NBR1 acts as an autophagy receptor for peroxisomes

Peroxisomes are critical metabolic organelles that are required for oxidation of fatty acids, and reduction of hydrogen peroxide produced during lipid oxidation. They are required for synthesis of essential cellular components such as plasmalogens, isoprenoids and lysine. Unlike other metabolic organelles, peroxisome numbers are highly regulated, varying in numbers depending on cellular needs (Till et al., 2012). This plasticity is most readily observed in the liver of rodents. Upon activation of the nuclear receptor peroxisome proliferator-activated receptor-alpha (PPARα), peroxisome proliferation occurs by growth and division of pre-existing peroxisomes (Platta and Erdmann, 2007). However, upon removing the PPARα ligand, peroxisome numbers are quickly reduced to basal levels. This reduction is mediated by macro-autophagy, a catabolic mechanism of delivering large cytosolic material to lysosomes for degradation (Iwata et al., 2006). Little is known about the precise mechanism of macro-autophagy of peroxisomes, or macro-pexophagy (hereafter referred to as pexophagy). However, it is known that the inability to maintain peroxisome numbers is linked to various neurodegenerative and developmental disorders such as X-linked adrenoleukodystrophy, and Krabbe disease (Ribeiro et al., 2012; Singh et al., 2009). In particular, in some leukodystrophies, the loss of peroxisomes during neuroinflammation is thought to exacerbate the cellular inflammation, eventually leading to cell death (Kassmann and Nave, 2008).

Selective macro-autophagy (hereafter selective autophagy) is a catabolic cellular process by which large cytoplasmic materials are selectively sequestered and degraded in lysosomes. In the mammalian system, peroxisomes along with other cytosolic components such as protein aggregates, damaged mitochondria, ER, ribosomes, membrane remnants, midbody rings, intracellular bacteria, bacteriocidal precursor and viral capsid proteins are substrates of selective autophagy (reviewed by Johansen and Lamark, 2011; Kirkin et al., 2009b; Komatsu and Ichimura, 2010; Kraft et al., 2010). The mechanism of selective autophagy is not well understood. However, recent studies of various substrates are coalescing into a common theme in the mechanism of targeting selective substrates to autophagosomes. In this mechanism the substrate is activated for degradation by accumulation of ubiquitin (Ub) on the cytosolic side of the substrate followed by recruitment of ubiquitin-binding autophagy receptor(s) (Johansen and Lamark, 2011). The autophagy receptors play an essential role in linking the substrate to nascent autophagosomes by binding to both the ubiquitin on the substrate and the autophagy factor LC3 on the autophagosomes.

At present four Ub-binding mammalian autophagy receptors have been identified: p62, NBR1, NDP52 and optineurin (Bjørkøy et al., 2005; Kirkin et al., 2009a; Pankiv et al., 2007; Thurston et al., 2009; Wild et al., 2011). Recent studies suggest that depending on the substrate, these Ub-binding receptor proteins can either act cooperatively or independently. For example, both p62 and NBR1 have been reported to act cooperatively to target polyubiquitylated aggregates to autophagosomes (Kirkin et al., 2009a). However, Salmonella targeting to autophagosomes does not require NBR1, but instead requires both p62 and NDP52 (Cemma et al., 2011; Zheng et al., 2009). These two autophagy receptors do not act cooperatively but appear to be recruited to Salmonella independently of each other. Similar results were observed for Listeria. However, for Shigella the recruitment of all three proteins was shown to be interdependent on each other (Mostowy et al., 2011).

The role of the various Ub-binding receptors in selective autophagy of organelles is not well understood. Several groups have reported that p62 is involved in mitophagy (the autophagic degradation of mitochondria); however, it is not clear whether p62 is required for targeting of ubiquitylated mitochondria to autophagosomes or serves a different functional role (Ding et al., 2010; Geisler et al., 2010; Narendra et al., 2010; Okatsu et al., 2010). Recently, we showed that p62 is involved in selective degradation of peroxisomes when peroxisomes were labeled with ubiquitin (Kim et al., 2008). Depleting cells of p62 using siRNA resulted in a decrease in peroxisome turnover, thus causing an increase in peroxisome numbers. However, the role of NBR1 in pexophagy is not known.

Here, we tested the hypothesis that autophagy receptors confer substrate selectivity in autophagy. For these studies we used peroxisomes as substrates for selective autophagy and examined the role of two autophagy receptors, NBR1 and p62. Peroxisomes were selected as the substrate to examine the role of these two receptors for two reasons. First, peroxisome numbers can be readily quantified, allowing for monitoring changes in their numbers both statically and dynamically. Second, except for our previous report implicating p62 in the clearance of peroxisomes (Kim et al., 2008), little is known about the role of other autophagy receptors, like NBR1, in mammalian pexophagy. Here, we characterize the roles of both NBR1 and p62 in targeting of peroxisomes to autophagosomes for their degradation. Based on our findings we present a mechanism for substrate specificity in selective autophagy. NBR1 uses a mechanism similar to ‘coincidence detection’, where the combination of the ubiquitin-binding UBA domain and the phospholipid-binding J domain functions as an effective ‘coincidence detector’ (Carlton and Cullen, 2005), allowing NBR1 to target peroxisomes for autophagic degradation.

Recently, NBR1 and p62 were shown to act together to selectively target polyubiquitylated protein aggregates to autophagosomes for degradation (Kirkin et al., 2009a). p62 was also shown to be involved in pexophagy (Kim et al., 2008). However, it is not known whether NBR1 is required for peroxisome degradation, and if so, whether it works cooperatively with p62. In order to address this question, we depleted HeLa cells of NBR1 using siRNA and examined whether its loss affected the endogenous level of peroxisomes. To quantify the change in peroxisome numbers, the total fluorescence intensity from a peroxisomal protein, catalase, was quantified by indirect immunofluorescence (Huang et al., 2011; Kim et al., 2008). To ensure that only cells with knockdown of p62 and/or NBR1 were quantified, they were co-immunostained for p62 and/or NBR1 expression.

