Proteaphagy in mammalian cells can function independent of ATG5/ATG7

. The degradation of intra- and extracellular proteins is essential in all cell types and mediated by two systems, the ubiquitin-proteasome system (UPS) and the autophagy-lysosome pathway. This study investigates the changes in autophagosomal and lysosomal proteomes upon inhibition of proteasomes by bortezomib (BTZ) or MG132. We find an increased abundance of more than 50 proteins in lysosomes of cells in which the proteasome is inhibited. Among those are dihydrofolate reductase (DHFR), ß-Catenin and 3-hydroxy-3-methylglutaryl-coenzym-A (HMGCoA)-reductase. Since these proteins are known to be degraded by the proteasome they seem to be compensatorily delivered to the autophagosomal pathway when the proteasome is inactivated. Surprisingly, most of the proteins which show increased amounts in the lysosomes of BTZ or MG132 treated cells are proteasomal subunits. Thus an inactivated, non-functional proteasome is delivered to the autophagic pathway. Native gel electrophoresis shows that the proteasome reaches the lysosome intact and not disassembled. Adaptor proteins, which target proteasomes to autophagy, have been described in Arabidopsis, Saccharomyces and upon starvation in mammalians. However, in cell lines deficient of these proteins or their mammalian orthologues, respectively, the transfer of proteasomes to the lysosome is not impaired. Obviously, these proteins do not play a role as autophagy adaptor proteins in mammalian cells. We can also show that chaperone-mediated autophagy (CMA) does not participate in the proteasome delivery to the lysosomes. In autophagy-related (ATG)-5 and ATG7 deficient cells the delivery of inactivated proteasomes to the autophagic pathway was only partially blocked, indicating the existence of at least two different pathways by which inactivated proteasomes can be delivered to the lysosome in mammalian cells. Stable amino lysosomes with magnetic beads : All SILAC reagents were obtained from Thermo United and For lysosomal proteome analysis, HEK293 cells were cultivated for six passages in SILAC-DMEM supplemented with 10% fetal bovine serum containing either 87.8 mg/ml l-arginine HCl, 181.2 mg/ml l-lysine for light labeling of cells or l-arginine 13 C 615 N 4 and l-lysine 13 C 615 N 2 for heavy labeling of cells. The isolation of lysosomal fractions with magnetic beads was performed as described recently 13 . % FA) in 60 min. Peptides eluting from the column were ionized in the positive ion mode using a capillary voltage of 1600 V and analyzed using a Thermo Orbitrap Velos mass spectrometer. One survey scan at a mass range of m/z 400 to m/z 1200 and a resolution of 30.000 was acquired in the Orbitrap mass analyzer followed by fragmentation of the 10 most abundant ions in the ion trap part of the instrument. The repeat count was set to one and the dynamic exclusion window to 60 s.


