sign in sign up
<<
Home , Peroxisome

Metabolism Article

Hif-2a Promotes Degradation of Mammalian Peroxisomes by Selective

Katharina M. Walter,1,2,10 Miriam J. Scho¨ nenberger,1,2,10 Martin Tro¨ tzmu¨ ller,3 Michael Horn,1 Hans-Peter Elsa¨ sser,4 Ann B. Moser,5 Miriam S. Lucas,6 Tobias Schwarz,6 Philipp A. Gerber,7 Phyllis L. Faust,8 Holger Moch,9 Harald C. Ko¨ feler,3 Wilhelm Krek,1,2,* and Werner J. Kovacs1,2,* 1Institute of Molecular Health Sciences 2Competence Center for Systems Physiology and Metabolic Diseases ETH Zurich, CH-8093 Zurich, Switzerland 3Core Facility for Mass Spectrometry, Center for Medical Research, Medical University of Graz, A-8010 Graz, Austria 4Department of Cytobiology, Philipps-University Marburg, D-35037 Marburg, Germany 5Kennedy Krieger Institute, Baltimore, MD 21205, USA 6ScopeM – Scientific Center for Optical and Electron Microscopy, ETH Zurich, CH-8093 Zurich, Switzerland 7Division of Endocrinology and Diabetes, University Hospital Zurich, CH-8091 Zurich, Switzerland 8Department of Pathology and , College of Physicians and Surgeons, Columbia University, New York, NY 10032, USA 9Institute of Surgical Pathology, University Hospital Zurich, CH-8091 Zurich, Switzerland 10Co-first Authors *Correspondence: [email protected] (W.K.), [email protected] (W.J.K.) http://dx.doi.org/10.1016/j.cmet.2014.09.017

SUMMARY expressed HIF-1b subunit and O2-regulated a subunits (HIF-1a and HIF-2a)(Keith et al., 2012). Under normoxia, HIF-a subunits Peroxisomes play a central role in metabolism, are hydroxylated and targeted for proteasomal degradation by and their function depends on molecular oxygen. an ligase complex containing the von Hippel-Lindau Low oxygen tension or von Hippel-Lindau (Vhl) tumor (VHL) tumor suppressor protein pVHL. Under hypoxia, HIF-a suppressor loss is known to stabilize hypoxia-induc- hydroxylation is inhibited, thereby stabilizing HIF-a protein. a b a ible factors alpha (Hif-1a and Hif-2a) to mediate HIF- dimerizes with HIF-1 to activate target genes. HIF-1 a adaptive responses, but it remains unknown if perox- and HIF-2 have both common and distinct target genes and are differentially regulated in various physiological and patho- isome homeostasis and metabolism are intercon- a physiological conditions, such as cancer and ischemic diseases nected with Hif- signaling. By studying liver-specific (Semenza, 2012; Keith et al., 2012). HIF-induced genes promote Vhl Vhl/Hif1a Vhl/Hif2a , , and knockout mice, we angiogenesis, erythropoiesis, metastasis, and metabolic reprog- demonstrate a regulatory function of Hif-2a signaling ramming, such as shifting metabolism from oxidative phos- on peroxisomes. Hif-2a activation augments peroxi- phorylation to . Activation of HIFs also reprograms lipid some turnover by selective autophagy (pexophagy) metabolism leading to lipid accumulation (Kucejova et al., 2011; and thereby changes lipid composition reminiscent Qu et al., 2011; Kim et al., 2006; Rankin et al., 2009; Krishnan of peroxisomal disorders. The autophagy receptor et al., 2009), but it remains controversial how they promote lipid Nbr1 localizes to peroxisomes and is likewise deg- accumulation. a raded by Hif-2a-mediated pexophagy. Furthermore, Abundant evidence suggests key roles for HIF-1 in the regu- we demonstrate that peroxisome abundance is lation of mitochondrial metabolism (Semenza, 2011). However, so far there is no evidence linking HIF signaling to peroxisomes, reduced in VHL-deficient human clear cell renal cell whose function depends highly on molecular O . In fact, perox- carcinomas with high HIF-2a levels. These results 2 isomes may be responsible for as much as 20% of O2 consump- establish Hif-2a as a negative regulator of peroxi- tion and 35% of H2O2 production in tissues such as the liver some abundance and metabolism and suggest a (Fransen et al., 2012). Peroxisomes are extremely versatile and mechanism by which cells attune peroxisomal func- dynamic whose number, size, and function are tion with oxygen availability. dependent on cell type and metabolic needs. They play key roles in the degradation of fatty acids (i.e., very long-chain fatty acids [VLCFAs] and polyunsaturated fatty acids [PUFAs]), INTRODUCTION synthesis, and synthesis, and metabolism of (Van Veldhoven, 2010; Faust and Ko-

Oxygen (O2) is an essential substrate in cellular metabolism and vacs, 2014; Fransen et al., 2012). The importance of peroxisomal signaling and, as such, is inseparably linked to survival and metabolism is illustrated by peroxisome biogenesis disorders in normal function of all metazoan. Central to the molecular mech- which functional peroxisomes are absent or disorders caused by anisms underlying O2 homeostasis are hypoxia-inducible factors single peroxisomal deficiencies (Raymond et al., 2009). (HIFs) that function as master regulators of the adaptive response Lack of peroxisomal metabolism leads to extensive cellular to hypoxia. HIFs are heterodimers composed of a constitutively metabolic disarray, and patients with peroxisomal disorders,

882 Cell Metabolism 22, 882–897, November 4, 2014 ª2014 Elsevier Inc. Cell Metabolism Hif-2a Promotes Mammalian Pexophagy

(legend on next page)

Cell Metabolism 22, 882–897, November 4, 2014 ª2014 Elsevier Inc. 883 Cell Metabolism Hif-2a Promotes Mammalian Pexophagy

as well as peroxisome-deficient mice, display severe hepatic To characterize the peroxisomal population in more detail, we dysfunction with fatty liver, cholestasis, and eventually cirrhosis isolated peroxisomes from control and Vhl/ livers (Figure S2A). (Raymond et al., 2009; Baes and Van Veldhoven, 2012; Faust activity was significantly decreased in postnuclear su- and Kovacs, 2014). pernatants from Vhl/ livers (Figure S2B). Peroxisomal protein

Since peroxisomal function depends on O2, we hypothesized levels were significantly decreased in subcellular fractions from / that to minimize O2 consumption under hypoxic conditions, Vhl livers, whereas levels of the mitochondrial proteins HIF-a signaling may inhibit O2-dependent peroxisomal meta- Tomm20, Trap1, and Vdac were similar to controls (Figure S2C). bolism and/or decrease the number of peroxisomes. In addition, Immunoblots of gradient fractions from control livers showed HIF-mediated reduction in peroxisomal activity could promote that peroxisomes were enriched in fractions 17–20 (density metabolic reprogramming causing changes in lipid metabolism. range of 1.144–1.169 g/ml), whereas peroxisomal protein levels To test this hypothesis, we chose the mouse liver as model sys- were significantly decreased in gradient fractions from Vhl/ tem because peroxisomes are highly abundant in the liver, and livers (Figure S2D). The distribution pattern of catalase activity liver-specific loss of pVhl causes severe lipid accumulation. In displayed a major peak coincident with peroxisomal proteins in this study, we systematically investigated peroxisome homeo- control livers, and catalase activity was significantly decreased stasis and metabolism in wild-type and liver-specific Vhl, Vhl/ in Vhl/ livers (Figure S2E). These results argue for the presence Hif1a, and Vhl/Hif2a knockout mice and explored the role of of intact peroxisomes in Vhl/ livers, but the distribution pattern Hif-1a and Hif-2a in this context. of peroxisomal proteins suggests a heterogeneous peroxisome population. In summary, hepatic loss of Vhl function is associ- RESULTS ated with a reduction in peroxisome abundance.

