Unc-51 like kinase 3 (ULK3) contributes to autophagy and cell survival in multiple myeloma.

Текст сохранён. Часть иллюстраций пока доступна только в первоисточнике; загрузка будет проверена повторно.

Introduction

Multiple myeloma (MM) is an incurable bone marrow malignancy that causes bone destruction 1 – 3 . Although proteasome inhibitors (PIs) 4 , CAR-T and bispecific T-cell engagers (BiTEs) 5 , 6 have improved survival, relapse remains inevitable, highlighting the need for new therapies. To identify targets, we leveraged RNA-seq datasets generated at Moffitt from CD138 + MM cells isolated from bone marrow aspirates of newly diagnosed, relapsed/refractory patients, as well as those with monoclonal gammopathy of undetermined significance (MGUS) and smoldering multiple myeloma (SMOL) ( n = 813). Bioinformatic analyses showed that autophagy-related genes correlate with disease stage, supporting a role for autophagy in MM survival 7 . Clinical trials with the autophagy flux inhibitor chloroquine (CQ) are ongoing. Still, they are limited by off-target toxicities and incomplete mechanistic understanding 8 – 10 , underscoring the need for selective inhibitors of key autophagy regulators.

Autophagy initiation in mammalian cells is known to be driven by the ULK1/ULK2 kinase complex, activated upon mTORC1 inhibition and dephosphorylation of the ULK complex (ULK1, ATG13, and FIP200) during nutrient starvation 11 , 12 . Emerging evidence in other cancers suggests that ULK3 can promote autophagy independently of classical starvation signaling 12 , potentially functioning in adaptive or stress-induced autophagy (e.g., chemotherapy, proteotoxic stress), contributing to therapy resistance. Its Drosophila ortholog, ADUK, has been implicated in autophagy initiation upon chemotherapy-induced stress 13 , whereas ULK4 is a pseudokinase without clear autophagy function.

Although ULK1 inhibitors have been developed for cancers 14 – 16 , our patient datasets showed no correlation between ULK1 expression and MM progression. In contrast, ULK3 expression strongly correlated with advanced disease stage, suggesting a more relevant role in MM progression. ULK3 has also been implicated in cellular abscission via Beclin-1 17 , and oncogenic signaling through GLI2, regulating anti-apoptotic proteins like BCL-2 18 – 21 . Based on these findings and recent reports 22 , we investigated its functional role in MM autophagy and progression.

This work uncovers the role for ULK3 in driving MM progression and therapy resistance. We show that ULK3 expression correlates with advanced disease and contributes to MM cell survival through autophagy. Developing small-molecule inhibitors with potent activity against ULK3, we demonstrate that pharmacologic blockade not only suppresses tumor growth and preserves bone integrity in vivo but also restores sensitivity to proteasome inhibitors in resistant cells. Importantly, these inhibitors represent the first pharmacologic tools targeting ULK3, providing a foundation for the future development of more selective ULK3-directed therapies. Together, these findings establish ULK3 as a critical regulator of MM autophagy and a promising therapeutic target for both newly diagnosed and relapsed/refractory disease.

Results

ULK3 expression correlates with MM progression

Understanding cell survival programs and/or signaling pathways in MM progression and relapse 23 can reveal therapeutic targets. We leveraged RNA-seq data from CD138 + MM cells obtained from Moffitt patients classified as MGUS ( n = 64), smoldering (SMOL; n = 57), newly diagnosed (NDMM; treatment naïve, n = 207), early relapsed/refractory (ERMM; 1–3 lines of therapy, n = 303) and late relapsed/refractory (LRMM; >3 lines of therapy n = 182) 24 . A set of autophagy-associated genes (identified by Enrichr pathway analysis, KEGG 2021 Human, Reactome 2022; ontologies, GO biological process 2023) was evaluated in our database and correlated with MM progression stages (Fig. 1A ). Average-linkage hierarchical clustering and expression levels of single autophagy genes across MM stages were also assessed 25 . Fig. 1 Elevated ULK3 expression is a hallmark of MM disease progression and connotes worse outcome. A Heatmap representation of the average-linkage hierarchical clustering (correlation-based distance) and the expression levels (measured as median Z-score values) of single genes in the autophagy signature across MM disease stage cohorts (MGUS; n = 64), smoldering (SMOL; n = 57), newly diagnosed (NDMM; n = 207), early relapsed (ERMM; n = 303), and late relapsed MM (LRMM; n = 182). B ULK3 expression in Moffitt MM patients divided in disease stage cohorts. Representation as violin plots. Adjusted p value: (****) p < 0.0001, (***) p = 0.001, (*) p = 0.0349, ordinary one-way ANOVA assuming normal distribution and sphericity using Tukey’s multi-test correction). Exact p values are provided in the accompanying Source Data file. C Representative images of a healthy donor’s bone marrow biopsy core (top) and multiple myeloma patient (bottom) from TCGA program tissue micro arrays (hematoxylin and eosin stain H&E stained, scale bar; 100 μm) and corresponding confocal microscopy images of the k λ light chains + /ULK3 immunofluorescence staining (anti-human k λ light chain Dako F0198, F0199 in green), Alexa Fluor 647 anti-ULK3 [EPR4888] (ab310161 in red), DAPI nuclear staining in blue. IgG controls were used (Millipore Gt X Hu Ig k and λ CAT# AP-502, AP-506). Acquisition by Zeiss 488/647. Arrows represent colocalization of ULK3 protein and anti-human k λ light chain positive cells, resulting in yellow staining. D Immunofluorescence quantification of ULK3 + k λ cells/total cells in biopsies of normal bone marrow ( n = 15) and MM bone marrow ( n = 34) (two-sided unpaired t-test, p = 0.0002). Box-and-whisker plots show the median (center line), 25th and 75th percentiles (box boundaries), and minimum and maximum values (whiskers); all data points are shown. E Immunoblot of ULK1 and ULK3 protein levels in primary isolated CD19 +ve human B-cells from 4 healthy donors (HD1-4), and MM cell lines (U266 and 8266). GAPDH was used as a loading control. F Immunoblotting analysis of the ATG13-FIP200-ULK protein complex upon ATG13 pulldown in parental U266 MM cell lysates. U266 cell lysates were subjected to immunoprecipitation with anti-ATG13 antibody or control IgG, followed by immunoblotting for ULK3 and FIP200. The “supernatant” lane corresponds to the post-IP flow-through fraction (unbound proteins). Molecular marker showed here as mk.

Dissecting our patient dataset, ULK3 expression correlated with disease stage (Fig. 1B ), unlike other ULK kinase family members, including ULK1 and ULK2 (Supplementary Fig. 1A, B ). We did, however, note that ULK4 expression was significantly higher in relapsed patients (Supplementary Fig. 1C ). We validated the increased expression of ULK3 compared to precursor states in independent data sets ( GSE5900 combined with GSE2658 ; Supplementary Fig. 1D–G ) 26 , 27 . However, the lack of LRMM patients prevented independent validation of our ULK4 findings. Of note, other autophagy components (e.g., Beclin-1, RAB7A, and SQSTM1) also increased expression with disease stage, but given ULK3’s potential as an initiator of the program and ULK4’s status as a pseudo kinase, we further interrogated ULK3. Taken together, the gene expression profile prioritized ULK3 due to its significant upregulation among kinases and its tractability as a kinase target for small molecule inhibition.

We observed that ULK3 levels were significantly elevated in MM patient samples compared to kλ light chain positive plasma cells in healthy bone marrow specimens (Fig. 1C, D ). Elevated ULK3 expression was also observed in MM cell lines compared with CD19 + B-cells from healthy donors (HD1-4, Fig. 1E and Supplementary Fig. 1H, I ). ATG13 immunoprecipitation followed by proteomic analysis identified components of the ATG13:FIP200 autophagy subcomplex among ULK3-associated proteins in MM cells, indicating that ULK3 engages core autophagy machinery (Fig. 1F and Supplementary Fig. 2A–C ) 28 . PRM analysis confirmed ULK3, ATG13, and FIP200 with high-confidence peptide matches (Sequest XCorr: ULK3 2.70, FIP200 2.80, ATG13 3.44) and precise precursor mass measurements (±0.1 ppm). Notably, this complex was not detectable in B cells from healthy donors, highlighting a disease-specific interaction (Supplementary Fig. 2D, E ). This interaction was further validated by reciprocal ULK3 immunoprecipitation in U266 cells (Supplementary Fig. 3A–C ).

ULK3 contributes to autophagy regulation in MM

To evaluate whether ULK3 influences autophagy in MM, we used CRISPR-CAS9 to ablate ULK3 in U266 and MM.1S cells (Fig. 2A, B ). Cyto-ID staining revealed a significant reduction in autophagic vesicle formation in ULK3-knockdown cells (Fig. 2C, D ). Independent assessment of autophagic flux using adherent WM1366 cells expressing m-Cherry-LC3B/GFP-LC3B reporters confirmed that ULK3 knockdown impairs autophagy (Fig. 2E–G ), leading to reduced cell survival. Genetic reduction of ULK3 also significantly impaired cell growth over time (Fig. 2H ). Of note, other common kinases upregulated in MM, such as BRD4 and JAK2, are not impacted by ULK3 knockdown (Fig. 2I ). Importantly, re-expression of ULK3 in knockdown U266 cells restored both ULK3 protein levels and autophagic vesicle formation, as measured by CYTO-ID staining (Fig. 3A–E ). Quantification across multiple fields of view showed that ULK3 rescue restored cytoplasmic autophagic vesicles to levels comparable to those in parental cells, confirming that the observed autophagy defects are ULK3-dependent. Fig. 2 ULK3 contributes to regulate autophagy and survival of MM cell lines. A Immunoblotting of wild-type (WT) vs ULK3 knockdown using CRISPR gene editing (KD ULK3 ) in U266 and MM1.S MM cells (24 h from transfection). B Representative immunofluorescence images of ULK3 staining (anti-ULK3 antibody, EPR4888) in U266 non-targeting control (NTC) and U266 ULK3 knockdown (KD ULK3 ) cells, as well as MM1.S NTC and MM1.S KD ULK3 cells. ULK3 is shown in red, and nuclei are counterstained with DAPI (blue). Full-field images are shown with scale bars of 25 μm (U266) and 10 μm (MM1.S). C , D Confocal imaging and flow cytometry quantification of autophagy in U266 and MM1.S multiple myeloma cells and their corresponding KD ULK3 counterparts, measured using the CYTO-ID® Autophagy Detection Kit. Data represent two independent biological replicates generated from two independent CRISPR knockout experiments (U266 NTC vs. U266 KD ULK3 , n = 2 biological replicates, two-sided unpaired t-test, p = 0.0203; MM1.S NTC vs. MM1.S KD ULK3 , n = 2 biological replicates, two-sided unpaired t-test, p = 0.0223). For each biological replicate, the fluorescence signal was quantified by averaging all analyzed fields. Exact p values are shown. E m-Cherry-GFP-LC3B expressing WM1366 melanoma cells were silenced for ULK3 (KD ULK3 ), and the impact on autophagy inhibition was assessed by ( F ) confocal microscopy analysis ( G ). Red puncta indicate ongoing autophagy, while yellow mcherry-GFP-LC3 colocalization indicates autophagy inhibition since GFP is not degraded. Each plotted value represents the median of three independent experiments ( n = 5 fields/fluorodish, two-sided unpaired t test, p = 0.0267). Quantification was as follows: normalized colocalization: (sum red and green overlap area (µm 2 )/cell count × 100/average red area per cell). Median is annotated; asterisks denote statistical significance. For immunoblots, numbers indicate molecular weight in kDa, and actin was used as a loading control. H MM cell growth curves of parental vs. KD ULK3 counterparts (U266 NTC vs. U266 KD ULK3 and MM1.S NTC vs. MM1.S KD ULK3 ). Cell counts were determined by trypan blue dye exclusion over 12 days post transduction. Each dot represents the mean value ( n = 3) of two biological replicates. Bars represent mean values, and error bars indicate SD. U266 two-way mixed ANOVA, adjusted p values: *** p = 0.0009; **** p < 0.0001). MM1.S two-way mixed ANOVA, adjusted p values: ** p = 0.005). Statistical significance was determined using repeated-measures two-way ANOVA with Šidák’s multiple comparisons test. All tests were two-sided. I Immunoblotting of no template control (NTC) vs ULK3 knockdown using CRISPR gene editing (KD ULK3 ) in U266 cells (24 h from transfection). BRD4 and JAK2 levels are not impacted by ULK3 knockdown (KD1 and KD2). Of note, representative confocal images from one of three independent experiments with similar results are shown. Fig. 3 Rescue experiment demonstrating ULK3-dependent autophagy phenotype in U266 cells. A Immunoblot analysis of ULK3 protein expression in parental U266 cells, ULK3 knockdown cells generated using a CRISPR/Cas9 system (KD1 ULK3 NTC), and ULK3 re-expression in KD1 ULK3 NTC cells (OE1 ULK3 ), confirming successful ablation and rescue of ULK3 expression. B Quantification of cytoplasmic autophagic vesicles labeled with Cyto-ID by confocal microscopy, measured as the mean CYTO-ID fluorescence intensity per area, where each area represents an independent field of view ( n = 5 technical replicates). Across the five analyzed fields of view, a total of 595 U266 parental cells, 442 ULK3 knockdown cells, and 517 OE1 ULK3 cells were quantified. The experiment was independently repeated three times with similar results. C – E Representative confocal images of each cell line acquired using a Leica SP8 confocal microscope. Of note, representative confocal images from one of three independent experiments with similar results are shown. Scale bars were added during figure assembly but correspond to the original image calibration. Insets show magnified regions highlighting cytoplasmic CytoID-positive vesicles.

Collectively, these results demonstrate that ULK3 contributes to autophagy regulation and promotes MM cell survival, establishing it as a critical node in MM pathophysiology.

Development of small-molecule kinase inhibitors targeting ULK3

Given ULK3’s role in MM autophagy and the lack of selective inhibitors, we developed multi-kinase compounds targeting ULK3 alongside other kinases and BET proteins, such as BRD4 29 , 30 . Accordingly, our findings support a role for ULK3 but may reflect combined target inhibition; selective inhibitors are currently in development for definitive validation. Compound SG3-014 (SG3; Supplementary Fig. 4A ) inhibits BRD4, JAK2 and ULK3 with IC 50 of 9.3 ± 6.2, 4.6 ± 0.4 and 204 ± 45 nM, respectively (Supplementary Fig. 5A ). Inhibition of ULK1 was considerably weaker (IC 50 = 539 ± 103 nM). Of note, BRD4 is highly expressed in MM 31 – 33 , and we validated these observations in our dataset (Supplementary Fig. 4B ).

BRD4 regulates MYC and is inhibited by JQ1, but resistance emerges rapidly 34 – 36 . We therefore hypothesized that dual ULK3/BRD4 inhibition could overcome this. SG3 rapidly reduced ULK3/ULK1 levels (6 h), altered early autophagy proteins (e.g., Beclin-1), and suppressed autophagic flux and cell survival effects not seen with JQ1 (Supplementary Fig. 4C–H ). Importantly, treatment with a generic receptor tyrosine kinase (RTKs) inhibitor like SU6668 (0–3 μM) does not impact ULK3 protein levels in U266 MM cells (6 h) (Supplementary Fig. 4D ). Additionally, SG3 inhibited osteoclastogenesis without affecting MSCs or osteoblasts (Supplementary Fig. 4I ), suggesting potential protection against MM-induced bone disease. Thus, we hypothesized that these inhibitors could protect against cancer- induced bone disease.