As shown previously (Kim et al., 2008), knocking down p62 expression resulted in an increase in catalase levels relative to the cells treated with non-targeting siRNA, suggesting that p62 depletion inhibited the endogenous turnover of peroxisomes (Fig. 1A,B). Similarly, in cells where NBR1 was depleted, an increase in catalase levels was also observed, suggesting that NBR1 is also involved in the endogenous turnover of peroxisomes (Fig. 1A,B). The increase in peroxisomal catalase levels in p62 and NBR1 knockdown cells was confirmed by western blot analysis, which showed elevated catalase levels in cells where either NBR1 or p62 was depleted (Fig. 1C). We also examined the autophagy receptor NDP52, which was previously shown to be involved in autophagy of the bacterial pathogen Salmonella typhimurium (Thurston et al., 2009). When NDP52 expression was knocked down using siRNA, the endogenous level of peroxisomes was not affected (supplementary material Fig. S1). Hence, NDP52 is not required for basal peroxisome turnover. Together, these experiments suggest that NBR1 and p62, but not NDP52, are required for basal turnover of peroxisomes.

Interestingly, the increase in catalase levels in NBR1 knockdown cells was significantly higher than those observed in p62 knockdown cells (Fig. 1B,C). A possible explanation for this observation is that p62 and NBR1 mediate independent but parallel peroxisome degradation pathways, like that found with NDP52 and p62 in S. typhimurium (Cemma et al., 2011). Conversely, both may be involved in the same pathway but at different stages. To determine whether p62 and NBR1 act in two distinct parallel pathways, we knocked down the expression of both p62 and NBR1 simultaneously and quantified peroxisomal catalase levels (Fig. 1B). If NBR1 and p62 are involved in different but parallel pathways, depletion of both proteins should result in an additive increase in peroxisome levels. If NBR1 and p62 are involved in the same pathway but at different steps, then no additive increase should be observed when both proteins are depleted. As shown in Fig. 1B, when both NBR1 and p62 were knocked down, the increase in peroxisome levels was not additive but remained similar to NBR1-only knockdown conditions. This result was further verified by western blot of endogenous catalase levels that showed no further increase in catalase upon knockdown of both p62 and NBR1 compared to NBR1 alone (Fig. 1C). Thus, NBR1 and p62 act in the same pathway.

If p62 and NBR1 are involved in the same pathway, they may function cooperatively to target peroxisomes to autophagosomes, similarly to polyubiquitylated protein aggregates. If this is the case, then the differential inhibition of pexophagy observed between NBR1 and p62 knockdown conditions (Fig. 1B) suggests that NBR1 may play the initiating or prominent role in mediating pexophagy. This hypothesis is supported by the observation that knockdown of NBR1 resulted in a similar increase in catalase levels compared to cells where autophagy was inhibited by knockdown of the autophagic factor ATG12 (Fig. 1B,C; supplementary material Fig S2). A possible explanation for these results is that NBR1 may be recruited to peroxisomes to activate the targeting of peroxisomes to autophagosomes before p62. If this model is true, then overexpression of NBR1 should upregulate pexophagy by increasing the amount of free NBR1 that can bind to peroxisomes.

We tested this hypothesis by first asking whether exogenously expressed NBR1 colocalized with endogenous peroxisomes. Peroxisomes were identified using antibodies against endogenous catalase or PMP70 (Fig. 2A,F). When we expressed a Cherry–NBR1 chimera protein, we observed that peroxisomes clustered around NBR1-positive punctate structures (Fig. 2B,C). Peroxisome clustering was not observed in cells expressing Cherry–p62 (Fig. 2D,E). In order to illustrate the differences in peroxisome binding between the two autophagy receptors, we quantified the Manders' colocalization coefficient of Cherry signal overlapping with catalase/PMP70 (M_Cherry) for both autophagy receptors. For these quantifications, only cells with similar expression levels (as determined by the total Cherry fluorescence in each cell) were quantified. In comparison to Cherry–p62, Cherry–NBR1 showed a significantly higher colocalization coefficient with endogenous catalase and PMP70 (Fig. 2G). Similar NBR1/peroxisome clusters were observed in other cell lines, such as COS7, HEK293 and wild-type mouse embryonic fibroblasts (MEFs), showing that the observed phenomenon was not cell type specific (data not shown).

To determine the nature of these NBR1/peroxisome clusters, we examined HeLa cells expressing NBR1 by correlative light–electron microscopy (CLEM). As seen in the representative images (Fig. 3A–D), we found two populations of NBR1-positive structures. One group consisted of clusters of single membrane vesicular-like structures with a diameter of 40–70 nm (Fig. 3D, arrow). The second population clearly shows these vesicular-like structures sequestered within a double-membrane structure resembling typical autophagosomes (Fig. 3D, arrowhead). The identity of these NBR1 clusters was further verified by immuno-EM on cells transfected with GFP-NBR1 (Fig. 3E). To establish the relationship between these NBR1 vesicle structures and peroxisomes, we transfected HeLa cells with Cherry–NBR1 and the peroxisomal marker GFP–PMP34 (Fig. 3F–H). At high magnification we observed many peroxisome-like vesicles, often localized at the periphery of the aggregates (Fig. 3I,K). These results confirm the clustering of peroxisomes with NBR1 observed by confocal fluorescence microscopy.

