Mitochondrial Stasis Reveals p62-mediated Ubiquitination in Parkin-independent Mitophagy and Mitigates Nonalcoholic Fatty Liver Disease

Summary

It is unknown what occurs if both mitochondrial division and fusion are completely blocked. Here, we introduced mitochondrial stasis by deleting two dynamin-related GTPases for division (Drp1) and fusion (Opa1) in livers. Mitochondrial stasis rescues liver damage and hypotrophy caused by the single knockout (KO). At the cellular level, mitochondrial stasis reestablishes mitochondrial size and rescues mitophagy defects caused by division deficiency. Using Drp1KO livers, we found that the autophagy adaptor protein p62/sequestosome-1—which is thought to function downstream of ubiquitination—promotes mitochondrial ubiquitination. p62 recruits two subunits of a cullin-RING ubiquitin E3 ligase complex, Keap1 and Rbx1, to mitochondria. Resembling Drp1KO, diet-induced nonalcoholic fatty livers enlarge mitochondria and accumulate mitophagy intermediates. Resembling Drp1Opa1KO, Opa1KO rescues liver damage in this disease model. Our data provide a new concept that mitochondrial stasis leads the spatial dimension of mitochondria to a stationary equilibrium and a new mechanism for mitochondrial ubiquitination in mitophagy.

eTOC

Yamada et al show that mitochondrial size is antagonistically regulated by division and fusion and that extreme mitochondrial size is deleterious in the liver. They identify a new parkin-independent mitophagy pathway, which is inhibited by mega mitochondria present in nonalcoholic liver hepatocytes. Restoring mitophagy could be beneficial in NAFLD.

Introduction

A hallmark of mitochondria is their dynamic nature: they divide and fuse in most, if not all, cells ranging from unicellular organisms such as yeast to multicellular organisms such as humans (Friedman and Nunnari, 2014; Kameoka et al., 2018; Liesa and Shirihai, 2013; Roy et al., 2015). Although the exact roles of mitochondrial division and fusion remain to be determined, these processes have been shown to play critical roles in the function, distribution and turnover of mitochondria.

The most direct consequence of mitochondrial division and fusion is an alteration in the size of this critical organelle (Elgass et al., 2013; Karbowski and Neutzner, 2012; Nagdas and Kashatus, 2017; Pernas and Scorrano, 2016; Shutt and McBride, 2012). Mitochondrial division creates small mitochondria, and mitochondrial fusion produces large mitochondria. The frequency of division and fusion is balanced to maintain mitochondrial size in healthy cells. However, under pathological conditions, significantly larger or smaller mitochondria have been observed. For example, in nonalcoholic fatty livers, malnutrition, myopathy and cancer cells, enlarged mitochondria, termed megamitochondria, have been reported (Caldwell et al., 1999; Kleiner and Makhlouf, 2016; Lotowska et al., 2014; Neuman et al., 2014; Noureddin et al., 2013; Wakabayashi, 2002). It is largely unknown whether an altered mitochondrial size or imbalanced activities of division or fusion contribute to these pathological outcomes.

Conserved GTPases that belong to the dynamin superfamily mediate mitochondrial division and fusion. Dynamin-related protein 1 (Drp1) and dynamin-2 constrict and cut mitochondria, and mitofusin 1 and 2 and optic atrophy 1 (Opa1) fuse the mitochondrial membranes (Bui and Shaw, 2013; Ishihara et al., 2013; Lee et al., 2016; McBride and Frost, 2016; Mishra and Chan, 2014; Tamura et al., 2011; van der Bliek et al., 2013). All of these GTPases are mutated in human diseases: mutations in Drp1 cause neurodevelopmental disorders, dynamin-2 and mitofusin 2 are defective in Charcot-Marie-Tooth neuropathy and Opa1 is defective in dominant optic atrophy (Itoh et al., 2013; MacVicar and Langer, 2016; Serasinghe and Chipuk, 2017). To counteract unwanted, unequal mitochondrial division and fusion, cells appear to have mechanisms that readjust their equilibrium. The loss of mitochondrial division results in decreased levels of mitofusin 1 and 2 through proteosomal degradation and inactivation of Opa1 through excess proteolytic cleavage (Saita et al., 2016; Wakabayashi et al., 2009).

One of the key roles of mitochondrial division appears to be to facilitate turnover of mitochondria (Lemasters, 2014; Rambold and Lippincott-Schwartz, 2011; Tanaka et al., 2010; Twig et al., 2008; Williams and Ding, 2015). We have shown that Drp1-mediated mitochondrial division promotes mitophagy that involves ubiquitination of mitochondrial proteins in a manner independent of the ubiquitin E3 ligase parkin (Kageyama et al., 2014; Kageyama et al., 2012). In the absence of Drp1, mitophagy is decreased and cells accumulate mitophagy intermediates containing ubiquitinated proteins, p62 and LC3, in the brain and heart in tissue-specific Drp1-knockout (KO) mice. Such a Drp1 loss causes cell death due to neurodegeneration and animal death due to lethal heart failure (Kageyama et al., 2014; Kageyama et al., 2012; Yamada et al., 2016; Yamada et al., 2017), making it difficult to further analyze mitochondrial dynamics. We are currently lacking experimental data to define the exact function of mitochondrial division in this parkin-independent mitophagy pathway. We also do not understand the molecular underpinnings of this new mitophagy pathway.

On the other hand, mitochondrial fusion is believed to facilitate the mixing of mitochondrial contents such as proteins, lipids and DNAs among individual mitochondria to maintain their homeostasis (Ishihara et al., 2013; Roy et al., 2015). However, the functional importance of such content mixing has not been directly tested since the inhibition of mitochondrial fusion also decreases the size of mitochondria. It is necessary to inhibit content mixing without compromising the size of mitochondria to unambiguously determine the function of mitochondrial fusion.

In the current study, we introduced mitochondrial stasis by completely blocking dynamin-mediated mitochondrial division and fusion using liver-specific KO mice for Drp1 and Opa1. We showed that mitochondrial stasis reestablishes mitochondrial size and restores the healthy status of the liver by alleviating pathological effects caused by single division or fusion deficiency. Using this animal system, we also showed that accumulation of mitophagy intermediates are cleared by reestablishing mitochondrial size. As a critical component in the parkin-independent mitophagy, we found that the autophagy adaptor protein p62/sequestosome-1 promotes mitochondrial ubiquitination by recruiting Rbx1 and Keap1, two components of a cullin-RING E3 ligase complex, to mitochondria. These findings formulate a new model for parkin-independent mitophagy because Rbx1 and Keap1 have not been implicated in mitophagy and because this novel function of p62 in mitochondrial ubiquitination is distinct from its known roles in targeting ubiquitinated mitochondria to autophagosomes. Furthermore, using a mouse model for nonalcoholic fatty liver disease (NAFLD), we showed that NAFLD hepatocytes accumulate large mitochondria that are associated with p62, Keap1, and ubiquitin, similar to Drp1KO hepatocytes. Knockout of Opa1 removed these mitophagy intermediates and rescued liver damage in this NAFLD mouse model. Therefore, balanced mitochondrial size, independent of division and fusion, controls organ integrity and parkin-independent mitophagy in the liver.

Results

To determine the impact of mitochondrial stasis on mitochondrial size and organ integrity, we deleted Drp1 and Opa1 individually and simultaneously in hepatocytes of mouse livers using the floxed DRP1 and OPA1 alleles (Wakabayashi et al., 2009; Zhang et al., 2011) in combination with Cre recombinase expressed from the hepatocyte specific albumin promoter (Postic et al., 1999) in Alb-Drp1KO mice (Alb-Cre^+/−::Drp1^flox/flox), Alb-Opa1KO mice (Alb-Cre^+/−::Opa1^flox/flox) and Alb-Drp1Opa1KO mice (Alb-Cre^+/−::Drp1^flox/flox::Opa1^flox/flox). As a control, Alb-Cre^+/− and Drp1^flox/flox mice were used. We focused on hepatocytes since Drp1 loss in hepatocytes does not kill cells or animals, unlike Drp1 loss in neurons or cardiomyocytes, which leads to neurodegeneration and lethal heart failure, respectively (Ishihara et al., 2015; Kageyama et al., 2014; Kageyama et al., 2012; Song et al., 2015; Wang et al., 2015). The level of Drp1 and Opa1 in the liver is modest compared with levels in the brain and heart (Fig. 1A–C). Upon Cre-mediated excision of their floxed alleles, levels of Drp1 and Opa1 were reduced by ~30% (Fig. S1A), revealing that the DRP1 and OPA1 genes are effectively deleted in hepatocytes because hepatocytes constitute ~70% of the total cells in the liver (Stanger, 2015). The levels of most of the tested mitochondrial proteins were not affected by the loss of Drp1 and Opa1 (Fig. S1A and B).

