Cellular quality control by the ubiquitin-proteasome system and autophagy

See allHide authors and affiliations

Science  15 Nov 2019:
Vol. 366, Issue 6467, pp. 818-822
DOI: 10.1126/science.aax3769


To achieve homeostasis, cells evolved dynamic and self-regulating quality control processes to adapt to new environmental conditions and to prevent prolonged damage. We discuss the importance of two major quality control systems responsible for degradation of proteins and organelles in eukaryotic cells: the ubiquitin-proteasome system (UPS) and autophagy. The UPS and autophagy form an interconnected quality control network where decision-making is self-organized on the basis of biophysical parameters (binding affinities, local concentrations, and avidity) and compartmentalization (through membranes, liquid-liquid phase separation, or the formation of aggregates). We highlight cellular quality control factors that delineate their differential deployment toward macromolecular complexes, liquid-liquid phase-separated subcellular structures, or membrane-bound organelles. Finally, we emphasize the need for continuous promotion of quantitative and mechanistic research into the roles of the UPS and autophagy in human pathophysiology.

The widely adopted anthropomorphic view of cellular quality control often considers quality control to be a decision-making process that discriminates between normal and malfunctioning proteins or subcellular structures. Its purpose, similar to industrial quality control, is to “achieve and maintain the fitness for use of its products or services” (1). Indeed, mechanisms analogous to those of industrial quality control are found in cellular systems, such as surveillance, feedback control, and adaptation to altered conditions. However, cells do not operate on management systems that choose between quality control measures. Instead, the main cellular quality control systems for proteins and organelles, the ubiquitin-proteasome system (UPS) and autophagy, are controlled by biophysical properties and coordinated at multiple layers securing cellular well-being and appropriate responses to stress and damage.

Features of the UPS

The UPS and autophagy were both initially thought to act as unspecific bulk recycling routes. It is now recognized that they exhibit a high degree of specificity involving selective enzymatic reactions and discriminatory receptors, with modular domain structures that deliver substrates either to proteasomes or to lysosomes, respectively (Fig. 1). Substrate size critically influences pathway choice: The UPS mostly degrades single, unfolded polypeptides able to enter into the narrow channel of the proteasome, whereas autophagy primarily deals with larger, cytosolic structures such as protein complexes, cellular aggregates, organelles, or pathogens (2).

“Despite their distinct degradation mechanisms, the UPS and autophagy share molecular determinants…”

Fig. 1 Overview and comparison of UPS and autophagy.

Left: The UPS is characterized by the strong dependence on Ub as a degradation signal and on the size limit of substrates. Soluble single proteins are polyubiquitinated in an inducible and reversible manner. Lys48 (K48)–linked polyUb attached by the substrate-specific Ub E3 ligase is recognized either by intrinsic Ub receptors of the proteasome (e.g., Rpn10 or Rpn13) or shuttle factors that are equipped with both a Ub-binding domain and a domain that binds to the proteasome (e.g., Ubl, PB1). Before degradation in the inside of the barrel-shaped proteasome, substrates are deubiquitinated and unfolded. Right: Selective autophagy is able to degrade large and heterogeneous cytosolic material, including aggregated proteins, organelles, bacteria, and molecular machines. Substrate labels recognized by the autophagic machinery are more diverse and include Ub, lipid-based signals (e.g., galectins), or organelle-intrinsic autophagy receptors that become exposed to the cytosol. A growing phagophore engulfs autophagic cargo that is connected to the ATG8/LC3-decorated membrane via LIR motif–containing autophagy receptors. Eventually, the phagophore closes around the cargo to give rise to the autophagosome that finally fuses with the lysosome.

The UPS is predominantly driven by ubiquitin (Ub) as a degradation tag, which is controlled by multilayered, reversible enzymatic reactions. These reactions show specific kinetics for each substrate, resulting in protein half-lives ranging from seconds to hours. Ub conjugation is carried out by a hierarchically acting enzymatic cascade. First, adenosine triphosphate–dependent Ub activation has to occur (E1s, n ≥ 2 in humans), followed by transfer of a Ub thioester to a Ub-conjugating enzyme (E2s, n > 50) and formation of an isopeptide bond catalyzed by Ub ligases (E3s, n > 500 in humans). These latter enzymes are endowed with substrate specificity (2, 3). When substrates are progressively modified with Ub, either at the N terminus (Met1) or at a lysine side chain of Ub (Lys6, Lys11, Lys27, Lys29, Lys33, Lys48, Lys63), various linear or branched Ub chains are built. Thus, Ub tags can be diverse in structure and will dictate the outcome of protein modification, because they can recruit accessory factors or receptors harboring Ub-binding domains (UBDs), including shuttle factors that deliver substrates to the proteasome. Poly-Ub Lys48-linked and branched Lys48-Lys11 chains serve as the most potent signals for degradation by the proteasome. Monoubiquitination can alter protein localization and protein complex stoichiometries; however, recent systematic analyses have revealed that monoubiquitination can also target up to 40% of proteins for degradation by the proteasome (4), particularly proteins of 20 to 150 residues. Ub conjugation is readily reversible owing to the existence of numerous deubiquitinating enzymes (3).