To determine whether the peroxisome/NBR1 clusters lead to targeting of peroxisomes to lysosomes we looked for colocalization with the lysosomal marker Lamp1 (supplementary material Fig. S3). To prevent degradation of peroxisomes and NBR1 within lysosomes, and allow their visualization, cells were treated with the lysosomal protease inhibitors E-64 and leupeptin. The majority of the NBR1 punctate structures colocalized with Lamp1 (supplementary material Fig. S3A). This colocalization was abolished upon treating the cells with the autophagy inhibitor 3-methyladenine (3-MA) (supplementary material Fig. S3B), suggesting that the colocalization of NBR1 with lysosomes was mediated by autophagy. We also examined whether peroxisome/NBR1 clusters also colocalized with lysosomes. When Cherry–NBR1/Lamp1–GFP-expressing cells were probed for endogenous catalase, the majority of peroxisomes were found clustered with NBR1 within lysosomes (supplementary material Fig. S3C–E). To determine the extent of this clustering we quantified the number of the NBR1/peroxisome clusters that colocalized with the lysosomal marker Lamp1. In these cells 85±13% of these clusters were found inside lysosomes. However, there was a small population of cells (∼10%) that had large NBR1 aggregates with peroxisomes clustered around them but not localized inside lysosomes. The other 5% on average did not colocalize with peroxisomes. Time-lapse imaging of cells expressing Cherry–NBR1 and the peroxisomal marker PMP34-GFP verified that peroxisomes were targeted to lysosomes for degradation when NBR1 was overexpressed. As the expression of NBR1 increased, peroxisomes clustered around NBR1, and eventually disappeared (supplementary material Movie 1). Next we asked whether NBR1/peroxisome clusters localize with autophagosomes. In cells expressing GFP–NBR1 along with the autophagosome marker Cherry–LC3 we found that NBR1/peroxisome clusters colocalized with Cherry–LC3-positive structures (supplementary material Fig. S3F,G). Together, these results strongly suggest that exogenously expressing NBR1 causes an increased targeting of peroxisomes to lysosomes by autophagy.

In order to visualize and quantify pexophagy activity due to elevated NBR1 levels, we employed a pH-sensitive live-cell assay hereafter referred to as the ‘RG-lysosome assay’ (Pankiv et al., 2007). In this assay a tandem chimera of mCherry and EGFP fused to the peroxisome membrane targeting sequence of PEX26 (GFP–Cherry–PEX26_TM) is expressed in HeLa cells. Since the fluorescence of mCherry is more acid resistant than EGFP, peroxisomes targeted to lysosomes maintain their red fluorescent signal while the GFP fluorescence gets quenched. Thus, GFP–Cherry–PEX26_TM localized inside lysosomes exhibits red fluorescence only, while cytoplasmic protein appears yellow due to the combination of red and green fluorescence (supplementary material Fig. S4). Time lapse imaging of cells expressing Cer–NBR1 and GFP–Cherry–PEX26_TM over 2 days revealed an NBR1-mediated gradual movement of peroxisomes into lysosomes, followed by disappearance of peroxisomes (supplementary material Movie 2). The disappearance of peroxisomes was preceded by an increase in ‘Red’ peroxisomes (which appear slightly purple due to overlap with the blue signal from Cer–NBR1). Unlike GFP, Cerulean has a low pK_a (4.7), thus it maintains its fluorescence signal within the low pH environment of lysosomes (Rizzo et al., 2004).

In order to assess the ability of exogenously expressed NBR1 to induce pexophagy we quantified pexophagy using the RG-lysosome assay in cells coexpressing GFP–Cherry–PEX26_TM with Cer–NBR1, Cer–p62 or Cerulean (Cer). We considered a cell to show increased pexophagy if 20% or more of its total Cherry pixels were ‘red-only’ (supplementary material Fig. S4). A pixel was considered to be ‘red-only’ if the fluorescent signal from Cherry was at least three times higher than the GFP signal (see Materials and Methods). There was a significant increase in the number of cells with elevated pexophagy upon expressing Cer–NBR1 compared to Cer–p62 or Cerulean alone (Fig. 4A,B). Interestingly, no significant difference was observed between cells expressing Cer–p62 compared to Cerulean alone (Fig. 4B). To demonstrate that the increase in the red-only signal was due to pexophagy, we treated the cells with 3-MA and found that it prevented the accumulation of red-only signal (Fig. 4A,B).

To further verify that exogenously expressing NBR1 leads to pexophagy we repeated these experiments in ATG5 knockout MEFs, an autophagy deficient cell line. To quantify changes in peroxisome numbers, we immunostained the MEFs for endogenous catalase and determined the peroxisome density (number of peroxisomes per cell volume). There was no significant change in peroxisome numbers in Atg5^−/− MEFs when Cherry–NBR1 was exogenously expressed, whereas a significant loss of peroxisomes was observed in wild-type (Atg5^+/+) MEFs compared to non-transfected cells (Fig. 4C,D). These results were further verified by immunoblot analysis for changes in peroxisomal proteins. We found that both catalase and PMP70 levels decreased in the wild-type cells but not in the autophagy deficient cell line (Fig. 4E). Therefore, these results suggest that NBR1 can promote the activation of pexophagy. Interestingly, we found that there is no difference in the percentage of NBR1 clustered with peroxisomes between the two cell lines (Fig. 4F), yet there is a significant decrease in peroxisomes (Fig. 4E). These results may indicate that the clustering of NBR1 with peroxisomes precedes targeting to autophagosomes/lysosomes.

The activation of pexophagy by elevated NBR1 levels suggests that the amount of NBR1 in the cell may partly regulate pexophagy. The fact that NBR1 upregulated pexophagy at endogenous p62 levels suggests that p62 may not be strictly required for pexophagy during elevated NBR1 levels. Hence, we asked whether NBR1 could target peroxisomes to lysosomes independently of p62. To ascertain the requirement for p62 in NBR1-dependent pexophagy, we repeated the NBR1 RG-lysosome assay in cells where p62 expression was knocked down. In p62 siRNA-treated cells, Cer–NBR1-expressing cells still showed only a slight decrease in pexophagy compared to Cer–NBR1-expressing cells treated with non-targeting siRNA (Fig. 5A). However, these cells showed a significant increase in ‘red-only’ signal compared to control (Cer) or Cer–p62-expressing cells (Fig. 5A). Again there was no significant difference between cells expressing the siRNA resistant Cer–p62 and Cerulean alone, in both control siRNA and p62 knockdown conditions. These results suggest that NBR1 is sufficient to target peroxisomes to lysosomes for degradation in the absence of p62.

