Loss of LUC7L2 and U1 snRNP Subunits Shifts Energy Metabolism from Glycolysis to OXPHOS

SUMMARY

Oxidative phosphorylation (OXPHOS) and glycolysis are the two major pathways for ATP production. The reliance on each varies across tissues and cell states, and can influence susceptibility to disease. At present, the full set of molecular mechanisms governing the relative expression and balance of these two pathways is unknown. Here, we focus on genes whose loss leads to an increase in OXPHOS activity. Unexpectedly, this class of genes is enriched for components of the pre-mRNA splicing machinery and in particular for subunits of the U1 snRNP. Among them, we show that LUC7L2 represses OXPHOS and promotes glycolysis by a coordinated mechanism, involving (1) splicing of the glycolytic enzyme PFKM to suppress glycogen synthesis, (2) splicing of the cystine/glutamate antiporter SLC7A11 (xCT) to suppress glutamate oxidation, and (3) secondary repression of mitochondrial respiratory supercomplex formation. Our results connect LUC7L2 and, more generally, the U1 snRNP to cellular energy metabolism.

Graphical Abstract

eTOC

Jourdain et al., report the identification of OXPHOS repressors: genes whose loss shifts metabolism from glycolysis to OXPHOS. Prominent in this set are members of the U1 snRNP, including LUC7L2, which the authors show is required for splicing of PFKM and SLC7A11 (xCT) and controls glycogen and glutamate metabolism.

INTRODUCTION

Human cells employ two chief pathways for generating ATP: glycolysis and oxidative phosphorylation (OXPHOS) (Fig. 1A). Use of these metabolic routes is associated with key tradeoffs, and while glycolysis tends to be kinetically favorable, the ATP yield from OXPHOS is higher (Pfeiffer et al., 2001). The relative balance of OXPHOS and glycolysis varies across tissues. For example, while cardiac tissue is rich in mitochondria and highly oxidative in its metabolism, proliferating cells from the thymus are highly glycolytic (Warburg, 1924). Balance between these two programs can vary during cellular differentiation and in response to environmental stimuli. Activation of immune cells is often accompanied by rewiring towards glycolysis, while stem cell differentiation results in increased OXPHOS (Pearce et al., 2013, Ito and Suda, 2014). Cells also acutely respond to nutrient and oxygen availability to adjust flux through these pathways within minutes (Crabtree, 1929, Pasteur, 1861), while oncogenic transformation promotes aerobic glycolysis (Warburg, 1924). Notably, differential reliance on OXPHOS versus glycolysis can be exploited for therapeutic benefit (Bonnet et al., 2007, Gohil et al., 2010).

Figure 1: Identification of Pre-mRNA Splicing Components as Repressors of OXPHOS. See also Figure S1 and Table S1.

(A) Overview of the main ATP-generating pathways in human cells. OXPHOS: oxidative phosphorylation. ETC: electron transport chain. TCA: tricarboxylic acid cycle. (B) Gene-level analysis of a genome-wide CRISPR/Cas9 screen in glucose and galactose. Each dot represents an expressed, non-essential gene (n = 9,189). (C) Gene ontology analysis generated using a gene list ranked by viability in galactose against GO components. (D–F) Functional validation of the screening results. Basal whole-cell oxygen consumption rates (OCR), extracellular acidification rates (ECAR) and OCR/ECAR ratios were simultaneously measured after CRISPR/Cas9-mediated gene depletion in K562 cells grown in glucose-containing media. Data are shown as mean ± SEM (n≥3 independent experiments. *P<0.05, **P<0.01, ***P<0.001, t-test relative to control (GFP) sgRNA-treated cells. NDUFB5 is a control with known role in OXPHOS.

A small number of genomic programs have been identified that can influence the balance of cellular energy metabolism. For example, the transcriptional co-activator PGC-1α integrates nutrient levels and physiological inputs to orchestrate a genomic program that induces the expression of the OXPHOS machinery (Puigserver et al., 1998). Conversely, the transcription factor HIF-1α promotes expression of a set of genes, including glycolytic enzymes, in response to a decline in oxygen levels (Huang et al., 1998). Post-transcriptionally, RNA-binding proteins such as CLUH bind to a large number of mRNAs encoding mitochondrial proteins (Gao et al., 2014). A few instances of alternative splicing regulating individual metabolic enzymes have also been reported. For example, alternative splicing to generate the PKM2 isoform of the pyruvate kinase, re-routing lower glycolytic carbon flux, is believed to contribute to cancer progression (Christofk et al., 2008).

High-throughput approaches can provide insight into the regulation and patterning of cellular metabolic programs (Hillenmeyer et al., 2008, VanderSluis et al., 2014). We previously reported a nutrient-sensitized screen for small molecules that impact the fitness of cells in galactose, a poor substrate for glycolysis, and focused on dozens of small molecules that induce a shift from OXPHOS to glycolysis (Gohil et al., 2010). Recently, we reported the genome-wide identification of genes necessary to sustain OXPHOS (Arroyo et al., 2016). We systematically catalogued genes whose loss impaired OXPHOS in human cells, including 72 that underlie known OXPHOS diseases. However, that report did not explore the opposite side of the screen, which could in principle include pathways that tonically suppress OXPHOS.

Here, we report the genome-wide identification of “OXPHOS repressors”, defined as genes whose knockout promotes relative fitness in the absence of glucose as a fuel for glycolysis. We validate top-scoring genes and show that their depletion augments OXPHOS activity. OXPHOS repressors are enriched for components of the pre-mRNA splicing machinery, including subunits of the U1 snRNP. Among them, we show that LUC7L2 encodes a U1 snRNP subunit involved in pre-mRNA splicing and gene expression. Amongst LUC7L2 gene targets, we focus on two genes, PFKM and SLC7A11 (xCT), which we show represent two metabolic crossovers that influence the bioenergetic state of the cell in a LUC7L2-dependent manner.

RESULTS

Genome-Wide Search for Factors that Limit OXPHOS Identifies Components of the Pre-mRNA Splicing Machinery

To nominate genes whose depletion promotes OXPHOS, we reanalyzed the results of our genome-wide “death screen” that compared viability of CRISPR/Cas9 mutagenized K562 cells that are shifted to glucose or galactose conditions for 24 hours (Arroyo et al., 2016). Loss of genes required for OXPHOS is tolerated in the presence of glucose, whereas these genes are conditionally essential in galactose, as it is a poor substrate for glycolysis (Robinson et al., 1992). We reanalyzed 9,189 expressed non-essential genes in K562 cells, calculating a z-score of viability of the gene knockout under each condition (Fig. 1B, Table S1). In this analysis, sgRNAs targeting 3,726 unexpressed genes were used as negative controls. Gene Ontology (GO) analysis confirmed our previous result: that the depletion of genes encoding subunits of the mitochondrial respiratory chain caused loss of viability, as expected (Fig. S1A).

Unexpectedly, the genes whose depletion promotes relative viability in galactose were enriched in splicing-related GO terms, including “spliceosomal complex” (P<10⁻⁵) and “U1 snRNP” (P<10⁻⁴) (Fig. 1C, S1B). Hits with these GO terms included U1 snRNP-specific subunits (LUC7L2, SNRPA, SNRNP70), heterogeneous nuclear ribonucleoproteins (HNRNPD, HNRNPU), splicing factors (SF3B2, SFPQ), RNA helicases (DHX8, DDX47), an LSm-family protein (LSM1), an exon junction complex protein (ACIN1), and a polyadenylate-binding protein (PABPC1). Other gene expression pathways were not significantly enriched.

We used CRISPR/Cas9 and sgRNA sequences from the screening library to disrupt the expression of six representative genes identified in the screen that were not previously linked to energy metabolism, including LUC7L2, SNRPA (U1A), SNRNP70 (U1–70K), ACIN1, HNRNPD and PABPC1. We included a gene encoding a subunit of respiratory complex I (NDUFB5) as a control with known impact on metabolism, and measured oxygen consumption rates (OCR), a proxy for OXPHOS, and extracellular acidification rates (ECAR), a proxy for glycolysis, in transduced K562 cells. Notably, depletion of several of these genes significantly increased basal, maximal and ATP-linked OCR (Fig. 1D, S1C–D), while also decreasing ECAR (Fig. 1E), suggesting rewiring of metabolism from glycolysis to OXPHOS. In fact, the OCR/ECAR ratio was significantly increased upon depletion of all six selected genes (Fig. 1F), whereas it decreased in NDUFB5-depleted cells, as expected. Together, these results confirm that silencing certain pre-mRNA splicing genes boosts oxidative energy metabolism.

Expression of LUC7L2 Represses OXPHOS

Among the validated screening hits, LUC7L2 showed the most robust phenotype. This relatively unstudied gene belongs to the LUC7 family together with LUC7L and LUC7L3, which are all homologs of yeast LUC7, a U1 snRNP protein involved in pre-mRNA splicing (Fortes et al., 1999). Mutations, haploinsufficiency, and complete loss of LUC7L2 are all associated with poorer survival in myelodysplastic syndromes (MDS) (Singh et al., 2013), while Arabidopsis LUC7 genes contribute to development and stress responses (de Francisco Amorim et al., 2018). None of the LUC7 family members have been previously linked to energy metabolism.

To further investigate the function of LUC7L2, we used CRISPR/Cas9 to generate single-cell clones in which the expression of LUC7L2 was ablated (Fig. S2A–C). We observed that LUC7L2^KO K562 cells grew more slowly relative to wild-type in standard 25mM glucose cell culture conditions (Fig. 2A). However, they grew comparatively faster than LUC7L2^WT cells when glycolysis was limiting, either pharmacologically by treatment with 2-deoxyglucose (2-DG) or when glucose was replaced by galactose (Fig. 2B). In contrast, LUC7L2 depletion sensitized cells to killing by pharmacologic inhibition of OXPHOS (Fig. 2C).

Figure 2: LUC7L2 Impacts Metabolic State-Dependent Cell Growth and Bioenergetics. See also Figure S2.

(A–C) Cell proliferation of LUC7L2^KO K562 cells grown in (A) glucose and (B) treated with 2-deoxyglucose (2-DG) or when glucose was replaced by galactose or (C) glucose with OXPHOS inhibitors. (D) Respiratory parameters of LUC7L2^KO cells as determined by oxygen consumption rate (OCR). (E) Basal glycolytic activity in LUC7L2^KO cells as determined by extracellular acidification rate (ECAR). (F) Relative mtDNA abundance and (G) citrate synthase activity of LUC7L2^KO cells. All data are shown as mean ±SEM (n≥3). *P< 0.05, **P<0.01, ***P<0.001, t-test relative to LUC7L2^WT cells.

Next, we characterized the bioenergetic consequences of LUC7L2 depletion. OCR measurement confirmed our initial observation that LUC7L2 represses OXPHOS and, accordingly, all measured respiratory parameters in LUC7L2^KO K562 and HAP1 cells were increased relative to controls (Fig. 2D, S2D–F). As expected from our initial validation, LUC7L2-depleted K562 clones exhibited less ECAR (Fig. 2E). To test whether the global abundance of mitochondria was affected by LUC7L2 depletion, we quantified mitochondrial DNA copy number and citrate synthase activity and observed no differences (Fig. 2F–G). Electron microscopy also confirmed the absence of gross differences in mitochondrial abundance or ultrastructure in these cells (Fig. S2G). Collectively, our results indicate that LUC7L2 impacts metabolic state-dependent cell growth and bioenergetics. While LUC7L2 loss does not appear to affect the gross abundance of mitochondria, it influences the balance between activity of glycolysis and OXPHOS.

