The Genomic Landscape of Renal Oncocytoma Identifies a Metabolic Barrier to Tumorigenesis Article The Genomic Landscape o f Renal Oncocytoma Identifies a Metabolic Barrier to Tumorigenesis Graphical Abstract Highlights d Type 2 renal oncocytoma may precede eosinophilic ChRCC d Mitochondrial defects disrupt Golgi organization and trafficking d Mitochondrial accumulation in oncocytomas is due to defective elimination d Impaired mitochondrial function is a barrier to tumorigenesis Joshi et al., 2015, Cell Reports 13, 1895–1908 December 1, 2015 ª2015 The Authors http://dx.doi.org/10.1016/j.celrep.2015.10.059 Authors Shilpy Joshi, Denis Tolkunov, Hana Aviv, ..., Shridar Ganesan, Chang S. Chan, Eileen White Correspondence chanc3@cinj.rutgers.edu (C.S.C.), epwhite@cinj.rutgers.edu (E.W.) In Brief Mechanisms that restrict tumors to benign disease inform approaches to cancer therapy. Joshi et al. report that genetic defects in mitochondrial respiration in benign oncocytomas block trafficking and activate p53, limiting tumor growth to benign disease. mailto:chanc3@cinj.rutgers.edu mailto:epwhite@cinj.rutgers.edu http://dx.doi.org/10.1016/j.celrep.2015.10.059 http://crossmark.crossref.org/dialog/?doi=10.1016/j.celrep.2015.10.059&domain=pdf Cell Reports Article The Genomic Landscape of Renal Oncocytoma Identifies a Metabolic Barrier to Tumorigenesis Shilpy Joshi,1,7 Denis Tolkunov,1,7 Hana Aviv,2 Abraham A. Hakimi,3 Ming Yao,1 James J. Hsieh,4 Shridar Ganesan,1,5 Chang S. Chan,1,5,* and Eileen White1,6,* 1Rutgers Cancer Institute of New Jersey (CINJ), 195 Little Albany Street, New Brunswick, NJ 08903, USA 2Department of Pathology and Laboratory Medicine, Rutgers Robert Wood Johnson Medical School, One Robert Wood Johnson Place, MEB 212, New Brunswick, NJ 08901, USA 3Urology Service, Department of Surgery, Memorial Sloan Kettering Cancer Center, 1275 York Avenue, New York, NY 10065, USA 4Human Oncology and Pathogenesis Program, Memorial Sloan Kettering Cancer Center, 1275 York Avenue, New York, NY 10065, USA 5Department of Medicine, Robert Wood Johnson Medical School, Rutgers, the State University of New Jersey, 1 Robert Wood Johnson Place, New Brunswick, NJ 08901, USA 6Department of Molecular Biology and Biochemistry, Rutgers University, 604 Allison Road, Piscataway, NJ 08854, USA 7Co-first author *Correspondence: chanc3@cinj.rutgers.edu (C.S.C.), epwhite@cinj.rutgers.edu (E.W.) http://dx.doi.org/10.1016/j.celrep.2015.10.059 This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). SUMMARY Oncocytomas are predominantly benign neoplasms possessing pathogenic mitochondrial mutations and accumulation of respiration-defective mitochon- dria, characteristics of unknown significance. Using exome and transcriptome sequencing, we identified two main subtypes of renal oncocytoma. Type 1 is diploid with CCND1 rearrangements, whereas type 2 is aneuploid with recurrent loss of chromosome 1, X or Y, and/or 14 and 21, which may proceed to more aggressive eosinophilic chromophobe renal cell carcinoma (ChRCC). Oncocytomas activate 50 adenosine monophosphate-activated protein kinase (AMPK) and Tp53 (p53) and display disruption of Golgi and autophagy/lysosome trafficking, events attributed to defective mitochondrial function. This suggests that the genetic defects in mitochondria activate a metabolic checkpoint, producing auto- phagy impairment and mitochondrial accumulation that limit tumor progression, revealing a novel tu- mor-suppressive mechanism for mitochondrial inhi- bition with metformin. Alleviation of this metabolic checkpoint in type 2 by p53 mutations may allow pro- gression to eosinophilic ChRCC, indicating that they represent higher risk. INTRODUCTION Altered metabolism in cancer that promotes the fermentation of glucose was proposed more than 50 years ago by Otto Warburg to originate from respiratory system damage (Warburg, 1956). Only recently with the discovery of cancer driver mutations in nuclear genes encoding mitochondrial tricarboxylic acid (TCA) Cell Re cycle enzymes have we begun to unravel the underlying onco- genic functions linked to altered metabolism (Ward and Thomp- son, 2012). Moreover, it is now clear that many oncogenic events common in cancer indirectly regulate cellular metabolism (Van- der Heiden et al., 2009). These metabolic alterations may provide substrates and energy necessary for cell growth, promote anti- oxidant defense, and produce signals that alter gene expression and drive pro-oncogenic signaling pathways. These discoveries are revealing new approaches to cancer therapy that target metabolic functions important for tumorigenesis. One underex- plored area within the sphere of metabolic alterations in cancer is represented by the tumor type oncocytoma. Oncocytomas are rare, predominantly benign neoplasms mostly of epithelia that have inactivating mutations in mitochon- drial genome-encoded enzymes or control regions that cause respiration defects. Oncocytomas are also characterized by the dramatic accumulation of these defective mitochondria (Gasparre et al., 2011). Whether the mitochondrial impairments are functionally neutral, promote tumor growth as Warburg envi- sioned, or to the contrary are a liability is unknown. It is also un- clear why defective mitochondria accumulate and if this plays any role in disease. Oncocytomas may represent an evolutionary dead end or an intermediate that progresses to more aggressive cancer. Identifying mechanisms that restrict some tumors such as these to benign disease can inform novel approaches to can- cer therapy. Mitochondria are essential for eukaryotic cell function. Cells contain hundreds of mitochondria, with their numbers controlled by both a transcription program of biogenesis (Wallace, 2012) and removal of defective mitochondria through mitophagy, a selective form of autophagy (Youle and Narendra, 2011). Thus, the presence of tumor cells with the accumulation of defective mitochondria suggests increased biogenesis, possibly as a compensatory mechanism (Simonnet et al., 2003), or an under- lying defect in removal by mitophagy (Guo et al., 2013b). As autophagy deficiency in genetically engineered mouse models for KrasG12D- and BRAFV600E-driven lung cancers converts ports 13, 1895–1908, December 1, 2015 ª2015 The Authors 1895 mailto:chanc3@cinj.rutgers.edu mailto:epwhite@cinj.rutgers.edu http://dx.doi.org/10.1016/j.celrep.2015.10.059 http://creativecommons.org/licenses/by-nc-nd/4.0/ http://crossmark.crossref.org/dialog/?doi=10.1016/j.celrep.2015.10.059&domain=pdf Figure 1. Identification of Two Main Subtypes of Renal Oncocytoma (A) Copy-number variations in renal oncocytomas in comparison to eosinophilic ChRCC (eosinophilic ChRCC sequencing data were downloaded from the TCGA CGHub repository [https://cghub.ucsc.edu/]) (Davis et al., 2014). Blue corresponds to loss of one copy, red corresponds to a gain, and gray on chromosome Y indicates a female. (B) Somatic mutations in renal oncocytomas in comparison to ChRCC (ChRCC sequencing data were downloaded from the TCGA CGHub repository [https:// cghub.ucsc.edu/]) (Davis et al., 2014). Only SNVs common between oncocytoma and ChRCC are shown; the complete list of detected somatic SNVs and indels is reported in Tables S2 and S3. (C) Expression levels of key genes from the transcriptome analysis. Each bar represents log2-transformed transcript abundance. Error bars show SEs. (D) Principal component analysis (PCA) of gene expression identifies two subtypes of oncocytoma. The first principal component (PC1) separates tumor and normal samples; the second principal component (PC2) separates oncocytomas into two subtypes. (legend continued on next page) 1896 Cell Reports 13, 1895–1908, December 1, 2015 ª2015 The Authors https://cghub.ucsc.edu/ https://cghub.ucsc.edu/ https://cghub.ucsc.edu/ carcinomas to more benign oncocytomas, defects in mitochon- drial removal may underlie the oncocytoma phenotype and pro- duce benign instead of malignant disease (Guo et al., 2013a; Karsli-Uzunbas et al., 2014; Strohecker et al., 2013). This also suggests that autophagy and mitochondrial quality control are important for promoting tumorigenesis (Guo et al., 2013b). To address the cause and role of oncocytomas in the spec- trum of cancer, we sequenced the exome and transcriptome of the most common oncocytoma, those that arise in the kidney (Linehan and Ricketts, 2013). We identified two major subtypes, one without (type 1) and the other with (type 2) recurrent whole- chromosome losses that contain few other mutations. Both subtypes had recurrent inactivating mutations in nuclear- and mitochondria-encoded proteins critical for respiration and dis- played activation of AMPK and p53. Recurrent CCND1 chromo- somal rearrangement and overexpression was a likely cancer driver mutation in type 1. Both subtypes had evidence of defec- tive autophagy attributed to metabolic-deficiency-induced Golgi