Differential Aspartate Usage Identifies a Subset of Cancer Cells Particularly Dependent on OGDH

Article

Differential Aspartate Usage Identifies a Subset of Cancer Cells Particularly Dependent on OGDH Graphical Abstract

Authors Eric L. Allen, Danielle B. Ulanet, David Pirman, ..., Marion Dorsch, Shengfang Jin, Gromoslaw A. Smolen

Correspondence [email protected]

In Brief Allen et al. show that knockdown of alpha-ketoglutarate dehydrogenase E1 subunit, OGDH, results in profound growth defects in a subset of cancer cell lines in vitro and in vivo. Differential aspartate utilization, via malate-aspartate shuttle, serves as a marker of OGDH dependence and suggests broader utility of nutrient utilization screens to identify cancer targets.

Highlights d

Cancer cells display a broad range of sensitivities to OGDH knockdown

d

Predictive 3D in vitro assays allow for in vivo translation of metabolic vulnerabilities

d

OGDH-dependent cell lines display low levels of malateaspartate shuttle activity

Allen et al., 2016, Cell Reports 17, 876–890 October 11, 2016 ª 2016 The Author(s). http://dx.doi.org/10.1016/j.celrep.2016.09.052

Cell Reports

Article Differential Aspartate Usage Identifies a Subset of Cancer Cells Particularly Dependent on OGDH Eric L. Allen,1,3 Danielle B. Ulanet,1,3 David Pirman,1,3 Christopher E. Mahoney,1 John Coco,1 Yaguang Si,1 Ying Chen,2 Lingling Huang,2 Jinmin Ren,2 Sung Choe,1 Michelle F. Clasquin,1 Erin Artin,1 Zi Peng Fan,1 Giovanni Cianchetta,1 Joshua Murtie,1 Marion Dorsch,1 Shengfang Jin,1 and Gromoslaw A. Smolen1,4,* 1Agios

Pharmaceuticals, 88 Sidney Street, Cambridge, MA 02139, USA ChemPartner Co. Ltd., 998 Halei Road, Pudong, 201203 Shanghai, China 3Co-first author 4Lead Contact *Correspondence: [email protected] http://dx.doi.org/10.1016/j.celrep.2016.09.052 2Shanghai

SUMMARY

Although aberrant metabolism in tumors has been well described, the identification of cancer subsets with particular metabolic vulnerabilities has remained challenging. Here, we conducted an siRNA screen focusing on enzymes involved in the tricarboxylic acid (TCA) cycle and uncovered a striking range of cancer cell dependencies on OGDH, the E1 subunit of the alpha-ketoglutarate dehydrogenase complex. Using an integrative metabolomics approach, we identified differential aspartate utilization, via the malate-aspartate shuttle, as a predictor of whether OGDH is required for proliferation in 3D culture assays and for the growth of xenograft tumors. These findings highlight an anaplerotic role of aspartate and, more broadly, suggest that differential nutrient utilization patterns can identify subsets of cancers with distinct metabolic dependencies for potential pharmacological intervention. INTRODUCTION The genetic changes underlying aberrant cancer cell proliferation signal a rewiring of metabolism to balance macromolecular biosynthesis with ATP production (Ward and Thompson, 2012). Since metabolic reprogramming has gained acceptance as a hallmark of most cancers (Hanahan and Weinberg, 2011), the development of therapeutic strategies to target key metabolic dependencies has attracted much attention. While the clinical feasibility of targeting metabolism has long been established by multiple approved agents targeting nucleic acid metabolism (Vander Heiden, 2011), a goal for the next wave of novel therapeutic agents targeting deregulated tumor metabolism will be to use the personalized medicine paradigm to identify more selective agents that can be utilized in defined cancer subtypes. Large-scale cancer sequencing efforts have identified relatively few recurrent mutations in metabolic genes and fewer still with clear gain-of-function roles. The notable exceptions are IDH1 and IDH2, where mutations lead to a novel enzymatic activity

producing the oncometabolite, 2-hydroxyglutarate (Dang et al., 2009). In fact, the first inhibitors of mutant IDH1 and IDH2 have demonstrated promising signs of efficacy in early clinical trials (Stein and Tallman, 2016). Since other mutations in metabolic genes, across diverse tumor types, are associated with a loss of function, synthetic lethal approaches have been undertaken to identify specific dependencies in these contexts, such as enolase 2 in ENO1-deleted tumors (Muller et al., 2012) and heme oxygenase in FH-null tumors (Frezza et al., 2011). Alternatively, exploiting cancer-specific auxotrophies has shown promise, such as dependence on arginine in the context of ASS1-negative tumors (Qiu et al., 2014). The general approach of identifying nodes of particular metabolic sensitivity has been expanded to more common cancer genotypes, such as PRPS2 in Myc-driven tumors (Cunningham et al., 2014) and glutaminase in VHL-null renal tumors (Gameiro et al., 2013). Recent studies have also pointed to enhanced dependency on glutaminase in mesenchymal-type tumor cells (Ulanet et al., 2014), and inhibitors targeting glutaminase are currently being tested in clinical trials (Gross et al., 2014). While the success of therapeutic strategies targeting nutrient utilization pathways remains to be determined, the complex in vivo microenvironment and metabolic plasticity of tumors may prevent sufficient crippling of tumors by blocking a singlenutrient utilization pathway. For instance, although glutamine is recognized as an important anaplerotic source in cancer cells in vitro, glucose oxidation via pyruvate carboxylase may predominate in the in vivo tumor setting (Davidson et al., 2016; Sellers et al., 2015). A lack of glutamine oxidation in the tricarboxylic acid (TCA) cycle was reported for brain tumors, which were demonstrated to simultaneously oxidize glucose and acetate (Mashimo et al., 2014). Internalization of extracellular protein via macropinocytosis was identified as a feature of Ras-transformed cells, resulting in yet another way to acquire glutamine and other amino acids (Commisso et al., 2013). Due to the multitude of potential nutrient sources and routes of feeding carbon into the TCA cycle, we investigated whether directly targeting TCA cycle enzymes could serve as a more definitive means of blocking this pathway. In an small interfering RNA (siRNA) screen, we identified OGDH, the E1 subunit of alpha-ketoglutarate dehydrogenase complex (OGDHC), as the top hit. OGDH catalyzes the conversion of a-ketoglutarate

876 Cell Reports 17, 876–890, October 11, 2016 ª 2016 The Author(s). This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

(aKG) to succinyl-CoA and significantly contributes to mitochondrial NADH production (Mullen et al., 2014). While OGDH has been reported to contribute to fumarate accumulation and optimal growth of FH-deficient cells (Sullivan et al., 2013), we wanted to determine its role in a broader array of cancer cell lines. A recent report demonstrated differential sensitivity of several cell lines to the phosphonate analog of alpha-ketoglutarate and OGDH inhibitor, succinyl phosphonate (SP) (Bunik et al., 2016). We used highly validated small hairpin RNA (shRNA) tools in an expanded panel of cancer cell lines and identified a range of cellular dependence on OGDH, with some cell lines being completely OGDH independent. The reliance on OGDH was context dependent and in a more stringent, modified version of a 3D culture assay (m3D), only a small subset of lines was sensitive to OGDH knockdown. Notably, significantly impaired growth upon OGDH knockdown in m3D translated to a striking dependence on OGDH for tumor growth in vivo and was associated with a distinct pattern of aspartate carbon and nitrogen utilization. These findings lend support for further investigation of OGDH as a therapeutic target in a metabolically distinct subset of cancers. RESULTS siRNA Screen Focused on TCA Cycle Components Highlights OGDH To functionally understand TCA cycle and identify steps particularly critical for cell viability, we undertook a focused siRNA-based screen of TCA cycle genes in A549 cells, a model cancer cell line with fast growth kinetics in vitro and in vivo (Figure 1A). For completeness, we included additional genes encoding homologs performing similar enzymatic functions, but localized to the cytosol. To ensure that we thoroughly assess the contribution of each gene to the cell proliferation, we implemented a multi-transfection protocol designed to start the growth assay with the cells already under conditions of significant target gene knockdown (Figure S1A). Using a very abundant target, GAPDH, as a control, we showed that significant knockdown is maintained for the entire duration of the assay (Figure S1B). Using this multi-transfection procedure, we successfully achieved knockdown of all genes analyzed, as measured by qPCR at the beginning of the assay and the amount of residual mRNA was, on average, 3.5% (Figure S1C). Surprisingly, a wide spectrum of growth defects was observed, ranging from none to severe (Figure 1B). The strongest hit was OGDH, the E1 subunit of alpha-ketoglutarate dehydrogenase complex. To further validate this hit, we deconvoluted the initial OGDH siRNA pool by repeating the knockdown experiment with individual siRNA duplexes. Similar findings were observed with multiple duplexes, suggesting that the results are likely to be on target (Figure S1D). To increase confidence in our findings, we developed multiple doxycycline (dox)-inducible shRNAs capable of knocking down OGDH: sh815, sh885, sh1137 (Figure 1C), which resulted in profound dox-dependent growth defects in a colony formation assay (Figure 1D). To conclusively show that the effects observed were on target, we conducted rescue experiments with wild-type OGDH that