Macroautophagy and its link to ubiquitin

The flux of substrates through the autophagy pathway is a multistep process that relies on protein-protein– and protein-lipid–driven interactions. The best-characterized form of autophagy is macroautophagy (hereafter called autophagy), which is characterized by the engulfment of cellular material by a double-membrane structure, the autophagosome. Autophagy is the major response to cellular stress, maintains metabolic building blocks during periods of nutrient deprivation, and eliminates unwanted cellular contents such as toxic protein aggregates, damaged organelles, or intracellular pathogens (2). Autophagy is initiated by the activation of the ULK1 kinase complex, which gives rise to a spatiotemporally highly restricted enzymatic cascade that promotes the local assembly of multiprotein complexes. These complexes promote the conjugation of members of the ATG8 protein family [in higher eukaryotes, LC3s (microtubule-associated protein 1 light chain 3) and GABARAPs (γ-aminobutyric acid receptor–associated protein)] to phosphatidylethanolamine or phosphatidylserine, thereby anchoring them into membranes.

ATG8s exposed on growing autophagic membranes can also sequester substrates delivered by selective autophagic receptors that contain LC3-interacting regions or GABARAP-interacting motifs (LIRs or GIMs, respectively). Very often, these receptors, such as p62/SQSTM1, NBR1, NDP52, OPTN, and TAX1BP1, are also equipped with UBDs, bridging ubiquitinated proteins and autophagic membranes (5). It is estimated that more than half of all selective autophagic substrates rely on Ub as an “eat me” signal, whereas other substrates are recognized in a Ub-independent manner (6).

Ubiquitin and p62 at the crossroads of the UPS and autophagy

Despite their distinct degradation mechanisms, the UPS and autophagy share molecular determinants and substrates. Accumulating evidence indicates a dynamic cross-talk between the two pathways, with Ub playing a role in most of the multiple layers of communication.

First, both systems use Ub as a degradation signal for substrates; thus, cellular Ub concentration as well as accessibility of Ub receptors both appear crucial. The most studied Ub receptor is p62/SQSTM1, which binds to ubiquitinated substrates via its Ub-associated (UBA) domain, to the proteasome via its Phox and Bem1p domain (PB1), and to autophagic membranes via its LIR domain (5). Hence, p62 can escort ubiquitinated substrates to the proteasome and can also act as an autophagy receptor recruiting Ub-conjugated substrates to autophagosomes (Fig. 2). Recent studies have revealed that competition between proteasomal and lysosomal degradation is governed by avidity determinants of a complex, rather than by individual affinity of receptors for Ub (7). Thus, bona fide UPS-dedicated receptors can be transformed into autophagy-dedicated receptors upon formation of oligomers. A switch of these receptors between the dimeric and oligomeric or filamentous state determines which pathway is used by p62-delivered substrates (8). This switch can also be controlled by Ub itself: Under conditions of Ub stress (e.g., at high concentrations of free Ub as observed upon heat shock or prolonged proteasome inhibition), ubiquitination can block UBA domain–dependent dimerization of p62. This favors oligomerization and substrate channeling to autophagy (9).

Fig. 2 Ub and p62 at the crossroads of UPS and autophagy.

The p62 protein can serve both as a selective autophagy receptor (via its LIR domain) and as a proteasomal shuttle factor (via its PB domain) for ubiquitinated substrates. The pathway choice depends on the oligomeric state of p62. Left: The p62 dimer recognizes single protein molecules labeled with K48-linked polyUb via its UBA domain and shuttles them to the proteasome via interaction of its PB domain. Middle left: Binding to K63-linked polyUb chains induces oligomerization of p62 via its PB domain and liquid-liquid phase separation, resulting in the formation of p62 bodies that subsequently recruit autophagic membranes via the LIR domain. Middle right: A high concentration of cellular Ub can lead to oligomerization of p62. The process involves ubiquitination of p62 at multiple lysines, resulting in destabilization of UBA dimerization and PB1-mediated oligomerization, which favors LIR-mediated autophagic targeting of polyubiquitinated cargo. Right: Oligomerization can also be induced independently of Ub: N-degrons bound by the ZZ domain of p62 promote formation of p62 oligomers stabilized by disulfide bridges between PB1 domains. Again, oligomeric p62 favors the autophagic route via LIR-mediated interactions.