disassembly and lysosome dysfunction that blocked LAMP-2 trafficking and lysosomal protease activation. Pharmacologic inhibition of mitochondrial complex I with metformin to repli- cate mitochondrial dysfunction in oncocytomas caused AMPK activation and Golgi disassembly and blocked LAMP-2 traf- ficking and autophagy. Thus, mitochondrial respiration defects that arise early in renal tumorigenesis trigger a metabolic check- point, trafficking and lysosome defects, and failure of mitochon- drial quality control, activating p53 and AMPK as a barrier to tumor progression. Moreover, the similarities in the mutational landscape and transcriptome suggest that type 2 oncocytoma and eosinophilic subtype of chromophobe renal cell carcinoma (ChRCC) are closely related, with ChRCC having acquired addi- tional driver mutations in p53 and PTEN and further genetic instability. Thus, treatment of type 2 oncocytomas requires more consideration. Genomic Analysis Indicates Two Main Subtypes of Renal Oncocytoma Whole-exome sequencing of 12 renal oncocytomas and matched adjacent normal tissue and RNA sequencing of 9 of these pairs were performed. Copy-number variation (CNV) analysis from exome sequencing indicates that tumors are clonal, with two distinct subtypes based on the pattern of aneuploidy. One tumor was an outlier that was excluded from further analysis. Type 1 is diploid and type 2 is hypodiploid with loss of chromosome 1, X or Y, and/or 14 and 21 (Figures 1A and S1A; Table S1A). Type 1 shows no sex bias, but type 2 has male predominance (five male/one female), suggesting strong selection for loss of the Y chromosome. The one type 2 tumor arising in a female has lost one X chromosome (Figure 1A). Strong selection for loss of Y in males or X in females suggests possible involvement of gene loss on the pseudoautosomal region (PAR). Few and no recurrent single-nucleotide variants (SNVs) (162 nonsynonymous SNVs in 11 tumors; Tables S1B and S2) or in- (E) Type 1 and type 2 oncocytomas have distinct histology, with type 2 having a la (F) Gene expression similarities between oncocytoma and ChRCC. PCA of gene identify the three RCC subtypes as separate clusters with the corresponding ma demonstrating similarities in gene expression patterns. Cell Re sertions/deletions were detected (Table S3). There were no mu- tations in common cancer-causing genes (p53, PTEN) with the exception of one tumor with an MLL3 mutation (Figure 1B). Three tumors have mutations in nuclear-encoded mitochondrial genes NDUFV1, UQCRC2 (type 1), and SLC25A5 (ANT2) (type 2). ANT2 is located on the X chromosome, and this ANT2 mutant tumor likely has a loss-of-function mutation as it has a frameshift muta- tion with X chromosome loss (Figure 1B; Table S3). Mutations in FLCN, VHL, FH, SDHB, SDHC, or SDHD, known drivers in other renal cell carcinomas (RCCs), were not detected. Despite loss of chromosome 1 in type 2, expression levels of FH, SDHB, and SDHC located on chromosome 1 were not affected (Figure 1C). The nuclear genome mutation spectrum observed in onco- cytomas was C > A (132), C > G (98), C > T (329), T > A (64), T > C (147), and T > G (97). The analysis showed transition to transversion ratio (Ts/Tv) of 1.22 in somatic mutations in contrast to 2.33 for germline mutations (found in dbSNP) consistent with expected values (Ebersberger et al., 2002). The comparison displayed strong enrichment for transversion mutations among somatic mutations, suggesting reactive oxygen species (ROS) exposure at some point in oncocytoma evolution. Oncocytomas have mitochondrial genome mutations respon- sible for respiration defects (Gasparre et al., 2007; Mayr et al., 2008; Simonnet et al., 2003), with all oncocytomas examined here having mutations, and mutations in MT-COX1, MT-COX2, MT-COX3, MT-ND4, MT-ND5, and MT-CYTB found in three or more tumors (Figures 1B and S1B; Table S4). Thus, the distin- guishing feature of both subtypes of renal oncocytoma is recur- rent inactivating mutations in mitochondrial genes encoded by both the nuclear or mitochondrial genomes. The mitochondrial point mutation spectrum for oncocytomas was C > T (46) and A > G (10) on the heavy strand and C > T (10) and A > G (3) on the light strand. The only other point muta- tions detected were C > A (7) on the light strand. The dominance and strand bias of C > T and A > G mutations on the heavy strand are consistent with findings from other human cancers and postulated to result from mitochondrial DNA replication errors (Ju et al., 2014). In contrast, the C > A mutations are consistent with damage by ROS (Thilly, 2003), which were elevated (9%) compared to other cancers (4%), (Ju et al., 2014). While the majority of oncocytoma mitochondrial mutations showed the signature of DNA replication error indicating that DNA replication errors are the dominant source of mutations, most of the patho- genic mutations with high allelic frequencies (>75%) are indels and occure in the complex I subunits. In summary, 7 out of 11 on- cocytomas have a pathogenic mitochondrial DNA mutation fre- quency greater than 75% (Table S4). The gene expression pattern also represents type 1 and 2 oncocytomas as distinct groups and indicates that the matched normal adjacent tissue (NAT) from sample 5 is likely contami- nated with tumor (Figures S1C and S1D, A). Principle component analysis (PCA) shows type 1 and 2 grouping separately with type 2 forming a tighter cluster showing greater similarity (Figure 1D). rger volume of eosinophilic cytoplasm relative to the nucleus. Error bar, 50 mm. expression for oncocytoma, ChRCC, clear cell RCC, and papillary cell RCC tched normal samples forming one cluster. Oncocytomas cluster with ChRCC ports 13, 1895–1908, December 1, 2015 ª2015 The Authors 1897 Figure 2. Chromosome Rearrangements and Overexpression of CCND1 in Type 1 and p53 Activation in Type 1 and 2 Oncocytomas (A) Overexpression of a single allele of CCND1. The RNA-sequencing coverage tracks for four tumors (T2, T6, T7, and T9) and matched normal samples (N2, N6, N7, and N9) are shown in gray. The numbers in brackets display the coverage for this region and reflect the level of the gene expression, showing that CCND1 is over 10-fold higher in tumors. SNPs present in the exons of CCND1 are shown as colored bars. Some of the SNPs are heterozygous and are indicated by the two- color bars showing the relative expression of the two alleles. The appearance of homozygous SNPs in tumor where the same SNPs appear heterozygous in the matched normal sample, as in T2 versus N2, is a result of the overexpression of a single allele of CCND1. (B) Western blot showing elevated CCND1 protein expression in tumors. The asterisks marks sample 5, which has normal (N) tissue mixed with oncocytoma (O) tissue. Multiple bands represent different phosphorylated forms of CCND1. (C) Representative fluorescence images of CCND1 FISH show rearrangement in type 1 (sample 2), but not type 2 (sample 3), tumors. White arrowheads show segregation of CCND1 probe. (D) Karyotype of renal oncocytomas shows chromosomal rearrangement including translocations, inversions, and deletions near CCND1 in type 1 and loss of chromosome 1 and Y in type 2. (E) Summary of genetic alterations in all 39 oncocytoma samples. (legend continued on next page) 1898 Cell Reports 13, 1895–1908, December 1, 2015 ª2015 The Authors The separate groups persist using only gene expression from diploid chromosomes from types 1 and 2 (Figure S1D, B and C). Both types are distinct from normal controls (Figures 1D and S1C). Type 1 and 2 are distinguished histologically, with type 2 having a more expansive eosinophilic cytoplasm (Figures 1E and S2), suggesting greater accumulation of defective mito- chondria, thought to be a main contributing factor to the histo- logical appearance of these tumors. Type 2 Oncocytoma May Be the Precursor of Eosinophilic ChRCC The relationship between oncocytomas and other renal tumors is not currently known. Thus, we compared the genomic and transcriptomic signature of oncocytomas with RCC subtypes. The pattern of gene expression in oncocytoma was clearly distinct from clear cell and papillary cell RCC (Figure 1F), consis- tent with the vastly different mutational landscape of these tumors (Gao et al., 2013; Varela et al., 2011). In contrast, ChRCC has some histologic similarities and common biomarkers with oncocytoma, and distinguishing these tumor types is important, as ChRCC is a more aggressive disease with higher risk of metastasis (Ng et al., 2014). There are two types of ChRCC, classic and eosinophilic, with the eosinophilic subtype possessing some similarities with onco- cytoma (Davis et al., 2014). The global gene expression pattern of both oncocytoma and ChRCC are similar (Figure 1F), and both have a gene expression signature of distal tubule, suggest- ing that they originate from a common cell type (Figure S1E). Eosinophilic