Peroxisome Abundance Is Reduced in Vhl–/– Livers Changes in Lipid Composition in Vhl–/– Livers Are To investigate whether peroxisomal abundance and/or meta- Reminiscent of Peroxisomal Disorders bolism are regulated by Hif-a, we inactivated the Vhl gene in Since peroxisomes play important roles in the degradation of the liver. As previously reported (Kucejova et al., 2011; Rankin VLCFAs and VLC-PUFAs, as well as in the synthesis of polyun- et al., 2009), 4-week-old (P28) liver-specific Vhl/ mice were saturated FAs, we investigated whether the decrease in peroxi- significantly smaller than control mice but had an increased liver some abundance had consequences on the lipid composition in to bodyweight ratio, displayed severe steatosis, were hypo- Vhl/ livers. We quantified the major lipid classes and the distri- glycemic, and Hif target gene expression was increased (Fig- bution of fatty acids within these classes in P28 control and ures S1A–S1G available online). Intriguingly, the protein levels Vhl/ livers and observed profound lipid alterations in Vhl/ of the peroxisomal matrix proteins catalase, acyl-CoA oxidase livers. The profile of triacylglycerol (TAG) species showed a (Acox1), multifunctional protein 1 (Mfp1), (Uox), disproportionate increase of TAGs with VLCFAs and VLC-PUFAs / as well as the peroxisomal membrane proteins Pex14p and in Vhl livers (Figures 1D and 1E; Table S1). C62 to C64 species Pex3p were significantly decreased in livers of P28 Vhl/ mice were hardly detectable in control livers but reached concentra- (Figure 1A), a phenomenon already evident at 2 weeks of age tions between 10–100 nmol/g liver tissue in Vhl/ livers. Most (Figure S1H). The effects of pVhl loss on peroxisomal protein TAG species showed an accumulation of 10-fold in Vhl/ levels were highly selective, as representative ER and mitochon- livers, but TAG species with a higher number of double bonds drial protein levels were similar in control and Vhl/ mice (Fig- were disproportionately increased 30- to 100-fold (Figures 1D ures 1A and S1H). Only the mitochondrial outer membrane and 1E). Next, we determined the distribution of fatty acids in / protein Vdac was moderately reduced in P28 Vhl mice. Anal- the liver. The hepatic levels of docosahexaenoic (DHA, C22:6n3) ysis of the number of peroxisomes by cytochemical localization and arachidonic acid (AA, C20:4n6) were significantly decreased of catalase and immunofluorescence of Pex14p and Acox1 in Vhl/ mice, whereas the parent fatty acids, linolenic revealed a strong reduction in peroxisome abundance in P14– (C18:3n3) and linoleic acid (C18:2n6), respectively, as well as doco- / P28 Vhl livers (Figures 1B and S1I). Catalase activity was sapentaenoic acid (C22:5n3) upstream to DHA synthesis were decreased 60%–70% in Vhl/ livers compared with control significantly increased (Figure 1F; Table S2). Blood levels of livers (Figures 1C and S1J). Electron microscopic comparisons VLCFA C26:0 and C24:0 lyso-phosphatidyl cholines were of control and Vhl/ livers revealed that peroxisomes are similar increased in Vhl/ mice (Figure S1L). Hence, VLCFAs and in size and shape (Figure S1K). VLC-PUFAs accumulations as well as DHA and AA deficiencies

Figure 1. Peroxisome Abundance Is Decreased in Vhl–/– Livers Peroxisomes were analyzed in P28 Vhlf/f (control) and liver-specific Vhl/ mice. (A) Immunoblots of liver lysates with antibodies against peroxisomal matrix (Catalase, Acox1, Mfp1, and Uox) and membrane proteins (Pex14p and Pex3p), mitochondrial (Vdac), and ER (ERp57 and Grp94) proteins. (B) Cytochemical localization of catalase and immunofluorescence of Pex14p and Acox1 in livers. (C) Catalase activity in liver lysates. (D) Concentration spectrum of TAG species. TAGs are designated numerically as carbon and double bond number (after the colon) in acyl chains. (See Table S1 for detailed quantitative lipid profiles.) (E) Relative abundance of the TAG lipid species in Vhl/ livers normalized to control values. (F) profiles in total lipid extracts from livers were determined by gas chromatography-mass spectrometry (GC-MS). Relative signal intensities were normalized to an internal standard and expressed as percentage of total fatty acids. Data are mean ± SD (n = 6). *p < 0.05; **p < 0.01; ***p < 0.001 versus control mice.

884 Cell Metabolism 22, 882–897, November 4, 2014 ª2014 Elsevier Inc. Cell Metabolism Hif-2a Promotes Mammalian Pexophagy

Figure 2. Hif-2a-Dependent Decrease of Peroxisome Abundance in Vhl–/– Livers Peroxisomes were analyzed in P28 control (Cre: Vhlf/f, Vhlf/f/Hif1af/f, and Vhlf/f/Hif2af/f) and liver-specific Vhl/, Vhl//Hif1a/, and Vhl//Hif2a/ mice. (A) Immunoblots of peroxisomal proteins, Hif-2a, and pVhl in liver lysates. (B) Immunofluorescence of Pex14p in livers. (C) Catalase activity in liver lysates. (legend continued on next page) Cell Metabolism 22, 882–897, November 4, 2014 ª2014 Elsevier Inc. 885 Cell Metabolism Hif-2a Promotes Mammalian Pexophagy

are characteristic features of hepatic Vhl loss of function. These droxycholestanoic acid (DHCA) and 3a,7a,12a-trihydroxycho- changes in lipid composition are reminiscent of alterations due to lestanoic acid (THCA) were significantly increased in the defective peroxisomal b-oxidation in peroxisome-deficient mice plasma of these mice (Figures S4B and S4C). C27-bile acids and patients with peroxisomal disorders (Baes and Van Veld- accumulate in plasma of patients and mice with peroxisome hoven, 2012; Raymond et al., 2009; Van Veldhoven, 2010). deficiency disorders; they are toxic and have been suggested to contribute to liver disease associated with peroxisomal disor- Hif-2a but Not Hif-1a Activation Mediates the Decrease ders (Keane et al., 2007; Ferdinandusse et al., 2009; Hiebler of Peroxisome Abundance in Vhl–/– Livers et al., 2014). CA and CDCA levels were similar in the plasma To examine if the reduction of peroxisome number upon pVhl of control and Vhl//Hif2a/ mice, and DHCA and THCA loss is mediated by Hif-1a or Hif-2a, we analyzed peroxisome levels were decreased in the plasma of Vhl//Hif2a/ mice abundance in liver-specific Vhl/Hif1a and Vhl/Hif2a double compared to controls (Figure S4D). knockout mice (Figures 2 and S3). P28–P42 Vhl//Hif1a/ mice were significantly smaller than control mice, they were hy- Peroxisome Proliferation Can Be Induced in Vhl–/– Livers poglycemic, and their liver was enlarged and revealed severe Peroxisome homeostasis is achieved by balancing biogenesis steatosis similar to Vhl/ mice (Figures S3A–S3C). In contrast, and degradation of peroxisomes. To explore potential mecha- Vhl//Hif2a/ mice were phenotypically similar to control nisms underlying the pronounced reduction of peroxisome mice (Figure S3A). With respect to the peroxisomal phenotype, abundance in Vhl/ livers, we assessed the expression of Pex we observed a striking rescue in Vhl//Hif2a/ but not genes, which encode essential for peroxisome biogen- Vhl//Hif1a/ mice. Peroxisomal protein levels were signifi- esis (Smith and Aitchison, 2013). The expression of Pex genes cantly decreased in Vhl//Hif1a/ livers (Figures 2A and was similar in P14–P28 control and Vhl/ livers (Figures 3A S3D), whereas their levels were similar in control and Vhl// and S1M). Pex11a was the only that showed significant Hif2a/ livers (Figures 2A and S3E). Immunofluorescence of transcriptional upregulation in Vhl/ (Figures 3A and S1M) and Pex14p revealed a strong reduction of peroxisome number in Vhl//Hif1a/ (Figure 3B) livers, whereas Pex11a was similarly Vhl//Hif1a/ but not Vhl//Hif2a/ livers (Figure 2B). Cata- expressed in control and Vhl//Hif2a/ livers (Figure 3B). Since lase activity was similarly decreased in P28 Vhl/ and Vhl// overexpression of Pex11a promotes peroxisome division and in- Hif1a/ livers compared to controls, whereas catalase activity duces peroxisome abundance (Schrader et al., 2012), this result was similar in control and Vhl//Hif2a/ livers (Figures 2C may reflect a feedback loop in which Vhl/ and Vhl//Hif1a/ and S3F). Hepatic TAG levels were significantly increased in hepatocytes sense the reduction of peroxisomes and attempt to Vhl/ and Vhl//Hif1a/ livers and only marginally increased counterbalance their lack by inducing a positive regulator of in Vhl//Hif2a/ livers compared to controls (Figure 2D). The peroxisome biogenesis. expression of Hif target genes was, with the exception of Epo, Given the fact that the expression of the peroxisome biogen- similarly increased in P28–P42 Vhl/, Vhl//Hif1a/, and esis machinery was not impaired in Vhl/ livers and that perox- Vhl//Hif2a/ livers (Figures 2E and S3G). Thus, both Hif-1a isome biogenesis proceeds, at least in part, through division of and Hif-2a were stabilized in Vhl/ livers, but the reduction in pre-existing peroxisomes, we tested whether pharmacological peroxisome number depends exclusively on Hif-2a activity. induction of peroxisome proliferation was possible in Vhl/ Liver-specific loss of Hif-1a or Hif-2a alone did not affect perox- livers. We treated Vhl/ mice either with WY-14,643 (WY), which isomal protein levels (Figure S3H). is a peroxisome proliferator-activated receptor a (Ppara) ligand, Next, we investigated the composition of hepatic TAGs in or 4-phenylbutyrate (PBA), which acts independently of Ppara. Vhl//Hif1a/ and Vhl//Hif2a/ livers. TAG species with Treatment of mice for 3 days with PBA resulted in a strong induc- VLCFAs and VLC-PUFAs increased in Vhl//Hif1a/ livers tion of catalase activity (Figure 3C) and peroxisomal protein (Figure 2F; Table S1) but changed only moderately in Vhl// levels (Figure 3E) in control and Vhl/ livers. However, PBA Hif2a/ livers (Figure S3I; Table S1). Changes in the fatty acid treatment was stopped after 3 days due to a negative effect on profile of Vhl//Hif1a/ livers were similar as in Vhl/ livers the health state of Vhl/ mice. WY treatment for 7 days strongly (Figure 2G; Table S2). Importantly, fatty acid profiles from Vhl/ increased catalase activity (Figure 3D) and peroxisomal protein and Vhl//Hif1a/ livers were comparable to P13 peroxisome- levels in control livers and to a lesser extent in Vhl/ livers (Fig- deficient Pex2/ livers that also displayed steatosis (Figure S4A) ure 3F). Cytochemical staining of catalase activity revealed that (Keane et al., 2007; Kovacs et al., 2012). peroxisome number was significantly increased both in control Bile acids were analyzed in plasma from our mouse models. and Vhl/ livers upon WY treatment (Figure 3G), but in Vhl/ The plasma of Vhl/ and Vhl//Hif1a/ mice was yellowish livers, peroxisome abundance was still lower than in untreated with increased levels of cholic acid (CA) and chenodeoxycholic controls. Electron microscopy demonstrated that peroxi- acid (CDCA) (Figures S4B and S4C), which is a sign of chole- somes from Vhl/ livers were similarly shaped as those from stasis. Importantly, the levels of the C27-bile acids 3a,7a-dihy- control mice, although generally smaller in size (Figure 3G).