Figure 1. Mitochondrial stasis reestablishes mitochondrial size and liver integrity.

(A) Western blotting of tissues isolated from mice was performed using antibodies to Drp1, Opa1, PDH (a mitochondrial protein) and GAPDH (a cytosolic protein). Equal amounts of proteins were loaded for each tissue. (B and C) Quantification of band intensity of Drp1 (B) and Opa1 (C). Values are average ± SEM (n=3 mice). (D) Mitochondria were analyzed by confocal immunofluorescence microscopy of liver sections of the indicated mice at 3 months of age with anti-PDH antibodies. Boxed regions are enlarged. (E) Individual mitochondrial size (n=846 mitochondria from 4 control mice, 640 from 3 Alb-Drp1KO mice, 625 from 3 Alb-Opa1KO mice, 633 from 3 Alb-Drp1Opa1KO mice). (F) Average size of the mitochondria. Error bars indicate SEM (n=3–4 mice). (G) Body weight of the male mice at 3 months. Values are average ± SEM (n=3–6 mice). (H) Liver weight relative to body weight (n=5–8 mice). (I) Blood levels of ALT (n=5 mice). (J) Blood levels of triglyceride (n=5 mice). (K) Blood levels of cholesterol (n=5 mice). (L) Histological images of liver sections using PAS staining. (M) Liver sections of the indicated mice were analyzed by confocal microscopy with anti-E-cadherin antibodies and DAPI. (N) Individual size of cells. Red lines indicate averages (n=352 cells from 4 control mice, 264 from 3 Alb-Drp1KO mice, 264 from 3 Alb-Opa1KO mice, 264 from 3 Alb-Drp1Opa1KO mice). Statistical analysis was performed using Student’s t-test: *p< 0.05, **p< 0.01, ***p< 0.001. See also Figure S1.

Simultaneous loss of Drp1 and Opa1 reestablishes mitochondrial size in hepatocytes

To assess mitochondrial size, we performed confocal immunofluorescence microscopy of frozen liver sections with anti-pyruvate dehydrogenase (PDH) antibodies. Mitochondrial size in Drp1KO hepatocytes was significantly increased due to unopposed fusion in the absence of Drp1-mediated mitochondrial division, and the mitochondrial size in Opa1KO hepatocytes was significantly decreased due to unopposed division (Fig. 1D–F). When we knocked out both Drp1 and Opa1, mitochondria in Drp1Opa1KO hepatocytes returned to roughly their original size, similar to that of the control hepatocytes (Fig. 1D–F). These data show that Drp1-mediated division and Opa1-mediated fusion antagonistically maintain mitochondrial size, and that mitochondrial stasis reestablishes mitochondrial size in the absence of these activities.

To further examine mitochondrial structure in these knockout mice, we examined their livers using electron microscopy. Consistent with the immunofluorescence microscopy results, while mitochondrial size was increased in Drp1KO hepatocytes and decreased in Opa1KO hepatocytes, hepatocyte size appeared to be restored to normal levels in Drp1Opa1KO hepatocytes (Fig. S1C). We also found that the mitochondria in Drp1KO hepatocytes were elongated, with decreased circularity and increased aspect ratios (Fig. S1C–E). In contrast, the mitochondria in Opa1KO hepatocytes were rounder, with increased circularity and decreased aspect ratios. The shapes of the mitochondria in Drp1Opa1KO hepatocytes were between these two KO hepatocytes, with rounder shapes than control hepatocytes (Fig. S1C–E). In addition to mitochondrial fusion, Opa1 also controls the integrity of the inner membrane cristae (Pernas and Scorrano, 2016). We observed fewer cristae in Opa1KO hepatocytes (Fig. S1C and F). In contrast to mitochondrial size, this cristae defect was not rescued in Drp1Opa1KO hepatocytes (Fig. S1C and F).

Simultaneous loss of Drp1 and Opa1 restores the integrity and function of the liver

To examine the physiological consequence of individual and combined loss of Drp1 and Opa1, we first measured the weight of the mice and their livers. We found no significant difference in the body weight in control, Alb-Drp1KO, Alb-Opa1KO and Alb-Drp1Opa1KO mice (Fig. 1G). Although the overall appearance of the livers was similar among the mice (Fig. S1G), liver weight was significantly decreased in Alb-Opa1KO mice (Fig. 1H). Importantly, the normal weight of the liver was restored in Alb-Drp1Opa1KO mice (Fig. 1H). Second, blood levels of alanine aminotransferase (ALT), which is released from the liver upon liver injury, were significantly increased in Alb-Drp1KO mice (Fig. 1I). Simultaneous loss of Drp1 and Opa1 restored normal levels of ALT in Alb-Drp1Opa1KO mice (Fig. 1I). Third, livers regulate lipid homeostasis. We found that both Alb-Drp1KO and Alb-Opa1KO mice had increased levels of two major lipids, triglyceride and cholesterol (Fig. 1J and K); the levels of both lipids were restored in Alb-Drp1Opa1KO mice (Fig. 1J and K). In contrast, glucose homeostasis and insulin sensitivity were not affected in Alb-Drp1KO, Alb-Opa1KO, or Alb-Drp1Opa1KO mice (Fig. S1H and I). Fourth, during histological analysis, we noticed that the hepatocytes appeared to be larger in both the Alb-Drp1KO and Alb-Opa1KO mice (Fig. 1L). Hepatocytes increase in size upon liver damage (Gentric and Desdouets, 2014); indeed, confocal immunofluorescence microscopy with anti-cadherin antibodies showed significantly increased cell size in Drp1KO and Opa1KO hepatocytes, which was rescued in the Alb-Drp1Opa1KO mice (Fig. 1M and N). These results reveal that the loss of Drp1 or Opa1 alone compromises the integrity and function of the liver and that the concomitant loss of Drp1 and Opa1 rescues both pathologies. It appears that a balance of mitochondrial sizes, rather than mitochondrial division or fusion per se, is fundamental to the maintenance of a healthy liver.

Simultaneous loss of Drp1 and Opa1 restores parkin-independent mitophagy

In the absence of Drp1, neurons and cardiomyocytes increase mitochondrial size and accumulate mitophagy intermediates containing ubiquitin, the autophagy adaptor protein p62 and the autophagosomal protein LC3 (Kageyama et al., 2014; Kageyama et al., 2012). It has been suggested that mitochondrial division may separate the damaged part of mitochondria from the rest of mitochondria during mitophagy (Roy et al., 2015; Shirihai et al., 2015). However, it remains to be determined whether the accumulation of the mitophagy intermediates results from the loss of mitochondrial division or the increase in mitochondrial size. To disentangle the mechanistic contribution of mitochondrial division and size to the accumulation of mitophagy intermediates, we took advantage of mitochondrial stasis in which mitochondrial size is reestablished. We first investigated whether Drp1KO hepatocytes accumulate mitophagy intermediates using confocal microscopy with antibodies to PDH, p62, ubiquitin and LC3. We found that p62 accumulated around mitochondria in Drp1KO hepatocytes (Fig. 2A and B). p62 was colocalized with ubiquitin (Fig. 2C) and LC3 (Fig. 2D). The accumulation of p62 (Fig. 2A and B), ubiquitin (Fig. 2C) or LC3 (Fig. 2D) was barely detected in Drp1Opa1KO hepatocytes. In addition, Western blotting showed p62 levels were increased in Drp1KO livers and restored in Drp1Opa1KO livers (Fig. 2E). These results suggest that mitophagy intermediates accumulate due to the increase in mitochondrial size, rather than the loss of mitochondrial division (Fig. 2F). Ruling out the possibility that autophagy is globally induced in the absence of Drp1, we found no induction of macroautophagy, which is induced by starvation, in Drp1KO hepatocytes (Fig. S2).

Figure 2. Mitophagy intermediates containing p62, ubiquitin and LC3 accumulate in Drp1KO hepatocytes due to increase in mitochondrial size, not the lack of mitochondrial division.