Metabolic Basis of the Shift from Glycolysis to OXPHOS in LUC7L2-Depleted Cells

The rewiring of cellular bioenergetics upon LUC7L2 depletion prompted us to analyze the abundance of metabolic intermediates central to glycolysis and OXPHOS. We used LC-MS to quantify the relative steady-state levels of 122 metabolites in cell pellets as well as the absolute consumption/release rates of 22 media metabolites (Table S2–3). Consistent with our ECAR results, we observed reduced rates of glucose uptake, lactate secretion, and media acidification as well as a dramatically decreased media lactate/pyruvate ratio in LUC7L2^KO K562 cells, all consistent with decreased glycolysis (Fig. 3A–C, S3C, S3F). In pellets of LUC7L2^KO, we observed significant accumulation of glucose, glucose-6-phosphate and of the fructose-6-phosphate/glucose-1-phosphate isomers. Notably, levels of fructose-1,6-bisphosphate, the fourth intermediate of glycolysis and the product of phosphofructokinase (PFK), were significantly reduced in the absence of LUC7L2. Thus, we observed accumulation of the substrate of PFK and depletion of its product, identifying this enzyme as a metabolic crossover in LUC7L2^KO cells. Accordingly, a loss of function mutation in PFKM leads to skeletal muscle glycogen accumulation (Tarui et al., 1965), and we observed significant accumulation of glycogen and its precursor UDP-glucose in these cells (Fig. 3A).

Figure 3: Metabolite Analysis in LUC7L2-Depleted Cells Reveals Crossovers at Phosphofructokinase and System X_c⁻. See also Figure S3 and Table S2–3.

(A) Intracellular levels of metabolites in LUC7L2^KO K562 cells as determined by LC-MS. (B) Extracellular levels of metabolites as determined by LC-MS analysis of the spent media from (A). Positive and negative values illustrate metabolite secretion and consumption by the cells, respectively. All data are shown as mean ±SEM (n = 5–8). *P<0.05, **P<0.01, ***P<0.001, t-test relative to LUC7L2^WT. (C) Media acidification of LUC7L2^KO K562 cells grown in glucose.

Our metabolite analysis also provided insight into how mitochondrial metabolism is rewired in LUC7L2^KO cells. We observed accumulation of four of the five TCA metabolites analyzed (2-ketoglutarate, succinate, fumarate, malate) as well as two TCA-cycle derived metabolites (2-hydroxyglutarate, aspartate) (Fig. 3A, S3A). Glutamine is an important fuel that can contribute glutamate to the TCA cycle (Reitzer et al., 1979), but glutamine consumption and intracellular glutamine levels remained unchanged (Fig. 3A–B). In contrast, intracellular glutamate accumulated in LUC7L2^KO cells. Intracellular glutamate may either be converted into 2-ketoglutarate to serve as an anaplerotic input into the TCA cycle or be exported out of the cell in exchange for cystine via the system X_c⁻, a plasma membrane antiporter encoded by two subunits, SLC7A11 (xCT) and SLC3A2 (4F2) (Sato et al., 1999). The system X_c⁻ has previously been implicated in the survival of cells in low glucose conditions (Shin et al., 2017, Koppula et al., 2017). Importantly, we found that while intracellular glutamate accumulated in LUC7L2-depleted cells, its secretion to the culture media was significantly reduced (Fig. 3B). These observations pointed to a second crossover at the level of the system X_c⁻, as LUC7L2-depleted cells also consumed less media cystine (Fig. 3B, S3E).

Collectively, analysis of steady-state intracellular metabolites as well as consumption and release of media metabolites indicate that the loss of LUC7L2 modulates two key crossovers in energy metabolism: (1) upper glycolysis/glycogen storage at PFK, and (2) glutamate oxidation/secretion via the exchange of cystine and glutamate at the transporter system X_c⁻. The direction of these metabolic changes is concordant with the observed shift from glycolysis to OXPHOS in cells lacking LUC7L2.

LUC7L2 is a U1 snRNP Subunit Involved in Pre-mRNA Splicing

Next, we sought to determine how loss of LUC7L2 leads to remodeling of cellular energy metabolism. LUC7L2 is not a well-studied gene, but its yeast and plant homologs are components of the U1 snRNP. Hence, we predicted that LUC7L2 encodes a component of the mammalian snRNP, impacting splicing and expression of genes that might be influencing energy metabolism.

We used confocal microscopy to confirm the nuclear localization of LUC7L2 and observed that it localized in SRSF2-positive nuclear speckles, which are themselves enriched for the pre-mRNA splicing machinery (Rino et al., 2007) (Fig. 4A). We next immunoprecipitated LUC7L2 and discovered 29 interacting proteins using mass spectrometry, including our validated screening hits SNRPA and SNRNP70 and most other known U1 subunits (Fig. 4B–C, Table S4).

Figure 4: LUC7L2 Encodes a U1 snRNP Subunit Involved in Pre-mRNA Splicing. See also Figure S4 and Table S4–7.

(A) Confocal microscopy of a single nucleus from a HeLa cell expressing LUC7L2-GFP and immunolabeled with antibodies to SRSF2. (B) LUC7L2-interacting proteins as determined by IP-MS (n = 2). (C) Representation of LUC7, SNRPA and SNRNP70 on the yeast U1 snRNP (PDB 5UZ5) (Li et al., 2017). (D) Proportion of eCLIP peaks mapping to splicing snRNAs in HeLa and K562 cells (eCLIP n = 2, each). (E) Representation of the genes bound by LUC7L2 at P<10⁻⁴. (F) Proportion of LUC7L2 eCLIP peaks in pre-mRNAs at P<10⁻⁴. (G) Meta-analysis of LUC7L2 binding sites across shared eCLIP peaks at P<10⁻⁴. (H) Differential gene expression in LUC7L2^KO cells (n = 3 for each cell type and each genotype) as determined by RNA deep-sequencing at FDR<10⁻⁴ and >|1.5| fold change. (I) Alternative splicing events seen in LUC7L2^KO cells as determined by rMATS at FDR<0.1 and |Δψ|>0.05 (n = 3 for each cell type and each genotype). (J) Types of alternative splicing in LUC7L2^KO cells with SE: skipped exon; MXE: mutually-exclusive exons; A5SS: alternative 5′ splice site; A3SS: alternative 3′ splice site; RI: retained intron. (K) Alternative events presenting an eCLIP peak at a 250-nucleotide distance from splicing events at P<10⁻² (in darker shade).

We next sought to identify the transcripts that are bound by endogenous LUC7L2 and used enhanced crosslinking and immunoprecipitation coupled to RNA deep-sequencing (eCLIP) (Van Nostrand et al., 2016). In agreement with the proposed role for LUC7L2 as a U1 snRNP subunit, we found that LUC7L2 crosslinks to the U1 snRNA, and to a lesser extent to the U6 snRNA, which comes in close proximity during the transfer of the 5′ splice site (5′SS) during spliceosome assembly (Plaschka et al., 2018) (Fig. 4D). We identified eCLIP peaks for LUC7L2 in 5,595 genes in HeLa cells and 3,378 genes in K562 cells. Of these, 2,614 were shared (expression-corrected overlap significance P<10⁻³, Poisson) (Fig. 4E, Table S5). Within pre-mRNAs, we found that LUC7L2 bound mainly to exons and introns (Fig. 4F). A meta-analysis revealed that LUC7L2 preferentially bound near splice sites (Fig. 4G). The pattern of binding near splice sites is consistent with association of LUC7L2 with U1 snRNP complexes, which recognize 5′SS motifs and also interact with U2 snRNP bound upstream of 3′ splice sites (De Conti et al., 2013)

To understand the impact of LUC7L2^KO on gene expression, we profiled the transcriptomes of LUC7L2^KO HeLa and K562 cells. Loss of LUC7L2 changed the expression of ~1000–1500 genes in each cell type (FDR<10⁻⁴ and >50% absolute fold change), 149 of which were shared (overlap not significant) (Fig. 4H, S4, Table S6). Splicing analysis using rMATS (Shen et al., 2014) identified 4,815 and 4,704 alternative splicing events in HeLa and K562 cells, respectively (FDR<0.1 and |Δψ|>0.05, where ψ represents “percent spliced in” and Δψ is the change in ψ following gene depletion) (Fig. 4I, Table S7). In all, 379 splicing changes were shared between both cell types (expression-corrected overlap significance P<10⁻³, Poisson) (Fig. 4I). Skipped exons (SE) comprised the majority of LUC7L2-induced alternative splicing events (Fig. 4J), and approximately 20–25% of these splicing events had a LUC7L2 binding site within a distance of 250 nucleotides from a splice site (Fig. 4K). The binding of LUC7L2 in close proximity to a SE event supports a direct effect of LUC7L2 on these exons, although absence of an adjacent CLIP peak does not imply that regulation is not direct, since CLIP does not detect all binding, and some regulation may occur across longer distances (Lovci et al., 2013, Van Nostrand et al., 2020).

Our RNA-seq and eCLIP analyses indicate that LUC7L2 impacts pre-mRNA splicing and gene expression. Importantly, among the genes most significantly impacted in the splicing analysis was PFKM (Fig. 4I), and amongst the most differentially expressed genes was SLC7A11 (Fig. 4H). Both of these correspond to metabolic crossovers identified above (Fig. 3). We next sought to validate these splicing and gene expression changes as being downstream of LUC7L2 to determine whether they might contribute to the observed metabolic phenotype.

LUC7L2 Promotes Expression of PFKM and Suppresses Glycogen Storage

We first investigated the changes in PFKM and other glycolytic enzymes following LUC7L2 depletion. Most transcripts encoding glycolytic enzymes were bound by LUC7L2 in eCLIP (FDR<10⁻⁴), and we found alternative splicing in PFKM, PKM, ALDOA and ENO3 (FDR<0.1, Table S5, 7). Of all glycolytic enzymes, PFKM was the most significantly alternatively spliced in both LUC7L2^KO HeLa and K562 cells, with two altered splicing events detected: increased usage of an alternative 5′SS between exons 11 and 12 (Δψ = 0.29, FDR<10⁻¹¹), and exon 12 skipping (Δψ = −0.19, FDR<10⁻¹²). LUC7L2 binding was observed near the 5′SS of both alternatively spliced exons (Fig. 5A). Notably, our unbiased metabolomics experiment suggested decreased PFKM activity in LUC7L2^KO cells (Fig. 3A), and both alternative splicing events were indeed expected to reduce PFKM activity. First, the alternative 5′SS whose usage increases upon LUC7L2-depletion results in inclusion of a premature termination codon (PTC) (Fig. 5A). Second, skipping of exon 12 deletes 30 amino acids in the catalytic site of the enzyme (Fig. S5A). We validated the increased skipping of exon 12 in LUC7L2^KO cells and observed a global decrease in the abundance of PFKM protein (Fig. 5B). To confirm the effects of this splicing change on PFKM levels, we designed an antisense oligonucleotide (ASO) targeting the 5′SS of this exon. Similar to LUC7L2 depletion, acute ASO treatment led to exon 12 skipping and decreased PFKM protein abundance, indicating that the skipping of PFKM exon 12 likely yields a less stable protein (Fig. 5B).

Figure 5: Role of LUC7L2-Mediated PFKM and SLC7A11 Alternative Splicing in Energy Metabolism. See also Figure S5.