was engineered to be resistant to multiple shRNAs. To formally demonstrate that an enzymatic activity, rather than any potential scaffolding role, is critical for the growth of cells, we designed a catalytically dead version of OGDH. Since an X-ray structure of human OGDH was not available, a homology model was built using the crystal structure of Mycobacterium smegmatis alpha-ketoglutarate decarboxylase homodimer, and two corresponding human mutants were predicted to be catalytically dead, R509K and H513F. Indeed, overexpression of wild-type OGDH, but not either of the catalytically dead versions, led to a consistent rescue of growth phenotypes of all three shRNAs (Figures 1E and 1F; data not shown). Collectively, this establishes a high degree of confidence in the genetic tools used for all the subsequent studies and validates OGDH as a point of particular sensitivity in A549 cells. OGDH Is Necessary for TCA Cycle Metabolism In Vitro and In Vivo To investigate the metabolic consequences of OGDH knockdown, we conducted multiple analyses that take advantage of the inducible shRNA system, allowing investigation of how OGDH knockdown affects cells, but before any overt loss of cell viability. First, we measured steady-state pool levels of TCA cycle intermediates and observed significant decreases in malate abundance with concomitant increases in aKG levels (Figure 2A). These changes were consistent with the metabolites being downstream and upstream of OGDH, respectively. As expected, aspartate levels also significantly decreased (Figure 2A), since aspartate is rapidly synthesized by cancer cells from oxaloacetate via aminotransferase activity and has been used as a proxy measure of TCA cycle activity (DeBerardinis et al., 2007). Interestingly, total citrate levels were only slightly decreased and total glutamate levels were unaffected by OGDH knockdown (Figure S2A). Second, we used a well-established in vitro anaplerotic substrate, 13C5-glutamine, to trace carbon flow in the TCA cycle in cells with OGDH knockdown (Figures 2B and S2B). These data show a clear constriction in the TCA cycle at the OGDH step. Significant decreases in the abundance of 13C4 isotopomers of malate, aspartate, and citrate were observed (Figures 2B and S2B). Increases in the abundance of 13 C5 isotopomers of aKG and citrate were also observed, consistent with lower alpha-ketoglutarate dehydrogenase enzyme activity. To further characterize these metabolic phenotypes, we conducted time course labeling experiments using 13C5glutamine and 13C6-glucose showing that both glucose and glutamine contribute carbon to the TCA, and effects of OGDH knockdown can be monitored with both labeling strategies (Figure S2C). Consistent with the previous results, OGDH appears to affect glucose entry into the TCA, as the abundance of 13C2citrate isotopomer was dramatically decreased upon OGDH knockdown in a 13C6-glucose-labeling experiment. Collectively, these data suggest a functional E1 (OGDH) component of the alpha-ketoglutarate dehydrogenase complex is necessary for normal oxidative TCA cycle function. To investigate whether OGDH was required for tumor maintenance in vivo, we conducted a xenograft study where doxycycline was added to animals harboring already established tumors of approximately 200 mm3. OGDH knockdown mediated

Cell Reports 17, 876–890, October 11, 2016 877

AcetylCoA

A

B

CS Citrate

Oxaloacetate

ACO1 ACO2

MDH1 MDH2

TCA cycle genes NT controls KIF11

Malate

Relative growth (%)

140 Isocitrate

IDH1 IDH2 IDH3A IDH3B IDH3G

FH

Fumarate

aKG

120 100 80 60 40 20

OGDH

0

SDHA SDHB SDHC SDHD

Succinate

0

OGDH DLST DLD

SuccinylCoA

5

10

15 Genes

20

25

30

SUCLA2 SUCLAG1 SUCLAG2

C

OGDH

D OGDH shNT

–

sh815

+

–

shNT

sh885

–

+

sh815

sh885

sh1137

sh1137

–

+

+

– DOX

DOX OGDH

+ DOX

-actin

F

E shNT control vector

–

+

OGDH wild type

–

+

sh885 OGDH R509K

–

+

vector

–

+

OGDH wild type

–

+

vector

OGDH R509K

–

+

shNT control OGDH wild type

sh885 OGDH R509K

vector

OGDH wild type

OGDH R509K – DOX

DOX OGDH -actin

+ DOX

Figure 1. siRNA Screen of TCA Cycle Genes Highlights the Key Role of OGDH (A) Schematic illustrating the TCA cycle. All the genes included in the screen are shown in blue, and homologs localized to the cytosol are denoted in italics. (B) siRNA screen in A549 cells. Relative growth defects after knockdown of individual genes are normalized to the average of several experiments where cells were transfected with NT control siRNA pool. (C) Immunoblot analysis showing effective OGDH knockdown using shRNAs after 4 days of dox treatment. (D) Dox-dependent OGDH knockdown leads to growth defects in a colony formation assay. (E) Immunoblot analysis showing overexpression of shRNA-resistant OGDH, wild-type, and R509K mutant in shRNA knockdown lines (shNT and sh885). (F) Rescue of sh885-induced cell growth phenotypes by wild-type OGDH but not the R509K mutant.

by doxycycline-inducible sh885 resulted in significant growth defects (Figure 2C). The effect was rescued by overexpression of an shRNA-resistant wild-type OGDH, but not the catalytically dead R509K mutant OGDH. Similar to in vitro studies, OGDH protein was efficiently knocked down in vivo (Figure 2D), which resulted in a significant decrease of enzymatic activity in tumor lysates at the end of the study (Figure 2E). Metabolic evaluation of tumors revealed similar TCA cycle perturbations in vivo as compared to in vitro experiments (Figure 2F). Collectively, these data suggest that OGDH, and thus the alpha-ketoglutarate dehydrogenase complex, contributes to TCA cycle metabolism and is necessary for growth of A549 cells in vitro and in vivo.

878 Cell Reports 17, 876–890, October 11, 2016

Cells Display a Broad Range of Sensitivities to OGDH Knockdown To define cancer subtypes particularly susceptible to reduction of OGDH function, we opted for an expanded shRNA-based screen using the two highly validated shRNAs targeting OGDH (sh815 and sh885) and an additional control, shNT. Consequences of OGDH knockdown on cell proliferation were assessed using several different assays. The standard assay consisting of plating cells on 2D plastic plates under low-density conditions was used. While easy to conduct, this assay does not recapitulate many aspects of tumor cells in vivo and consequently has relatively limited predictive power of in vitro-to-in vivo

A

+

Empty OGDHwt

_

+

13

+

DOX

C4 Malate

*

*

13

*

1.5 1.0 0.5 +

_

+

Empty OGDHwt

_

+

_

_

+

+

_

*

*

_

+

+

DOX

200

D

sh885 + empty reexpression vector sh885 + OGDH-wt reexpression

sh885 + OGDH-R509K reexpression

0 25

E

F 2.5

+ DOX Day 24

-actin

OGDH -actin

OGDH -actin

*

*

*

*

2.0

OGDH

+

_

_

+

1.5

_

+

DOX

sh885

36% TGI (p = 0.0047)

0

5

10 15 20 25 Days Aspartate

*

#

3

#

*

2 1 0

–

+

Empty

1.0

–

+

OGDH-wt

–

+

OGDHR509K

DOX Reexpression vector

sh885

0.5 0.0

+

Empty OGDH- Reexpression vector wt

600

0

20

ns

– DOX + DOX

800

10 15 Days

#

sh885 + OGDH-R509K reexpression

200

– DOX

_

shNT

(NS)

5

C5 aKG

Empty OGDHwt

sh885

600

0

sh885

#

0

0

25

DOX

0.2

200 20

+

0.4

400

10 15 Days

_

+

Empty OGDH- Reexpression vector wt

0.6

400

5

ns

0.8

400

0

_

+

1

Empty OGDH- Reexpression vector wt

– DOX + DOX

800

_

shNT

sh885 + OGDH-wt reexpression

49% TGI (p < 0.0001)

+

Empty OGDHwt

13

C4 Aspartate

shNT

600

_

DOX

sh885

Empty OGDHwt

sh885

– DOX + DOX

+

0.5 0

DOX

_

+

1

nmol/min/mg

Tumor size (mm3)

_

Empty OGDH- Reexpression vector wt

*

sh885 + empty reexpression vector 800

+

2

Empty OGDH- Reexpression vector wt

shNT

C

+

_

_

+

shNT

2.0

_

_

Empty OGDHwt

sh885

2.5

0

1 0

Empty OGDH- Reexpression vector wt

shNT

B

_

2

Pool total (AU x 1E+6)

+

_

#

ns

–

+

Empty

–

+

OGDH-wt sh885

–

+

OGDHR509K

DOX Reexpression vector

Pool total (AU x 1E+6)

_

*

3

Pool total (AU x 1E+6)

Pool total (AU x 1E+6)

2

aKG

*

*

Relative isotopomer (AU x 1E+6)

Pool total (AU x 1E+6)

Aspartate

*

4

0

Relative isotopomer (AU x 1E+6)

*

*

Relative isotopomer (AU x 1E+6)

Malate 6

aKG

1.0

*

*

0.8

ns

#

0.6 0.4 0.2 0

–

+

Empty

–

+

OGDH-wt

–

+

OGDHR509K

DOX Reexpression vector

sh885

Figure 2. OGDH Is Necessary for TCA Cycle Metabolism In Vitro and In Vivo (A) OGDH knockdown affects steady-state pool size of multiple metabolites in vitro. (B) OGDH knockdown affects 13C5-glutamine labeling patterns in vitro. (C) Xenograft growth of A549 cells expressing sh885 with several versions of the overexpression vector: empty control, OGDH wild-type rescue, OGDH R509K catalytically dead mutant. Error bars represent SEM. Percentage of tumor growth inhibition (TGI) is indicated on each of the panels, along with the corresponding p value. (D) Immunoblot analysis showing effective OGDH knockdown at the end of the in vivo efficacy study. Five individual tumor samples from each cohort are shown. (E) Alpha-ketoglutarate dehydrogenase enzymatic activity assay conducted on tumors shown in (D). (F) OGDH knockdown affects steady-state pool size of multiple metabolites in vivo. (A, B, E, and F) Error bars represent the SD from a representative experiment. *p < 0.01, #p < 0.05 (Student’s t test; n = 3).