Second, the proteasome itself can both be regulated by and targeted for autophagic degradation by ubiquitination. Depending on the trigger and the type of ubiquitination, ubiquitination can lead to autoinhibition of proteasomes (10) or autophagic degradation of proteasomes, also called proteaphagy (11). Proteaphagy is observed under chronic amino acid starvation and, at lower capacity, under basal metabolic conditions. This suggests that proteaphagy may act as a key mechanism for proteasome turnover (11). Proteaphagy was first described in plants and yeast, and it can be further regulated by compartmentalization upon specific stresses. Upon nitrogen starvation in yeast, proteasomes can be degraded by autophagy, whereas during carbon starvation, proteasomes are sequestered in proteasome storage granules where specific factors (including the HEAT-repeat protein Blm10 and the deubiquitinating enzymes Rpn11 and Ubp3/Bre5) seem to protect them from autophagy (12).

Third, the UPS and autophagy are also linked by various properties of p62 and Ub. For example, p62 can recognize proteins that expose positively charged or bulky N-terminal residues, so-called degradation signals (degrons) for the N-end rule pathway, which usually trigger proteasomal degradation (13). This binding triggers p62 aggregation through formation of disulfide bonds between p62 molecules, ultimately leading to autophagic degradation of the N-end rule substrate. This corroborates the notion that receptor oligomerization can determine pathway choice.

The interaction between the UPS and autophagy is exemplified well when studying nutrient shortage or upon inhibition of one of the two systems. Acute nutrient stress will lead to replenishment of the cellular amino acid pool via the proteasomal route before autophagy is activated (14), and cells will up-regulate proteasomal activity at both transcriptional and posttranslational levels (15). The critical importance of this fast response becomes evident upon proteasome inhibition, which eventually leads to cell death caused by amino acid deprivation (16). In fact, amino acid scarcity has been shown to be the critical signal that blocks amino acid consumption by protein synthesis and activates autophagy through an mTOR (mechanistic target of rapamycin)–dependent pathway. This results in the bulk degradation of cytosolic material in an attempt to rescue the amino acid pool (16). Furthermore, proteasome inhibition causes rapid, Nrf1-dependent induction of proteasomal subunits (15) as well as expression of p62 and GABARAPs, which help to sequester ubiquitinated proteins into inclusions (17). Similarly, mTOR signaling regulates autophagic gene transcription, mainly through phosphorylation and cytoplasmic retention of the MiT/TFE-type transcription factors TFEB, TFE3, and MITF. Upon nutrient stress, dephosphorylation of MiT/TFE transcription factors leads to their nuclear translocation and promotes the transcription of autophagy genes, Ub receptors (including p62), and lysosomal genes (18), also through Nrf1, which is a direct TFEB target gene itself (19).

Thus, Ub can act on several regulatory layers to affect substrate channeling, specifically at the level of Ub receptors and more generally at the level of proteasome activity. This latter type of regulation leads to a transcriptional feedback, which is still not well understood under physiological conditions. Moreover, autophagy is directly affected by adaptations in the UPS and is subject to an apparently more complex feedback regulation. For instance, the autophagy-initiating kinase ULK1 is targeted for proteasomal degradation during autophagy progression (20). Similarly, Beclin1, a component of the lipid kinase complex immediately downstream of ULK1, requires the activity of deubiquitinating enzymes for its maintenance at steady-state levels (21). Although these two examples require (de)ubiquitination, Ub-independent degradation of autophagy pathway components that mediate autophagosome-lysosome fusion has also been reported (22). Further study of the cross-talk and feedback between the UPS and autophagy will most likely reveal an even greater level of mutual control.

Quality control of cytosolic membraneless structures

As discussed above, substrate processing through the UPS or autophagy critically depends on the biophysical state of the substrate. Although the focus has long been on the two states of soluble versus aggregated, more recently it has become clear that many proteinaceous, membraneless subcellular structures are better described by a third biophysical state, a phase-separated condensate assembled by liquid-liquid phase separation (LLPS). Such structures include the centrosome, the nucleolus, Cajal bodies, P-bodies, and stress granules. During LLPS, a stimulus-responsive condensation drives the spontaneous separation of fluid- or gel-like phases when one phase is composed of molecules that harbor intrinsic, associative properties. Such molecules must be multivalent, either harboring multiple adhesive domains or linear motifs or harboring polyvalent, intrinsically disordered regions (23). LLPS can have outcomes that are hard to predict, and molecules showing LLPS need to be evaluated systematically to uncover the stimulus or regulatory mechanisms that lead to condensation.

