94]Fig. 4
   [95]Open in a new tab

   A The heat map illustrates DEGs following CELF1 depletion. B The top 20
   enriched GO terms among DEGs. C The top 20 enriched KEGG pathways among
   DEGs. D Volcano plot showing 240 DEGs following CELF1 knockdown. E Venn
   diagram displays the common genes between the downregulated coding
   genes and the HADb. F Spearman’s correlation analysis of ATG5, CDKN1A
   with CELF1 expression in the AML profiles of TCGA dataset. G, H RT-qPCR
   and Western blot analysis the expression of ATG5 and CDKN1A following
   CELF1 depletion.

CELF1 interacts with ATG5 mRNA and maintains its stability

   To further explore the potential molecular mechanism of CELF1 in
   regulating ATG5 expression, we conducted RIP assays to validate the
   targeting of ATG5 mRNA by CELF1 protein. The combined results of RIP
   and RT-qPCR showed that the CELF1-specific antibody effectively
   co-precipitated ATG5 mRNA compared to the negative IgG control (Fig.
   [96]5A). Additionally, using biotin-labeled ATG5 mRNA sequences and
   their reverse sequences in RNA pull-down experiments, Western blot
   results demonstrated specific interaction between ATG5 mRNA and CELF1
   protein, confirming their relationship (Fig. [97]5B). In mRNA decay
   experiments, after actinomycin D treatment, the abundance of ATG5 mRNA
   in the CELF1 knockdown group decreased significantly over time compared
   to the control group, indicating that CELF1 specifically binds to ATG5
   mRNA in AML cells and increases its half-life (Fig. [98]5C, D).

Fig. 5. CELF1 binds to ATG5 mRNA and preserves its stability.

   [99]Fig. 5
   [100]Open in a new tab

   A RIP-qPCR analysis confirmed CELF1 binding to ATG5 mRNA, normalized
   against the input group with IgG as the negative control. B The RNA
   pull-down assay and western blot analysis confirmed the direct binding
   of biotinylated ATG5 to CELF1 proteins. C, D CELF1 depletion was
   studied for its effect on ATG5 mRNA stability in AML cells by
   actinomycin D-chase experiments. E, F Schematic representation of CELF1
   binding motifs on ATG5 and the fragmented ATG5 mRNA biotinylated
   transcripts. G Representative CELF1 immunoblots of the material pulled
   down by the different biotinylated fragments of the ATG5 mRNA. H
   Binding sites of CELF1 on ATG5 and mutant sequences. I Dual-luciferase
   reporter assay demonstrating CELF1-mediated regulation of ATG5 mRNA
   abundance.

   Previous studies have shown that CELF1, as a crucial RNA-binding
   protein, can specifically recognize CUG, UG, and UA repeat sequences
   (such as CUGCUG, UGUGUG, UAUAUA, etc.) [[101]25–[102]27]. Through
   sequence alignment, we identified two UGUGUG sites in the CDS region
   and one CUGCUG site in the 3’UTR region of ATG5 mRNA (Fig. [103]5E).
   Subsequently, we fragmented ATG5 mRNA into smaller transcripts to
   determine specific binding sites for CELF1. By constructing
   biotin-labeled full-length ATG5 transcript (FL) and three region
   fragments (F1, F2, and F3), we conducted RNA pull-down analysis (Fig.
   [104]5F). The results showed the strongest binding affinity of CELF1 to
   the F1 region, weaker binding to the F2 region, and no observed binding
   to F3 (Fig. [105]5G). These findings clearly demonstrate that CELF1 can
   specifically bind to ATG5 mRNA, with its highest affinity spanning the
   CDS region.