Cell Reports 17, 876–890, October 11, 2016 879

translation (Breslin and O’Driscoll, 2013). To further characterize these stable cell lines in assays of increasing stringency, we utilized a modified soft agar assay format, m3D (Figure 3A). Given that multiple agents have shown diminished potency in the context of 3D cultures, we intended to introduce the OGDH knockdown only in the setting of already established colonies. First, we plated the cells without doxycycline, thus keeping the shRNA expression off, and allowed the colonies to form to roughly 100- to 120-mm size. At that time, doxycycline was added, and the cells were further incubated in the presence of doxycycline until the colonies in the control plates were clearly visible to the naked eye (500–700 mm). While the classic method of scoring the effects on growth relies on comparing the experimental arms with doxycycline to those without doxycycline, we focused on comparing the size of the preformed colonies at the start of the assay to the colonies with doxycycline at the end of the assay. Strong hits were defined as cell lines that did not show any evidence of further growth after doxycycline addition. We constructed three stable lines (shNT, sh815, sh885) in each of the 59 cell lines and observed a broad range of the sensitivities in multiple assays (Figures 3B and 3C). Because not all the cell lines were capable of forming colonies in soft agar, the direct comparison of 2D versus m3D effects was possible only in 38 cell lines. Interestingly, although most of the cell lines were sensitive to OGDH knockdown in a 2D assay, and thus termed OGDH-dependent, significantly fewer lines displayed OGDH dependence in the context of the m3D assay. In fact, when using the most stringent criteria of looking for absolute growth after doxycycline addition to preformed colonies, only four cell lines were scored as OGDH dependent. Representative images of multiple cell lines within each category are shown in Figure 3B. Importantly, a similar extent of OGDH knockdown was observed in the vast majority of cell lines (Figure 3D; data not shown), as well as in both 2D and m3D assay formats (Figure S3A). Interestingly, some cell lines appear to display higher levels of OGDH expression in the context of soft agar colonies (Figure S3B), but this observation does not appear to correlate with sensitivity to OGDH knockdown. To further show that differential effects on cell growth were due to inherent sensitivities rather than to differential extent of OGDH inhibition, we conducted 13C5-glutamine labeling experiments in several representative cell lines. Consistent with comparable OGDH depletion in these cell lines, the TCA cycle intermediates were similarly affected showing significant decreases in 13C4-malate, aspartate, and citrate isotopomers. While some cell lines displayed effects on 13C5-glutamate isotopomer levels, those were not correlated with sensitivity to OGDH knockdown (Figure 3E). Collectively, these data suggest that, while OGDH activity is comparably inhibited in OGDH-dependent and -independent cell lines, the cellular response can be dramatically different. To functionally characterize the cellular consequence of OGDH knockdown in OGDH-dependent cells, we have conducted a time-course experiment analyzing growth of 22RV1 and H508 cells lines (Figure S3C). OGDH knockdown resulted in profound growth defects and cell-cycle abnormalities, such as a decrease in S phase population and a concurrent expansion of

880 Cell Reports 17, 876–890, October 11, 2016

the subG1 population, suggesting that both cell death and reduced proliferation contribute to the phenotype (Figure S3D). Characterization of Putative alpha-Ketoglutarate Dehydrogenase Inhibitors Using the highly annotated set of OGDH-dependent and OGDHindependent cell lines, we evaluated published putative small molecule inhibitors of alpha-ketoglutarate dehydrogenase complex, succinyl phosphonate (SP), and CPI-613 (Bunik et al., 2005; Stuart et al., 2014). Using several different criteria, we showed that these molecules cannot be used as specific OGDHC inhibitors in cell-based assays. First, dose titration of SP and CPI-613 showed similar sensitivity of representative OGDH-dependent and OGDH-independent cell lines, A549 and H838, respectively (Figure S4A). Second, metabolomics studies of A549 cells treated with SP clearly showed no major impact on overall pool levels of TCA cycle intermediates or labeling patterns using 13 C5-glutamine tracer (Figure S4B). Treatment of A549 cells with CPI-613 resulted in significant TCA cycle perturbations, as judged by overall metabolite pool levels and 13C5-glutamine tracer labeling patterns (Figure S4C), but those were inconsistent with the shRNA studies shown in Figure 2. Third, we assessed the effects of these compounds on OGDHC enzymatic activity in A549 cell lysates. Although SP was able to inhibit OGDHC lysate activity with IC50 of 3.9 mM, CPI-613 was able to do it only at 248 mM (Figure S4D). Collectively, these results point to some of the key limitations of these tool compounds (cell penetration for SP and lack of specificity for CPI-613) and highlight the insights that can be gained when using the highly validated shRNA tools. In fact, CPI-613 has been previously described to also inhibit pyruvate dehydrogenase (Zachar et al., 2011). In Vivo Dependence on OGDH To investigate the in vitro to in vivo translation of our findings, we conducted multiple xenograft efficacy studies. One of the most sensitive cell lines to OGDH knockdown in the stringent m3D assay, 22RV1, showed a dramatic response in vivo (Figure 4A). To further demonstrate the specificity of this response, we conducted an expanded study with additional shRNAs and a wildtype OGDH rescue arm of sh885 (Figure S5A). Similar efficacy was observed with additional shRNAs and, as with the in vitro studies, functional rescue of sh885 cell line was observed with overexpression of wild-type OGDH (Figure S5B). Extending these results, striking in vivo dependence on OGDH was also demonstrated for H508 cells (Figure 4B). Additional cell lines evaluated (HEC1A and HT1080) gave similar results to A549 cells (data not shown) or showed no efficacy (U87MG) (Figure S5C). Interestingly, when compiling all in vitro and in vivo results, it appeared that the most pronounced in vivo responses tracked with the stringent m3D assay in vitro (Figure 4C). As such, this work highlights the importance of assessing growth in multiple formats and presents a promising in vitro assay that closely tracks with dramatic in vivo responses. Aspartate Is an Alternative Anaplerotic Source for the TCA Cycle To elucidate factors affecting sensitivity to OGDH knockdown, we decided to develop deeper understanding of metabolic

A

C Preformed colonies

Seed

Key Strong growth defect Moderate growth defect No growth defect Does not form soft agar colonies

– DOX

+ DOX Cell Line

B 22RV1

H508

A549

HEC1A

HT1080

U87MG

HCT15

DLD1

– DOX

2D + DOX

Seed Preformed colonies

3D

+ DOX

– DOX

D

H508

22RV1

–

+

–

A549

+

HEC1A

–

+

–

+

DOX OGDH -actin

HT1080

–

U87MG

+

–

DLD1

+

–

HCT15

+

–

+

DOX OGDH -actin

13

C4 Malate

*

*

*

*

*

120 80 40

50 A5 8 H 49 EC H 1A T1 0 U 80 87 M G D LD H 1 C T1 5

160

H

22

RV 1

− + − + − + − + − + − + − + − + DOX

160

*

*

*

0

− + − + − + − + − + − + − + − + DOX

13

*

*

*

120 80 40 0

50 A5 8 H 49 EC H 1A T1 0 U 80 87 M G D LD H 1 C T1 5

H

22

RV 1

− + − + − + − + − + − + − + − + DOX

*

#

*

C5 Glutamate

ns

*

*

ns

#

*

120 80 40 0

− + − + − + − + − + − + − + − + DOX

HCC1500 A427 SW1573 NCI-H1048 TCCSUP Detroit 562 T84 MDA-MB-453 MDA-MB-361 Ca Ski SW579 BXPC3 HCC827 HT-1197 H1975 HCC1954 OVISE TOV21G H838 H1563 HCC202

50 A5 8 H 49 EC H 1A T1 0 U 80 87 M G D LD H 1 C T1 5

*

*

40

RV 1

C4 Citrate

*

H

*

*

80

22

*

*

120

13

*

*

RV H 1 50 8 A5 4 H 9 EC H 1A T1 0 U 80 87 M D G LD H 1 C T1 5

0

m3D

C4 Aspartate

Change with OGDH KD (%)

*

22

160

*

*

Change with OGDH KD (%)

Change with OGDH KD (%)

160

Change with OGDH KD (%)

13

E

2D

22RV1 BT-474 H508 MOLT4 A549 HEC1A HT1080 BT-20 KS1 H1703 T47D RKO AGS SK-OV-3 H460 MCF7 LS180 C-33 A H1395 H520 U87MG J82 HT29 U937 GP2D H2126 H1155 MDA-MB-468 P3HR-1 HT SK-UT-1 HCT116 KNS42 CA46 Raji HMCB HCT15 DLD1

(legend on next page)

Cell Reports 17, 876–890, October 11, 2016 881

A Tumor volume (mm3)

2500

22RV1 shNT – DOX 22RV1 shNT + DOX 22RV1 sh885 – DOX 22RV1 sh885 + DOX

2000 1500 1000 500 0 0

4

8

12

16 Days

20

24

28

32

16 Days

20

24

28

32

B Tumor volume (mm3)

1600 H508 shNT – DOX H508 shNT + DOX H508 sh885 – DOX H508 sh885 + DOX

1400 1200 1000 800 600 400 200 0 0

4

8

12

C 2D 22RV1 H508 HEC1A A549 HT1080 U87MG

m3D

% TGI in vivo

p-values

104 / 91 99 / 98 47 / 41 40 / 49 54 / 61 0

1.20e-9 / 5.40e-7 3.04e-16 / 3.08e-12 1.98e-5 / 1.13e-2 4.16e-2 / 1.72e-4 5.00e-4 / 6.00e-6 NS

Figure 4. A Subset of Cancer Cell Lines Displays Profound OGDH Dependence In Vivo (A) Xenograft growth of 22RV1 cells engineered to harbor sh885 and shNT constructs. Doxycycline was added to induce shRNA expression once tumors were approximately 200 mm3. Error bars represent SEM. (B) Xenograft growth of H508 cells engineered to harbor sh885 and shNT constructs. Doxycycline was added to induce shRNA expression once tumors were approximately 200 mm3. Error bars represent SEM. (C) Correlation between different in vitro assay formats and the in vivo efficacy. Percentage of tumor growth inhibition (TGI) and the associated p values are shown from two independent replicate experiments, where available. NS, no statistical significance.