The observations that p62 was not necessary in the presence of elevated NBR1 expression (Fig. 5A) but was required at endogenous NBR1 levels (Fig. 1) suggest that p62 may play a secondary role in the targeting of peroxisomes to lysosomes. To test this hypothesis, we compared the efficiency of pexophagy in cells expressing a mutant of NBR1, which is deficient in p62-binding (NBR1 D50R) to wild-type NBR1. NBR1 interacts with p62 via its PB1 domain and a single residue mutation within the PB1 domain (D50R) prevents this interaction (Kirkin et al., 2009a). To directly compare the D50R mutant with wild-type NBR1, siRNA-resistant (SR) versions of NBR1 constructs were expressed in cells where endogenous NBR1 was knocked down using siRNA. Under these conditions the NBR1 D50R mutant was less efficient in inducing pexophagy than wild-type NBR1 (Fig. 5B). The decrease in efficiency was not due to the instability of the mutant protein, since the expression of this and subsequently discussed mutant NBR1 proteins was checked by western blot analysis (supplementary material Fig. S5B).

To verify that p62 does interact with NBR1 during pexophagy, we examined the cellular localization of endogenous p62 in cells expressing Cherry–NBR1. We found that endogenous p62 colocalized with the NBR1/peroxisome clusters (Fig. 5C). Interestingly, close examination of these NBR1/p62/peroxisome clusters showed that p62 did not completely overlap with NBR1, but also formed punctate structures juxtaposed to the NBR1/peroxisome cluster structures (Fig. 5C).

Next we examined whether p62 was required for NBR1/peroxisome cluster formation. We found that even upon knocking down the endogenous p62, NBR1 was able to induce clustering of peroxisomes (Fig. 5D). Combined with the observation that elevated p62 expression alone did not cause peroxisome degradation (Fig. 4B; Fig. 5A), but was required for pexophagy at endogenous levels of NBR1 (Fig. 1), the above results suggest that p62 is not involved in the clustering and may not be required for NBR1 interaction with peroxisomes. Its role in pexophagy may be to increase the efficiency of NBR1-mediated targeting of peroxisomes to autophagosomes.

The homo-oligomerization of p62 is required for its targeting to nascent autophagosomes (Itakura and Mizushima, 2011). p62 homo-oligomerizes through the self-interaction of its PB1 domain (Lamark et al., 2003; Wilson et al., 2003). Although NBR1 also possesses a PB1 domain (through which it interacts with p62), it instead forms homo-oligomers via its coiled-coil domain (CC1) (Kirkin et al., 2009a). It is not known, however, whether self-oligomerization is required for targeting its cargo to autophagosomes. In order to address this question, we expressed the NBR1 construct, NBR1 ΔCC1, where the coiled-coil domain was deleted. This construct is defective in homo-oligomerization (Kirkin et al., 2009a). Although this mutant can still cause peroxisome clustering, clustering was less efficient in the absence of endogenous wild-type NBR1 (Fig. 6A,B). Additionally, the RG-lysosome assay indicated that it targeted peroxisomes to lysosomes much less efficiently than wild-type NBR1 (Fig. 6C). In fact, there was no significant difference between NBR1 ΔCC1 compared to the control vector (Cer). Therefore, these results suggest that homo-oligomerization of NBR1 is required for efficient targeting of peroxisomes to autophagosomes.

One characteristic of autophagy receptors is their ability to interface between the substrate and autophagosomes. The latter is accomplished by the LC3-interacting region (LIR), which is required for autophagy receptors to bind the autophagy factor LC3. It has been previously shown that NBR1 mediates targeting of polyubiquitylated aggregates to autophagosomes and that NBR1 contains two LIRs (LIR1 and LIR2) (Kirkin et al., 2009a). To determine whether the LIRs of NBR1 are required for pexophagy we constructed an NBR1 mutant where LIR1 is mutated and LIR2 deleted. This construct, called NBR1(LIR_mut), formed peroxisome clusters at similar levels as wild-type NBR1 (Fig. 6D). However, its pexophagy activity was significantly decreased compared to the wild-type NBR1 (Fig. 6E), suggesting that the LIR motifs are required for pexophagy and not for NBR1/peroxisome clustering.

Previously we showed that an increase in ubiquitylated proteins on peroxisomes resulted in their targeting to autophagosomes for degradation (Kim et al., 2008). This suggests that the ubiquitin binding domain of NBR1, the UBA domain, may be required for binding to peroxisomes. Recently, the JUBA domain (hereafter for simplicity referred to as the J domain) – a novel, membrane-interacting, amphipathic α-helix region preceding the UBA domain – was found to be necessary for binding to endosomes (Mardakheh et al., 2010). Hence, we wanted to test if the J domain is also required for peroxisome binding.

First, we determined whether these domains were required for clustering of peroxisomes with NBR1. When both the J and UBA domains were deleted (GFP–NBR1 ΔJ-ΔUBA), clustering was not observed (Fig. 7A,E). Similarly, deletion of the J domain (NBR1 ΔJ) also abrogated formation of peroxisome/NBR1 clusters (Fig. 7B,E). However, when only the UBA was deleted and the J domain retained (NBR1 ΔUBA), clustering occurred (Fig. 7C). The extent of clustering was quantified by calculating the percentage of NBR1 that colocalized with peroxisomes, using the Manders' colocalization coefficient. The colocalization of NBR1 ΔUBA with peroxisomes was similar to that of wild-type NBR1 (Fig. 7E).