(A) Confocal immunofluorescence microscopy of liver sections from the indicated mice at 3 months of age using antibodies to p62 and the mitochondrial protein PDH. Boxed regions are enlarged. (B) Cells that show p62 accumulation are quantified. Values represent the average ± SEM (n=3 mice). (C and D) Confocal microscopy of liver sections using antibodies to p62 and PDH along with ubiquitin (C) or LC3 (D). (E) Western blotting of livers isolated from the indicated mice at 3 months of age using antibodies p62 and GAPDH. Quantification of band intensity. Values are average ± SEM (n=3 mice). (F) Summary of the data. (G) Plasmids carrying Su9-mCherry-GFP were delivered to the livers of control, Alb-Drp1KO, and Alb-Drp1Opa1KO mice via hydrodynamic tail vein injection. Four days after injection, the livers were analyzed by confocal microscopy. (H) The mitophagy index was determined by measuring the relative area of mCherry fluorescence that did not overlap with GFP fluorescence over the total area of mCherry fluorescence in each cell using NIH ImageJ software and set 1 in the control mice. Values are average ± SEM (n=3–4 mice). Statistical analysis was performed using Student’s t-test: *p<0.05, **p< 0.01, ***p< 0.001. See also Figure S2.

To more directly examine mitophagy, we delivered plasmids that expressed a mitophagy biosensor (matrix-targeted Su9-mCherry-GFP) (McWilliams et al., 2016; Rojansky et al., 2016) to the livers of control, Alb-Drp1KO, and Alb-Drp1Opa1KO mice by hydrodynamic tail vein injection. Su9-mCherry-GFP is first imported into mitochondria, in which both mCherry and GFP signals are observed (McWilliams et al., 2016; Rojansky et al., 2016). When mitochondria are subjected to mitophagy and transported to lysosomes, only the mCherry signal is detected because the GFP signal is lost due to the acidic pH of the lysosomes. Four days after injection, we quantified the mCherry and GFP signals in liver sections. The results show that mitophagy was significantly decreased in Alb-Drp1KO mice (Fig. 2G and H), and this mitophagy defect was rescued in Alb-Drp1Opa1KO mice (Fig. 2G and H).

We have previously shown that Drp1 loss in neurons and cardiomyocytes decreases mitophagy that is independent of parkin, a Parkinson’s disease-associated ubiquitin ligase E3 (Kageyama et al., 2014; Kageyama et al., 2012). To test whether the accumulation of mitophagy intermediates requires parkin in the liver, we deleted PARK2 in addition to DRP1 in Alb-Drp1ParkinKO mice (Alb-Cre^+/−::Drp1^flox/flox::Parkin^−/−). We first tested the impact of parkin loss on mitochondria size. ParkinKO (Drp1^flox/flox::Parkin^−/−) hepatocytes showed normal mitochondrial sizes, and Drp1ParkinKO hepatocytes displayed increased mitochondrial sizes that were similar to those of Drp1KO hepatocytes (Fig. 3A–C). We then examined the accumulation of p62, ubiquitin and LC3 on mitochondria. In contrast to Alb-Drp1Opa1KO mice, Alb-Drp1ParkinKO mice persistently showed the accumulation of p62 (Fig. 3D and E), ubiquitin (Fig. 3F and G) and LC3 (Fig. 3H) around mitochondria. Therefore, mitochondrial ubiquitination is independent of parkin in Drp1KO hepatocytes (Fig. 3I).

Figure 3. Mitochondrial ubiquitination is independent of parkin in Alb-Drp1KO mice.

(A) Liver sections of the indicated mice were analyzed at 3 months of age using confocal microscopy using anti-PDH antibodies. Boxed regions are enlarged. (B) Individual mitochondrial size (n=887 mitochondria from 4 control mice, 654 from 3 Alb-Drp1KO mice, 611 from 3 ParkinKO mice, 640 from 3 Alb-Drp1ParkinKO mice). (C) Average size of mitochondria. Error bars indicate SEM (n=3–4 mice). (D) Confocal microscopy of liver sections using antibodies to p62 and PDH. (E) Cells that showed p62 accumulation are quantified. Values represent the average ± SEM (n=3 mice). (F and H) Confocal microscopy of liver sections using antibodies to p62 and PDH together with ubiquitin (F) or LC3 (H). (G) Cells that showed ubiquitin accumulation are quantified. Values represent the average ± SEM (n=3 mice). (I) Summary of the data. Statistical analysis was performed using Student’s t-test: **p< 0.01, ***p< 0.001. See also Figure S3.

Furthermore, we determined the impact of parkin loss in the integrity of the liver. The single loss of parkin did not affect body weight (Fig. S3A), liver weight (Fig. S3B), the level of blood ALT (Fig. S3C) or the size of hepatocytes (Fig. S3D–F) in ParkinKO mice. While additional loss of parkin in Drp1KO hepatocytes did not change ALT levels (Fig. S3C) in Alb-Drp1ParkinKO mice, cell size further increased in Drp1ParkinKO hepatocytes compared with single Drp1KO hepatocytes (Fig. S3E and F).

Mitochondrial stasis restores normal level of mitofusin 1 and 2

Instead of mitochondrial ubiquitination, we found that parkin regulates the levels of mitofusin 1 and 2 in the absence of Drp1. It has been shown that, upon the loss of Drp1-mediated mitochondrial division, three other dynamin-related proteins that control mitochondrial fusion are partially inactivated; the levels of two outer membrane proteins mitofusin 1 and 2 are decreased by proteosomal degradation, and the inner membrane protein Opa1 undergoes increased proteolytic cleavage by Oma1 (Saita et al., 2016; Wakabayashi et al., 2009) (Fig. S3G). Although this regulation of mitochondrial fusion proteins has been suggested as a mechanism for sensing and readjusting the balance between mitochondrial division and fusion, it is unclear how the loss of mitochondrial division is recognized in cells. Western blotting of livers showed that levels of both mitofusin 1 and 2 were significantly decreased in Drp1KO livers, and returned to roughly their original levels in Drp1Opa1KO livers (Fig. S3G). Interestingly, we found that the levels of mitofusin 1 and 2 were significantly increased in Alb-Drp1ParkinKO mice compared with Alb-Drp1KO mice (Fig. S3G). In contrast, we observed no changes in Opa1 processing in Alb-Drp1ParkinKO mice compared with Alb-Drp1KO mice (Fig. S3G). These results suggest that parkin specifically controls mitofusin 1 and 2 in response to increases in mitochondrial size. Both Alb-Drp1ParkinKO and Alb-Drp1Opa1KO mice showed restored levels of mitofusin 1 and 2 (Fig. S3G); however, the liver damage caused by Drp1 deficiency was only rescued in the Alb-Drp1Opa1KO mice. Therefore, the rescuing effect of Opa1KO in Drp1KO is not derived from restoring mitofusin levels.

p62 promotes mitochondrial ubiquitination

p62 contains a ubiquitin binding site and a LC3 binding site (Bitto et al., 2014). It has been shown that p62 connects ubiquitinated mitochondria to LC3-associated autophagosomes in mitophagy (Geetha et al., 2012; Katsuragi et al., 2015; Manley et al., 2013). The accumulation of p62 on mitochondria in Drp1KO hepatocytes prompted us to test this current view of the function of p62 in the parkin-independent mitophagy pathway. By breeding p62^−/− mice (Komatsu et al., 2007) with Alb-Drp1KO mice, we generated p62KO (Drp1^flox/flox::p62^−/−) mice and Alb-Drp1p62KO (Alb-Cre^+/−::Drp1^flox/flox::p62^−/−) mice. p62KO hepatocytes displayed normal mitochondrial sizes and Drp1p62KO hepatocytes exhibited increased mitochondrial sizes that were similar to those of Drp1KO hepatocytes (Fig. 4A–C). Also, similar to mitochondria in Drp1KO hepatocytes, mitochondria in Drp1p62KO hepatocytes were elongated with decreased circularity and increased aspect ratios (Fig. S1C–E). We then examined ubiquitin using confocal immunofluorescence microscopy. Unexpectedly, we found that mitochondrial ubiquitination was significantly decreased in Drp1p62KO hepatocytes compared with Drp1KO hepatocytes (Fig. 4D–F).

Figure 4. p62 promotes mitochondrial ubiquitination in Alb-Drp1KO mice.

