p62/SQSTM1 in autophagic clearance of a non-ubiquitylated substrate

Proteolytic systems and the aggresome pathway contribute to preventing accumulation of cytotoxic aggregation-prone proteins. Although polyubiquitylation is usually required for degradation or aggresome formation, several substrates are processed independently of ubiquitin through a poorly understood mechanism. Here, we found that p62/SQSTM1, a multifunctional adaptor protein, was involved in the selective autophagic clearance of a non-ubiquitylated substrate, namely an aggregation-prone isoform of STAT5A (STAT5A_ΔE18). By using a cell line that stably expressed STAT5A_ΔE18, we investigated the properties of its aggregation and degradation. We found that STAT5A_ΔE18 formed non-ubiquitylated aggresomes and/or aggregates by impairment of proteasome functioning or autophagy. Transport of these aggregates to the perinuclear region was inhibited by trichostatin A or tubacin, inhibitors of histone deacetylase (HDAC), indicating that the non-ubiquitylated aggregates of STAT5A_ΔE18 were sequestered into aggresomes in an HDAC6-dependent manner. Moreover, p62 was bound to STAT5A_ΔE18 through its PB1 domain, and the oligomerization of p62 was required for this interaction. In p62-knockdown experiments, we found that p62 was required for autophagic clearance of STAT5A_ΔE18 but not for its aggregate formation, suggesting that the binding of p62 to non-ubiquitylated substrates might trigger their autophagic clearance.


Introduction
Misfolded or aggregated proteins are usually refolded with molecular chaperones or degraded by the proteolytic machinery consisting of two main systems, designated the ubiquitinproteasome system (UPS) and the autophagy-lysosome pathway (Goldberg, 2003). The formation of aggregates and inclusion bodies in the brain is a pathognomonic feature of many neurodegenerative diseases, including Alzheimer's disease, Parkinson's disease and Huntington's disease, suggesting that dysfunctions of the UPS and the autophagy-lysosome pathway are involved in these diseases (Ross and Poirier, 2005;Rubinsztein, 2006). Indeed, impairment of these proteolytic systems causes neurodegeneration. For example, conditional knockout mice for a 26S proteasome subunit gene (Psmc1) or autophagy-related genes (Atg5 or Atg7) form ubiquitin-positive inclusion bodies and exhibit neurodegeneration (Bedford et al., 2008;Hara et al., 2006;Komatsu et al., 2007a). Furthermore, it was recently reported that parkin, which is the causal gene of familial Parkinson's disease and an E3 ubiquitin ligase, is selectively recruited to damaged or uncoupled mitochondria for clearance by the autophagy-lysosome pathway (Narendra et al., 2008;Vives-Bauza et al., 2010). Although these proteolytic systems are very important for quality control of proteins or organelles, some proteins that escape degradation often accumulate in the cytoplasm. To protect against the cytotoxicity of indigestible aggregates, these proteins are immediately sequestered within inclusion bodies via an alternative system called the aggresome pathway (Johnston et al., 1998;Kopito, 2000). This pathway involves histone deacetylase (HDAC) 6-mediated transport, which recruits the aggregates to dynein motors for transport to aggresomes located in close proximity to the microtubule-organizing center (MTOC) (Kawaguchi et al., 2003).