   To elucidate whether CELF1 regulates the stability of ATG5 mRNA by
   specifically recognizing the UGUGUG motif and to demonstrate the
   efficiency and effect of the binding site, we designed mutated
   sequences for the binding site and created a model diagram (Fig.
   [106]5H). Subsequently, we constructed luciferase reporter vectors
   containing either the wild-type ATG5 mRNA (with the UGUGUG motif) or
   the mutated sequence (ACACAC) and transfected them into 293T cells. By
   co-transfecting with the OE-CELF1 plasmid and measuring changes in
   luciferase activity, we found that CELF1 overexpression significantly
   increased luciferase activity in the wild-type group, indicating an
   enhancement in mRNA stability or translation efficiency, while this
   effect was completely abolished in the MUT1 ([250]UGUGUG[255]) (Fig.
   [107]5H). These results provide direct evidence that CELF1 regulates
   the stability of ATG5 mRNA by binding to its UGUGUG motif, offering
   molecular-level insights into their functional interaction.

CELF1 regulates autophagy in AML cells via ATG5 to promote ADR resistance

   To confirm whether CELF1 influences cellular autophagy and ADR
   resistance by targeting ATG5, we conducted rescue experiments, with the
   overexpression efficiency of ATG5 demonstrated in Supplementary Fig.
   [108]S2C, D. The results from the CCK-8 assay showed that Compared to
   the si-NC group, the IC[50] of the si-CELF1 group was significantly
   reduced. Furthermore, following the knockdown of CELF1, overexpression
   of ATG5 resulted in a significant increase in the IC[50] value in the
   si-CELF1 + OE-ATG5 group compared to the si-CELF1 group, indicating
   that overexpression of ATG5 partially reversed the increased
   sensitivity to ADR in CELF1-knocked down KG-1/ADR and HL-60/ADR cells
   (Fig. [109]6A, B). Similarly, flow cytometry results demonstrated that
   in KG-1/ADR and HL-60/ADR cells with CELF1 knockdown alone, apoptosis
   was significantly higher compared to the control group, while in the
   rescue group with ATG5 overexpression on the basis of CELF1 knockdown,
   this phenomenon was significantly reversed (Fig. [110]6C, D). This
   suggests that CELF1 likely mediates ADR resistance through targeting
   ATG5. Furthermore, we used Western blot analysis to examine
   autophagy-related proteins. The results showed that overexpression of
   ATG5 reversed the decreased ratio of LC3-II/LC3-I and Beclin-1 levels,
   and reduced the upregulation of p62 in CELF1-knocked down KG-1/ADR and
   HL-60/ADR cells (Fig. [111]6E). Additionally, transmission electron
   microscopy results indicated that knocking down CELF1 reduced the
   formation of autophagosomes in cells, while a slight increase in the
   number of autophagosomes was observed in cells overexpressing ATG5 on
   the basis of CELF1 knockdown (Fig. [112]6F, G). The tandem mRFP-GFP-LC3
   reporter assay demonstrated that CELF1 knockdown significantly
   suppressed cellular autophagic flux. Notably, this suppression was
   partially reversed by ATG5 overexpression in CELF1- depleted cells
   (Fig. [113]6H, I). This suggests that CELF1 can regulate cellular
   autophagy through ATG5.

Fig. 6. CELF1 modulates AML cell autophagy via ATG5 to combat ADR resistance.

   [114]Fig. 6
   [115]Open in a new tab

   A, B CCK8 assess IC[50] values and inhibition curves after CELF1
   knockdown and ATG5 upregulation in AML cells. C, D Flow cytometry
   analysis of apoptosis in cells following CELF1 knockdown and ATG5
   upregulation. E Western blot analysis of autophagy markers in cells
   after CELF1 knockdown and ATG5 upregulation. F, G Autophagic organelles
   visualized by transmission electron microscopy in cells after CELF1
   knockdown and ATG5 upregulation. H, I Tandem mRFP-GFP-LC3 fluorescence
   assay performed in cells after CELF1 knockdown and ATG5 upregulated to
   monitor autophagic flux.