(A) Representation of PFKM exons 10–13, LUC7L2 binding sites as determined by eCLIP, antisense oligonucleotides (ASO) targeting sites and the expected transcripts. Ψ: percent spliced in reported by rMATS in K562 cells. E: exon. A negative Δψ value indicates exon skipping. PTC: premature termination codons. (B) RT-PCR (top) and immunoblot (bottom) of LUC7L2^KO K562 cells (left) or HAP1 cells treated for 48h with ASO targeting the 5′SS of PFKM exon 12 (right). (C) Relative ECAR (n = 3–5) and (D) glycogen in LUC7L2^KO K562 cells expressing control cDNAs (GFP) or PFKM cDNA (n = 2–4). (E) Representation of SLC7A11 exons 6–10 as in (A). (F) RT-PCR of LUC7L2^KO K562 cells with primers amplifying transcripts corresponding to SLC7A11 exons 6–12. SE: skipped exon. (G) Immunoblot on LUC7L2^KO K562 cells with antibodies to SLC7A11 and ACTIN. (H) Cell viability of LUC7L2^Rescue (corresponds to LUC72L2^KO expressing LUC7L2 cDNA) and LUC7L2^KO HAP1 cells grown for 24h in galactose relative to glucose (n = 3). SAS: 500μM sulfasalazine. (I) RT-PCR (top) and immunoblot (bottom) of HAP1 cells treated for 48h with ASOs targeting the 5′SS of exon 7 and/or exon 9 of SLC7A11. (J) Media glutamate (n = 4), (K) representative seahorse trace (shown as mean ±SD), and (L) viability in galactose of HAP1 cells treated for 48h with the indicated ASOs (n = 3). All data are shown as mean ±SEM (unless otherwise stated) with * P<0.05, ** P<0.01, ***P<0.001, t-test relative to control.

The short time frame of the acute ASO treatment and the potential for long-term toxicity are not compatible with studying slow processes such as glycogen storage, so we opted for cDNA rescue to experimentally address the role of PFKM in mediating aspects of the phenotype of LUC7L2-depleted cells (Fig. S5B). We observed that overexpression of PFKM alone was insufficient to restore the bulk of glycolysis in a LUC7L2^KO background, possibly due to the aforementioned consequences of LUC7L2 depletion on other glycolytic enzymes. However, it was sufficient to restore normal glycogen storage (Fig. 5C–D). Together, our data indicate that LUC7L2 is required for normal splicing and expression of full-length PFKM, and that its absence favors glycogenesis.

Expression of LUC7L2 is Limiting for Splicing of the Cystine/Glutamate Antiporter SLC7A11 (xCT)

Next, we addressed the genetic basis of the second crossover observed in LUC7L2-depleted cells at the level of the cystine/glutamate antiporter. Previous studies have shown that this antiporter controls survival in glucose-limiting conditions and OXPHOS activity (Shin et al., 2017, Koppula et al., 2017). Accordingly, we found that while inhibition with sulfasalazine prevented glutamate secretion and promoted maximal respiration, over-expression of the SLC7A11 subunit alone was sufficient to restore glutamate secretion in LUC7L2^KO K562 cells (Fig. S5C–G), confirming that SLC7A11 is limiting for glutamate oxidation and OXPHOS. We next examined whether subunits of the system X_c⁻ require LUC7L2 for their expression. SLC3A2 transcripts were not affected by LUC7L2 depletion but we found a significant reduction in transcript abundance of SLC7A11 in both LUC7L2^KO HeLa and K562 cells (Fig. S5H). Reduced inclusion of SLC7A11 exon 7 was detected in these cells (Δψ = −0.25, FDR<6 × 10⁻¹¹) and we noticed that exon 9 also showed reduced inclusion, which was previously annotated and which we confirmed by RT-PCR and Sanger sequencing (Fig. 5E–F, S5I). Skipping of exons 7 and/or 9 yield PTCs in SLC7A11 transcripts, likely reducing mRNA abundance via nonsense-mediated mRNA decay (NMD) (Fig. S5J). LUC7L2 binding was also observed at the 5′SS of these exons by eCLIP (Fig. 5E), suggesting a direct effect on their splicing. Accordingly, depletion of LUC7L2 led to a decrease in SLC7A11 expression in K562, HeLa and HAP1 cells (Fig. 5G, S5K). Reduced expression of SLC7A11 was also observed upon depletion of SNRPA and SNRNP70, the two other U1 snRNP subunits identified in our screen (Fig. S5L), and LUC7L2 over-expression was sufficient to stabilize SLC7A11 transcript and protein in K562 cells (Fig. S5M–N).

To test for a direct contribution of SLC7A11 splicing to the metabolic phenotypes observed in LUC7L2^KO, we focused on LUC7L2^KO HAP1 cells, a cell line in which the role of this antiporter in antagonizing viability in low glucose conditions is well characterized (Shin et al., 2017). As in K562 cells, we found that LUC7L2 depletion increased OXPHOS activity (Fig. S2F) and viability in galactose (Fig. 5H). We then treated wild-type HAP1 cells with ASOs targeting the 5′SS of SLC7A11 exons 7 and 9, where LUC7L2 and the U1 snRNP bind. Similar to the phenotype of LUC7L2^KO cells, we found that acute ASO treatment led to skipping of these exons, prevented SLC7A11 expression and glutamate secretion, and boosted maximal respiration and viability in galactose (Fig. 5I–L), all resembling the phenotype observed upon LUC7L2 depletion. We conclude that splicing of SLC7A11 (xCT) is sensitive to perturbations of U1 snRNP components including LUC7L2, and that altered splicing of SLC7A11 induced by LUC7L2 depletion contributes to the metabolic and viability phenotypes observed in LUC7L2^KO cells (Fig. 1, 3).

LUC7L2 Depletion Causes Secondary Accumulation of Respiratory Chain Complexes

To obtain a comprehensive view of proteome remodeling in LUC7L2-depleted K562 cells, we next performed global quantitative proteomics (Fig. 6A, S6A–D). As expected, we observed depletion of PFKM in LUC7L2^KO cells as well as differential expression of multiple glycolytic enzymes, including rate-determining factors (Tanner et al., 2018) such as hexokinases HK1/2 (P<10⁻²), the glucose transporter GLUT1 (P<10⁻⁵) and the lactate transporter SLC16A1 (P<0.02) (Table S8). Transcripts of these factors were generally bound by LUC7L2 but splicing changes were not always observed in LUC7L2^KO cells, suggesting regulation by both direct and indirect mechanisms. SLC7A11 was not detected in the proteomics analysis, possibly owing to its hydrophobicity. Importantly, while abundances of mitochondria and splicing-related proteins were not globally affected, GO analysis revealed a strong enrichment for OXPHOS proteins in LUC7L2^KO cells, with “NADH dehydrogenase complex” as the top-scoring term associated with up-regulated proteins (P<10⁻⁴) (Fig. S6C).

Figure 6: Proteomic Analysis of LUC7L2^KO and Galactose-Grown Cells Reveals Secondary Complexes I+III+IV Accumulation. See also Figure S6 and Table S8.

(A) Volcano plots and immunoblots of OXPHOS protein expression in LUC7L2^KO and (B) in galactose-grown K562 cells. > indicates a protein not shown but reported in Table S8. (C) Blue-Native PAGE on a mitochondria-rich fraction isolated from LUC7L2^KO K562 cells and stained with coomassie or (D) immunoblotted with the indicated antibodies. Parallel blots in which the same lysate was loaded were used to avoid antibodies cross-reactivity. SCs: supercomplexes. CI-V: complexes I to V. (E) Immunoblot on K562 cells expressing Cas9 and treated with sgRNAs targeting glycolytic enzymes with the indicated antibodies. (F) Model of the secondary regulation of the respiratory chain by LUC7L2 and galactose.

Given this inverse relationship between abundance of LUC7L2 and OXPHOS proteins, we analyzed their relative expression in published large-scale proteomics studies. Similar to our experimental observation in cells, LUC7L2 was significantly anti-correlated with OXPHOS protein expression across organs in two in vivo mouse tissue proteomic atlases (P<10⁻⁴², Wilcoxon) (Geiger et al., 2013, Huttlin et al., 2010) (Fig. S6E). A similar anti-correlation was observed in a proteomics study of brains from healthy subjects or from patients with neurodegenerative disease and showed that LUC7L2 protein accumulates in patients, whereas OXPHOS proteins are decreased in Parkinson’s disease (Fig. S6F) (Ping et al., 2018). OXPHOS protein abundance also decreases in skeletal muscle with aging, as recently confirmed by a proteomic study on 58 skeletal muscle biopsies (Ubaida-Mohien et al., 2019). In this study too, we observed significant up-regulation of LUC7L2 with age and anti-correlation with OXPHOS (P<10⁻²⁶, Wilcoxon) (Fig. S6E). Thus, in these four in vivo datasets we observed inverse relationships between the abundance of LUC7L2 and OXPHOS proteins resembling LUC7L2^KO depletion.

OXPHOS proteins also accumulate in glucose-limiting conditions (Rossignol et al., 2004). To compare this condition to LUC7L2^KO, we performed proteomics analysis of galactose-grown cells and observed strong up-regulation, as expected (Fig. 6B). Importantly, we noticed a similar pattern in both LUC7L2^KO cells and in galactose: the up-regulation of subunits of the respiratory chain (RC) complex I, III, IV. In contrast, protein subunits of RC complex II and the ATP synthase (V), as well as from other mitochondrial protein complexes and the gene expression machinery were not affected, or rather decreased (Fig. 6A–B, S6A–B,G). Transcripts of RC complexes I+III+IV did not accumulate in LUC7L2^KO, suggesting an effect occurring after RNA processing (Fig. S6A). However, these same RC complexes I+III+IV have the ability to interact within the mitochondrial inner membrane to form “supercomplexes” (SCs), and their proposed roles in reinforcing RC complex stability (Acin-Perez et al., 2004) could explain their accumulation in LUC7L2^KO cells. Accordingly, blue-native PAGE confirmed accumulation of higher molecular weight RC complexes in both K562 and HeLa LUC7L2^KO and galactose-grown cells (Fig. 6C–D, S6H).

Galactose growth and LUC7L2 depletion both attenuate glycolysis, and we next directly tested whether lower glycolytic rates could explain increased RC abundance. We used CRISPR/Cas9 to acutely deplete six glycolytic enzymes and subsequently measured OXPHOS proteins (Fig. 6E, S6I). Importantly, we found that depletion of these genes generally led to the accumulation of the same RC subunits as LUC7L2^KO and galactose, which was particularly apparent upon depletion of ENO1 and TPI1. While the mechanism by which attenuated glycolysis leads to RC complex accumulation was not investigated here, our observations indicate that the increased abundance of RC complexes in LUC7L2^KO cells is likely secondary to the effect of this gene on glycolysis (Fig. 6F), and possibly involves the stabilization of mitochondrial SCs.

Cross-Regulation and Partial Redundancy Within the LUC7 Family

Finally, we investigated the role of LUC7L2 paralogs LUC7L and LUC7L3 in energy metabolism. All three LUC7 genes encode similar proteins (Tufarelli et al., 2001) (Fig. 7A). Analysis of LUC7 proteins across a mouse proteomics atlas (Geiger et al., 2013) revealed that although a general anti-correlation with OXPHOS was observed for all three proteins, the expression patterns of individual members of the LUC7 family were not identical (Fig. S7). For example, LUC7L was present at higher levels in the brain, while expression of LUC7L3 was not detectable in the spleen, suggesting tissue-specific roles in alternative splicing.

Figure 7: Pre-mRNA Splicing and Partial Redundancy Within the LUC7 Family. See also Figure S7.

(A) Phylogenetic tree of the LUC7 protein family. LUC7 is from S. cerevisiae. ZnF: Zinc-finger domain. S/R-rich: Serine and arginine-rich domain. (B) Quantitative PCR detecting LUC7 family transcripts in LUC7L2^KO K562 cells (n = 3). (C) Representation of LUC7L exons 1–3, LUC7L2 binding sites as determined by eCLIP, and the expected transcripts. Ψ: percent spliced in reported by rMATS in K562 cells. I: intron. E: exon. (D) RT-PCR amplifying LUC7L exon 1 to exon 2. Arrowheads: retained entities in LUC7L. (E) Immunoblot of LUC7 proteins in cell lines expressing Cas9 and sgRNAs targeting the indicated genes using the indicated antibodies and (F) number of cells after 4 days of growth in glucose-containing media. (G) Oxygen consumption analysis of LUC7L2^KO K562 cells expressing cDNAs of LUC7 family members. All data are shown as mean ±SEM with * P<0.05, ** P<0.01, ***P<0.001, t-test relative to control.