Recently, p62/SQSTM1, which is linked to classical adult-onset Paget disease of the bone, was reported to function in aggregate formation and proteolysis (Kirkin et al., 2009a). p62 can bind to ubiquitylated substrates and microtubule-associated protein 1 light chain 3 (LC3) on autophagosomes, and is itself degraded by autophagy (Bjørkøy et al., 2005;Pankiv et al., 2007;Komatsu et al., 2007b;Ichimura et al., 2008). Moreover, it was demonstrated that the accumulation of ubiquitin-positive inclusion bodies in Atg7-deficient hepatocytes and neurons is suppressed by an additional deficiency of p62 (Komatsu et al., 2007b). p62 has been proposed to mediate the targeting of ubiquitylated proteins for degradation by selective autophagy and aggregate formation (Ichimura et al., 2008). Although many misfolded proteins undergo ubiquitylation, which is also an important signal for target recognition by HDAC6 and p62 via ubiquitin-binding domains, several inclusion bodies, such as GFP-250 and mutant superoxide dismutase (SOD), do not contain appreciable polyubiquitin (Kirkin et al., 2009a;García-Mata et al., 1999;Johnston et al., 2000). It remains unclear how non-ubiquitylated proteins are recognized, degraded and transported to aggresomes. In this study we took advantage of an aggregation-prone protein, comprising a novel STAT5A isoform (STAT5A_⌬E18), to investigate aggresome formation and autophagic clearance of non-ubiquitylated aggregates. Our results revealed that non-ubiquitylated STAT5A_⌬E18 underwent p62-dependent autophagic clearance, suggesting that p62 is also involved in the selective autophagic clearance of non-ubiquitylated substrates.

Aggresome formation of STAT5A_⌬E18
It was previously reported that STAT5A_E18 shows a strong tendency to aggregate . To understand the properties of the resulting aggregates, we verified whether this abnormal protein was degraded by proteolytic systems, such as the UPS or the autophagy-lysosome pathway, using a HeLa cell line stably expressing STAT5A_E18-EGFP at a very low level (Fig.  1A). When this stable cell line was treated with the proteasome inhibitor MG132 (10 M) for 12 hours, STAT5A_E18 gradually accumulated and was localized to perinuclear regions as large inclusion bodies (Fig. 1B). Similar bodies were observed after inhibition of autophagy by the phosphatidylinositol 3-kinase inhibitor wortmannin (Fig. 1C). The half-life of STAT5A_E18 under these conditions was measured by cycloheximide (CHX)chase analysis (supplementary material Fig. S1A). Degradation of STAT5A_E18 was markedly delayed in the presence of MG132 and wortmannin. Furthermore, specific inhibition of autophagy by knockdown of ATG5 or ATG7 also led to the accumulation of STAT5A_E18 (Fig. 2). When HeLa cells stably expressing STAT5A_⌬E18-EGFP were transfected with short interfering RNAs (siRNAs) against ATG5 or ATG7, starvation-induced autophagy was inhibited in both cell types ( Fig. 2A,B). After 24 hours of ATG5 or ATG7 knockdown, STAT5A_⌬E18-EGFP accumulated and formed aggregates (Fig. 2C). It is known that indigestible aggregates are immediately transported to the MTOC by the HDAC6-dynein transport system and sequestered within inclusion bodies called aggresomes (Kawaguchi et al., 2003). Therefore, we examined whether the perinuclear inclusion bodies of STAT5A_E18 were formed via the HDAC6-dependent aggresome pathway. For inhibition of the aggresome pathway, the stable cell line was treated with trichostatin A (TSA), a broadspectrum inhibitor of HDACs (Furumai et al., 2001). The STAT5A_E18 protein rapidly began to accumulate in the cytoplasm (Fig. 1D) because TSA causes not only HDAC6 inhibition but also elevation of STAT5A_E18-EGFP gene expression through epigenetic changes of the DNA (Yoshida et al., 1990). However, despite the abundant protein accumulation, aggregates of STAT5A_E18 were not transported to the perinuclear region and remained as dot-like structures in the presence of TSA (Fig. 1D). It was confirmed that the MG132induced aggresome formation was also blocked by tubacin, a specific inhibitor of HDAC6 ( Fig. 1E) but not blocked by niltubacin, a nonactive derivative of tubacin (Fig. 1F). In addition, the cells were categorized into four groups on the basis of the sizes and distributions of the STAT5A_⌬E18-EGFP signals (Fig. 1G). hours. Some cells were treated with 10M tubacin (E) or 10M niltubacin (F) for 8 hours before treatment with 5M MG132. The cells were then fixed, stained with DAPI to visualize the nuclei (blue) and observed by confocal laser microscopy. Scale bar: 10m. (G)Images of cells (n>100) from each treatment were examined and the STAT5A_⌬E18-EGFP signals were categorized into four types on the basis of size and distribution pattern: aggresomes, aggregates, diffuse distribution and no accumulation. After 12 hours of treatment with MG132 or wortmannin, 55-65% of the STAT5A_⌬E18-expressing cells had aggresomes of ≥3 m in diameter. By contrast, aggresome formation was observed in only 2-8% of TSA-or tubacin-treated cells, although numerous EGFP-positive aggregates of 0.1-0.5 m in diameter accumulated in 54% or 40% of the cells, respectively (Fig. 1G). This analysis was in agreement with biochemical characterization of STAT5A_⌬E18: that its insoluble form was markedly increased by proteolytic inhibitions (supplementary material Fig. S1B). These results indicate that STAT5A_E18 is accumulated by proteasomal and autophagic impairment and forms aggresomes via HDAC6dependent retrograde transport.