   In addition, we explored the effect of Rapa on the mTOR pathway in
   drug-resistant cells to determine whether the resistance and the effect
   of rapamycin arise from different pathways. Rapa treatment
   significantly reduced phosphorylated mTOR (p-mTOR) levels in
   drug-resistant cells while markedly inducing autophagy (Supplementary
   Fig. [116]S3A). These findings indicate that Rapa effectively inhibits
   mTOR activity in drug-resistant cells, thereby relieving its
   suppression on autophagy. Further investigation demonstrated that CELF1
   knockdown did not alter the expression of mTOR pathway-related proteins
   (Supplementary Fig. [117]S3B), suggesting that the drug resistance
   mechanism regulates autophagy through the CELF1-ATG5 axis independently
   of the mTOR pathway. This evidence further supports the critical
   regulatory role of the CELF1-ATG5 axis in AML drug resistance.

CELF1 knockdown combined with ADR treatment improves AML xenograft model
survival

   To further elucidate the role of CELF1 in AML cell resistance in vivo,
   we established a disseminated AML mouse model using HL-60/ADR cells
   stably expressing luciferase. Three days after cell injection, ADR
   treatment was administered, and bioluminescence imaging was performed
   on days 7, 14, and 28 to track the fate of the target cells. The Blank
   group received no drug treatment, and the In Vivo Imaging system
   revealed rapid growth of AML cells. The remaining three groups
   underwent ADR treatment under the same conditions. In vivo fluorescence
   imaging showed that stable knockdown of CELF1 significantly reduced
   fluorescence intensity compared to the NC group, indicating that CELF1
   knockdown effectively increased tumor cell killing and sensitivity to
   ADR. Overexpression of ATG5 on the basis of CELF1 knockdown increased
   fluorescence intensity, suggesting that the rescue effect of ATG5
   partially reversed cell ADR resistance (Fig. [118]7A, B). Furthermore,
   bone marrow cells from some mice after 28 days were subjected to flow
   cytometry analysis. The results showed that the Blank group had a high
   positivity rate of CD45^+ cells at 74%, while the NC group had 46.6%
   CD45^+ cells. In the CELF1 knockdown group, the CD45^+ cell proportion
   was only 7.85%, significantly lower than the NC group, and in the ATG5
   rescue group, the CD45^+ cell proportion recovered to 24.6% (Fig.
   [119]7C, D). Histological analysis of the spleen, lung, liver, and
   kidney tissue revealed differential levels of infiltration and tissue
   damage among the groups. The Blank and NC groups displayed severe
   tissue infiltration and structural damage, while the CELF1 knockdown
   group exhibited significantly reduced infiltration and preservation of
   tissue architecture. In contrast, the ATG5 rescue partially reversed
   these protective effects, as evidenced by increased infiltration and
   damage in comparison to the CELF1 knockdown group (Fig. [120]7E).
   Additionally, all ADR treatment groups showed extended survival, with
   the CELF1 knockdown combined treatment group exhibiting the best
   survival, while ATG5 rescue weakened some of the treatment effects(Fig.
   [121]7F). It is evident that knockdown CELF1 expression effectively
   enhances the therapeutic effect of ADR on AML xenografts, and this
   effect is mediated through the regulation of ATG5 expression.

Fig. 7. CELF1 depletion combined with ADR treatment reduced leukemia
progression in an AML mouse model.

   [122]Fig. 7
   [123]Open in a new tab

   A, B Comparison of the evolution of bioluminescence emission in mice
   treated with ADR in combination with each group or the untreated group
   after injection of HL-60/ADR Luci cells, measured by the IVIS Spectrum.
   C, D CD45^+ cell proportion in mouse bone marrow assessed by flow
   cytometry in each group. E H&E staining shows that CELF1 knockdown
   reduces leukemic infiltration and preserves tissue structure in the
   spleen, lung, liver and kidney, while ATG5 overexpression partially
   reverses this effect. F Kaplan–Meier survival curves of the four
   different treatment groups.

