An essential role of the mitochondrial electron transport chain in cell proliferation is to enable aspartate synthesis

A CRISPR-based genetic screen for metabolic genes that when lost sensitize human cells to phenformin

Pharmacological or genetic inhibition of the ETC greatly suppresses cell proliferation (Santidrian et al., 2013; Wheaton et al., 2014) (Figure 1A), but exactly why is unclear. To study this question, we performed a CRISPR-based negative selection screen for genes whose loss potentiates the anti-proliferative effects of mild ETC inhibition. Such genes should reveal processes that help cells adapt to ETC impairment and thus pinpoint key ETC functions in proliferating cells. Given the central role of mitochondria in metabolism, we generated a library consisting of ~30,000 sgRNAs targeting ~3,000 metabolic enzymes, small molecule transporters, and metabolism-related transcription factors (~10 sgRNA/gene) as well as 500 control sgRNAs in a Cas9-expressing lentiviral vector (Figure 1B). We transduced human Jurkat leukemic T-cells with the sgRNA library, and passaged the pool of knockout cells in pyruvate-free RPMI media for 14 population doublings in the presence or absence of 0.5 uM phenformin. This lipophilic biguanide inhibits complex I of the ETC (Birsoy et al., 2014; Owen et al., 2000; Wheaton et al., 2014) and only mildly slowed proliferation at the concentration used (Figure 1A). As expected for an ETC inhibitor, phenformin dose dependently suppressed respiration as measured by oxygen consumption (Figure S1D).

Using massively parallel sequencing, we measured the abundances of all the sgRNAs in the vehicle- and phenformin-treated Jurkat cells at the beginning and at the end of the culture period. For each gene we calculated its score as the median log₂ fold change in the abundance of the 10 sgRNAs targeting the gene. As expected, most genes, as well as the control sgRNAs, scored similarly in the presence or absence of phenformin (Figure 1C).

Among the genes selectively required in the presence of phenformin, were several encoding ETC-related proteins, including complex I subunits (NDUFA11, NDUFB2, NDUFA10, NDUFB9), mitochondrial RNA processing and ubiquinone synthesis enzymes (PDE12 and COQ4), and ETC assembly factors (ACAD9 and SCO2) (Figure 1D, Figure S1B). Unlike genes encoding core ETC components, these genes were not essential in the absence of phenformin (Figure S1A). This suggests that they are not absolutely required for ETC function, explaining why their loss synergize with low-dose phenformin. These results are consistent with the previous finding that cells having heteroplasmic mitochondrial DNA mutations that partially impair the ETC are hypersensitive to phenformin (Birsoy et al., 2014).

The best scoring gene in the screen was GOT1, which encodes the cytosolic aspartate aminotransferase that is part of the malate-aspartate shuttle for transferring reducing equivalents to the mitochondrial matrix (Figure 1D–E). GOT1 catalyzes the reversible transfer of an amino group between aspartate and glutamate, and like other transaminases, requires pyridoxal-5-phosphate (PLP) as a cofactor (Toney, 2014) (Figure 1D). Interestingly, the fourth highest scoring gene, PDXK, encodes a pyridoxal kinase, which converts vitamin B6 to PLP (Figure S1C). These data strongly suggest that a GOT1-catalyzed reaction is important for maintaining cellular fitness upon mild phenformin treatment. Given its high score and unexplored role during ETC inhibition, we focused our attention on GOT1.

ETC inhibition kills cells lacking GOT1

To begin to understand how GOT1 loss sensitizes cells to phenformin, we used the CRISPR-Cas9 system to generate two clonal Jurkat cell lines in which the GOT1 protein was undetectable (GOT1_KO1 and GOT1_KO2) (Figure 1F). Under normal culture conditions, GOT1-null cells are viable and proliferate at slightly slower rates than wild type cells (Figure S1E). Consistent with the results of the screen, low doses of phenformin inhibited the proliferation of GOT1-null cells to a much greater extent than that of wild type cells (Figure 1F). Remarkably, at the higher concentrations of phenformin that strongly repress respiration (Figure S1D), GOT1-null cells arrested and died, while the wild type counterparts did not (Figure 1F, Figure S1E). Importantly, expression of an sgGOT1-resistant human GOT1 cDNA in the null cells eliminated their hypersensitivity to phenformin (Figure 1G). These findings validate the screening results and reveal that GOT1 loss and phenformin interact synergistically in Jurkat cells.

