A distinctive form of autophagy induced by oncogenic RAS

ABSTRACT

RAS mutations enhance macroautophagy/autophagy in tumor cells, crucial for their growth and survival, making autophagy a promising therapeutic target for RAS-mutant cancers. However, the distinction between RAS-induced autophagy and physiological autophagy is not well understood. We recently identified a unique form of autophagy, RAS-induced non-canonical autophagy via ATG8ylation (RINCAA), which differs from starvation-induced autophagy. RINCAA is regulated by different sets of autophagic factors and forms structures distinct from the double-membrane autophagosome known as RAS-induced multivesicular/multilaminar bodies of ATG8ylation (RIMMBA). A key feature of RINCAA is the phosphorylation of PI4KB by ULK1, and inhibiting this phosphorylation shows superior effects compared to general autophagy inhibitors. This work suggests a potential for specifically targeting autophagy in RAS-driven cancers as a therapeutic strategy.

Macroautophagy (hereafter referred to as autophagy) is a conserved metabolic pathway in eukaryotic cells, crucial for maintaining intracellular homeostasis and cellular survival. Dysregulation of autophagy is associated with various diseases, including cancer, prompting efforts to target autophagy for cancer therapy. However, the clinical efficacy of general autophagy inhibitors has been limited due to two main challenges: (1) autophagy plays a dual role in cancer progression, and (2) inhibiting autophagy indiscriminately also affects physiological autophagy, which has beneficial effects for the host. Therefore, a more detailed understanding of the molecular mechanisms underlying autophagy dysregulation, particularly the differences between cancer-associated autophagy and physiological forms (e.g., starvation-induced autophagy), is needed.

The current study [1] established a cellular model to examine RAS mutation-induced autophagy. Through a systematic comparison of the molecular requirements for physiological autophagy, we identified a distinct form of autophagy induced by hyperactive RAS mutations, characterized by ATG8ylation. We term this process “RAS-induced non-canonical autophagy via ATG8ylation” (RINCAA). RINCAA shares significant overlap with physiological autophagy, involving key factors such as the ULK1 complex, lipidation machinery, WIPI2, and SNARE proteins. However, certain factors essential for starvation-induced autophagy, including the class III phosphatidylinositol 3-kinase (PtdIns3K), ATG9A, and ATG2, are not required for RINCAA. Notably, PI4KB substitutes for the class III PtdIns3K, generating phosphatidylinositol-4-phosphate (PtdIns4P) as a lipid signal to activate WIPI2 and downstream autophagic processes.

Due to the absence of key components involved in the biogenesis of typical double-membrane autophagosomes, such as the PtdIns3K complex, ATG9A, and ATG2, we observed the formation of multivesicular and multilaminar autophagosomes in RINCAA, marked by mammalian Atg8-family proteins (ATG8s). These structures are designated RAS-induced multivesicular/multilaminar bodies of ATG8ylation (RIMMBA), which likely serve as substitutes for double-membrane autophagosomes.

Signaling pathway analysis revealed that the RAS-MAPK/p38-ULK1-PI4KB-WIPI2 cascade regulates this process. KRAS activates the MAPK/p38 pathway, promoting ULK1 activation. ULK1, in turn, phosphorylates PI4KB at S256 and T263, enhancing PtdIns4P production. WIPI2, acting as a PtdIns4P effector, recruits the ATG12–ATG5-ATG16L1 conjugation complex to facilitate ATG8ylation and RIMMBA formation.

Importantly, PI4KB phosphorylation at S256 and T263 is upregulated in RAS-mutant cancer cells and in colorectal cancer patient samples, suggesting that ULK1-mediated PI4KB phosphorylation could serve as a promising therapeutic target for RAS-mutant cancers (Figure 1). Indeed, blocking PI4KB phosphorylation using a peptide (Peptide 1) that competes with ULK1-mediated phosphorylation effectively inhibits autophagy in RAS-mutant cancer cells and suppresses tumor growth in xenograft and KPC pancreatic cancer models.