TCA cycle inputs and outputs. In particular, we focused on alternative anaplerotic TCA cycle sources in OGDH-independent cells where such pathways are predicted to be active. We chose H838 cells because they showed clear OGDH independence and, using a CRISPR system, we derived KO clones for DLST, a unique E2 subunit of alpha-ketoglutarate dehydrogenase complex. Multiple individual clones showed a clear loss of DLST protein and one was chosen for further characterization (Figure 5A).

Consistent with the initial OGDH knockdown results, the DLSTKO cells showed only a very mild proliferation defect phenotype (Figure 5B). To demonstrate that the cells lacked an OGDHC activity, we conducted a 13C5-glutamine labeling experiment and showed that despite comparable levels of 13C5-glutamate and 13 C5-aKG in both cell lines, no 13C4-fumarate, 13C4-malate, or 13 C4-aspartate was formed (Figure 5C). Importantly, OGDH knockdown resulted in qualitatively similar 13C5-glutamine labeling results of most metabolites consistent with only partial loss of OGDHC function (Figure S6A). Interestingly, we observed a difference in 13C5-aKG pattern in the context of DLST-KO cells versus OGDH knockdown that might reflect differences in chronic versus acute loss of OGDHC function, respectively (Figures 5C and S6A). Therefore, the above results suggest that DLST-KO cells represent a useful tool of cells with a complete break in the oxidative TCA cycle, while confirming that DLST is a non-redundant member of the OGDHC complex. The DLST-KO cells provided a unique opportunity to investigate additional anaplerotic carbon sources that may be used by OGDH-independent cells in the event of TCA cycle disruption. As such, uniformly 13C-labeled glucose, glutamine, aspartate, tyrosine, arginine, lysine, isoleucine, leucine, and valine were individually supplied for 4 hr to H838 wild-type and DLST-KO cells, and labeling patterns were assessed based on predicted carbon flow pathways (Figure 5D). DLST-KO cells showed a greater percentage of 13C3-malate derived from 13C6-glucose, suggestive of a differential pyruvate carboxylase involvement (Figure S6B). However, the absolute abundance of 13C3-malate was not changed in these cells, and the signal was thus primarily driven by decreased overall malate pool levels in DLST-KO cells. Differential pyruvate dehydrogenase involvement was also suggested by the dramatic decrease in 13C2-citrate levels (Figure S6C). Interestingly, total pools of citrate remained the same in both cell lines, suggesting an alternative anaplerotic source being used. No difference between the two cell lines and very little net flow of carbon to the TCA cycle were observed with leucine, isoleucine, valine, tyrosine, and lysine. In contrast, there was substantial contribution from 13C4-asparate to TCA cycle carbon pools as shown for malate, and the fractional contribution from aspartate was increased by 2.5-fold upon DLST knockout (Figure 5E). While the total levels of malate were somewhat reduced in DLST-KO cells, the 13C4-malate abundance actually increased, suggesting more aspartate-derived carbon was converted to malate. To further elucidate the mechanisms by which aspartate contributes to TCA cycle, we used 15N,13C4-aspartate labeling experiments. Specifically, we examined three possible aspartate

Figure 3. Cancer Cells Display a Broad Range of Sensitivities to OGDH Knockdown (A) Schematic representation of the modified 3D soft agar assay used (m3D). 22RV1 cells with OGDH knockdown (sh885) are shown as an example. (B) Representative images of cells grown in 2D and m3D assays are shown to illustrate differential sensitivities to OGDH knockdown. All cell lines displayed harbor the sh885 construct allowing inducible OGDH knockdown. Colored bars on the bottom of the panel denote clustering of cells based on m3D results. Detailed legend can be found in (C). (C) Summary of OGDH sensitivities in 2D and m3D assays observed across 59 cancer cell lines. (D) Immunoblot analysis showing similar levels of OGDH knockdown in cell lines of different OGDH sensitivities. All cell lines displayed harbor the sh885 construct allowing inducible OGDH knockdown. Expression was monitored after 4 days of dox treatment. (E) TCA cycle is similarly affected by OGDH knockdown in cell lines of different OGDH sensitivities as demonstrated by a 13C5-glutamine labeling experiment. All cell lines displayed harbor the sh885 construct allowing inducible OGDH knockdown. Results are plotted as percentage of change as compared to the same cell line control without dox treatment. Error bars represent the SD from a representative experiment. *p < 0.01, #p < 0.05 (Student’s t test; n = 3).

882 Cell Reports 17, 876–890, October 11, 2016

A w

t ild

yp

e

GLUCOSE

D

O

-K ST

LYSINE LEUCINE

DL

AcetylCoA

ASPARTATE

DLST

ISOLEUCINE

Citrate

Oxaloacetate GAPDH

Malate

B

Isocitrate

#

6

Fold growth in 72 hours

wild type DLST KO 4

Fumarate

GLUTAMINE

aKG OGDH complex

TYROSINE ASPARTATE

2

Succinate

ISOLEUCINE

SuccinylCoA

VALINE

0

10

300

13

C4 Fumarate

wild type

0

50

0

*

30

20

10

0

0

13

C4 Malate

*

5 0

25 20

*

0

F

C4 Malate

G 80

wild type DLST KO

15 10 5 0

Relative isotopomer (AU x 1E+6)

100

20 0.5

2

Isotopomer (%)

150

DLST KO

10

4

13

*

15

1.0

Malate

C4 Aspartate

20

40

6

E

*

60

1.5

200 100

13

C4 Malate

*

2.0

8

0

Pool total (AU x 1E+6)

ns

Relative isotopomer (AU x 1E+6)

Relative isotopomer (AU x 1E+6)

C5 aKG

ns

400

13

13

C5 Glutamate

2. 15N IMP

60

3. 15N 13C4 Argininosuccinate

40 20 0

1

2

3

15

15

1. 15N Glutamate Isotopomer (%)

13

C

N Glutamate

*

wild type DLST KO

10

5

0

Figure 5. Aspartate Is an Alternative Anaplerotic Source for the TCA Cycle (A) Immunoblot analysis showing no DLST expression in the DLST-KO cells in comparison to H838 wild-type cells. (B) Relative growth rates of wild-type H838 and DLST-KO cells. (C) DLST-KO cells display no OGDHC activity, as determined by 13C5-glutamine-labeling patterns. (D) Schematic illustration of the labeling experiment showing potential anaplerotic sources of carbon into the TCA cycle. (E) Analysis of malate illustrates differential handling of 13C4-aspartate in DLST-KO cells as compared to the H838 wild-type cells. Total malate pool size, 13 C4-malate, and percentage of 13C4-malate isotopomer abundance are shown. (F) Diagnostic 15N transfer onto glutamate in the 15N,13C4-aspartate labeling experiment, indicating significant malate-aspartate shuttle activity in H838 wild-type cells. (G) DLST-KO cells display higher level of malate-aspartate shuttle activity than H838 wild-type cells, as measurement by percentage of 15N-glutamate isotopomer abundance after 15N,13C4-aspartate labeling. (B–G) Error bars represent the SD from a representative experiment. *p < 0.01, #p < 0.05 (Student’s t test; n = 3).

pathways, each potentially generating 13C4-fumarate or 13C4-malate, and followed the 15N transfer as a diagnostic marker of the specific pathway utilized (urea cycle usage generating 15 13 N, C4-argininosuccinate and ultimately 15N-arginine; malateaspartate shuttle usage generating 15N-glutamate; purine biosynthetic pathway usage generating 15N-AICAR and 15N-IMP) (Figures S6D–S6F). We observed clear evidence of malate-aspartate

shuttle activity and no evidence of the other two pathways being involved in parental H838 cells (Figure 5F). To directly compare the malate-aspartate shuttle activity in DLST wild-type and DLST-KO cells, we used 15N,13C4-aspartate labeling experiments. DLST-KO cells displayed significantly higher levels of 15 N-glutamate isotopomer as compared to DLST wild-type cells (Figure 5G), consistent with the initial result observed in carbon