Recently, such systematic studies have unraveled the importance of LLPS for the UPS and autophagy. For example, the ability to oligomerize influences the recruitment of a Cul3 Ub ligase complex to liquid-like nuclear speckles. Self-association of the ligase is induced by binding of its substrate adaptor SPOP to substrates, which affects localization of the ligase and its activity (24). In stress granule formation, autophagy pathway components can co-partition with stress granules. LC3s/GABARAPs, in particular, show a high degree of direct interactions with other granule components (25). In addition, dynamic aspects and a cross-talk between the UPS and autophagy have been documented for LLPS. The ability of p62 to oligomerize and the multivalence of poly-Ub chains recognized by it can drive both phase separation and segregation of the substrates into autophagic degradation (26). In contrast, for proteasomal Ub receptors such as ubiquilins, partitioning through LLPS is inhibited by binding to mono-Ub or Ub chains, which block multivalent interactions between ubiquilin moieties (27). Ubiquilins can affect autophagy in more indirect ways (most likely not requiring LLPS) by regulating lysosomal acidification (28). Hence, to evaluate dynamic changes and steady states of proteasomal versus autophagy-dependent degradation of stimulus-responsive condensates, a proper characterization of their biophysical properties will be crucial. In most of these systems, the physiological relevance of LLPS in vivo still needs to be addressed.

“…there is accumulating evidence that autophagy and the UPS are simultaneously affected and may mutually affect each other in a variety of diseases.”

Quality control of organelles

Quality control of damaged or superfluous organelles is largely driven by selective autophagy (also called organellophagy) (Fig. 3). This is a critical step in cellular surveillance because multiple cellular functions depend on proper functioning of organelles. Exhausted or damaged organelles are recognized by dedicated autophagy receptors that are localized in organelle membranes or are recruited from the cytosol. These receptors harbor LIR or GIM motifs that facilitate incorporation into nascent autophagosomes. Autophagy-mediated turnover of mitochondria (mitophagy) is a particularly well-studied example. Mitophagy occurs during different physiological situations that rely on different types of autophagy receptors: The family of NIX/BNIP3 mitochondrial autophagy receptors promote mitophagy during development or erythrocyte maturation independently of Ub signals (29). In contrast, mitophagy of damaged or depolarized mitochondria relies on ubiquitination of several proteins of the outer mitochondrial membrane driven by the PINK1-parkin pathway (30). UBD-containing autophagy receptors, including OPTN, NDP52, p62, NBR1, and TAXBP1, link ubiquitinated mitochondria with autophagic membranes (31, 32). In addition, OPTN and NDP52 can also recruit ULK1 directly to ubiquitinated mitochondria to initiate autophagosome formation locally (31). Such spatiotemporal control of ULK1 activation is dependent on the TBK1 kinase, but not on AMPK (adenosine monophosphate–activated protein kinase) or mTOR activities or the presence of ATG8/LC3, and is common for multiple selective autophagy processes, including, for example, bacterial clearance (xenophagy) (33).

Fig. 3 Summary of Ub dependency of organellophagy.

Depending on the type of organelle, selective autophagic degradation of exhausted or damaged organelles may involve both Ub-dependent and -independent autophagy receptors and may also involve proteasomal activity.

Another recently characterized example of organellar quality control is the autophagic degradation of topologically distinct domains of the endoplasmic reticulum (ER), or ER-phagy. Deficiencies in ER-phagy have been linked to several human diseases, including neuropathies (34). Several ER-phagy receptors have been identified that mediate the delivery of ER tubules (RTN3, ATL3), ER tubular junctions (TEX264), ER-luminal components (CCPG1), and ER sheets (FAM134B) to autophagic membranes (34). ER-phagy is also indirectly linked with UPS functions because it degrades substrates of the ER-associated degradation (ERAD) pathway. FAM134B, through the cooperation with the ER chaperone calnexin, mediates the quality control–dependent turnover of ER luminal proteins that are refractory to ERAD, such as misfolded procollagen (35) or α1-antitrypsin Z (ATZ) (36). A recent study has revealed the requirement for some components of the COPII secretory pathway for ER-phagy of procollagen (37) and ATZ in the lysosome (38). Thus, multiple ER-phagy pathways may deliver protein aggregates within the ER to the lysosome. Interestingly, growing ER-phagy vesicles at ER exit sites (ERESs) (containing procollagen, Ub, and COPII elements) may be directly bridged to the lysosome (37).