In preparation for in vivo studies, in vitro hepatocyte microsomal stability assays revealed a short half-life for SG3, prompting the development of MA9-060 (MA9) (Fig. 4A and Supplementary Fig. 5A, B ) 37 . The molecular binding mode of MA9:ULK3 was determined by X-ray crystallography. The structure indicates that MA9 is a type-1 inhibitor that engages the hinge Cys94 residue via hydrogen bonding to its pyrimidine N1 and 2-anilinyl NH groups (Supplementary Fig. 5C ), as unequivocally confirmed by the polder map and the 2fo-fc map (Supplementary Fig. 6A, B ). Structural superposition of the MA9 complex with those of botsutinib (6FDY) and momelotinib (6FDZ) highlights conformational changes in the alpha C helix, the activation loop, and the DFG motif (Supplementary Fig. 6C–F ), confirming that the MA9:ULK3 structure indeed adopts the typical inactive DFG-in conformation. Like SG3, MA9 treatment of U266 MM cells also provoked rapid reductions in ULK3 levels and autophagy effectors (Fig. 4B ). Notably, U266 cells express little to no MYC, but we validated the impact of MA9060 on ULK3 and MYC in the MM.1S cell line and showed MA9060 was not toxic to normal primary human MSCs in the nM concentration (Supplementary Fig. 5E–H ). We also demonstrated that MA9 treatment was comparable to CQ in halting autophagy (Fig. 4C and Supplementary Fig. 7 ), whereas JQ1 had no effect (Fig. 4C, D ). Importantly, CQ inhibition of autophagic flux results in the accumulation of large vesicles, whereas MA9 appeared to inhibit the proper formation of autophagolysosomes. As with SG3, MA9 treatment rapidly compromised cell survival in vitro, whereas ruxolitinib, a JAK2 inhibitor control, had no effect (Fig. 4E ). These results suggest that MA9’s mechanism of action likely involves autophagy blockade in the U266 model. Fig. 4 Anti-myeloma activity of the MA9-060 autophagy inhibitor. A Chemical structure of the autophagy inhibitor MA9-060. B Effects of MA9-060 (1 µM) and JQ1 (10 µM) (6 h treatment) on the steady state levels of autophagy proteins in U266 MM cells. Chloroquine (CQ) (1 µM) served as a positive control for autophagy inhibition. C Representative confocal images (scale 1–75 μm) and flow cytometry quantification ( D ) of autophagy in U266 MM cells, treated with MA9-060 (1 μM), JQ1 (1 μM), or CQ (1 μM), measured by a cationic amphiphilic tracer (CAT) autophagic vacuoles dye (biological replicates n = 4) (CYTO-ID® Autophagy detection kit). Hoechst 33342 dye appears here as the nuclear blue staining (ordinary one-way ANOVA assuming normal distribution and sphericity using Tukey’s multi-test correction, adjusted p value * p = 0.0354, ** p = 0.0081, and *** p = 0.0009). Bars represent mean values, and error bars indicate SD. Exact p -values are provided in the accompanying Source Data file. E MTS assay demonstrating the inhibitory profile of MA9-060 (10 −4 –10 −9 M) on MM cells (5TGM1, 8226, and U266). Data represent the mean of technical replicates. Each compound was assessed in three independent experiments. Concentration ranges were optimized for individual experiments while producing consistent dose–response trends.

MA9 significantly reduces MM growth and myeloma-associated bone disease in vivo

To evaluate MA9 in vivo while avoiding MYC-related confounding, we used the U266 cell line, which homes to bone in NSG mice 38 . Following bioluminescence-confirmed engraftment, mice were randomized on day 20, including a JQ1 arm to assess BRD4 inhibition.

MA9 dosing was based on that of JQ1 34 , 39 . Our results demonstrate that MA9 treatment significantly impaired tumor growth compared to control (Fig. 5A, B ). At endpoint, MA9 had nearly double the median survival (112 days) compared to vehicle (70 days) (Fig. 5C, D ), outperforming other treatment arms, including JQ1 + CQ (Fig. 5D ). Importantly, there were no overt toxicities (weight loss, lethargy) with the applied dosing regimen (Fig. 5E ). Consistent with in vitro results, MA9 significantly decreased ULK3 levels in κ λ light chain-expressing MM cells (Supplementary Fig. 8A, B ). Moreover, MA9 reduced MM proliferation, although no impact on apoptosis was observed at the study endpoint, when tumor burdens were similar across groups (Fig. 5F–I ). X-ray and histological analyses revealed that MA9 significantly protected against MM-induced osteolysis (Fig. 6A–D ). Although osteoclast numbers were higher, this reflected severe trabecular bone loss in controls (Fig. 6E, F ). Overall, MA9 exhibits strong anti-myeloma activity and preserves bone integrity. Fig. 5 MA9-060 has potent in vivo anti-myeloma activity. A Representative bioluminescent imaging (BLI) of the indicated cohorts of mice ( n = 10/group), tail vein injected with 1 × 10 6 U266-Luc MM cells and treated with vehicle control ( β -HCD Captisol) or MA9-060. Treatment started at day 20 (M-F, 10 mg/kg/day subcutaneously). B Bioluminescence quantification of tumor burden over 100 days post inoculation (VEH vs MA9060, two-way mixed ANOVA, with Bonferroni’s multi-test correction, adjusted p = 0.0073). Graph indicates mean, and error bars represent the SEM. C MA9-060 treatment improves overall survival (median 111 days post-tumor inoculation) compared to vehicle control (CTRL, median 72.5 days). Of note, MA9-060 treatment did not show any sign of toxicity (weight loss <20%, ruffled hair) at the chosen dose. D Table indicating median survival (days) and significance of VEH CTRL vs treatment groups (log rank Mantel Cox test, p < 0.0001). E Weight monitoring (weekly) of the indicated cohorts ( n = 10). Representative confocal microscopy images ( F ) and quantification ( G ) of formalin-fixed, paraffin-embedded tumors (5-micron sections), stained for Ki67 and human MM cells (hMITO) colocalization. Alexa Fluor Plus 647 goat anti-rabbit and Alexa Fluor Plus 488 goat anti-mouse were used as secondary antibodies. Confocal immunofluorescent images were acquired with Leica TCS SP8 system with 63× objective (Leica) and quantified as the number of Ki67+/hMITO+ cells, normalized to the total nuclei number of analyzed area (two-sided unpaired t-test, p = 0.0003, scatter dot plots, mean with SD). Exact p -values are provided in the accompanying Source Data file. Representative confocal microscopy images ( H ) and quantification ( I ) of formalin-fixed, paraffin- embedded tumors (5-micron sections, CTRL, n = 6 biological replicates; MA9060 n = 9 biological replicates), stained for CC3 and human MM cells (hMITO) colocalization. Alexa Fluor Plus 647 goat anti-rabbit and Alexa Fluor Plus 488 goat anti-mouse were used as secondary antibodies. I Confocal immunofluorescent images were acquired with the Leica TCS SP8 system with 40× objective (Leica) and quantified as the number of CC3+/hMITO+ cells, normalized to the total nuclei number of analyzed area (two-sided unpaired t-test, p = 0.2915, scatter dot plots, mean with SD). Of note, representative images are shown from biological samples representative of the indicated experimental group (CTRL, n = 6 biological replicates; MA9060 n = 9 biological replicates). Fig. 6 MA9060 treatment protects from bone-associated MM disease. Ex vivo quantification of MM associated bone destruction, measured in untreated mice (VEH) vs. treatment groups (MA9-060, JQ1, CQ, and JQ1 + CQ cohorts). A Representative cortical osteolytic lesions imaging and quantification ( n = 10/treatment group), that were measured ( B ), by Faxitron X-rays (multiple comparisons ANOVA, Bonferroni’s adjusted P value (*** P = 0.0002, **** P < 0.0001). Exact p -values are provided in the accompanying Source Data file. C Histological analyses of bone sections (H&E staining, ( n = 10/ treatment group), and D relative quantification of trabecular bone (multiple comparisons ANOVA, Bonferroni’s adjusted P value **** P < 0.0001). E Histological analyses of osteoclast activity (TRAcP staining, n = 10/treatment group; indicated by arrows), and F , quantification of osteoclast number (ordinary one-way ANOVA assuming normal distribution and sphericity using Tukey’s mult-test correction, adjusted p value * p = 0.0246, ** p = 0.0013, *** P = 0.0001, and **** p < 0.0001). Of note, the MA9-060-treated group shows higher actively resorbing osteoclast number on trabecular bone, that is not observed in the vehicle cohort, since the aggressive bone marrow MM destroyed the internal bone structure. Graphs represent scatter dot plots, mean with SD. Of note, representative images are shown from biological samples representative of the indicated experimental group (CTRL, n = 6; MA9060 n = 9, JQ1 n = 9, CQ n = 10, JQ1 + CQ n = 9 biological replicates).

To further demonstrate the importance of ULK3 in MM progression, we inoculated mice with wildtype and KD ULK3 cells (Fig. 7A, B ). Our data show that loss of ULK3 resulted in a delayed onset of disease and significantly higher overall survival (Fig. 7B, C ). We also demonstrated the selectivity of MA9 for ULK3 (Fig. 7D ), showing that KD ULK3 MM-bearing mice received no further benefit in terms of overall survival (Fig. 7E, F ). These findings collectively highlight ULK3 as a functionally relevant therapeutic target in MM and support the efficacy and specificity of MA9 in vivo. Fig. 7 ULK3 ablation suppresses MM cell growth. A Immunoblotting for ULK3 expression in parental, NTC CTRL and KD ULK3 U266 cells. Of note, NTC CTRL and KD ULK3 were used for the in vivo KD ULK3 model. Densitometric quantification values are annotated on the blot. B Mice ( n = 20/group) were tail vein injected with 1 × 10 6 NTC CTRL and KD ULK3 U266 MM cells. Mice injected with KD ULK3 U266 MM cells had a delayed onset of the disease compared to NTC CTRL . Statistical significance between survival curves was determined using the log-rank (Mantel–Cox) test ( p < 0.0001). Exact p -values are provided in the accompanying Source Data file. C Representative confocal microscopy images of formalin fixed, paraffin embedded tumors (5-micron sections), stained for human mitochondria (hMITO) as a readout of MM cells colocalization. Alexa Fluor Plus 488 goat anti-mouse were used as secondary antibody. Of note, representative confocal images from one of three independent experiments with similar results are shown. Scale bars were added during figure assembly but correspond to the original image calibration. D MTS showing the inhibitory profile of MA9060 on NTC CTRL and KD ULK3 for U266 cell line. Treatment with MA9060 (0–100 μM, n = 3) for 48 h show reduced activity of MA9060 in KD ULK3 cell lines. E Mice were divided respectively in two groups, vehicle control ( n = 10, β -HCD Captisol) or MA9-060 ( n = 10). Treatment started at day 20 (M-F, 10 mg/kg/day subcutaneously). MA9-060 treatment improves overall survival in the NTC CTRL arm (median 41 days post-tumor inoculation) compared to vehicle control (CTRL, median 36 days). Statistical significance between survival curves was determined using the log-rank (Mantel–Cox) test ( p = 0.0001). F MA9-060 did not affect the overall survival of the group of mice injected with KD ULK3 U266 MM cells (median 91 days for KD ULK3 CTRL vs 87 days for MA9-060 treatment). MA9-060 treatment did not show any signs of toxicity (weight loss <20%, ruffled hair) at the chosen dose. Statistical significance was not achieved between survival curves, as determined using the log-rank (Mantel–Cox) test ( p = 0.07).

Drug-resistant MM cells express increased levels of ULK3 and are sensitive to MA9

Proteasome inhibitors (PI), such as bortezomib (BTZ) and carfilzomib (CFZ), are standard treatments for MM, but resistance often develops 39 . Interestingly, analysis of Moffitt RNA-Seq data showed a significant increase in ULK3 expression in MM patients who had failed multiple lines of therapy, including PIs, compared with those who were newly diagnosed (Fig. 8A ). Fig. 8 MA9-060 shows potency as a monotherapy and synergizes with proteasome inhibitors (BTZ) in vitro. A ULK3 expression in divided in NDMM ( n = 182) vs. post-proteasome inhibitor treatment ( n = 27) in the Moffitt PMRC MM Patient Dataset. Asterisks denote statistical significance (two-sided unpaired t-test, p = 0.005). Exact p -values are provided in the accompanying Source Data file. B Representative confocal microscopy images of a NDMM bone marrow biopsy core ( left ) and RRMM ( right ) from the Moffitt TCGA program tissue microarrays (H&E stained, scale bar; 100 μm), (120 cores, 40 cases, 10 normal bone marrow tissue cores, 3 MM cell lines as CTRL, 0.6-mm biopsies) generated at the H. Lee Moffitt Cancer Center. k λ light chains +/ ULK3+ immunofluorescence staining (k λ light chain in green and ULK3 in red), DAPI nuclear staining is in blue. Control IgGs were also used (Millipore Gt X Hu Ig k and λ CAT# AP-502, AP-506). Arrows represent colocalization of ULK3 protein and anti-human k λ light chain positive cells (scale bar: 0–75 μm). C Quantitative analysis of TMAs (NDMM, n = 14 and RRMM, n = 11). Box-and-whisker plots show the median (center line), 25th and 75th percentiles (box boundaries), and minimum and maximum values (whiskers); all data points are shown. Asterisks denote significance (two-sided unpaired t-test, p = 0.0098). D ULK3 expression in proteasome inhibitor (bortezomib) sensitive (U266, 8226) vs. resistant (U266-PSR, 8226-B25) MM cell lines. E MTS assays of the indicated MM cell lines following treatment with MA9-060 after 48 h, expressed as IC 50 . Data represent the mean of three technical replicates ( n = 3) from one representative experiment. The experiment was independently repeated three times with similar results. CQ is here used as reference CTRL for autophagy flux inhibition. F Proteasome inhibitor (bortezomib, BTZ) sensitive (U266, ANBL6) and resistant counterpart (U266-PSR, ANBL6-V10R) MM cells were treated with BTZ (0–100 nM) in combination with fixed doses of MA9-060 (50, 100, and 500 nM, n = 3/treatment condition). Data represent the mean of three technical replicates ( n = 3) from one representative experiment. The experiment was independently repeated three times with similar results. Bluer colors indicate synergism at the defined doses. Synergy data is mapped as dose-response D-R (LOEWE) with Combenefit Software.

Tissue microarray (TMA) analysis confirmed elevated ULK3 protein in relapsed patients and PI- resistant MM cell lines (Fig. 8B–D ) 40 – 42 . Moreover, in vitro cytotoxicity assays revealed that PI-resistant MM cells were highly sensitive to MA9 compared to CQ (Fig. 8E ). We also noted that the combination of MA9 with BTZ could significantly enhance MM toxicity in PI-resistant cell lines, which is in keeping with reports showing roles for autophagy in facilitating proteasome inhibitor resistance (Fig. 8F ) 43 – 51 .

MA9 displays potent anti-myeloma activity against patient-derived MM cells ex vivo

For clinical robustness, we tested the activity of MA9 in MM patient samples using our ex vivo mathematical myeloma advisor (EMMA) platform (Supplementary Table 1 ) 6 , 24 , 52 , 53 . EMMA allows drug testing as single therapies or in combination, on patient-derived CD138 + selected MM cells cultured with collagen I and patient-derived bone marrow stroma over 6 days using live-cell imaging. An area under the curve (AUC) is calculated up to 96 h for each dose, and results are displayed as a mean efficacy AUC. Using this approach, we noted that MA9 treatment potently reduced the viability of patient-derived MM cells (Fig. 9A ). Parallel samples from 15 MM patients exposed to different chemotherapeutics showed that MA9 is a potent single agent, even outperforming the duplet JQ1 + CQ combination. Further, MA9 showed comparable potency versus PI (CFZ) across the same MM patient specimens (Fig. 9B ). These findings were independently validated in a second cohort of MM patients ( n = 36), demonstrating that both MA9 and CFZ treatment were similarly potent in compromising MM cell survival, and that the combination of MA9 with CFZ deepened the cytotoxic response, in a manner comparable to the triplet treatment of JQ1 + CQ + CFZ (Fig. 9C and Supplementary Tables 2 , 3 ). Moreover, comparing the MA9 response in NDMM ( n = 18) and RRMM ( n = 27) patients revealed that MA9 was effective as single agent but, when combined with CFZ, significantly improved cytotoxicity in both groups (Fig. 9D, E ). A waterfall plot showed CFZ combined with MA9 yielded an efficacy benefit over CFZ alone in 88% of NDMM and 74% of RRMM patients (Fig. 10A, B ). We also demonstrated MA9 potency in provoking the rapid suppression of ULK3 within a single patient specimen (Fig. 10C ). Fig. 9 MA9-060 treatment compromises the survival of patient NDMM and RRMM cells ex vivo. A Representative viability (live cell imaging) response of CD138 + cells from NDMM patient (Pt620) cultured and tested in autologous stromal microenvironment in EMMA’s ex vivo platform, exposed to three different concentrations of MA9-060 (0.7, 2.2, 6.7 μM, n = 3/ treatment condition). Of note, no toxicity was observed for co-cultured stromal cells, even at the highest concentration of MA9-060. (not shown here—20 μM). B Ex vivo EMMA platform assay to quantify the chemosensitivity of primary CD138 + MM cells plated in a collagen matrix with bone marrow stroma and patient plasma. Tumor cells were treated with carfilzomib (CFZ), MA9060, JQ1, CQ and JQ1 + CQ and efficacy (live cell imaging) was measured as area under the curve (AUC % to vehicle control) for 96 h (EVOS FL Auto) ( n = 15 patients). Exact p -values are provided in the accompanying Source Data file. C The duplet combination therapy of MA9-060 and CFZ was tested vs. the triplet JQ1 + CQ + CFZ combination, showing equal efficacy in MM patients ( n = 36). Combination therapy of MA9-060 and CFZ is highly efficacious in paired analysis of NDMM ( n = 18) ( D ) and RRMM ( n = 27) ( E ) patients. B – E Statistical differences between treatment groups were evaluated using a repeated measures (paired) one-way analysis of variance (ANOVA), assuming sphericity and a normal distribution. To evaluate differences between all possible pairs of columns, Tukey’s multiple comparison test was applied post hoc. To maintain control over the experiment-wide false-positive rate, multiplicity-adjusted p -values were computed directly using the studentized range distribution based on the total number of treatment groups and the residual degrees of freedom. Pairwise comparisons were considered statistically significant when the adjusted p -value was less than 0.05. Fig. 10 Patient-derived multiple myeloma samples reveal ULK3-associated response to MA9-060 and combination treatment strategies. A , B NDMM (F) and RRMM (G) are represented as single patient response to combination of MA-9060 + CFZ. The contribution effect of MA9060 + CFZ treatment over CFZ single agent independent treatment (0) is quantified and displayed as positive (blue bars) (NDMM, n = 18; RRMM, n = 20) or negative (red bars) (NDMM, n = 2; RRMM, n = 7), AUC % difference MA9060 + CFZ vs. CFZ. C Immunoblot showing levels of indicated proteins in CD138 + cells isolated from RRMM patients (Pt971, Pt1050). Pt1050 CD138 + cells were assessed for ULK3-mediated autophagy related proteins (ATGs), following treatment with MA9-060 (1 μM at 2 and 6 h). D RNA-seq data of Moffitt MM patients ( n = 44, 19 NDMM, 25 RRMM) divided in ULK3 high and ULK3 low (≤ median) expression (Z-score) and correlation to MA9-060 treatment response (AUC) (two-sided unpaired t-test, p = 0.0042; two-sided Wilcoxon rank-sum test p = 0.0182). Exact p -values are provided in the accompanying Source Data file. E RNA-seq data of Moffitt MM patients ( n = 44, 19 NDMM, 25 RRMM) divided in BRD4 high and BRD4 low (≤ median) expression (Z-score) and correlation to MA9-060 treatment response (AUC) (two-sided unpaired t-test p = 0.4331). Box-and-whisker plots show the median (center line), 25th and 75th percentiles (box boundaries), and minimum and maximum values (whiskers); all data points are shown. F RNA-seq data of Moffitt MM patients ( n = 44, 19 NDMM, 25 RRMM) divided in ULK1 high and ULK1 low (≤ median) expression (Z-score) and correlation to MA9-060 treatment response (AUC) (two-sided unpaired t-test p = 0.0418).