Interestingly, the NBR1 ΔUBA mutant localized to both punctate and tubular structures that resembled mitochondria (Fig. 7C). Upon further investigation, this mutant was indeed found to colocalize with the mitochondrial marker mito-RFP in HeLa cells (Fig. 7D) and COS7 cells (data not shown). However, NBR1, NBR1 ΔJ and NBR1 ΔJ-ΔUBA were not found to colocalize with mitochondria (data not shown). We also examined the colocalization of the NBR1 ΔUBA mutant with other organelles. A comparison of colocalization coefficients of NBR1 ΔUBA versus the wild-type NBR1 show that there is increased colocalization of NBR1 ΔUBA with the ER, lysosomes, the Golgi apparatus and mitochondria (supplementary material Fig. S6). p62 does not have any domain homologous to the J domain. When we fused the J domain to p62 with its own UBA domain deleted, we also found that p62-J_N-ΔUBA_N localized to mitochondria (supplementary material Fig. S7D–F). The wild-type p62 or another variant where the J domain is located N-terminal to the UBA domain of p62 does not localize to mitochondria (supplementary material Fig. S7A–C). Therefore, these observations suggest that the J domain is capable of binding membrane bilayers other than peroxisomes in the absence of the UBA domain. Hence, the UBA domain may play a regulatory role in the targeting of NBR1 to specific membranes.

To determine whether the various NBR1 deletion mutants can promote pexophagy, we coexpressed them with GFP–Cherry–PEX26_TM and subjected them to the RG-lysosome assay. The three mutants NBR1 ΔUBA, NBR1 ΔJ and NBR1 ΔJ-ΔUBA did not show any significant increase in pexophagy compared to control (Fig. 7F). Only the wild-type NBR1, containing both the J and UBA domains, resulted in a significant increase in pexophagy. Therefore, these results show that although the J domain is required for peroxisome binding, both the J domain and UBA domains are required for targeting peroxisomes to autophagosomes for lysosomal degradation.

To further test the role of the UBA and J domains in pexophagy, we examined whether they can confer pexophagy activity on p62. p62 has similar domain architecture to NBR1; it contains all the domains required to induce pexophagy, except for the amphipathic J domain preceding the UBA domain (Fig. 8A). Like NBR1, p62 forms homo-oligomers, possesses a UBA domain and possesses an LIR domain. Yet p62 cannot induce pexophagy upon overexpression in cells (Fig. 4B). Therefore, we asked whether the J-domain of NBR1 confers its specificity for peroxisomes. To test this hypothesis, we constructed a number of domain-swap mutants of p62 where its UBA domain was replaced with the J and/or UBA domains of NBR1 (Fig. 8A). We also inserted the J domain of NBR1 immediately upstream of the UBA domain in p62 (p62-J_N-UBA_p). All chimeras containing the J domain exhibited clustering with peroxisomes (Fig. 8B), but when they were subjected to the RG-lysosome assay, only the p62 chimeras containing both the J domain and a UBA domain (of either NBR1 or p62) were able to cause an increase in pexophagy (Fig. 8C; supplementary material Fig. S8). Notably, the UBA domains of NBR1 and p62 were interchangeable and both promoted pexophagy, provided that the J domain was also present. The constructs containing only the J domain or UBA domain of NBR1 did not induce pexophagy above the level found for wild-type p62 and Cer control. Together with the analysis of NBR1 mutants these data suggest that both the J and UBA domains are required to promote selective degradation of peroxisomes.

Next we asked how NBR1 might be recruited to peroxisomes. The importance of the UBA domain for selective binding of NBR1 to peroxisomes suggests that NBR1 is likely binding to a ubiquitylated protein on peroxisomes. One possible candidate is PEX5, a cytosolic receptor for peroxisomal matrix proteins. Peroxisome matrix proteins are imported into peroxisomes by the formation of a transient pore that consists of at least two proteins, PEX14 and PEX5. PEX14 is the membrane receptor that recruits the matrix protein receptor PEX5 to the peroxisomal membrane, and together they form a transient pore (Meinecke et al., 2010). It is thought that this transient pore is disassembled by ubiquitylation of PEX5 (Platta et al., 2007). We wanted to test if NBR1 is recruited to peroxisomes upon binding to ubiquitylated PEX5. To test this hypothesis we prevented PEX5 recruitment to peroxisomes by knocking down the expression of PEX14. Using quantitative PCR we confirmed that there was a 74% decrease in PEX14 mRNA upon treating HeLa cells with siRNA against PEX14 (Fig. 8D). Furthermore we observed that knocking down PEX14 disrupted the import of a PEX5 cargo, catalase, into peroxisomes, (supplementary material Fig. S9A, punctate versus cytosolic), suggesting that PEX14 knockdown prevents PEX5 recruitment to peroxisomes.

To determine whether NBR1 required ubiquitylated PEX5 to target to peroxisomes, we exogenously expressed Cherry–NBR1 in PEX14 siRNA knockdown cells. While Cherry–NBR1 clustered with PMP70-positive structures in PEX14-containing cells, in PEX14 depleted cells we found that Cherry–NBR1 did not colocalize with PMP70 positive structures, suggesting that it did not bind to peroxisomes (supplementary material Fig. S9B–E). When we examined these cells for changes in pexophagy upon NBR1 expression, we found that there was a significant decrease in pexophagy compared to control siRNA-treated cells. Yet, pexophagy was still higher than control (Fig. 8E). This may be the result of the incomplete knockdown of PEX14 (Fig. 8D; supplementary material Fig. S9), or it is equally likely that NBR1 is also binding to another ubiquitylated peroxisomal membrane protein. Nevertheless, these experiments suggest that the ubiquitylated PEX5 may be one component that recruits NBR1 to peroxisomes.