Introduction
Two distinct mechanisms, the autophagy-lysosome pathway and the UPS, mediate protein degradation within the cell. Lysosomes are membrane-bound organelles mediating the degradation of intracellular macromolecules like proteins, lipids and oligosaccharides.
Endocytosed cargo and autophagocytosed cell components 1 are degraded by a wide array of acid hydrolases. The delivery of substrates to the lysosomal lumen is mostly accomplished by vesicle transport, through endosomes and phagosomes for extracellular material or autophagosomes for the degradation of intracellular material. Alternatively, lysosomal substrates can be delivered by CMA or endosomal microautophagy 2 . The generation of autophagosomes starts by the formation and growth of a double-layered isolation membrane enabled by specialized autophagy proteins like the microtubuleassociated protein light chain 3 (LC3) in the cytoplasm, engulfing cargo that is to be degraded 3 . This engulfment can be mediated by adaptor proteins for specific cargo in case of selective autophagy, as has been shown for ribosomes in case of the nuclear FMR1interacting protein 1 (NUFIP1) 4 or nuclear receptor coactivator 4 (NCOA4) 5 for ferritin degradation, respectively. Often selective autophagy of a specific cargo is preceded by its ubiquitination, which in turn enables attachment to ubiquitin-binding domain-containing proteins like sequestosome-1/p62 (SQSTM1) or Neighbor of BRCA1 gene 1 protein (NBR1).
After closing of the autophagosomal membrane, autophagosomes and lysosomes fuse by the action of several proteins like soluble N-ethylmaleimide-sensitive-factor attachment receptors (SNARES) 6 and the inner autophagosomal membrane including its content is degraded by lysosomal hydrolases.
In contrast to the organelles lysosomes and autophagosomes, proteasomes reside as 2.5 mega Dalton protein complexes within the cytoplasm. They are composed of the 20S core complex containing 4 hetero-hepteromeric rings, each containing either seven α-or βsubunits. Six catalytic subunits in total, two of each β 1 , β 2 and β 5 subunits with caspase-like, trypsin-like and chymotrypsin-like proteolytic properties mediate the cleavage of the peptide bonds within the barrel of the 20S core complex. The 19S regulatory complex can be further by guest on  divided into lid and base parts, and its subunits mediate e.g. binding of ubiquitinated substrates and the removal of ubiquitin before entry of the substrate into the narrow barrel of the 20S proteasome 7 . Proteasomes exclusively degrade proteins and due to their structure can only do so when proteins are unfolded and selectively labeled with polyubiquitin chains on their lysine residues. This ubiquitination is mediated by a sequential action of the E1ubiquitin-activating enzymes, which activate the C-terminal Gly residue of ubiquitin and then transfers it to the E2-ubiquitin carrier family. From there it is attached to a lysine residue of the target protein by highly specific members of the E3-ubiquitin ligase family 8 . While initially the autophagy/lysosome system and the proteasome were regarded as two independent systems within the cell, in recent years it has become increasingly obvious that those systems are interconnected as both can degrade ubiquitinated substrates. Many proteins have been shown to be degraded by both pathways, and an increased autophagic protein degradation in case of proteasome impairment indicates compensatory mechanisms 9 .
In this study, we analyzed the impact of proteasomal inhibition on the composition of the proteome of lysosomes and autophagosomes. The activity of the proteasome was inhibited by either BTZ or MG132, and the composition of a lysosome-enriched fraction was analyzed by mass spectrometry. Among other proteins this revealed an increased presence of proteasome complexes within the lysosomal compartment where they are degraded. An indepth investigation to specify the detailed mechanism revealed that inactivated proteasomes are most likely degraded by macroautophagy and not by chaperone-mediated autophagy, but do still reach lysosomes when either ATG5 or ATG7 are deficient. This process is neither Sample preparation for mass spectrometry: For proteome analysis of lysosomal, autophagosomal and proteasomal fractions, samples were processed for mass spectrometry analysis using in gel digestion as described previously 13 . In brief, isolated fractions were denatured and alkylated with acrylamide 15  Waltham, United States for 1h. For immunostaining cells were exposed to appropriate primary antibodies in 1% FCS in PBS overnight at 4°C. Cells were washed thrice in PBS and incubated with appropriate fluorescence dye-coupled secondary antibodies for 1h at room temperature. After two times washing in PBS and once in water cells were mounted in DAPIcontaining mounting solution. Cells were imaged with either Zeiss LSM700 confocal microscope or Zeiss LSM800 with airyscan processing. Wild type and knockout cell samples were exposed to laser for identical time spans. Dynamic ranges were adjusted for optimal color representation.
Lysosome counting: We quantified lysosome structures from separate microscopic images, each containing one single cell. By using CellProfiler software (3.1.8) we detected individual lysosome structures by applying gaussian smoothing (sigma=1), automatic thresholding (minimum cross entropy method) and an intensity-based watershed for separating touching objects (sigma=12 for gaussian smoothing). For each detected lysosome object, the area (nr of pixels) and the average marker intensity was calculated. Objects with a diameter below 10 pixels were rejected.