Pre-aggresomal forms rather than aggresomes of STAT5A_⌬E18 are predominantly targeted for the autophagy-lysosome pathway As described above, we found that inhibition of autophagy led to increased numbers of aggregates and aggresomes of STAT5A_E18. Many aggresomes of proteins, such as mutant huntingtin (Q150Htt), filamentous actin (F-actin) and the ⌬F508 mutant of the cystic fibrosis transmembrane conductance regulator (⌬F508-CFTR), are known to be degraded by the autophagy-2694 Journal of Cell Science 124 (16) lysosome pathway (Iwata et al., 2005;Lázaro-Diéguez et al., 2008). Therefore, we examined whether the aggresomes and aggregates of STAT5A_⌬E18 were degraded by the autophagy-lysosome pathway. First, accumulation of soluble or insoluble STAT5A_⌬E18 during TSA treatment (0-12 hours) was investigated by immunoblotting (Fig. 3A). Soluble STAT5A_⌬E18 was present after 6 hours of TSA treatment (Fig. 3A, lane 3), and both soluble and insoluble forms had increased after 12 hours (Fig. 3A, lane 4). Under these conditions, the autophagic activity was evaluated by immunoblotting analysis using an anti-LC3 antibody (Fig. 3B). The relative intensity of the LC3-II band, which was C-terminally conjugated to phosphatidylethanolamine, increased after 6 and 12 hours of treatment with TSA in STAT5A_⌬E18-expressing cells (Fig. 3B, left panel). The presence of autophagic flux was also confirmed using bafilomycin A1 (Baf A1), a specific inhibitor of V-ATPase. Therefore, these observations indicate that STAT5A_⌬E18 accumulation activated autophagic clearance. These findings are consistent with immunocytochemical analyses using the anti-LC3 antibody. As shown in Fig. 3E,J, the number of LC3positive puncta was substantially increased in the cytoplasm after MG132 or TSA treatment compared with the control cells. However, the LC3-positive puncta showed little overlap with the Cells were lysed with 0.5% Triton X-100, followed by separation of the soluble (upper panel) and insoluble (lower panel) fractions. These fractions were subjected to immunoblotting with anti-GFP antibody. TSA-treated stable cells were observed with a fluorescence microscope (right panels). Scale bar: 50 m. (B) The extent of autophagy was analyzed by assessing the levels of LC3-II by immunoblotting with LC3 antibody. STAT5A_⌬E18 stable cells (left panels) or HeLa (right panels) were treated with TSA for 0-12 hours, and 100 nM bafilomycin A1 (Baf A1) was added for the last 2 hours. The LC3-II levels relative to that of actin (LC3-II/Actin) were quantified using ImageJ software. These observations were confirmed using DsRed-LC3 stable transformants (supplementary material Fig. S2). To monitor autophagy, HEK293 cells stably expressing DsRed-LC3 were established using the Flp-In T-REx system. Although DsRed-LC3 puncta appeared in the cytoplasm after transient expression of STAT5A_E18, neither aggregates nor aggresomes were incorporated into the DsRed-LC3 puncta (supplementary material Fig. S2C,F). Considering that the fluorescence of EGFP is lost in acidic compartments, we could not rule out the possibility of failure to detect aggregates incorporated into autophagolysosomes. Therefore, aggregates of STAT5A_E18 were immunostained with an anti-GFP antibody (Fig. 3D,I,N). The immunocytochemical staining also indicated that LC3-positive autophagosomes did not sequester aggresomes and aggregates, and were predominantly distributed with diffuse soluble STAT5A_E18 staining (Fig.  3G,L). Accordingly, our observations that neither aggregates nor aggresomes were incorporated into LC3 puncta were not artifacts caused by quenching of EGFP signals in acidic compartments. Furthermore, we examined the distribution of STAT5A_E18 and lysosomes ( Fig. 4), and confirmed that most of the aggregates and aggresomes were not delivered to lysosomes for degradation (Fig. 4D,E), although a portion of the aggregates were colocalized with lysosomes as visualized by LysoTracker Red (Fig. 4I,J). In addition to the finding that accumulation of soluble STAT5A_E18 led to the activation of autophagy (Fig. 3), soluble or pre-aggresomal forms but not aggresomes of STAT5A_⌬E18 are probably targeted to the autophagy-lysosome pathway.
