ATG4D role in mAtg8s delipidation and neuroprotection

The Atg4-Atg8 system is an important piece of the molecular machinery of macroautophagy/autophagy in yeast. Atg4 protease catalyzes the initial activating cleavage of Atg8, which is required for the binding of this protein to a molecule of phosphatidylethanolamine at the phagophore. Once at the phagophore membrane, Atg8 drive its expansion and also mediates the selective recognition of autophagic substrates. Additionally, Atg4 also deconjugates Atg8 from autophagosomal membranes when it is no longer required so it can be reincorporated into new nascent autophagosomes. In mammalian cells, four different ATG4 proteases (ATG4A to ATG4D) and multiple Atg8-like proteins have been identified, which suggests that these molecules may have distinct roles in autophagy. In fact, two Atg8 subfamilies (MAP1LC3 and GABARAP) have been described, acting at different stages of autophagosome formation and maturation. As for the ATG4 proteases, an overlapping redundancy of their activity in cell-free systems has been proposed, although the specific roles for each of them in living cells remain unclear. Driven by these questions, we have generated and characterized mice lacking ATG4D to better understand the in vivo contributions of this protein to autophagy, cell homeostasis and disease [1].

atg4d^−/- cells and tissues show markers linked to a blockage in autophagosome degradation or transport, such as the accumulation of both SQSTM1/p62 and the lipidated forms of Atg8-like proteins. Moreover, these alterations are associated with the presence of a higher number of autophagosomes, which are reduced in size in comparison to those in control cells. Surprisingly, specific experiments using lysosomal degradation inhibitors unequivocally show that ATG4D loss does not block autophagic flux. Moreover, flow cytometry-based analyses showed that starvation-induced degradation of ectopically-expressed GFP-LC3 is comparable in atg4d^−/- and WT cells. Consistently, autophagy-dependent degradation of long-lived proteins is not altered by ATG4D loss. Intrigued by these unusual autophagy alterations, we hypothesized that the observed increase in membrane-bound forms of mAtg8s could be due to defects in their delipidation upon ATG4D loss. By developing two complementary strategies (pH-dependent detection of mKeima-LC3B and fluorescent labeling of the accessible pool of LC3B upon selective permeabilization of the plasma membrane) we could demonstrate that cells lacking ATG4D accumulate LC3B at the cytosolic leaflet of the external membrane of autophagosomes and autolysosomes (Figure 1). Thus, these analyses yielded two interesting conclusions: (1) mAtg8 deconjugation is not necessary for autophagosome-lysosome fusion and (2) ATG4D loss leads to defects in mAtg8 delipidation from autophagosomal and autolysosomal membranes in living cells. This defect is not observed in cells lacking either ATG4A, ATG4B or ATG4C, even when the delipidation of a C-terminal deletion mutant of LC3B (MYC-LC3B[ΔC22]) that bypasses the initial proteolytic cleavage was analyzed. Together, these results show that ATG4D is the main mAtg8 delipidating enzyme in vivo. However, the reasons linking ATG4D loss to increased autophagosomal content, reduced autophagosome average size and the rise in SQSTM1 levels when autophagic flux is unaffected remain yet to be elucidated.

Figure 1.

Scheme depicting the consequences of ATG4D loss. Cells deficient for ATG4D protease show increased levels of membrane-bound mAtg8s and a higher number of smaller autophagosomes, without noticeable impact on autophagic flux. In mice, ATG4D loss leads to alterations in GABA_A receptor trafficking and clustering, progressive Purkinje cell loss and cerebellar ataxia

Through the analysis of atg4d^−/- mice and cells, we have previously shown that the initial cleavage of mAtg8 proteins is mainly carried out by ATG4B. Indeed, ATG4B loss leads to a partial inhibition of autophagic flux due to a reduction in the rate of autophagosome formation. In a similar way that mAtg8 priming is mainly carried out by a single ATG4 protease (ATG4B), our recent study shows that mAtg8 deconjugation from autophagic membranes is predominantly performed by ATG4D in mammalian cells. It is interesting that these two roles, which are played by the only Atg4 protease in less complex eukaryotes, are preferentially carried out by distinct members of the ATG4 family in more complex organisms. This could be one of the reasons underlying the increased complexity of the ATG4-mAtg8 system through evolution, although how these two processes are timely and spatially regulated remains to be solved.

At the organismal level, atg4d^−/- mice are viable and show no major histological alterations at a young age. However, they show progressive loss of Purkinje cells (PCs) accompanied by motor coordination defects characteristic of cerebellar ataxia. Surprisingly, atg4d^−/- PCs do not die through apoptosis, but display ultrastructural features resembling those described during dark cell generation, a type of neuronal death normally associated with excitotoxicity. As GABARAP is an essential effector in GABA_A receptor trafficking, we hypothesized that altered dynamics of this mAtg8 by ATG4D deficiency could affect the targeting of GABA_A receptors to the membrane, likely having an impact on GABA signaling. In fact, GABARAP interacts with GABRG2/GABA_A receptor γ2 subunit. Our experiments revealed that ATG4D loss leads to an increase in this interaction and that a high percentage of GABA_A receptors are targeted to lysosomal degradation in atg4d^−/- neurons. This defect alters the subcellular localization of different GABA_A receptor subunits, reducing their presence both at synaptic clusters and at extrasynaptic locations. The consequence of this alteration is likely a decrease of GABA-mediated inhibitory inputs, which could be the basis of PC damage and ultimately cerebellar ataxia and neurodegeneration (Figure 1), as PCs are highly sensitive to excitotoxicity. This hypothesis is supported by the fact that treatment with gaboxadol/THIP, a GABA_A receptor agonist, can rescue the impaired motor coordination of atg4d^−/- mice even at a young age, before the onset of neurodegeneration.

The evolutionarily conserved protective role for ATG4D against neurodegeneration is highlighted by the existence of two human gene variants linked to cerebellar ataxia in a patient with ATG4D compound heterozygosity, (http://neurodegenerativeatg4d.blogspot.com). Interestingly, both mutations interfere with the two mAtg8s docking sites of ATG4D, and the and the variant p.Y280C fails to completely revert the accumulation of lipidated mAtg8s in atg4d-deficient cells, hinting at a possible link between the roles of ATG4D in mAtg8 deconjugation and neuroprotection in humans. More studies are undoubtedly required to clarify how, if so, the accumulation of conjugated mAtg8s by ATGD deficiency leads to neurodegeneration, as well as fully understand the different roles of ATG4 proteases in vivo.

Funding Statement

This work was supported by grants from Ministerio de Economía y Competitividad (Spain) (BFU2015-68539-R), Principality of Asturias Government (IDI/2018/000159)

Disclosure statement

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