Analysis of proteomic changes in lysosome-enriched fractions after proteasome inhibition
To determine changes in the lysosomal protein composition caused by an impairment of the ubiquitin-proteasome system, a SILAC-based mass spectrometric analysis 18

Known macroautophagy adaptor proteins and CMA are not involved in proteaphagy
As specific autophagy usually relies on binding of the cargo by specific adaptor proteins, we aimed to identify potential adaptor proteins. Known adaptor proteins that have been identified previously have shown enhanced association with cargo under degradationinducing conditions, as reported for the proteaphagy receptors Rpn10 in Arabidopsis thaliana 21 , coupling of ubiquitin conjugation to ER degradation protein 5 (Cue5) in yeast 23 , and SQSTM1 in human cells 24 .
It was reported earlier that SQSTM1 may be involved in degradation of mammalian proteasomes under nutrient stress conditions 24 , and this may also hold true for inactivated proteasomes. We generated HEK CRISPR-Cas knock-out cell lines for SQSTM1 In isolated autophagosomes, SQSTM1 was present but showed a BTZ/control ratio of 0.9 and was therefore not enriched together with proteasomes.
In yeast, the aggrephagy receptor protein Cue5 was shown to bind ubiquitinated proteasomes accumulated in cytoplasmic aggregates upon MG132 treatment, and deletion of Cue5 abolished transport of proteasomes to the vacuole 23 . The mammalian homologue of Cue5 is the toll-interacting protein (TOLLIP), a cytoplasmic soluble protein interacting with by guest on  proteins of the Atg8 family 25 . We generated TOLLIP CRISPR-Cas knockout cells (Supplementary Figure 3), but in western blots of isolated lysosome-enriched fractions we could not detect differences in the protein expression levels of PSMA7 or PSMB5 (Figure 6 B) in comparison to wildtype cells. As PSMD4, the human orthologue of Arabidopsis Rpn10, does not possess the domain necessary for proteaphagy in Arabidopsis and its role in mammalian proteaphagy was already excluded by Marshall and colleagues 21 , we did not include it in our analysis.
Another pathway to deliver proteins to the lysosome is chaperone-mediated autophagy (CMA), and it has been suggested that proteasomes may be delivered to lysosomes via this mechanism 20 . Since CMA depends on the lysosomal membrane protein LAMP2, we investigated a possible involvement of CMA in LAMP2-deficient mouse embryonic fibroblast (MEF) cells 26 . We incubated the cells with BOD-TMR-Epox and analyzed the accumulation of the inhibited proteasomes by microscopy. No differences were observed between wildtype and LAMP2-deficient cells (Figure 6 C). Thus, CMA is not involved in delivery of proteasomes to lysosomes.

No new proteaphagy adaptor protein could be identified
To identify yet unknown possible adaptor proteins we isolated proteasomes by anti-PSMB5 affinity purification and compared those proteins associated to the isolated proteasomes with and without proteasome inhibition by MG132 (Supplementary Table S8 with a dataset generated by Besche and colleagues 27 , who isolated proteasomes from BTZtreated and non-treated HEK293 cells expressing tagged proteasome subunits. Nine proteins were commonly upregulated in both datasets (Supplementary Table S8), among them the RING-type E3 ubiquitin transferase RNF181 and the proteasome activator complex subunit 3.