(D) Hepatic TAG content. (E) Expression of Hif target genes in P28 livers. Each value represents the amount of mRNA relative to that in control mice, which was arbitrarily definedas1. (F) Concentration spectrum of TAG species in livers. Data are mean ± SD (n = 6; see Table S1 for detailed quantitative lipid profiles). (G) Fatty acid profiles in total lipid extracts from livers were determined by GC-MS (n = 6 mice for each genotype). Relative signal intensities were normalized to an internal standard and expressed as percentage of total fatty acids. Data are mean ± SD (n = 7 for Vhlf/f, n = 5 for Vhl/, n = 3 for Vhlf/f/Hif1af/f, n = 4 for Vhl// Hif1a/, n = 4 for Vhlf/f/Hif2af/f, n = 5 for Vhl//Hif2a/ mice). *p < 0.05; **p < 0.01; ***p < 0.001 versus control mice.

886 Cell Metabolism 22, 882–897, November 4, 2014 ª2014 Elsevier Inc. Cell Metabolism Hif-2a Promotes Mammalian Pexophagy

Figure 3. WY and PBA Induce Peroxisome Proliferation in Vhl–/– Mice (A) Expression of Pex genes in P28 livers (n = 5–7 mice). (B) Expression of Pex11a in P28–P42 livers (n = 10 mice). (C and D) Catalase activity in liver lysates.

(legend continued on next page) Cell Metabolism 22, 882–897, November 4, 2014 ª2014 Elsevier Inc. 887 Cell Metabolism Hif-2a Promotes Mammalian Pexophagy

The short-term induction of peroxisomes in Vhl/ livers upon structures containing either single or clusters of peroxisomes PBA or WY treatment was insufficient to reduce hepatic TAG were detected by electron microscopy in hepatocytes of Vhl/ levels and rescue the massive steatosis in Vhl/ mice (Figures but not in control mice (Figures 4E and 4F), arguing for increased 3H and 3I). These results argue for a functional peroxisome pexophagy in Vhl/ livers. biogenesis machinery in Vhl/ livers. Lonp2 expression was similar in control and Vhl/, Vhl// Since peroxisome proliferation in Vhl/ mice was only moder- Hif1a/, and Vhl//Hif2a/ livers (Figures S5F–S5H). Hepatic ately increased in response to WY treatment and accumulation mRNA levels of Alox15 were very low and only moderately of VLCFAs and VLC-PUFAs is known to induce Ppara signaling increased in Vhl/ and Vhl//Hif1a/ livers compared to con- and peroxisome proliferation, we determined the mRNA levels trols (Figures S5F and S5G). Protein levels of Alox15 were similar of Ppara target genes in WY-treated mice (Figure 3J). The in control and Vhl/ livers (Figure S5E). Hence, Lonp2 and mRNA levels of Ppara and its target genes were significantly Alox15 do not behave as Hif target genes whose induction could increased both in control and Vhl/ mice upon WY treatment promote peroxisome turnover. (Figure 3J), but WY-mediated transcriptional induction of Ppara target genes was much weaker in Vhl/ mice (Figure 3J). Except The Autophagy Machinery Is Required for for Pex11a, Ppara target gene expression was similar in vehicle- Hif-2a-Mediated Pexophagy treated control and Vhl/ mice (Figure 3J). Likewise, induction To directly assess whether autophagy is responsible for the of Ppara target genes was impaired upon fasting in Vhl/ peroxisomal phenotype in Vhl/ livers, we examined the effect mice, whereas their expression was similar in ad-libitum-fed of constitutive Hif-2a signaling on peroxisome abundance in control and Vhl/ mice (Figure 3K). Thus, Hif-2a signaling has liver-specific, autophagy-deficient Atg7/ mice. We generated a repressive effect on ligand-dependent Ppara transcriptional an adenovirus coexpressing GFP and HIF-2a P405/531A (here- activity and by extension on stimulated peroxisome biogenesis. after called Ad-HIF-2a(mt)), a HIF-2a variant that escapes degra- dation by pVHL. We injected Atg7f/f and liver-specific Atg7/ Autophagy Is Induced in Vhl–/– Mice mice either with Ad-HIF-2a(mt) or an adenovirus expressing Since peroxisome biogenesis and proliferation are functional, we GFP (Ad-GFP) as control and sacrificed mice after 6 days. addressed if increased degradation of peroxisomes underlies Expression of HIF-2a(mt) caused hepatomegaly and severe stea- the pronounced reduction of peroxisome abundance in Vhl/ tosis in Atg7f/f livers, while Atg7/ livers remained unaffected livers. Three mechanisms for peroxisome degradation have (Figure 5A). Almost all hepatocytes in Atg7f/f livers expressing been described in mammalian cells, which include selective HIF-2a(mt) displayed macrovesicular steatosis (Figure 5A), an ef- autophagy (pexophagy), proteolysis by Lon protease 2 fect not seen in HIF-2a(mt)-expressing Atg7/ livers (Figure 5A). (Lonp2), and 15-lipoxygenase-1 (Alox15)-mediated autolysis Peroxisome abundance and peroxisomal protein levels were (Till et al., 2012). To quantify autophagic flux in vivo, control significantly decreased in Atg7f/f livers expressing HIF-2a(mt) and Vhl/ mice were infected with an adenovirus expressing compared to controls (Figures 5B and 5C), while peroxisome GFP-tagged LC3 (Ad-GFP-LC3). The number of GFP-LC3-posi- homeostasis was similar in Atg7/ hepatocytes expressing tive vesicles was increased 3-fold in Vhl/ livers compared to HIF-2a(mt) compared to GFP-expressing Atg7/ hepatocytes. control mice (Figure 4A), indicating increased autophagic flux. Overexpression of HIF-2a(mt) did not affect mitochondrial and Conversion of nonlipidated LC3-I into the phosphatidylethanol- ER protein levels (Figure 5C). In accord with the histological anal- amine-conjugated LC3-II form was significantly increased and ysis revealing severe steatosis, lipid droplet protein Adfp (adipose protein levels of the autophagic substrate p62 were decreased differentiation-related protein) levels increased significantly in in Vhl/ livers (Figure 4B). In summary, these results show Atg7f/f but not Atg7/ livers expressing HIF-2a(mt) (Figure 5C). that autophagic flux is increased in Vhl/ livers. Catalase activity was decreased 60% in Atg7f/f and remained Immunofluorescence of catalase revealed that GFP-LC3-pos- unchanged in Atg7/ livers expressing HIF-2a(mt) as compared itive colocalized with peroxisomes in Vhl/ to controls (Figure 5D). livers (Figures 4C, S5A, and S5B). We analyzed the autophagic Although Atg7f/f and Atg7/ mice received identical amounts sequestration of peroxisomes with superresolution microscopy of Ad-HIF-2a(mt), HIF-2a(mt) protein levels were higher in Atg7f/f capable of 3D-structured illumination (3D-SIM), which provides compared to Atg7/ livers (Figure 5C). A possible explanation is an increase of both axial and lateral resolution by a factor of that autophagy is important for efficient adenoviral infection and two compared to confocal microscopy. Reconstructed 3D-SIM replication (Zeng and Carlin, 2013). Hence, one could argue that images show that both single and multiple peroxisomes were se- the amounts of HIF-2a(mt) produced in Atg7/ liver cells are questrated in autophagosomes and surrounded by an autopha- insufficient to promote pexophagy. To address this, we gosomal membrane decorated with GFP-LC3 (Figures 4D, S5C, compared Hif-2a levels in Vhl/ and Vhl//Hif1a/ livers and S5D; Movies S1 and S2). Importantly, -like with HIF-2a(mt) levels in Atg7/ livers (Figure 5E). Hif-2a levels

