BAG3-HSP70 Autophagy Axis Emerges as Contested Target in Leukemia

Acute myeloid leukemia has long resisted the kind of precision breakthroughs that transformed other cancers, and a sweeping new review argues that one of the field’s most seductive mechanisms may be built on shakier ground than anyone realized. Writing in Frontiers in Cell and Developmental Biology, Yang Liu and Tao Wu of the 940th Hospital in Lanzhou dissect the BAG3-HSP70 axis, a stress-responsive proteostasis node that has been proposed to drive the autophagy machinery sustaining leukemia stem cells. Their verdict is deliberately provocative: the axis is plausible, pharmacologically tractable and prognostically relevant, but the specific molecular couplings that would make it a precision target in AML remain unproven. The review arrives at a moment when proteostasis itself has been identified as a therapeutic vulnerability in AML, making the question of what exactly to target more urgent than ever.

The technical heart of the story is autophagy, an evolutionarily conserved lysosomal degradation pathway in which cytoplasmic material is sequestered in double-membrane autophagosomes and delivered to lysosomes for recycling. Its role in cancer is famously dual and context dependent: it can suppress early tumorigenesis yet sustain established tumors under metabolic, hypoxic or therapeutic stress. In AML, autophagy is intertwined with metabolism, mitochondrial quality control and leukemia stem cell maintenance, and it is frequently mobilized as a protective response to targeted therapy. The BAG3-HSP70 axis integrates proteotoxic stress with this machinery, most notably through chaperone-assisted selective autophagy, or CASA, in which the HSPB8-HSP70-BAG3-CHIP module ubiquitinates aggregation-prone clients and presents them to the cargo receptor p62 and membrane-bound LC3-II for sequestration. BAG3, a stress-inducible co-chaperone and nucleotide exchange factor for HSP70, is barely detectable under physiological conditions in many tissues but is strongly induced by heat shock, proteasome inhibition and oxidative or proteotoxic stress through heat-shock factor 1.

A well-documented feature positions BAG3 as the inducible, autophagy-directed arm of the HSP70 system: the BAG1-to-BAG3 switch in protein quality control. BAG1 preferentially routes HSP70 clients to the ubiquitin-proteasome system, whereas BAG3 recruits the autophagy-lysosomal pathway, and during aging or under proteasome impairment the BAG3-to-BAG1 ratio rises and macroautophagy becomes dominant. Whether this switch is engaged in AML has never been tested. The review also insists on a terminological constraint that runs throughout the literature: demonstrating p62- or OPTN-dependent autophagy or mitophagy in AML, or observing changes in LC3-II or p62 abundance, does not by itself constitute evidence for BAG3-dependent CASA, because those receptors and that machinery can operate independently of BAG3. A claim of CASA requires evidence that the cargo flux is BAG3-dependent, ideally through BAG3 perturbation together with a flux readout such as LC3-II turnover in the presence of a lysosomal inhibitor or a tandem mRFP-GFP-LC3 reporter.

To impose order on a sprawling evidence base, the authors apply a four-tier framework: Tier A for direct AML evidence from patient samples, cell lines or genetic models; Tier B for other hematological settings; Tier C for non-AML evidence from solid tumors, muscle or non-mammalian systems; and Tier D for proposed or hypothetical mechanisms without direct supporting data. Within Tier A, BAG3 has been linked to anti-apoptotic stability, therapeutic resistance and adverse prognosis rather than to autophagy coupling. In AML cell lines, BAG1 and BAG3 cooperate to maintain the anti-apoptotic proteins BCL2 and MCL1, and BAG3 overexpression rescues their levels when BAG1 is depleted. More recently, elevated BAG3 was identified as a component of a venetoclax-resistance prognostic signature, with BAG3 abundance positively associated with venetoclax IC50, and a nomogram combining the immune checkpoint CD276, BAG3 and SRC reported that high expression predicts inferior outcome. Tier B studies reinforce an anti-apoptotic and proteostasis role: BAG3 gene silencing sensitized leukemic cells to bortezomib-induced apoptosis, including in a nude-mouse xenograft setting, the leukemic oncoprotein WT1 transcriptionally activates BAG3, and BAG3 is overexpressed and proposed as a therapeutic target in chronic lymphocytic leukemia.