Discussion

   Acute Myeloid Leukemia (AML), a severe malignant hematologic tumor,
   originates from hematopoietic stem cell progenitors [[124]2]. Among the
   eight subtypes of AML (M0-M7), only Acute Promyelocytic Leukemia (APL,
   M3 subtype) can achieve clinical remission rates of over 95% with the
   use of cytarabine (Ara-C) or all-trans retinoic acid (ATRA) [[125]28,
   [126]29]. However, other types of AML, while controllable with standard
   chemotherapy, lead to relapse or drug resistance in around 40% of young
   patients and the majority of elderly patients [[127]30]. The issue of
   chemoresistance in AML poses a critical challenge in clinical
   treatment. Numerous studies suggest that RNA-binding proteins (RBPs)
   play crucial regulatory roles in cells, participating in processes such
   as gene transcription, splicing, and translation. Aberrant expression
   of RBPs in tumor cells can alter cellular responses to chemotherapy
   [[128]31]. For instance, abnormal RBP expression can affect the
   expression or regulation of drug target genes, leading to reduced
   sensitivity of tumor cells to drugs [[129]32]. It may also disrupt
   normal cellular signaling pathways, promoting cell proliferation or
   inhibiting apoptosis, thereby enhancing drug resistance [[130]33,
   [131]34]. Therefore, there is an urgent need to broaden our
   understanding of the relationship between RBPs and AML drug resistance,
   identify new therapeutic targets for AML clinical resistance, and
   provide a new theoretical basis for treating relapsed AML.

   CELF1, also known as CUG-repeat binding protein 1 (CUGBP1), is a novel
   RNA-binding protein containing three RNA recognition motifs (RRMs) that
   bind to specific mRNAs and determine their fate [[132]7, [133]27].
   Studies have shown that CELF1 is primarily found in skeletal muscle,
   heart, and brain, initially discovered to be involved in the selective
   translation of muscle atrophy kinase in myotonic dystrophy [[134]8,
   [135]35, [136]36]. Additionally, CELF1 plays important roles in the
   occurrence and development of various tumors. For example, CELF1 is
   overexpressed in gastric and lung cancers, and knocking down CELF1
   significantly inhibits tumor cell proliferation [[137]37, [138]38]. In
   gliomas, CELF1 is upregulated and promotes cell proliferation by
   inhibiting CDKN1B [[139]39]. Furthermore, CELF1 enhances cell
   migration, invasion, and chemoresistance in colorectal cancer by
   targeting ETS2 [[140]10]. In this study, analysis of TCGA data revealed
   significantly higher expression of CELF1 in AML patients compared to
   normal individuals, suggesting a potential important role for CELF1 in
   AML progression. To further investigate the relationship between CELF1
   expression levels in AML cells and ADR resistance, we established two
   AML cell lines with acquired resistance, analyzed the differential
   expression of CELF1 between these cells and their parental cell lines,
   and found significantly elevated expression of CELF1 in the
   drug-resistant AML cells. Using CCK-8 assays and siRNA knockdown
   techniques, we identified CELF1 as a key molecule involved in acquired
   AML drug resistance. Additionally, we found that knocking down CELF1
   affects cellular autophagy levels, which is a critical factor in tumor
   drug resistance. Therefore, in-depth exploration of CELF1’s regulation
   of autophagy-mediated AML drug resistance holds significant clinical
   significance.