(A) Liver sections of the indicated mice were analyzed at 3 months of age using confocal microscopy with anti-PDH antibodies. Boxed regions are enlarged. (B) Individual mitochondrial size (n=619 mitochondria from 3 control mice, 595 from 3 Alb-Drp1KO mice, 583 from 3 p62KO mice, 614 from 3 Alb-Drp1p62KO mice). (C) Average size of mitochondria. Error bars indicate SEM (n=3 mice). (D) Confocal microscopy of liver sections using antibodies to p62 and PDH along with ubiquitin. The p62 signals disappeared in p62KO and Drp1p62KO hepatocytes, demonstrating the specificity of the anti-p62 antibodies. (E) Confocal microscopy of liver sections using antibodies to ubiquitin and LC3. Dotted lines outline single cells. (F) Quantification of cells showing the accumulation of ubiquitin. Values are average ± SEM (n=3–4 mice). (G) Quantification of cells that show colocalization of ubiquitin to LC3 (n=3 mice). (H) Summary of the data. Statistical analysis was performed using Student’s t-test: *p< 0.05, **p< 0.01.

In addition to this new function of p62 in mitochondrial ubiquitination, we also confirmed the role of p62 in connecting ubiquitinated mitochondria to autophagosomes. We found that a small fraction of mitochondria still contained ubiquitinated proteins in Drp1p62KO hepatocytes (Fig. 4E and F). These mitochondria showed a significant decrease in association with LC3 (Fig. 4E and G). Taken together, our data suggest that p62 functions at two distinct steps in parkin-independent mitophagy: p62 promotes mitochondrial ubiquitination and recruits LC3-bound autophagosomes to the ubiquitinated mitochondria (Fig. 4H).

p62 recruits Keap1 to mitochondria in Drp1KO hepatocytes

How does p62 induce mitochondrial ubiquitination? We found that p62 on mitochondria is subjected to phosphorylation at serine 351 in Alb-Drp1KO mice (Fig. 5A), which is known to promote the direct interaction of p62 with Keap1 (Ichimura et al., 2013). Keap1 forms a cullin-RING ubiquitin ligase complex with the E3 ligase Rbx1 (Bitto et al., 2014; Katsuragi et al., 2015; Zhang et al., 2004). Confocal microscopy of liver sections showed that Keap1 is recruited to the mitochondria and colocalized with p62 in Drp1KO hepatocytes, but not in control hepatocytes (Fig. 5B and C). Importantly, this recruitment of Keap1 to mitochondria no longer persisted in Drp1p62KO and Drp1Opa1KO hepatocytes (Fig. 5B and C). Additional loss of parkin did not affect the mitochondrial localization of Keap1 in Drp1ParkinKO hepatocytes (Fig. 5B and C). In contrast to p62, whose levels were increased in Drp1KO hepatocytes, the Keap1 levels were not increased in Drp1KO hepatocytes as shown by Western blot analysis (Fig. 5D). Thus, Keap1 is recruited to mitochondria in a p62-dependent manner in Drp1KO hepatocytes (Fig. 5E).

Figure 5. Keap1 is recruited to mitochondria by interactions with p62 in Drp1KO hepatocytes and mediates mitochondrial ubiquitination.

(A) Liver sections of the indicated mice were analyzed at 3 months of age using confocal microscopy using antibodies to PDH, phospho-p62(S351) and p62. Boxed regions are enlarged. 87.9% of p62-positive mitochondria were also positive for the p62 phosphorylation (n=1048 mitochondria). (B) Liver sections were analyzed using confocal microscopy using anti-PDH antibodies, Keap1, and p62. Boxed regions are enlarged. (C) Cells that show Keap1 accumulation on mitochondria are quantified. Values represent the average ± SEM (n=3–5 mice). (D) Western blotting of livers using antibodies to p62, Keap1 and GAPDH. Quantification of band intensity is shown. Values are average ± SEM (n=3 mice). (E) Summary of the data. (F and G) Plasmids carrying WT p62-MffTM or Keap1-interaction-defective (KID) p62-MffTM were delivered to the livers of Alb-Drp1p62KO mice via hydrodynamic tail vein injection. Four days after injection, the livers were analyzed by confocal fluorescence microscopy. Quantification of p62 that overlapped with Keap1 (F) and ubiquitin (G) is shown. Values are average ± SEM (n=3 mice). Statistical analysis was performed using Student’s t-test: *p< 0.05, **p< 0.01, ***p< 0.001.

p62-Keap1 interactions promote mitochondrial ubiquitination

To determine whether the recruitment of Keap1 to mitochondria is required for mitochondrial ubiquitination, we anchored p62 to mitochondria using a transmembrane domain of the outer membrane protein Mff (p62-MffTM) and tested mitochondrial ubiquitination in the hepatocytes of Alb-Drp1p62KO mice. Forced tethering of p62 to mitochondria was critical since p62 can be recruited to mitochondria through interactions with ubiquitin on mitochondria (Katsuragi et al., 2015). We found that p62-MffTM recruited Keap1 and ubiquitin to mitochondria (Fig. 5F and G). We then expressed a Keap1-interaction-defective p62-MffTM in Drp1p62KO hepatocytes (Ichimura et al., 2013). In contrast to WT p62-MffTM, the Keap1-interaction-defective p62-MffTM failed to recruit Keap1 and ubiquitin (Fig. 5F and G). These data suggest that p62 promotes mitochondrial ubiquitination via association with Keap1.

Rbx1 is associated with mitochondria in Drp1KO hepatocytes and its knockdown decreases mitochondrial ubiquitination

We also tested whether Rbx1 is recruited to the mitochondria in Drp1KO hepatocytes. We delivered plasmids expressing HA-Rbx1 to the livers of control, Alb-Drp1KO, Alb-Drp1p62KO and Alb-Drp1Opa1KO mice by hydrodynamic tail vein injection. Eight hours after injection, the localization of HA-Rbx1 was analyzed using confocal immunofluorescence microscopy of liver sections with antibodies to HA, PDH, and p62. We found that HA-Rbx1 was associated with mitochondria and colocalized with p62 in Drp1KO hepatocytes, but not in control or Drp1p62KO hepatocytes (Fig. 6A). Furthermore, HA-Rbx1 was not associated with mitochondria in Drp1Opa1KO hepatocytes, similar to p62 and mitochondrial ubiquitination (Fig. 6A). Therefore, Rbx1 appears to fulfill the criteria for the E3 ligase to ubiquitinate mitochondrial proteins in a p62-dependent manner in Drp1KO hepatocytes.

Figure 6. Rbx1 is recruited to mitochondria in a p62-dependent manner in Drp1KO hepatocytes and mediates mitochondrial ubiquitination.

(A) Plasmids carrying HA-Rbx1 were delivered to the livers of control, Alb-Drp1KO, Alb-Drp1p62KO, and Alb-Drp1Opa1KO mice via hydrodynamic tail vein injection. Eight hours after injection, the livers were analyzed by confocal immunofluorescence microscopy with antibodies to PDH, HA, and p62. (B) Mouse embryonic fibroblasts were infected with lentiviruses expressing scramble or two independent Rbx1 shRNAs. Six days after infection, whole cell lysates were analyzed by Western blotting with antibodies to Rbx1, PDH and GAPDH. Quantification of band intensity is shown. Values are average ± SEM (n=3 experiments). (C and D) Plasmids expressing the indicated shRNAs along with GFP were introduced to the livers of Alb-Drp1KO mice (C) and Alb-Drp1ParkinKO mice (D) using tail vein injection. Four days after injection, the livers were analyzed by confocal microscopy with antibodies to GFP and HA. Quantification of GFP-positive cells that have the accumulation of ubiquitin is shown. Values are average ± SEM (n=3–4 mice). (E) Model for p62-mediated mitochondrial ubiquitination. p62 recruits Keap1 and Rbx1 to mitochondria to promote ubiquitination in Drp1KO hepatocytes. p62 also connects mitochondria to autophagosomes through interactions with ubiquitin and LC3. Statistical analysis was performed using Student’s t-test: *p<0.05, **p< 0.01. See also Figure S4.

To test the role of Rbx1 in mitochondrial ubiquitination, we knocked down Rbx1 using shRNA approaches. We identified two effective shRNAs against Rbx1 using Western blotting of mouse embryonic fibroblasts expressing either scramble or Rbx1-targeted shRNAs (Fig. 6B). Plasmids expressing these shRNAs, along with GFP, were delivered into the livers of Alb-Drp1KO and Alb-Drp1ParkinKO mice by hydrodynamic tail vein injection. Four days after injection, we assessed the impact of Rbx1 knockdown on ubiquitination using confocal microscopy of liver sections with antibodies to ubiquitin and GFP. The results showed that Rbx1 knockdown significantly decreased mitochondrial ubiquitination in both Drp1KO hepatocytes (Fig. 6C) and Drp1ParkinKO hepatocytes (Fig. 6D). These data suggest that Rbx1 is involved in mitochondrial ubiquitination in hepatocytes (Fig. 6E).