Cell Reports 17, 876–890, October 11, 2016 883

tracing studies (Figure 5E). Collectively, these results highlight the cellular flexibility in using anaplerotic sources and suggest that the ability to utilize aspartate might represent a compensatory mechanism in situations where TCA cycle is interrupted, in this case at the alpha-ketoglutarate dehydrogenase step. Differential Aspartate Utilization by OGDH-Dependent and OGDH-Independent Cell Lines To further investigate the differential utilization of aspartate, we screened a number of cell lines with 15N,13C4-aspartate and monitored the incorporation of 15N-label into 15N,13C4-argininosuccinate, 15N-IMP, and 15N-glutamate. Consistent with our initial results, we observed a correlation between cell lines that displayed pronounced OGDH dependence in the m3D assay and the decreased levels of the malate-aspartate shuttle activity (Figure 6A). This correlation was further confirmed by investigating the fate of carbon derived from aspartate by specifically monitoring 13C4-citrate isotopomer levels (Figure 6B). The differences between cell lines were not accounted for by differential ability to take up the 15N,13C4-aspartate from the media (Figure 6C). Similarly to the results observed in H838 cells, we did not observe any 15N-label incorporation into 15N,13C4-argininosuccinate and 15N-IMP (data not shown). Importantly, the cells that were dependent on OGDH in 2D, but became less dependent on OGDH in the m3D assays, still displayed a significant level of 15N-glutamate and 13C4-citrate signal and hence tracked with the OGDH-independent group. This observation is noteworthy as it suggests the feasibility of an expanded nutrient utilization screen to identify additional OGDH-dependent cell lines, a preferable option over the labor-intensive shRNA-based approach. To specifically test the contribution of aspartate to the OGDH phenotypes observed, we conducted a series of experiments involving alterations of media composition. As predicted, supplementation of media with high levels of aspartate was able to rescue growth defects conferred by OGDH knockdown in multiple cell lines that were sensitive to OGDH knockdown in 2D and that maintained some basal level of malate-aspartate shuttle activity (Figure 6D; data not shown). Interestingly, some minimal threshold of malate-aspartate shuttle activity might be required, as aspartate rescue did not effectively rescue OGDH-mediated growth defects in H508 cells (data not shown). Conversely, we showed that OGDH-independent cells, such DLST-KO, are exquisitely sensitive to media aspartate levels, as they displayed a profound growth defect in the context of aspartate dropout media (Figure 6E). To demonstrate that this role of aspartate requires functional malate-aspartate shuttle, we knocked down its critical components—glutamic-oxaloacetic transaminases, GOT1 and GOT2. In the context of media containing aspartate, we observed specific growth defects only in DLST-KO cells upon GOT2 knockdown (Figure 6F). Since the levels of GOT1 and GOT2 knockdown were roughly equivalent in both cell lines (Figure 6G), we undertook a metabolomics approach to dissect the differences between GOT1 and GOT2. We observed the expected increases in overall aspartate total pool levels upon GOT1 knockdown and decreases upon GOT2 knockdown (Figure 6H), consistent with the previously observed aspartate-consuming role of GOT1

884 Cell Reports 17, 876–890, October 11, 2016

and aspartate-producing role of GOT2 (Son et al., 2013). Next, we conducted 15N,13C4-aspartate labeling experiments to monitor the levels of 15N-label transfer onto glutamate and observed modest effects upon GOT2 knockdown (Figure 6I). Since the interpretation of these results is complicated by the fact that the assay does not distinguish between the individual aminotransferases, we employed an alternative 13C5-glutamine labeling strategy to monitor 13C4-aspartate levels. Similarly to total aspartate pool levels, GOT1 knockdown increased 13 C4-aspartate isotopomer abundance, and GOT2 knockdown decreased 13C4-aspartate isotopomer abundance (Figure 6J). Interestingly, the lowest levels of aspartate were observed in DLST-KO cells upon GOT2 knockdown, thus correlating with growth defects observed (Figure 6F). Collectively, these data support the critical role of aspartate, and the malateaspartate shuttle, in mediating cellular response to OGDH knockdown. Differential Effects of OGDH Knockdown on Cellular Respiration To functionally demonstrate the aspartate contribution to respiration, we directly measured the mitochondrial oxygen consumption rate (OCR) of several OGDH-dependent and -independent cell lines. Importantly, the measurements were done in RPMI media containing much higher concentrations of glucose and glutamine as compared to aspartate (11 mM, 2 mM, 150 mM, respectively). Several observations highlight the fundamental differences between OGDH-dependent and -independent cell lines. First, we observed a significantly higher basal respiration rate in OGDH-dependent cell lines (22RV1 and H508) as compared to the OGDH-independent ones (HCT15, DLD1, and H838) (Figure 7A). Second, the OGDH-dependent cell lines tended to have a larger decrease in basal OCR upon OGDH knockdown. This relative impact of OGDH knockdown was also observed in OGDH-dependent cell lines with more intermediate levels basal OCR (A549 and AGS), further supporting their particular reliance on OGDHC activity. To further understand the respiration in these cells we used FCCP, a well-established uncoupling agent that collapses the proton gradient. The result of mitochondrial membrane potential disruption is uninhibited electron flow through the electron transport chain (ETC) and maximal oxygen consumption by complex IV. The difference between basal respiration and FCCP-stimulated respiration can be used to assess reserve respiratory capacity, a measure of cellular ability to respond to increased energy demand. Interestingly, we observed a larger response to FCCP in OGDH-dependent cell lines, and, most importantly, OGDH knockdown almost completely abrogated this response (Figure 7B). The OGDH-independent cell lines displayed significant residual reserve respiratory capacity after OGDH knockdown, consistent with additional mechanisms maintaining the TCA cycle and ETC activities. To evaluate the relative contribution of the malate-aspartate shuttle to this activity, we used a previously described aminotransferase inhibitor aminooxyacetate, AOA (Son et al., 2013). As predicted, we observed a significant reduction of FCCP-stimulated OCR by AOA in OGDH-independent cells (DLD1 and HCT15), as assessed in cells with OGDH knockdown (Figure 7C). Interestingly, even at the highest

15

A

13

B

N Glutamate p = 2.2 x 10-6

2 1 0

60

Isotopomer (%)

Isotopomer (%)

6 4 2

40

20

0

siNT siGOT1 siGOT2

100 75 50 25

G

dropout media + aspartate

+ DOX

dropout media – aspartate

100

50

0

H

H838 WT

I

Aspartate

*

*

*

15

*

8

4

siNT siGOT1 siGOT2

6

*

H838 WT

H838 DLST-KO

100

50

H838 DLST-KO

*

ns *

siNT siGOT1 siGOT2

4

2

H838 WT

H838 WT

13

J

N Glutamate

0

0

siNT siGOT1 siGOT2

150

0

H838 DLST-KO

Isotopomer (%)

H838 WT

H838 DLST-KO

– DOX

150

0

12

H838 WT

: aspartate

+

Relative GOT2 expression levels (%)

–

F

E

22RV1 (sh885)

+

Relative GOT1 expression levels (%)

–

22 R BT V1 47 H 4 M 50 O 8 HE LT C1 4 A A HT 54 1 9 U8 08 7M 0 H4 G HC 60 T D 15 LD H TO 8 1 3 HCV21 8 C2 G 02

22 R BT V1 47 H 4 M 50 O 8 L HE T C1 4 A A H 54 T1 9 U8 08 7M 0 H G HC 460 T DL 15 D TO H8 1 3 HCV21 8 C2 G 02

22 R BT V1 47 H 4 M 50 O 8 HE LT C1 4 A A HT 54 1 9 U8 08 7M 0 H G HC 460 T DL 15 D TO H8 1 3 HCV21 8 C2 G 02

A549 (sh885)

0

Relative isotopomer (AU)

Isotopomer (%)

3

D

N 13C4 Aspartate

p = 0.0016

8

4

Relative growth (%)

15

p = 5.9 x 10-7

5

Pool total (AU x 1E+8)

C

C4 Citrate

109 10

*

H838 DLST-KO

C4 Aspartate

*

*

*

8

siNT siGOT1 siGOT2

107 106 105

H838 WT

H838 DLST-KO

H838 DLST-KO

Figure 6. Differential Aspartate Utilization in OGDH-Dependent and OGDH-Independent Cell Lines (A) Aspartate utilization in vitro correlates with OGDH sensitivity in vivo. 15N,13C4-aspartate labeling was used to assess the malate-aspartate shuttle activity by following the diagnostic 15N-transfer to glutamate. (B) Alternative readout of malate-aspartate shuttle activity confirming carbon labeling of the TCA cycle intermediates. 15N,13C4-aspartate labeling was used to show the appearance of 13C4-citrate. (C) Percentage of 15N,13C4-aspartate uptake across multiple cell lines. (D) Aspartate can rescue growth defects induced by OGDH knockdown. OGDH-dependent cell lines were plated in 2D colony formation assay in regular RPMI media and supplemented with additional aspartate. (E) DLST-KO cells are particularly sensitive to media aspartate levels. OGDH-independent cells, H838 wild-type (WT) and H838 DLST-KO, were plated in 2D colony formation assays in a dropout media specifically testing the role of aspartate.

(legend continued on next page)

Cell Reports 17, 876–890, October 11, 2016 885

AOA concentrations tested we observed no effects on basal respiration rates in any of the cell lines tested (data not shown). These significant differences in how OGDH-dependent and OGDH-independent cells respire, prompted us to conduct several additional controls. First, we measured overall ATP levels and showed no differences between OGDH-dependent and OGDH-independent cell lines upon OGDH knockdown (Figure S7A). Second, we measured ROS levels using CMH2DCFDA dye and showed significantly higher basal ROS levels in OGDH-dependent cells compared to OGDH-independent cells, perhaps consistent with the differential oxygen consumption rates observed. Interestingly, we observed no significant differences in ROS levels upon OGDH knockdown (Figure S7B). Similarly, we investigated mitochondrial superoxide levels using MitoSOX dye and observed no significant differences upon OGDH knockdown (Figure S7C). Third, we measured NAD+/ NADH ratio and found that OGDH knockdown resulted in preferential decrease of NAD+/NADH ratio in OGDH-dependent cells (Figure S7D). Since the OGDHC reaction generates mitochondrial NADH, this result was unexpected and is likely to be an indirect reflection of the overall cellular energetic state of OGDHdependent cells responding to target knockdown rather than a direct readout of alpha-ketoglutarate dehydrogenase reaction. Collectively, the above results support the specificity of the observed effects on cellular respiration and are consistent with a model where OGDH-independent cells can use multiple mechanisms to maintain respiration, including the malate-aspartate shuttle, while the OGDH-dependent cell lines lack this metabolic flexibility (Figure 7D).