The inventory supports a precise and uncomfortable conclusion. BAG3 is expressed, prognostically relevant and anti-apoptotic or proteostasis-supportive in AML and other leukemic cells, and its perturbation can modulate sensitivity to bortezomib, venetoclax and proteasome inhibitors. However, none of these studies demonstrates in primary AML that BAG3 directly binds or activates ULK1, that BAG3 or HSP70 competes at the Beclin1 BH3 groove, that HSP70 directly shields ATG5 from cleavage, or that the relevant autophagic flux is BAG3-dependent. The molecular depiction of a BAG3-driven CASA cascade in AML therefore rests on Tier C and Tier D extrapolation rather than on direct AML coupling. This is the central evidence boundary of the review, and the authors state it bluntly: direct BAG3 coupling to ULK1, competitive BAG3/HSP70 binding at the Beclin1 BH3 domain, and direct HSP70 shielding of ATG5 from cleavage are unsupported and are more accurately framed as untested or tissue-specific hypotheses.

The duality of autophagy itself complicates any targeting strategy. At the leukemogenic and newly diagnosed stage, autophagy can restrain disease: AML blasts display comparatively low expression of autophagy and lysosomal genes, and heterozygous loss of Atg5 or Atg7 accelerates MLL-ENL-driven leukemogenesis, consistent with a tumor-suppressive brake in early transformation. In established and LSC-enriched disease, by contrast, autophagy is predominantly cytoprotective. Conditional deletion of Atg5 or Atg7 in a murine myeloid leukemia model cleared leukemia-initiating cells and prolonged survival, and cytarabine activated autophagy whereas Atg7 loss sensitized leukemic cells to chemotherapy. In human AML, AMPK/FIS1-mediated mitophagy is required for primitive-cell self-renewal, with FIS1 or AMPK depletion impairing mitophagy, promoting differentiation and reducing serial engraftment while relatively sparing normal hematopoietic progenitors. Pharmacological blockade with bafilomycin A1 preferentially targeted hypoxic, stem-like AML blasts by disrupting mitochondrial homeostasis and PINK1-dependent mitophagy, functional dissection identified OPTN as indispensable in AML, and a small-molecule inhibitor of the p62 ZZ domain attenuated leukemia-initiation potential. These studies establish LSC-selective mitophagy and receptor-mediated autophagy, but none tested BAG3 dependence.

Therapeutic stress can shift the balance further toward protective autophagy, and the review maps this crosstalk across recurrent AML driver mutations. Oncogenic FLT3-ITD sustains basal autophagy through an ATF4-dependent transcriptional program required for blast survival, identifying autophagy as a non-oncogene addiction in this subtype. Upon treatment with FLT3 inhibitors, leukemic cells mount a protective autophagic response that contributes to resistance, and autophagy inhibition restores sensitivity in vitro and in vivo; integrated translatome-proteome profiling independently identified autophagy as an on-target resistance mechanism to FLT3 inhibition. In NPM1-mutated leukemia, the cytoplasmic mutant maintains ULK1 protein abundance through TRAF6-dependent K63-linked ubiquitination to promote autophagic survival, and the glycolytic enzyme PKM2 drives Beclin1-associated autophagic activation to support cell survival. Critically, both lines of evidence engage the ULK1 and Beclin1 machinery but currently appear independent of demonstrated BAG3 involvement, defined through TRAF6 and PKM2 rather than through the BAG3-HSP70 module. The TP53 connection is hypothesis-generating: wild-type p53 represses BAG3 transcription under metabolic stress in non-hematopoietic cells, and recent solid-tumor work reports that acetylation promotes mutant TP53-HSPA8 and HSPA8-BAG3 interactions, routing mutant p53 to lysosomes preferentially through CASA. If chaperone inhibition stabilizes mutant p53 in TP53-mutant AML, it could paradoxically be detrimental, a possibility that must be tested before any genotype-specific use.

The therapeutic toolkit has broadened considerably, and the review stratifies direct axis-targeting from broad HSP70-family or proteostasis approaches. VER-155008, an adenosine-derived ATP-competitive inhibitor binding the HSP70 nucleotide-binding domain, inhibited proliferation and induced apoptosis in primary human AML blasts, and its combination with the HSP90 inhibitor 17-DMAG produced additive antileukemic effects without the compensatory heat-shock induction seen with HSP90 inhibition alone. PES, or pifithrin-μ, covalently modifies cysteine residues in the HSP70 substrate-binding domain and also disrupts lysosome integrity and autophagic flux; in AML and acute lymphoblastic leukemia cell lines it showed IC50 values of approximately 2.5 to 12.7 micromolar, with a median IC50 of about 8.9 micromolar in primary AML blasts and relative sparing of normal hematopoietic cells, and it sensitized leukemic cells to cytarabine, 17-AAG, vorinostat and sorafenib. The covalent HSP70 inhibitor QL47 reacts with Cys267 near the ATP pocket, promotes degradation of client proteins, triggers degradation of FLT3-ITD itself, overcomes resistance conferred by FLT3 tyrosine-kinase-domain mutations, and achieved 76.64 percent tumor growth inhibition at 10 milligrams per kilogram per day in a subcutaneous MV-4-11 xenograft model. The most direct axis-selective route, however, is the allosteric HSP70-BAG3 protein-protein interaction inhibitor JG-98, which disrupts HSP70-BAG3 binding with EC50 values of roughly 0.3 to 4 micromolar across non-AML cancer cell lines and destabilizes clients such as FoxM1, yet has never been evaluated in AML, and pathway profiling showed that some JG-98 responses, including changes in AKT and MYC, are BAG3-independent.