(E and F) Immunoblots of peroxisomal proteins in liver lysates. (G) Cytochemical localization of catalase in livers. Electron micrographs of ultrathin sections are shown in the right panel. (H and I) Hepatic TAG content in livers. (J) Expression of Ppara target genes in livers. (K) Expression of Ppara target genes in livers from ad-libitum-fed and overnight-fasted P28 mice. Data are mean ± SD (Ad-libitum-fed mice: n = 5–7; 12 hr fasted mice: n = 3) (Vehicle: n = 4–8 mice; WY: n = 4–8 mice) (Vehicle: n = 3 mice; PBA: n = 5 mice). *p < 0.05; **p < 0.01; ***p < 0.001 versus vehicle-treated or ad libitum- fed control mice. #, p < 0.05; ##, p < 0.01; ###, p < 0.001 versus WY- or PBA-treated or fasted control mice.

888 Cell Metabolism 22, 882–897, November 4, 2014 ª2014 Elsevier Inc. Cell Metabolism Hif-2a Promotes Mammalian Pexophagy

Figure 4. Autophagic Flux Is Increased in Vhl–/– Livers Control and Vhl/ mice were infected with Ad-GFP-LC3, treated with CQ, and sacrificed at the age of 4 weeks. (A) Representative pictures of GFP-LC3 punctate structures in hepatocytes and quantification of GFP-LC3 punctate structures per cell. Data are mean ± SD. **p < 0.01. (B) Immunoblots of autophagy-related proteins in liver lysates. (C) Immunofluorescence analysis for the colocalization of autophagosomes with peroxisomes. Liver sections were immunostained for GFP and catalase. Arrowheads: yellow dots represent colocalized GFP-LC3-positive autophagosomes with red fluorescence-labeled peroxisomes. See also Figures S5A and S5B for analysis of confocal z stack images with the orthogonal slicer tool of the Imaris software. (legend continued on next page) Cell Metabolism 22, 882–897, November 4, 2014 ª2014 Elsevier Inc. 889 Cell Metabolism Hif-2a Promotes Mammalian Pexophagy

in Vhl//Hif1a/ livers were comparable with HIF-2a(mt) levels was significantly increased in livers of 3-MA-treated Vhl/ in Atg7/ livers. Since Vhl//Hif1a/ livers revealed a strong mice (Figures 6A and 6B), while peroxisomal protein levels reduction in peroxisome abundance, HIF-2a(mt) levels in were similar in untreated and 3-MA-treated control livers. Resto- Atg7/ livers are likely sufficient to induce Hif-2a-mediated pex- ration of peroxisome abundance in 3-MA-treated Vhl/ mice ophagy, and the deficiency of autophagy prevents this outcome. confirmed that loss of peroxisomes in Vhl/ livers is due to To further corroborate this assertion, we asked whether lowering enhanced pexophagy, whereas lack of peroxisome accumula- HIF-2a(mt) levels in Atg7f/f mice by injection of less pfus could tion in treated control mice suggests that their turnover rate un- still decrease peroxisome abundance. Injection of different der basal conditions is low. Hif target gene expression was not pfus resulted in a dose-dependent decrease of HIF-2a(mt) levels affected by 3-MA treatment (Figure S7E). Adfp protein levels (Figure 5F) and a mosaic infection pattern in those livers (Fig- were decreased in 3-MA-treated Vhl/ mice compared to un- ure S6A). A direct comparison of peroxisome abundance in in- treated Vhl/ mice (Figure 6A), which is consistent with attenu- fected and uninfected hepatocytes revealed a strong reduction ated hepatic steatosis (Figure S7B). 3-MA treatment did not in peroxisome numbers in hepatocytes expressing HIF-2a(mt) affect ER and mitochondrial protein levels (Figure 6A). These compared to uninfected hepatocytes (Figures 5G and S6B). data corroborate the functional requirement for autophagy in We determined the expression of Hif target genes in Atg7f/f the decrease in peroxisome content downstream of Hif-2a stabi- and Atg7/ mice infected with different amounts of Ad-HIF- lization in Vhl/ livers. 2a(mt). The mRNA levels of Hif target genes were significantly Protein levels of the selective autophagy receptors p62 and increased in livers of Atg7f/f and Atg7/ mice infected with Nbr1 (neighbor of BRCA1 gene 1) (Rogov et al., 2014) were 3 3 109 pfu of Ad-HIF-2a(mt) compared to controls (Figure S6C), increased in 3-MA-treated Vhl/ mice compared to untreated but the induction was much weaker in Atg7/ livers. However, Vhl/ mice (Figure 6A) and were similar in untreated and 3- their expression was comparable to Atg7f/f mice infected with MA-treated control mice. Nbr1 and p62 levels and peroxisome 1 3 109 pfu of Ad-HIF-2a(mt) (Figure S6C). In mice infected abundance were concomitantly decreased in Vhl//Hif1a/ with low doses of Ad-HIF-2a(mt) (1 3 108 or 5 3 108 pfus), little, livers, whereas their levels were similar in control and Vhl// if any, induction of Hif target genes was observed. These results Hif2a/ livers (Figure 6C). Loss of Atg7 resulted as expected suggest that Atg7/ livers display a mosaic infection pattern in the accumulation of Nbr1 and p62, and Nbr1 levels were and that infected hepatocytes produce sufficient HIF-2a(mt) to decreased in Atg7f/f livers expressing HIF-2a(mt) (Figure 6C). induce pexophagy. Overexpression of NBR1 and p62 has been shown to induce Overexpression of HIF-2a(mt) in Atg7f/f mice led to similar degradation of peroxisomes (Deosaran et al., 2013), but mRNA changes in hepatic lipid composition as seen in Vhl/ and levels of Nbr1 and p62 were similar in control and Vhl/, Vhl//Hif1a/ livers. Most TAG species showed an accumula- Vhl//Hif1a/, and Vhl//Hif2a/ livers (Figures S7C and tion of 10-fold in HIF-2a(mt)-expressing Atg7f/f livers compared S7D). In addition, overexpression of HIF-2a in Atg7f/f livers to controls, and TAGs with VLCFAs and fatty acids with a high decreased mRNA levels of Nbr1 and p62 (Figure S6C). These re- number of double bonds were again disproportionately sults demonstrate that Nbr1 and p62 are not Hif target genes increased up to 160-fold (Figures 5H and S6D). In summary, whose induction could promote pexophagy. these results support the view that Hif-2a-mediated pexophagy Since our data suggested that Nbr1 and p62 are involved in requires a functional autophagy machinery, and HIF-2a(mt) gain- Hif-2a-mediated pexophagy, we determined the subcellular of-function leads to similar lipid changes as Vhl/ and Vhl// localization of Nbr1 and p62. The prediction would be that Hif1a/ livers. Nbr1 and/or p62 localize to peroxisomes in Vhl/ mice to mediate pexophagy. Immunofluorescence of catalase revealed Pharmacological Inhibition of Autophagy Increases colocalization of peroxisomes with Nbr1 and p62 in untreated Peroxisome Abundance in Vhl–/– Livers Vhl/ mice (Figures 6D and S7F). Interestingly, Nbr1 already Next, we tested whether pharmacological inhibition of auto- localized to peroxisomes in control livers, whereas p62 was phagy would restore peroxisome abundance in Vhl/ livers. barely detectable (Figures 6D and S7F). p62 accumulated as Treatment of Vhl/ mice with chloroquine (CQ) significantly focal aggregates in 3-MA-treated control livers, and some of increased peroxisomal protein levels in Vhl/ livers (Figure S7A). these aggregates showed colocalization with Nbr1 and catalase, To inhibit autophagy early at presequestration steps (Petiot et al., but in contrast to Nbr1, most peroxisomes did not colocalize with 2000; Seglen and Gordon, 1982), P21 control and Vhl/ mice p62 (Figure S7F). In addition to the restoration of peroxisome were treated with 3-methyladenine (3-MA) for 5 days. 3-MA- abundance, we observed significant clustering of Nbr1- and treated Vhl/ mice displayed improved tonus and activity p62-positive peroxisomes in 3-MA-treated Vhl/ mice (Figures compared to untreated Vhl/ mice. Peroxisome abundance 6D and S7F). Moreover, immunofluorescence revealed that