DISCUSSION Alpha-ketoglutarate dehydrogenase complex is composed of three well-described subunits: E1 encoded by OGDH, E2 encoded by DLST, and E3 encoded by DLD. An additional subunit of the OGDH complex, Kgd4, was recently identified in yeast and shown to be necessary for stable complex formation (Heublein et al., 2014). While mechanisms regulating the biochemical reaction catalyzed by OGDHC are beginning to emerge (Armstrong et al., 2014; Bunik and Fernie, 2009), the cellular and organismal roles of OGDHC remain relatively poorly defined. Profound loss of OGDHC function appears to be deleterious, as evidenced by embryonic lethality of Dlst KO mice (Yang et al., 2009), Dld KO mice (Johnson et al., 1997), and aKG aciduria found in human patients with severely decreased levels of alpha-ketoglutarate dehydrogenase activity (Dunckelmann et al., 2000).

Given the central role of the alpha-ketoglutarate dehydrogenase complex, it was surprising to uncover a subset of cell lines that could maintain growth in the setting of near complete OGDH knockdown. A greater range of cellular sensitivities to OGDH knockdown was observed in the in vitro m3D setting, where in fact only a small minority of cell lines displayed dramatic dependence on OGDH. Notably, the m3D growth data closely tracked with in vivo dependence on OGDH, highlighting that metabolic wiring is highly context dependent. This point was recently illustrated by several studies demonstrating alterations in the expression of metabolic genes (including OGDH) and drug sensitivity from the 2D to 3D cell culture setting (Mandujano-Tinoco et al., 2013; Vinci et al., 2012). The metabolic flexibility of tumor cells is exemplified by the DLST knockout experiment, in which cells were capable of proliferation despite a complete break in the TCA cycle, further supporting an emerging notion that the these reactions are not a simple unidirectional cycle. Other examples include conditions where HIF-1a is stabilized (Metallo et al., 2012; Wise et al., 2011) or the mitochondrial electron transport system is defective (Mullen et al., 2011), in which conversion of aKG to isocitrate via reductive carboxylation is particularly prominent as compared to the oxidative TCA cycle direction. The hypoxia-induced switch in carbon flow is in part controlled by HIF-1 controlled ubiquitin-mediated degradation of an OGDH splice variant (Sun and Denko, 2014), shunting carbon into citrate for lipid synthesis. Surprisingly, OGDH activity was also demonstrated to help drive reductive carboxylation by providing NADH as a substrate for nicotinamide nucleotide transhydrogenase (NNT)-dependent NADPH formation (Mullen et al., 2014). This report adds another dimension to our understanding of metabolic flexibility by highlighting differential utilization of the malate-aspartate shuttle, a system thought primarily to shuttle reducing equivalents from the cytosol into the mitochondria (Figure S6F). Inactive or low malate-aspartate shuttle activity could deprive the mitochondria of the cytosol-derived NADH sources, such as glycolysis. In turn, these cells could be particularly dependent on mitochondrial sources of NADH. The ability to deliver aspartate carbon into the mitochondria also provides cells with a potential anaplerotic source in the setting of disrupted TCA cycle carbon flow. While the mechanistic link between low malate-aspartate shuttle activity and dependence on OGDH remains to be fully elucidated, our data are consistent with a model wherein aspartate plays a dual role of providing NADH to the mitochondria and carbon equivalents to the TCA cycle. Further dissection of malate-aspartate shuttle involvement will require future development of label tracing methods

(F) DLST-KO cells display specific sensitivity to GOT2 knockdown in comparison to H838 WT cells. Error bars represent the SD of triplicate wells from a representative experiment. (G) GOT1 and GOT2 knockdown levels were tested by qPCR in H838 WT and H838 DLST-KO cells. Error bars represent the SD of triplicate measurement from a representative experiment. (H) GOT1 and GOT2 knockdown effects on total aspartate pool levels. (I) Aspartate aminotransferase activity, as measured by percentage of 15N-glutamate isotopomer abundance after 15N,13C4-aspartate labeling. (J) GOT1 and GOT2 knockdown effects on 13C4-aspartate relative isotopomer levels in 13C5-glutamine labeling experiment. The 13C4-aspartate isotopomer values are plotted on a log scale to allow comparison across a wide concentration range. (A–C) Cell lines are color-coded according to their phenotype in m3D assay, further described in the legend of Figure 3C. Error bars represent the SD of triplicate wells from a representative experiment. Student’s t test was performed, as indicated by the brackets. (H–J) Error bars represent the SD from a representative experiment. *p < 0.01, #p < 0.05 (Student’s t test; n = 3).

886 Cell Reports 17, 876–890, October 11, 2016

+ DOX

*

*

1

22

– DOX

1

0 0 39 1. 56 6. 25 25 10 0 0.

FCCP-stimulated OCR (pmol / min / 1E3 cells)

10

+ DOX ns ns

0

– DOX

15

+ DOX ns

10 #

5 0 0 39 1. 56 6. 25 25 10 0

25 25 10 0

56

H508 (sh885)

20

0.

6.

1.

0.

0 39

-5 0 0. 39 1. 56 6. 25 25 10 0

FCCP-stimulated OCR (pmol / min / 1E3 cells)

– DOX

AOA (µM)

AOA (µM)

D

*

*

AOA (µM)

22RV1 (sh885)

5

− +

+ DOX

2

AOA (µM) 15

− +

HCT15 (sh885)

3

25 10 0

1.

0.

0 0. 39 1. 56 6. 25

0

− +

25 10 0

2

− +

5

8 50 H

FCCP-stimulated OCR (pmol / min / 1E3 cells)

3

− +

− +

T1

5 T1

− +

DOX

C

– DOX

*

0 -2

H

D

LD

8 83

DLD1 (sh885)

4

0 39 56 6. 25 25 10 0

FCCP-stimulated OCR (pmol / min / 1E3 cells)

C

56%

*

C

− +

58%

1

− +

H

22

− + AG S

8 50 H

− + 49

− +

A5

− +

32% #

H

− +

5

*

LD

*

D

*

100%

8 5

8

22%

*

10

83

13%

1

*

91%

*

H

*

29%

100%

AG S

55%

15

0 0. 39 1. 56 6. 25

40% #

*

49

*

OGDH-independent in 2D

84%

A5

30%

10

0

OGDH-dependent in 2D

FCCP-stimulated OCR (pmol / min / 1E3 cells)

*

15

DOX

B 20

RV 1

Basal OCR (pmol / min / 1E3 cells)

OGDH-independent in 2D

0 0. 39 1. 56 6. 25 25 10 0

OGDH-dependent in 2D 55%

20

RV 1

A

OGDH-dependent cells

OGDH-independent cells

MAS

MAS TCA

TCA OGDH

OGDH

Figure 7. OGDH Knockdown Differentially Affects Respiration of OGDH-Dependent and OGDH-Independent Cell Lines (A) Basal oxygen consumption rates in multiple cell lines harboring sh885, as a function of OGDH knockdown. Percentages displayed indicate cell line-specific decreases in respiration after OGDH knockdown. Cell lines are color-coded according to their phenotype in m3D assay, further described in the legend of Figure 3C, and their OGDH dependence in 2D assays is denoted by the bracket above the panel. (B) FCCP-stimulated oxygen consumption rates in multiple cell lines harboring sh885, as a function of OGDH knockdown. FCCP-stimulated OCR was calculated based on the OCR difference before and after FCCP injection. Percentages displayed indicate cell-line-specific decreases in respiration after OGDH knockdown. Cell lines are color-coded according to their phenotype in m3D assay, further described in the legend of Figure 3C, and their OGDH dependence in 2D assays is denoted by the bracket above the panel.

(legend continued on next page)

Cell Reports 17, 876–890, October 11, 2016 887

to specifically track cytosolic and mitochondrial aspartate aminotransferase (GOT1 versus GOT2) and malate dehydrogenase (MDH1 versus MDH2) reactions. This is of particular interest, as some reports have suggested the possibility of reverse reaction direction, as driven by relative concentrations of individual intermediates and mitochondrial membrane potential in specific cellular contexts (Birsoy et al., 2015; Bremer and Davis, 1975). Identifying the mechanisms that control malate-aspartate shuttle activity will be of high interest, as such findings may lead to the development of clinically tractable markers for identifying tumors with this particular metabolic vulnerability. Recent studies have just begun to reveal such potential modes of malate-aspartate shuttle regulation by identifying a stimulatory role of GOT2 acetylation by SIRT3 (Yang et al., 2015). This future direction will be of particular importance, as no genetic lesion in cell lines analyzed was clearly associated with response to OGDH knockdown in m3D or in vivo (data not shown). Our initial proof-of-concept observation that OGDH dependence can be modulated by knockdown of malate-aspartate shuttle components, particularly GOT2 (Figure 6F), suggests that tumors with low or absent GOT2 activity might be particularly sensitive to OGDH inhibition. Further in vivo studies will be needed to validate these hypotheses. Recent reports have highlighted the essential role of the ETC in maintaining aspartate biosynthesis in proliferating cells (Birsoy et al., 2015; Sullivan et al., 2015). Our results build upon these studies by demonstrating, in a broader panel of cancer cell lines, that not only is a functional TCA cycle key for generating aspartate, but that a differential ability to utilize aspartate can dictate the ability of cells to survive and grow in the setting of a disrupted TCA cycle. Furthermore, the metabolite tracing experiments and oxygen consumption rate measurements suggest that the role of aspartate extends beyond the direct support of cellular biosynthetic needs to more broadly affecting mitochondrial function (Birsoy et al., 2015; Sullivan et al., 2015). Interestingly, our results are consistent with the concept of an aspartate threshold necessary for cellular viability (Birsoy et al., 2015), as the DLST-KO cells upon GOT2 knockdown displayed the lowest levels of aspartate and were the only ones to display growth defects. The circulating levels of aspartate are some of the lowest among all the circulating amino acids (Mayers and Vander Heiden, 2015); thus, the ability to take up and utilize aspartate as an anaplerotic backup mechanism remains to be further tested. Development of specific small molecule inhibitors targeting OGDH and the malate-aspartate shuttle components will enable deeper in vitro and in vivo analyses of these metabolic pathways. Collectively, our findings highlight OGDH as a potential therapeutic target in a metabolically distinct subset of cancers. The analysis of this group of cell lines revealed another mechanism underlying the general robustness and redundancy of metabolism through a role of the malate-aspartate shuttle, particularly