To assess whether ULK3 expression correlates with MA9 efficacy, we analyzed paired RNA-seq data from EMMA samples and found that patients with above-median ULK3 levels ( n = 44; 19 NDMM, 25 RRMM) showed significantly better responses to MA9 (Fig. 10D and Supplementary Table 4 ), consistent with ULK3 inhibition. In contrast, MA9 efficacy did not correlate with BRD4 expression (Fig. 10E ). Additionally, patients with low ULK1 expression responded better than those with high ULK1 (Fig. 10F ). Together, these data support ULK3 targeting as a promising strategy in both newly diagnosed and relapsed MM.

Discussion

Our findings identify ULK3 as a kinase that contributes to regulating autophagy in MM, cell survival and disease progression, highlighting its potential as a therapeutic target. Our functional studies indicate that ULK3 contributes to autophagy regulation in multiple myeloma, as its knockdown impairs autophagic flux and cell survival. At the same time, re-expression restores autophagy and growth, establishing ULK3 as a critical regulator of MM cell homeostasis.

An important limitation is the lack of selective ULK3 inhibitors, which prevent attributing observed effects exclusively to ULK3. While our data support its involvement, definitive validation will require selective compounds and comprehensive kinase profiling. Using the small-molecule inhibitor MA9, we show that ULK3 activity is suppressed at nanomolar concentrations, reducing MM tumor burden, preserving bone integrity, and improving overall survival in vivo. While this suggests a role for ULK3, we cannot exclude contributions from additional targets of the compound. Importantly, these findings are recapitulated in primary CD138 + MM cells, supporting clinical relevance.

Mechanistically, ULK3 substrates and signaling remain undefined. Rapid ULK3 loss upon inhibitor treatment suggests potential proteasomal regulation, linking it to the ubiquitin-proteasome system and autophagy crosstalk. This mechanism remains unclear and will require dedicated experiments to clarify whether MA9 induces ULK3 degradation or alters its stability indirectly. Notably, ULK1 does not compensate for ULK3 loss, indicating that ULK1 and ULK3 have non-redundant functions 13 . It should be noted that ULK3 has also been implicated in other important aspects of cellular biology, including acting as a quality control mechanism for cellular abscission and cell cycle progression 17 , 54 – 58 . Other reports have linked ULK3 to the expression of oncogenes, such as GLI2 59 , which induces the expression of anti-apoptotic proteins, again tying in with the role of autophagy in cancer cell survival 60 .

Therapeutically, targeting autophagy has been limited by non-specific agents, such as chloroquine, with poor tolerability 61 , 62 . More precise targeting is needed. Unlike ULK1/2, which respond to nutrient deprivation, ULK3 appears to be activated under therapeutic stress 13 , a relevant context in MM. Its association with ATG13 and FIP200 further supports a role in early autophagy regulation.

Genetic silencing of ULK3 significantly reduces autophagy in MM despite changes in ULK1 expression, again underscoring a lack of functional redundancy amongst ULK family members. Multi-kinase inhibitors used herein were originally intended to block both JAK2 and BRD4. Using the JAK2 inhibitor ruxolitinib as a starting point, SG3 and MA9 were found to retain their JAK2 inhibition profiles 29 . Notably, while SG3-014 corresponds to compound 2 described in Ember et al., MA9-060 (MA9) has not been reported in that study (US Patent 12,391,676). However, the MM cell lines tested here were not affected by the JAK2 inhibitor ruxolitinib, indicating the primary mechanism of action of SG3 and MA9 in compromising MM cell survival via inhibition of ULK1/3 or BRD4. JQ1 inhibition of BRD4 did not impact autophagy in a variety of assays performed.

In vivo analysis using the U266 cell line, characterized by negligible expression of the MYC oncogene, showed that MA9 outperformed JQ1 by extending median overall survival, suggesting that ULK3/ULK1 blockade is responsible for the improved efficacy. Notably, MA9 showed similar performance compared with the JQ1 plus CQ combination in vivo, but MA9 has the advantage of being administered as a single agent, thus avoiding potential pharmacokinetic/pharmacodynamic issues associated with the administration of multiple reagents. Of note, while we did observe reduced MM proliferation in the MA9 treatment arm, we only observed a trend towards increased apoptosis.

We posit that this may be due to clonal selection in vivo, and we would expect that removing the mice from the study at earlier time points would reveal greater apoptosis in the MA9 group than in the vehicle control. A caveat of our MA9 studies is that it also inhibits ULK1, albeit it at a higher IC 50 than ULK3 (227 ± 46 nM for ULK3 vs. 1046 ± 472 nM for ULK1). Thus, we cannot rule out that some of the observed in vivo effects are due to ULK1 inhibition. The ULK3 kinase domain crystal structure has recently been published and should enable the continued development of more specific ULK3 inhibitors 63 . In vivo studies with ULK3-ablated MM cells showed delayed disease progression and prolonged survival, demonstrating ULK3’s role in MM. In contrast, resistance to MA9 in KD ULK3 cells confirmed the inhibitor’s on-target activity and therapeutic potential.

Finally, MA9 shows activity across heavily pretreated patient samples, suggesting potential to overcome resistance. Given emerging links between autophagy and resistance to immunotherapies such as CAR-T 64 , targeting ULK3 may have broader therapeutic impact.

With the crystal structure of the ULK3 kinase domain having been resolved, our studies should initiate the development of selective ULK3 inhibitors that will be highly beneficial for the treatment of MM patients across the disease spectrum.

In summary, our data highlight ULK3 as a key mediator of MM autophagy regulation and cell survival. Targeting ULK3 represents a promising therapeutic strategy, particularly in resistant disease settings, and provides a foundation for future drug development efforts focused on autophagy-dependent cancers.

Methods

Study design

To uncover the role of ULK3-mediated autophagy in multiple myeloma, we used primary cells, TMA samples from patients and healthy donors, and preclinical validation in mice and cell lines.

Ethical considerations

The utilization of CD19+ B cells from donors at H Lee Moffitt Cancer Center was conducted under the Institutional Review Board (IRB) under protocol #: Pro00021733. Written informed consent was obtained from all donors. Consent was also obtained for the reporting and sharing of de-identified individual-level data. Similarly, CD138 + primary cells were sourced from multiple myeloma patients with written informed consent due to their enrollment in protocols MCC14690 and MCC18608 at H. Lee Moffitt Cancer Center and Research Institute. MCC14690 (Total Cancer Care—TCC) is designed to develop personalized medicine by collecting biological samples (blood, bone marrow aspirates, and core biopsies) and annotating clinical records for translational research. MCC18608 is an institutional protocol specifically dedicated to multiple myeloma research and tissue banking. It allows researchers to study drug sensitivities, cytogenetics, and tumor cell mutational profiles without requiring patients to provide additional samples beyond their clinically indicated procedures. These protocols were conducted in accordance with the ethical standards outlined in the Declaration of Helsinki, the International Ethical Guidelines for Biomedical Research Involving Human Subjects (CIOMS), the Belmont Report, and the U.S. Common Rule, ensuring adherence to patient privacy and ethical treatment.

For RNAseq data from multiple myeloma patients, patient data were handled in accordance with rigorous privacy measures as medical records were deidentified. Only clinically relevant information was scrutinized, including (A) details of administered treatments (therapeutic agents, doses, and schedule) preceding the biopsy, (B) cytogenetic information, and (C) results from serum and urine electrophoresis. This approach ensures both the ethical treatment of participants and the meticulous examination of pertinent clinical details in compliance with established ethical norms and regulations.

Animal experimentation

All animal experiments were performed with approval from the Moffitt/University of South Florida Institutional Animal Care and Use Committee (Protocol #10955R, Title: Role of ULK3 in Sensitive and Refractory Multiple Myeloma). Mice were maintained in a specific pathogen-free facility under a 12 h light/12 h dark cycle at 20–24 °C and 40–60% relative humidity, with food and water provided ad libitum. For in vivo validation of MA9 efficacy (Fig. 5 ), we initially included n = 10 mice per treatment group: suboptimal tail vein injection resulted in exclusion criteria. Randomization based on sex, weight, and initial tumor burden was conducted, followed by treatment with either vehicle control ( n = 6), single agents MA9 ( n = 9) and JQ1 ( n = 9), CQ ( n = 10), or the combination of JQ1 and CQ ( n = 9). Tumor progression was tracked via bioluminescence: endpoint was paralysis or >20% weight loss. Experiments were blinded. For the NTC CTRL vs ULK3 KD U266 in vivo model, mice ( n = 10/ group) received 1 × 10 6 cells via tail vein (Fig. 7 ). For MA9 therapy efficacy, mice were treated with vehicle ( n = 10) or MA9 (10 mg/kg/day, subcutaneously, M-F from day 20; n = 10) (Fig. 7E, F ). Survival and toxicity (weight, clinical signs) were monitored.

Study participants

Peripheral blood and bone marrow (BM) samples were obtained from IRB-consented patients (MCC14690 and MCC18608). Specimens were handled per the Declaration of Helsinki, the International Ethical Guidelines for Biomedical Research Involving Human Subjects (CIOMS), the Belmont Report, and the U.S. Common Rule. See Supplementary Table 1 —Patient Demographics for patient demographics.

CD138 + patient cell isolation and purity

Bone marrow (BM) aspirates (20 ml) were collected, and bone marrow mononuclear cells (BMMCs) were isolated using standard procedures. BMMCs were enriched for CD138 using Miltenyi (Bergisch Gladbach, Germany 130-051-301) antibody-coated magnetic beads and cytospin slides of the CD138 + cells were prepared. Post-isolation, the average purity was 95.3% (IC: 88.1% 100%) 65 .

Cell culture

Luciferase-labeled MM cells 5TGM1-Luc (RRID: CVCL_VI66) and U266-Luc (RRID: CVCL_0566) were obtained from the University of Texas, Health Science Center at San Antonio, TX (2012) and the University of Virginia, VA (2014), respectively. MM1.S cell line was obtained in 2015 from ATCC (RRID: CVCL_8792; ATCC CRL-2974). Isogenic PI-sensitive and resistant pairs U266 (RRID: CVCL_0566; ATCC TIB-196) and PSR 41 , RPMI8226 (RRID: CVCL_0014;

ATCC CCL-155) and B25 45 were obtained from Dr. Steven Grant’s laboratory (VCU-Massey), and ANBL-6 (RRID: CVCL5425) and V10-R 66 pairs were obtained from Dr. Kenneth Shain’s laboratory (Moffitt Cancer Center). PI-resistant cell lines were maintained by bi-weekly dosing with PI (bortezomib, 30 nM). Human MM.1S, U266, RPMI 8226 MM cells (all from the ATCC, USA) were routinely cultured at 37 °C in Roswell Park Memorial Institute (RPMI) 1640 medium supplemented with 10% heat-inactivated FBS (PEAK, CAT # PK-FBS1) and penicillin/streptomycin (hereafter, complete medium).

Cells were not used for 5 days after PI-treatment. All myeloma cell lines were maintained below 1 × 10 6 cells/ml in complete medium and used within 20 passages. IL-6-dependent cell lines ANBL-6 and V10-R are cultured in complete medium containing 1 ng/ml of IL-6 (R&D Systems, HZ-1019). Isolation and culture methods for mesenchymal stem cells (MSCs) were adapted from previously published protocols 67 . Briefly, hind limbs were collected in sterile PBS from 4 to 6-week-old male C57BL/6 Rag2 −/− tumor-bearing mice. Following removal of muscle tissue, epiphyses were removed, and bone marrow flushed three times with sterile PBS to deplete the hematopoietic cells. Flushed bones were then cut into 1–3 mm chips, digested with 1 mg/mL collagenase II (Invitrogen) in α-MEM with 15% FBS, and shaken at 150 RPM for 1 h at 37 °C.

Digested bone fragments were grown in 6-well tissue culture plates in α-MEM with 15% FCS. Medium was changed every 3 days 68 . Bone marrow cells were harvested by flushing the tibiae of 6–8-week-old male and female C57BL/6 Rag2 −/− mice. Cells were cultured for 3 days in α-MEM containing 25 ng/mL M-CSF (CAT# 315-02, PeproTech). The resulting non-adherent cells were then differentiated into osteoclasts by culture in osteoclastogenic medium supplemented with 100 ng/mL RANKL (CAT# 47187000, OYC Americas) and 20 ng/mL M-CSF for 7 days 69 . Cells used in these studies routinely tested negative for mycoplasma by PCR (polymerase chain reaction) (Bulldog Bio, CAT # 25233) and were additionally authenticated against ATCC, DSMZ or ExPASy STR profiles. WM1366 cell line stably transfected with m-cherry-GFP-LC3 was a kind gift of Dr. Keiran Smalley (H Lee Moffitt Cancer Center) and was generated and selected as previously described 70 .

RNA Sequencing (RNA-Seq)

After CD138 + cell enrichment, vials containing 1.0 × 10 6 viable frozen CD138 + cells were sent for RNA-Seq analysis.

Nucleic acid extraction

RNA extraction from frozen tissue was performed using the Qiagen RNeasy Plus Mini Kit, resulting in an average insert size of 216 bp.

RNA Sequencing

The Illumina TruSeq RNA Exome protocol was employed, involving single library hybridization, cDNA synthesis, library preparation, and sequencing with either 100 or 150 bp paired end reads, achieving a total coverage of 100 million reads or 50 million paired reads.

The RNA-Seq data were analyzed using the Tumor Pipeline, following the workflow below with the GRCh38/hg38 human genome reference and GenCode build version 32.

Adapter trimming

Adapter sequences were removed from the raw sequencing FASTQ files. This process involved k-mer matching for adapter trimming, along with quality trimming, contaminant removal, sequence masking, GC content filtering, length filtering, and entropy filtering. The processed FASTQ files were then used for reading alignment.

Read alignment

The trimmed FASTQ files were aligned to the human genome reference (GRCh38/hg38) and the Gencode genome annotation v32 using the STAR aligner. The STAR aligner produced several output files for subsequent gene fusion prediction and gene expression analysis.

RNA expression

Expression levels were calculated based on mapped reads, with values reported as fragments per kilobase of transcript per million mapped reads (FPKM) and transcripts per million mapped reads (TPM), both at the transcript and gene levels, using transcriptome alignments from STAR.

Gene expression data download and normalization

Gene expression data were retrieved from DNAnexus files containing FPKM and TPM values for 59,368 records. Of these, 19,933 were protein-coding genes, which were further analyzed, while the rest were excluded. For each gene/sample, log2(FPKM + 10 −3 ) values were calculated, and genes with identical values for the first and third quartiles (indicating expression in less than 25% of samples) were excluded. The remaining 16,738 genes were z-normalized across all samples using MATLAB’s normalization function.

Data mining

We extrapolated gene expression data from MM patients: MGUS, n = 64; SMM, n = 57; NDMM, n = 207; ERMM, n = 303; LRMM, n = 182 patient cohorts. We used the NCBI Gene Expression Omnibus (GEO; https://www.ncbi.nlm.nih.gov/geo/) to validate data for ULK1, ULK2, ULK3 , ULK4 mRNA expression from FPKM genome-wide analysis; GSE5900 combined with GSE2658 (MGUS, n = 44; SMM, n = 12; NDMM, n = 559 patient cohorts, accessed in July 2023) 26 , 71 , 72 .