(D) Superresolution microscopy shows that both single (left panel) or multiple (right panel) peroxisomes (shown in red) are sequestrated in autophagosomes and surrounded by an autophagosomal membrane decorated with GFP-LC3. Liver sections were stained with antibodies against GFP and catalase. See also Figures S5C and S5D and Movies S1 and S2. (E and F) Electron microscopy detection of peroxisomal autophagy in Vhl/ livers. Livers were fixed and stained with DAB for catalase, which generates an electron-dense precipitate over the . (E) The left panel shows a double-membrane structure, possibly a phagophore, enclosing a peroxisome. The right panel shows a peroxisome that is sequestered within a double-membrane autophagosome. (F) A cluster of peroxisomes is contained in a single-membrane-bound compartment, representing a putative autolysosome.

890 Cell Metabolism 22, 882–897, November 4, 2014 ª2014 Elsevier Inc. Cell Metabolism Hif-2a Promotes Mammalian Pexophagy

(legend on next page)

Cell Metabolism 22, 882–897, November 4, 2014 ª2014 Elsevier Inc. 891 Cell Metabolism Hif-2a Promotes Mammalian Pexophagy

GFP-LC3-positive autophagosomes colocalized with Nbr1 and in contrast to 39% with medium number and 46% with many catalase in Vhl/ livers (Figure S7G). peroxisomes (Figure 7D). Interestingly, peroxisome abundance Analysis of purified peroxisomes from livers of untreated and is reduced more frequently in well-differentiated grade 1 as in 3-MA-treated control and Vhl/ mice also confirmed the perox- grade 2 or 3 tumors (Figure 7E). In summary, the negative regu- isomal localization of Nbr1 (Figures S2, S7H, and S7I). 3-MA latory effect of HIF-2a signaling on peroxisome abundance is not treatment significantly increased catalase activity in peroxisomal restricted to hepatocytes and may extend to other tissues and peak fractions of control and Vhl/ livers (Figure S7H). Interest- disease settings that are characterized by high levels of HIF-2a ingly, the density of peroxisomes was increased in 3-MA-treated such as ccRCC. mice (Figure S7H). Peroxisomal protein levels were increased in gradient fractions of 3-MA-treated Vhl/ mice compared to un- DISCUSSION treated knockouts (Figure S7I). Nbr1 was present in peroxisomal fractions from untreated control livers, whereas in untreated The main finding of this report is that the transcription factor Hif- Vhl/ livers Nbr1 was not detected probably due to the low 2a, but not Hif-1a, is a negative regulator of peroxisome abun- amount of peroxisomes (Figures S2D and S7I). 3-MA treatment dance in hepatocytes. In fact, Hif-2a promotes pexophagy since increased Nbr1 and p62 protein levels and both autophagic re- Hif-2a-mediated downregulation of peroxisome abundance was ceptors were present in peroxisomal fractions of control and suppressed in liver-specific autophagy-deficient Atg7/ mice Vhl/ livers (Figure S7I). Collectively, we conclude that Nbr1 and after inhibition of autophagy with 3-MA in Vhl/ mice. The and p62 are recruited to peroxisomes and participate in Hif- autophagy receptor Nbr1 localizes to peroxisomes and is 2a-mediated pexophagy in Vhl/ livers, where they are deg- degraded together with peroxisomes by Hif-2a-mediated pex- raded together with peroxisomes in autolysosomes. ophagy. Reduced number of peroxisomes and the ensuing defi- ciency in peroxisomal function encompass major changes in the Peroxisome Abundance Is Reduced in Clear Cell Renal hepatic lipid profile that are reminiscent of peroxisomal disorders Cell Carcinomas with High HIF-2a Levels and include the accumulation of VLCFAs and VLC-PUFAs. To explore the relevance of HIF-2a-dependent pexophagy in These are activating ligands for the transcription factor human disease, peroxisome abundance was analyzed in more Ppara, a promoter of peroxisome proliferation. However, Hif- than 200 clear cell renal cell carcinoma (ccRCC) tissue samples 2a restrains ligand-induced Ppara-mediated peroxisome pro- on a tissue microarray (TMA) and compared with HIF-2a expres- liferation. Thus, by simultaneously inducing pexophagy and sion and differentiation grade. This disease setting was chosen counteracting Ppara, Hif-2a ensures efficient depletion of the for several reasons. First, loss of VHL function occurs in up to peroxisome pool (Figure 7F). Finally, we demonstrate that perox- 90% of sporadic ccRCC, the most common form of kidney can- isome abundance is significantly reduced in ccRCCs that are cer (Hakimi et al., 2013). Second, HIF-2a is considered to be a characterized by high HIF-2a levels, providing a link between driver oncoprotein for ccRCC. Third, peroxisomes are highly reduced peroxisome abundance and HIF-2a expression in hu- abundant in the kidney, and kidneys are affected in peroxisome man disease. biogenesis disorder patients. Fourth, at least in the context of Peroxisomes adapt rapidly to cellular demands, and their ho- steatosis and angiogenesis, ccRCC and Vhl/ livers behave meostasis is achieved by the counterbalance between organelle similarly. biogenesis and degradation (Smith and Aitchison, 2013; Till Peroxisome abundance varies between the different cell types et al., 2012). Peroxisomal function depends highly on molecular of the kidney. They are highly abundant in cells of proximal tu- O2, and the identification of Hif-2a as an activator of pexophagy bules and less abundant in distal tubules and collecting duct suggests that peroxisome homeostasis is directly coupled to O2 (Figure 7A). Organelle abundance was classified according to availability. Hence, Hif-2a-mediated pexophagy may well be part the situation in normal tissue: many organelles correspond to of adaptive responses to hypoxia. Hypoxia signaling is involved proximal tubules, a medium number to distal tubules and few or- in the pathogenesis of various liver diseases that are associated ganelles describe a number below the abundance observed in with changes in lipid metabolism (Nath and Szabo, 2012), and normal kidney tissue (Figure 7B). In tumor specimen with strong Hif-2a-mediated pexophagy might contribute to those changes HIF-2a staining, 42% had few peroxisomes and only 8% had in lipid metabolism. Indeed, we showed that Hif-2a-mediated many peroxisomes (Figure 7D). Among the specimens that pexophagy associated with decreased peroxisomal metabolism were negative for HIF-2a only 15% displayed few peroxisomes resulted in the accumulation of VLCFAs and VLC-PUFAs and

Figure 5. Overexpression of HIF-2a P405/531A (HIF-2a(mt)) Reduces Peroxisome Abundance in Atg7f/f but Not in Atg7–/– Livers (A) Hematoxylin and eosin staining on liver sections. (B) Immunofluorescence of Pex14p and GFP in livers. Some Atg7/ hepatocytes displayed large peroxisomal clusters, and confocal z stack analysis with 3D reconstruction revealed that these clusters represent a network of interconnected, reticular peroxisomal structures and individual peroxisomes (see Movies S3 and S4). (C) Immunoblots of liver lysates. The asterisk indicates a nonspecific band. (D) Catalase activity in liver lysates. Data are mean ± SD (Ad-GFP: n = 3–6; Ad-HIF-2a(mt): n = 7). ***p < 0.001 versus GFP-infected mice. (E and F) Immunoblots of Hif-2a in P28 liver lysates. (G) Immunofluorescence of Pex14p and GFP in livers from Atg7f/f mouse infected with 1 3 109 pfu of Ad-HIF-2a(mt). (H) Concentration spectrum of TAG species in livers of Atg7f/f mice infected with Ad-GFP or Ad-HIF-2a(mt). Data are mean ± SD (n = 6 to 7) (see Table S1 for detailed quantitative lipid profiles).

892 Cell Metabolism 22, 882–897, November 4, 2014 ª2014 Elsevier Inc. Cell Metabolism Hif-2a Promotes Mammalian Pexophagy

Figure 6. Inhibition of Autophagy with 3-MA Increases Peroxisome Abundance in Vhl–/– Livers (A) Immunoblots of liver lysates. (B) Immunofluorescence of Pex14p in livers.