The loss or downregulation of p62 prevents the liver damage caused by Drp1 deficiency

We found that cell size was smaller in Drp1p62KO hepatocytes than in Drp1KO (Fig. S4A and B). These data suggest that the additional loss of p62 may rescue liver damage caused by Drp1 deficiency. To directly test this idea, we measured ALT levels in blood. Indeed, ALT levels in Alb-Drp1p62KO mice were significantly decreased and similar to those in the control and p62KO mice (Fig. S4C). These findings suggest that liver damage in Alb-Drp1KO mice may result from the accumulation of p62.

It has been shown that increased levels of p62 sequester Keap1 from its function in the degradation of the stress-responsive transcription factor Nrf2, leading to the toxic overexpression of Nrf2 target genes in macroautophagy-deficient Atg7KO livers (Komatsu et al., 2007). To test whether the same mechanism leads to liver damage in Alb-Drp1KO mice, we performed qRT-PCR analysis of eight Nrf2 target genes. In contrast to Atg7KO livers (Komatsu et al., 2010), Drp1KO livers did not significantly increase the expression of the Nrf2 target genes (Fig. S4D). Since Nrf2 also regulates p62 (Jain et al., 2010), we knocked out Nrf2 in Alb-Drp1KO mice. The additional Nrf2KO restored p62 expression to normal levels and decreased blood levels of ALT in Alb-Drp1Nrf2KO mice (Alb-Cre^+/−::Drp1^flox/flox::Nrf2^−/−) (Fig. S4E and F). It appears that removing p62 by p62KO or decreasing p62 levels by Nrf2KO is sufficient to rescue liver damage in Alb-Drp1KO mice. p62-promoted mitochondrial ubiquitination may stimulate the degradation of outer membrane proteins and further aggravates damaged mitochondria, leading to liver damage.

Loss of Opa1 clears mitophagy intermediates and rescues liver damage in a NAFLD mouse model

Megamitochondria has been observed in the hepatocytes of patients with nonalcoholic fatty liver disease (NAFLD), the most common liver disease in Western countries (Neuman et al., 2014; Targher et al., 2018; Wakabayashi, 2002; Younossi et al., 2018). To test the link of our findings with megamitochondria in NAFLD, we examined mitochondria in a well-established diet-induced NAFLD mouse model (Anstee et al., 2010; Machado et al., 2015). Control and Alb-Opa1KO mice were fed a methionine- and choline-deficient diet for six weeks. Liver sections from these mice were analyzed by confocal microscopy with antibodies to PDH, ubiquitin, and p62. We found that the enlarged mitochondria (Fig. 7A and B) in NAFLD hepatocytes accumulate p62, ubiquitin and Keap1 (Fig. 7A, C, D and E), similar to mitochondria in Drp1KO hepatocytes. Remarkably, Opa1KO greatly decreased mitochondrial size (Fig. 7A and B) and cleared the accumulation of mitophagy intermediates (Fig. 7A, C, D and E) in the NAFLD model, similar to Drp1Opa1KO hepatocytes. Furthermore, Opa1KO also rescued liver damage in this liver disease model (Fig. 7F). These data suggest that mitophagy is decreased in NAFLD and that restoring mitochondrial size may serve as a potential treatment for NAFLD.

Figure 7. Megamitochondria are formed in a NAFLD mouse model.

(A) Control Opa1^flox/flox mice and Alb-Opa1KO mice were fed a methionine- and choline-deficient diet (MCD diet) for six weeks. Liver sections from these mice were analyzed by confocal microscopy with antibodies to PDH, ubiquitin, and p62. (B) Individual mitochondrial sizes were analyzed (n=602 mitochondria from 3 control mice, 639 from 3 MCD-diet-fed control mice, and 692 from 3 MCD-diet-fed Alb-Opa1KO mice). Red lines indicate averages. (C and D) Cells that show the accumulation of p62 (C) and ubiquitin (D) are quantified. Values represent the average ± SEM (n=3 mice). (E) Liver sections from the same set of mice in (A) were analyzed by confocal microscopy with antibodies to PDH, Keap1, and p62. 86.5% of p62-positive mitochondria were also positive for Keap1 (n=37 mitochondria). (F) Blood levels of ALT were measured before and during administration at 2, 4, and 6 weeks. ALT activity levels for each mouse were normalized to that obtained before MCD diet administration. Values are average ± SEM (n=3–5 mice). Statistical analysis was performed using Student’s t-test: *p< 0.05, **p< 0.01, ***p<0.001.

Discussion

It has been shown in the budding yeast S. cerevisiae that simultaneous loss of Dnm1 (a yeast homolog of Drp1) and one of three proteins essential for mitochondrial fusion, Mgm1 (a yeast homolog of Opa1), Fzo1 (a yeast homolog of mitofusin) and Ugo1 (a yeast homolog of SLC25A46), nearly reestablishes the wild-type shape of mitochondria (Sesaki and Jensen, 1999, 2001; Sesaki et al., 2003). In yeast double mutants, cells maintain mitochondrial respiration that is lost in the absence of mitochondrial fusion (Sesaki and Jensen, 1999, 2001; Sesaki et al., 2003). However, it remains unclear whether this is true in multicellular organisms such as mammals considering their much more complex cellular structure and expanded list of mitochondrial function. To address this fundamental question in cell biology, we have introduced mitochondrial stasis by blocking mitochondrial division and fusion in vivo. Using this animal system, we demonstrated that mitochondrial size is antagonistically regulated by division and fusion. Furthermore, our data show that, instead of the lack of mitochondrial division or fusion, an extreme mitochondrial size, whether large or small, is harmful to hepatocytes in the liver. Our current findings, together with previous data, provide a general concept that mitochondrial size plays a critical role in organelle function and that generating a balanced mitochondrial size distribution can obviate the necessity for mitochondrial division and fusion.

We further discuss that the function of mitochondrial division and fusion is not only limited to the regulation of mitochondrial size. As we have shown here, expression levels of Drp1 and Opa1 in the brain and heart are higher than those in the liver. In the brain, even partial decreases in both mitochondrial division and fusion—accomplished by decreasing the level of one of several Drp1 receptors proteins, mitochondrial fission factor, using gene trap and deleting one of two mitofusin 1 and 2—failed to maintain the viability of neurons (Chen et al., 2015). Interestingly, partial decreases in both mitochondrial division and fusion can support the function of the heart (Chen et al., 2015). However, the complete loss of mitochondrial division and fusion does not fully support heart function (Song et al., 2017). Therefore, the role that mitochondrial dynamics plays in the integrity of cells, tissues and organs likely depends on their architecture, environment and energy metabolism. Determining the exact role of mitochondrial division and fusion in each cell and organ type will provide critical information to reveal the functional landscape of mitochondrial dynamics in health and diseases.

Our findings provide a new mechanistic framework for parkin-independent mitophagy. We discovered a novel function of p62 in promoting mitochondrial ubiquitination in parkin-independent mitophagy (Fig. 6E). This finding was quite surprising because p62 has been shown to function downstream of ubiquitination and connect ubiquitinated proteins to autophagosomes by binding to ubiquitin and the autophagosomal protein LC3 in mitophagy as well as other types of autophagy (Bitto et al., 2014; Katsuragi et al., 2015). This new function of p62 is likely mediated by recruiting Rbx1—an ubiquitin E3 ligase that has not been implicated in mitophagy (Kensler et al., 2007; Zhang et al., 2004)—to mitochondria. In parallel to Rbx1, p62 also recruits Keap1—a p62-binding protein that can form a cullin-RING ubiquitin E3 ligase complex with Rbx1 (Bulatov and Ciulli, 2015; Genschik et al., 2013; Uruno et al., 2015; Zhang et al., 2004)—to mitochondria. We propose that that p62 brings this ligase complex to mitochondria through interactions with Keap1.

The size and shape of mitochondria change under different physiological and pathological conditions. One extreme example is the formation of megamitochondria in liver diseases, including NAFLD and alcoholic liver injury (Caldwell et al., 1999; Kleiner and Makhlouf, 2016; Lotowska et al., 2014; Neuman et al., 2014; Noureddin et al., 2013; Wakabayashi, 2002). These dramatic changes in mitochondrial morphology have been reported for many years; however, what these megamitochondria really are and how they contribute to liver damage are largely unknown. Our data suggest that megamitochondria are halted mitophagy intermediates. The formation of megamitochondria likely involves unbalanced mitochondrial division and fusion since Opa1KO can oppose megamitochondrion formation in the NAFLD model. NAFLD is the most common liver disease in Western countries—approximately 30% of adults in the US have NAFLD. As NAFLD progresses, it leads to more severe and potentially fatal conditions, such as liver failure and cancer. Therefore, the need to understand the disease’s mechanism and develop a treatment strategy is urgent. We hypothesize that mitophagy plays a critical role for both the pathogenesis and treatment of NAFLD. Mitophagy may be blocked in NAFLD due to increased mitochondrial size, and restoring mitophagy may enable us to maintain healthy livers in NAFLD patients.