   Cell autophagy is a fundamental physiological process through which
   cells self-digest damaged or excessive cellular components and
   structures to maintain intracellular homeostasis, energy balance, and
   promote cell renewal and remodeling [[141]40]. Under normal conditions,
   autophagy plays a crucial role in cell health and longevity. However,
   in certain abnormal states such as disease or aging, autophagy may
   become dysregulated, leading to cellular abnormalities [[142]41,
   [143]42]. Previous studies have shown that cellular autophagy can be a
   survival mechanism triggered by stress conditions like cancer
   treatment, often contributing to chemotherapy resistance and long-term
   dormancy of tumor cells, ultimately leading to advanced clinical
   symptoms of relapse and metastasis [[144]43]. For example, docetaxel
   treatment of triple-negative breast cancer induces autophagy,
   inactivating the NF-κB signaling pathway, reducing apoptosis rates, and
   increasing resistance to Epirubicin [[145]44]. Autophagy is involved in
   cadmium resistance in lung epithelial fibroblasts by inducing multidrug
   resistance protein 1 (MRP1) and balancing endoplasmic reticulum stress
   [[146]45]. Additionally, Docetaxel promotes the expression of high
   mobility group box 1 (HMGB1), enhancing the formation of the autophagy
   core complex Beclin-1-PI3K-III through the MEK/ERK1/2 pathway, leading
   to resistance to Docetaxel [[147]46]. Currently, several chemotherapy
   drugs used in AML treatment, such as ADR and cytarabine, can induce
   autophagy as a survival mechanism to resist cytotoxic stress and
   counteract the therapeutic effects of these drugs [[148]47]. Therefore,
   identifying the role and potential mechanisms of CELF1 in autophagy
   regulation may be an effective strategy to enhance efficacy and
   overcome resistance in AML treatment.

   To uncover the potential mechanisms by which CELF1 drives acquired drug
   resistance in AML cells, we conducted RNA-seq high-throughput
   sequencing analysis of KG-1/ADR cells before and after CELF1 knockdown,
   revealing significant changes in 987 genes upon CELF1 knockdown. Many
   of these genes were enriched in Gene Ontology terms related to DNA
   replication and the cell cycle. In the KEGG pathway enrichment
   analysis, pathways such as DNA replication, the cell cycle, pathways in
   cancer, the MAPK signaling pathway, and the PI3K-Akt signaling pathway
   were enriched. This suggests that CELF1 likely promotes AML cell
   resistance by regulating downstream target genes involved in these
   pathways. Subsequently, we identified two downstream target genes of
   CELF1 involved in regulating autophagy, ATG5 and CDKN1A, through
   analysis with the Human Autophagy Database. Expression correlation
   analysis revealed a strong positive correlation between CELF1 and ATG5,
   while no significant correlation was found between CELF1 and CDKN1A.
   Our RT-qPCR and Western blot results showed significant downregulation
   of ATG5 and no significant change in CDKN1A expression upon CELF1
   knockdown. Rescue experiments both in vitro and in vivo further
   confirmed that CELF1 enhances AML cell tolerance to ADR, with CELF1
   increasing cell resistance through regulation of ATG5 expression.
   Autophagy is a tightly regulated process, with key participants being
   autophagy-related (ATG) proteins. Among the identified 41 ATG genes,
   ATG5 is essential for autophagosome formation and positively influences
   autophagic activity [[149]48]. m^6A-modified USP13 induces autophagy
   and imatinib resistance in gastrointestinal stromal tumor cells by
   regulating ATG5 stability [[150]49]. miR-137 promotes sensitivity of
   pancreatic cancer cells to ADR by inhibiting ATG5-mediated autophagy
   [[151]50]. Additionally, exosomal LOC85009 regulates cell autophagy to
   inhibit resistance to docetaxel through the USP5/USF1/ATG5 axis
   [[152]51]. Therefore, ATG5 serves as a crucial intermediate link in
   CELF1-regulated autophagy and ADR resistance in AML.

   Subsequently, we performed RIP-qPCR and RNA pull-down experiments to
   confirm the direct interaction between CELF1 protein and ATG5 mRNA.
   Moreover, mRNA decay experiments demonstrated that knocking down CELF1
   expression significantly reduced the half-life of ATG5 mRNA. Previous
   studies have shown that CELF1, also known as CUG-repeat binding protein
   1 (CUGBP1), contains three RNA recognition motifs (RRMs) that
   specifically bind to CUG repeat sequences (such as CUGCUG) in the
   N-terminal RRM (RRM1 and RRM2), and UG repeat sequences (such as
   UGUGUG, CGUGUG) in the C-terminal RRM3, mediating alternative splicing
   events and mRNA stability [[153]8, [154]25–[155]27, [156]52]. Through
   sequence alignment, we found two UGUGUG sites in the CDS region and one
   CUGCUG site in the 3’UTR region of ATG5 mRNA. Binding activity
   experiments with different segments of CELF1 and ATG5 transcripts
   revealed significant differences in binding affinity between segments.
   Regions containing two UGUGUG sites showed the strongest binding
   capacity, while regions containing one CUGCUG site had weaker but
   synergistic binding capabilities. Therefore, CELF1 likely positively
   regulates ATG5 mRNA stability through specific binding to these repeat
   sequences, although the exact mechanism by which CELF1 promotes ATG5
   mRNA stability remains to be elucidated.