(legend continued on next page) Cell Metabolism 22, 882–897, November 4, 2014 ª2014 Elsevier Inc. 893 Cell Metabolism Hif-2a Promotes Mammalian Pexophagy

depletion of DHA and AA. Such lipid changes are characteristic Hif-2a signaling increases the rate of Nbr1 interaction with per- features of peroxisomal disorders (Van Veldhoven, 2010; Ray- oxisomes. The accumulation of Nbr1 on peroxisomes likely re- mond et al., 2009; Baes and Van Veldhoven, 2012). Decreased cruits p62 to achieve a critical mass required for pexophagy. DHA and AA levels have also been observed in patients with This may concentrate sufficient ubiquitin-like modifiers (e.g., nonalcoholic fatty liver disease and steatohepatitis (Puri et al., LC3 and Gabaraps) in close proximity to peroxisomes to prime 2007, 2009). Since substrates for peroxisomal b-oxidation also phagophore assembly. Hif-2a could also promote a posttran- include branched-chain and dicarboxylic FAs, C27-bile acids, slational modification of a peroxisomal membrane protein or and eicosanoids (Van Veldhoven, 2010), Hif-2a-mediated pex- Nbr1 that induces pexophagy. For example, hypoxia-induced ophagy likely contributes to additional lipid alterations. More- dephosphorylation of FUNDC1, a mitochondrial outer mem- over, alterations in lipid metabolism may derive also from indirect brane protein, enhances its interaction with the core autophagic consequences of reduced peroxisome function like activation of machinery and promotes mitophagy (Liu et al., 2012). Nbr1 ER stress pathways and mitochondrial dysfunction (Baes and phosphorylation by glycogen synthase kinase 3 prevents the ag- Van Veldhoven, 2012; Faust and Kovacs, 2014). gregation of ubiquitinated proteins and their selective autopha- Autophagy is an evolutionary conserved catabolic process gic degradation (Nicot et al., 2014). Hif-2a could activate for degradation of macromolecules and organelles. Both nonse- mRNA translation of a pexophagy receptor or a protein-modi- lective ‘‘bulk’’ autophagy and selective autophagy of specific fying gene as part of its recently identified transcription-indepen- proteins or organelles have been described (Mizushima et al., dent role in the regulation of translation (Uniacke et al., 2012). 2011; Schreiber and Peter, 2014). The key factor in selective The identification of Hif-2a as a major inducer of pexophagy pro- autophagy is specific cargo-recognizing receptors that connect vides now a path toward delineating the underlying mechanism. the cargo to the core autophagic machinery. Two pexophagy re- Low numbers of peroxisomes have been observed in various ceptors, Atg30 and Atg36, have been identified in (Farre´ tumor cells (Frederiks et al., 2010). Our analysis of more than et al., 2008; Motley et al., 2012), and their overexpression stimu- 200 ccRCC tissue samples revealed a striking correlation be- lates pexophagy even under peroxisome-inducing conditions. tween high HIF-2a levels and loss of peroxisomes, further sup- However, there are no orthologous genes in mammals, and no porting a direct association between activation of HIF-2a and pexophagy receptors have been identified. Previous studies pexophagy. Of particular interest in this regard is the relation- showed that hypoxia-induced macroautophagy or mitophagy ship between peroxisome abundance and the histological tu- are Hif-1a dependent (Semenza, 2011), but a specific Hif-2a mor differentiation grade, since peroxisome abundance is function in the context of autophagy has not been reported. reduced more frequently in well-differentiated tumors. The ma- While it remains an open question how Hif-2a induces pexoph- jority of ccRCC is associated with constitutive upregulation of agy at the detailed molecular level, several possibilities exist. HIF-a due to loss of pVHL function. HIF-2a is considered to Hif-2a could induce the expression of a pexophagy receptor. be a core oncogenic driver in ccRCC and is upregulated in Overexpression of NBR1 and p62, which are receptors for selec- the vast majority of solid tumors (Franovic et al., 2009; Qing tive autophagy of ubiquitinated targets, induces clustering and and Simon, 2009). Accumulation of neutral lipids and glycogen degradation of peroxisomes in cell lines (Deosaran et al., are characteristic features of ccRCCs. Thus, alterations in lipid 2013), but Nbr1 and p62 are not induced by Hif-2a. Artificial metabolism as a result of deregulated peroxisome homeostasis monoubiquitination of peroxisomal membrane proteins in play likely an important but unappreciated role in ccRCC devel- mammalian cells is sufficient to cause NBR1- and p62-depen- opment (Currie et al., 2013; Santos and Schulze, 2012). Fatty dent pexophagy (Kim et al., 2008), but it is unknown if a peroxi- acids support cancer growth by providing substrates for energy somal membrane protein is ubiquitinated under pexophagy- production or by generating building blocks for membranes and inducing conditions and whether subsequent interaction with signaling lipids in proliferating cells. VLC-PUFAs could be con- Nbr1 and/or p62 links ubiquitinated peroxisomes to the auto- verted to eicosanoids, biologically active lipids involved in phagic machinery. One possibility is that Hif-2a induces an E3 inflammation and carcinogenesis. Eicosanoids are degraded ubiquitin ligase that mediates ubiquitination of a peroxisomal in peroxisomes, and dysbalance of peroxisome homeostasis membrane protein. However, Hif-2a did not induce the expres- will affect eicosanoid signaling. Peroxisomes are essential for sion of peroxisomal E3 ligases (Pex2, Pex10, and Pex12). Alter- the synthesis of ether lipids, and they are increased in aggres- natively, Nbr1 could be recruited to peroxisomes independently sive cancers (Lodhi and Semenkovich, 2014). Changes in of ubiquitin, because Nbr1 (but not p62) contains a membrane- VLCFAs, ether lipids, and DHA levels affect membrane proper- interacting domain capable of binding to peroxisomes (Deosaran ties, thereby influencing membrane microdomain organization et al., 2013), and Nbr1 localized to peroxisomes even in control and trafficking pathways (Lodhi and Semenkovich, 2014; Obara livers. In addition, p62 is not required for pexophagy when et al., 2013). Nbr1 is in excess, but its binding to Nbr1 increases the efficiency In summary, the findings presented here provide now an of Nbr1-mediated pexophagy (Deosaran et al., 2013). Since impetus to further characterize the functional consequences of 3-MA treatment of Vhl/ mice induced significant clustering HIF-2a-driven pexophagy and associated lipid alterations in of Nbr1- and p62-positive peroxisomes, we hypothesize that cancer and to exploit this knowledge for therapy.

(C) Immunoblots of p62 and Nbr1 in liver lysates. (D) Immunofluorescence of catalase, Nbr1, and p62 (shown in Figure S7F). The right panel shows enlarged images of the boxed regions in the overlay panels. Arrowheads represent Nbr1-positive aggregates that colocalize with catalase. Arrows represent peroxisome clusters in livers of 3-MA-treated Vhl/ mice. Insets in the bottom panel show higher magnification of the colocalization of peroxisome clusters with Nbr1 in livers of 3-MA-treated Vhl/ mice.

894 Cell Metabolism 22, 882–897, November 4, 2014 ª2014 Elsevier Inc. Cell Metabolism Hif-2a Promotes Mammalian Pexophagy

Figure 7. Strong HIF-2a Staining Correlates with a Reduction of Peroxisome Abundance in ccRCCs (A) Characterization of peroxisome distribution and abundance in healthy kidney. Sequential sections of a biopsy from a healthy kidney were stained with specific markers for proximal tubules (p), glomeruli (g), distal tubules (d), collecting duct (c), and with catalase as marker for peroxisomes. (B) Immunofluorescence staining of a RCC TMA for catalase. Representative examples of ccRCC biopsies with varying abundance of peroxisomes are shown. (C) Immunohistochemical staining of a RCC TMA for HIF-2a. Shown are representative images of biopsies that were classified as HIF-2a negative, weak, and strong and correspond to biopsies that were deemed to contain many, medium, and few peroxisomes, respectively. (D) Correlation of peroxisome abundance with HIF-2a staining in ccRCC tumor biopsies in a TMA.