under conditions of TCA cycle disruption. More broadly, our findings linking OGDH dependence to differential aspartate usage suggest a wider utility of nutrient utilization screens to identify phenotypically distinct cancer subsets and to identify additional metabolic nodes for targeted cancer therapy. EXPERIMENTAL PROCEDURES Cell Lines All cell lines described in these studies were obtained from the American Type Culture Collection (ATCC), Japanese Collection of Research Bioresources (JCRB) Cell Bank, or Sigma. Cells were routinely cultured in their preferred media according to vendor recommendations. Cells harboring inducible shRNA constructs were maintained in appropriate media with tetracycline-free fetal bovine serum (631101, Clontech Laboratories). Xenograft Studies Experiments were carried out under an Institutional Animal Care and Use Committee-approved protocol, and institutional guidelines for the proper and humane use of animals were followed. See the Supplemental Experimental Procedures for additional details. siRNA Transfections The general scheme of the transfection protocol of A549 cells is illustrated in Figure S1A. 2e5 cells were plated in 6-cm dishes to achieve 40% confluence at the time of transfection. Cells were transfected with ON-TARGETplus SMARTpool siRNAs (Dharmacon) using Lipofectamine RNAiMAX (13778150, Life Technologies) for 4 hr at 10 nM in OPTI-MEM media (31985-070, Life Technologies). Subsequently, equal volume of cell media was added containing 20% FBS, to achieve 5 nM siRNA and 10% FBS final concentrations, and the transfection was allowed to go on overnight. Next day, media was exchanged over cells for fresh 10% FBS containing media. The following day, the transfection protocol was repeated again. 24 hr after the second transfection, cells were trypsinized, counted, and plated for 96-well-format growth assays. Matching control siRNA, NT, was purchased from Dharmacon (D-00181010-05). The protocol for siRNA transfection of H838 cells was the same as for A549 cells. GOT1 was knocked down using ON-TARGETplus siRNA (J-011673-10, GE Dharmacon). GOT2 was knocked down using ON-TARGETplus siRNA (J-011674-10, Dharmacon). Lentiviral Techniques Lentivirus-based (shRNAs or cDNA overexpression) constructs were made using the standard protocol from The RNAi Consortium (TRC) protocol from the Broad Institute (http://portals.broadinstitute.org/gpp/public/resources/ protocols). shRNA viruses were titered on individual target cell lines and infected at MOI no greater than 0.7. cDNA overexpression viruses were infected at higher MOI, where possible. To infect, cells were centrifuged for 1 hr at 2,250 rpm in the presence of the viruses and 8 mg/ml polybrene (H9268, Sigma). Media was preplaced after the spin, and drug selection was added 24 hr later (puromycin or neomycin, as appropriate). Selection was carried out until uninfected control cells were all dead. See the Supplemental Experimental Procedures for shRNA sequences and additional details. 13 C5-glutamine and 15N,13C4-aspartate Labeling Studies shRNA-containing cells were pre-treated for 3 days with 200 ng/mL doxycycline to induce target gene knockdown. Subsequently, cells were plated at

(C) Malate-aspartate shuttle contribution to FCCP-stimulated oxygen consumption rates was assessed using AOA tool compound. FCCP-stimulated OCR was calculated based on the OCR difference before and after FCCP injection. (D) Schematic illustrating the functional connection of malate-aspartate shuttle (MAS) and the tricarboxylic acid (TCA) cycle. The differential activity of MAS has profound consequences on the ability of cells to proliferate in the conditions of TCA cycle disruption, such as OGDH knockdown. OGDH-dependent cell lines display low levels of MAS, while OGDH-independent cell lines display higher levels of MAS activity. (A–C) Error bars represent the SD of six wells from a representative experiment. *p < 0.01, #p < 0.05 (Student’s t test).

888 Cell Reports 17, 876–890, October 11, 2016

60%–70% confluency in triplicate in fresh media containing the same concentration of doxycycline and incubated overnight. Water was used as the ‘‘-dox’’ control in all the experiments. The next day, media was replaced with a media containing 2 mM 13C5-glutamine. Labeling was done for 3 hr after which the cells were washed three times with 1 3 PBS (pH 7.4), and metabolites were extracted by the addition of ice cold 80% methanol. Cell debris was removed by centrifugation and samples were dried under a stream of nitrogen air. Internal standard (L-Glutamic acid-13C5-15N-d5, 749850, Sigma-Aldrich) at 1 mg/mL was added during the extraction step. 15 N,13C4-aspartate labeling studies were done similarly except a 4 hr incubation period was used. In all labeling studies, the reported amounts of labeled species have been corrected for natural isotope abundance. See the Supplemental Experimental Procedures for detailed description of sample processing, LC-MS, and data analysis. Statistical Analysis Analyses were performed using GraphPad software package. All data are expressed as mean of multiple measurements (n, indicating the number of replicates). Error bars represent SEM for all in vivo studies and SD for all the in vitro studies. Student’s t test was used to assess statistical significance. Exact values and cutoff are specified within each figure or figure legend.

Bunik, V.I., and Fernie, A.R. (2009). Metabolic control exerted by the 2-oxoglutarate dehydrogenase reaction: A cross-kingdom comparison of the crossroad between energy production and nitrogen assimilation. Biochem. J. 422, 405–421. Bunik, V.I., Denton, T.T., Xu, H., Thompson, C.M., Cooper, A.J., and Gibson, G.E. (2005). Phosphonate analogues of alpha-ketoglutarate inhibit the activity of the alpha-ketoglutarate dehydrogenase complex isolated from brain and in cultured cells. Biochemistry 44, 10552–10561. Bunik, V.I., Mkrtchyan, G., Grabarska, A., Oppermann, H., Daloso, D., Araujo, W.L., Juszczak, M., Rzeski, W., Bettendorff, L., Fernie, A.R., et al. (2016). Inhibition of mitochondrial 2-oxoglutarate dehydrogenase impairs viability of cancer cells in a cell-specific metabolism-dependent manner. Oncotarget 7, 26400–26421. Commisso, C., Davidson, S.M., Soydaner-Azeloglu, R.G., Parker, S.J., Kamphorst, J.J., Hackett, S., Grabocka, E., Nofal, M., Drebin, J.A., Thompson, C.B., et al. (2013). Macropinocytosis of protein is an amino acid supply route in Ras-transformed cells. Nature 497, 633–637. Cunningham, J.T., Moreno, M.V., Lodi, A., Ronen, S.M., and Ruggero, D. (2014). Protein and nucleotide biosynthesis are coupled by a single ratelimiting enzyme, PRPS2, to drive cancer. Cell 157, 1088–1103.

SUPPLEMENTAL INFORMATION

Dang, L., White, D.W., Gross, S., Bennett, B.D., Bittinger, M.A., Driggers, E.M., Fantin, V.R., Jang, H.G., Jin, S., Keenan, M.C., et al. (2009). Cancer-associated IDH1 mutations produce 2-hydroxyglutarate. Nature 462, 739–744.

Supplemental Information includes Supplemental Experimental Procedures and seven figures and can be found with this article online at http://dx.doi. org/10.1016/j.celrep.2016.09.052.

Davidson, S.M., Papagiannakopoulos, T., Olenchock, B.A., Heyman, J.E., Keibler, M.A., Luengo, A., Bauer, M.R., Jha, A.K., O’Brien, J.P., Pierce, K.A., et al. (2016). Environment Impacts the Metabolic Dependencies of Ras-Driven NonSmall Cell Lung Cancer. Cell Metab. 23, 517–528.

AUTHOR CONTRIBUTIONS Conceptualization, S.J. and G.A.S.; Methodology, E.L.A., D.B.U., D.P., Y.C., M.F.C., J.M., S.J., and G.A.S.; Investigation, E.L.A., D.B.U., D.P., C.E.M., J.C., Y.S., Y.C., L.H., J.R., S.C., M.F.C., E.A., Z.P.F., and G.C.; Writing – Original Draft, D.B.U., D.P., and G.A.S.; Writing – Review and Editing, D.B.U., D.P., M.F.C., M.D., and G.A.S.; Visualization, E.L.A., D.B.U., D.P., and G.A.S.; Supervision, M.D., S.J., and G.A.S. ACKNOWLEDGMENTS We would like to thank multiple members of the Agios scientific community for their technical assistance and helpful discussions as well as Matthew Vander Heiden for critical review of the manuscript. Editorial assistance with figure creation was provided by Helen Varley, PhD, CMPP, of Excel Scientific Solutions. Indicated authors are employees of and have ownership interest in Agios Pharmaceuticals. This work was supported by Agios Pharmaceuticals. Received: February 21, 2016 Revised: August 18, 2016 Accepted: September 16, 2016 Published: October 11, 2016 REFERENCES Armstrong, C.T., Anderson, J.L., and Denton, R.M. (2014). Studies on the regulation of the human E1 subunit of the 2-oxoglutarate dehydrogenase complex, including the identification of a novel calcium-binding site. Biochem. J. 459, 369–381. Birsoy, K., Wang, T., Chen, W.W., Freinkman, E., Abu-Remaileh, M., and Sabatini, D.M. (2015). An essential role of the mitochondrial electron transport chain in cell proliferation is to enable aspartate synthesis. Cell 162, 540–551. Bremer, J., and Davis, E.J. (1975). Studies on the active transfer of reducing equivalents into mitochondria via the malate-aspartate shuttle. Biochim. Biophys. Acta 376, 387–397. Breslin, S., and O’Driscoll, L. (2013). Three-dimensional cell culture: The missing link in drug discovery. Drug Discov. Today 18, 240–249.