CD19 + B cell isolation

Human normal CD19+ B cells were isolated from healthy donors’ peripheral blood mononuclear cells (IRB Pro00021733) using CD19-specific microbeads and LS columns (Miltenyi Biotech, CAT# 130-050-301, CAT# 130-042-401) following the manufacturer’s protocol.

Cloning, ULK3 protein expression and purification

The gene encoding human ULK3 (amino acids 8–274) was synthesized and subcloned into the pET28a (+)-TEV vector (GenScript) for expression as an N-terminal His 6 -GFP fusion protein. The construct was co-expressed with λ phosphatase in E. coli BL21(DE3) cells. Protein expression was induced with 0.1 mM IPTG, and cells were harvested 20 h post-induction. Cell pellets were resuspended in lysis buffer containing 50 mM HEPES (pH 7.4), 500 mM NaCl, 20 mM imidazole, 0.5 mM tris(2-carboxyethyl) phosphine (TCEP), 10 µM MnCl₂, 5% glycerol, 0.1% Triton X-100, and EDTA-free protease inhibitor cocktail (1 tablet per 50 mL; Roche). Clarified lysate was applied to Ni-NTA Superflow resin (Qiagen), and the fusion protein was eluted using a linear imidazole gradient in buffer containing 50 mM HEPES (pH 7.4), 500 mM NaCl, 0.5 mM TCEP and 5% glycerol. Fractions containing the tagged protein were pooled and desalted using a Sephadex G-25 column (GE Healthcare). The His 6 -GFP tag was removed by incubation with TEV protease at 4 °C overnight. The cleaved sample was reapplied to Ni-NTA resin to remove the tag and uncleaved protein. The untagged ULK3 protein was further purified by size-exclusion chromatography using a Superdex 75 26/60 column (GE Healthcare) equilibrated in 25 mM HEPES (pH 7.4), 150 mM NaCl and 0.5 mM TCEP.

Protein purity was assessed by SDS–PAGE, and concentration was determined using a NanoDrop 2000c spectrophotometer (Thermo Scientific). Purified protein was flash-frozen in liquid nitrogen and stored at −80 °C.

Co-immunoprecipitation

Co-immunoprecipitation (co-IP). U266 cells (5 × 10 6 cells; approximately 2 mg total protein) were lysed in RIPA buffer supplemented with protease inhibitors. Cell lysates were precleared with 10 μL Protein G agarose beads (CAT# 15920-010) before incubation with 1 μg anti-ATG13 antibody (CAT# MBS8564879) for 1 h at 4 °C. Immune complexes were captured by overnight incubation with Protein G agarose beads at 4 °C. Following extensive washing, bead-bound proteins were divided for either immunoblot validation or proteomic analysis. All co-IP experiments were performed using validated antibodies under conditions optimized to minimize nonspecific protein binding.

Sample preparation for proteomic analysis

For proteomic analysis, proteins bound to agarose beads were eluted in SDS-PAGE sample buffer consisting of 20 μL MOPS running buffer (Bio-Rad, CAT# 1610788), 10 μL loading buffer (Bio-Rad, CAT# 1610791), and 2 μL reducing agent (Bio-Rad, CAT# 1610792). Samples were denatured at 95 °C for 5 min before separation on 10% Bis-Tris Criterion XT precast gels (Bio-Rad, CAT# 345-0111) at 125 V for 2 h. Gels were stained with InstantBlue Coomassie (Abcam, CAT# ab119211), imaged, and excised for in-gel digestion.

Excised gel fragments were reduced with 2 mM tris(2-carboxyethyl)phosphine (TCEP), alkylated with 20 mM iodoacetamide (IAA), and digested overnight at 37 °C with 200 ng sequencing-grade trypsin. An additional 200 ng trypsin was added the following day, and digestion continued for 2 h. Peptides were extracted using 50% acetonitrile containing 0.1% trifluoroacetic acid (TFA), dried by vacuum centrifugation, and resuspended in 0.1% TFA. Peptide desalting was performed on SOLAμ C18 plates (Thermo Scientific, CAT# 60209-001), and peptide concentration was determined using the Pierce Peptide Quantitation Assay (Thermo Scientific, CAT# 23275). Samples were fractionated into four high-pH reverse-phase fractions (12.5%, 17.5%, 22.5%, and 50% acetonitrile) using Pierce High-pH Reverse Phase Spin Columns (Thermo Scientific, CAT# 84868). Aliquots containing 100 ng peptide were dried, resuspended in 30 μL of 0.1% TFA, and loaded onto EvoSep EvoTip Pure tips before LC-MS/MS analysis.

LC-MS/MS

Peptide separation was performed using an EvoSep One nano-UHPLC system coupled to a Q Exactive Plus Orbitrap mass spectrometer. Samples were loaded onto EvoTip Pure cartridges (EV2013) and resolved on an analytical column (EV1106; 15 cm × 150 μm internal diameter, 1.9 μm particles) using the manufacturer’s standard 88-min gradient with solvent A (0.1% formic acid in water) and solvent B (0.1% formic acid in acetonitrile). Electrospray voltage was maintained at 1.9 kV, the capillary temperature was 275 °C, and the S-lens RF level was set to 50. Data-dependent acquisition was performed using a Top16 method. Full MS and MS/MS spectra were acquired at resolutions of 70,000 and 17,500, respectively, with a dynamic exclusion time of 15 s.

LC-RPM

Parallel reaction monitoring (PRM) experiments were carried out using an RSLC nano-UHPLC system (Thermo Scientific) coupled to an Orbitrap Exploris 480 mass spectrometer. Peptides were first concentrated on a C18 PepMap100 trap column (100 μm × 2 cm, 5 μm particles, 100 Å pores) before separation on a C18 PepMap100 analytical column (75 μm × 25 cm, 2 μm particles, 100 Å pores). Peptides were resolved over a 135-min gradient at a flow rate of 300 nL/min. Data-dependent acquisition used a 3-s duty cycle. Spray voltage was set to 2.0 kV and capillary temperature to 280 °C. Full MS and PRM scans were acquired at resolutions of 120,000 and 30,000, respectively. Relative peptide quantification by PRM targeted the following peptides: ATG13, IYFGEVQLSGLGEGFQTVR (3+) and ASPHDVLETIFVR (3+); RBCC1, EScamCQSTVHQQDETTIDTK (3+) and IQDNNENYQVGLAELR (2+); ULK1, QLLLALLQR (2+), MFSAGPTGSASSSAR (2+), and ALLLLEGLQHMLSDQADIENVTK (3+); and ULK3, ASVENLLTEIEILK (2+) and DQEGDSAAALSLYcamCK (2+), where camC denotes carbamidomethylated cysteine. PRM acquisition employed a 2.0 m/z isolation window, AGC target of 2 × 10 5 , maximum injection time of 500 ms, normalized collision energy of 28, and unscheduled data acquisition.

Proteomic data analysis

Raw mass spectrometry files were processed in Proteome Discoverer v3.0 using Mascot and Sequest search engines. Database searches were performed against the SwissProt Human (2023_03) reference database using trypsin specificity with up to two missed cleavages, precursor and fragment mass tolerances of 10 ppm and 0.05 Da, respectively. Carbamidomethylation of cysteine and oxidation of methionine were included as variable modifications. Protein identification and validation were assessed using Scaffold v5.0, while PRM data were visualized and quantified in Skyline v23.1.

Immunoblotting

Following treatment, cell pellets were lysed in RIPA buffer supplemented with protease and phosphatase inhibitors (Cell Signaling Technology, CAT# 5872S). Protein concentrations were determined using the Pierce BCA Protein Assay (Thermo Scientific, CAT# PI23225). Equal amounts of protein (30 μg) were separated on 10–15% SDS-PAGE gels at 100 V for 2 h and transferred to PVDF membranes. Membranes were blocked for 1 h in TBST (10 mM Tris-HCl, pH 8.0, 120 mM NaCl, 0.1% Tween-20) containing 5% skim milk and 5% BSA. Primary antibodies (1:1000 dilution) were incubated overnight at 4 °C. After three 10-min washes with TBS, membranes were incubated with the appropriate HRP-conjugated secondary antibodies (1:1000 dilution) prepared in either skim milk or 2% BSA 73 . Membranes were washed three additional times with TBST before chemiluminescent detection using ECL reagents. Images were acquired and analyzed using LI-COR Image Studio Quant software. All immunoblotting experiments were independently repeated at least three times. Representative blots are shown in the Source Data files, and uncropped, unprocessed immunoblots together with the corresponding densitometric analyses are provided in the Source Data . Details of all commercial antibodies and working dilutions are listed in Supplementary Table 5 .

ULK3 overexpression in ULK3-knockdown U266 cells

ULK3-knockdown U266 cells were transduced with lentiviral particles encoding ULK3 or a non-targeting control using the pLenti-C-Myc-DDK-P2A-Puro vector (Origene, CAS PS100092 ) (Fig. 3 ). Cells were seeded in antibiotic-free RPMI and transduced at MOI 5 or 10 in the presence of Vectofusin-1. Medium was replaced the following day, and puromycin (5 µg mL − 1 ) selection was initiated on day 7 post-transduction.

Confocal microscopy and image analysis

Cells were stained with Hoechst and Cyto-ID (AF488) and imaged on a Leica SP8 confocal microscope (Leica Microsystems GmbH, Wetzlar, Germany) with 405 and 488 nm laser excitation. Single-plane images were acquired at 40×/1.3 NA, with and without 2× zoom, under sequential scanning mode using one PMT and one HyD detector. Imaging parameters were kept constant across samples. Images were exported as TIF files using LAS X (v3.1.5). Images were analyzed in CellProfiler (v4.2.6; Broad Institute) with a custom pipeline. Nuclei were segmented from the Hoechst channel, and cytoplasmic regions were defined for each cell. AF488 fluorescence intensity (Cyto-ID signal) was measured within the cytoplasm. Single-cell data were exported to Excel, and the mean cytoplasmic intensity per experimental group was calculated. Statistical significance was determined using the Kruskal–Wallis test.

Immunofluorescence staining

Immunofluorescent (IF) detection of ULK3 was performed in MM patient tissue microarrays, in vitro-fixed cell lines and ex vivo tissues derived from animal studies. We conducted IF staining for Ki67 and cleaved caspase-3 in ex vivo tissue sections derived from animal studies.

For TMA analyses, slides were deparaffinized and then underwent antigen retrieval by immersion in Antigen Retrieval Citra Plus solution (pH-6.2) (BioGenex, SKU: HK080-9K) for 45 min in steamer conditions. After cooling down, slides were blocked in PBS with 10% goat serum for 1 h at room temperature (RT), then incubated for 1 h with primary antibodies (FITC-conjugated kappa/lambda light chains, Dako Agilent F0198/ F0199, 1:500 dilution, Alexa Fluor647 anti-ULK3 [EPR4888], CAT# ab310161, 1:50). After 3 × 10 min washes after incubation, 0.1 µg/mL DAPI (CAT# 28718-90-3 Sigma Aldrich) with Vector TrueVIEW (Vector Labs) autofluorescence quenching reagent was used prior to mounting the slides. Confocal immunofluorescent images of each TMA core were acquired with a Leica SP8 system with 63× objective (Leica). Four hundred five, 488, and 638 laser lines were used to excite the fluorophores and emissions were tuned to detect fluorophore-specific wavelengths. A fourth channel was added using the 552-laser line to detect autofluorescence in the orange-to-red part of the spectrum. The autofluorescence channel was used to observe platelets and red blood cells (RBCs) within the tissue.

Quantification

The fluorescence LIF image file was imported into QuPath 0.4.3, where TMA core images were analyzed. First, tissue areas were manually annotated to exclude bone tissue, debris, and regions with a high platelets and RBC count. Corresponding hematoxylin and eosin (H&E) stained cores were used to guide this process. Second, the cell detection algorithm was applied to these annotated regions using the DAPI channel to segment every cell into an individual object. A threshold classifier was used with the autofluorescence channel to separate the platelets and RBCs from other cell types. Third, a second threshold classifier was used with the CD138 (green fluorescence) channel to identify plasma cells within the previously identified cell populations. Finally, a pixel-based classification was used to identify the ULK3 staining area within each cell using a minimum intensity threshold of 50 (on a scale of 0–255). Data for each cell, including area and staining intensity measurements, were exported to Microsoft Excel for data analysis.

For in vitro immunofluorescent detection of ULK3 in MM cell lines, cells were grown on glass coverslips, fixed with ice-cold methanol for 10 min, and blocked in 10% normal goat serum for 1 h. Cells were then incubated for 1 h with primary antibody (Recombinant Anti-ULK3 antibody [EPR4888], CAT# ab124947), washed and incubated with secondary antibodies for 1 h. After washing, cells were mounted with 10 µl of mounting media (VECTASHIELD antifade mounting medium with DAPI, Vector Laboratories, CAT# H-1200) and imaged by confocal microscopy. Antibodies used are listed in Supplementary Table 5 .

For ex vivo immunofluorescent detection of proteins in xenograft tumors (isolated from U266Luc model), formalin-fixed, paraffin-embedded tumors were cut at 5-microns sections. Deparaffinized tissue slices underwent rehydration and antigen retrieval by immersion in Antigen Retrieval Citra Plus solution (pH-6.2) (BioGenex, SKU: HK080-9K) for 45 min in steamer conditions. After cooling, slides were blocked in PBS with 10% goat serum and stained for ULK3 (Alexa Fluor® 647 Anti-ULK3 antibody [EPR4888] (CAS# ab310161)) and MM cells (anti-human k λ light chain Dako F0198, F0199 in green) to assess colocalization. Confocal immunofluorescent images were acquired with a Leica TCS SP8 system using a 40× objective (Leica) and quantified as the number of tumor cells positive for ULK3 divided the total number of cells. The mean of each antibody staining was used to determine the threshold of positivity.

For ki67 and cleaved caspase staining, slides were hydrated in xylene (10 min × 2), 100% ethanol (2 min × 2), 90% ethanol (2 min × 2), 70% ethanol (2 min × 2), 50% ethanol (2 min × 2), and rinsed in water for 3 min. Tissues were incubated in gently boiling pressure cooker buffer (Tris-EDTA Buffer, pH = 9) for 4 min/high pressure for the antigen retrieval step.

Slides were cooled down to RT and blocked with 10% normal goat serum in PBS for 1 h. Primary antibody Ki67 (rabbit, 1:1000, CAT# 91295) or CC3 (rabbit, 1:1000, CAT# 9664 L) and human MM cells (hMITO, mouse, 1:500, CAT# MAB1273) were diluted in 10% normal goat serum in PBS and incubated with the slides at RT for 1 h. Slides were washed with TBS for 5 min for 3 times. Alexa Fluor Plus 647 goat anti-rabbit (1:1000, Invitrogen CAT# A32733) and Alexa Fluor Plus 488 goat anti-mouse (1:1000, Invitrogen CAT# A32723) were used as secondary antibodies for 1 h at room temperature. After washing, cells were mounted with 10 µl of mounting media (VECTASHIELD antifade mounting medium with DAPI, Vector Laboratories, CAT# H-1200) and imaged by confocal microscopy.

Confocal microscopy

Tissue slides were imaged with a Leica SP8 confocal microscope (Leica Microsystems GmbH, Wetzlar, Germany). Samples were excited with 405, 488, and 638 laser lines, and emissions were tuned to the appropriate wavelengths for each dye. Images were captured with a 40×/1.3NA objective lens, with and without a 2× zoom using sequential acquisition. All system settings remained consistent for all samples within the experiment. Five different fields of view were captured for each biological sample and plotted as an average. Images were processed for maximum projections and exported as TIF files using LAS X software version 3.1.5 (Leica Microsystems GmbH, Wetzlar, Germany).

The TIF images were imported into Definiens Tissue Studio version 4.7 (Definiens AG, Munich, Germany) for image analysis. The software was used to segment individual nuclei using a nucleus detection algorithm applied to the Hoechst channel. Within each nucleus, the co-expression detection algorithm was used to identify areas containing both green fluorescence and red fluorescence, as well as areas without co-expression. The same intensity threshold was used for each image within an experiment. Nuclei count and area data for each nucleus were exported to Excel for further data analysis.

Reagents

Dual inhibitors SG3-014 and MA9-060, along with BRD4 inhibitor JQ1, were synthesized and characterized at the H Lee Moffitt Cancer Center, US patent issued 2018; US10,106,507B2. Chloroquine (CQ, CAT# C6628-25G) and Temsirolimus (Tems, CAT# PZ0020-5mg, ≥98% HPLC) were purchased from Millipore (Sigma-Aldrich). Ruxolitinib was purchased from Selleckchem (CAT# S1378). For in vivo use, all compounds were dissolved in PBS solution of 15% 2-hydroxypropyl-β-cyclodextrin (HPβCD; CAT#128446-35-5, Millipore Sigma). Proteasome inhibitors, bortezomib (PS-341 CAT# 179324-69-7) and carfilzomib (PR-171 CAT# 868540-17-4) were purchased from Selleckchem.