To ask if GOT1 loss sensitizes other cell types to ETC inhibition, we knocked out GOT1 in human Raji lymphoma and KMS-26 multiple myeloma cells, as well as immortalized mouse embryonic fibroblasts (MEFs) (Figure 2A, Figure S2A). Indeed, all the GOT1-null cells were far more sensitive to phenformin than their wild type counterparts, indicating a generalizable role for GOT1 in cells treated with phenformin (Figure 2A, Figure S2A).

Finally, inhibition of different complexes of the ETC can have pleiotropic effects on metabolism (Bell et al., 2007; Frezza et al., 2011). This raised the possibility that the sensitizing effect of GOT1 loss might be specific to complex I inhibition or even phenformin treatment, in particular. However, this is not the case because compared to wild type cells, those lacking GOT1 were substantially more sensitive to other complex I inhibitors (metformin and piericidin) as well as the complex III inhibitor, antimycin (Figure 2B, Figure S2B–C). These data indicate that GOT1 loss has a synthetic lethal interaction with ETC dysfunction, independently of which complex is inhibited.

Upon ETC inhibition GOT1 reverses flux and generates aspartate

GOT1 is part of the malate–aspartate shuttle (Figure 3A, Figure S3A), but no other component of the shuttle (MDH1, MDH2, SLC25A13, GOT2) scored in the screen (Figure S3B). Given these results, we focused on the reaction mediated by GOT1 itself rather than the overall function of the shuttle in transferring reducing equivalents into the mitochondrial matrix.

In normal cells GOT1 is thought to use aspartate and alpha-ketoglutarate to make oxaloacetate and glutamate (Safer, 1975). Consistent with GOT1 consuming aspartate, its levels are 4–5 fold higher in GOT1-null than wild type Jurkat cells (Figure 3B). Aspartate is normally synthesized in the mitochondrial matrix through the sequential actions of MDH2 and GOT2 and then transported to the cytosol for use by GOT1 and other enzymes (Figure 3A). Because MDH2 is an oxidoreductase, the drop in the NAD⁺/NADH ratio that occurs upon ETC dysfunction should inhibit MDH2 and thus mitochondrial aspartate synthesis. Indeed, in wild type Jurkat cells, phenformin caused aspartate levels to fall by ~3-fold (Figure 3B).

As GOT1 is bidirectional, the drop in aspartate levels might allow GOT1 to reverse flux so that in cells with ETC dysfunction GOT1 generates rather than consumes aspartate. If this were the case, ETC inhibition should cause aspartate to drop to a greater extent in GOT1-null than wild type cells. Indeed, in GOT1-null cells phenformin treatment lead to an almost complete loss of cellular aspartate (~30-fold reduction) without much effect on other amino acids (Figure 3B). Thus, upon ETC inhibition, cells use a GOT1-dependent pathway to generate aspartate. It is important to note, however, that this pathway does not fully compensate for loss of mitochondrial aspartate synthesis as aspartate is three-fold lower in cells without a functional ETC (Figure 3B). To understand if aspartate levels also drop upon ETC inhibition in a differentiated cell in vivo, we generated a previously reported mouse strain with the heart-specific deletion of mitochondrial transcription factor A (Tfam) (Hansson et al., 2004). Consistent with the severe ETC defect caused by Tfam loss, Tfam knock-out hearts had a two-fold lower ratio of aspartate to leucine than wild type hearts (Figure 3C).

Aspartate supplementation enables cells to proliferate under pharmacological ETC inhibition

Because aspartate is required for the synthesis of proteins as well as purines and pyrimidines (Lane and Fan, 2015), we reasoned that the drop in aspartate levels caused by ETC inhibition might lead to the concomitant suppression of cell proliferation. Consistent with this notion, supplementation of RPMI media with aspartate (10 mM) partially reversed the anti-proliferative effects of phenformin on wild type cells (Figure 3D). Aspartate addition also prevented the phenformininduced death of GOT1-null cells and even enabled these cells to proliferate in the presence of the drug (Figure 3D). Given these findings, we re-examined our screen hits and noted that several enzymes scored that use aspartate for purine and pyrimidine synthesis (Figure 1D, Figure S1B).