ChRCC has two subgroups: one is chromosomally diploid like type 1 but without rearrangement or overexpression of CCND1, and the other has a pattern of chromosome loss similar to type 2 with loss of chromosomes 1, X, or Y, but with further losses including 2, 6, 10, 13, and 17 (Figures 1A and 1B) (Davis et al., 2014). Like type 2, eosinophilic ChRCC also has a sex bias (15 male/4 female) whereas classic ChRCC has no gender bias (24 male/23 female) (Figure 1A; Table S5). The main nuclear genomic difference between oncocytomas and ChRCC are mutations in p53 and PTEN in ChRCC (Figure 1B). TERT promoter mutations and upregulation of telomerase reverse transcriptase (TERT) expression occur in 10% of ChRCC (Davis et al., 2014), and both oncocytoma subtypes lacked TERT expression (Figure 1C), indicating the absence of this driver mutation. The similarities in gene expression, aneuploidy pattern, and histology suggest type 2 may be closely related to eosinophilic ChRCC. One possibility is that ChRCC may have acquired additional mutations in TP53, PTEN, or TERT, leading to further genomic instability and a less indolent natural history. Alternatively, there could be a single cell type precursor for both type 2 oncocytoma and ChRCC, where chromosome 1 is lost followed by either pathogenic mitochondrial genome or p53 mutations, directing fate to either oncocytoma or ChRCC. This, however, does not account for the common mitochondrial (F) Oncocytomas have a gene expression signature similar to ChRCC without p53 gene expression values. The correlation is calculated for the set of genes (Table mutation in p53. (G) Western blot for p53 showing induction in oncocytomas with quantitation rel Cell Re mutations in type 1 and 2 oncocytoma without chromosome 1 loss in type 1. In contrast to the increased nuclear genomic mutation burden in ChRCC compared to oncocytoma, there was less mutational burden of the mitochondrial genome in ChRCC compared to on- cocytoma (p < 0.0001) (Figure 1B) (Davis et al., 2014). Hetero- plasmy may be a prerequisite to selection for a less defective pool of mitochondria during progression of benign type 2 to a more malignant ChRCC. Recurrent CCND1 Rearrangement and Amplification in Type 1 Oncocytomas Four of the five type 1 oncocytomas had high expression of CCND1 in comparison to type 2 and NAT (Figure 1C). CCND1 encodes the cell-cycle regulator and oncoprotein CyclinD1. All three tumors exhibit monoallelic expression of CCND1 consis- tent with chromosomal rearrangement that led to aberrant high expression of one CCND1 allele (Figure 2A) consistent with pre- vious observations (Sukov et al., 2009). Western blot shows high CyclinD1 protein expression in some tumors with high mRNA expression (Figure 2B). CyclinD1 expression is also elevated in some of the type 2 tumors, reflecting non-transcriptional mech- anisms for CyclinD1 activation (Figure 2B). Fluorescence in situ hybridization (FISH) analysis of three each of type 1 and 2 oncocytomas confirmed translocation of 11q13 at the CCND1 locus in type 1 and not in type 2 (Figure 2C). Cyto- genetic analysis of 28 additional renal oncocytomas showed eight tumors with translocations, inversions, or deletions near CCND1 (11q13) but an otherwise chromosomally intact genome (Figure 2D). These findings are consistent with the presence of two mutually exclusive categories of oncocytoma: type 1 with CCND1 rearrangement and no other gross chromosomal losses, and type 2 with chromosome 1 loss. Of the 28 oncocytomas, two were chromosomally stable, harboring a translocation between chromosomes 6 and 9 of unknown significance (Figure 2D) (Hudacko et al., 2011). Five others had unique chromosome re- arrangements and CNVs, suggesting low-level diversity of this tumor type (Figures 2D and 2E). The expanded characterization of 39 oncocytomas confirmed the association of type 1 with CCND1 rearrangements (12/13) and type 2 with chromosome 1 loss (19/19), sex chromosome loss (11/19), and male bias (12/19 male, 7/19 female) (Figure 2E). No aberrant high expres- sion of CCND1 was detected in chromosomally stable ChRCC (six tumors) so there is no evidence that type 1 progresses to a subset of ChRCC that also lacks chromosomal losses. Since translocation or rearrangement of CCND1 and overexpression of CyclinD1 is a well-known oncogenic driver, it is likely a driver mutation in type 1 oncocytomas. p53 Is a Barrier to Oncocytoma Progression p53 is mutated in 22 out of 66 (33%) ChRCC tumors (Davis et al., 2014) and not in oncocytoma (Figure 1B). We found a set of mutation as shown by correlation heat map between oncocytoma and ChRCC S6), which are differentially expressed between ChRCCs with and without a ative to ACTIN. ports 13, 1895–1908, December 1, 2015 ª2015 The Authors 1899 Figure 3. Oncocytomas Show a Gene Expression Signature of Oxidative Metabolism and Mitochondrial and Energetic Defects (A) Oncocytomas have increased mitochondrial DNA copy number. p value was calculated for a one-tailed t test. (B–D) Canonical pathway analysis. Each graph represents the top 15 canonical pathways enriched for the corresponding gene set (http://www.ingenuity.com/). Numbers on top of the bars indicate the number of genes in the pathway. Differential expression analysis of type 2 versus type 1 identifies 50 genes with FDR-corrected p value < 0.05. (B) Pathway analysis of a set of 1,042 genes differentially expressed between tumor and normal samples. The set of genes was selected based on the FDR-corrected p value < 0.05. Red corresponds to genes that are upregulated in oncocytomas, and green corresponds to genes downregulated in oncocytomas. (C) Pathway analysis of a set of 457 genes that are upregulated in oncocytomas compared to normal samples. The set of genes was selected based on the FDR-corrected p value < 0.05. Red arrowhead marks AMPK signaling. (D) Pathway analysis of a set of 500 genes differentially expressed between type 1 and type 2 oncocytoma. The set of genes was selected based on the lowest p values. Red corresponds to genes upregulated in type 2. Red arrowheads mark AMPK and p53 signaling. (legend continued on next page) 1900 Cell Reports 13, 1895–1908, December 1, 2015 ª2015 The Authors http://www.ingenuity.com/ genes differentially expressed between ChRCC with and without p53 mutation (Table S6) and oncocytomas have the gene signa- ture more similar to ChRCC without p53 mutation (Figure 2F). This finding is consistent with changes in gene expression linked to p53 mutational status in both ChRCC and oncocytomas. p53 protein is stabilized as a barrier to tumorigenesis by stress, oncogene activation including CCND1 amplification, and genome damage. To test if p53 was activated in benign on- cocytomas explaining its mutation upon progression to ChRCC in type 2, we examined p53 levels. In comparison to NAT, wild- type p53 protein was induced in most oncocytomas indepen- dent of subtype, suggesting that p53 activation in oncocytomas may contribute to their slow growth and benign characteristics (Figure 2G). Type 1 and 2 Oncocytomas Display Mitochondrial Defects and Metabolic Stress Since metabolic stress, triggered by mitochondrial defects, is a potential cause of p53 activation in oncocytomas, we assessed mitochondrial genome copy number, gene expression signa- ture, mitochondrial complex protein expression, and metabolic signaling. Both type 1 and 2 have elevated mitochondrial genome copy number relative to NAT, with type 2 being greater, suggesting increased mitochondrial accumulation (Figure 3A; Table S7). Pathway analysis of the transcriptome indicated that in com- parison to NAT, oncocytomas have a predominance of downre- gulated genes that are represented in metabolic and degradative pathways (Figure 3B; Tables S8 and S9). Mitochondrial accumu- lation in oncocytomas may be due to compensatory increased biogenesis or decreased clearance by autophagy, concepts that are not mutually exclusive. Indeed, PPARGC1a expression is induced several fold in oncocytomas compared with NAT (Table S9), indicating activation of mitochondrial biogenesis. Pathway analysis of genes upregulated in oncocytomas identi- fied AMPK, creatine-phosphate biosynthesis, and ketolysis pathways suggestive of energy crisis (Figure 3C; Table S9). On- cocytomas have high expression of genes encoding the TCA cy- cle and other mitochondria-expressed gene products involved in respiration (Figure 3C; Table S10). Comparison of differential gene expression between types 1 and 2 identified metabolic and cancer pathways such as AMPK and p53 (Figure 3D; Tables S11 and S12). Both type 1 and type 2 oncocytomas have AMPK pathway upregulated compared to NAT, with upregulation more pronounced in type 1. Type 1 has 3.4-, 6.7-, 2.4-, and 4-fold higher levels of PRKAB2, PRKAG2, ACACA (ACC1), and ACACB (ACC2), whereas type 2 has 8-, 2.2-, and 3.3-fold higher levels of PRKAG2, ACACA, and