Viability, proliferation, and dose response assays

Myeloma cell lines were plated in 96-well plates (2 × 10 5 cells/mL) or 384-well plates (3 × 10 5 cells/mL) and treated with vehicle control (HPβCD or DMSO) and test compounds. Cell viability was assessed at 48 h using the MTS assay (CellTiter 96® Aqueous One Solution, CAT# G3582, Promega) or the CellTiter-Blue® Cell Viability Assay (CAT# G8080, Promega). For MTS, absorbance was measured at 490 nm; for CellTiter-Blue, fluorescence (560 nm excitation/590 nm emission) was recorded, each following a 2-h incubation at 37 °C. IC 50 values were calculated using Prism software (version 9.3.1) for 96-well data or XLfit (version 5.50, IDBS) for 384-well data, and synergy matrices were generated with Combenefit software.

Genetic manipulation of ULK3 expression

ULK3 expression in MM cell lines was reduced using CRISPR–Cas9 and siRNA approaches. For CRISPR-mediated knockdown, cells were electroporated with pooled sgRNAs targeting ULK3 (Hs.Cas9.ULK3.1.AA: CGCACAACACATGTCCCCGT; Hs.Cas9.ULK3.1.AB: AGGACTTTTTTGCGCACCCC; Integrated DNA Technologies) using the Neon® Transfection System (Thermo Fisher Scientific) according to the manufacturer’s instructions 74 . An Alt-R CRISPR-Cas9 Negative Control RNA (#1079138) served as non-targeting control.

Electroporation was performed on cells at ~80% confluence, resuspended in Buffer R at 1 × 10 7 cells/ml, and 1 × 10 6 cells were electroporated in 10 µl reactions before seeding in antibiotic-free RPMI. Cells were maintained as a pooled population without single-cell cloning. Knockdown efficiency was assessed by immunoblotting and showed a partial but consistent reduction in ULK3 protein levels, consistent with pooled editing and explaining the presence of residual ULK3 bands.

Measurement of autophagic activity—CytoID assay

Autophagic activity was monitored using the Cyto-ID Green Autophagy Detection Kit (Enzo Life Sciences, ENZ-51031-K200, USA) as per the manufacturer’s instructions.

The Cyto-ID Green autophagy dye served as a selective marker of autolysosomes and early autophagic compartments. Cells were washed and treated with 1 μM tested compound or vehicle control for 6 h under standard tissue culture conditions. Upon completion of the treatment period, cells were washed with PBS and seeded at 10% confluency in FluoroDish (WPI FD35-100, glass bottom for confocal microscopy) or 1 × 10 5 cells/100 μl (flow cytometry). Microscopy: for every 1 mL of 1× Assay Buffer or complete cell growth medium, 2 µL of CYTO-ID® Green Detection Reagent and 1 µL of Hoechst 33342 Nuclear Stain were added and supplemented with 1% FBS.

Flow cytometry

The assay buffer containing 1% FBS, 1 μl of Cyto-ID Green Detection Reagent in 1 ml total volume and 100 ng/ml DAPI live/dead staining was added. Suspension cells were mixed well and incubated under standard tissue culture conditions in the dark. Thirty minutes later, cells were washed and assessed by flow cytometry. Stained controls and samples were analyzed using BD Biosciences LSRII flow cytometer or BD FACS Canto. Appropriate compensation and fluorescence-minus-one (FMO) controls were generated in parallel. The median fluorescence intensity (MFI) and percentage of Cyto-ID Green-expressing MM singlet live (DAPI negative) human MM cells in FCS Express 7. Gating strategy is reported in Supplementary Fig. 9 .

Confocal microscopy

Cells stained with Hoechst 33342 Nuclear Stain and Cyto-ID were imaged with a Leica SP8 confocal microscope (Leica Microsystems GmbH, Wetzlar, Germany). Samples were excited with 405 and 488 laser lines, and emissions were tuned to the appropriate wavelengths for each dye. Images were captured with a 40×/1.3NA objective lens and a 2× zoom using sequential acquisition on two PMT detectors. All system settings remained consistent for all samples within the experiment. Images were exported in TIF format with LAS X software version 3.1.5 (Leica Microsystems GmbH, Wetzlar, Germany).

Analysis

The TIF images were imported into Definiens Tissue Studio version 4.7 (Definiens AG, Munich, Germany) for image analysis. The software was used to segment individual nuclei using a nucleus detection algorithm applied to the Hoechst channel. Within each nucleus, the spot detection algorithm was used to detect green fluorescence spots with identical intensity and size thresholds applied to each image within an experiment. Spot count and area data for each nucleus were exported to Excel for further data analysis.

Microsome stability assay

The assay was conducted by the Exploratory Pharmacology Core at the Sanford Burnham Prebys Medical Discovery Institute in Orlando, Florida, following established procedures as previously reported 37 . Compounds, at a final concentration of 1 μM, were incubated with hepatic microsomes (1 mg/mL) for 1 h, as detailed in the referenced protocol 37 . The quantification of the remaining compound was determined through high-performance liquid chromatography (HPLC).

ULK3-MA9060 crystallization and structure determination

The frozen protein sample was thawed, concentrated to ~12 mg/ml, mixed with ~500 µM compound MA9-060 and incubated for at least 1 h. Crystallization screenings were carried out using the sitting-drop vapor diffusion method. Initial crystal hits were obtained using a reservoir solution containing 100 mM amino acid cocktail (L-Na-glutamate; alanine (racemic); glycine; lysine HCl (racemic); serine (racemic)), 100 mM buffer 1 cocktail (imidazole; sodium cacodylate; MES (acid); Bis-Tris), and 30% glycerol and PEG 4000 mix. The crystals were further optimized and flash-frozen in liquid nitrogen directly without additional cryo-protectant. X-ray diffraction data were collected at 100 K, on beamline 19-ID of NLSL-II synchrotron of Brookhaven National Laboratory (BNL). The data were integrated and scaled using XDS 75 . Structure solution was obtained by molecular replacement using Phaser as implemented in the Phenix programs suite, using a search model PDB:6FDY published previously 63 . Iterative rounds of model building and refinement were performed with COOT 76 and Phenix.Refine 77 to obtain a clear electron density map (1Fo-Fc) for compound MA9-060, which was built into the structure near the end of iterative cycles of model building and refinement. Crystal parameters, data collection statistics and refinement statistics are summarized in Supplementary Fig. 5D and deposited in the Protein Data Bank with PDB code 10NN.

Kinase profiling assay

Compounds were profiled by Reaction Biology Corporation 78 against four kinases (JAK2, ULK1, ULK2 and ULK3). Compounds were supplied as 10 mM stock solutions and gently heated to facilitate dissolution.

Kinase activity was measured using a 10-point IC 50 format with 3-fold serial dilutions starting at 10 µM. The control compound, staurosporine, was tested in a 10-point IC 50 format using 4-fold serial dilutions starting at 20 µM. Reactions were performed at a final ATP concentration of 10 µM. Assay outputs included raw data, percent enzyme activity relative to DMSO controls, and curve fitting. IC 50 values were determined when enzyme activity at the highest compound concentration was ≤65%. Data represent the mean ± s.d. of triplicate measurements. The summary of kinase inhibition is shown in Supplementary Fig. 5A .

Bromodomain profiling assay

Briefly, four compounds (SG3-014, MA9-060, JQ1, CQ; stock concentration of 10 mM) were evaluated against Bromodomain 1 by Reaction Biology Corporation.

The compounds were subjected to testing using a 10-dose IC50 singlet approach, involving a 3-fold serial dilution starting at 10 μM, specifically against BRD4-1. Additionally, as a benchmark, the control compound JQ-1 underwent testing in a 10-dose IC50 mode, employing a 3-fold serial dilution starting at 10 μM.

For ligand and detection purposes, the H4/4Ac ligand, representing the histone H4 peptide (1–21) with acetylation at lysine residues K5, K8, K12, and K16, and tagged with Biotin, was utilized.

Detection was carried out through the AlphaScreen Binding assay with excitation/emission wavelengths set at 680/520–620 nm, respectively, using the Enspire platform. The data acquisition included raw data, represented as signal-background*, % binding (relative to DMSO controls), and curve fits. Notably, curve fits were performed where the % binding at the highest concentration of compounds exhibited significant variation from DMSO controls. The IC 50 summary, detailing the inhibitory concentrations of the compounds against the respective kinases, is presented in Supplementary Fig. 5A .

mCherry-GFP-LC3 fluorescence intensity quantification and imaging

WM1366 mCherry-GFP-LC3 cells were obtained as a kind gift of Dr. Keiran Smalley’s collaboration with The Wistar Institute and produced as previously described 70 .

For cell imaging, WM1366 mCherry-GFP-LC3 cells were seeded onto glass coverslips. The next day, medium was refreshed with RPMI containing 10% fetal bovine serum, and treated with vehicle control, SG3-014, MA9-060, JQ1, and CQ (0.1 µM). After 6 h, cells were fixed and permeabilized with ice-cold MeOH.

Imaging was performed with an inverted Leica SP8 confocal microscope equipped with an 40×/1.3NA oil lens objective. Cells were mounted in a 35 mm FluoroDish Cell Culture Dish (WPI, Sarasota, FL), and z-stacks were acquired at a speed of 400 Hz, approximately 30 min post-treatment and every 2 h thereafter. Two to three fields were acquired per sample. Field size was 388 × 388 µm in x and y, and 9 µm in z, with a voxel size of 0.4 × 0.4 × 1 µm. Excitation and emission wavelengths were at 488 and 552 nm, and 500–538 nm and 587–656 nm, respectively. Z-stacks were then max-projected for visualization. (From Moffitt Microscopy Shared Resources). Analysis of images was performed with Zen software (Zeiss) or ImageJ software (NIH, National Institutes of Health).

In vitro osteoclast differentiation assays

Bone marrow cells were obtained by flushing the tibiae of 6–8-week-old male and female C57BL/6 Rag2 − / − mice. Cells were cultured in α-MEM supplemented with 25 ng/mL macrophage colony-stimulating factor (M-CSF; CAT# 315-02, PeproTech) for 3 days. The non-adherent cell fraction was then transferred to osteoclast differentiation medium containing 100 ng/mL receptor activator of nuclear factor κB ligand (RANKL; CAT# 47187000, OYC Americas) and 20 ng/mL M-CSF and maintained for an additional 7 days to generate mature osteoclasts 69 , 79 . SG3-014 or JQ1 was subsequently administered at concentrations ranging from 10 to 50 nM for 24 h. Control cultures received vehicle alone (0.5% DMSO), corresponding to the final DMSO concentration used in the highest drug treatment (50 nM). Following treatment, cells were fixed in 4% paraformaldehyde and stained for tartrate-resistant acid phosphatase (TRAcP) as previously described. TRAcP-positive multinucleated osteoclasts, defined as cells containing three or more nuclei, were quantified using ImageJ software 80 .

In vivo preclinical multiple myeloma murine models

For the U266-Luc model, 1 × 10 6 cells (100 µl/PBS) were tail vein injected into 4–6-week-old female and male NSG (NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ; JAX #005557, both sexes) mice. Upon engraftment, mice were randomized into vehicle ( n = 6), MA9- 060, JQ1 ( n = 9, 10 mg/kg, 5×/week, Monday–Friday), CQ ( n = 10, 10 mg/kg, 5×/week, Monday–Friday), and JQ1 + CQ ( n = 9, 10 mg/kg JQ1 + 10 mg/kg CQ, 5×/week, Monday–Friday) groups. Endpoint was represented by hind limb paralysis or >20% weight loss, as per protocol. Tumor burden/size limits established by the approved animal protocol were not exceeded. When sacrificed, the tibiae were harvested and fixed in 10% neutral buffered formalin for 24 h. The bioluminescent signal is reported as the average of the sum of the dorsal and ventral signal of mice/group/timepoint.

For the NTC CTRL vs ULK3 KD U266 in vivo model, mice ( n = 10/group) received 1 × 10 6 cells via tail vein (Fig. 7 ). For MA9 therapy efficacy, mice were treated with vehicle ( n = 10) or MA9 (10 mg/kg/day, subcutaneously, M-F from day 20; n = 10) (Fig. 7E, F ). Survival and toxicity (weight, clinical signs) were monitored.

Bone sectioning and histology

The harvested tibiae were fixed with 10% neutral formalin for 24 h and decalcified with 10% EDTA for 4 weeks, prior to dehydrating by a graded alcohol series, and embedding in paraffin. Tibia tissue sections (5 μm) were stained with hematoxylin and eosin (H&E) using well-established and standardized protocols as previously outlined 79 .

Osteoclast TRAcP staining

For TRAcP staining and osteoclast measurements, slides were deparaffinized and rehydrated to water, then incubated in basic stock incubation medium (112 mM anhydrous sodium acetate, 49 mM dibasic dehydrate sodium tartrate, 0.28% glacial acetic acid) containing 1% naphthol-phosphate substrate (2% naphthol AS-BI phosphate in 2-ethoxyethanol) for 1 h at 37 °C. Slides were then transferred to basic stock incubation medium containing 250 μL pararosaniline dye (5% pararosaniline dye in 2 N HCl) and 250 μL of sodium nitrite solution (4% sodium nitrite in distilled water) at 37 °C and monitored for the development of red-stained osteoclasts. After developing, the slides were rinsed in distilled water and counterstained with hematoxylin, blued, and aqueously mounted. The number of multi-nucleated, TRAcP-positive osteoclasts was quantified from multiple tissue sections.

Ex vivo X-rays bone imaging

Radiographic images were captured using the Faxitron X-ray Corp system, employing an energy setting of 35 kVp and an exposure time of 8 ms. The spatial resolution was 10 lp/mm (48 μm). The tumor area (TuA) was calculated as a function of the total tissue area (TotA) of the tibial medullary canal by using ImageJ software 80 , 81 .

Ex vivo drug sensitivity characterization and synergy studies

Primary multiple myeloma (MM) cell sensitivity to drug treatment was evaluated using an established ex vivo assay. Fresh bone marrow aspirates were processed to isolate CD138 + plasma cells by magnetic bead selection (Miltenyi Biotec, Bergisch Gladbach, Germany; CAT# 130-051-301). Purified MM cells were plated in collagen I-coated Corning CellBIND 384-well plates together with a previously established human stromal cell layer, at approximately 4000 MM cells and 1000 stromal cells per well. Cultures were maintained in 80 μL RPMI 1640 medium supplemented with heat-inactivated fetal bovine serum (FBS), penicillin/streptomycin, and 10% freshly collected, filtered autologous patient plasma obtained from the corresponding bone marrow aspirate. Plates were incubated overnight to allow stromal cell attachment before drug treatment.

Drug dispensing was performed using an automated robotic plate handler. Single agents and drug combinations (fixed-ratio combinations) were evaluated across five concentrations generated by 1:3 serial dilutions, with each condition tested in duplicate. Vehicle-treated wells containing supplemented culture medium with or without dimethyl sulfoxide (DMSO) served as negative controls, whereas MM1.S cells exposed to the highest concentration of each drug were included as positive controls. Following treatment, plates were transferred to an EVOS Auto FL live-cell imaging system (Life Technologies, Carlsbad, CA) equipped with environmental control and maintained at 37 °C in a humidified atmosphere containing 5% CO 2 . Brightfield images of every well were acquired at 30-min intervals for up to 6 days.

Digital image analysis

Drug responses were quantified using the EMMA digital image analysis platform, which determines MM cell viability longitudinally without disrupting co-cultures 82 . The algorithm compares sequential time-lapse images to identify viable myeloma cells based on continuous plasma membrane deformation generated through interactions with the surrounding extracellular matrix. Loss of these dynamic membrane movements was used as an indicator of cell death. By analyzing all 288 images collected for each well throughout the experiment, EMMA quantified drug-induced cytotoxicity over time without requiring physical separation of stromal and myeloma cells. For each patient and treatment condition, a dose–time response surface was generated and summarized as the area under the curve (AUC; Supplementary Table 2 ), representing the average treatment response across all tested concentrations during the initial 96 h of drug exposure.

Ex vivo combination effect analysis

Synergy was determined by using the method described by Sudalagunta et al. 53 , in which percent live cells across time, and five serially diluted concentrations (1:3) when treated with each drug are used to compute additive response using the Bliss Independence Model. The additive response serves as a reference to determine the extent of synergy observed in each patient sample by comparing it with the percent live cells measured when treated with the combination (at a fixed ratio of the 2 constituent single agents). Two metrics of drug sensitivity was used, area under the curve (Supplementary Table 2 ) and median lethal dose (Supplementary Table 3 ), in which the additive area under the curve/median lethal dose is compared with that of the combination to quantify the synergy seen in each patient sample. Due to the potency of the MA9 as a single agent, we expressed data as positive and negative contribution to the effect of single agent PI carfilzomib (CFZ), by calculating the combo benefit = (AUC CFZ + MA9) − (AUC CFZ). This ex vivo drug sensitivity characterization and combination effect framework using MM patient samples was developed and pioneered at H. Lee Moffitt Cancer Center and Research Institute, which resulted in several significant contributions in the field of hematological malignancies over the past 5 years 52 , 83 – 86 .