Next, we examined whether aggregates or aggresomes of STAT5A_⌬E18 were bound to p62, an ubiquitin-binding protein (Zatloukal et al., 2002;Nagaoka et al., 2004). Immunocytochemical analyses showed that intense p62 staining was colocalized with both aggresomes and aggregates generated after treatment with MG132 ( Fig. 6D-F) or TSA (Fig. 6G-I). Moreover, the production of aggregates after TSA treatment increased the amount of insoluble p62 (Fig. 6J, lane 3). Similarly, p62 binding to STAT5A_⌬E18 was also confirmed by coimmunoprecipitation analysis (Fig. 6K). These results indicate that ubiquitin-negative aggregates or aggresomes were also targeted by p62. In general, p62 binds to substrates via a C-terminal ubiquitin-association (UBA) domain (Seibenhener et al., 2004), raising the question of how non-ubiquitylated substrates are recognized by p62. It is known that p62 is a scaffolding protein with several interaction modules that include a Phox and Bem 1 domain (PB1), a zinc finger domain (ZZ), a SOD1 mutant interaction region (SMIR), an LC3-interacting region (LIR/LRS: L3-interacting region/L3-recognition sequence) and a ubiquitinassociated domain (UBA; Fig. 7A). For identification of the region of p62 binding to the STAT5A_⌬E18 aggregate, we constructed deletion mutants of p62 fused to a myc-His tag at the C-terminal end (Fig. 7B). These constructs and STAT5A_⌬E18 were cotransfected into p62KO MEF cells. The colocalization of STAT5A_⌬E18 and mutant p62 was immunocytochemically examined with anti-myc antibody (Fig. 7C). A deletion mutant of the PB1 domain (p62⌬PB1) was not colocalized with STAT5A_⌬E18 aggregates, whereas other deletion mutants (p62⌬ZZ, p62⌬SMIR, p62⌬LRS and p62⌬UBA) and a truncated form (PB1 1-125 ) were (Fig. 7C). These results indicate that the nonubiquitylated substrate STAT5A_⌬E18 is bound to p62 via the PB1 domain but not the UBA domain. Moreover, the p62 K7AD69A mutant, lacking self-oligomerization activity, could not colocalized with STAT5A_⌬E18 aggregates (Fig. 7C; p62 K7AD69A ), suggesting that oligomer formation of p62 through the PB1 domain was required for p62-binding to the STAT5A_⌬E18 aggregate.

Aggregate formation and autophagic clearance of STAT5A_⌬E18 pre-aggresomal forms in p62-knockdown cells
Several studies have suggested that p62 is involved in aggregate formation and shuttling of ubiquitylated proteins to proteolytic systems (Komatsu et al., 2007b;Ichimura et al., 2008). To elucidate this we observed STAT5A_⌬E18 aggresome formation in p62knockdown cells. To achieve high-efficiency silencing of p62, transfection of the siRNAs was repeated every 48 hours for 5 days before treatment with 10 M MG132 or 1 M TSA for 18 hours. p62 knockdown was confirmed by RT-PCR and immunoblotting (Fig. 8A,B), and the generation of aggregates or aggresomes was observed by confocal laser scanning microscopy ( Fig. 8C-H).