Peroxisome numbers are highly regulated. Pexophagy rapidly removes superfluous or damaged peroxisomes (Till et al., 2012). Peroxisomes are also turned over in response to various cellular signals, such as starvation conditions. During amino acid deprivation peroxisomes are believed to be selectively degraded while mitochondria are protected from autophagic degradation (Gomes et al., 2011; Hara-Kuge and Fujiki, 2008; Rambold et al., 2011). The ubiquitin binding autophagy receptors are involved in selective targeting of organelles to autophagosomes for degradation. Here we demonstrate that substrate selectivity may be partly conferred by the autophagy receptors themselves. Using peroxisomes as substrates for selective autophagy we showed that NBR1, and not p62, is necessary and sufficient for the turnover of endogenous peroxisomes.

We report that the J and UBA domains of NBR1 are required for its specificity for peroxisomes. These domains confer specificity by mediating the selective binding of NBR1 to peroxisomes. An essential role for the J domain is supported by the observation that mutants lacking the J domain (NBR1 ΔJ and NBR1 ΔJΔUBA) did not colocalize with peroxisomes (Fig. 7). However, based on our results we hypothesize that the UBA domain is required for the specific recruitment of NBR1 to the peroxisomal membrane. NBR1 without the UBA domain (NBR1 ΔUBA) formed peroxisome/NBR1 clusters, but also exhibited increased colocalization with other organelle markers, such as the mitochondria, which was not observed with full length NBR1 (supplementary material Fig. S6). Similar non-selective binding to various bilayers via the J domain was also observed with the p62-J_N-ΔUBA construct, where the UBA domain of p62 was replaced with the J domain of NBR1 (Fig. 8; supplementary material Fig. S7). Adding a UBA domain from either p62 or NBR1 to the J domain (p62-J_N-UBA_P/N) prevented the nonspecific binding to mitochondria (supplementary material Fig. S7). These results suggest that the UBA domain may act as a regulator of NBR1 binding to peroxisomes, since the J domain in the absence of the UBA domain appears to bind non-specifically to any lipid bilayer.

There is a precedent for this type of membrane binding mechanism. A number of PI3P binding proteins, such as EEA1 and Vps27, target and bind to PI3P rich membranes via a two-step binding mechanism (reviewed in Kutateladze, 2006). Both EEA1 and Vps27 possess the PI3P binding domain, FYVE, and a membrane interaction loop (MIL). These proteins are initially targeted to a specific membrane by interaction of the FYVE domain with PI3P, resulting in a conformational change exposing the membrane interaction loop domain to insert into the lipid bilayer to anchor the protein to the target organelle. In fact, phosphoinositide-binding proteins are known to act as detectors of coincident localization signals to allow localization to specific membranes and membrane domains (reviewed in Carlton and Cullen, 2005). So, NBR1 may act as a coincidence detector requiring both the coincident biding of the UBA domain and the J domain to peroxisomes to effectively target peroxisomes for autophagic degradation. We favor this mechanism as it explains the non-specific binding of the UBA deletion mutant (NBR1 ΔUBA) and the inability of the J domain deletion mutant (NBR1 ΔJ) to bind to peroxisomes.

Interestingly, although NBR1 ΔUBA was able to bind peroxisomes, it was not able to increase the pexophagy levels. Instead, an increase in pexophagy required both the J and UBA domains. The inability of the J domain alone to induce pexophagy may in part be explained by the fact that it binds to more different organelles (substrates) than the wild type, which appears to bind mainly to peroxisomes. The non-selective binding of this mutant to a number of subcellular membranes may result in an increase in autophagy substrates, causing an effective decrease in peroxisome degradation.

The need for a UBA domain on NBR1 to induce pexophagy further supports our previous report showing that increasing ubiquitin moieties on the cytosolic face of peroxisomes induced pexophagy (Kim et al., 2008). However, the question remains as to which peroxisomal protein(s) is ubiquitylated to induce pexophagy. Here we demonstrate that one possible peroxisomal protein target is PEX5, a component of the peroxisomal transient pore complex that is required for the import of peroxisomal matrix proteins. PEX5 is likely the most frequently ubiquitylated protein on the peroxisomal membrane, as it must be removed to disassemble the transient pore and be recycled back to the cytosol to import other peroxisomal matrix proteins (Platta et al., 2007; Meinecke et al., 2010). We postulate that the overexpression of NBR1 increases the rate of NBR1 interaction with the ubiquitylated PEX5. This stabilizes the J-domain-mediated binding of NBR1 to the bilayer, resulting in the accumulation of NBR1 on the peroxisomal membrane. Pexophagy is likely induced when there is sufficient accumulation of NBR1 on peroxisomes. In other words, in order to target peroxisomes to autophagosomes, a ‘critical mass,’ or number of autophagy receptors on peroxisomes is likely required to target these organelles to nascent autophagosomes.

We found that knocking down p62 expression results in an increase in peroxisome numbers, but not to the same levels as observed for NBR1 or ATG12 knockdown (Fig. 1). Although we find that p62 is not required for pexophagy in an NBR1 overexpression system, it does increase efficiency. Thus, based on our critical mass hypothesis, NBR1 on peroxisomes likely recruits p62 in order to increase the number of autophagy receptors on the peroxisome surface and achieve the ‘critical mass’ required for efficient pexophagy. This is further supported by the observation that although both NBR1 and p62 can form independent punctate structures (Kirkin et al., 2009a), they colocalize with each other on the substrate (Fig. 6C). Interestingly, the colocalization of p62 with NBR1 was only partial as there were distinct NBR1 and p62 structures juxtaposed to each other around peroxisomes. These structures may be analogous to the microdomains formed by the autophagy receptors p62 and NDP52 around Salmonella inside autophagosomes (Cemma et al., 2011; Mostowy et al., 2011; Zheng et al., 2009). However, the nature of these microdomains is not known.