Figure 1.

Models for physiological autophagy and RINCAA. (A) Under starvation-induced autophagy (a physiological form of autophagy), ULK1 is activated by inhibiting MTOR and meanwhile activating AMPK. ULK1 can phosphorylate the class III PtdIns3K complex. The PtdIns3K complex phosphorylates phosphatidylinositol (PtdIns) lipids to generate PtdIns3P, which recruits WIPI2. WIPI2 recruits the conventional ATGs to build the autophagosome, which maintains cell homeostasis. (B) RINCAA operates independently of several autophagic factors typically associated with starvation-induced autophagy. RAS activates the MAPK/p38 pathway, which promotes ULK1 activation. ULK1 selects PI4KB instead of PtdIns3K as the substrate; it phosphorylates PI4KB at S256 and T263 and enhances PI4KB activity, leading to PtdIns4P production, which recruits WIPI2. WIPI2 recruits the unconventional ATGs to build the RIMMBA, which promotes tumor progression.

Notably, Peptide 1 demonstrates superior antitumor efficacy and a prolonged lifespan in comparison to the general lysosomal inhibitor chloroquine when combined with the MAPK/ERK inhibitor trametinib. This enhanced efficacy is due to the RINCAA-specific inhibition, which promotes antitumor immunity by increasing MHC I levels. In RAS-mutant cancers, MHC I is degraded by autophagy, and the inhibition of RINCAA preserves immune cell (CD8⁺ T-cell) autophagy, leading to increased CD8⁺ T-cell infiltration into tumors-a process blocked by chloroquine. Therefore, targeting RAS-specific autophagy appears to be a more promising strategy for treating RAS-mutant cancers, which account for approximately 25% of all cancer cases. Furthermore, it would be valuable to explore whether RINCAA-specific inhibition could be combined with immune modulation strategies, such as PDCD1/PD-1 antibody treatment, as many RAS-mutant tumors are immune-suppressive.

A key feature of RINCAA is the formation of non-canonical autophagosomes, specifically characterized by the non-double-membrane RIMMBA. The mechanism underlying the formation of this structure warrants further investigation. ATG2 facilitates lipid transport during autophagosome expansion, and ATG9A functions as a scramblase in starvation-induced autophagosome formation. These factors are critical for the directed growth of typical double-membrane autophagosomes, and ATG9 deficiency leads to the formation of irregularly shaped membranes positive for ATG8ylation. Interestingly, in RINCAA, neither ATG2 nor ATG9A is essential, which may explain the absence of typical double-membrane autophagosomes. Instead, we found that VPS13B may substitute for ATG2, while TMEM41B or VMP1 might replace ATG9A in classical autophagosome formation. The coordination of these lipid transfer proteins and scramblases in RIMMBA formation remains to be explored.

Understanding the membrane origin of autophagosomes has been a critical area of research. In starvation-induced autophagy, a phagophore cradle is generated in the ER through calcium flux and RB1CC1/FIP200 phase separation, followed by membrane acquisition from multiple organelles, including the ER, ER-Golgi intermediate compartment (ERGIC), and plasma membrane. It remains to be determined which membrane sources contribute to RIMMBA formation. Previous studies suggest that PtdIns4P originates from the Golgi apparatus, raising the possibility that the Golgi, or other organelles such as the ERGIC, may contribute to RIMMBA formation. Further investigation is needed to elucidate the potential roles of VPS13B, TMEM41B, and VMP1 in RINCAA and their relationship to the membrane origin of RIMMBA.

Funding Statement

The work is funded by National Natural Science Foundation of China (92254302, 32225013, 32130023, 32450388, 32370728), National Key R&D Program of China (2021YFA0804802, 2024YFA1802600), Tsinghua University Dushi Program, New Cornerstone Science Foundation, China Postdoctoral Science Foundation (BX20240186 and 2024M761616), Shuimu Tsinghua Scholar Program.

Disclosure statement

No potential conflict of interest was reported by the author(s).