With 85% protein identity, LUC7L is the closest homolog to LUC7L2. This gene was also one of the most impacted genes in LUC7L2^KO (Fig. 7B) and we found decreased inclusion of an annotated exon located in an intron between canonical exons 1 and 2 of LUC7L as well as decreased retention of the associated intron, all within a region highly bound by LUC7L2 (Δψ = −0.30 and −0.17, respectively) (Fig. 7C,D). Importantly, inclusion of the exon with or without intron retention results in a PTC in LUC7L, which likely leads to transcript degradation by NMD. Accordingly, protein analysis revealed strong up-regulation of LUC7L in LUC7L2^KO cells, indicating a repressive role for LUC7L2 (Fig. 7E, Table S8), an observation that is relatively common among paralogous RBPs (Spellman et al., 2007, Ni et al., 2007, Lareau et al., 2007)

To experimentally address the function of each of the LUC7 genes, we depleted each individually, as well as LUC7L and LUC7L2 together (Fig. 7E). We found that depletion of LUC7L or LUC7L2 individually led to a mild defect in cell proliferation (Fig. 7F) while depletion of LUC7L3 caused a strong growth defect. Simultaneous depletion of LUC7L and LUC7L2 showed a synthetic lethal phenotype, consistent with some degree of functional redundancy and cross-regulation. To directly address redundancy, we investigated whether members of the LUC7 family could rescue the metabolic phenotype observed in LUC7L2^KO cells. For this purpose, we stably expressed LUC7L, LUC7L2 or LUC7L3 in these cells and measured oxygen consumption. Importantly, we found that while expression of LUC7L2 alone was able to decrease OCR in wild-type cells (Fig. 7G), re-expression of any of the members of the family could restore normal oxygen consumption in LUC7L2^KO cells, to varying degrees (Fig. 5D). Together, our observations implicate all members of the LUC7 family in energy metabolism.

DISCUSSION

We have found that the expression of genes related to pre-mRNA splicing and the U1 snRNP impact the balance between glycolysis and OXPHOS. Genetic loss of any of three U1 snRNP-specific subunits (SNRPA, SNRNP70, LUC7L2) boosts cellular fitness when glycolysis is compromised. The U1 snRNP plays an essential role in pre-mRNA splicing and modulates other nuclear RNA-related processes such as cleavage and polyadenylation (Berg et al., 2012) and chromatin retention of non-coding RNAs (Yin et al., 2020). To our knowledge, our work is the first to establish a link between the expression of U1 snRNP components and the bioenergetic state of the cell.

In the current study, we were able to confirm that LUC7L2 is a genuine component of the U1 snRNP in humans, as has been observed for LUC7 homologs in yeast and plants (de Francisco Amorim et al., 2018, Fortes et al., 1999). Because LUC7 genes are expressed in a tissue-specific manner (Fig. S7), as are some of the transcripts of metabolic genes bound by LUC7L2, we speculate that the relative abundance of these paralogs in different cells may contribute to differences in levels of OXPHOS and glycolysis between different cell types.

We show that loss of LUC7L2 results in an activation of OXPHOS. The mechanisms linking LUC7L2 expression to energy metabolism are likely complex. Given that the U1 snRNP plays central roles in splicing and gene expression, we searched for changes in our transcriptome analysis that might help explain the observed metabolic phenotype. In the current study we have prioritized observations from our metabolic characterization of LUC7L2^KO cells (Fig. 1–3) and investigated in detail PFKM and SLC7A11, which we show contribute to key aspects of the metabolic phenotype following depletion of LUC7L2 (Fig. 5).

First, we found that LUC7L2 binds to and contributes to splicing and expression of PFKM (Fig. 5A). This enzyme alone did not explain the global decrease in glycolysis in LUC7L2^KO cells, as we found multiple instances of altered expression in glycolytic proteins in LUC7L2^KO (Table S8), including in rate-limiting factors (Tanner et al., 2018). We propose that these changes may collectively lead to the decreased glycolytic activity in LUC7L2^KO cells, and likely originate from both primary and secondary mechanisms. However, we show that altered splicing in PFKM is likely directly responsible for the glycogen storage phenotype in LUC7L2^KO cells. The differential role of PFKM in glycogen storage versus glycolysis is not unexpected from observations of patients with Tarui disease, who present minimal effects on circulating lactate at rest (Piirila et al., 2016), possibly due to the complex interplay between PFKM, its paralogs and allostery.

Second, we have identified alternative splicing events in SLC7A11 whereby skipping of exons 7 and 9 leads to loss of SLC7A11 (xCT) protein, which has a direct impact on OXPHOS. xCT has emerged as a critical regulator of metabolism and cell viability, and while high expression of this cystine/glutamate antiporter blunts glutamine anaplerosis and creates a dependency on glycolysis for ATP production (Koppula et al., 2017, Shin et al., 2017), genetic ablation of this cancer-related gene in tumors induces death by lipid peroxidation (Badgley et al., 2020). Expression of SLC7A11 is highly regulated and is known to be transcriptionally activated by ATF4 (Sato et al., 2004) and NRF2 (Shin et al., 2017) and repressed by P53 (Jiang et al., 2015). SLC7A11 activity is also regulated by direct mTORC2 phosphorylation (Gu et al., 2017). Our work adds another layer of complexity to the regulation of this transporter. In the future, it will be interesting to determine whether these splicing events are regulated in disease states to influence the sensitivity to lipid peroxidation.

In addition, we report that, in general, shifts from glycolysis to OXPHOS appear to be associated with an accumulation of RC complexes (Fig. 6). When cells are grown in galactose, or in glucose-grown cells following genetic ablation of LUC7L2 or of glycolytic enzymes, we observed a characteristic pattern of RC I+III+IV accumulation which prompted us to investigate assembly of the SCs. Future studies are needed to determine how the reduction in glycolysis impacts RC assembly.

Our work predicts that the existence of shifts in energy metabolism will accompany human conditions associated with mis-expression of LUC7L2 and the U1 snRNP. U1 dysregulation and mutations have been reported to occur in cancer (Shuai et al., 2019, Suzuki et al., 2019, Oh et al., 2020). Similarly, mutations in LUC7L2 or haploinsufficiency through loss of chromosome 7q are associated with MDS (Singh et al., 2013). 7q- models recapitulate differentiation defects observed in MDS (Kotini et al., 2015) and reintroduction of LUC7L2 is sufficient to restore differentiation. How LUC7L2 loss impacts hematopoietic stem cell differentiation and MDS is not clear, but our work raises the possibility of a bioenergetics-related mechanism. It is notable that other splicing factors associated with MDS also participate in the splicing of metabolic enzymes; for example, MDS-associated mutations in SRSF2 give rise to the same splicing change in PFKM (Zhang et al., 2015) as we observe in LUC7L2^KO. Our work has revealed a novel metabolic vulnerability of LUC7L2-depleted cells (Fig. 2C) and predicts that pharmacological blockade of OXPHOS using drugs such as metformin may be beneficial in these disorders.

While the current analysis has focused on the effects of LUC7L2 gene depletion, an exciting future direction is to determine whether metabolism is regulated by changes in expression of LUC7 family members in normal development or physiology. We observe a consistent, inverse correlation between the protein levels of LUC7L2 and OXPHOS across organs and during disease and aging (Fig. S6) Expression of LUC7L2 and its paralogs is known to respond to changes in the environment, including changes in oxygen tension (Kimura et al., 2004, Gao et al., 2011). It is notable that previous proteomics studies have established that LUC7L2 is post-translationally modified via phosphorylation and hydroxylation (Webby et al., 2009, Dephoure et al., 2008). In this context, it is possible that proteins of the LUC7 family, as well as other U1 snRNP components, might integrate signals such as oxygen and nutrients to balance the activity of major pathways in energy metabolism.

LIMITATION OF STUDY

Pre-mRNA splicing, gene expression and energy metabolism are cell type-specific processes. We report here that LUC7L2 impacts energy metabolism in three cellular models, but there are differences across cell types. For example, while we observed decreased levels of SLC7A11 in three LUC7L2^KO cell lines investigated (Fig. 5G, S5K), it led to an increase in net oxygen consumption only in K562 and HAP1 cells, but not in HeLa cells (Fig. S2E). It is possible that cells with different metabolic programs, for instance, based on low SLC7A11 and PFKM expression or cells with high expression of LUC7L and LUC7L3 might be indifferent to LUC7L2 depletion. Furthermore, although we observed that LUC7L2 depletion results in altered splicing in PFKM, SLC7A11 and other genes, and that LUC7L2 crosslinks near the regulated exons in PFKM and SLC7A11, this does not formally prove that the effect on splicing is direct. Our profiling experiments of LUC7L2^KO cells were likely biased towards abundant molecules, including in our sequencing analysis, as well as for the detection of peptides and metabolites by mass spectrometry. Thus, it is possible that additional low-abundance, splicing events in LUC7L2^KO cells may have been missed and that those events may affect additional metabolic aspects in LUC7L2-depleted cells.

STAR METHOD

RESOURCE AVAILABILITY

Lead Contact

Vamsi K. Mootha (vamsi@hms.harvard.edu)

Material Availability

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact.

Data and Code Availability

RNA sequencing data were deposited at GEO (GSE157917). Proteomics data were deposited at PRIDE (PXD021917). Plasmids were deposited at Addgene. Unedited gel scans are available in the supplemental information file (Data S1).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Cell Lines

K562 (ATCC CCL-243), HeLa (ATCC CCL-2) and 293T (ATCC CRL-3216) were obtained from ATCC and were re-authenticated by STR profiling at ATCC prior submission of the manuscript. HAP1 cells were from Horizon Discovery (C631). Cells were periodically tested to ensure absence of mycoplasma.

METHOD DETAILS

CRISPR Screen Re-Analysis

CRISPR screen analysis was performed either using a normalized Z-score approach (To et al., 2019) or using MAGeCK (Li et al., 2014). Raw sgRNA read counts were normalized to reads per million and then log₂ transformed using the following formula: Log₂(reads from an individual sgRNA / total reads in the sample10⁶ + 1) (To et al., 2019). Log₂ fold-change of each sgRNA was determined relative to the pre-swap control. For each gene in each replicate, the mean log₂ fold-change in the abundance of all 4 sgRNAs was calculated. Genes with low expression (log₂ FPKM<0) according to publicly available K562 RNA-seq dataset (sample GSM854403 in GEO seriesGSE34740) and essential genes previously reported (Arroyo et al., 2016) were removed. Log2 fold-changes were averaged by taking the mean across replicates. For each treatment, a null distribution was defined by the 3,726 genes with lowest expression. To score each gene within each treatment, its mean log₂ fold-change across replicates was Z-score transformed, using the statistics of the null distribution defined as above. To score each gene using MAGeCK, normalized sgRNA read counts from the duplicate in each condition were used as input for MAGeCK v0.5.3 to obtain a p-value and FDR for gene enrichment or depletion relative to the reference samples (pre-swap). MAGeCK was run with default parameters. For representation purpose, viability in galactose was defined as the - annexin V value.

Cell Culture and Cell Growth Assays

Unless otherwise specified, cells were maintained in DMEM containing 1mM sodium pyruvate (ThermoFisher Scientific) with 25 mM glucose, 10% fetal bovine serum (FBS, ThermoFisher Scientific), 50 μg/mL uridine (Sigma), and 100 U/mL penicillin/streptomycin (ThermoFisher Scientific) under 5% CO₂ at 37°C. Cells were counted using a ViCell Counter (Beckman) and only viable cells were considered. Drugs were diluted in the same culture media for cell growth assays and compared to the solvent control (DMSO or water). For galactose growth assays, FBS was replaced by dialyzed FBS (Life Technologies) and glucose was replaced by an equivalent amount of galactose.