Our finding that NBR1 is a receptor for pexophagy in mammalian cells suggests a mechanistic difference between the yeast and mammalian systems. Recently, Atg30 has been described to be the autophagy receptor for peroxisome turnover in Pichia pastoris (Farré et al., 2008). This protein is found localized on the peroxisomal membrane via its interaction with two peroxisomal membrane proteins, Pex14p and Pex3p. Unlike mammalian autophagy receptors it does not directly interact with Atg8, the yeast LC3 homologue, but instead interacts with Atg11 and Atg17 in order to target peroxisomes to autophagosomes. This interaction with the Atg11–Atg17 complex requires the phosphorylation of Atg30. Furthermore, Atg30 does not have an obvious ubiquitin binding domain, suggesting that its targeting to peroxisomes is not ubiquitin dependent. However, a pexophagy receptor, Atg36, that binds to Pex3 on peroxisomes and to Atg8 and the adaptor Atg11 was recently discovered in Saccharomyces cerevisiae (Motley et al., 2012). Apart from the lack of ubiquitin in the targeting role, this is more similar to the situation with NBR1 in mammalian pexophagy described here.

Cherry-NBR1, GFP-NBR1(ΔCC1), Cherry-NBR1(D50R), Cherry-p62, Lamp1-GFP, Lamp1-Cer, ss-GFP-KDEL, mito-RFP, mito-GFP, GFP-LC3 and PMP34-GFP were described previously (Kim et al., 2008; Kirkin et al., 2009a; Mitra et al., 2009). All other plasmids were constructed using standard protocols or Gateway recombination cloning, and were sequenced for confirmation (The Centre for Applied Genomics, Toronto, ON, Canada, and Sequencing Core Facility, University of Tromsø, Norway). Primers were produced by Sigma Genosys (Oakville, ON, Canada) and Invitrogen. Cloning strategies and primer sequences used are available upon request.

The GFP-NBR1(ΔJΔUBA) construct contains the first 875 codons of the NBR1 gene, while GFP-NBR1(JΔUBA) has the first 919 codons of NBR1 within Clontech pEGFP-C1. NBR1(D50R)(ΔJ) group of plasmids were constructed from NBR1(D50R) plasmid but the base pairs pertaining to residues 877–902 were deleted. For simplicity, these constructs, although D50R mutants, are referred to as NBR1(ΔJ) in the text and figures. SiRNA-resistant (SR) alleles of wild-type NBR1 and the NBR1(ΔCC1) and NBR1(D50R) mutants were constructed by introducing a silent mutation at L839 by site-directed mutagenesis. A schematic representation of the constructs of Cer-p62-J_NUBA_N, p62-UBA_N, p62-J_N-ΔUBA and p62-J_N-UBA_p are shown in Fig. 8A.

GFP-Cherry-PEX26_TM was constructed in several steps. First, the PEX26 ORF from pSPORT-PEX26 (Invitrogen) was cloned into the EcoRI and SalI sites of pEGFP-C1, and then it was truncated by PCR. Finally, the mCherry gene was inserted into the BglII and BamHI sites of the vector to generate GFP-Cherry-PEX26_TM.

Leupeptin (Bioshop, or Enzo Life Sciences) was used at 0.5 mM and E-64 (Enzo Life Sciences) at 2 µM. 3-methyladenine (3-MA; Sigma-Aldrich) was dissolved directly in media at 10 mM. siRNA targeting NBR1 (5′-GGAGUGGAUUUACCAGUUAUU-3′), p62 (5′-GCAUUGAAGUUGAUAUCGAUTT-3′), ATG12 (5′-GUGGGCAGUAGAGCGAACATT-3′), and non-targeting controls (5′-UUCUCCGAACGUGUCACGUTT-3′ or 5′-UAAGGCUAUGAAGAGAUACTT-3′) were produced by GenePharma (Shanghai, China). The NDP52 siRNA (5′-UUCAGUUGAAGCAGCUCUGUCUCCC-3′) and the rabbit polyclonal anti-NDP52 was a gift from Dr John Brumell.

Primary antibodies used in this study were rabbit polyclonal anti-catalase (EMD Biosciences, Calbiochem, Darmstadt, Germany), mouse monoclonal anti-p62 (BD Biosciences), rabbit polyclonal anti-NBR1 (Kirkin et al., 2009a), mouse monoclonal anti-NBR1 (Abnova, Taipei City, Taiwan), rabbit polyclonal anti-GFP (T. Johansen) and mouse anti-actin (Sigma-Aldrich). All fluorescent anti-mouse and anti-rabbit Alexa Fluor 488, 568 and 633 secondary antibodies were from Invitrogen. HRP-conjugated goat anti-mouse IgG and goat anti-rabbit IgG secondary antibodies were from Cedarlane. 10 nm Protein A gold (CMC, Utrecht, The Netherlands).

HeLa and MEF cells were grown in DMEM supplemented with 10% fetal bovine serum (FBS; Invitrogen). For live-cell imaging, cells were grown, manipulated and imaged in LabTek chamber slides (Nalgene Nunc, Rochester, NY). For immunofluorescence, cells were grown either in LabTek chamber slides or on no. 1 glass coverslips. Plasmids were transfected using Fugene-HD (Roche) or Lipofectamine-2000 (Invitrogen) or TransIT LT1 (Mirus Bio). siRNAs were transfected using Lipofectamine-2000 according to the manufacturer's instructions. For simultaneous siRNA knockdown and plasmid overexpression, siRNA alone was transfected on the first day, and siRNA and plasmids were transfected simultaneously on the second day, using Lipofectamine-2000. For the RG-lysosome assay, cells were treated with leupeptin and E-64 beginning on the day after plasmid transfection, to prevent degradation of protein inside lysosomes and allow detection of Cherry signal. When specified, cells were treated with 3-MA, beginning on the day after plasmid transfection. Cells were imaged 2 and/or 3 days after plasmid transfection. Prior to live imaging, the medium was changed to CO₂-independent medium (Invitrogen) containing 10% FBS, and leupeptin, E-64, and 3-MA, as needed.