Flow cytometry quantification of ULK3

U266, MM1.S or RPMI8226 MM cells were stained at 1 × 10 6 cells/ml in PBS with Zombie NIR (1:250; Biolegend Cat# 423106) for 30 min. Cells were washed thrice in PBS, followed by fixation and permeabilization steps with Cyto-Fast™ Fix/Perm Buffer Set (Biolegend; Cat# 426803) according to the manufacturer’s instructions. Cells were then stained with Alexa Fluor® 647 Anti ULK3 antibody [EPR4888] (Abcam; Cat# ab310161) at a dilution of 1:200 for 30 min before washing and acquisition. Stained controls and samples were acquired and analyzed using BD Biosciences LSRII flow cytometer. Appropriate compensation and fluorescence-minus-one (FMO). Controls were generated in parallel and analyzed in FCS Express 7 (Supplementary Fig. 1H, I ).

Statistical analysis

Statistical significance in this study was assessed using MATLAB, Prism (GraphPad software), and IBM Statistical Package for the Social Sciences (SPSS) software. Graphs were generated using Prism (GraphPad) version 11. All sequencing data underwent analysis using R version 4.1.2 and Python version 3.10.1.

For comparisons between two groups, paired or unpaired two-sided Student’s t tests were used as appropriate. Non-parametric comparisons were performed using the Wilcoxon rank-sum or Wilcoxon signed-rank test.

For comparisons involving three or more groups, one-way analysis of variance (ANOVA) with Bonferroni post hoc correction was applied. Bioluminescence imaging data were analyzed using two-way mixed ANOVA in IBM SPSS, followed by Bonferroni-adjusted pairwise comparisons when a significant group × time interaction was observed. Non-parametric comparisons across multiple groups were performed using the Kruskal–Wallis test.

Survival analyses were conducted using Kaplan–Meier estimates and the log-rank (Mantel–Cox) test implemented in the R survival package. Holm–Bonferroni post hoc correction was applied on survival curves. Data exploration, including assessment of outliers, normality, and homogeneity of variance, as well as correlation and down-sampling analyses, was performed using IBM SPSS. Differences were considered statistically significant at p < 0.05. Significance levels are indicated as follows: p < 0.05 (*), p < 0.01 (**), p < 0.001 (***), and p < 0.0001 (****); n.s., not significant.

For Fig. 9B–E , statistical differences between treatment groups were evaluated using a repeated measures (paired) one-way analysis of variance (ANOVA), assuming sphericity and a normal distribution. To evaluate differences between all possible pairs of columns, Tukey’s multiple comparison test was applied post hoc. To maintain control over the experiment-wide false-positive rate, multiplicity-adjusted p -values were computed directly using the studentized range distribution based on the total number of treatment groups and the residual degrees of freedom. Pairwise comparisons were considered statistically significant when the adjusted p -value was less than 0.05.

Exact P values (including non-significant comparisons, where applicable) are provided in the Source Data files. Confidence intervals are reported for statistical analyses in which they are generated by the corresponding test.

Immunoblot analyses were independently performed at least three times using the same or overlapping biological samples. Samples collected independently under identical experimental conditions yielded consistent results across separate blots and experiments.

MA9-060 satisfies most criteria of a chemical probe when appropriate controls are applied to assess selectivity, as discussed in this study.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Supplementary information

Supplementary Information Reporting Summary Transparent Peer Review File

Source data

Source Data

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-76711-0 .

Acknowledgements

This research was made possible through the Oncology Research Information Exchange Network (ORIEN) Avatar Project in collaboration with Aster Insights, the Total Cancer Care protocol and supported by the Moffitt Small Animal Imaging Laboratory, Proteomics and Metabolomics, Flow Cytometry, Tissue Core, Analytical Microscopy, Molecular Genomics, Cancer Pharmacokinetics and Pharmacodynamics, Biostatistics and Bioinformatics and Collaborative Data Services Shared Resource Facilities at the H. Lee Moffitt Cancer Center & Research Institute; an NCI designated Comprehensive Cancer Center (P30-CA076292). This work was supported by philanthropic funding from the Pentecost Family Myeloma Research Center. We are thankful for the help of Drs. Manali S. Phadke, Soraya Zorro Manrique, Julieta Abraham Miranda, Karen L. Burger, Mario R. Fernandez and Massimiliano Aceti for sharing their expertise in the field.

Author contributions

M.T. conceived, designed, performed, analyzed in vitro, in vivo and ex vivo studies, analyzed patient datasets and wrote the manuscript. T.L. and N.N. performed in vitro experiments and assisted with animal studies. P.S. and R.R.C. analyzed and interpreted EMMA platform studies and RNA-seq data. M.M. performed EMMA platform studies. R.R.A. and R.R.C. analyzed patient samples and served as an honest broker for the PMRC Multiple Myeloma patient dataset. R.A. and F.L.L. provided and performed analysis of data derived from healthy donor samples. N.J.L., H.R.L., S.G., M.A., P.N., S.Y.J. and M.P.T. designed and characterized inhibitors. Y.Y. expressed and purified the ULK3 protein. T.H.T. redesigned the ULK3 construct, crystallized, and solved the structure of the MA9:ULK3 complex. E.S. biochemically characterized the inhibitors. R.T.B. assisted with Cyto-ID experiments. G.S. provided the osteoclast culture. M.M.N. assisted with animal studies. J.L.C. assisted with experimental design and edited the manuscript. A.S.S. conceived and supervised clinical sample RNA-seq and EMMA platform analyses. K.H.S. conceived and supervised patient samples and EMMA platform studies, interpreted and provided guidance on clinical data. C.C.L. conceived and designed studies, provided overall supervision and funding, and wrote the manuscript. All authors were involved in reviewing and editing the final version of the manuscript.

Peer review

Peer review information

Nature Communications thanks Xuebin Liao and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

These studies were supported by: the National Cancer Institute/National Institute of Health (NCI/NIH) grant R01CA269721 (C.C.L.). National Cancer Institute (NCI) Comprehensive Cancer Center Grant P30 CA076292, by the Cortner-Couch Endowed Chair for Cancer Research from the University of South Florida School of Medicine (J.L.C.). State of Florida funding to the H. Lee Moffitt Cancer Center & Research Institute. State of Florida Bankhead-Coley (20B10 and 24B06): Leukemia & Lymphoma Society Screen-to-Lead Grant for funding the development and crystallography of MA9-060 and SG3-014 (N.J.L., H.R.L., and E.S.). The Pentecost Family Myeloma Research Center (PMRC). The NIH grant R50 CA211447-06A1 (H.R.L.). The structural part of this publication was conducted at NYX beamline 19-ID, supported by the New York Structural Biology Center, at the National Synchrotron Light Source II, a U.S. Department of Energy (DOE) Office of Science User Facility operated for the DOE Office of Science by Brookhaven National Laboratory under Contract No. DE-SC0012704. The NYX detector instrumentation was supported by grant S10OD030394 through the Office of the Director of the National Institutes of Health. We are grateful to beamline scientist Dr. Kevin Battaile for his help with remote data collection.

Data availability

The authors confirm that the data supporting the findings of this study are available within the article and its Supplementary Materials. Source data are provided with this paper. ULK3-MA9060 crystal parameters, data collection statistics and refinement statistics are summarized in Supplementary Fig. 5D and deposited in the Protein Data Bank under PDB code 10NN.

Bulk WES and RNA-seq data . Raw sequencing data (FASTQ/BAM) and associated clinical metadata have been deposited in dbGaP under accession number phs003892 ( https://dbgap.ncbi.nlm.nih.gov/beta/study/phs003892.v1.p1/#study) . Access to the dbGaP repository is controlled because it contains human genomic data generated from participants enrolled under the Moffitt Cancer Center protocol MCC14690. Researchers seeking access should follow the instructions provided on the dbGaP study page. Scientific inquiries regarding the dataset may be directed to Dr. Ariosto S. Silva (Moffitt Cancer Center; ariosto.silva@moffitt.org) . Post-analysis files, including the MAF file, log2(FPKM), and z-score normalized RNA-seq expression data, are available through both dbGaP ( https://dbgap.ncbi.nlm.nih.gov/beta/study/phs003892.v1.p1/#study) and Synapse ( RRID:SCR_006307; https://doi.org/10.7303/syn52316987 ). The mapping between RNA-seq datasets and disease samples is provided in Supplementary Table 6 .

Proteomics data have been made available through PRIDE/ProteomeXchange under accession PXD067362 and https://doi.org/10.6019/PXD067362 . Source data are provided with this paper.

Competing interests

K.H.S. reports honoraria from Bristol Myers Squibb/Celgene, Janssen, Adaptive, Sanofi, GlaxoSmithKline, Takeda, and Amgen, as well as research funding to the institution from AbbVie and Karyopharm. F.L.L. reports scientific advisory roles and/or consulting fees from Adaptive Biotechnologies, Allogene, Amgen, AstraZeneca, Bluebird Bio, Bristol Myers Squibb (BMS), Calibr, Caribou, EcoR1, Gerson Lehrman Group (GLG), Iovance, Kite Pharma, Janssen, Legend Biotech, Miltenyi, Novartis, Sana, Pfizer, and Poseida. He serves on the Data and Safety Monitoring Board for the NCI Safety Oversight CAR T-cell Therapies Committee. His institution has received research funding or contracts from 2SeventyBio, Allogene, BMS, Incyte, Kite Pharma, the Leukemia & Lymphoma Society (Scholar in Clinical Research award; PI: Locke), the Mark Foundation, and the National Cancer Institute. He is named on several unlicensed institutional patents in the field of cellular immunotherapy. He has also engaged in educational or editorial activities with Aptitude Health, the American Society of Hematology (ASH), the American Society for Transplantation and Cellular Therapy (ASTCT), Clinical Care Options Oncology, and the Society for Immunotherapy of Cancer (SITC). H.R.L. and N.J.L. are named as inventors on the issued patents for Dual BRD4 kinase inhibitors, issued patents US10,738,016 and US10,106,507B2. H.R.L., N.J.L., C.L., and M.T. are named inventors on ULK3 inhibitors, patent application: WO2024035771, US Patent 12,391,676, which describes MA9-060. The other authors declare that they have no competing interests.

References

1. Adamik J Galson DL Roodman GD Osteoblast suppression in multiple myeloma bone disease J. Bone Oncol. 2018 13 62 70 10.1016/j.jbo.2018.09.001 30591859 PMC6303385 Adamik, J., Galson, D. L. & Roodman, G. D. Osteoblast suppression in multiple myeloma bone disease. J. Bone Oncol. 13 , 62–70 (2018). 10.1016/j.jbo.2018.09.001 PMC6303385 30591859

2. Reagan MR Liaw L Rosen CJ Ghobrial IM Dynamic interplay between bone and multiple myeloma: emerging roles of the osteoblast Bone 2015 75 161 169 10.1016/j.bone.2015.02.021 25725265 PMC4580250 Reagan, M. R., Liaw, L., Rosen, C. J. & Ghobrial, I. M. Dynamic interplay between bone and multiple myeloma: emerging roles of the osteoblast. Bone 75 , 161–169 (2015). 10.1016/j.bone.2015.02.021 PMC4580250 25725265

3. Terpos E Ntanasis-Stathopoulos I Gavriatopoulou M Dimopoulos MA Pathogenesis of bone disease in multiple myeloma: from bench to bedside Blood Cancer J. 2018 8 7 10.1038/s41408-017-0037-4 29330358 PMC5802524 Terpos, E., Ntanasis-Stathopoulos, I., Gavriatopoulou, M. & Dimopoulos, M. A. Pathogenesis of bone disease in multiple myeloma: from bench to bedside. Blood Cancer J. 8 , 7 (2018). 10.1038/s41408-017-0037-4 PMC5802524 29330358

4. Moreau P Proteasome inhibitors in multiple myeloma: 10 years later Blood 2012 120 947 959 10.1182/blood-2012-04-403733 22645181 PMC4123429 Moreau, P. et al. Proteasome inhibitors in multiple myeloma: 10 years later. Blood 120 , 947–959 (2012). 10.1182/blood-2012-04-403733 PMC4123429 22645181

5. Wang D A phase 1 study of a novel fully human BCMA-targeting CAR (CT103A) in patients with relapsed/refractory multiple myeloma Blood 2021 137 2890 2901 10.1182/blood.2020008936 33512480 Wang, D. et al. A phase 1 study of a novel fully human BCMA-targeting CAR (CT103A) in patients with relapsed/refractory multiple myeloma. Blood 137 , 2890–2901 (2021). 10.1182/blood.2020008936 33512480

6. Munshi NC Idecabtagene vicleucel in relapsed and refractory multiple myeloma New Engl. J. Med. 2021 384 705 716 10.1056/NEJMoa2024850 33626253 Munshi, N. C. et al. Idecabtagene vicleucel in relapsed and refractory multiple myeloma. New Engl. J. Med. 384 , 705–716 (2021). 10.1056/NEJMoa2024850 33626253

7. Bashiri H Tabatabaeian H Autophagy: a potential therapeutic target to tackle drug resistance in multiple myeloma Int. J. Mol. Sci. 2023 24 6019 10.3390/ijms24076019 37046991 PMC10094562 Bashiri, H. & Tabatabaeian, H. Autophagy: a potential therapeutic target to tackle drug resistance in multiple myeloma. Int. J. Mol. Sci. 24 , 6019 (2023). 10.3390/ijms24076019 PMC10094562 37046991

8. Della Porta A Acute chloroquine and hydroxychloroquine toxicity: a review for emergency clinicians Am. J. Emerg. Med. 2020 38 2209 2217 10.1016/j.ajem.2020.07.030 33071096 PMC7369162 Della Porta, A. et al. Acute chloroquine and hydroxychloroquine toxicity: a review for emergency clinicians. Am. J. Emerg. Med. 38 , 2209–2217 (2020). 10.1016/j.ajem.2020.07.030 PMC7369162 33071096

9. Mubagwa K Cardiac effects and toxicity of chloroquine: a short update Int. J. Antimicrob. Agents 2020 56 106057 10.1016/j.ijantimicag.2020.106057 32565195 PMC7303034 Mubagwa, K. Cardiac effects and toxicity of chloroquine: a short update. Int. J. Antimicrob. Agents 56 , 106057 (2020). 10.1016/j.ijantimicag.2020.106057 PMC7303034 32565195

10. Muller R Systemic toxicity of chloroquine and hydroxychloroquine: prevalence, mechanisms, risk factors, prognostic and screening possibilities Rheumatol. Int. 2021 41 1189 1202 10.1007/s00296-021-04868-6 33893862 PMC8064887 Muller, R. Systemic toxicity of chloroquine and hydroxychloroquine: prevalence, mechanisms, risk factors, prognostic and screening possibilities. Rheumatol. Int. 41 , 1189–1202 (2021). 10.1007/s00296-021-04868-6 PMC8064887 33893862

11. Puente C Hendrickson RC Jiang X Nutrient-regulated phosphorylation of ATG13 inhibits starvation-induced autophagy J. Biol. Chem. 2016 291 6026 6035 10.1074/jbc.M115.689646 26801615 PMC4786734 Puente, C., Hendrickson, R. C. & Jiang, X. Nutrient-regulated phosphorylation of ATG13 inhibits starvation-induced autophagy. J. Biol. Chem. 291 , 6026–6035 (2016). 10.1074/jbc.M115.689646 PMC4786734 26801615

12. Chen R The general amino acid control pathway regulates mTOR and autophagy during serum/glutamine starvation J. Cell Biol. 2014 206 173 182 10.1083/jcb.201403009 25049270 PMC4107793 Chen, R. et al. The general amino acid control pathway regulates mTOR and autophagy during serum/glutamine starvation. J. Cell Biol. 206 , 173–182 (2014). 10.1083/jcb.201403009 PMC4107793 25049270

13. Braden CR Neufeld TP Atg1-independent induction of autophagy by the Drosophila Ulk3 homolog, ADUK FEBS J. 2016 283 3889 3897 10.1111/febs.13906 27717182 PMC5123689 Braden, C. R. & Neufeld, T. P. Atg1-independent induction of autophagy by the Drosophila Ulk3 homolog, ADUK. FEBS J. 283 , 3889–3897 (2016). 10.1111/febs.13906 PMC5123689 27717182

14. Sun D Discovery of 5-bromo-4-phenoxy-N-phenylpyrimidin-2-amine derivatives as novel ULK1 inhibitors that block autophagy and induce apoptosis in non-small cell lung cancer Eur. J. Med. Chem. 2020 208 112782 10.1016/j.ejmech.2020.112782 32961380 Sun, D. et al. Discovery of 5-bromo-4-phenoxy-N-phenylpyrimidin-2-amine derivatives as novel ULK1 inhibitors that block autophagy and induce apoptosis in non-small cell lung cancer. Eur. J. Med. Chem. 208 , 112782 (2020). 10.1016/j.ejmech.2020.112782 32961380

15. Mu P Karuppasamy R Discovery of human autophagy initiation kinase ULK1 inhibitors by multi-directional in silico screening strategies J. Recept. Signal Transduct. Res. 2019 39 122 133 10.1080/10799893.2019.1638401 31311432 Mu, P. & Karuppasamy, R. Discovery of human autophagy initiation kinase ULK1 inhibitors by multi-directional in silico screening strategies. J. Recept. Signal Transduct. Res. 39 , 122–133 (2019). 10.1080/10799893.2019.1638401 31311432