Although immunocytochemistry with an anti-p62 antibody also showed that the p62 level was drastically reduced in p62 siRNAtransfected cells, p62 knockdown did not affect the formation of STAT5A_⌬E18 aggresomes and aggregates (Fig. 8C,D), indicating another role for p62 binding to STAT5A_⌬E18. This result was reconfirmed in p62KO MEF cells (Fig. 8I-N). It has been proposed that p62 directly binds to LC3 via an LIR/LRS domain and acts as a cargo receptor or adaptor for autophagic degradation of ubiquitylated substrates (Pankiv et al., 2007;Ichimura et al., 2008). Therefore, we investigated whether p62 was involved in the selective autophagic clearance of STAT5A_⌬E18. First, autophagic induction by TSA treatment was observed in cells transfected with a negative control siRNA (Fig. 9D,E). Quantification of the average cellular LC3 puncta as a measure of autophagy was performed (Fig. 9G). The number of LC3 puncta was considerably increased in p62-expressing cells owing to the accumulation of STAT5A_⌬E18 aggregates (15-fold) after TSA treatment. By contrast, in p62-knockdown cells the number of LC3 puncta was substantially decreased (Fig. 9A,B,G). These observations were clearly consistent with the conversion of LC3-I to LC3-II in control versus p62-knockdown cells (Fig. 9H), suggesting that aggregateinduced autophagy was reduced in p62-knockdown cells.

Aggresome formation of STAT5A_⌬E18 occurs in an HDAC6-dependent manner
Unfolded or misfolded proteins are refolded, degraded or sequestered into aggresomes for protection of the cell because the accumulation of such proteins is highly toxic and often causes cell death (Goldberg, 2003). Although it is commonly known that ubiquitin signals play important roles in protein homeostatic processes, there is little understanding of how non-ubiquitylated substrates are processed (Rubinsztein, 2006). In this study, we discovered the existence of HDAC6-dependent aggresome formation and p62-mediated autophagic clearance of nonubiquitylated substrates using STAT5A_⌬E18, an aggregationprone isoform. Many protein aggregates are known to form large protein inclusion bodies called aggresomes through HDAC6dependent retrograde transport (Kawaguchi et al., 2003). HDAC6 contains two catalytic domains and a ubiquitin-binding domain (BUZ). Ubiquitylated misfolded proteins, such as Q150Htt and ⌬F508-CFTR, are recognized via the BUZ domain and transported to aggresomes by dynein motors (Kawaguchi et al., 2003;Iwata et al., 2005). From these previous reports, STAT5A_⌬E18 was expected to undergo ubiquitylation; however, its aggresomes were unexpectedly ubiquitin-negative despite the HDAC6-dependent aggresome formation. These findings are also in contrast to those of another non-ubiquitylated protein, GFP-250, which is considered to form aggresomes in an HDAC6-independent manner, on the basis of experiments using HDAC6-knockdown cells (Kawaguchi et al., 2003). Although the mechanism for the ubiquitin-independent HDAC6 binding remains unclear, the inhibition of aggresome formation by treatment with TSA or tubacin indicates that HDAC6 deacetylase activity is required for the formation of STAT5A_⌬E18 aggresomes as well as that of ubiquitylated substrates. It was 2698 Journal of Cell Science 124 (16) recently shown that MyD88, a universal adaptor molecule for Tolllike receptors and interleukin-1 receptor, could interact with HDAC6 mutants in the BUZ domain or deacetylase domain, suggesting the existence of ubiquitylation-independent binding of HDAC6 with STAT5A_⌬E18 . However, the precise binding mechanism is not yet clear. Further investigations are required to resolve how HDAC6 acts as an adaptor protein that selectively links non-ubiquitylated substrates to motor complexes.