Gene-Specific CRISPR-Cas9 Knockouts

The two best sgRNAs from the Avana-library were ordered as complementary oligonucleotides (Integrated DNA Technologies) and cloned in pLentiCRISPRv2. An sgRNA targeting EGFP was used as a negative control. Lentiviruses were produced according to Addgene’s protocol (Sanjana et al., 2014) and 24h post-infection cells were selected with 2mg/mL puromycin (ThermoFisher Scientific) for 48h. Cells were then maintained in routine culture media for 10–20 addition days before analysis. Gene disruption efficiency was verified by qPCR and/or immunoblotting. For HAP1 cells, a LUC7L2^KO cell line expressing LUC7L2 cDNA was used as control (LUC7L2^Rescue). For acute treatment, K562 cells were transduced with a high titer of sgRNAs targeting LUC7L2 and analyzed after 7 days. Sequences of the sgRNAs used are in the Key Resources Table.

KEY RESOURCE TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Actin (AC40)	Sigma	Cat # A4700; RRID: AB_476730
Actin (EPR16769)	Abcam	Cat # ab179467; RRID: AB_2737344
Anti-SC35 [SC-35]	Abcam	Cat # ab11826; RRID: AB_298608
COX1	Abcam	Cat # ab14705; RRID AB_2084810
Enolase-1	Cell Signaling Technology	Cat #3810; RRID AB_2246524
FLAG M2	Sigma	Cat # F1804; RRID: RRID: AB_262044
Goat Anti-Mouse IgG H&L (Alexa Fluor® 488)	Abcam	Cat # ab150113; RRID AB_2576208
Goat Anti-Rabbit IgG H&L (Alexa Fluor® 647)	Abcam	Cat # ab150079; RRID AB_2722623
Hexokinase II (C64G5) Rabbit mAb	Cell Signaling Technology	Cat # 2867; RRID AB_2232946
HRP-linked anti-mouse IgG	GE Healthcare	Cat # NA934; RRID: AB_772206
HRP-linked anti-rabbit IgG	GE Healthcare	Cat # NXA931; RRID: AB_772209
Human OXPHOS cocktail (ATP5A, SDHB, UQCRC2, COX2, NDUFB8)	Abcam	Cat # ab110411; RRID: AB_2756818
IRDye 680RD Goat anti-Mouse IgG (H + L)	LI-COR Biosciences	Cat # 926–68070; RRID: RRID: AB_10956588
IRDye 680RD Goat anti-Rabbit IgG (H + L)	LI-COR Biosciences	Cat # 926–68071; RRID: RRID: AB_10956166
IRDye 800CW Goat anti-Mouse IgG (H + L)	LI-COR Biosciences	Cat # 926–32210; RRID: RRID: AB_621842
IRDye 800CW Goat anti-Rabbit IgG (H + L)	LI-COR Biosciences	Cat # 926–32211; RRID: RRID: AB_621843
LUC7L (43-I)	Santa Cruz	Cat # 101075; RRID: AB_2139473
LUC7L2	Sigma	Cat # HPA051631; RRID: AB_2681558
LUC7L3	Sigma	Cat # HPA018484; RRID: AB_1847248
NDUFS3	ThermoFisher Scientific	Cat # 15066–1-AP; RRID: AB_2151109
PFKM	Abcam	Cat # ab97353; RRID: AB_10680060
PFKP (D4B2) Rabbit mAb	Cell Signaling Technology	Cat # 8164; RRID AB_2713957
PGAM1 (D3J9T) Rabbit mAb	Cell Signaling Technology	Cat # 12098; RRID AB_2736922
SLC7A11 (D2M7A)	Cell signaling	Cat # 12691; RRID: AB_2687474
TOMM20	Santa Cruz	Cat # SC-11415; RRID: AB_2207533
TUBB	ThermoFisher Scientific	Cat # MA5–16308; RRID: AB_2537819
Chemicals, peptides, and recombinant proteins
2-Deoxy-D-glucose	Sigma	Cat # D8375
Agilent Seahorse XF Base Medium	Agilent	Cat #03334–100
Ambion RNase I, cloned, 100 U/μL	Ambion	Cat # AM2295
Anti-FLAG® M2 Magnetic Beads	Sigma	Cat # M8823
Antimycin A	Sigma	Cat # A8674
Carbonyl cyanide 3-chlorophenylhydrazone	Sigma	Cat # C2759
Chloramphenicol	Sigma	Cat # C0378
Coomassie Brilliant Blue R-250 Staining Solution	Biorad	Cat # 1610436
Cycloheximide	Sigma	Cat # C4859
DMEM, high glucose, pyruvate	ThermoFisher Scientific	Cat # 11995073
DMEM, no glucose, no glutamine, no phenol red	ThermoFisher Scientific	Cat # A1443001
Dulbecco’s PBS	Sigma	Cat # D8537
Endo-Porter (PEG)	Gene Tools	Cat # OT-EP-PEG-1
ExoSAP-IT PCR Product Cleanup Reagent	ThermoFisher Scientific	Cat # 78201.1.ML
EXPRESS [35S]-protein labeling mix	Perkin Elmer	Cat # NEG772
Fetal Bovine Serum	ThermoFisher Scientific	Cat #26140079
Fetal Bovine Serum, dialyzed	ThermoFisher Scientific	Cat # 26400044
FLAG peptide	Sigma	Cat # F3290
FluorSave	EMD Millipore	Cat # 345789
G-418 solution	ThermoFisher Scientific	Cat # 10131035
Galactose	Sigma	Cat # G5388
Glucose	Sigma	Cat # G7021
Hoescht 33342	ThermoFisher Scientific	Cat # H1399
Pierce 16% Formaldehyde (w/v), Methanol-free	ThermoFisher Scientific	Cat # 28906
L-Glutamine	ThermoFisher Scientific	Cat # 25030081
M-MLV Reverse Transcriptase	Promega	Cat # M1701
Meclizine dihydrochloride	Sigma	Cat # SML0950
met/cys-free DMEM	ThermoFisher Scientific	Cat #21013024
Metformin / 1,1-Dimethylbiguanide HCl	Sigma	Cat # D150959
NEBNext® Ultra II Q5® Master Mix	New England BioLabs	Cat # M0544
Odyssey Blocking Buffer	LI-COR Biosciences	Cat # 927–40000
Odyssey® Blocking Buffer (PBS)	LI-COR Biosciences	Cat # 927–40100
Oligomycin A	Sigma	Cat # 75351
Penicillin-Streptomycin	ThermoFisher Scientific	Cat # 15140122
Protease Inhibitor Cocktail (100X)	Cell Signaling	Cat # 5871
Proteinase K, Molecular Biology Grade	New England BioLabs	Cat # P8107
Puromycin Dihydrochloride	ThermoFisher Scientific	Cat # A1113803
RNase-Free DNase Set	QIAGEN	Cat # 79254
RNaseA	Takara	Cat # 740505
RNaseOUT Recombinant Ribonuclease Inhibitor	ThermoFisher Scientific	Cat # 10777019
Sulfasalasine	Sigma	Cat # S0883
T4 Polynucleotide Kinase	New England BioLabs	Cat # M0201
T4 RNA Ligase 1 (ssRNA Ligase)	New England BioLabs	Cat # M0204
Universal Nuclease (Pierce)	ThermoFisher Scientific	Cat # 88702
Uridine	Sigma	Cat # U3003
XBridge BEH Amide column	Water	Cat # 186006091
ZIC-philic column	Merck	Cat # 150460
Critical commercial assays
Amplex red glutamic acid/glutamate oxidase assay kit	ThermoFisher Scientific	Cat # A22189
Citrate Synthase Activity Assay Kit	Abcam	Cat # ab119692
DC Protein Assay	Biorad	Cat # 5000112
Glycogen Assay Kit II (Colorimetric)	Abcam	Cat # ab169558
Hoechst 33342, Trihydrochloride, Trihydrate	ThermoFisher Scientific	Cat # H3570
NativeMark Unstained Protein Standard	ThermoFisher Scientific	Cat # LC0725
NativePAGE 3 to 12%, Bis-Tris, 1.0 mm, Mini Protein Gel, 10-well	ThermoFisher Scientific	Cat # BN1001BOX
NativePAGE Cathode Buffer Additive (20X)	ThermoFisher Scientific	Cat # BN2002
NativePAGE Running Buffer (20X)	ThermoFisher Scientific	Cat # BN2001
NativePAGE Sample Buffer (4X)	ThermoFisher Scientific	Cat # BN2003
Novex 10–20% Tris-Glycine Mini Gels	ThermoFisher Scientific	Cat # XP10200BOX
Novex 4–12% Tris-Glycine Mini Gels	ThermoFisher Scientific	Cat # XP04122BOX
Precision Plus Protein Kaleidoscope Prestained Protein Standards	Biorad	Cat # 1610375
RNeasy Mini Kit	QIAGEN	Cat # 74106
Seahorse XFe96 FluxPaks	Agilent	Cat # 102416–100
Trans-Blot Turbo Midi Nitrocellulose Transfer Packs	Biorad	Cat # 1704159
LUC7L Taqman Assay	ThermoFisher Scientific	Hs00216077
LUC7L2 Taqman Assay	ThermoFisher Scientific	Hs00255388
LUC7L3 Taqman Assay	ThermoFisher Scientific	Hs00895240
SLC7A11 Taqman Assay	ThermoFisher Scientific	Hs00921938
TBP Taqman Assay	ThermoFisher Scientific	Hs00427620
Deposited data
RNA-seq	GEO	GEO: GSE157917
Proteomics	PRIDE	PRIDE: PXD021917
Experimental models: cell lines
K562	ATCC	CCL-243
293T	ATCC	CRL-3216
HAP1	Horizon Discovery	C631
HeLa	ATCC	CCL-2
Oligonucleotides
ACIN1_sg1_as	Integrated DNA Technologies	AAACCAGAGATGCGAGAGTCATCAC
ACIN1_sg1_s	Integrated DNA Technologies	CACCGTGATGACTCTCGCATCTCTG
ACIN1_sg2_as	Integrated DNA Technologies	AAACCCCTCTGTGTCACTGTTTCC
ACIN1_sg2_s	Integrated DNA Technologies	CACCGGAAACAGTGACACAGAGGG
AluYb8_as	Integrated DNA Technologies	GAGACGGAGTCTCGCTCTGTC
AluYb8_probe	Integrated DNA Technologies	VIC-ACTGCAGTCCGCAGTCCGGCCT-MGBNFQ
AluYb8_s	Integrated DNA Technologies	CTTGCAGTGAGCCGAGATT
ENO1_sg1_as	Integrated DNA Technologies	AAACTGAAGTTTGCTGCACCGACTC
ENO1_sg1_s	Integrated DNA Technologies	CACCGAGTCGGTGCAGCAAACTTCA
ENO1_sg2_as	Integrated DNA Technologies	AAACAGTGGTGCTTCAACTGGTATC
ENO1_sg2_s	Integrated DNA Technologies	CACCGATACCAGTTGAAGCACCACT
HK2_sg1_as	Integrated DNA Technologies	AAACCGCCTTTGTTCTCCTTGATGC
HK2_sg1_s	Integrated DNA Technologies	CACCGCATCAAGGAGAACAAAGGCG
HK2_sg2_as	Integrated DNA Technologies	AAACTGTGCTTTGGGTGAAAGTAAC
HK2_sg2_s	Integrated DNA Technologies	CACCGTTACTTTCACCCAAAGCACA
HNRNPD_sg1_as	Integrated DNA Technologies	AAACCGTTCTTACTGGCGTCAATC
HNRNPD_sg1_s	Integrated DNA Technologies	CACCGATTGACGCCAGTAAGAACG
HNRNPD_sg2_as	Integrated DNA Technologies	AAACCCGAACTGCTCCTCCGACATC
HNRNPD_sg2_s	Integrated DNA Technologies	CACCGATGTCGGAGGAGCAGTTCGG
LUC7L_sg1_as	Integrated DNA Technologies	AAACCAAGTGTAACTTGCCACCGAC
LUC7L_sg1_s	Integrated DNA Technologies	CACCGTCGGTGGCAAGTTACACTTG
LUC7L_sg2_as	Integrated DNA Technologies	AAACAGTCGTGGATTTTGGTACATC
LUC7L_sg2_s	Integrated DNA Technologies	CACCGATGTACCAAAATCCACGACT
LUC7L2_sg1_as	Integrated DNA Technologies	AAACCTCTTAAAGCCAGGTCATGGG
LUC7L2_sg1_s	Integrated DNA Technologies	CACCGCCATGACCTGGCTTTAAGAG
LUC7L2_sg2_as	Integrated DNA Technologies	AAACTTACTTTCTGGGATTCCTCC
LUC7L2_sg2_s	Integrated DNA Technologies	CACCGGAGGAATCCCAGAAAGTAA
LUC7L3_sg1_as	Integrated DNA Technologies	AAACCGTCTGATCCTACGTTCTAC
LUC7L3_sg1_s	Integrated DNA Technologies	CACCGTAGAACGTAGGATCAGACG
LUC7L3_sg2_as	Integrated DNA Technologies	AAACCGAGAGCGTAAGTCCCGCGGC
LUC7L3_sg2_s	Integrated DNA Technologies	CACCGCCGCGGGACTTACGCTCTCG
ND2_as	Integrated DNA Technologies	CCTGCAAAGATGGTAGAGTAGATGA
ND2_probe	Integrated DNA Technologies	FAM-CCCTGGCCCAACCC-MGBNFQ
ND2_s	Integrated DNA Technologies	TGTTGGTTATACCCTTCCCGTACTA
NDUFB5_sg1_as	Integrated DNA Technologies	AAACAAACTCTGGAATTTCTGCTAGTTC
NDUFB5_sg1_s	Integrated DNA Technologies	CACCGAACTAGCAGAAATTCCAGA
NDUFB5_sg2_as	Integrated DNA Technologies	AAACAAACTCTTGATATGGGATGCTACAC
NDUFB5_sg2_s	Integrated DNA Technologies	CACCGTGTAGCATCCCATATCAAGA
PABPC1_sg1_as	Integrated DNA Technologies	AAACCCTTTTTCTGAGCTCGACCAC
PABPC1_sg1_s	Integrated DNA Technologies	CACCGTGGTCGAGCTCAGAAAAAGG
PABPC1_sg2_as	Integrated DNA Technologies	AAACCTCCTTGGGCTACGCGTATGC
PABPC1_sg2_s	Integrated DNA Technologies	CACCGCATACGCGTAGCCCAAGGAG
PFKM_exon12_5SS Antisense Oligonucleotide	Gene Tools	GGCCACACAGCCCAGTGACTTACCA
PFKM_RT_as	Integrated DNA Technologies	TCATGGAATGTGTCCAGGTG
PFKM_RT_s	Integrated DNA Technologies	CATCATCATTGTGGCTGAGG
PFKM_RT2_as	Integrated DNA Technologies	ACACATTCCATGAGGGGCAG
PFKM_RT2_s	Integrated DNA Technologies	CCGTGGTTCTCGTCTCAACA
PFKP_sg1_as	Integrated DNA Technologies	AAACCCGAGAGTTTGACACACATC
PFKP_sg1_s	Integrated DNA Technologies	CACCGATGTGTGTCAAACTCTCGG
PFKP_sg2_as	Integrated DNA Technologies	AAACTTGGGATCTGATCATCCGGC
PFKP_sg2_s	Integrated DNA Technologies	CACCGCCGGATGATCAGATCCCAA
PGAM1_sg1_as	Integrated DNA Technologies	AAACACCTGAGCCCGGCGGGCCAC
PGAM1_sg1_s	Integrated DNA Technologies	CACCGTGGCCCGCCGGGCTCAGGT
PGAM1_sg2_as	Integrated DNA Technologies	AAACCTGAAGCGGTTCTCCAGGTTC
PGAM1_sg2_s	Integrated DNA Technologies	CACCGAACCTGGAGAACCGCTTCAG
Random Primers	ThermoFisher Scientific	Cat # 48190011
SLC7A11_exon7_5SS Antisense Oligonucleotide	Gene Tools	CCAACTTGGACTTACCACTGCCACT
SLC7A11_exon9_5SS Antisense Oligonucleotide	Gene Tools	ATATACTTGTTAATATGCATTACCA
SLC7A11_RT_as	Integrated DNA Technologies	GGCAGATTGCCAAGATCTCAAG
SLC7A11_RT_s	Integrated DNA Technologies	TGCTGGCTGGTTTTACCTCAA
SNRNP70_sg1_as	Integrated DNA Technologies	AAACCCCGCTACGATGAGAGGTAAC
SNRNP70_sg1_s	Integrated DNA Technologies	CACCGTTACCTCTCATCGTAGCGGG
SNRNP70_sg2_as	Integrated DNA Technologies	AAACCTTACAAACACGCAGATGGC
SNRNP70_sg2_s	Integrated DNA Technologies	CACCGCCATCTGCGTGTTTGTAAG
SNRPA_sg1_as	Integrated DNA Technologies	AAACCCGCCTTGCACAGCCTTCTTC
SNRPA_sg1_s	Integrated DNA Technologies	CACCGAAGAAGGCTGTGCAAGGCGG
SNRPA_sg2_as	Integrated DNA Technologies	AAACGGCCTTTGTCATCTTCAAGG
SNRPA_sg2_s	Integrated DNA Technologies	CACCGGCCTTTGTCATCTTCAAGG
Standard Control Antisense Oligonucleotide	Gene Tools	CCTCTTACCTCAGTTACAATTTATA
TPI1_sg1_as	Integrated DNA Technologies	AAACCAGTCTTTGATCATGCCAGGC
TPI1_sg1_s	Integrated DNA Technologies	CACCGCCTGGCATGATCAAAGACTG
TPI1_sg2_as	Integrated DNA Technologies	AAACCGTGGGTGGTCCTGGGGCAC
TPI1_sg2_s	Integrated DNA Technologies	CACCGTGCCCCAGGACCACCCACG
Recombinant DNA
plentiCRISPR v2	Addgene	Plasmid # 52961
pLV-EF1a-IRES-Puro	Addgene	Plasmid # 85132
pLV-EF1a-IRES-Puro GFP-3xFLAG	This study	N/A
pLV-EF1a-IRES-Puro LUC7L2–3xFLAG	This study	N/A
pLV-EF1a-IRES-Puro PFKM	This study	N/A
pMD2.G	Addgene	Plasmid # 12259
psPAX2	Addgene	Plasmid # 12260
pWPI/Neo	Addgene	Plasmid # 35385
pWPI/Neo GFP	This study	N/A
pWPI/Neo LUC7L-3xFLAG	This study	N/A
pWPI/Neo LUC7L2–3xFLAG	This study	N/A
pWPI/Neo LUC7L3–3xFLAG	This study	N/A
pWPI/Neo SLC7A11-HA	This study	N/A
pLENTICRISPR_v2_PFKP_sgRNA1	To et al., 2019	N/A
pLENTICRISPR_v2_PFKP_sgRNA1	To et al., 2019	N/A
Software and algorithms
ImageJ	NIH	https://imagej.nih.gov/ij/
CLIPper	Yeo lab	https://github.com/YeoLab/clipper/wiki/CLIPper-Home
Clustal Omega	EMBL-EBI	https://www.ebi.ac.uk/Tools/msa/clustalo/
Gene Ontology GOrilla	Eden et al., 2009	http://cbl-gorilla.cs.technion.ac.il/
Gene set enrichment analysis (GSEA)	Broad Institute	https://www.gsea-msigdb.org/gsea/index.jsp
IGV	Broad Institute	https://software.broadinstitute.org/software/igv/
MAGeCK	Li et al., 2014	https://sourceforge.net/p/mageck/wiki/Home/
Prism	GraphPad	https://www.graphpad.com/scientific-software/prism/
rMATS	Shen et al., 2014	http://rnaseq-mats.sourceforge.net/
Seahorse Wave Desktop Software	Agilent	https://www.agilent.com/en/product/cell-analysis/real-time-cell-metabolic-analysis/xf-software/seahorse-wave-desktop-software-740897
STAR	Dobin et al., 2013	https://github.com/alexdobin/STAR
Kallisto	Bray et al., 2016	https://github.com/pachterlab/kallisto
R	The R Project	https://www.r-project.org/
DESeq2	Love et al., 2014	https://bioconductor.org/packages/release/bioc/html/DESeq2.html