16. Lazarus MB Novotny CJ Shokat KM Structure of the human autophagy initiating kinase ULK1 in complex with potent inhibitors ACS Chem. Biol. 2015 10 257 261 10.1021/cb500835z 25551253 PMC4301081 Lazarus, M. B., Novotny, C. J. & Shokat, K. M. Structure of the human autophagy initiating kinase ULK1 in complex with potent inhibitors. ACS Chem. Biol. 10 , 257–261 (2015). 10.1021/cb500835z PMC4301081 25551253

17. Caballe A ULK3 regulates cytokinetic abscission by phosphorylating ESCRT-III proteins eLife 2015 4 e06547 10.7554/eLife.06547 26011858 PMC4475061 Caballe, A. et al. ULK3 regulates cytokinetic abscission by phosphorylating ESCRT-III proteins. eLife 4 , e06547 (2015). 10.7554/eLife.06547 PMC4475061 26011858

18. Maloverjan A Dual function of UNC-51-like kinase 3 (ULK3) in the Sonic Hedgehog signaling pathway J. Biol. Chem. 2010 285 30079 30090 10.1074/jbc.M110.133991 20643644 PMC2943318 Maloverjan, A. et al. Dual function of UNC-51-like kinase 3 (ULK3) in the Sonic Hedgehog signaling pathway. J. Biol. Chem. 285 , 30079–30090 (2010). 10.1074/jbc.M110.133991 PMC2943318 20643644

19. Pantazi E GLI2 induces genomic instability in human keratinocytes by inhibiting apoptosis Cell Death Dis. 2014 5 e1028 10.1038/cddis.2013.535 24481442 PMC4040660 Pantazi, E. et al. GLI2 induces genomic instability in human keratinocytes by inhibiting apoptosis. Cell Death Dis. 5 , e1028 (2014). 10.1038/cddis.2013.535 PMC4040660 24481442

20. Degenhardt K Autophagy promotes tumor cell survival and restricts necrosis, inflammation, and tumorigenesis Cancer Cell 2006 10 51 64 10.1016/j.ccr.2006.06.001 16843265 PMC2857533 Degenhardt, K. et al. Autophagy promotes tumor cell survival and restricts necrosis, inflammation, and tumorigenesis. Cancer Cell 10 , 51–64 (2006). 10.1016/j.ccr.2006.06.001 PMC2857533 16843265

21. Mathew R Karantza-Wadsworth V White E Role of autophagy in cancer Nat. Rev. Cancer 2007 7 961 967 10.1038/nrc2254 17972889 PMC2866167 Mathew, R., Karantza-Wadsworth, V. & White, E. Role of autophagy in cancer. Nat. Rev. Cancer 7 , 961–967 (2007). 10.1038/nrc2254 PMC2866167 17972889

22. Goruppi S The ULK3 kinase is a determinant of keratinocyte self-renewal and tumorigenesis targeting the arginine methylome Nat. Commun. 2023 14 887 10.1038/s41467-023-36410-6 36797248 PMC9935893 Goruppi, S. et al. The ULK3 kinase is a determinant of keratinocyte self-renewal and tumorigenesis targeting the arginine methylome. Nat. Commun. 14 , 887 (2023). 10.1038/s41467-023-36410-6 PMC9935893 36797248

23. Sessa M Cavazzini F Cavallari M Rigolin GM Cuneo A A tangle of genomic aberrations drives multiple myeloma and correlates with clinical aggressiveness of the disease: a comprehensive review from a biological perspective to clinical trial results Genes 2020 11 1453 10.3390/genes11121453 33287156 PMC7761770 Sessa, M., Cavazzini, F., Cavallari, M., Rigolin, G. M. & Cuneo, A. A tangle of genomic aberrations drives multiple myeloma and correlates with clinical aggressiveness of the disease: a comprehensive review from a biological perspective to clinical trial results. Genes 11 , 1453 (2020). 10.3390/genes11121453 PMC7761770 33287156

24. Burger KL CK1δ and CK1ε signaling sustains mitochondrial metabolism and cell survival in multiple myeloma Cancer Res. 2023 83 3901 3919 10.1158/0008-5472.CAN-22-2350 37702657 PMC10690099 Burger, K. L. et al. CK1δ and CK1ε signaling sustains mitochondrial metabolism and cell survival in multiple myeloma. Cancer Res. 83 , 3901–3919 (2023). 10.1158/0008-5472.CAN-22-2350 PMC10690099 37702657

25. Fernandez MR Disrupting the MYC-TFEB circuit impairs amino acid homeostasis and provokes metabolic anergy Cancer Res. 2022 82 1234 1250 10.1158/0008-5472.CAN-21-1168 35149590 PMC9004540 Fernandez, M. R. et al. Disrupting the MYC-TFEB circuit impairs amino acid homeostasis and provokes metabolic anergy. Cancer Res. 82 , 1234–1250 (2022). 10.1158/0008-5472.CAN-21-1168 PMC9004540 35149590

26. Das DS Blockade of deubiquitylating enzyme USP1 inhibits DNA repair and triggers apoptosis in multiple myeloma cells Clin. Cancer Res. 2017 23 4280 4289 10.1158/1078-0432.CCR-16-2692 28270494 PMC5540781 Das, D. S. et al. Blockade of deubiquitylating enzyme USP1 inhibits DNA repair and triggers apoptosis in multiple myeloma cells. Clin. Cancer Res. 23 , 4280–4289 (2017). 10.1158/1078-0432.CCR-16-2692 PMC5540781 28270494

27. Bai H Wu S Wang R Xu J Chen L Bone marrow IRF4 level in multiple myeloma: an indicator of peripheral blood Th17 and disease Oncotarget 2017 8 85392 85400 10.18632/oncotarget.19907 29156727 PMC5689617 Bai, H., Wu, S., Wang, R., Xu, J. & Chen, L. Bone marrow IRF4 level in multiple myeloma: an indicator of peripheral blood Th17 and disease. Oncotarget 8 , 85392–85400 (2017). 10.18632/oncotarget.19907 PMC5689617 29156727

28. Bischoff ME Selective MAP1LC3C (LC3C) autophagy requires noncanonical regulators and the C-terminal peptide J. Cell Biol. 2021 220 e202010075 10.1083/jcb.202004182 PMC8129795 33988680 Bischoff, M. E. et al. Selective MAP1LC3C (LC3C) autophagy requires noncanonical regulators and the C-terminal peptide. J. Cell Biol. 220 , e202010075 (2021). 10.1083/jcb.202004182 PMC8129795 33988680

29. Ember SW Potent dual BET bromodomain-kinase inhibitors as value-added multitargeted chemical probes and cancer therapeutics Mol. Cancer Ther. 2017 16 1054 1067 10.1158/1535-7163.MCT-16-0568-T 28336808 PMC5457702 Ember, S. W. et al. Potent dual BET bromodomain-kinase inhibitors as value-added multitargeted chemical probes and cancer therapeutics. Mol. Cancer Ther. 16 , 1054–1067 (2017). 10.1158/1535-7163.MCT-16-0568-T PMC5457702 28336808

30. Davis RR Structural insights into JAK2 inhibition by ruxolitinib, fedratinib, and derivatives thereof J. Med. Chem. 2021 64 2228 2241 10.1021/acs.jmedchem.0c01952 33570945 PMC8327781 Davis, R. R. et al. Structural insights into JAK2 inhibition by ruxolitinib, fedratinib, and derivatives thereof. J. Med. Chem. 64 , 2228–2241 (2021). 10.1021/acs.jmedchem.0c01952 PMC8327781 33570945

31. Moller HEH Myc protein overexpression is a feature of progression and adverse prognosis in multiple myeloma Eur. J. Haematol. 2018 101 585 593 10.1111/ejh.13141 29999206 Moller, H. E. H. et al. Myc protein overexpression is a feature of progression and adverse prognosis in multiple myeloma. Eur. J. Haematol. 101 , 585–593 (2018). 10.1111/ejh.13141 29999206

32. Chng WJ Clinical and biological implications of MYC activation: a common difference between MGUS and newly diagnosed multiple myeloma Leukemia 2011 25 1026 1035 10.1038/leu.2011.53 21468039 PMC3432644 Chng, W. J. et al. Clinical and biological implications of MYC activation: a common difference between MGUS and newly diagnosed multiple myeloma. Leukemia 25 , 1026–1035 (2011). 10.1038/leu.2011.53 PMC3432644 21468039

33. Szabo AG Overexpression of c-Myc is associated with adverse clinical features and worse overall survival in multiple myeloma Leuk. Lymphoma 2016 57 2526 2534 10.1080/10428194.2016.1187275 27243588 Szabo, A. G. et al. Overexpression of c-Myc is associated with adverse clinical features and worse overall survival in multiple myeloma. Leuk. Lymphoma 57 , 2526–2534 (2016). 10.1080/10428194.2016.1187275 27243588

34. Delmore JE BET bromodomain inhibition as a therapeutic strategy to target c-Myc Cell 2011 146 904 917 10.1016/j.cell.2011.08.017 21889194 PMC3187920 Delmore, J. E. et al. BET bromodomain inhibition as a therapeutic strategy to target c-Myc. Cell 146 , 904–917 (2011). 10.1016/j.cell.2011.08.017 PMC3187920 21889194

35. Calder J Resistance to BET inhibitors in lung adenocarcinoma is mediated by casein kinase phosphorylation of BRD4 Oncogenesis 2021 10 27 10.1038/s41389-021-00316-z 33712563 PMC7955060 Calder, J. et al. Resistance to BET inhibitors in lung adenocarcinoma is mediated by casein kinase phosphorylation of BRD4. Oncogenesis 10 , 27 (2021). 10.1038/s41389-021-00316-z PMC7955060 33712563

36. Kurimchak AM Resistance to BET bromodomain inhibitors is mediated by kinome reprogramming in ovarian cancer Cell Rep. 2016 16 1273 1286 10.1016/j.celrep.2016.06.091 27452461 PMC4972668 Kurimchak, A. M. et al. Resistance to BET bromodomain inhibitors is mediated by kinome reprogramming in ovarian cancer. Cell Rep. 16 , 1273–1286 (2016). 10.1016/j.celrep.2016.06.091 PMC4972668 27452461

37. McAnally D Repurposing antimalarial aminoquinolines and related compounds for treatment of retinal neovascularization PLoS ONE 2018 13 e0202436 10.1371/journal.pone.0202436 30208056 PMC6135396 McAnally, D. et al. Repurposing antimalarial aminoquinolines and related compounds for treatment of retinal neovascularization. PLoS ONE 13 , e0202436 (2018). 10.1371/journal.pone.0202436 PMC6135396 30208056

38. Fowler JA Mundy GR Lwin ST Lynch CC Edwards CM A murine model of myeloma that allows genetic manipulation of the host microenvironment Dis. Models Mech. 2009 2 604 611 10.1242/dmm.003160 PMC2776114 19779066 Fowler, J. A., Mundy, G. R., Lwin, S. T., Lynch, C. C. & Edwards, C. M. A murine model of myeloma that allows genetic manipulation of the host microenvironment. Dis. Models Mech. 2 , 604–611 (2009). 10.1242/dmm.003160 PMC2776114 19779066

39. Wang L Small molecule JQ1 promotes prostate cancer invasion via BET-independent inactivation of FOXA1 J. Clin. Investig. 2020 130 1782 1792 10.1172/JCI126327 31874106 PMC7108891 Wang, L. et al. Small molecule JQ1 promotes prostate cancer invasion via BET-independent inactivation of FOXA1. J. Clin. Investig. 130 , 1782–1792 (2020). 10.1172/JCI126327 PMC7108891 31874106

40. Nishihori T Shain K Insights on genomic and molecular alterations in multiple myeloma and their incorporation towards risk-adapted treatment strategy Int. J. Genom. 2017 2017 6934183 10.1155/2017/6934183 PMC5698810 29250532 Nishihori, T. & Shain, K. Insights on genomic and molecular alterations in multiple myeloma and their incorporation towards risk-adapted treatment strategy. Int. J. Genom. 2017 , 6934183 (2017). 10.1155/2017/6934183 PMC5698810 29250532

41. Chen S CDK inhibitors upregulate BH3-only proteins to sensitize human myeloma cells to BH3 mimetic therapies Cancer Res. 2012 72 4225 4237 10.1158/0008-5472.CAN-12-1118 22693249 PMC3421040 Chen, S. et al. CDK inhibitors upregulate BH3-only proteins to sensitize human myeloma cells to BH3 mimetic therapies. Cancer Res. 72 , 4225–4237 (2012). 10.1158/0008-5472.CAN-12-1118 PMC3421040 22693249

42. Turner JG Treatment of acquired drug resistance in multiple myeloma by combination therapy with XPO1 and topoisomerase II inhibitors J. Hematol. Oncol. 2016 9 73 10.1186/s13045-016-0304-z 27557643 PMC4997728 Turner, J. G. et al. Treatment of acquired drug resistance in multiple myeloma by combination therapy with XPO1 and topoisomerase II inhibitors. J. Hematol. Oncol. 9 , 73 (2016. 10.1186/s13045-016-0304-z PMC4997728 27557643

43. Zeng R Chen Y Zhao S Cui GH Autophagy counteracts apoptosis in human multiple myeloma cells exposed to oridonin in vitro via regulating intracellular ROS and SIRT1 Acta Pharmacol. Sin. 2012 33 91 100 10.1038/aps.2011.143 22158107 PMC4010261 Zeng, R., Chen, Y., Zhao, S. & Cui, G. H. Autophagy counteracts apoptosis in human multiple myeloma cells exposed to oridonin in vitro via regulating intracellular ROS and SIRT1. Acta Pharmacol. Sin. 33 , 91–100 (2012). 10.1038/aps.2011.143 PMC4010261 22158107

44. Dykstra KM Allen C Born EJ Tong H Holstein SA Mechanisms for autophagy modulation by isoprenoid biosynthetic pathway inhibitors in multiple myeloma cells Oncotarget 2015 6 41535 41549 10.18632/oncotarget.6365 26595805 PMC4747172 Dykstra, K. M., Allen, C., Born, E. J., Tong, H. & Holstein, S. A. Mechanisms for autophagy modulation by isoprenoid biosynthetic pathway inhibitors in multiple myeloma cells. Oncotarget 6 , 41535–41549 (2015). 10.18632/oncotarget.6365 PMC4747172 26595805

45. Chen S A Bim-targeting strategy overcomes adaptive bortezomib resistance in myeloma through a novel link between autophagy and apoptosis Blood 2014 124 2687 2697 10.1182/blood-2014-03-564534 25208888 PMC4208284 Chen, S. et al. A Bim-targeting strategy overcomes adaptive bortezomib resistance in myeloma through a novel link between autophagy and apoptosis. Blood 124 , 2687–2697 (2014). 10.1182/blood-2014-03-564534 PMC4208284 25208888

46. Dai Y Grant S BCL2L11/Bim as a dual-agent regulating autophagy and apoptosis in drug resistance Autophagy 2015 11 416 418 10.1080/15548627.2014.998892 25700997 PMC4502657 Dai, Y. & Grant, S. BCL2L11/Bim as a dual-agent regulating autophagy and apoptosis in drug resistance. Autophagy 11 , 416–418 (2015). 10.1080/15548627.2014.998892 PMC4502657 25700997

47. Vogl DT Combined autophagy and proteasome inhibition: a phase 1 trial of hydroxychloroquine and bortezomib in patients with relapsed/refractory myeloma Autophagy 2014 10 1380 1390 10.4161/auto.29264 24991834 PMC4203515 Vogl, D. T. et al. Combined autophagy and proteasome inhibition: a phase 1 trial of hydroxychloroquine and bortezomib in patients with relapsed/refractory myeloma. Autophagy 10 , 1380–1390 (2014). 10.4161/auto.29264 PMC4203515 24991834

48. Zhang H Knockdown of DEPTOR induces apoptosis, increases chemosensitivity to doxorubicin and suppresses autophagy in RPMI-8226 human multiple myeloma cells in vitro Int. J. Mol. Med. 2013 31 1127 1134 10.3892/ijmm.2013.1299 23503641 Zhang, H. et al. Knockdown of DEPTOR induces apoptosis, increases chemosensitivity to doxorubicin and suppresses autophagy in RPMI-8226 human multiple myeloma cells in vitro. Int. J. Mol. Med. 31 , 1127–1134 (2013). 10.3892/ijmm.2013.1299 23503641

49. Aronson LI Davies FE DangER: protein ovERload. Targeting protein degradation to treat myeloma Haematologica 2012 97 1119 1130 10.3324/haematol.2012.064923 22580998 PMC3409807 Aronson, L. I. & Davies, F. E. DangER: protein ovERload. Targeting protein degradation to treat myeloma. Haematologica 97 , 1119–1130 (2012). 10.3324/haematol.2012.064923 PMC3409807 22580998