The aspartate rescue experiments required supplementation of the media with concentrations of aspartate (10 mM) that are higher than those found in standard media (Figure 3D), which is likely a consequence of Jurkat cells having poor aspartate transport. To test this possibility, we cultured GOT1-null cells stably expressing the SLC1A3 glutamate-aspartate transporter in standard RPMI, containing 150 uM aspartate. SLC1A3 imports aspartate (Storck et al., 1992) and is highly expressed in neuronal tissues but not in Jurkat cells (Figure S3C–D). Remarkably, SLC1A3 overexpression prevented the death of the GOT1-null cells caused by phenformin (Figure 3E, Figure S3E) and other ETC inhibitors (Figure S3F), and, like high dose aspartate, enabled the proliferation of these cells (Figure 3D, Figure S3E–F). Thus, we conclude that upon ETC inhibition, aspartate becomes limiting for maintaining the viability and proliferation of cells.

Metabolic route for aspartate synthesis in cells with ETC inhibition

To generate aspartate in cells with ETC inhibition, GOT1 must have a source of oxaloacetate (Figure S4A), which can be made from (i) malate by the cytosolic (MDH1) or mitochondrial (MDH2) malate dehydrogenases; (ii) pyruvate by pyruvate carboxylase (PC) in mitochondria; or (iii) citrate by ATP-citrate lyase (ACL) in the cytosol. MDH1, MDH2 or PC did not score in our screen as differentially essential upon mild phenformin treatment (Figure S3B, Figure S4B), so we focused on the possible generation of oxaloacetate from citrate upon ETC inhibition. In cells with ETC dysfunction, glutamine is a major source of citrate through the “reductive carboxylation” pathway (Metallo et al., 2012; Mullen et al., 2012). In reductive carboxylation, the mitochondrial citrate carrier (SLC25A1) transports glutaminederived citrate to the cytosol, where ACL cleaves it into oxaloacetate. Consistent with oxaloacetate being generated from citrate in cells with ETC inhibition, SLC25A1 scored as differentially required in our phenformin screen (Figure 1D). The differential requirement of ACL could not be assessed because it scored as essential under all conditions (Figure S4B), likely because it is required for fatty acid synthesis.

To directly determine how aspartate is made, we measured its generation from [U-¹³C]-L-glutamine in wild type and GOT1-null Jurkat cells treated with or without phenformin. Oxidative metabolism of the uniformly labeled glutamine will generate aspartate with four ¹³C atoms (m+4), while aspartate made by reductive carboxylation will have three ¹³C atoms (m+3) (Figure 4A). Under normal conditions, we found that oxidative glutamine metabolism was the predominant source of aspartate (~50% of total aspartate pool) in both wild type and GOT1-null Jurkat cells (Figure 4B). Upon ETC inhibition, aspartate synthesis dropped, and what remained came primarily from reductive glutamine metabolism with almost no contribution from the oxidative pathway (Figure 4B). Reductive formation of aspartate completely depends on GOT1 activity, as in phenformin-treated GOT1-null cells there was almost no generation of aspartate or aspartate-derived nucleotide precursors from labeled glutamine (Figure 4B, Figure S4C). These results indicate that upon ETC inhibition cells use the reductive carboxylation of glutamine to fuel compensatory aspartate synthesis via a GOT1-requiring pathway.