Antisense Oligonucleotides (ASOs) Treatment

2 × 10⁵/mL HAP1 cells were seeded in a culture or seahorse plate. 24h later, the media was replaced by fresh media containing 10μM of the specific or control ASO, and 6μL/mL of PEG-based endoporter (GeneTools). Cells were analyzed 48h–72h later and recounter after the experiment.

Oxygen Consumption and Extracellular Acidification Rates by Seahorse XF Analyzer

1.25 × 10⁵ K562 cells were plated on a Seahorse plate in Seahorse XF DMEM media (Agilent) containing 25mM glucose and 4mM glutamine (ThermoFisher Scientific). Oxygen consumption and extracellular acidification rates were simultaneously recorded by a Seahorse XFe96 Analyzer (Agilent) using the mito stress test protocol, in which cells were sequentially perturbed by 2mM oligomycin, 1μM CCCP and 0.5mM antimycin (Sigma). Data were analyzed using the Seahorse Wave Desktop Software (Agilent). Data were not corrected for carbonic acid derived from respiratory CO₂. For seahorse in HeLa and HAP1 cells, 5 × 10⁴ and 1 × 10⁵ cells, respectively, were plated in a 96-well seahorse plate the day before the experiment. Cells were trypsinized after the experiment and recounted and the data was normalized to cell number.

Media Acidification

To visualize media acidification, 3 × 10⁶ LUC7L2^WT or LUC7L2^KO K562 cells were collected by centrifugation, washed in PBS and incubate in 1mL of DMEM media (containing 25mM glucose and phenol red, a pH indicator) for 3h before imaging.

Mitochondrial and Nuclear DNA Determination

Mitochondrial and nuclear DNA determination was carried as previously described (Bao et al., 2016). LUC7L2^WT and LUC7L2^KO K562 cells were grown for 24h in fresh cell culture media and counted. 1 × 10⁵ cells from each condition (n = 3) were harvested and lysed in 100uL mtDNA lysis buffer (25mM NaOH, 0.2mM EDTA) before incubation at 95°C for 15min. 100uL of 40mM Tris-HCl pH 7.5 was added to neutralize the reaction on ice. Samples were diluted 50x and the ratio between mitochondrial and nuclear DNA was determined using a custom Taqman based assay and qPCR using a CFx96 quantitative PCR machine (Biorad). Relative mtDNA abundance was determine using the ΔΔCt method.

RNA-Extraction, Reverse Transcription and qPCR

qPCR was performed using the TaqMan assays (ThermoFisher Scientific). RNA was extracted from total cells with an RNeasy kit (QIAGEN) and DNase-I digested before murine leukemia virus (MLV) reverse transcription using random primers (Promega) and a CFx96 quantitative PCR machine (rad). All data were normalized to TBP using ΔΔCt method.

Citrate Synthase Activity Determination

Citrate synthase activity was determined using a commercially available kit (Abcam). LUC7L2^WT and LUC7L2^KO K562 cells were grown for 48h in fresh cell culture media and counted. 5 × 10⁶ cells from each condition (n = 3) were harvested, washed in PBS and resuspended in lysis buffer (provided by the kit) and completed with protease inhibitor. 100uL of lysate was used and the experiment was performed as described in the kit protocol. Protein abundance was determined using a DC protein assay (Biorad) and the citrate synthase activity signal was normalized to the protein abundance of each sample.