50. Tucci M An imbalance between Beclin-1 and p62 expression promotes the proliferation of myeloma cells through autophagy regulation Exp. Hematol. 2014 42 897 908 10.1016/j.exphem.2014.06.005 24971696 Tucci, M. et al. An imbalance between Beclin-1 and p62 expression promotes the proliferation of myeloma cells through autophagy regulation. Exp. Hematol. 42 , 897–908 (2014). 10.1016/j.exphem.2014.06.005 24971696

51. Aronson LI Davenport EL Mirabella F Morgan GJ Davies FE Understanding the interplay between the proteasome pathway and autophagy in response to dual PI3K/mTOR inhibition in myeloma cells Leukemia 2013 27 2397 2403 10.1038/leu.2013.150 23670295 PMC3865535 Aronson, L. I., Davenport, E. L., Mirabella, F., Morgan, G. J. & Davies, F. E. Understanding the interplay between the proteasome pathway and autophagy in response to dual PI3K/mTOR inhibition in myeloma cells. Leukemia 27 , 2397–2403 (2013). 10.1038/leu.2013.150 PMC3865535 23670295

52. Silva A An ex vivo platform for the prediction of clinical response in multiple myeloma Cancer Res. 2017 77 3336 3351 10.1158/0008-5472.CAN-17-0502 28400475 PMC7819642 Silva, A. et al. An ex vivo platform for the prediction of clinical response in multiple myeloma. Cancer Res. 77 , 3336–3351 (2017). 10.1158/0008-5472.CAN-17-0502 PMC7819642 28400475

53. Sudalagunta P A pharmacodynamic model of clinical synergy in multiple myeloma EBioMedicine 2020 54 102716 10.1016/j.ebiom.2020.102716 32268267 PMC7136599 Sudalagunta, P. et al. A pharmacodynamic model of clinical synergy in multiple myeloma. EBioMedicine 54 , 102716 (2020). 10.1016/j.ebiom.2020.102716 PMC7136599 32268267

54. Zhang Y Coordinated regulation of protein synthesis and degradation by mTORC1 Nature 2014 513 440 443 10.1038/nature13492 25043031 PMC4402229 Zhang, Y. et al. Coordinated regulation of protein synthesis and degradation by mTORC1. Nature 513 , 440–443 (2014). 10.1038/nature13492 PMC4402229 25043031

55. Mathiassen SG De Zio D Cecconi F Autophagy and the cell cycle: a complex landscape Front. Oncol. 2017 7 51 10.3389/fonc.2017.00051 28409123 PMC5374984 Mathiassen, S. G., De Zio, D. & Cecconi, F. Autophagy and the cell cycle: a complex landscape. Front. Oncol. 7 , 51 (2017). 10.3389/fonc.2017.00051 PMC5374984 28409123

56. Odle RI An mTORC1-to-CDK1 switch maintains autophagy suppression during mitosis Mol. Cell 2020 77 228 240 10.1016/j.molcel.2019.10.016 31733992 PMC6964153 Odle, R. I. et al. An mTORC1-to-CDK1 switch maintains autophagy suppression during mitosis. Mol. Cell 77 , 228–240 (2020). 10.1016/j.molcel.2019.10.016 PMC6964153 31733992

57. Furuya T Negative regulation of Vps34 by Cdk-mediated phosphorylation Mol. Cell 2010 38 500 511 10.1016/j.molcel.2010.05.009 20513426 PMC2888511 Furuya, T. et al. Negative regulation of Vps34 by Cdk-mediated phosphorylation. Mol. Cell 38 , 500–511 (2010). 10.1016/j.molcel.2010.05.009 PMC2888511 20513426

58. Kang R Zeh HJ Lotze MT Tang D The Beclin 1 network regulates autophagy and apoptosis Cell Death Differ. 2011 18 571 580 10.1038/cdd.2010.191 21311563 PMC3131912 Kang, R., Zeh, H. J., Lotze, M. T. & Tang, D. The Beclin 1 network regulates autophagy and apoptosis. Cell Death Differ. 18 , 571–580 (2011). 10.1038/cdd.2010.191 PMC3131912 21311563

59. Goruppi S The ULK3 kinase is critical for convergent control of cancer-associated fibroblast activation by CSL and GLI Cell Rep. 2017 20 2468 2479 10.1016/j.celrep.2017.08.048 28877478 PMC5616185 Goruppi, S. et al. The ULK3 kinase is critical for convergent control of cancer-associated fibroblast activation by CSL and GLI. Cell Rep. 20 , 2468–2479 (2017). 10.1016/j.celrep.2017.08.048 PMC5616185 28877478

60. Zhang D shRNA-mediated silencing of Gli2 gene inhibits proliferation and sensitizes human hepatocellular carcinoma cells towards TRAIL-induced apoptosis J. Cell Biochem. 2011 112 3140 3150 10.1002/jcb.23240 21695716 Zhang, D. et al. shRNA-mediated silencing of Gli2 gene inhibits proliferation and sensitizes human hepatocellular carcinoma cells towards TRAIL-induced apoptosis. J. Cell Biochem. 112 , 3140–3150 (2011). 10.1002/jcb.23240 21695716

61. Joyce E Fabre A Mahon N Hydroxychloroquine cardiotoxicity presenting as a rapidly evolving biventricular cardiomyopathy Eur. Heart J. Acute Cardiovasc. Care 2013 2 77 83 24062937 10.1177/2048872612471215 PMC3760572 Joyce, E., Fabre, A. & Mahon, N. Hydroxychloroquine cardiotoxicity presenting as a rapidly evolving biventricular cardiomyopathy. Eur. Heart J. Acute Cardiovasc. Care 2 , 77–83 (2013). 10.1177/2048872612471215 PMC3760572 24062937

62. Tleyjeh IM Cardiac toxicity of chloroquine or hydroxychloroquine in patients with COVID-19 Mayo Clin. Proc. Innov. Qual. Outcomes 2021 5 137 150 10.1016/j.mayocpiqo.2020.10.005 33163895 PMC7605861 Tleyjeh, I. M. et al. Cardiac toxicity of chloroquine or hydroxychloroquine in patients with COVID-19. Mayo Clin. Proc. Innov. Qual. Outcomes 5 , 137–150 (2021). 10.1016/j.mayocpiqo.2020.10.005 PMC7605861 33163895

63. Mathea S Conformational plasticity of the ULK3 kinase domain Biochem. J. 2021 478 2811 2823 10.1042/BCJ20210257 34190988 Mathea, S. et al. Conformational plasticity of the ULK3 kinase domain. Biochem. J. 478 , 2811–2823 (2021). 10.1042/BCJ20210257 34190988

64. Lei Y Zhang E Bai L Li Y Autophagy in cancer immunotherapy Cells 2022 11 2996 10.3390/cells11192996 36230955 PMC9564118 Lei, Y., Zhang, E., Bai, L. & Li, Y. Autophagy in cancer immunotherapy. Cells 11 , 2996 (2022). 10.3390/cells11192996 PMC9564118 36230955

65. Meads MB Target antigen and plasma cell phenotype are critical factors for sensitivity to response-adapted daratumumab therapy Blood 2026 147 497 512 10.1182/blood.2025029921 40983035 Meads, M. B. et al. Target antigen and plasma cell phenotype are critical factors for sensitivity to response-adapted daratumumab therapy. Blood 147 , 497–512 (2026). 10.1182/blood.2025029921 40983035

66. Zhuang J Ubiquitin-activating enzyme inhibition induces an unfolded protein response and overcomes drug resistance in myeloma Blood 2019 133 1572 1584 10.1182/blood-2018-06-859686 30737236 PMC6450433 Zhuang, J. et al. Ubiquitin-activating enzyme inhibition induces an unfolded protein response and overcomes drug resistance in myeloma. Blood 133 , 1572–1584 (2019). 10.1182/blood-2018-06-859686 PMC6450433 30737236

67. Zhu W Mesenchymal stem cells derived from bone marrow favor tumor cell growth in vivo Exp. Mol. Pathol. 2006 80 267 274 10.1016/j.yexmp.2005.07.004 16214129 Zhu, W. et al. Mesenchymal stem cells derived from bone marrow favor tumor cell growth in vivo. Exp. Mol. Pathol. 80 , 267–274 (2006). 10.1016/j.yexmp.2005.07.004 16214129

68. McGuire JJ Mesenchymal stem cell-derived interleukin-28 drives the selection of apoptosis-resistant bone metastatic prostate cancer Nat. Commun. 2021 12 723 10.1038/s41467-021-20962-6 33526787 PMC7851397 McGuire, J. J. et al. Mesenchymal stem cell-derived interleukin-28 drives the selection of apoptosis-resistant bone metastatic prostate cancer. Nat. Commun. 12 , 723 (2021). 10.1038/s41467-021-20962-6 PMC7851397 33526787

69. Shay G Selective inhibition of matrix metalloproteinase-2 in the multiple myeloma-bone microenvironment Oncotarget 2017 8 41827 41840 10.18632/oncotarget.18103 28611279 PMC5522031 Shay, G. et al. Selective inhibition of matrix metalloproteinase-2 in the multiple myeloma-bone microenvironment. Oncotarget 8 , 41827–41840 (2017). 10.18632/oncotarget.18103 PMC5522031 28611279

70. Kraya AA Identification of secreted proteins that reflect autophagy dynamics within tumor cells Autophagy 2015 11 60 74 10.4161/15548627.2014.984273 25484078 PMC4502670 Kraya, A. A. et al. Identification of secreted proteins that reflect autophagy dynamics within tumor cells. Autophagy 11 , 60–74 (2015). 10.4161/15548627.2014.984273 PMC4502670 25484078

71. Grieve S TAZ functions as a tumor suppressor in multiple myeloma by downregulating MYC Blood Adv. 2019 3 3613 3625 10.1182/bloodadvances.2019000374 31743393 PMC6880893 Grieve, S. et al. TAZ functions as a tumor suppressor in multiple myeloma by downregulating MYC. Blood Adv. 3 , 3613–3625 (2019). 10.1182/bloodadvances.2019000374 PMC6880893 31743393

72. Zhan F Gene-expression signature of benign monoclonal gammopathy evident in multiple myeloma is linked to good prognosis Blood 2007 109 1692 1700 10.1182/blood-2006-07-037077 17023574 PMC1794073 Zhan, F. et al. Gene-expression signature of benign monoclonal gammopathy evident in multiple myeloma is linked to good prognosis. Blood 109 , 1692–1700 (2007). 10.1182/blood-2006-07-037077 PMC1794073 17023574

73. Mahmood T Yang PC Western blot: technique, theory, and troubleshooting N. Am. J. Med. Sci. 2012 4 429 434 10.4103/1947-2714.100998 23050259 PMC3456489 Mahmood, T. & Yang, P. C. Western blot: technique, theory, and troubleshooting. N. Am. J. Med. Sci. 4 , 429–434 (2012). 10.4103/1947-2714.100998 PMC3456489 23050259

74. Beyer M Kramer OH RNA interference protocol to silence oncogenic drivers in leukemia cell lines STAR Protoc. 2022 3 101512 10.1016/j.xpro.2022.101512 35779262 PMC9254495 Beyer, M. & Kramer, O. H. RNA interference protocol to silence oncogenic drivers in leukemia cell lines. STAR Protoc. 3 , 101512 (2022). 10.1016/j.xpro.2022.101512 PMC9254495 35779262

75. Kabsch W Integration, scaling, space-group assignment and post-refinement Acta Crystallogr. D Struct. Biol. 2010 66 133 144 10.1107/S0907444909047374 PMC2815666 20124693 Kabsch, W. Integration, scaling, space-group assignment and post-refinement. Acta Crystallogr. D Struct. Biol. 66 , 133–144 (2010). 10.1107/S0907444909047374 PMC2815666 20124693

76. Emsley P Lohkamp B Scott WG Cowtan K Features and development of Coot Acta Crystallogr. D Struct. Biol. 2010 66 486 501 10.1107/S0907444910007493 PMC2852313 20383002 Emsley, P., Lohkamp, B., Scott, W. G. & Cowtan, K. Features and development of Coot. Acta Crystallogr. D Struct. Biol. 66 , 486–501 (2010). 10.1107/S0907444910007493 PMC2852313 20383002

77. Adams PD PHENIX: a comprehensive Python-based system for macromolecular structure solution Acta Crystallogr. D Struct. Biol. 2010 66 213 221 10.1107/S0907444909052925 PMC2815670 20124702 Adams, P. D. et al. PHENIX: a comprehensive Python-based system for macromolecular structure solution. Acta Crystallogr. D Struct. Biol. 66 , 213–221 (2010). 10.1107/S0907444909052925 PMC2815670 20124702

78. Anastassiadis T Deacon SW Devarajan K Ma H Peterson JR Comprehensive assay of kinase catalytic activity reveals features of kinase inhibitor selectivity Nat. Biotechnol. 2011 29 1039 1045 10.1038/nbt.2017 22037377 PMC3230241 Anastassiadis, T., Deacon, S. W., Devarajan, K., Ma, H. & Peterson, J. R. Comprehensive assay of kinase catalytic activity reveals features of kinase inhibitor selectivity. Nat. Biotechnol. 29 , 1039–1045 (2011). 10.1038/nbt.2017 PMC3230241 22037377

79. Tauro M Bone-seeking matrix metalloproteinase-2 inhibitors prevent bone metastatic breast cancer growth Mol. Cancer Ther. 2017 16 494 505 10.1158/1535-7163.MCT-16-0315-T 28069877 PMC5337166 Tauro, M. et al. Bone-seeking matrix metalloproteinase-2 inhibitors prevent bone metastatic breast cancer growth. Mol. Cancer Ther. 16 , 494–505 (2017). 10.1158/1535-7163.MCT-16-0315-T PMC5337166 28069877

80. Schneider CA Rasband WS Eliceiri KW NIH Image to ImageJ: 25 years of image analysis Nat. Methods 2012 9 671 675 10.1038/nmeth.2089 22930834 PMC5554542 Schneider, C. A., Rasband, W. S. & Eliceiri, K. W. NIH Image to ImageJ: 25 years of image analysis. Nat. Methods 9 , 671–675 (2012). 10.1038/nmeth.2089 PMC5554542 22930834

81. Kim Y Brodt MD Tang SY Silva MJ MicroCT for scanning and analysis of mouse bones Methods Mol. Biol. 2021 2230 169 198 10.1007/978-1-0716-1028-2_11 33197015 PMC8409170 Kim, Y., Brodt, M. D., Tang, S. Y. & Silva, M. J. MicroCT for scanning and analysis of mouse bones. Methods Mol. Biol. 2230 , 169–198 (2021). 10.1007/978-1-0716-1028-2_11 PMC8409170 33197015

82. Silva A Jacobson T Meads M Distler A Shain K An organotypic high throughput system for characterization of drug sensitivity of primary multiple myeloma cells J. Vis. Exp. 2015 101 e53070 10.3791/53070 PMC4544963 26274375 Silva, A., Jacobson, T., Meads, M., Distler, A. & Shain, K. An organotypic high throughput system for characterization of drug sensitivity of primary multiple myeloma cells. J. Vis. Exp. 101 , e53070 (2015). 10.3791/53070 PMC4544963 26274375

83. Mostofa AGM Plasma cell dependence on histone/protein deacetylase 11 reveals a therapeutic target in multiple myeloma JCI Insight 2021 6 e151713 10.1172/jci.insight.151713 34793338 PMC8783683 Mostofa, A. G. M. et al. Plasma cell dependence on histone/protein deacetylase 11 reveals a therapeutic target in multiple myeloma. JCI Insight 6 , e151713 (2021). 10.1172/jci.insight.151713 PMC8783683 34793338

84. Ren Y PLK1 stabilizes a MYC-dependent kinase network in aggressive B cell lymphomas J. Clin. Investig. 2018 128 5517 5530 10.1172/JCI122533 30260324 PMC6264635 Ren, Y. et al. PLK1 stabilizes a MYC-dependent kinase network in aggressive B cell lymphomas. J. Clin. Investig. 128 , 5517–5530 (2018). 10.1172/JCI122533 PMC6264635 30260324

85. Zhao X BCL2 amplicon loss and transcriptional remodeling drives ABT-199 resistance in B cell lymphoma models Cancer Cell 2019 35 752 766 10.1016/j.ccell.2019.04.005 31085176 PMC6945775 Zhao, X. et al. BCL2 amplicon loss and transcriptional remodeling drives ABT-199 resistance in B cell lymphoma models. Cancer Cell 35 , 752–766 (2019). 10.1016/j.ccell.2019.04.005 PMC6945775 31085176

86. Zhou L IAP and HDAC inhibitors interact synergistically in myeloma cells through noncanonical NF-κB- and caspase-8-dependent mechanisms Blood Adv. 2021 5 3776 3788 10.1182/bloodadvances.2020003597 34464977 PMC8679669 Zhou, L. et al. IAP and HDAC inhibitors interact synergistically in myeloma cells through noncanonical NF-κB- and caspase-8-dependent mechanisms. Blood Adv. 5 , 3776–3788 (2021). 10.1182/bloodadvances.2020003597 PMC8679669 34464977