Discussion
The connections of the UPS and the lysosome-autophagy pathways are numerous and diverse. Both systems can partially compensate each other, as it has been extensively shown that inhibition of the proteasome can lead to upregulation of ATG genes.
Pharmacological enhancement of 20S proteasome activity impairs autophagic flux by increasing degradation of Synaptosomal-associated protein 29 and Syntaxin-17 28 . The present study revealed even more interconnections between both pathways.
Proteotoxic stress conditions can lead to a shift from proteasomal to autophagosomal degradation mediated by BAG family molecular chaperone regulator 3 29 . Proteins that would normally be degraded by the UPS can be deflected to autophagic degradation upon inhibition of the proteasome by either MG132 or BTZ. Dihydrofolate reductase is a wellknown substrate of the proteasome [30][31][32] . Its degradation is obviously re-routed to the lysosome as the proteasome is blocked, but it could not be identified in our mass spectrometric analysis of autophagic vesicles. As it has been described to be able to enter the lysosome by CMA 33 , this is likely its route of transport into the lysosomal lumen for degradation. Β-catenin degradation has been extensively described and is usually performed by the proteasome 34,35 . In our analysis it is not only enriched in lysosomes from MG132-and BTZ-treated cells, but also in isolated autophagosomes. This is in line with an earlier study by Petherick and colleagues 36 , who have described a direct interaction of βcatenin with LC3 followed by autolysosomal degradation.
The autophagic degradation of proteasomes by lysosomes has initially been described in Arabidopsis thaliana 21 , where Rpn10, a subunit of the proteasome itself and whose human homologue is PSMD4, has been elegantly identified as the responsible adaptor protein by guest on  which binds to ubiquitinated proteasomes and Atg8 with ubiquitin-interacting motifs.
Additionally to being part of the proteasome complex, it exists as a cytosolic pool binding directly to Atg8 and proteasomes upon their inhibition. As the region identified as being responsible for this process is not present in the yeast/mammalian counterparts of Rpn10, and it has been shown to not or only weakly bind to yeast and human versions of Atg8, its function as a proteaphagy adaptor is probably limited to plants 21,37 .
In yeast, two general pathways of proteasome clearance can be distinguished: either by nutrient stress conditions like nitrogen starvation, or induced by proteasome inhibitor treatment. The latter seems to be a stepwise process starting with deposition of inactivated proteasomes in cytoplasmic aggregates followed by autophagic degradation 23,38 .
In mammalian cells, the degradation of proteasomes also seems to follow different pathways. While the degradation of misassembled proteasome subunits is performed preferentially by the proteasome itself 38 , the degradation of whole proteasome by autophagy during steady-state and starvation conditions was documented in several studies 39,40 . Our results clearly show that also proteasomes inactivated by the inhibitors MG132 or BTZ are accumulated in autophagosomes and subsequently in lysosomes. Proteasome subunits were identified in a proteomic analysis of autophagosome-associated proteins 39 , and 19S and 20S subunits showed reduced degradation rates in autophagy deficient cells lacking ATG5 or ATG7 40 . The degradation of proteasomes by autophagy induced by prolonged starvation periods in HeLa cells was shown to be at least partially dependent on SQSTM1, whose siRNA knockdown led to a 25% reduction in colocalization between the proteasome and LC3B 24 . This could not be verified by our study regarding inhibitor-induced proteaphagy, as we did not find any differences in the degree of proteasome accumulation in lysosomes after inhibitor treatment in SQSTM1 deficient cells. Therefore, the shuttling of proteasome subunits to the autophagosome by SQSTM1 may be restricted to nutrient-stress induced degradation of proteasomes.
The same study also reported an ubiquitination preceding proteaphagy, a mechanism that is probably common to both pathways as Besche and colleagues 27  that has been implicated in aggrephagy and lipophagy 25,44 does, according to our results, not seem to be responsible for delivery of inactivated proteasomes to the autophagosome.
Nevertheless, a role of TOLLIP or SQSTM1 in proteaphagy can still not completely excluded, as other macroautophagy adaptors could substitute for them in the respective knockout cells. Furthermore, it should be noted that also nuclear proteins could be involved in proteasome delivery to the lysosomes, which could not be detected by our study as we used postnuclear supernatants for the isolation of organelles.
Dengjel and colleagues investigated the possibility of proteasome binding directly to LC3, but no direct interaction between proteasome subunits and LC3 39 could be detected. Direct interactions between those proteins upregulated after proteasome inhibition and LC3 were also not reported so far, but could be investigated by immunoprecipitation or immunomagnetic separation.
Interestingly, in the model organism Dictyostelium discoideum it could recently be shown that the 19S subunits PSMD1 and PSMD2 directly interact with ATG16, a component of the core autophagy machinery, and are degraded by autophagy dependent on ATG16 45 . This interaction and the following autophagic degradation is dependent on the N-terminal portion of Atg16 containing an Atg5-interacting motif.
In conclusion, proteaphagy of inhibitor-inactivated proteasomes in human cells seems to follow an at least partially different pathway from that utilized by proteasomes in nutrient stress conditions. The conventional autophagy pathway probably mediates most of this degradation, but even if LC3-conjugation is blocked by ATG5 or ATG7 knockout, inactivated proteasomes still reach the lysosome by alternative pathways. The utilization of adaptor proteins for selective autophagy of the proteasome is species-specific, and there may very well be a direct binding of inactivated, probably ubiquitinated proteasomes to components of the forming autophagosome.

Data Availability
The mass spectrometry datasets have been deposited to the Proteome Xchange Consortium (27)