ACACB, respectively. The mitochon- drial accumulation and genome mutations, oxidative meta- bolism gene expression signature, and patterns of chromosome loss suggest similarities between oncocytomas and eosinophilic (E) Oncocytomas show upregulation of a gene expression signature for mitochon encoded genes as for ChRCC (Davis et al., 2014) for ease of comparison. Ye mutations. Star indicates somatic mutations in nuclear-encoded genes. (F) Deregulation of mitochondrial protein expression in oncocytomas determined (G) Oncocytomas have loss of mTOR as indicated by reduction of P-S6 (Ser 23 (a-subunit, Thr 172) in western blots. Cell Re ChRCC (Figure 3E) (Davis et al., 2014). This may reflect an increased number of mitochondria and not necessarily increased mitochondrial metabolism. Increased mitochondrial genome copy number was reflected in high cytochrome c levels in oncocytomas relative to NAT in all tumors with the exception of sample 5, where normal tissue was contaminated with tumor (Figure 3F). Mitochondrial genome mutations resulting in defects in complex I of the electron trans- port chain (ETC) are common in oncocytomas (Figures 1B and S1B) (Gasparre et al., 2007; Mayr et al., 2008; Simonnet et al., 2003). Low or undetectable levels of the subunit NDUFS1 (com- plex I), UQCRC2 (complex III), and MT-COX2 (complex IV) were observed in a subset of oncocytomas indicative of impairment in mitochondrial respiration at multiple points in the ETC (Figure 3F). Compared to NAT, ANT2, and SLC25A6 (ANT3), protein expres- sion levels were low or undetectable in type 2 in contrast to the increased mitochondrial mass (Figure 3F). ANT3 is on the PAR and had a reduced mRNA expression (log2 fold change = �0.9, uncorrected p = 0.04) associated with loss of the Y or X chromosome and correspondingly reduced protein expression (Figure 3F). Tumor 8 had no detectable ANT2, likely due to a frameshift mutation and X chromosome loss containing the other allele (Figures 1B and 3F). Interestingly, mutations in ANT2 also occur in ChRCC (Figure 1B). ANT2 and ANT3 modulate the mitochondrial permeability transition pore, and their deficiency impairs mitochondrial function (Kokoszka et al., 2004). These findings are consistent with unifying defects in mitochondrial function in oncocytomas. To examine the potential manifestation of mitochondrial metabolic defects on oncocytoma, we assessed the activity of the nutrient sensor and master regulator of cell growth, the mammalian target of rapamycin (mTOR), and the metabolic stress response regulator AMPK. Both subtypes had reduced phospho(P)-S6, indicating loss of mTOR activity, and increased phospho(P)-AMPK (Thr 172) (Figures 3G and S1F), indicating activation of AMPK signaling, consistent with upregulation of the gene expression signature (Figure 3C, arrow). These findings reflect metabolic and growth impairment consistent with defec- tive mitochondrial function in oncocytomas. Oncocytomas Manifest Phenotypes of Defective Autophagy To test if autophagy defects underlie the accumulation of defec- tive mitochondria in oncocytomas, we examined the levels of autophagy substrates that accumulate with autophagy inactiva- tion. Oncocytomas show accumulation of mitochondria that are normally degraded by autophagy as determined by immuno- histochemistry (IHC) and western blotting for the mitochon- drial proteins TOM20 and cytochrome c (Figures 4A and 4B). Levels of the essential autophagy regulator BECLIN1 were reduced (Figures 4B and S2), and puncta of LC3 indicative of drial respiration. Red and blue correspond to up- or downregulation of nuclear- llow and gray correspond to mitochondrial genes with and without somatic by western blots. The asterisk marks sample 5 with mixed normal and tumor. 5/236) and induction of AMPK signaling as indicated by induction of P-AMPK ports 13, 1895–1908, December 1, 2015 ª2015 The Authors 1901 Figure 4. Oncocytomas Lack Autophagosomes, Accumulate Autophagy Substrates and Dysfunctional Mitochondria, and Are Primed for Mitophagy (A) Human oncocytomas accumulate mitochondria (TOM20) and p62 (autophagy substrate) and lack LC3 puncta (autophagosomes) indicative of autophagy inactivation. Representative IHC for each of the two subtypes with all in Figure S2. (B) Western blots of indicated mitochondrial and autophagy proteins in normal tissues and oncocytomas. Quantitation of LC3-I/II and p62 levels is shown below. (C) Oncocytomas have evidence for activation of the fission and the PINK1/PARKIN machinery. Western blots of normal and oncocytoma tissues for indicated proteins. FL, full length; DMTS, without mitochondrial translocation sequence. (legend continued on next page) 1902 Cell Reports 13, 1895–1908, December 1, 2015 ª2015 The Authors autophagosomes were present in NAT and absent in oncocyto- mas (Figure 4A). Consistent with the absence of LC3 puncta and autophagy, oncocytomas had higher levels of the unprocessed form of LC3 (LC3-I) and lower levels of the processed, autopha- gosome-associated form of LC3 (LC3-II) than NAT as indicated by western blotting and LC3I/II ratios (Figure 4B). The absence of LC3-II and autophagosomes and the accumulation of LC3-I is a characteristic of genetic ablation of an essential auto- phagy gene in mice (Karsli-Uzunbas et al., 2014). Oncocytomas also displayed high levels of the autophagy substrate p62 that accumulates in autophagy-defective cells and tissues (Figures 4A and 4B). Mitophagy Is Primed, but Not Executed, in Oncocytomas Since oncocytomas accumulate defective mitochondria, we examined oncocytomas for evidence of mitophagy dysfunc- tion. Mitochondria undergo fission in response to stress to allow their segregation and elimination by mitophagy. Fission is induced when the protease OMA1 cleaves the dynamin- related GTPase OPA1 (McBride and Soubannier, 2010). Mi- tophagy is subsequently activated by loss of mitochondrial membrane potential that prevents cleavage and inactivation of the kinase PTEN-induced putative kinase 1 (PINK1), which recruits the E3 ligase PARKIN to mitochondria (Youle and Narendra, 2011). PARKIN then adds ubiquitin chains to mitochondrial outer membrane proteins to signal recruitment of autophagy adaptors such as p62, which interact with LC3-II in autophagosome membranes, allowing their engulf- ment within autophagosomes. In normal conditions, OMA1 is cleaved and OPA1 is not, allow- ing fusion and preventing fission and mitophagy. This was the case in NAT, where cleaved OMA1 (40 kDa) and less cleaved OPA1 (S-form) were apparent (Figure 4C). In contrast, oncocyto- mas have uncleaved OMA1 (60 kDa) and accumulation of cleaved OPA1 (S-form) consistent with activation of fission. Since LC3-II-positive autophagosomes were not detected in oncocytomas (Figures 4A and 4B), this suggests that the initial steps of mitochondrial fission are initiated but the process of autophagosome sequestration is not. Oncocytomas have un- cleaved PINK1 (FL-PINK1) and accumulation of PARKIN, consis- tent with priming of mitophagy without autophagosome forma- tion (Figure 4C). PINK1 is cleaved by PARL and further cleaved to the p45kDa fragment to inactivate fission and mitophagy when mitochondria are healthy. Importantly, normal tissue but not oncocytomas have the p45kDa cleaved PINK1, consistent with inactivation of PINK1 and priming of mitophagy (Figure 4C). Interestingly, cells with pathogenic mitochondrial DNA mutations are defective for mitochondrial membrane fusion (Mishra et al., 2014), consistent with activation of the fission machinery and the PINK1/PARKIN system shown here in oncocytomas (Fig- ure 4C). Thus, upstream signaling of mitophagy at mitochondria (D) Primary cultures of oncocytoma cells display autophagy substrate accumu oncocytoma cells (O159/O144) accumulate LC3 and p62 aggregates, mitochond (E) Primary cultures of oncocytoma cells (0159/O144) show accumulation of dys icantly lower relative membrane potential than normal cell controls NK159 and N (F) Primary oncocytoma cells (O159) have reduced OCR and increased ECAR wh (N159). Cell Re occurs in oncocytomas, but sequestration is absent, suggesting a block in autophagy. Impaired Mitochondrial Respiration in Primary Oncocytoma Cells Attempts to generate cell lines from renal oncocytomas by others and us have been unsuccessful. To address mitochon- drial function, two pairs of normal adjacent kidney (NK159/ N144) and oncocytoma-derived (O159/O144) primary cell cul- tures were examined. Primary oncocytoma cells remain viable in short-term culture (<1 week) without undergoing cell division, permitting limited analysis. In contrast to NK159 and N144, O159 and O144 primary cultures were composed of large cells with LC3 and p62 aggregates and increased mitochondria, glycogen, and lipids indicative of defective autophagy (Figure 4D). Even though primary oncocytoma cells had severalfold greater mito- chondrial mass, relative mitochondrial