   In conclusion, this study demonstrates that CELF1 is a novel AML
   chemotherapy resistance-related protein that directly targets and
   regulates ATG5 mRNA stability, promoting its expression and triggering
   protective autophagy-mediated drug resistance in cells. Our findings
   provide new insights into the potential mechanisms of AML chemotherapy
   resistance, which will help clarify the precise regulatory mechanisms
   of AML chemotherapy resistance and offer a new perspective on CELF1 and
   other RBPs as novel therapeutic targets for AML.

Materials and methods

Bioinformatics analysis

   The pan-cancer analysis of the CELF1 gene was conducted using data from
   The Cancer Genome Atlas (TCGA) database. CELF1 expression and clinical
   data for AML were acquired from the TCGA database. Preprocessing steps
   were performed, including normalization and filtering of low-expressed
   genes. To examine the correlation between CELF1 expression and ATG5 and
   CDKN1A in TCGA, Spearman correlation analysis was performed on
   log-transformed and normalized RNA-Seq data. Autophagy genes from the
   Human Autophagy Database ([157]http://www.autophagy.lu/) were used for
   screening downstream target genes of CELF1. Analyzed data are available
   in Supplementary Table [158]S1.

Cell culture

   The AML cell lines (HL-60, MOLM-13, U937, THP-1, and KG-1) and the
   human bone marrow stromal cell line (HS-5) were obtained from ATCC and
   cultured in appropriate growth media. The AML cells were maintained in
   RPMI-1640 medium supplemented with 10% FBS and 1%
   penicillin-streptomycin, while the HS-5 cells were cultured in DMEM
   supplemented with 10% FBS and 1% penicillin-streptomycin. Media were
   changed regularly to sustain optimal cell growth. All cells were
   incubated at 37 °C with 5% CO[2] in a humidified atmosphere.

   ADR-resistant cell lines were established using a stepwise selection
   method. The parental cell lines (HL-60 and KG-1) were gradually exposed
   to increasing concentrations of ADR over several months. Starting with
   a low concentration, the cells were cultured and passaged in
   progressively higher ADR concentrations until an ADR-resistant cell
   line was obtained.

Transfection of cells

   Cells were transfected with either siRNA targeting CELF1 to achieve
   knockdown or a ATG5 overexpression plasmid using Lipofectamine 2000.
   the specific siRNA and a negative control or overexpression plasmid and
   an empty vector control were diluted in serum-free growth medium and
   added to cells at 50–70% confluency. Following incubation, the cells
   were cultured at 37 °C for the appropriate duration to facilitate
   efficient siRNA and DNA uptake.

Cell viability assay

   Cell viability was assessed using the Cell Counting Kit-8 (CCK-8) assay
   according to the manufacturer’s instructions. Briefly, cells in each
   group were seeded into 96-well plates at a density of 2 × 10^3
   cells/well. After incubation for 48 h, the absorbance was determined by
   using a microplate reader set at a wavelength of 450 nm. The
   experiments were repeated in triplicate, independently.

Cell apoptosis assay

   Cell apoptosis was assessed using the Annexin V-FITC/PI staining assay
   followed by flow cytometry analysis. Briefly, cells subjected to
   experimental treatments were collected, washed with PBS, and stained
   with Annexin V-FITC and PI according to the assay kit instructions.
   Flow cytometry analysis was then performed to determine the percentage
   of apoptotic cells based on Annexin V-FITC and PI fluorescence signals.
   Statistical analysis was conducted to compare the percentage of
   apoptotic cells between experimental and control groups, with a
   significance level set at P < 0.05.