Pyruvate stimulates aspartate synthesis in a GOT1-dependent fashion

It has long been appreciated that human cells with ETC dysfunction can proliferate when cultured in media containing supra-physiological concentrations of pyruvate (Harris, 1980). Remarkably, pyruvate even enables the proliferation of cells that have deleterious mutations in their mitochondrial DNA or lack it altogether (143B ρ⁰ cells) (King and Attardi, 1989). Pyruvate has been hypothesized to act as a biosynthetic substrate or to maintain the cellular redox state in cells with ETC dysfunction via reduction by lactate dehydrogenase, which helps regenerate the NAD⁺ that is lost upon ETC inhibition (Harris, 1980; Wilkins et al., 2014). The NAD⁺ made through pyruvate reduction should facilitate glycolytic flux and thus ATP production in cells lacking ETC function, but the key metabolic consequence of pyruvate addition that allows such cells to proliferate is unclear.

As both pyruvate and aspartate supplementation enables the proliferation of cells with ETC dysfunction, we explored a possible link between the two by culturing wild type and GOT1-null cells in the presence or absence of pyruvate and treating them with ETC inhibitors. Consistent with previous reports (Harris, 1980), pyruvate almost completely blocked the anti-proliferative effects of several ETC inhibitors (phenformin, piericidin, and antimycin) on wild type Jurkat cells (Figure 5A). In contrast, pyruvate had no beneficial effect on the GOT1-null cells (Figure 5A), suggesting that the pyruvate-mediated rescue requires aspartate synthesis via GOT1. Indeed, pyruvate restored aspartate levels to normal in phenformin-treated wild type cells, but had no effect on aspartate in phenformin-treated GOT1-null cells (Figure 5B). Collectively, these data indicate that a key mechanism through which pyruvates restores the proliferation of cells with ETC inhibition is to promote GOT1-catalyzed aspartate synthesis.

Metabolic path of pyruvate-induced aspartate synthesis

To understand how pyruvate stimulates aspartate synthesis in cells with ETC inhibition, we considered two pathways through which pyruvate might contribute to the aspartate pool. (i) One possibility is that pyruvate carboxylase (PC) directly carboxylates pyruvate into oxaloacetate, which GOT1 then uses to generate aspartate. However, we quickly ruled out this pathway as in PC-null Jurkat cell lines pyruvate still prevented the anti-proliferative effects of ETC inhibitors (Figure S5A). (ii) As described earlier, pyruvate can promote the regeneration of NAD⁺ in the cytosol. Given this, we hypothesized that NAD⁺ might activate the cytosolic malate dehydrogenase (MDH1) to generate oxaloacetate that then drives aspartate synthesis by GOT1. To investigate this possibility, we generated MDH1-null Jurkat cells (MDH1_KO1 and MDH1_KO2) (Figure 6A). In the absence of ETC inhibitors, these cells had similar levels of aspartate as GOT1-null cells, consistent with MDH1 normally consuming the oxaloacetate generated by GOT1 from aspartate (Figure 6B). Upon phenformin treatment, aspartate in the MDH1-null cells dropped to the same level as it did in wild type cells (Figure 6B). Pyruvate, however, had no effect on the MDH1-null cells: in cells treated with ETC inhibitors it did not stimulate aspartate synthesis or rescue their proliferation (Figure 6A–B). Importantly, aspartate supplementation or expression in the null cells of an sgRNA-resistant human MDH1 cDNA reversed this proliferation defect (Figure 6A–B, Figure 5SC). Our findings suggest a model in which pyruvate-induced NAD⁺ activates MDH1 to produce oxaloacetate from malate and drive aspartate synthesis via GOT1. In addition, the reductive pathway to oxaloacetate synthesis that operates in the absence of pyruvate (Figure 6C) is still operational, so in the presence of pyruvate there are likely multiple ways of fueling GOT1 for aspartate synthesis (Figure 6C). Supporting the idea that redox balance is upstream of aspartate synthesis, aspartate addition, unlike that of pyruvate, did not restore the NAD/NADH ratio in cells with ETC inhibition (Figure S5B).

Although under normal conditions both GOT1- and MDH1-null cells have high aspartate levels, they behave strikingly differently upon ETC inhibition: aspartate levels become almost undetectable in GOT1-null cells and the cells die, while in MDH1-null cells aspartate and proliferation drop to the same extent as in wild type cells (Figure 6D). This major difference in aspartate levels upon ETC inhibition is a consequence of the MDH1-null (and also wild type) cells being able to generate some aspartate even in the absence of pyruvate through the reductive carboxylation pathway described earlier while the GOT1-null cells cannot (Figure 4). In other words, in the absence of ETC function, the loss of GOT1 eliminates all routes to aspartate synthesis while the reductive carboxylation path is still available in the MDH1-null and wild type cells.