Accumulating evidence suggests that some forms of selective autophagy can be mediated by noncanonical autophagy pathways. A crucial indication of the existence of such pathways came somewhat unexpectedly from the reexamination of drugs claimed to block canonical autophagosome formation (including ionophores, such as CCCP, or basic lipophiles, such as chloroquine). However, these chemicals do not block flux in the canonical pathway, as previously thought, but rather induce lysosomal LC3/GABARAP lipidation at single membranes and do so independently of ATG9, ATG13, or PI3P generation (39). Direct lipidation of LC3s/GABARAPs on single membranes in the form of LC3-associated phagocytosis (LAP) functions also in a wide range of physiological processes, such as the vision cycle, neurodegenerative disease, tumor cell immunotolerance, and bacterial infection (40). Further analysis of the functions of organellophagy in normal physiology might thus allow us to modify key hypotheses in the field, especially concerning the existence of physiologically relevant noncanonical pathways that act in parallel to or instead of the canonical autophagy pathway.

Implications for pathophysiology

Although often found in the context of the same pathophysiologic conditions, the UPS and autophagy are mostly examined as discrete entities. Indeed, the simultaneous investigation of the two pathways in any given pathology entails a huge experimental effort. Nonetheless, there is accumulating evidence that autophagy and the UPS are simultaneously affected and may mutually affect each other in a variety of diseases (41). This has been particularly evident in disorders of the central nervous system that involve inflammatory or immune alteration, such as Parkinson’s, Alzheimer’s, and Huntington’s diseases (42). Also, in several types of cancer, molecular links between the UPS and autophagy have been discovered (2), but general concepts have not yet been demonstrated in clinical settings. A main reason for this is that the cell-autonomous and -nonautonomous roles of autophagy are not yet fully understood, not only with respect to metabolic adaptations within the tumor tissue itself but also in regulating host tissue adaptation (43). Moreover, immune responses to tumors seem also to rely on noncanonical forms of autophagy (40). Thus, understanding the cross-talk of the two systems is highly relevant for therapeutic approaches involving the inhibition or activation of either one. For example, the proteasome inhibitor bortezomib has been used successfully in the treatment of cancers such as multiple myeloma (41). However, as discussed above, prolonged inhibition of the proteasome will eventually lead to activation of autophagy, which can in turn be exploited as a prosurvival mechanism by cancer cells. Several reports have also indicated the highly contextual role of autophagy, suggestive of a highly dynamic and complex influence of autophagy in different types of cells as well as at distinct stages of cancer development (42).


The UPS and autophagy act as an integrated quality control network embedded in the general cellular stress response. However, the UPS and autophagy are still primarily investigated from a unicellular perspective that neglects the complexities arising in multicellular contexts. Thus, in the physiological context, many aspects regarding the mechanisms of selectivity and cross-talk remain elusive. Specifically, unraveling the kinetics, feedback controls, and membrane dynamics will require the development of methods to locally determine protein quantities and stoichiometries for the UPS and autophagy machineries in vivo. Also, it will be crucial to better understand the interplay among LLPS, ubiquitin, and autophagy, eventually allowing researchers and physicians to target the UPS and autophagy in pathophysiological processes that are associated with defective LLPS, like those observed in neurodegeneration. For the latter to be successful, a combination of biophysical experiments and modeling should help us to identify the constraints of autophagic membrane formation around phase-separated subcellular structures and how this is modulated by Ub-dependent mechanisms. Finally, the recent discoveries on organellophagy and the growing list of physiological processes depending on noncanonical autophagy pathways necessitate a reexamination of quality control in vivo. This will require the adaptation of quantitative, molecular tools from cellular to animal and organoid models, as well as engineering tools for the administration of chronic and low-dose stress in vivo to better mimic human pathophysiology in cancer, chronic infection, diabetes, and neurodegeneration. Ultimately, these approaches will provide much deeper insights into physiological and medical aspects of integrated cellular quality control pathways.

References and Notes

Acknowledgments: We thank D. Hoeller, K. Koch, A. Gubas, L. Alves, and P. Grumati for critical comments on the manuscript and D. Hoeller for help with preparation of figures. Author contributions: C.P. and I.D. designed the structure for the review, wrote the manuscript, and prepared the figures. Funding: Supported by the DFG-funded Collaborative Research Centre on Selective Autophagy (SFB 1177) (I.D. and C.P.) and by the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation program (grant agreement 742720) (I.D.). Competing interests: Authors declare no competing interests.

Stay Connected to Science

Navigate This Article