membrane potential was significantly lower when compared to normal cells, consis- tent with ETC defects (Figure 4E). One of the primary cultures (O159) provided sufficient material for determination of the oxy- gen consumption rate (OCR) and extracellular acidification rate (ECAR). Compared to normal cells, O159 cells were defective for respiration and more glycolytic, consistent with pathogenic mitochondrial genome mutations (Figures 4F and S1B). Oncocytomas Display LAMP-2 Trafficking, Golgi, and Lysosome Defects Autophagy delivers cargo to lysosomes for degradation, and interference with lysosome function and cargo degradation feeds back to inhibit autophagy (Tanaka et al., 2000). Moreover, muta- tionsinthemajorandessential lysosome transmembranecompo- nent lysosome-associated membrane protein-2 (LAMP-2) cause lysosomal storage disease and interfere with autophagy (Eskeli- nen, 2006). To test if lysosome defects in oncocytomas interfere with autophagy, lysosome numbers and localization were exam- ined by monitoring LAMP-2. LAMP-2-positive vesicle distribution in types 1 and 2 was distinct from NAT (Figure 5A). In contrast, and suggestive of lysosome abnormality, localization of LAMP- 2-positive vesicles was perinuclear in type 1 and peripheral in type 2 (Figure 5A). LAMP-2 is synthesized and heavily polylactosamine glycosy- lated in the Golgi apparatus prior to transit to lysosomes, which serves to protect the protein from proteolysis in the lysosome lumen where most of the polypeptide resides. Oncocytomas have enhanced glycosylation of LAMP-2 compared to NAT as indicated by its higher molecular weight (Figure 5B). Increased glycosylation of LAMP-2 occurs when it is retained in the Golgi, where this post-translational modification takes place (Nabi and Rodriguez-Boulan, 1993). In contrast to perinuclear stacks of Golgi in NAT, oncocytomas displayed disrupted Golgi organiza- tion with LAMP-2 co-localization (Figures 5A and 5C), consistent lation. In contrast to primary cells derived from NAT (NK159/N144), primary ria, glycogen, and lipids, characteristics of autophagy impairment. functional mitochondria with several fold more mitochondrial mass but signif- 144. en provided a fermentable carbon source in comparison to normal control cells ports 13, 1895–1908, December 1, 2015 ª2015 The Authors 1903 Figure 5. Golgi Disassembly, Altered LAMP-2, and Defective Cathepsin Activation in Oncocytomas (A) Golgi disassembly and altered LAMP-2 localization in human oncocyto- mas. Representative IHC for each subtype with all shown in Figure S2. (B) Western blots showing enhanced LAMP-2 glycosylation (increased molecular weight) in human oncocytomas. Oncocytomas have defective CTS-B and -L activation. Low molecular weight, cleaved, activated forms of cathepsins predominate in normal tissue. Pro, proenzyme; Int, intermediate. (C) Golgi disassembly and LAMP-2 co-localization in human oncocy- tomas. Representative fluorescence images of each subtype stained for Golgi (GM130), LAMP-2, and nuclei (DAPI). The fluorescence images of all are shown in Figure S2. 1904 Cell Reports 13, 1895–1908, December 1, 2015 ª2015 The Aut with Golgi disassembly and LAMP-2 retention and enhanced glycosylation (Figures 5B and 5C). The major lysosomal proteases, cathepsins (CTSs), are also processed in the Golgi and cleaved and activated to mature forms upon reaching the acidic environment of lysosomes (Ishi- doh and Kominami, 2002). In contrast to NAT, most oncocyto- mas had predominantly inactive CTS-B and -L and lacked the cleaved, active forms, indicative of defective trafficking and lyso- some function (Figure 5B). CTS-B and -L are cysteine proteases involved in lysosomal protein turnover, and failure to activate them is consistent with the accumulation of autophagic sub- strates in oncocytomas. These findings suggested that mito- chondrial defects and energy crisis in oncocytomas produce a general disruption of Golgi-lysosome protein trafficking that impairs autophagy. The Golgi is a highly structurally organized organelle essential for protein sorting, trafficking, and post-translational modifica- tion. The Golgi undergoes disassembly at mitosis for proper par- titioning in daughter cells and in response to metabolic stress to conserve energy (Mao et al., 2013). To test if metabolic stress triggers Golgi disassembly and enhances LAMP-2 glycosylation that is alleviated by autophagy, isogenic tumor cells wild-type or null for the essential autophagy gene ATG7 (Guo et al., 2013a) were starved and Golgi and LAMP-2 localization and glycosyla- tion were examined. Wild-type cells displayed no changes in LAMP-2 glycosylation but had increased P-AMPK, inhibition of P-S6, perinuclear Golgi with distinct LAMP-2 localization, and only modest Golgi disruption with prolonged starvation (Figures 6A and 6B). Autophagy-deficient ATG7 null cells, however, showed Golgi disruption and enhanced LAMP-2 glycosylation even under normal growth conditions, which was increased by starvation (Figures 6A and 6B). LAMP-2 co-localized with the Golgi in the starved ATG7-deficient cells, consistent with Golgi retention (Figure 6B). Thus, metabolic crisis induced by either starvation or autophagy deficiency disrupts the Golgi and alters LAMP-2 glycosylation. Complex I Inhibition and Golgi Disassembly Block Autophagy Metabolic stress and AMPK activation triggers Golgi disas- sembly by phosphorylation of Golgi-specific brefeldin A resis- tance factor 1 (GBF1), a guanine nucleotide exchange factor (GEF) for Arf-GTPases. Phosphorylation of GBF1 disassociates GBF1 from the Golgi membrane, producing Golgi fragmentation (Miyamoto et al., 2008). To test the hypothesis that mitochon- drial mutations impairing complex I in oncocytomas produce metabolic crisis that disrupts Golgi function, thereby inhibiting autophagy, we examined the cellular response to the complex I inhibitor metformin (Bridges et al., 2014). Metformin potently induced AMPK activation (increased P-AMPK), inhibited mTOR (reduced P-S6), disrupted the Golgi, produced Golgi-LAMP-2 co-localization, and enhanced LAMP-2 glycosylation (Figure 6C). Importantly, metformin reduced autophagic flux as measured by inhibition of LC3-I to -II processing and LC3 puncta in the pres- ence of BA1 (Figure 6D). The other complex I inhibitors phenfor- min and rotenone acted similarly (Figure 6E). Treatment with the AMPK activator AICAR confirmed that these changes are AMPK mediated (Figure 6E). Pharmacological disruption of the Golgi hors Figure 6. Sustained Metabolic Stress Induces AMPK Signaling, Golgi Disassembly, Golgi-LAMP-2 Co-localization, and LAMP-2 Hyper- glycosylation and Inhibits Autophagy Flux (A) Autophagy deficiency and starvation increase LAMP-2 glycosylation. Western blots of autophagy-competent and autophagy-deficient TDCLs following 0, 4, and 8 hr of starvation (HBSS) for the indicated proteins. Starvation induces AMPK signaling (increased P-AMPK) and reduces mTOR activity (decreased P-S6). (B) Autophagy deficiency and starvation cause Golgi disassembly and LAMP-2 co-localization. Representative fluorescence images of autophagy-competent and autophagy-deficient TDCLs stained for Golgi (GM130), LAMP-2, and DNA (nuclear DAPI). (C) Metformin (Met) disrupts Golgi and causes LAMP-2 co-localization. Autophagy-competent TDCLs stained for Golgi (GM130), LAMP-2, and nuclei (DAPI) following metformin treatment for the indicated times. White arrowheads show Golgi-LAMP2 co-localization. Bottom panel: western blots from samples. (D) Metformin disrupts autophagic flux. Western blots of metformin-treated wild-type TDCLs for LC3 showing less LC3-II in BA1. Metformin disrupts autophagic flux indicated by reduced puncta of endogenous LC3 with BA1 with quantitation. (E) Phenformin (Phen), Rotenone (Rot), 5-aminoimidazole 4-carboxamide ribonucleotide (AICAR), and brefeldin A (Bref A) disrupt Golgi and cause Golgi-LAMP-2 co-localization and LAMP-2 hyper-glycosylation. Autophagy-competent TDCLs stained for Golgi (GM130) and LAMP-2 following Phen, Rot, AICAR or Bref A treatment for indicated times. Nuclei (DAPI). Bottom panel: corresponding western blots. Cell Reports 13, 1895–1908, December 1, 2015 ª2015 The Authors 1905 Figure 7. Model for Oncocytoma Genesis Chronic mitochondrial metabolic defect causes mitochondrial accumulation. with brefeldin A induced Golgi-LAMP2 co-localization and LAMP-2 hyper-glycosylation (Figure 6E). Thus, mutational inac- tivation of the mitochondrial ETC in oncocytomas is likely the cause of AMPK activation and disruption of Golgi, autophagy, and lysosome function. DISCUSSION Molecular characterization of benign renal oncocytomas re- vealed two main subtypes that share recurrent inactivating mutations in mitochondrial genes encoded by both nuclear and mitochondrial genomes. Clonal expansion of somatic mito- chondrial mutations in oncocytomas suggests that they are an early