Proteasomal and autophagic clearance of nonubiquitylated STAT5A_⌬E18
Inhibition experiments using MG132 or wortmannin revealed that the non-ubiquitylated STAT5A_⌬E18 protein was degraded by both the proteasomal and autophagy systems. Several proteasomal substrates, such as ornithine decarboxylase, thymidylate synthase and c-Jun, are known to be degraded without any requirement for ubiquitin modification (Murakami et al., 1992;Hoyt and Coffino, 2004). Targeting of non-ubiquitylated substrates to the proteasome is brought about by accessory molecules or by a sequence within the substrates themselves (Orlowski and Wilk, 2003). Recently, the first proline-rich domain of thymidylate synthase (amino acids 9-15) was identified as a degron of the ubiquitin-independent proteasomal degradation pathway (Peña et al., 2009). STAT5A_⌬E18 also possesses a proline-rich region at its Cterminus , suggesting that this region might be involved in proteasomal targeting. Moreover, we observed that STAT5A_⌬E18 was also targeted for autophagic clearance. However, the aggresomes of STAT5A_⌬E18 were never sequestered into LC3-positive autophagosomes and lysosomes, whereas aggresomes of other proteins such as Htt-Q103 and Tau-P301L are surrounded by these vacuoles (Iwata et al., 2005;Wong et al., 2008). In the case of STAT5A_⌬E18, these proteolytic  (lower) were also examined by immunoblotting. For autophagy flux assay, these cells were treated with 100 nM bafilomycin A1 for 2 hours. Relative LC3-II levels (LC3-II/Actin) were quantified using ImageJ software. compartments were predominantly distributed in soluble forms, suggesting that the autophagy-lysosome pathway probably degrades soluble or pre-aggresomal forms rather than aggresomes. In agreement with this hypothesis, we observed that the autophagic response was accelerated by accumulation of pre-aggresomal structures after TSA treatment.

p62 is involved in the autophagic clearance of non-ubiquitylated STAT5A_⌬E18
Similar to HDAC6, the multifunctional adaptor protein p62 is known to possess a UBA domain at its C-terminus and to bind to ubiquitylated substrates (Kirkin et al., 2009a;Lamark et al., 2003). However, our present observation that UBA-truncated p62 was able to bind to STAT5A_⌬E18 suggests the existence of ubiquitinindependent p62 binding for the recognition of certain substrates. Although there is a previous report that p62 can bind to mutant SOD1 in a ubiquitin-independent manner through SMIR (Gal et al., 2007;Gal et al., 2009), the PB1 domain and its oligomerization were essential for binding to non-ubiquitylated STAT5A_⌬E18 in our study. Our result suggests a multiple-substrate-recognition mechanism of p62. The p62-STAT5A_⌬E18 complex might be generated via the Src homology 2 (SH2) domain of STAT5A because p62 is known to interact with the SH2 domain of p56 lck (Park et al., 1995). This interaction is abolished by deletion of the N-terminal 21 amino acids (Pro29 to Arg50) within the PB1 domain of p62 (Joung et al., 1996). In the case of a specific interaction such as that between SH2 and PB1, however, it would be difficult for p62 to be adapted to multiple substrates because most substrates do not possess an SH2 domain. As an alternative, it is possible that indirect binding occurs through molecular chaperones. p62 interacts with BAG3, a HSC70-HSP70 co-chaperone, which mediates the recruitment of unfolded proteins to the macroautophagy pathway for protein quality control (Gamerdinger et al., 2009). The p62-BAG3-HSP70 complex might be involved in the recognition mechanism for non-ubiquitylated substrates because molecular chaperones can interact directly with a variety of unfolded proteins.