Beyond small molecules, the review surveys BAG3-directed agents, targeted protein degradation and delivery innovation. Small molecules that trigger CRBN-dependent BAG3 degradation have been described in non-AML settings, though AZD7762 is a pleiotropic kinase inhibitor and has not established a selective BAG3-degrader mechanism or AML efficacy. The terminology of degraders must be precise: the reported HSP70-interactome-mediated PROTAC recruits Hsc70/HSP70 machinery to degrade proteins such as GPX4 or BRD4; it does not degrade HSP70 itself. A carrier-free, peptide-directed nanoassembly co-delivering daunorubicin and an siRNA against the AML target LILRB4 showed targeted activity in THP-1 cells and an AML mouse model, illustrating a platform that could in principle carry BAG3 or HSP70-directed cargo to leukemic cells while limiting cardiac exposure, though application to the axis itself remains hypothetical. Immunotherapy is an emerging but currently indirect direction: secreted or extracellular BAG3 activates macrophages through IFITM2 in pancreatic ductal adenocarcinoma, CAR-NK cells targeting membrane HSP70 are in preclinical development, and autophagy is coupled to MHC-I antigen presentation in solid-tumor models, but a specific BAG3-dependent AML component of these immunological links is not established.

The translational challenges are as stark as the mechanistic ones. BAG3 has essential functions in cardiomyocytes, where it maintains proteostasis, mitochondrial function and inflammasome regulation, and BAG3 deficiency or mutation causes cardiomyopathy and heart failure; BAG3 additionally regulates mitophagy and TGF-beta receptor signaling in cardiac fibroblasts, so systemic BAG3 targeting carries a cardiac-safety liability that may require cardiac-sparing or leukemic-directed delivery. Normal hematopoietic stem cells offer a possible basis for selectivity, since Atg7 is essential for their maintenance, FOXO3A directs a protective autophagy program in them under stress, and notably normal HSCs preferentially traffic misfolded proteins through aggresomes and aggrephagy rather than CASA, but that distinction is not yet evidence that BAG3-CASA inhibition will spare normal cells. Patient selection and biomarker development are further priorities: HSP70 overexpression and circulating HSP70 have been associated with adverse features and tumor burden in AML, and prospective biomarkers should include basal and drug-induced autophagic flux measured as flux, HSF1 target induction, BAG3 and HSP70 paralog expression, the LSC and minimal residual disease fraction, and FLT3, NPM1 and TP53 genotype. No HSP70 or BAG3 inhibitor has yet entered AML clinical practice, and the authors argue that trials should be designed around biomarker-enriched, relapsed-refractory and LSC-high populations. Whether the BAG3-HSP70 axis can be converted from a plausible mechanism into a precise, biomarker-driven therapeutic target will now depend on CRISPR-based BAG3 knockout in patient-derived AML and xenograft models with rescue, direct flux measurements with BAG3 perturbation, and structural or biochemical tests of the three proposed couplings, which should either be demonstrated or explicitly retired from the mechanistic model.

Subject of Research: BAG3-HSP70 axis-mediated autophagy in acute myeloid leukemia

Article Title: Research progress on the BAG3-HSP70 axis-mediated autophagy in acute myeloid leukemia

Article References: Liu, Y., & Wu, T. (2026). Research progress on the BAG3-HSP70 axis-mediated autophagy in acute myeloid leukemia. Frontiers in Cell and Developmental Biology, 14, Article 1980188. https://doi.org/10.3389/fcell.2026.1980188

Image Credits: AI Generated

DOI: 10.3389/fcell.2026.1980188

Keywords: Research, progress, BAG3-HSP70, axis-mediated, autophagy, acute, myeloid, leukemia, scientific research