Previous studies have shown that, in parkin-mediated mitophagy, the E3 ligase parkin ubiquitinates mitofusin 1 and 2 upon the loss of the mitochondrial membrane potential induced by uncouplers such as carbonilcyanide p-triflouromethoxyphenylhydrazone, which recruits parkin to mitochondria (Gegg et al., 2010). Since the loss of Drp1 does not result in the loss of the membrane potential (Kageyama et al., 2014; Roy et al., 2016; Wakabayashi et al., 2009), we suggest that the degradation of mitofusin 1 and 2 in Drp1KO cells (Fig. S3G) is controlled by a mechanism other than mitophagy. For example, parkin controls the formation of mitochondria-derived vesicles that are delivered to lysosomes as a mitochondrial quality control pathway in response to oxidative stress (Sugiura et al., 2014). Since mitochondrial division deficiency increases oxidative stress in mitochondria (Kageyama et al., 2014; Kageyama et al., 2012), parkin may promote the degradation of mitofusin 1 and 2 as a part of this stress-induced pathway. Furthermore, we recently found that the membrane potential is transiently decreased in connected mitochondria in Drp1KO cells (Roy et al., 2016). This transient loss of the membrane potential may transiently recruit parkin to mitochondria and induce ubiquitination of mitofusin 1 and 2. We expect that our identification of parkin as a critical E3 ligase in the degradation of mitofusin 1 and 2 in response to increased mitochondrial size will help us further define mechanisms that sense and rebalance mitochondrial division and fusion in future studies. Such a readjustment of mitochondrial dynamics would be important for understanding the pathogenesis of Parkinson’s disease since parkin is mutated in a familial form of this devastating neurological disorder.

Limitations of Study

This study uncovers a new mitophagy pathway mediated by p62, Keap1, and Rbx1. Inhibiting interactions of p62 with Keap1 nearly prevents p62-mediated mitochondrial ubiquitination. However, ubiquitination is only partially quenched by Rbx1 knockdown. The apparent modest effect can be attributed to residual Rbx1 present in Rbx1 knockdown hepatocytes or another ubiquitin E3 ligase that works with p62 and Keap1. This study does not distinguish between the two. Future studies incorporating knockout approaches can address this important question. Another unanswered key question is the impact of Drp1KO on Nrf2. Although most Nrf2 target genes are not upregulated in Drp1KO hepatocytes in which Keap1 is recruited to mitochondria, p62 levels are increased in an Nrf2-dependent manner. There may be a mechanism by which Nrf2 regulates p62 differently than other target genes when Keap1 is associated with mitochondria. Using a mouse model for NAFLD, this study also suggests that this new mitophagy pathway is decreased as a result of megamitochondria formation in NAFLD. It is not known how these giant mitochondria are formed. Furthermore, translating these discoveries to human NAFLD is an overarching future challenge.

STAR Methods

Contact for reagent and resource sharing

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Hiromi Sesaki (hsesaki@jhmi.edu).

Experimental model and subject details

All of the work with animals was conducted according to the guidelines established by the Johns Hopkins University Committee on Animal Care and Use. Drp1^flox/flox (Wakabayashi et al., 2009), Parkin^−/− (Kageyama et al., 2014), p62^−/− (Komatsu et al., 2007) and Opa1 ^flox/flox (Zhang et al., 2011) mice have been described previously. Alb-Cre mice (Postic et al., 1999) and Nrf2^−/− mice (Chan et al., 1996) were obtained from the Jackson Laboratory. By breeding, we generated control (Drp1^flox/flox, and Alb-Cre^+/−), Alb-Drp1KO (Alb-Cre^+/−::Drp1^flox/flox), ParkinKO (Drp1^flox/flox::Parkin^−/−), Alb-Drp1ParkinKO (Alb-Cre^+/−::Drp1^flox/flox::Parkin^−/−), Alb-Opa1KO (Alb-Cre^+/−::Opa1^flox/flox), Alb-Drp1Opa1KO (Alb-Cre^+/−::Drp1^flox/flox::Opa1^flox/flox), p62KO (Drp1^flox/flox::p62^−/−), Alb-Drp1p62KO (Alb-Cre^+/−::Drp1^flox/flox::p62^−/−), Nrf2KO (Drp1^flox/flox::Nrf2^−/−) and Alb-Drp1Nrf2KO (Alb-Cre^+/−::Drp1^flox/flox::Nrf2^−/−) mice.

To induce NAFLD, control Opa1^flox/flox mice and Alb-Opa1KO mice were fed a methionine- and choline-deficient diet (A02082002BR, Research Diets) for 6 weeks (Anstee et al., 2010; Machado et al., 2015). Blood samples were collected before and during administration at 2, 4, and 6 weeks. Samples were centrifuged at 10,000 × g for 1 min at 4°C and the supernatants were used to determine ALT activity levels. Livers were examined using immunofluorescence microscopy at 6 weeks.

Method Details

Western blotting

Mouse tissues were harvested, flash-frozen in liquid nitrogen, and homogenized in RIPA buffer (Cell Signaling Technology, MA, USA) containing complete mini protease inhibitor (Roche, Basel, Switzerland). Lysates were centrifuged at 14,000 × g for 10 min, and the supernatants were collected. Protein concentrations were determined by the Bradford method using a Bio-Rad protein assay (CA, USA). Proteins were separated by SDS–PAGE and transferred onto Immobilon-FL (Millipore, Darmstadt, Germany). After blocking in 3% BSA/PBS/Tween-20 for 1 h at room temperature, the blots were incubated with primary antibodies. Immunocomplexes were visualized by appropriate secondary antibodies conjugated with fluorophores using a PharosFX Plus Molecular Imager (Bio-Rad). Band intensity was determined using NIH ImageJ software.

Antibodies

The following primary antibodies were used: Drp1 (611113; BD Biosciences), Opa1 (612607; BD Biosciences), PDH (ab110333; Abcam), GAPDH (MA5-15738; Thermo), Tim44 (612582; BD Transduction Laboratories), Tim23 (611223; BD Transduction Laboratories), Tom20 (sc-11415; Santa Cruz Biotechnology), mitofusin 1 (ab57602; Abcam), mitofusin 2 (ab56889; Abcam), HA (600-401-384; Rockland), VDAC (4866; Cell Signaling Technology), HSP60 (12165; Cell Signaling Technology), E-cadherin (14472; Cell Signaling Technology), p62 (GP-62C; Progen), ubiquitin (z0458; DAKO), phospho-p62 (serine 351) (PM074; MBL International), LC3 (PM036; MBL) and Keap1 (10503-2-AP; Proteintech). The following secondary antibodies were purchased from Invitrogen: Alexa 488 anti-Rabbit IgG (A21206), Alexa 488 anti-Mouse IgG (A21202), Alexa 488 anti-guinia pig IgG (A11073), Alexa 568 anti-mouse IgG (A10037), Alexa 647 anti-mouse IgG (A31571) and Alexa 647 anti-guinia pig IgG (A21450).

ALT activity

ALT activity in the blood was determined by a coupled enzyme assay, which resulted in a colorimetric (570 nm) product proportional to the pyruvate generated. We used a kinetic colorimetric assay kit from Sigma-Aldrich (MO, USA) according to the manufacturer’s protocol.

Histology and immunofluorescence

Mice were anesthetized by intraperitoneal injection of Avertin and fixed by cardiac perfusion of ice-cold 4% paraformaldehyde in PBS as previously described (Kageyama et al., 2014; Yamada et al., 2016). The livers were dissected and further fixed in 4% paraformaldehyde in PBS for 2 hs at 4°C. For the PAS stain, the samples were dehydrated and embedded in paraffin. Paraffin sections were cut and PAS stained at Johns Hopkins School of Medicine Pathology Core. The samples were observed using a microscope (model BX51; Olympus) equipped with a DP-70 color camera and 10× (0.3 NA) UIS2 objectives. For immunofluorescence microscopy, the samples were further incubated in PBS containing 30% sucrose overnight and frozen in OCT compound (Sakura Fintek). Immunofluorescence microscopy of liver sections was performed as previously described (Kageyama et al., 2014) with some modifications (Yamada et al., 2016). The frozen sections were cut, washed in PBS and blocked in 10% donkey or sheep serum. The sections were then incubated with primary antibodies followed by fluorescently labeled secondary antibodies.