(legend continued on next page) Cell Metabolism 22, 882–897, November 4, 2014 ª2014 Elsevier Inc. 895 Cell Metabolism Hif-2a Promotes Mammalian Pexophagy

EXPERIMENTAL PROCEDURES ney were stained against catalase and marker proteins for proximal tubules and glomeruli, distal tubules, and collecting ducts, according to the standard Mice protocol at the pathology of the University Hospital Zu¨ rich. Albumin-Cre (B6.Cg-Tg(Alb-cre)21Mgn/J), Vhlf/f (C;129S-Vhl < tm1Jae > /J), f/f f/f Hif1a (B6.129-Hif1a < tm3Rsjo > /J), and Hif2a (B6.Epas1 < tm1Mcs > /J) Statistical Analyses f/f mice were purchased from the Jackson Laboratory. Atg7 mice (B6.Cg- Data are expressed as mean ± SD. Statistical significance was evaluated by an Atg7 < tm1Tchi >) were obtained from the RIKEN Bio Resource Center (Ibar- unpaired Student’s t test. aki). Liver-specific inactivation of Vhl, Hif1a, Hif2a, and Atg7 was achieved by mating with Albumin-Cre mice. Cre-negative littermates were used as con- SUPPLEMENTAL INFORMATION trols. WY (50 mg/kg body weight/day) or PBA (1 g/kg/day) was administered by gavage. CQ (10 mg/kg/day) or 3-MA (2 mg/kg/day) was administered by Supplemental Information includes seven figures, three tables, four movies, intraperitoneal injection. All protocols for animal use and experiments were and Supplemental Experimental Procedures and can be found with this article approved by the Veterinary Office of Zurich (Switzerland). online at http://dx.doi.org/10.1016/j.cmet.2014.09.017. Lipid Measurements by Liquid Chromatography-Tandem Mass Spectrometry AUTHOR CONTRIBUTIONS Lipids from livers were extracted with the MTBE method, and data acquisition was performed as described previously (Fauland et al., 2011). W.J.K., K.M.W., and W.K. initiated the project. K.M.W., M.J.S., and W.J.K. de- signed experiments. K.M.W., M.J.S., P.A.G., and W.J.K. carried out experi- Adenovirus ments. K.M.W., M.J.S., W.K., and W.J.K. analyzed and interpreted the data. Adenovirus expressing HIF-2a P405/531A was generated by using the Rapid M.H., M.S.L., and H.-P.E. carried out electron microscopy analysis. A.B.M. Adenovirus Production System (Viraquest). The HA-HIF-2a P405/531A cDNA carried out plasma lipid and bile acid analysis. M.T. and H.C.K. carried out lip- was excised from pcDNA3.0-HA-HIF2a P405A/P531A (Addgene). The virus idomics analysis. T.S. assisted in superresolution microscopy. P.L.F. provided expresses GFP from an independent promoter in addition to HIF-2a P405/ lipid data from Pex2 mice. H.M. provided and analyzed human renal cancer 531A. Virus expressing only GFP served as control. TMAs. W.J.K., M.J.S., K.M.W., and W.K. wrote the manuscript and all authors commented on it. Analysis of Autophagy by Microscopy The GFP-LC3-expressing adenovirus was delivered to P21 Vhlf/f and Vhl/ ACKNOWLEDGMENTS mice by tail-vein injection. After 7 days, mice received CQ (10 mg/kg/day) by intraperitoneal injection for 2 days. Livers were harvested and immunostained Imaging and electron microscopy was performed on instruments of ScopeM, on paraffin sections against GFP. Punctuated GFP-LC3 structures were quan- the Scientific Center for Optical and Electron Microscopy of ETH Zu¨ rich. The tified using the particle analysis tool of Imaris software. Thirty to fifty single cells excellent technical assistance of J. Hehl, M. Gu¨ nthert, and F. Lucas from Sco- were evaluated in three Vhlf/f and Vhl/ mice. peM is gratefully acknowledged. Ad-GFP-LC3 was a generous gift from A.M. Tolkovsky. We would like to thank M. Stoffel for help with mouse work. We Quantitative RT-PCR and Biochemistry Measurements thank J.P. Theurillat and S.M. Mu¨ hl for help with TMA analysis. We thank H. Analyses were performed as described in Kovacs et al. (2012). Antibodies are Stangl for helpful discussions. This work was supported by the Swiss National listed in Table S3. Science Foundation (SNF) grant 31003A_132982 to W.J.K. and by grants from the Gebert Ruef Foundation and the Swiss National Science Foundation Histology and Immunohistochemistry (310030B_138669) to W.K. Experiments were performed as described in Kovacs et al. (2012). Antibodies are listed in Table S3. Images were taken with a Leica SP2-AOBS confocal Received: July 30, 2014 laser scanning microscope. Fluorescent dyes were imaged sequentially in Revised: August 26, 2014 frame interlace mode to eliminate crosstalk between the channels. For 3D visu- Accepted: September 24, 2014 alization (movie generation, orthogonal slicer tool), Imaris software was used. Published: November 4, 2014 3D-SIM experiments were realized on a DeltaVision OMX V4 Blaze (GE Healthcare). 3D-SIM data were acquired with a fixed z distance of 125 nm, REFERENCES 15 images per z stack (three different illumination angles at 60 interval and five different phases/angle). The superresolution stack was reconstructed in Baes, M., and Van Veldhoven, P.P. (2012). Mouse models for peroxisome softWoRx 6.0 (GE Healthcare) with channel-specific measured optical transfer biogenesis defects and b-oxidation enzyme deficiencies. Biochim. Biophys. functions. As comparison, wide-field images were acquired (same z range and Acta 1822, 1489–1500. z distance for 3D-SIM dates) and deconvolved with softWoRx 6.0. Image Currie, E., Schulze, A., Zechner, R., Walther, T.C., and Farese, R.V., Jr. (2013). registration for the different acquired channels was performed in softWoRx Cellular fatty acid metabolism and cancer. Cell Metab. 18, 153–161. 6.0 with alignment parameters obtained from a TetraSpeck bead preparate Deosaran, E., Larsen, K.B., Hua, R., Sargent, G., Wang, Y., Kim, S., Lamark, T., (bead diameter 0.2 mm). Reconstructed and registered images were 3D Jauregui, M., Law, K., Lippincott-Schwartz, J., et al. (2013). NBR1 acts as an rendered in Imaris. From the fluorescence information, surfaces were created autophagy receptor for peroxisomes. J. Cell Sci. 126, 939–952. (surface area detail level: 0.08 mm). Farre´ , J.-C., Manjithaya, R., Mathewson, R.D.,and Subramani,S. (2008). PpAtg30 Renal Tumor TMA tags peroxisomes for turnover by selective autophagy. Dev. Cell 14, 365–376. The construction of a renal tumor TMA and the tumors included have been Fauland, A., Ko¨ feler, H., Tro¨ tzmu¨ ller, M., Knopf, A., Hartler, J., Eberl, A., described previously (Mertz et al., 2008). Sequential sections of a healthy kid- Chitraju, C., Lankmayr, E., and Spener, F. (2011). A comprehensive method

(E) Correlation of peroxisome abundance with histological differentiation grade in ccRCC. (F) Model for Hif-2a-mediated decrease in peroxisome abundance. Hypoxia or loss of pVhl results in stabilization of Hif-a. Hif-2a signaling promotes pexophagy, whereas it does not affect basal peroxisome biogenesis. Consequently, the reduced number of peroxisomes and hence decreased peroxisomal metabolic activity leads to significant changes in the lipid profile, such as accumulation of VLCFAs and VLC-PUFAs. These lipids are potent activators of Ppara, a nuclear receptor that induces peroxisome proliferation. Hif-2a signaling represses this ligand-dependent activation of Ppara and by this means prevents Ppara-mediated peroxisome proliferation and consequential restoration of peroxisome homeostasis. In summary, Hif-2a is a negative regulator of peroxisome abundance and metabolic activity. P, peroxisome.

896 Cell Metabolism 22, 882–897, November 4, 2014 ª2014 Elsevier Inc. Cell Metabolism Hif-2a Promotes Mammalian Pexophagy