Glycogen Determination

Glycogen synthase activity was determined using a commercially available kit (Abcam). LUC7L2^WT and LUC7L2^KO K562 cells were grown for 24h in fresh cell culture media and counted. 1 × 10⁶ cells from each condition (n = 8) were incubated in fresh media for another 6h before being harvested, washed in ice-cold PBS and resuspended in lysis buffer (provided by the kit). 25uL of lysate was used and the experiment was performed as described in the kit protocol and normalized to total protein abundance.

Glutamate Determination

Glutamate levels were determined using a commercially available AMPLEX kit (Life Technologies). LUC7L2^WT, LUC7L2^Rescue (LUC7L2^KO cells were a LUC7L2 cDNA was stably expressed), LUC7L2^KO and ASO treated K562 or HAP1 cells were grown for 24h in fresh culture media and counted. Cells were then washed with PBS, the media was replaced and the cells were incubated for another 3h. Medias were then collected and centrifuged at 2,000g for 3min, and glutamate concentrations were determined from the supernatant.

Cell Viability Assay in Galactose

To measure viability in galactose, cells were washed in PBS, counted and an equal number of cells was seeded in culture media containing 25mM glucose or 25mM galactose. 24h later, cells were collected and viable cells were determined using a Vi-Cell Counter (Beckman).

Electron Microscopy

5 × 10⁶ LUC7L2^WT and LUC7L2^KO K562 cells were grown for 24h in fresh cell culture media and were fixed in 2.5% gluteraldehyde, 3% paraformaldehyde with 5% sucrose in 0.1M sodium cacodylate buffer (pH 7.4), pelleted, and post fixed in 1% OsO₄ in veronal-acetate buffer. Cells were stained en block overnight with 0.5% uranyl acetate in veronal-acetate buffer (pH6.0), dehydrated and embedded in Embed-812 resin. Sections were cut on a Leica EM UC7 ultra microtome with a Diatome diamond knife at a thickness setting of 50 nm, stained with 2% uranyl acetate, and lead citrate. The sections were examined using a FEI Tecnai spirit at 80KV and photographed with an AMT CCD camera.

Confocal Microscopy and Immunofluorescence

HeLa cells were transduced with pWPI-LUC7L2-GFP at least 48h before the experiment and grown on coverslips until 30–50% confluent. Cells were successively fixed in 4% paraformaldehyde in cell culture media at room temperature for 30min, blocked/permeabilized for 30min in Abdil buffer (PBS + 0.1% Triton X-100 + 3% bovine serum albumin (w/v)), incubated with primary antibodies (1:200) in Abdil buffer for 1h, washed 3× 5min in PBS + 0.1% Triton X-100, incubated in fluorophore-coupled secondary antibodies (1:1000) and hoescht (1:10000) in Abdil for 30min, washed 3x in PBS + 0.1% Triton X-100 and mounted on a slide using FluorSave (EMD Millipore). Cells were imaged using a Zeiss LSM700 confocal microscope

Polyacrylamide Gel Electrophoresis and Immunoblotting

Cells were harvested, washed in PBS and lysed for 5min on ice in RIPA buffer (25mM Tris pH 7.5, 150mM NaCl, 0.1% SDS, 0.1% sodium deoxycholate, 1% NP40 analog, 1x protease (Cell Signaling) and 1:500 Universal Nuclease (ThermoFisher Scientific). Protein concentration was determined from total cell lysates using DC protein assay (Biorad). Gel electrophoresis was done on Novex Tris-Glycine gels (ThermoFisher Scientific) before transfer using the Trans-Blot Turbo blotting system and nitrocellulose membranes (Biorad). All immunoblotting was performed in Intercept Protein blocking buffer (Licor). Washes were done in TBS + 0.1% Tween-20 (Sigma). Specific primary antibodies were diluted 1:100–1:5000 in blocking buffer. Fluorescent-coupled secondary antibodies were diluted 1:10,000 in blocking buffer. Membranes were imagined with an Odyssey CLx analyzer (Licor) or by chemiluminescence. In a few instances, the same lysates were loaded on parallel gels to avoid antibody cross-reactivity, or because the proteins of interest could not be resolved on the same percentage gels. In these cases, a loading control is provided for all gels, and panels from the same immunoblots are connected with a dotted line on the figure. All raw immunoblots pictures are provided in the supplemental information document.

Liquid Chromatography-Mass Spectrometry (LC-MS)

Intracellular Metabolite Profiling

LUC7L2^WT and LUC7L2^KO K562 cells were pre-incubated overnight in profiling media containing glucose-free DMEM media (ThermoFisher Scientific),10% dialyzed FBS (ThermoFisher Scientific), penicillin and streptomycin (ThermoFisher Scientific) and 25mM glucose (Sigma) but omitting supplemental pyruvate or uridine. On the day of the experiment, 2.5 × 10⁶ cells were seeded in 3mL of profiling media in a 6-well plate (n = 6 replicate plates for each genotype). An additional well containing 3mL of media but no cells was included as control. After 8h of incubation, cells were centrifuged at 300g for 3min at room temperature and the culture media was saved and frozen at −80°C until further analysis (described below). The cell pellet was briefly washed in ice-cold 150mM NaCl and centrifuged again. 1mL dry ice-cold 80% methanol was then added to quench metabolism. Cells were incubated on ice for >20min, centrifuged at 20,000g (4°C) and the supernatant was saved and dried down in a speed vacuum concentrator (Savant SPD 1010, ThermoFisher Scientific) and stored at −80°C until analysis. On the day of analysis, samples were re-suspended in 120μL of 60/40 acetonitrile/water, vortexed, sonicated in ice-cold water for 1min, incubated on ice for 20min and the supernatant was collected in an autosampler vial after centrifugation at 21,000g for 20min at 4°C. Pooled quality control (PooledQC) samples were generated by combining ~20μL of each sample. Metabolite profiling was performed using a Dionex Ultimate 3000 UHPLC system coupled to a Q-Exactive Plus orbitrap mass spectrometer (ThermoFisher Scientific, Waltham, MA) with an Ion Max source and HESI II probe operating in polarity switching mode. A zwitterionic zic pHilic column (150 × 2.1mm, 5μm, Merck KGaA) was used for polar metabolite separation. Mobile phase A (MPA) was 20mM ammonium carbonate in water, pH9.6 (adjusted with ammonium hydroxide) and MPB was acetonitrile. The column was held at 27°C, with an injection volume of 5μL, and an autosampler temperature of 4°C. The LC conditions at a flow rate of 0.15 mL/min were: 0min: 80% B, 0.5min: 80% B, 20.5min: 20% B, 21.3min: 20%B, 21.5min: 80% B with 7.5min of column equilibration time. MS parameters were: sheath gas flow = 30, aux gas flow = 7, sweep gas flow = 2, spray voltage = 2.80 for negative & 3.80 for positive ion modes, capillary temperature = 310°C, S-lens RF level = 50 and aux gas heater temp 370°C. Data acquisition was performed using Xcalibur 4.1 (ThermoFisher Scientific) in full scan mode with a range of 70–1000m/z, a resolving power of 70,000, an AGC target of 1 × 10⁶, and a maximum injection time of 80ms. Data analysis was done using Compound Discoverer 3.0. Samples were injected in a randomized order and pooled QC samples were injected regularly throughout the analytical batches. Metabolite annotation was based on accurate mass (±5ppm) and matching retention time (±0.3min) as well as MS/MS fragmentation pattern from the pooled QC samples against in-house retention time +MSMS library of reference chemical standards. Metabolites which had a pooled QC CV<20% were used for the statistical analysis.

Media Profiling

30μL of control or spent media was mixed with 120μL of ice-cold acetonitrile containing the metabolomics amino acid mix from Cambridge Isotope Labs (MSK-A2–1.2), ¹³C₆ -glucose, ¹³C₃ -pyruvate, and ¹³C₃ -lactate as internal standards, was vortexed, incubated on ice for 20min, centrifuged at 21,000g for 20min at 4°C and the supernatant was transferred to an autosampler vial for LC-MS analysis. Calibration curves were prepared in water at varying concentration levels depending on the amino acid level in the DMEM media formulation. Metabolite separation was done using XBridge BEH amide (2.1 × 150mm, 1.7 μm, Waters Corporation, MA). Mobile phase A was 90/5/5 water/acetonitrile/methanol, 20mM ammonium acetate, 0.2% acetic acid and mobile phase B was 90/10 acetonitrile/water, 10mM ammonium acetate, 0.2% acetic acid. The column temperature was 40°C and flow rate was 0.3 mL/min. The chromatographic gradient was: 0min: 95% B, 9min : 70% B, 9.75min: 40% B, 12min: 40% B, 13min: 30% B, 14min : 30%B, 14.1min: 10% B,17min: 10% B, 17.5min: 95% B, 22min : 95% B. MS parameters were: sheath gas flow = 50, aux gas flow = 12, sweep gas flow = 2, spray voltage = 2.80 for negative (3.50 for positive), Capillary temperature = 320°C, S-lens RF level = 50 and aux gas heater temperature 380°C. Data acquisition was done using Xcalibur 4.1 (ThermoFisher Scientific) and performed in full scan mode with a range of 70–1000m/z, a resolving power of 70,000, an AGC target 10⁶, and a maximum injection time of 100ms. Tracefinder 4.1 was used for quantitation analysis. One LUC7L2^KO sample gave aberrant spectra and was excluded.

LUC7L2–3xFLAG Immunoprecipitation and Mass Spectrometry

For immunoprecipitation, a 3xFLAG-tagged version of LUC7L2 was cloned into pWPI-Neo (Addgene), and viruses were produced. pWPI-GFP served as control. 293T cells were infected and expanded for at least 48h. An equal number of cells from each condition in duplicate was lysed in IP lysis buffer (50 mM Tris/HCl (pH 7.5), 150 mM NaCl, 1 mM MgCl₂, 1% NP-40, 0.1% sodium deoxycholate, 1× protease (Cell Signaling). Lysates were cleared by centrifugation at 20,000g for 20min and the supernatants were saved. Washed FLAG M2 magnetic beads (Sigma) were added to the lysate and incubated overnight at 4°C. Beads were recovered after extensive washing, and the protein/RNA complexes were eluted with 100 μg/mL 3xFLAG peptide (Sigma). For protein isolation, the eluate was run on an SDS-PAGE gel until the whole lysate entered the gel. Single bands containing all proteins from the sample were then cut and analyzed by mass spectrometry at the Whitehead proteomics facility. Peptides were identified and quantified using the Top 3 total ion current (TIC) method (Scaffold4). Interacting proteins were considered positive when they were enriched >2-fold over either control and identified only by unique peptides.

Enhanced Crosslinking and Immunoprecipitation (eCLIP)

Libraries were generated using standard eCLIP methods according to published protocols (Van Nostrand et al., 2017). In brief, K562 and Hela cells (2 × 10⁷ for each replicate plate) were UV crosslinked (254 nm, 400 mJ/cm2), then lysed and sonicated (Bioruptor) in eCLIP lysis buffer (50 mM Tris–HCl pH 7.4, 100 mM NaCl, 1% NP-40 (Igepal CA630), 0.1% SDS, 0.5% sodium deoxycholate, 1:200 Protease Inhibitor Cocktail I, in RNase/DNase-free H₂O). RNA fragments were created by incubating lysates with RNase I (Ambion) and LUC7L2:RNA complexes were immunoprecipitated for 2h at 4°C using Dynabeads bound to 4μg of LUC7L2-specific affinity-purified antibody. In parallel, libraries were generated from size-matched input (SMInput) samples containing RNAs present in the whole cell lysates, i.e. sans RBP-specific IP. For the IPs, a series of stringent washes (high salt wash buffer: 50 mM Tris–HCl pH 7.4, 1 M NaCl, 1 mM EDTA, 1% NP-40, 0.1% SDS, 0.5% sodium deoxycholate, in RNase/DNase-free H₂O; wash buffer: 20 mM Tris–HCl pH 7.4, 10 mM MgCl₂, 0.2% Tween-20, in RNase/DNase-free H₂O) was followed by RNA dephosphorylation with FastAP (ThermoFisher Scientific) and T4 PNK (NEB) then ligation of an adaptor to the 3′ ends of the RNAs with T4 RNA ligase 1 (NEB). Protein:RNA complexes were separated on 4–12% polyacrylamide gels, transferred to a nitrocellulose membranes and RNA was extracted from the membranes using Proteinase K (NEB). Immunoprecipitation was confirmed by parallel western blotting of fractions of each sample with the antibody described previously. Following purification, SMInput RNA were dephosphorylated and 3′-ligated and all samples were reverse transcribed with Superscript III (Invitrogen). Free primers were removed with ExoSap-IT (Affymetrix) and a DNA adaptor was ligated to the 3′ ends of the cDNA with T4 RNA ligase 1. cDNA was quantified by qPCR and PCR amplified using Q5 Master Mix (NEB) and resulting libraries were purified prior to Illumina sequencing.