RT-qPCR

   TRIzol reagent was utilized to extract total RNA from cells. The
   concentration and purity of the extracted RNAs were assessed using a
   NanoDrop One Spectrophotometer. Subsequently, cDNA synthesis was
   carried out following the manufacturer’s instructions provided with the
   PrimeScript RT reagent Kit. Real-time quantitative PCR (RT-qPCR) was
   performed utilizing the TB Green Premix Ex Taq kit. The primers used
   were as follows: CELF1 F: ATGGCACAGACGGCTATCAAGG, CELF1 R:
   CACAGATGCTGCGCTGATTTGC; ATG5 F: GCAGATGGACAGTTGCACACAC, ATG5 R:
   GAGGTGTTTCCAACATTGGCTCA; β-actin F: TGGCACCCAGCACAATGAA, β-actin R:
   CTAAGTCATAGTCCGCCTAGAAGCA. Relative gene expression was normalized to
   β-actin based on the 2^−ΔΔCt method.

Western blot assays

   Cells were lysed in RIPA buffer supplemented with protease and
   phosphatase inhibitors. The protein concentration was determined using
   a BCA protein assay kit. Equal amounts of protein samples were
   separated by SDS-PAGE and transferred onto PVDF membrane. After
   blocking non-specific binding, the membrane was probed overnight at
   4 °C with primary antibodies(CELF1, CST #95084; Autophagy Antibody
   Sampler Kit, CST #4445; mTOR, CST #2983; p-mTOR, CST #5536; CDKN1A, CST
   #2947) specific to the target proteins. Following washing, the membrane
   was incubated with HRP-conjugated secondary antibodies, and protein
   bands were visualized using ECL detection reagents.

Transmission electron microscopy

   Autophagic structures were observed using electron microscopy. Cells
   were seeded onto culture dishes and treated accordingly. Following
   treatment, cells were fixed, dehydrated, and embedded in resin.
   Ultrathin sections were cut, stained, and mounted on copper grids.
   Autophagic structures, including autophagosomes and autolysosomes, were
   examined under a transmission electron microscope operated.
   Representative images were captured to visualize the observed
   structures associated with autophagy.

Tandem mRFP-GFP-LC3 reporter assay

   Cells were transfected with the mRFP-GFP-LC3 plasmid, followed by
   designated pharmacological treatments (BafA1, rapamycin) or genetic
   manipulations (siRNA-mediated CELF1 knockdown, ATG5 overexpression).
   After 24–48 h, cells were fixed with 4% paraformaldehyde, permeabilized
   with 0.1% Triton X-100, and imaged using a confocal microscope. Yellow
   (GFP^+/mRFP^+) and red (GFP^−/mRFP^+) puncta representing
   autophagosomes and autolysosomes, were counted in cells per group using
   ImageJ, respectively. Autophagic flux was calculated as the ratio of
   red-to-yellow puncta across experimental conditions.

RNA sequencing

   Total RNA was isolated from CELF1 knockdown or control HL-60/ADR cells
   by using Trizol reagent. The quality and quantity of the RNA samples
   were assessed with a spectrophotometer. The high‐throughput sequencing
   and analyses for mRNA were carried out by Majorbio Technology
   (Shanghai, China). Analyzed data are available in Supplementary Table
   [159]S1.

RIP-qPCR

   Approximately 1 × 10^7 HL-60/ADR cells were harvested and rinsed with
   PBS. RNA immunoprecipitation (RIP) was conducted using the Magna RIP
   Kit (Millipore) following the manufacturer’s instructions. In brief,
   cells were lysed for 60 min on ice in RIP Lysis Buffer supplemented
   with protease inhibitor and RNase inhibitor. Then, 50 μL of magnetic
   beads were coupled with either anti-CELF1 antibodies or anti-rabbit IgG
   antibodies. The cell lysates were incubated with the bead pellet
   overnight at 4 °C. Immunoprecipitated RNA was extracted from the pellet
   using phenol chloroform, while the immunoprecipitate underwent
   proteinase K treatment and subsequent RT-qPCR analysis.