for lipid profiling by liquid chromatography-ion cyclotron resonance mass Nath, B., and Szabo, G. (2012). Hypoxia and hypoxia inducible factors: diverse spectrometry. J. Lipid Res. 52, 2314–2322. roles in liver diseases. Hepatology 55, 622–633. Faust, P.L., and Kovacs, W.J. (2014). Cholesterol biosynthesis and ER stress in Nicot, A.-S., Lo Verso, F., Ratti, F., Pilot-Storck, F., Streichenberger, N., peroxisome deficiency. Biochimie 98, 75–85. Sandri, M., Schaeffer, L., and Goillot, E. (2014). Phosphorylation of NBR1 by Ferdinandusse, S., Denis, S., Faust, P.L., and Wanders, R.J.A. (2009). Bile GSK3 modulates protein aggregation. Autophagy 10, 1036–1053. acids: the role of peroxisomes. J. Lipid Res. 50, 2139–2147. Obara, K., Kojima, R., and Kihara, A. (2013). Effects on vesicular transport Franovic, A., Holterman, C.E., Payette, J., and Lee, S. (2009). Human cancers pathways at the late in cells with limited very long-chain fatty acids. converge at the HIF-2a oncogenic axis. Proc. Natl. Acad. Sci. USA 106, J. Lipid Res. 54, 831–842. 21306–21311. Petiot, A., Ogier-Denis, E., Blommaart, E.F.C., Meijer, A.J., and Codogno, P. 0 Fransen, M., Nordgren, M., Wang, B., and Apanasets, O. (2012). Role of per- (2000). Distinct classes of phosphatidylinositol 3 -kinases are involved in oxisomes in ROS/RNS-metabolism: implications for human disease. signaling pathways that control macroautophagy in HT-29 cells. J. Biol. Biochim. Biophys. Acta 1822, 1363–1373. Chem. 275, 992–998. Puri, P., Baillie, R.A., Wiest, M.M., Mirshahi, F., Choudhury, J., Cheung, O., Frederiks, W.M., Bosch, K.S., Hoeben, K.A., van Marle, J., and Langbein, S. Sargeant, C., Contos, M.J., and Sanyal, A.J. (2007). A lipidomic analysis of (2010). Renal cell carcinoma and oxidative stress: The lack of peroxisomes. nonalcoholic fatty liver disease. Hepatology 46, 1081–1090. Acta Histochem. 112, 364–371. Puri, P., Wiest, M.M., Cheung, O., Mirshahi, F., Sargeant, C., Min, H.-K., Hakimi, A.A., Pham, C.G., and Hsieh, J.J. (2013). A clear picture of renal cell Contos, M.J., Sterling, R.K., Fuchs, M., Zhou, H., et al. (2009). The plasma lip- carcinoma. Nat. Genet. 45, 849–850. idomic signature of nonalcoholic steatohepatitis. Hepatology 50, 1827–1838. Hiebler, S., Masuda, T., Hacia, J.G., Moser, A.B., Faust, P.L., Liu, A., Qing, G., and Simon, M.C. (2009). Hypoxia inducible factor-2a: a critical medi- Chowdhury, N., Huang, N., Lauer, A., Bennett, J., et al. (2014). The Pex1- ator of aggressive tumor phenotypes. Curr. Opin. Genet. Dev. 19, 60–66. G844D mouse: a model for mild human Zellweger spectrum disorder. Mol. Genet. Metab. 111, 522–532. Qu, A., Taylor, M., Xue, X., Matsubara, T., Metzger, D., Chambon, P., Gonzalez, F.J., and Shah, Y.M. (2011). Hypoxia-inducible transcription factor Keane, M.H., Overmars, H., Wikander, T.M., Ferdinandusse, S., Duran, M., 2a promotes steatohepatitis through augmenting lipid accumulation, inflam- Wanders, R.J.A., and Faust, P.L. (2007). Bile acid treatment alters hepatic dis- mation, and fibrosis. Hepatology 54, 472–483. ease and bile acid transport in peroxisome-deficient PEX2 Zellweger mice. Hepatology 45, 982–997. Rankin, E.B., Rha, J., Selak, M.A., Unger, T.L., Keith, B., Liu, Q., and Haase, V.H. (2009). Hypoxia-inducible factor 2 regulates hepatic lipid metabolism. Keith, B., Johnson, R.S., and Simon, M.C. (2012). HIF1a and HIF2a: sibling ri- Mol. Cell. Biol. 29, 4527–4538. valry in hypoxic tumour growth and progression. Nat. Rev. Cancer 12, 9–22. Raymond, G.V., Watkins, P., Steinberg, S., and Powers, J. (2009). Peroxisomal Kim, W.Y., Safran, M., Buckley, M.R.M., Ebert, B.L., Glickman, J., Bosenberg, disorders. In Handbook of Neurochemistry and Molecular Neurobiology: M., Regan, M., and Kaelin, W.G., Jr. (2006). Failure to prolyl hydroxylate hyp- Neural Lipids, A. Lajtha, G. Tettamanti, and G. Goracci, eds. (Berlin: oxia-inducible factor alpha phenocopies VHL inactivation in vivo. EMBO J. 25, Springer Science+Business Media), pp. 631–670. 4650–4662. Rogov, V., Do¨ tsch, V., Johansen, T., and Kirkin, V. (2014). Interactions between Kim, P.K., Hailey, D.W., Mullen, R.T., and Lippincott-Schwartz, J. (2008). autophagy receptors and ubiquitin-like proteins form the molecular basis for Ubiquitin signals autophagic degradation of cytosolic proteins and peroxi- selective autophagy. Mol. Cell 53, 167–178. somes. Proc. Natl. Acad. Sci. USA 105, 20567–20574. Santos, C.R., and Schulze, A. (2012). Lipid metabolism in cancer. FEBS J. 279, Kovacs, W.J., Charles, K.N., Walter, K.M., Shackelford, J.E., Wikander, T.M., 2610–2623. Richards, M.J., Fliesler, S.J., Krisans, S.K., and Faust, P.L. (2012). Peroxisome deficiency-induced ER stress and SREBP-2 pathway activation in the liver of Schrader, M., Bonekamp, N.A., and Islinger, M. (2012). Fission and prolifera- newborn PEX2 knock-out mice. Biochim. Biophys. Acta 1821, 895–907. tion of peroxisomes. Biochim. Biophys. Acta 1822, 1343–1357. Krishnan, J., Suter, M., Windak, R., Krebs, T., Felley, A., Montessuit, C., Schreiber, A., and Peter, M. (2014). Substrate recognition in selective auto- Tokarska-Schlattner, M., Aasum, E., Bogdanova, A., Perriard, E., et al. phagy and the ubiquitin- system. Biochim. Biophys. Acta 1843, (2009). Activation of a HIF1a-PPARgamma axis underlies the integration of 163–181. glycolytic and lipid anabolic pathways in pathologic cardiac hypertrophy. Seglen, P.O., and Gordon, P.B. (1982). 3-Methyladenine: specific inhibitor of Cell Metab. 9, 512–524. autophagic/lysosomal protein degradation in isolated rat hepatocytes. Proc. Kucejova, B., Sunny, N.E., Nguyen, A.D., Hallac, R., Fu, X., Pen˜ a-Llopis, S., Natl. Acad. Sci. USA 79, 1889–1892. Mason, R.P., Deberardinis, R.J., Xie, X.J., Debose-Boyd, R., et al. (2011). Semenza, G.L. (2011). Hypoxia-inducible factor 1: regulator of mitochondrial Uncoupling hypoxia signaling from oxygen sensing in the liver results in hypo- metabolism and mediator of ischemic preconditioning. Biochim. Biophys. ketotic hypoglycemic death. Oncogene 30, 2147–2160. Acta 1813, 1263–1268. Liu, L., Feng, D., Chen, G., Chen, M., Zheng, Q., Song, P., Ma, Q., Zhu, C., Wang, Semenza, G.L. (2012). Hypoxia-inducible factors in physiology and medicine. R., Qi, W., et al. (2012). Mitochondrial outer-membrane protein FUNDC1 medi- Cell 148, 399–408. ates hypoxia-induced mitophagy in mammalian cells. Nat. Cell Biol. 14, 177–185. Smith, J.J., and Aitchison, J.D. (2013). Peroxisomes take shape. Nat. Rev. Mol. Lodhi, I.J., and Semenkovich, C.F. (2014). Peroxisomes: a nexus for lipid meta- Cell Biol. 14, 803–817. bolism and cellular signaling. Cell Metab. 19, 380–392. Till, A., Lakhani, R., Burnett, S.F., and Subramani, S. (2012). Pexophagy: the Mertz, K.D., Demichelis, F., Sboner, A., Hirsch, M.S., Dal Cin, P., Struckmann, selective degradation of peroxisomes. Int. J. Cell Biol. http://dx.doi.org/10. K., Storz, M., Scherrer, S., Schmid, D.M., Strebel, R.T., et al. (2008). 1155/2012/512721. Association of cytokeratin 7 and 19 expression with genomic stability and Uniacke, J., Holterman, C.E., Lachance, G., Franovic, A., Jacob, M.D., Fabian, favorable prognosis in clear cell renal cell cancer. Int. J. Cancer 123, 569–576. M.R., Payette, J., Holcik, M., Pause, A., and Lee, S. (2012). An oxygen-regu- Mizushima, N., Yoshimori, T., and Ohsumi, Y. (2011). The role of Atg proteins in lated switch in the protein synthesis machinery. Nature 486, 126–129. autophagosome formation. Annu. Rev. Cell Dev. Biol. 27, 107–132. Van Veldhoven, P.P. (2010). Biochemistry and genetics of inherited disorders Motley, A.M., Nuttall, J.M., and Hettema, E.H. (2012). Pex3-anchored Atg36 of peroxisomal fatty acid metabolism. J. Lipid Res. 51, 2863–2895. tags peroxisomes for degradation in . EMBO J. Zeng, X., and Carlin, C.R. (2013). Host cell autophagy modulates early stages 31, 2852–2868. of adenovirus infections in airway epithelial cells. J. Virol. 87, 2307–2319.

Cell Metabolism 22, 882–897, November 4, 2014 ª2014 Elsevier Inc. 897