DeBerardinis, R.J., Mancuso, A., Daikhin, E., Nissim, I., Yudkoff, M., Wehrli, S., and Thompson, C.B. (2007). Beyond aerobic glycolysis: Transformed cells can engage in glutamine metabolism that exceeds the requirement for protein and nucleotide synthesis. Proc. Natl. Acad. Sci. USA 104, 19345–19350. Dunckelmann, R.J., Ebinger, F., Schulze, A., Wanders, R.J., Rating, D., and Mayatepek, E. (2000). 2-ketoglutarate dehydrogenase deficiency with intermittent 2-ketoglutaric aciduria. Neuropediatrics 31, 35–38. Frezza, C., Zheng, L., Folger, O., Rajagopalan, K.N., MacKenzie, E.D., Jerby, L., Micaroni, M., Chaneton, B., Adam, J., Hedley, A., et al. (2011). Haem oxygenase is synthetically lethal with the tumour suppressor fumarate hydratase. Nature 477, 225–228. Gameiro, P.A., Yang, J., Metelo, A.M., Pe´rez-Carro, R., Baker, R., Wang, Z., Arreola, A., Rathmell, W.K., Olumi, A., Lo´pez-Larrubia, P., et al. (2013). In vivo HIF-mediated reductive carboxylation is regulated by citrate levels and sensitizes VHL-deficient cells to glutamine deprivation. Cell Metab. 17, 372–385. Gross, M.I., Demo, S.D., Dennison, J.B., Chen, L., Chernov-Rogan, T., Goyal, B., Janes, J.R., Laidig, G.J., Lewis, E.R., Li, J., et al. (2014). Antitumor activity of the glutaminase inhibitor CB-839 in triple-negative breast cancer. Mol. Cancer Ther. 13, 890–901. Hanahan, D., and Weinberg, R.A. (2011). Hallmarks of cancer: The next generation. Cell 144, 646–674. Heublein, M., Burguillos, M.A., Vo¨gtle, F.N., Teixeira, P.F., Imhof, A., Meisinger, C., and Ott, M. (2014). The novel component Kgd4 recruits the E3 subunit to the mitochondrial a-ketoglutarate dehydrogenase. Mol. Biol. Cell 25, 3342–3349. Johnson, M.T., Yang, H.S., Magnuson, T., and Patel, M.S. (1997). Targeted disruption of the murine dihydrolipoamide dehydrogenase gene (Dld) results in perigastrulation lethality. Proc. Natl. Acad. Sci. USA 94, 14512–14517. Mandujano-Tinoco, E.A., Gallardo-Pe´rez, J.C., Marı´n-Herna´ndez, A., MorenoSa´nchez, R., and Rodrı´guez-Enrı´quez, S. (2013). Anti-mitochondrial therapy in human breast cancer multi-cellular spheroids. Biochim. Biophys. Acta 1833, 541–551. Mashimo, T., Pichumani, K., Vemireddy, V., Hatanpaa, K.J., Singh, D.K., Sirasanagandla, S., Nannepaga, S., Piccirillo, S.G., Kovacs, Z., Foong, C., et al. (2014). Acetate is a bioenergetic substrate for human glioblastoma and brain metastases. Cell 159, 1603–1614.

Cell Reports 17, 876–890, October 11, 2016 889

Mayers, J.R., and Vander Heiden, M.G. (2015). Famine versus feast: Understanding the metabolism of tumors in vivo. Trends Biochem. Sci. 40, 130–140.

oncometabolite fumarate binds glutathione to amplify ROS-dependent signaling. Mol. Cell 51, 236–248.

Metallo, C.M., Gameiro, P.A., Bell, E.L., Mattaini, K.R., Yang, J., Hiller, K., Jewell, C.M., Johnson, Z.R., Irvine, D.J., Guarente, L., et al. (2012). Reductive glutamine metabolism by IDH1 mediates lipogenesis under hypoxia. Nature 481, 380–384.

Sullivan, L.B., Gui, D.Y., Hosios, A.M., Bush, L.N., Freinkman, E., and Vander Heiden, M.G. (2015). Supporting aspartate biosynthesis is an essential function of respiration in proliferating cells. cell 162, 552–563.

Mullen, A.R., Wheaton, W.W., Jin, E.S., Chen, P.H., Sullivan, L.B., Cheng, T., Yang, Y., Linehan, W.M., Chandel, N.S., and DeBerardinis, R.J. (2011). Reductive carboxylation supports growth in tumour cells with defective mitochondria. Nature 481, 385–388. Mullen, A.R., Hu, Z., Shi, X., Jiang, L., Boroughs, L.K., Kovacs, Z., Boriack, R., Rakheja, D., Sullivan, L.B., Linehan, W.M., et al. (2014). Oxidation of alpha-ketoglutarate is required for reductive carboxylation in cancer cells with mitochondrial defects. Cell Rep. 7, 1679–1690. Muller, F.L., Colla, S., Aquilanti, E., Manzo, V.E., Genovese, G., Lee, J., Eisenson, D., Narurkar, R., Deng, P., Nezi, L., et al. (2012). Passenger deletions generate therapeutic vulnerabilities in cancer. Nature 488, 337–342.

Sun, R.C., and Denko, N.C. (2014). Hypoxic regulation of glutamine metabolism through HIF1 and SIAH2 supports lipid synthesis that is necessary for tumor growth. Cell Metab. 19, 285–292. Ulanet, D.B., Couto, K., Jha, A., Choe, S., Wang, A., Woo, H.K., Steadman, M., DeLaBarre, B., Gross, S., Driggers, E., et al. (2014). Mesenchymal phenotype predisposes lung cancer cells to impaired proliferation and redox stress in response to glutaminase inhibition. PLoS ONE 9, e115144. Vander Heiden, M.G. (2011). Targeting cancer metabolism: A therapeutic window opens. Nat. Rev. Drug Discov. 10, 671–684. Vinci, M., Gowan, S., Boxall, F., Patterson, L., Zimmermann, M., Court, W., Lomas, C., Mendiola, M., Hardisson, D., and Eccles, S.A. (2012). Advances in establishment and analysis of three-dimensional tumor spheroid-based functional assays for target validation and drug evaluation. BMC Biol. 10, 29.

Qiu, F., Chen, Y.R., Liu, X., Chu, C.Y., Shen, L.J., Xu, J., Gaur, S., Forman, H.J., Zhang, H., Zheng, S., et al. (2014). Arginine starvation impairs mitochondrial respiratory function in ASS1-deficient breast cancer cells. Sci. Signal. 7, ra31.

Ward, P.S., and Thompson, C.B. (2012). Metabolic reprogramming: A cancer hallmark even Warburg did not anticipate. Cancer Cell 21, 297–308.

Sellers, K., Fox, M.P., Bousamra, M., 2nd, Slone, S.P., Higashi, R.M., Miller, D.M., Wang, Y., Yan, J., Yuneva, M.O., Deshpande, R., et al. (2015). Pyruvate carboxylase is critical for non-small-cell lung cancer proliferation. J. Clin. Invest. 125, 687–698.

Wise, D.R., Ward, P.S., Shay, J.E., Cross, J.R., Gruber, J.J., Sachdeva, U.M., Platt, J.M., DeMatteo, R.G., Simon, M.C., and Thompson, C.B. (2011). Hypoxia promotes isocitrate dehydrogenase-dependent carboxylation of a-ketoglutarate to citrate to support cell growth and viability. Proc. Natl. Acad. Sci. USA 108, 19611–19616.

Son, J., Lyssiotis, C.A., Ying, H., Wang, X., Hua, S., Ligorio, M., Perera, R.M., Ferrone, C.R., Mullarky, E., Shyh-Chang, N., et al. (2013). Glutamine supports pancreatic cancer growth through a KRAS-regulated metabolic pathway. Nature 496, 101–105. Stein, E.M., and Tallman, M.S. (2016). Emerging therapeutic drugs for AML. Blood 127, 71–78. Stuart, S.D., Schauble, A., Gupta, S., Kennedy, A.D., Keppler, B.R., Bingham, P.M., and Zachar, Z. (2014). A strategically designed small molecule attacks alpha-ketoglutarate dehydrogenase in tumor cells through a redox process. Cancer Metab. 2, 4. Sullivan, L.B., Martinez-Garcia, E., Nguyen, H., Mullen, A.R., Dufour, E., Sudarshan, S., Licht, J.D., Deberardinis, R.J., and Chandel, N.S. (2013). The proto-

890 Cell Reports 17, 876–890, October 11, 2016

Yang, L., Shi, Q., Ho, D.J., Starkov, A.A., Wille, E.J., Xu, H., Chen, H.L., Zhang, S., Stack, C.M., Calingasan, N.Y., et al. (2009). Mice deficient in dihydrolipoyl succinyl transferase show increased vulnerability to mitochondrial toxins. Neurobiol. Dis. 36, 320–330. Yang, H., Zhou, L., Shi, Q., Zhao, Y., Lin, H., Zhang, M., Zhao, S., Yang, Y., Ling, Z.Q., Guan, K.L., et al. (2015). SIRT3-dependent GOT2 acetylation status affects the malate-aspartate NADH shuttle activity and pancreatic tumor growth. EMBO J. 34, 1110–1125. Zachar, Z., Marecek, J., Maturo, C., Gupta, S., Stuart, S.D., Howell, K., Schauble, A., Lem, J., Piramzadian, A., Karnik, S., et al. (2011). Non-redox-active lipoate derivates disrupt cancer cell mitochondrial metabolism and are potent anticancer agents in vivo. J. Mol. Med. 89, 1137–1148.