RNA pull-down assay

   ATG5 or antisense RNA was transcribed in vitro from the pcDNA3.1 vector
   using the TranscriptAid T7 High Yield Transcription Kit. The
   transcribed ATG5 was purified with the PureLink RNA Mini Kit and
   subsequently biotin-labeled using the Biotin 3’ End DNA Labeling Kit.
   Next, the biotinylated RNA was incubated with whole-cell lysate from
   HL-60/ADR cells at room temperature for 1 h. The RNA-protein complex
   was then isolated using streptavidin magnetic beads, and the bound
   proteins were further extracted for subsequent Western blot analysis.

   Three biotinylated RNA fragments spanning the ATG5 mRNA sequence were
   generated. Whole cell lysates from 293T cells overexpressing CELF1 were
   incubated with the biotinylated RNA fragments for 1 h at room
   temperature to allow for RNA-protein complex formation. Streptavidin
   magnetic beads were then used to isolate the RNA-protein complexes, and
   the bound proteins were extracted for subsequent Western blot analysis.

RNA stability assay

   To assess ATG5 mRNA stability, HL-60/ADR cells were treated with
   Actinomycin D, a transcriptional inhibitor. After seeding, Actinomycin
   D was added at 5 μM, while control cells received vehicle treatment.
   Cells were then incubated for various time intervals to allow for RNA
   degradation. Total RNA was extracted from harvested cells using a
   TRIzol and assessed for quality and quantity. Subsequently, cDNA
   synthesis was performed, and the expression levels of target RNAs were
   measured at each time point using quantitative real-time PCR or other
   appropriate techniques. Data were normalized to housekeeping genes and
   analyzed to determine the rate of RNA degradation under Actinomycin D
   treatment.

CELF1-mediated drug resistance of AML cells in vivo

   Immunocompromised NSG mice (5 weeks old) were purchased from Shanghai
   SLAC Laboratory Animal Co., Ltd. HL-60/ADR cells transfected with
   sh-CELF1 or OE-ATG5 resuspended in PBS were injected directly into via
   tail vein to mimic systemic dissemination. Tumor growth and progression
   were monitored by measuring leukemic burden in peripheral blood samples
   obtained at different time points. Fluorescence imaging was employed to
   visualize the localization and intensity of fluorescently labeled AML
   cells within live animals. All the mice were performed euthanasia after
   the experiment to ensure the welfare and ethical treatment of the
   animals involved.

H&E staining

   Tissues (spleen, lung, liver, and kidney) were collected, fixed in 4%
   paraformaldehyde, and embedded in paraffin. Sections (4 µm) were
   deparaffinized, rehydrated, and stained with hematoxylin for 5 min,
   followed by differentiation in 1% acid-ethanol and bluing in tap water.
   Slides were counterstained with eosin for 2 min, dehydrated, cleared in
   xylene, and mounted with neutral resin. Images were captured using a
   light microscope to assess tissue infiltration and structural integrity
   across experimental groups.

Statistical analysis

   All statistical analyses were performed using GraphPad Prism 10
   software. Data were expressed as mean ± standard deviation (SD) or as
   indicated. A Student’s t test was used to analyze two group
   comparisons. Additionally, correlation analysis was conducted to
   evaluate the association between variables of interest. Differences
   were considered statistically significant at P < 0.05.

Supplementary information

   [160]Table S1^ (13.4MB, xlsx)
   [161]41419_2025_7926_MOESM2_ESM.docx^ (10MB, docx)

   Supplementary Figures and Figure legends
   [162]Original data files of Western blot^ (4.6MB, pdf)
   [163]41419_2025_7926_MOESM4_ESM.xlsx^ (17.5KB, xlsx)

   Original data files (qPCR original Ct values)

Acknowledgements