Although it remains unknown how pre-aggresomal structures undergo selective clearance, recent studies have shown that p62 is involved in inclusion body formation and selective autophagic clearance (Bjørkøy et al., 2005;Pankiv et al., 2007;Ichimura et al., 2008). p62 can directly interact with ubiquitin and LC3, an autophagy-related protein, suggesting that p62 functions as a shuttle protein to transport ubiquitylated proteins into the autophagosome or lysosome (Bjørkøy et al., 2005;Pankiv et al., 2007;Ichimura et al., 2008). In the present study, we found that p62 associated with STAT5A_⌬E18 in a ubiquitylation-independent manner, suggesting that p62 could also be involved in selective autophagic clearance of non-ubiquitylated substrates. This hypothesis is supported by our observation that p62-knockdown cells had significantly less aggregate-induced autophagy. Recently, p62 was also reported to be necessary for autophagic clearance of ubiquitylated substrates, peroxisomes or mitochondria (Kim et al., 2008;Geisler et al., 2010). A notable finding of our study is that p62 knockdown inhibited LC3 lipidation and autophagosome formation, suggesting that the binding of p62 to non-ubiquitylated substrates triggers their autophagic clearance. If this is the case, p62 knockdown would be expected to induce the accumulation of STAT5A_⌬E18 aggregates. However, this accumulation was not observed in p62knockdown cells. We reasoned that p62 knockdown would not be sufficient to completely inhibit autophagy. Indeed, Komatsu et al. reported that other p62-like receptors might compensate for the lack of p62 because of the mild accumulation of protein aggregates in p62-deficient mice (Komatsu et al., 2007b). Furthermore, it was recently reported that Nbr1, another adaptor protein, is recruited to ubiquitin-positive protein aggregates and degraded by autophagy, which was dependent on an LIR/LRS domain, similar to the case for p62 (Kirkin et al., 2009b). Thus, multiple cargo receptors such as p62 and Nbr1 might be involved in the selective autophagic clearance of not only ubiquitylated substrates but also certain types of non-ubiquitylated substrates. In future studies it will be important to elucidate how cargo receptors select substrates in a ubiquitylation-independent manner.
An EGFP-parkin plasmid was constructed by PCR using 5Ј-AAGCTTTGATAGT-GTTTGTCAGGTTCAAC-3Ј and 5Ј-GGATCCTACACGTCGAACCAGTGGTC-3Ј primers. The PCR product was inserted into the HindIII-BamHI site of pEGFP-C1. This plasmid was transfected with Lipofectamine LTX into HeLa cells for immunocytochemistry or immunoprecipitation.

Antibodies
The complete list of antibodies and sources used is given in supplementary material Table S1.

Drug treatment and immunocytochemistry
Cells (1ϫ10 5 ) were plated on round cover glasses (13 mm in diameter) and grown for 18 hours at 37°C in 5% CO 2 in an incubator. For analysis of autophagy, pDsRed-LC3 was transfected into cells using Lipofectamine LTX before the drug treatment. The cells were then treated with 10 M MG132, 10 M wortmannin or 1 M TSA for 12-24 hours. Because wortmannin has a short half-life in aqueous solution, the growth medium containing 10 M wortmannin was changed every 3 hours. For autophagy flux assays, cells were treated with 100 nM bafilomycin A1 for 2 hours. Immunocytochemistry was carried out as described previously . Cells were treated with 10 M tubacin or 5 M MG132 for 12 hours. All cells were fixed with 4% paraformaldehyde after the various drug treatments. The fixed cells were permeabilized in phosphate-buffered saline (PBS) containing 0.1% Triton X-100 and blocked with blocking solution (Nacalai Tesque, Kyoto, Japan). The cells were then incubated with anti-ubiquitin, anti-LC3, anti-His tag or anti-p62 antibodies (1:500 dilution) for 12 hours at 25°C. After three washes with PBS, the cells were incubated with tetramethylrhodamine-labeled goat anti-mouse IgG (Invitrogen), Cy3-labeled goat anti-rabbit IgG (Millipore) or Alexa-Fluor-350-labeled goat antimouse IgG (Invitrogen) for 4 hours at 25°C. After washing and staining with 4Ј,6diamidino-2-phenylindole (DAPI) (Dojindo, Kumamoto, Japan), the cells on the cover glasses were mounted on glass slides in an aqueous mounting medium (Gel Mount; Biomeda, Foster City, CA) and examined under a fluorescence microscope (Olympus, Tokyo, Japan) or a confocal laser microscope (LSM 510 META; Carl Zeiss, Oberkochen, Germany). EGFP was excited using a 488 nm argon laser, and emission was recorded through a BP 500-530 nm filter. Rhodamine was excited with a 543 nm helium-neon laser, and emission was recorded through an LP 560-nm filter. DAPI or Alexa Fluor 350 was excited at 800 or 780 nm, respectively, with a Mai Tai two-photon laser, and fluorescence was recorded through a 390-465 nm BP filter. Images were obtained with a Plan-Apochromat 63ϫ/1.4 or 100ϫ/1.4 oil DIC lens.