Confocal microscopy and image analysis

Samples were viewed using a Zeiss LSM510-Meta, LSM780-FCS and LSM800 GaAsP laser scanning confocal microscopes equipped with a 63× objective, as described previously (Kageyama et al., 2014; Yamada et al., 2016). Image analysis was performed using NIH ImageJ software.

Plasmids

To generate the Su9-mCherry-GFP plasmid, mCherry was PCR-amplified from pmCherry-N1 (Clontech) and cloned into the BamHI site of pSu9-GFP (Eura et al., 2003) using Gibson assembly (New England Biolabs). To fuse p62 to the transmembrane domain of Mff (MffTM: amino acids 268–291), p62 and MffTM were PCR-amplified from plasmids carrying human p62 and mouse Mff, respectively. p62 and MffTM were fused through a GGSG linker by PCR and cloned into a pcDNA3.1/V5-His plasmid. A Keap1-interaction-defective mutation (T350A) of human p62 (Ichimura et al., 2013) was introduced to p62-MffTM by PCR using the following DNA oligos. 5′-GTGAACTCCAGTCCCTACAGATGCCAGAATCC-3′ and 5′-GGGACTGGAGTTCACCTGCAGACGGGTCCACTTC-3′. Plasmids carrying HA-Rbx1 were obtained from Addgene (#19897) (Ohta et al., 1999). To create pSUPER-GFP, GFP was subcloned into the BamHI/HindIII sites of pRK5 (556104, BD PharMingen) from pEGFP-C3 (6082-1, Clontech) and the CMV promoter-GFP- SV40 poly A cassette was PCR amplified from pRK5-GFP and subcloned into the XbaI/NotI sites of pSUPER (VEC-PBS-0002, Oligoengine). To create shRNA plasmids, the following target sequence were cloned into pLKO1 for lentivirus production and pSUPER-GFP for tail vein injection. Rbx1#1: TGCATCTCTCGATGGCTCAAAttcaagagaTTTGAGCCATCGAGAGATGCA. Rbx1#2: CCACATTATGGATCTTTGTATttcaagagaATACAAAGATCCATAATGTGG. Scramble: CCTAAGGTTAAGTCGCCCTCGttcaagagaCGAGGGCGACTTAACCTTAGG (Sarbassov et al., 2005).

Hydrodynamic tail vein injection of HA-Rbx1 and shRNAs

50 μg of endotoxin-free plasmids in 2 ml of 0.9% NaCl were injected into mice via the tail vein. Mice were fixed by cardiac perfusion of ice-cold 4% parafomaldehyde in PBS at 8 hours and 4 days after injection of HA-Rbx1 and shRNAs, respectively.

Electron microscopy

Mice were anesthetized by intraperitoneal injection of Avertin and fixed by cardiac perfusion of 2% glutaraldehyde, 3 mM CaCl₂ and 0.1 M cacodylate buffer, pH 7.4, as previously described (Wakabayashi et al., 2009). Livers were dissected, cut into small pieces, and further fixed for 1 h. After washes, samples were post-fixed in 2.7% OsO4 and 167 mM cacodylate, pH 7.4, for 1 h on ice. After washes in water, samples were incubated in 2% uranyl acetate for 30 min. After dehydration using 50, 70, 90, and 100% ethanol and 100% propylene oxide, samples were embedded in EPON resin. Ultrathin sections were obtained using a Reichert-Jung ultracut E, stained with 2% uranyl acetate and lead citrate, and viewed using a transmission electron microscope (H-7600; Hitachi) equipped with a dual CCD camera (Advanced Microscopy Techniques). ImageJ software was used to measure the circularity and aspect ratios of the mitochondria and the number of cristae.

Glucose and lipid measurements

Non-fasting and fasting blood glucose concentrations were determined using a OneTouch UltraMini Meter (Lifescan), as previously reported (Zhang et al., 2011). To examine insulin sensitivity, insulin tolerance tests were performed as previously described (Zhang et al., 2011). After fasting for 5 hs, mice were subjected to intraperitoneal insulin injections (0.6 U/kg body weight). Blood glucose concentrations were determined at different time points. To analyze blood lipid levels, triglyceride and cholesterol levels were measured using triglyceride reagent (TR22421, Thermo Scientific) and total cholesterol reagent (TR13421, Thermo Scientific) per the manufacturer’s instructions.

In vivo mitophagy assay

50 μg of Su9-mCherry-GFP plasmid were injected into control, Alb-Drp1KO, and Alb-Drp1Opa1KO mice via their tail veins. Four days after injection, mice were anesthetized by intraperitoneal injection of Avertin and fixed by cardiac perfusion of ice-cold 4% paraformaldehyde. Livers were dissected and immersed in 4% paraformaldehyde in PBS for 2 hs. Fixed livers were incubated in 30% sucrose in PBS overnight and frozen in OCT compound. Frozen sections were cut into 7-μm-thick slices and these were washed in PBS 3 times for 5 min per wash. Sections were mounted on slide glasses with antifade medium (Vectashield). The mitophagy index was determined by measuring the relative area of mCherry fluorescence that did not overlap with GFP fluorescence over the total area of mCherry fluorescence in each cell using NIH ImageJ software and set 1 in the control mice.

Lentiviruses

Mouse embryonic fibroblasts and HEK293T cells were cultured as described (Adachi et al., 2016). Lentiviruses were generated as described previously (Kageyama et al., 2014; Stewart et al., 2003). Briefly, pLKO1 carrying shRNAs was cotransfected into HEK293T cells with two other constructs, pHR-CMV8.2ΔR and pCMV-VSVG, using Lipofectamine 2000 (Invitrogen). Two days after transfection, the supernatant of transduced cells containing released viruses was collected. The viruses were snap-frozen in liquid nitrogen and stored at −80°C.

qRT-PCR analysis

mRNAs were prepared from snap-frozen livers using a RNease mini kit (Qiagen) and analyzed using PowerUp SYBR Green Master Mix (Applied Biosystems). Levels of albumin mRNA were used as an internal standard. The following DNA oligos were used (Chanas et al., 2002; Wakabayashi et al., 2010). Nqo1: 5′-GGCATCCTGCGTTTCTGTG-3′ and 5′-GGTTTCCAGACGTTTCTTCCAT-3′. Gsta1: 5′-CGCAGACCAGAGCCATTCTC-3′ and 5′-TTGCCCAATCATTTCAGTCAGAT. Gsta2: 5′-CCCCTTTCCCTCTGCTGAAG-3′ and 5′-TGCAGCCACACTAAAACTTGAAAA-3′. Gsta3: 5′-CCTTGCCAAGATCAAGGAAC-3′ and 5′-ACCTGAATTGACACAGACG-3′. Gstm1: 5′-TTGAGAAGACCACAGCACCAG-3′ and 5′-TCCACATAGGTGACCTTGTCC-3′. Gclr: 5′-ACCTGGCCTCCTGCTGTGTG-3′ and 5′-GGTCGGTGAGCTGTGGGTGT-3′. Gclc: 5′-GCACGGCATCCTCCAGTTCCT-3′ and 5′-TCGGATGGTTGGGGTTTGTCC-3′. Prdx1 (Msp23): 5′-CTGCCAAGTGATTGGCGCTTCTGTGGATTCTC-3′ and 5′-TCCAGCCAGCTGGACACACTTCACC-3′. Albumin: 5′-TGCTTTTTCCAGGGGTGTGTT and 5′-TTACTTCCTGCACTAATTTGGCA-3′.