event (Gasparre et al., 2008). This might indicate a causal role for mitochondrial defects in tumor promotion as previously suggested (Gasparre et al., 2011). The genomic analysis of renal oncocytomas performed here was not able to resolve this issue; however, alterations of the nuclear genome consistent with oncogene activation and tumor suppressor gene inactiva- tion were found in almost all cases. Several issues we were able to address were the relationship between oncocytomas and other renal tumors, the cause of the metabolic defects, and the reasons for mitochondrial accumula- tion. Diploid type 1 has chromosomal rearrangements and upre- gulation of CCND1, a known oncogene. While type 2 lacked a clear driver mutation, it was hypodiploid, which could be the underlying basis for tumor formation, with the consistent loss of chromosome 1 being the prime candidate for location of a tu- mor suppressor gene. Loss of X or Y also suggests the presence of a tumor suppressor on the PAR that escapes X inactiva- tion. Despite differences at the genomic level, the two subtypes were qualitatively similar in phenotype. Our data suggest that oncocytomas have activated p53, and type 2 may progress to more malignant eosinophilic ChRCC upon acquiring p53, PTEN, and other mutations. Recent sequencing findings also support similarities between renal oncocytomas and ChRCC and the absence of p53 mutations 1906 Cell Reports 13, 1895–1908, December 1, 2015 ª2015 The Authors in oncocytomas (Durinck et al., 2015). Autophagy deficiency in genetically engi- neered mouse models causes energy crisis, activates p53, and suppresses tumor growth, effects lessened by p53 deficiency (Guo et al., 2013b). Analo- gously, oncocytomas may represent benign disease whose growth is limited by metabolic crisis and activation of p53. Subsequent mutation in p53 in type 2 may restore growth capacity and pro- gression to eosinophilic ChRCC (Guo et al., 2013a; Strohecker et al., 2013). It is also worth noting that ChRCCs have significantly fewer mitochondrial muta- tions than oncocytomas, suggesting that progression from onco- cytoma might be aided by heteroplasmy and the ability to partly recover the functional pool of mitochondria (Ju et al., 2014). This might be achieved by restoration of autophagy and lyso- somal biogenesis by TFE3 or TFEB activation (Linehan and Rick- etts, 2013). Alternatively, tumor cells depleted of mitochondrial DNA recover respiration, ETC function, and tumor growth by horizontal transfer of normal mitochondria from non-tumor to tumor tissue (Tan et al., 2015). These findings point to the need for tumors to preserve mitochondrial ETC function. In contrast, type 1 may represent an evolutionary dead end, as these tumors with CCND1 rearrangements are less susceptible to recurrence (Sukov et al., 2009). Recurrent pathogenic mitochondrial gene mutations en- coded within the mitochondrial and nuclear genomes are the likely cause of the respiration defects in oncocytomas, consis- tent with and expanding on previous work. Inactivating muta- tions are found in complex I, III, IV, and control regions, in nuclear encoded NDUFA13, and now also in complex V, ANT2, NDUFV1, and UQCRC2. Alterations in key mitochon- drial components are associated with the accumulation of defective mitochondria, suggesting a link between chronic metabolic impairment and autophagy suppression. Indeed, depletion of mitochondrial DNA and the resulting respiration defect inactivates autophagy in budding yeast and mamma- lian cells (Graef and Nunnari, 2011; Tan et al., 2015). In agree- ment with these findings, the pathogenic mitochondrial muta- tions and respiration defects in the oncocytomas reported here similarly suppress autophagy with the resulting metabolic insufficiency attributed to disruption of Golgi and lysosome function (Figure 7). A major consequence of compromised mitochondrial respi- ration in oncocytomas is chronic metabolic deficiency exem- plified by AMPK activation and loss of mTOR signaling. This is also seen in genetically engineered mouse models for cancer with autophagy inactivation (Guo et al., 2013b; Karsli- Uzunbas et al., 2014) and in response to pharmacologic inhibition of complex I (e.g., metformin) in cancer cells in vitro. Normally, transient energy crisis and AMPK activation cause Golgi disassembly to block trafficking and secretion that is thought to conserve energy. In contrast, the dysfunctional mitochondria in oncocytomas causes a chronic metabolic defect, sustained energy crisis, AMPK activation, and disrup- tion of Golgi and lysosome function essential for autophagic cargo degradation (Figure 7). Importantly, inhibition of com- plex I is sufficient to disrupt the Golgi, trafficking, and auto- phagic flux. Thus, cancer cells need to sustain mitochondrial respiration to prevent metabolic crisis that inhibits trafficking. It is worth noting that PTEN and p53 mutations, which alleviate AMPK activation, occur in ChRCC and may facilitate progres- sion of type 2 tumors by restoring Golgi, autophagy, and lyso- some function. Oncocytic features are common to many different subsets of renal tumors such as those with SDHB, SDHC, SDHD, and FLCN mutations and in hybrid ChRCC tumors (Linehan and Rick- etts, 2013). This may be due to the occurrence of numerous means to impair mitochondrial respiration, producing chronic metabolic and trafficking defects, and the resulting mitochon- drial accumulation as described here in oncocytomas. Metfor- min and other means to inhibit mitochondrial function have anti-cancer activity (Pollak, 2013; Viale et al., 2014; Weinberg et al., 2010). Our findings suggest that this may be due in part to chronic metabolic stress-induced trafficking defects, necessi- tating the preservation of ETC function to overcome a metabolic barrier to tumorigenesis. EXPERIMENTAL PROCEDURES Tumor Samples Tumor and adjacent normal kidney tissues from ten sporadic oncocytomas at MSKCC were collected. All patients had previously consented to a tissue protocol. This study was preapproved by the institutional review board (IRB), and all patients signed informed consent. All samples were reviewed by a dedi- cated uropathologist. Paired fresh-frozen normal and primary tumor tissue blocks were identified, marked, and macrodissected for maximal tumor den- sity. The two tumor specimens with normal adjacent tissue from CINJ were obtained from a de-identified tissue bank maintained at the Biospecimen Repository Service under the auspices of an IRB-approved protocol. All spec- imens were pre-existing and de-identified. Tissues for cytogenetics were obtained by reviewing records to identify all cases of renal oncocytomas and ChRCC for which cytogenetics was performed at Robert Wood Johnson University Hospital in the past 10 years. Whole-Exome and RNA-Sequencing Data Analysis The Agilent SureSelect Human All Exon V5 Enrichment protocol was used for exome capture. The exome was sequenced on Illumina HiSeq 2500, 2 3 100 bp paired-end reads to a minimum depth of 100 million reads per sample. The reads were preprocessed with Picard to remove PCR duplicates and aligned to the human reference genome (hg19) with Bowtie2. Somatic point mutations and small indels were called with MuTect and GATK Somatic Indel Detector algorithms. CNVs were detected with exomeCNV. Library prepara- tion and RNA sequencing were carried out using the Illumina TruSeq RNA Sample Prep Kit v2, according to the manufacturer’s protocol (Illumina), and samples were sequenced on the Illumina HiSeq 2500 (2 3 100 bp paired- end reads) to a minimum depth of 50 million reads per sample. Sequence reads were mapped using TopHat2/Bowtie2 algorithm. Analysis of gene expression was performed using Cufflinks and DESeq. The complete descrip- tion of the sequencing and genomic analysis and all other methods can be found in Supplemental Experimental Procedures. Cell Re ACCESSION NUMBERS The accession number for the sequencing data reported in this paper is SRA: SRP051509. SUPPLEMENTAL INFORMATION Supplemental Information includes Supplemental Experimental Procedures, two figures, and twelve tables and can be found with this article online at http://dx.doi.org/10.1016/j.celrep.2015.10.059. AUTHOR CONTRIBUTIONS S.J. designed and performed the histology and biological experiments. D.T. performed sequencing data analysis. H.A. performed karyotype analysis. A.A.H. and J.J.H. guided the renal tumor biology and tumor procurement. M.Y. helped perform FISH analysis. S.G. helped conceive and direct the proj- ect and human tissue and chromosome analysis. C.S.C. directed project and bioinformatics analysis. E.W. conceived and directed the project. ACKNOWLEDGMENTS E.W. acknowledges support from the NIH (grants R01 CA163591, R01 CA130893, and CA188096), the Val Skinner Foundation, and CINJ (grant P30 CA072720). C.S.C. and D.T. acknowledge support from CINJ. S.G. ac- knowledges