This conclusion also provides a rationale for why MDH1 did not score like GOT1 in our screen, which was performed in RPMI, a pyruvate-free medium (Figure 1B). In the absence of pyruvate and when treated with ETC inhibitors, MDH1-null and wild type cells have comparable aspartate levels and proliferate equally poorly (Figure 6A–B). Interestingly, aspartate levels in wild type, GOT1-null, and MDH1-null cells treated with ETC inhibitors are a good predictor of whether the cells proliferate, arrest, or die (Figure 6D).

Aspartate supplementation enables the proliferation of cybrids with patient-derived mtDNA mutations even in the absence of pyruvate

Transmitochondrial cytoplasmic hybrid cells (cybrids) are commonly used to study the mechanisms through which mutations in mitochondrial DNA (mtDNA) impair cellular function (Schon et al., 2012). These cells harbor patient-derived mitochondrial genomes with pathogenic mutations, and, like cells treated with ETC inhibitors, require pyruvate to proliferate in culture (King and Attardi, 1989, 1996). Complex III deficient cybrid cells also require the addition of uridine to the media as complex III is needed for its synthesis (Loffler et al., 1997). To expand upon our work with pharmacological inhibition of the ETC, we asked if aspartate supplementation rescues the proliferation of cells with genetic ETC defects. We used two patientderived cybrid cells, the first with a homoplasmic microdeletion in the cytochrome b subunit of complex III (CYTB) associated with a form of parkinsonism and encephalopathy (De Coo et al., 1999; Rana et al., 2000), and the second with a homoplasmic point mutation in the mitochondrially encoded tRNA lysine (MT-TK) associated with myoclonic epilepsy with ragged red fibers (MERRF) (Wallace et al., 1988). To characterize the metabolic needs of these cybrids, as well as cells lacking mtDNA (143B ρ⁰), we cultured them in media with or without pyruvate or aspartate. In agreement with previous reports, the ETC-deficient cybrids and ρ⁰ cells proliferated in media supplemented with pyruvate (Figure 7A). Consistent with the data obtained with pharmacological inhibitors, pyruvate stimulated aspartate synthesis in the ETC-defective cells (Figure 7B) and aspartate on its own enabled their proliferation (Figure 7A). The pyruvate-mediated rescue requires GOT1 as knocking it out in the ETC-defective cybrids severely impaired their proliferation (Figure 7C). Importantly, aspartate supplementation bypassed the need for GOT1 (Figure 7C).

To test the sufficiency of aspartate in enabling the proliferation of cells with genetic ETC defects, we stably expressed the glutamate-aspartate SLC1A3 transporter in the CYTB cybrids (Figure 7D). Strikingly, SLC1A3 expression was sufficient to enable the cybrids to proliferate in standard RPMI media, which has a low aspartate concentration (150 uM) and no pyruvate (Figure 7D). To rule out the possibility that SLC1A3 enables proliferation by transporting a molecule other than aspartate, we cultured the cells in aspartate-free RPMI. Now, the SLC1A3-expressing CYTB cybrids failed to proliferate in the absence of pyruvate. Thus, we conclude that aspartate is sufficient to enable the proliferation of cells with a genetic ETC defect.

CRISPR-based screen

The metabolism-focused sgRNA library was designed and performed as previously described (Wang et al., 2014). Oligonucleotides for sgRNAs were synthesized by CustomArray Inc. and amplified by PCR (Wang et al., 2014). Gene scores of all the metabolic genes in the screen can be found in Table S1.

Proliferation assays

Indicated cell lines were cultured in replicates of three in 96-well plates at 2–3,000 cells per well in 200 uL RPMI base media under the conditions described in each experiment, and a separate group of 3 wells was also plated for each cell line with no treatment for an initial time point. After 5 hours (untreated cells for initial time point) or after 5 days (with varying treatment conditions), 40 uL of Cell Titer Glo reagent (Promega) was added to each well, mixed briefly, and the luminescence read on a Luminometer (Molecular Devices). For each well, the fold change in luminescence relative to the initial luminescence was measured and reported in a log₂ scale. For Figure 6, relative fold change in luminescence relative to initial was calculated.