Laser-scanning confocal microscopy was performed on a Zeiss LSM710 with a 63×/1.4 Plan-Apochromat oil objective or on a Leica TCS SP5 with a 63×/1.2 NA water immersion objective. When required, images were acquired in Z-sections of 0.5 to 1.0 µM thickness. Live cell imaging was performed at 37°C in CO₂-independent medium containing 10% FBS, and leupeptin, E-64 and 3-MA, as specified. For immunofluorescence, cells were fixed using 3.7% paraformaldehyde (Electron Microscopy Sciences) in PBS for 15 minutes and permeabilized using 0.1% Triton X-100 (Fisher Scientific) in PBS for 15 minutes, followed by incubation with the appropriate primary and secondary antibodies for 2 hours (or overnight at 4°C), and 1 hour, respectively. For all NBR1/peroxisome clustering experiments, the same pre-set microscopy acquisition setting was always used (laser power, master gain, etc.) to ensure that images from experiments performed on different days could be directly compared.

Correlative light and electron microscopy (CLEM) was performed on HeLa cells grown on gridded coverslips (EMS, Hatfield, PA, USA) and transfected with GFP-NBR1 alone or double transfected with GFP-PMP34 and Cherry-NBR1 for 24 hours as described earlier. Cells were fixed in 4% formaldehyde, 0.2% glutaraldehyde in 0.1 M phosphate buffer, pH 7.2, embedded in Moviol and observed on a confocal microscope (LSM 710, Carl Zeiss MicroImaging, Inc.) with 63× or 20× objectives, using the appropriate channels and differential interference contrast (DIC) microscopy. We used maximum projections from Z-stacks (seven slices, 4–6 µm total height) taken at 63× magnification for the final overlay images. The localization of cells of interest on the gridded coverslip was recorded using DIC microscopy. After detaching the coverslips in phosphate buffer (1–2 hours at room temperature) they were fixed in 2% glutaraldehyde, followed by 2% OsO₄ and 1.5% KFeCN in the same buffer. Coverslips were then stained en bloc with 4% uranyl acetate for 1 hour, dehydrated in graded ethanol series and embedded with Epon filled gelatin capsules placed on top of the coverslip. The coverslips were removed after polymerization with 48% hydrofluoric acid (Merck, Germany) and the blocks were then trimmed down to the regions previously identified on the confocal microscope and now imprinted on the Epon block. Sections of 150–200 nm thickness were then cut parallel to the substratum on a Leica Ultracut and post-stained with lead citrate (2 minutes, Fig. 8B–D) or not stained (Fig. 8F–I). Electron micrographs were taken at indicated magnifications and overlaid with the confocal images using Adobe Photoshop. Importantly, we would like to point out that the confocal images represent total fluorescence from the whole cell volume, whereas electron micrographs only represent 150–200 nm thick sections, thereby explaining the fluorescent signals outside apparent cell borders and above the nuclei. Cells for immuno-EM were prepared as described previously (Bjørkøy et al., 2005) and immunolabeled with antibodies against GFP followed by 10 nm Protein A gold.

Images were analyzed using Volocity software (Perkin Elmer). For quantification of peroxisome intensity by catalase staining, all conditions were treated simultaneously using the same antibody conditions. All mean intensity was normalized against the control siRNA performed on the same day. Furthermore, to capture the mean intensity from the catalase staining within the whole cell, images were acquired with an open pinhole. At least 100 cells were quantified, normalized against the control and averaged per experiment. For quantification of peroxisome density in MEFs, the number of catalase-stained structures was divided by the volume of each cell, and the averages were normalized to the Atg5^+/+ untransfected controls. At least 50 cells were analyzed for each condition.

For quantification of the RG-lysosome assay, all images were acquired using settings calibrated such that the GFP and mCherry fluorescence intensities (I_G/R) were approximately equal (i.e. I_GFP/I_mCherry = 1). We found that this acquisition setting gave equivalent readings even on different days. Therefore a preset acquisition setting was used for all RG-lysosome assays. A cell was considered to be undergoing increased pexophagy if the number of ‘red-only’ pixels were greater than 20% of the total ‘red’ pixels. To calculate the red-only pixels, all the pixels with Cherry signal above a predetermined threshold were first determined, and both the GFP (green) and Cherry (red) signals were collected for each of these pixels. A pixel was considered red-only if the Cherry fluorescent signal was at least three times higher than the GFP fluorescent signal. A cell was considered to have an increase in pexophagy if at least 20% of the Cherry containing pixels were red-only (supplementary material Fig. 4). The percentage of cells with 20% or more red-only pixels was reported.

Manders' thresholded colocalization coefficients were determined for at least 50 cells for each condition using ImageJ (NIH, Bethesda) and the JaCoP plugin (Bolte and Cordelières, 2006). Cells above a pre-determined expression level, as determined by the fluorescence signal from NBR1 expression levels, were used. The threshold was set to exclude structures not corresponding to NBR1/p62 puncta or peroxisomes. The percentages reported correspond to the amount of mCherry–positive pixels that are also catalase-positive. Error bars represent standard error of the mean. For statistical significance a two-tailed t-test was performed using Microsoft Excel 2010.

Cells were lysed with 100 mM Tris pH 9 containing 1% SDS (Bio-Rad) and Halt protease inhibitor cocktail (Thermo Scientific), and the lysate was heated at 100°C with vortexing for 20 minutes, then centrifuged at 13,000 g for 20 minutes. The protein concentration in the supernatant was determined by BCA assay (EMD Novagen), equivalent sample amounts were subjected to SDS-PAGE, and protein was transferred to 0.45 µm BioTrace PVDF membrane (Pall) and probed with the appropriate primary and HRP-conjugated secondary antibodies following standard protocols. Blots were developed using Luminata Crescendo (Millipore), ECL Plus or ECL Advance reagents (GE Healthcare) and detected on blue film (AGFA) or Hyperfilm (GE Healthcare).

We are indebted to Dr John Brumell for reagents and helpful discussion pertaining to the project and the BioImaging core facility at the Institute of Medical Biology, University of Tromsø for the use of instrumentation and expert assistance.