Quantification and statistical analysis

Data are expressed as means with SEM as indicated in the figure legends of at least three independent experiments. Statistical analysis was performed using the Student’s t-test.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Mouse monoclonal anti-Drp1	BD Biosciences	Cat# 611113; RRID: AB_398424
Mouse monoclonal anti-Opa1	BD Biosciences	Cat# 612607; RRID: AB_399889
Mouse monoclonal anti-PDH	abcam	Cat# ab110333; RRID: AB_10862029
Mouse monoclonal anti-GAPDH	Thermo Fisher Scientific	Cat# MA5-15738; RRID: AB_10977387
Rabbit polyclonal anti-Tom20	Santa Cruz Biotechnology	Cat# sc-11415; RRID: AB_2207533
Mouse monoclonal anti-Tim44	BD Biosciences	Cat# 612582; RRID: AB_399869
Mouse monoclonal anti-Tim23	BD Biosciences	Cat# 611223; RRID: AB_398755
Rabbit polyclonal anti-VDAC	Cell Signaling Technology	Cat# 4866; RRID: AB_2272627
Rabbit monoclonal anti-HSP60	Cell Signaling Technology	Cat# 12165; RRID: AB_2636980
Mouse monoclonal anti-E-cadherin	Cell Signaling Technology	Cat# 14472; RRID: AB_2728770
Guinea pig polyclonal anti-p62	Progen	Cat# GP62-C; RRID: AB_2687531
Rabbit polyclonal anti-ubiquitin	DAKO	Cat# Z0458; RRID: AB_2315524
Rabbit polyclonal anti-phospho-p62 (Ser351)	MBL International	Cat# PM074
Rabbit polyclonal anti-LC3	MBL International	Cat# PM036; RRID: AB_2274121
Rabbit polyclonal anti-Keap1	Proteintech	Cat# 10503-2-AP; RRID: AB_2132625
Mouse monoclonal anti-Mitofusin 1	abcam	Cat# ab57602; RRID: AB_2142624
Mouse monoclonal anti-Mitofusin 2	abcam	Cat# ab56889; RRID: AB_2142629
Mouse monoclonal anti-Rbx1	Santa Cruz Biotechnology	Cat# sc-393640; RRID: AB_2722527
Alexa 488 anti-rabbit IgG	Thermo Fisher Scientific	Cat# A-21206; RRID: AB_2535792
Alexa 488 anti-mouse IgG	Thermo Fisher Scientific	Cat# A-21202; RRID: AB_141607
Alexa 488 anti-guinea pig IgG	Thermo Fisher Scientific	Cat# A-11073; RRID: AB_2534117
Alexa 568 anti-mouse IgG	Thermo Fisher Scientific	Cat# A10037; RRID: AB_2534013
Alexa 647 anti-rabbit IgG	Thermo Fisher Scientific	Cat# A-31573; RRID: AB_2536183
Alexa 647 anti-mouse IgG	Thermo Fisher Scientific	Cat# A-31571; RRID: AB_162542
Alexa 647 anti-guinia pig IgG	Thermo Fisher Scientific	Cat# A-21450; RRID: AB_141882
Chemicals, Peptides, and Recombinant Proteins
Insulin	Gibco	Cat# 12585014
2-Methyl-2-butanol	Sigma-Aldrich	Cat# 240486
2,2,2-Tribromoethanol	Sigma-Aldrich	Cat# T48402
Paraformaldehyde	Sigma-Aldrich	Cat# P6148
10XPBS	QUALITY BIOLOGICAL	Cat# 119-069-131
cOmplete™, Mini, EDTA-free Protease Inhibitor Cocktail	Roche	Cat# 11836170001
Bio-Rad Protein Assay Dye Reagent Concentrate	Bio-Rad	Cat# 5000006
Immobilon-FL PVDF Membrane	Millipore	Cat# IPFL00010
IMDM	Gibco	Cat# 12440053
Dulbecco’s Modified Eagle’s Medium - high glucose	Sigma-Aldrich	Cat# D5796
Dulbecco’s Phosphate Buffered Saline	Sigma-Aldrich	Cat# D8537
Fetal Bovine Serum	Sigma-Aldrich	Cat# F6178
Lipofectamine 2000 Transfection Reagent	Invitrogen	Cat# 11668019
Tissue-Tek® O.C.T. Compound	Sakura Finetech	Cat# 4583
RIPA Buffer (10X)	Cell Signaling Technology	Cat# 9806
Methionine Choline deficient diet	Research Diets	Cat# A02082002BR
Critical Commercial Assays
ALT Activity Assay	Sigma-Aldrich	Cat# MAK052
Triglyceride Reagent	Thermo Scientific	Cat# TR22421
Cholesterol Reagent	Thermo Scientific	Cat# TR13421
RNeasy® Mini Kit	Qiagen	Cat# 74104
Quick Ligation™ Kit	New England Biolabs	Cat# M2200L
Gibson Assembly® Master Mix	New England Biolabs	Cat# M5510AA
PowerUp™ SYBR™ Green Master Mix	Applied Biosystem	Cat# A25742
ReadyScript® cDNA Synthesis Mix	Sigma-Aldrich	Cat# RDRT
PureYield™ Plasmid Miniprep System	Primega	Cat# A1222
Experimental Models: Cell Lines
Mouse: embryonic fibroblast	Wakabayashi et al., J. Cell Biol. 2009	N/A
Human: HEK293T cells	ATCC	CRL-3216
Experimental Models: Organisms/Strains
Mouse: Drp1flox/flox	Wakabayashi et al., 2009	N/A
Mouse: Parkin−/−	Kageyama et al., 2014	N/A
Mouse: p62−/−	Komatsu et al., 2007	N/A
Mouse: Opa1 flox/flox	Zhang et al., 2011	N/A
Mouse: Alb-Cre: B6.FVB(129)-Tg(Alb1-cre)1Dlr/J	The Jackson Laboratory	# 016832
Mouse: Nrf2−/−: B6.129X1-Nfe2l2tm1Ywk/J	The Jackson Laboratory	# 017009
Mouse: Alb-Cre+/−::Drp1flox/flox	This paper	N/A
Mouse: Drp1flox/flox::Parkin−/−	This paper	N/A
Mouse: Alb-Cre+/−::Drp1flox/flox::Parkin−/−	This paper	N/A
Mouse: Alb-Cre+/−::Opa1flox/flox	This paper	N/A
Mouse: Alb-Cre+/−::Drp1flox/flox::Opa1flox/flox	This paper	N/A
Mouse: Drp1flox/flox::p62−/−	This paper	N/A
Mouse: Alb-Cre+/−::Drp1flox/flox::p62−/−	This paper	N/A
Mouse: Drp1flox/flox::Nrf2−/−	This paper	N/A
Mouse: Alb-Cre+/−::Drp1flox/flox::Nrf2−/−	This paper	N/A
Oligonucleotides
qRT-PCR primers for Nqo1	Wakabayashi et al., Sci. Signal. 2010	PrimerBank ID 6679080a3
qRT-PCR primers for Gsta1	Wakabayashi et al., Sci. Signal. 2010	Primer Bank ID 7110611a2
qRT-PCR primers for Gsta2	Chanas et al., Biochem. J. 2002	N/A
qRT-PCR primers for Gsta3, Gstm1, Gclr, Gclc and Prdx1	This paper	N/A
qRT-PCR primers for Albumin	Wakabayashi et al., Sci. Signal. 2010	Primer Bank ID 33859506a1
shRNA targeting sequence: Rbx1 #1: TGCATCTCTCGATGGCTCAAA	Sigma-Aldrich	TRC clone ID: TRCN0000273778
shRNA targeting sequence: Rbx1 #2: CCACATTATGGATCTTTGTAT	Sigma-Aldrich	TRC clone ID: TRCN0000273726
Scrambled shRNA targeting sequence: CCTAAGGTTAAGTCGCCCTCG	Sarbassov et al., Science 2005	Addgene plasmid # 1864
Recombinant DNA
pSu9-GFP	Eura et al., 2003	N/A
Su9-mCherry-GFP	This paper	N/A
pcDNA3.1/V5-His TOPO	Invitrogen	K480001
p62-MffTM	This paper	N/A
KID p62-MffTM	This paper	N/A
pcDNA3-HA2-ROC1 (HA-Rbx1)	Ohata et al., 1999	Addgene Plasmid #19897
pSUPER-GFP	This paper	N/A
pSUPER-GFP-Rbx1 #1	This paper	N/A
pSUPER-GFP-Rbx1 #2	This paper	N/A
pLKO.1	Open Biosystems	# RHS4078
pHR-CMV8.2ΔR	Stewart et al., RNA 2003	Addgene plasmid # 8455
pCMV-VSVG	Stewart et al., RNA 2003	Addgene plasmid #8454
Software and Algorithms
Fiji	Fiji	https://fiji.sc/

Highlights.

• Balanced mitochondrial dynamics are important for liver health

• p62-Keap1-Rbx1 complex ubiquitinates mitochondria in parkin-independent mitophagy

• Fatty livers generate megamitochondria and block parkin-independent mitophagy

• Suppressing megamitochondria formation by Opa1KO rescues damage in fatty livers

Abbreviations

KO

knockout

Drp1

dynamin-related protein 1

Opa1

optic atrophy 1

PDH

pyruvate dehydrogenase

NAFLD

nonalcoholic fatty liver disease