Cell Counting Assays

Cybrids or 143B ρ0 cells were plated in triplicate in 12 well plates at 5,000–20,000 cells per well in 1.5 mL as described in each experiment. After six days, the entire contents of the well was trypsinized and counted using a Beckman Z2 Coulter Counter with a size selection setting of 8–30 um.

Generation of knock-out and cDNA overexpression cell lines

sgRNAs (oligonucleotide sequences are indicated above) were cloned into lentiCRISPR-v1 linearized with BsmBI by Gibson Assembly (NEB). sgRNA expressing vector along with lentiviral packaging vectors Delta-VPR and CMV VSV-G were transfected into HEK-293T cells using the XTremeGene 9 transfection reagent (Roche). Similarly, for overexpression cell lines, cDNA vectors along with retroviral packaging vectors gag-pol and CMV VSV-G were transfected into HEK-293T cells. Media was changed 24 hours after transfection. The virus-containing supernatant was collected 48 and 72 hours after transfection and passed through a 0.45 µm filter to eliminate cells. Target cells in 6-well tissue culture plates were infected in media containing 8 µg/mL of polybrene and a spin infection was performed by centrifugation at 2,200 rpm for 1 hour. Post-infection, virus was removed and cells were selected with puromycin or blasticidin. For knock out cells, after selection, cells were single-cell sorted with a flow cytometer into the wells of a 96-well plate containing 200 ul of RPMI supplemented with 20% FBS. Cells were grown for two weeks, and the resultant colonies were trypsinized and expanded. Clones were validated for loss of the relevant protein via immunoblotting.

Metabolite profiling and isotope tracing

LC/MS analyses were conducted on a QExactive benchtop orbitrap mass spectrometer equipped with an Ion Max source and a HESI II probe, which was coupled to a Dionex UltiMate 3000 UPLC system (Thermo Fisher Scientific, San Jose, CA). External mass calibration was performed using the standard calibration mixture every 7 days.

For metabolite profiling experiments, Jurkat cells (2 million per sample) were incubated with 10 uM phenformin for 24 hours. Similarly, for glutamine tracing experiments, Jurkat cells (2 million per sample) were incubated with 10 uM phenformin for 18 hours; the cells were spun down and media was replaced with RPMI supplemented with 1 mM [U-¹³C]-L-glutamine for 7 hours. Polar metabolites were extracted using 1 ml of ice-cold 80% methanol with 10 ng/ml valine-d₈ as an internal standard. After a 10 min vortex and centrifugation for 10 min at 4°C at 10,000g, samples were dried under nitrogen gas. Dried samples were stored at −80° C and then resuspended in 100 µL water; 1 µL of each sample was injected onto a ZIC-pHILIC 2.1 × 150 mm (5 µm particle size) column (EMD Millipore). Buffer A was 20 mM ammonium carbonate, 0.1% ammonium hydroxide; buffer B was acetonitrile. The chromatographic gradient was run at a flow rate of 0.150 ml/min as follows: 0–20 min.: linear gradient from 80% to 20% B; 20–20.5 min.: linear gradient from 20% to 80% B; 20.5–28 min.: hold at 80% B. The mass spectrometer was operated in full-scan, polarity switching mode with the spray voltage set to 3.0 kV, the heated capillary held at 275°C, and the HESI probe held at 350°C. The sheath gas flow was set to 40 units, the auxiliary gas flow was set to 15 units, and the sweep gas flow was set to 1 unit. The MS data acquisition was performed in a range of 70–1000 m/z, with the resolution set at 70,000, the AGC target at 10⁶, and the maximum injection time at 80 msec. Relative quantitation of polar metabolites was performed with XCalibur QuanBrowser 2.2 (Thermo Fisher Scientific) using a 5 ppm mass tolerance and referencing an in-house library of chemical standards.