support from the Triple Negative Breast Cancer Foundation, the Hugs for Brady Foundation, and the NIH (grants R01 CA169182 and R01 CA130893). S.J. received support from a postdoctoral fellowship from the New Jersey Commission for Cancer Research (DFHS13PPCO24). This work used the Functional Genomics/Rutgers University Cell and DNA Repository, Biospecimen Repository Service, and Flow Cytometry/Cell Sorting CINJ core facilities supported by grant P30 CA72720. We thank R. Muthuramam for technical assistance. E.W. is on the SAB of Forma Therapeutics. J.H. is on the EAB of Chugai Pharmaceuticals, is a consultant for Novartis, and re- ceives research support from Pfizer and Novartis. Received: July 23, 2015 Revised: September 28, 2015 Accepted: October 19, 2015 Published: November 19, 2015 REFERENCES Bridges, H.R., Jones, A.J., Pollak, M.N., and Hirst, J. (2014). Effects of metfor- min and other biguanides on oxidative phosphorylation in mitochondria. Bio- chem. J. 462, 475–487. Davis, C.F., Ricketts, C.J., Wang, M., Yang, L., Cherniack, A.D., Shen, H., Bu- hay, C., Kang, H., Kim, S.C., Fahey, C.C., et al.; Cancer Genome Atlas Research Network (2014). The somatic genomic landscape of chromophobe renal cell carcinoma. Cancer Cell 26, 319–330. Durinck, S., Stawiski, E.W., Pavia-Jimenez, A., Modrusan, Z., Kapur, P., Jaiswal, B.S., Zhang, N., Toffessi-Tcheuyap, V., Nguyen, T.T., Pahuja, K.B., et al. (2015). Spectrum of diverse genomic alterations define non-clear cell renal carcinoma subtypes. Nat. Genet. 47, 13–21. Ebersberger, I., Metzler, D., Schwarz, C., and Pääbo, S. (2002). Genomewide comparison of DNA sequences between humans and chimpanzees. Am. J. Hum. Genet. 70, 1490–1497. Eskelinen, E.L. (2006). Roles of LAMP-1 and LAMP-2 in lysosome biogenesis and autophagy. Mol. Aspects Med. 27, 495–502. Gao, J., Aksoy, B.A., Dogrusoz, U., Dresdner, G., Gross, B., Sumer, S.O., Sun, Y., Jacobsen, A., Sinha, R., Larsson, E., et al. (2013). Integrative analysis of complex cancer genomics and clinical profiles using the cBioPortal. Sci. Signal. 6, pl1. Gasparre, G., Porcelli, A.M., Bonora, E., Pennisi, L.F., Toller, M., Iommarini, L., Ghelli, A., Moretti, M., Betts, C.M., Martinelli, G.N., et al. (2007). Disruptive ports 13, 1895–1908, December 1, 2015 ª2015 The Authors 1907 http://dx.doi.org/10.1016/j.celrep.2015.10.059 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref1 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref1 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref1 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref2 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref2 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref2 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref2 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref3 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref3 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref3 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref3 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref4 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref4 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref4 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref5 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref5 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref6 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref6 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref6 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref6 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref7 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref7 mitochondrial DNA mutations in complex I subunits are markers of oncocytic phenotype in thyroid tumors. Proc. Natl. Acad. Sci. USA 104, 9001–9006. Gasparre, G., Hervouet, E., de Laplanche, E., Demont, J., Pennisi, L.F., Colom- bel, M., Mège-Lechevallier, F., Scoazec, J.Y., Bonora, E., Smeets, R., et al. (2008). Clonal expansion of mutated mitochondrial DNA is associated with tumor formation and complex I deficiency in the benign renal oncocytoma. Hum. Mol. Genet. 17, 986–995. Gasparre, G., Romeo, G., Rugolo, M., and Porcelli, A.M. (2011). Learning from oncocytic tumors: Why choose inefficient mitochondria? Biochim. Biophys. Acta 1807, 633–642. Graef, M., and Nunnari, J. (2011). Mitochondria regulate autophagy by conserved signalling pathways. EMBO J. 30, 2101–2114. Guo, J.Y., Karsli-Uzunbas, G., Mathew, R., Aisner, S.C., Kamphorst, J.J., Stro- hecker, A.M., Chen, G., Price, S., Lu, W., Teng, X., et al. (2013a). Autophagy suppresses progression of K-ras-induced lung tumors to oncocytomas and maintains lipid homeostasis. Genes Dev. 27, 1447–1461. Guo, J.Y., Xia, B., and White, E. (2013b). Autophagy-mediated tumor promo- tion. Cell 155, 1216–1219. Hudacko, R., May, M., and Aviv, H. (2011). A new translocation between chro- mosomes 6 and 9 helps to establish diagnosis of renal oncocytoma. Ann. Diagn. Pathol. 15, 278–281. Ishidoh, K., and Kominami, E. (2002). Processing and activation of lysosomal proteinases. Biol. Chem. 383, 1827–1831. Ju, Y.S., Alexandrov, L.B., Gerstung, M., Martincorena, I., Nik-Zainal, S., Ram- akrishna, M., Davies, H.R., Papaemmanuil, E., Gundem, G., Shlien, A., et al.; ICGC Breast Cancer Group; ICGC Chronic Myeloid Disorders Group; ICGC Prostate Cancer Group (2014). Origins and functional consequences of somatic mitochondrial DNA mutations in human cancer. eLife 3, 3. Karsli-Uzunbas, G., Guo, J.Y., Price, S., Teng, X., Laddha, S.V., Khor, S., Kalaany, N.Y., Jacks, T., Chan, C.S., Rabinowitz, J.D., and White, E. (2014). Autophagy is required for glucose homeostasis and lung tumor maintenance. Cancer Discov. 4, 914–927. Kokoszka, J.E., Waymire, K.G., Levy, S.E., Sligh, J.E., Cai, J., Jones, D.P., MacGregor, G.R., and Wallace, D.C. (2004). The ADP/ATP translocator is not essential for the mitochondrial permeability transition pore. Nature 427, 461–465. Linehan, W.M., and Ricketts, C.J. (2013). The metabolic basis of kidney can- cer. Semin. Cancer Biol. 23, 46–55. Mao, L., Li, N., Guo, Y., Xu, X., Gao, L., Xu, Y., Zhou, L., and Liu, W. (2013). AMPK phosphorylates GBF1 for mitotic Golgi disassembly. J. Cell Sci. 126, 1498–1505. Mayr, J.A., Meierhofer, D., Zimmermann, F., Feichtinger, R., Kogler, C., Rat- schek, M., Schmeller, N., Sperl, W., and Kofler, B. (2008). Loss of complex I due to mitochondrial DNA mutations in renal oncocytoma. Clin. Cancer Res. 14, 2270–2275. McBride, H., and Soubannier, V. (2010). Mitochondrial function: OMA1 and OPA1, the grandmasters of mitochondrial health. Curr. Biol. 20, R274–R276. Mishra, P., Carelli, V., Manfredi, G., and Chan, D.C. (2014). Proteolytic cleav- age of Opa1 stimulates mitochondrial inner membrane fusion and couples fusion to oxidative phosphorylation. Cell Metab. 19, 630–641. Miyamoto, T., Oshiro, N., Yoshino, K., Nakashima, A., Eguchi, S., Takahashi, M., Ono, Y., Kikkawa, U., and Yonezawa, K. (2008). AMP-activated protein ki- nase phosphorylates Golgi-specific brefeldin A resistance factor 1 at Thr1337 to induce disassembly of Golgi apparatus. J. Biol. Chem. 283, 4430–4438. 1908 Cell Reports 13, 1895–1908, December 1, 2015 ª2015 The Aut Nabi, I.R., and Rodriguez-Boulan, E. (1993). Increased LAMP-2 polylactos- amine glycosylation is associated with its slower Golgi transit during establish- ment of a polarized MDCK epithelial monolayer. Mol. Biol. Cell 4, 627–635. Ng, K.L., Rajandram, R., Morais, C., Yap, N.Y., Samaratunga, H., Gobe, G.C., and Wood, S.T. (2014). Differentiation of oncocytoma from chromophobe renal cell carcinoma (RCC): can novel molecular biomarkers help solve an old prob- lem? J. Clin. Pathol. 67, 97–104. Pollak, M. (2013). Potential applications for biguanides in oncology. J. Clin. Invest. 123, 3693–3700. Simonnet, H., Demont, J., Pfeiffer, K., Guenaneche, L., Bouvier, R., Brandt, U., Schagger, H., and Godinot, C. (2003). Mitochondrial complex I is deficient in renal oncocytomas. Carcinogenesis 24, 1461–1466. Strohecker, A.M., Guo, J.Y., Karsli-Uzunbas, G., Price, S.M., Chen, G.J., Mathew, R., McMahon, M., and White, E. (2013). Autophagy sustains mito- chondrial glutamine metabolism and growth of BrafV600E-driven lung tumors. Cancer Discov. 3, 1272–1285. Sukov, W.R., Ketterling, R.P., Lager, D.J., Carlson, A.W., Sinnwell, J.P., Chow, G.K., Jenkins, R.B., and Cheville, J.C. (2009). CCND1 rearrangements and cy- clin D1 overexpression in renal oncocytomas: frequency, clinicopathologic features, and utility in differentiation from chromophobe renal cell carcinoma. Hum. Pathol. 40, 1296–1303. Tan, A.S., Baty, J.W., Dong, L.F., Bezawork-Geleta, A., Endaya, B., Goodwin, J., Bajzikova, M., Kovarova, J., Peterka, M., Yan, B., et al. (2015). Mitochon- drial genome acquisition restores respiratory function and tumorigenic poten- tial of cancer cells without mitochondrial DNA. Cell Metab. 