Half-life measurement of STAT5A_⌬E18-EGFP
Cells were pre-treated with 10 M MG132 or 10 M wortmannin for 8 hours before treatment with 50 g/ml CHX. After CHX treatment, cells were washed with PBS and lysed with Laemmli sample buffer. Protein samples were subjected to immunoblotting analysis with anti-GFP or anti-actin antibody.
RNA preparation, cDNA synthesis and real-time PCR analysis were performed as described previously (Yamaguchi et al., 2008).

Immunoblotting
Cells were washed with PBS and extracted with ice-cold PBS containing 0.5% Triton X-100 and a protease inhibitor cocktail (Nacalai Tesque). For immunoblotting analysis of LC3, detergent extracts were separated in 15% SDS-polyacrylamide gels using a Tris-glycine buffer system. For analysis of ubiquitylation, Triton X-100 extracts were fractionated by centrifugation (20,000 g, for 15 minutes at 4°C) after sonication. The detergent-insoluble fraction was washed with PBS and resuspended in Laemmli SDS sample buffer. Protein samples were separated in 15% or 12.5% SDS-polyacrylamide gels using the Tris-glycine buffer system. After electrophoresis, the separated proteins were transferred to Immobilon-P membranes (Millipore). After blocking with 10% skimmed milk, the membranes were incubated with anti-LC3, anti-EGFP, anti-ubiquitin or anti-p62 antibodies (1:1000 dilution) for 12 hours at 25°C. The membranes were then washed with PBS and incubated with alkalinephosphatase-conjugated anti-rabbit IgG (1:5000 dilution; Millipore) or anti-mouse IgG (1:2000 dilution; Jackson ImmunoResearch Laboratories Inc., West Grove, PA) for 1 hour. Immunopositive signals were detected with nitroblue tetrazolium chloride and 5-bromo-4-chloro-3-indolylphosphate P-toluidine salt reagents.

Immunoprecipitation
Transfected cells were washed with ice-cold PBS and lysed with PBS containing 0.5% Triton X-100 and a protease inhibitor cocktail, followed by sonication. The cell extracts were incubated with 1 g of rabbit polyclonal anti-GFP antibody or goat polyclonal anti-myc antibody at 4°C. After the incubation, 25 l of protein G Mag Sepharose (GE Healthcare Bio-Sciences, Uppsala, Sweden) was added to the extracts and mixed for 3 hours at 4°C. The Sepharose beads were then washed three times with PBS containing 0.5% Triton X-100 and resuspended in Laemmli SDS sample buffer.
p62 knockdown p62 (p62; siTrio Full Set) and negative control (siTrio Negative Control) siRNAs were purchased from B-Bridge International Inc. (Mountain View, CA). The siRNAs were transfected into cells using the siPORT NeoFX Transfection Agent (Applied Biosystems, Carlsbad, CA) according to the manufacturer's instructions. The transfections were repeated every 48 hours for 5 days before treatment with MG132 or TSA. For confirmation of p62 knockdown, RT-PCR and immunoblotting analyses were performed. RNA preparation and cDNA synthesis were carried out as described previously (Watanabe et al., 2004). PCR amplifications were carried out using the p62_Fw and p62_Rv1 primers (5Ј-GGTACCCGCCACCATGGCGTCGCTCAC -CGTGAA-3Ј and 5Ј-TCTAGACAACGGCGGGGGATGCTTTG-3Ј, respectively) and the following conditions: 28 cycles of 95°C for 15 seconds, 60°C for 15 seconds and 68°C for 90 seconds.

Statistical analysis
The numbers of LC3 puncta per cell were quantified (n25). The statistical significance of differences among values was determined by one-way ANOVA with a post-hoc Tukey's test using KaleidaGraph 4.0 software (Synergy Software, Reading, PA).