21, 81–94. Tanaka, Y., Guhde, G., Suter, A., Eskelinen, E.L., Hartmann, D., L€ullmann- Rauch, R., Janssen, P.M., Blanz, J., von Figura, K., and Saftig, P. (2000). Accu- mulation of autophagic vacuoles and cardiomyopathy in LAMP-2-deficient mice. Nature 406, 902–906. Thilly, W.G. (2003). Have environmental mutagens caused oncomutations in people? Nat. Genet. 34, 255–259. Vander Heiden, M.G., Cantley, L.C., and Thompson, C.B. (2009). Understand- ing the Warburg effect: the metabolic requirements of cell proliferation. Science 324, 1029–1033. Varela, I., Tarpey, P., Raine, K., Huang, D., Ong, C.K., Stephens, P., Davies, H., Jones, D., Lin, M.L., Teague, J., et al. (2011). Exome sequencing identifies frequent mutation of the SWI/SNF complex gene PBRM1 in renal carcinoma. Nature 469, 539–542. Viale, A., Pettazzoni, P., Lyssiotis, C.A., Ying, H., Sánchez, N., Marchesini, M., Carugo, A., Green, T., Seth, S., Giuliani, V., et al. (2014). Oncogene ablation- resistant pancreatic cancer cells depend on mitochondrial function. Nature 514, 628–632. Wallace, D.C. (2012). Mitochondria and cancer. Nat. Rev. Cancer 12, 685–698. Warburg, O. (1956). On respiratory impairment in cancer cells. Science 124, 269–270. Ward, P.S., and Thompson, C.B. (2012). Metabolic reprogramming: a cancer hallmark even warburg did not anticipate. Cancer Cell 21, 297–308. Weinberg, F., Hamanaka, R., Wheaton, W.W., Weinberg, S., Joseph, J., Lopez, M., Kalyanaraman, B., Mutlu, G.M., Budinger, G.R., and Chandel, N.S. (2010). Mitochondrial metabolism and ROS generation are essential for Kras-mediated tumorigenicity. Proc. Natl. Acad. Sci. USA 107, 8788–8793. Youle, R.J., and Narendra, D.P. (2011). Mechanisms of mitophagy. Nat. Rev. Mol. Cell Biol. 12, 9–14. hors http://refhub.elsevier.com/S2211-1247(15)01248-6/sref7 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref7 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref8 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref8 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref8 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref8 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref8 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref9 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref9 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref9 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref10 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref10 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref11 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref11 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref11 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref11 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref12 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref12 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref13 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref13 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref13 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref14 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref14 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref15 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref15 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref15 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref15 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref15 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref16 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref16 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref16 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref16 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref17 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref17 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref17 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref17 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref18 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref18 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref19 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref19 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref19 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref20 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref20 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref20 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref20 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref21 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref21 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref22 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref22 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref22 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref23 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref23 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref23 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref23 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref24 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref24 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref24 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref25 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref25 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref25 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref25 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref26 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref26 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref27 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref27 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref27 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref28 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref28 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref28 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref28 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref29 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref29 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref29 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref29 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref29 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref30 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref30 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref30 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref30 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref31 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref31 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref31 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref31 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref31 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref32 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref32 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref33 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref33 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref33 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref34 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref34 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref34 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref34 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref35 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref35 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref35 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref35 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref36 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref37 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref37 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref38 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref38 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref39 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref39 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref39 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref39 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref40 http://refhub.elsevier.com/S2211-1247(15)01248-6/sref40 The Genomic Landscape of Renal Oncocytoma Identifies a Metabolic Barrier to Tumorigenesis Introduction Genomic Analysis Indicates Two Main Subtypes of Renal Oncocytoma Type 2 Oncocytoma May Be the Precursor of Eosinophilic ChRCC Recurrent CCND1 Rearrangement and Amplification in Type 1 Oncocytomas p53 Is a Barrier to Oncocytoma Progression Type 1 and 2 Oncocytomas Display Mitochondrial Defects and Metabolic Stress Oncocytomas Manifest Phenotypes of Defective Autophagy Mitophagy Is Primed, but Not Executed, in Oncocytomas Impaired Mitochondrial Respiration in Primary Oncocytoma Cells Oncocytomas Display LAMP-2 Trafficking, Golgi, and Lysosome Defects Complex I Inhibition and Golgi Disassembly Block Autophagy Discussion Experimental Procedures Tumor Samples Whole-Exome and RNA-Sequencing Data Analysis Accession Numbers Supplemental Information Author Contributions Acknowledgments References