Blue-Native PAGE

For blue-native PAGE, a mitochondria-rich fraction (Jourdain et al., 2013) was isolated from LUC7L2^WT and LUC7L2^KO K562 and HeLa cells grown for two weeks in glucose or galactose-containing by differential centrifugation. An equal amount of mitochondria were resuspended in blue-native loading buffer containing 1% digitonin (Life Technologies) before electrophoresis on a 3 to 12% Native PAGE (Life Technologies) according to the manufacturer’s instruction. Gels were then fixed and stained with coomassie R-250, or transferred to PVDF membranes, denatured by 3% acetic acid treatment, destained with methanol, blocked and an immunodetection was performed with the indicated antibodies and secondary HRP-coupled antibodies. Samples were loaded on parallel gels to avoid cross-reactivity between antibodies.

Mitochondrial Translation

Determination of mitochondrial translation products in LUC7L2^WT and LUC7L2^KO cells was performed as previously described (Jourdain et al., 2013). K562s cells were incubated for 20 min in methionine/cysteine-free DMEM (Sigma) complemented with dialyzed serum and 2mM L-glutamine. Cells were then incubated for 1h in the same medium in the presence of 100μg/ml emetine and 100μCi/μL ³⁵S-labeled methionine/cysteine (PerkinElmer). Total protein concentration of cell lysates was measured, and lysates were resolved on an acrylamide gel, transferred to a nitrocellulose membrane, and analyzed by autoradiography. A mitochondrial translation inhibitor (chloramphenicol) was used as a control.

Gene-Specific cDNA Cloning and Expression

cDNAs of interest were custom designed (Genewiz or IDT) and cloned into pWPI-Neo or pLV-lenti-puro (Visanji et al., 2011, Hayer et al., 2016) using BamHI/SpeI and BamHI/NotI (NEB), respectively. cDNA sequences were:

GFP-3xFLAG

atggtgagcaagggcgaggagctgttcaccggggtggtgcccatcctggtcgagctggacggcgacgtaaacggccacaagttcagcgtgtccggcgagggcgagggcgatgccacctacggcaagctgaccctgaagttcatctgcaccaccggcaagctgcccgtgccctggcccaccctcgtgaccaccctgacctacggcgtgcagtgcttcagccgctaccccgaccacatgaagcagcacgacttcttcaagtccgccatgcccgaaggctacgtccaggagcgcaccatcttcttcaaggacgacggcaactacaagacccgcgccgaggtgaagttcgagggcgacaccctggtgaaccgcatcgagctgaagggcatcgacttcaaggaggacggcaacatcctggggcacaagctggagtacaactacaacagccacaacgtctatatcatggccgacaagcagaagaacggcatcaaggtgaacttcaagatccgccacaacatcgaggacggcagcgtgcagctcgccgaccactaccagcagaacacccccatcggcgacggccccgtgctgctgcccgacaaccactacctgagcacccagtccgccctgagcaaagaccccaacgagaagcgcgatcacatggtcctgctggagttcgtgaccgccgccgggatcactctcggcatggacgagctgtacaaggattataaagatcatgatggcgattataaagatcatgatattgattataaagatgatgatgataaataatagtgagcggccgc

LUC7L-3xFLAG

atgtccgcccaggcgcagatgcgggccctgctggaccagctcatgggcacggctcgggacggagacgaaaccagacagagggtcaagtttacagatgaccgtgtctgcaagagtcaccttctggactgctgcccccatgacatcctggctgggacgcgcatggatttaggagaatgtaccaaaatccacgacttggccctccgagcagattatgagattgcaagtaaagaaagagacctgttttttgaattagatgcaatggatcacttggagtcctttattgctgaatgtgatcggagaactgagctcgccaagaagcggctggcagaaacacaggaggaaatcagtgcggaagtttctgcaaaggcagaaaaagtacatgagttaaatgaagaaataggaaaactccttgctaaagccgaacagctaggggctgaaggtaatgtggatgaatcccagaagattcttatggaagtggaaaaagttcgtgcgaagaaaaaagaagctgaggaagaatacagaaattccatgcctgcatccagttttcagcagcaaaagctgcgtgtctgcgaggtctgttcagcctaccttggtctccatgacaatgaccgtcgcctggcagaccacttcggtggcaagttacacttggggttcattcagatccgagagaagcttgatcagttgaggaaaactgtcgctgaaaagcaggagaagagaaatcaggatcgcttgaggaggagagaggagagggaacgggaggagcgtctgagcaggaggtcgggatcaagaaccagagatcgcaggaggtcacgctcccgggatcggcgtcggaggcggtcaagatctacctcccgagagcgacggaaattgtcccggtcccggtcccgagatagacatcggcgccaccgcagccgttcccggagccacagccggggacatcgtcgggcttcccgggaccgaagtgcgaaatacaagttctccagagagcgggcatccagagaggagtcctgggagagcgggcggagcgagcgagggcccccggactggaggcttgagagctccaacgggaagatggcttcacggaggtcagaagagaaggaggccggcgagatcgattataaagatcatgatggcgattataaagatcatgatattgattataaagatgatgatgataaataatagtgagcggccgc

LUC7L2–3xFLAG

atgtcggcgcaggcccagatgcgcgcgatgctggaccagttgatgggcacctcccgggacggagatacaactcgtcaacgaatcaaattcagtgatgacagagtatgcaagagtcaccttctcaactgttgtcctcatgatgtcctttctggaactagaatggatcttggagaatgtctgaaagtccatgacctggctttaagagcggattatgaaattgcatccaaagaacaagattttttctttgaacttgatgccatggatcatctgcagtcattcattgcagattgtgatcgtagaacagaagtggccaagaaaagattagcagaaactcaagaagagattagtgctgaagtagcagcaaaggcagaacgtgttcatgagttaaatgaagaaattggtaaattgttagccaaggtggaacaactaggagctgaagggaatgtggaggaatcccagaaagtaatggatgaagtagagaaagcacgggcaaagaaaagagaagcagaggaagtttatcggaattctatgccagcttccagttttcagcagcagaaacttcgagtctgtgaagtctgctctgcctatttaggacttcatgataatgacagacgactggctgatcattttgggggtaaactgcacctgggatttattgaaataagagagaagcttgaagaattaaagagagtcgtagctgagaagcaggagaaaagaaaccaggaacggctgaaacgaagagaagagagagagagagaagaaagggagaagctgaggaggtcccgatcacacagcaagaatccaaaaagatccaggtccagagagcatcgcagacatcgatctcgctccatgtcacgtgaacgcaagaggagaactcgatccaaatctcgggagaaacgccatcgccacaggtcccgctccagcagccgtagccgcagccgtagccaccagagaagtcggcacagttctagagataggagcagagaacgatccaagaggagatcctcaaaagaaagattcagagaccaagacttagcatcatgtgacagagacaggagttcaagagacagatcacctcgtgacagagatcggaaagataagaagcggtcctatgagagtgctaatggcagatcagaagacaggaggagctctgaagagcgcgaagcaggggagatcggagggggtgattataaagatcatgatggcgattataaagatcatgatattgattataaagatgatgatgataaataatagtga

LUC7L3–3xFLAG

atgatttcggccgcgcagttgttggatgagttaatgggccgggaccgaaacctagccccggacgagaagcgcagcaacgtgcggtgggaccacgagagcgtttgtaaatattatctctgtggtttttgtcctgcggaattgttcacaaatacacgttctgatcttggtccgtgtgaaaaaattcatgatgaaaatctacgaaaacagtatgagaagagctctcgtttcatgaaagttggctatgagagagattttttgcgatacttacagagcttacttgcagaagtagaacgtaggatcagacgaggccatgctcgtttggcattatctcaaaaccagcagtcttctggggccgctggcccaacaggcaaaaatgaagaaaaaattcaggttctaacagacaaaattgatgtacttctgcaacagattgaagaattagggtctgaaggaaaagtagaagaagcccaggggatgatgaaattagttgagcaattaaaagaagagagagaactgctaaggtccacaacgtcgacaattgaaagctttgctgcacaagaaaaacaaatggaagtttgtgaagtatgtggagcctttttaatagtaggagatgcccagtcccgggtagatgaccatttgatgggaaaacaacacatgggctatgccaaaattaaagctactgtagaagaattaaaagaaaagttaaggaaaagaaccgaagaacctgatcgtgatgagcgtctaaaaaaggagaagcaagaaagagaagaaagagaaaaagaacgggagagagaaagggaagaaagagaaaggaaaagacgaagggaagaggaagaaagagaaaaagaaagggctcgtgacagagaaagaagaaagagaagtcgttcacgaagtagacactcaagccgaacatcagacagaagatgcagcaggtctcgggaccacaaaaggtcacgaagtagagaaagaaggcggagcagaagtagagatcgacgaagaagcagaagccatgatcgatcagaaagaaaacacagatctcgaagtcgggatcgaagaagatcaaaaagccgggatcgaaagtcatataagcacaggagcaaaagtcgggacagagaacaagatagaaaatccaaggagaaagaaaagaggggatctgatgataaaaaaagtagtgtgaagtccggtagtcgagaaaagcagagtgaagacacaaacactgaatcgaaggaaagtgatactaagaatgaggtcaatgggaccagtgaagacattaaatctgaaggtgacactcagtccaatgattataaagatcatgatggcgattataaagatcatgatattgattataaagatgatgatgataaataatagtgagcggccgc

SLC7A11-HA

atggtcagaaagcctgttgtgtccaccatctccaaaggaggttacctgcagggaaatgttaacgggaggctgccttccctgggcaacaaggagccacctgggcaggagaaagtgcagctgaagaggaaagtcactttactgaggggagtctccattatcattggcaccatcattggagcaggaatcttcatctctcctaagggcgtgctccagaacacgggcagcgtgggcatgtctctgaccatctggacggtgtgtggggtcctgtcactatttggagctttgtcttatgctgaattgggaacaactataaagaaatctggaggtcattacacatatattttggaagtctttggtccattaccagcttttgtacgagtctgggtggaactcctcataatacgccctgcagctactgctgtgatatccctggcatttggacgctacattctggaaccattttttattcaatgtgaaatccctgaacttgcgatcaagctcattacagctgtgggcataactgtagtgatggtcctaaatagcatgagtgtcagctggagcgcccgaatccagattttcttaaccttttgcaagctcacagcaattctgataattatagtccctggagttatgcagctaattaaaggtcaaacgcagaactttaaagacgccttttcaggaagagattcaagtattacgcggttgccactggctttttattatggaatgtatgcatatgctggctggttttacctcaactttgttactgaagaagtagaaaaccctgaaaaaaccattccccttgcaatatgtatatccatggccattgtcaccattggctatgtgctgacaaatgtggcctactttacgaccattaatgctgaggagctgctgctttcaaatgcagtggcagtgaccttttctgagcggctactgggaaatttctcattagcagttccgatctttgttgccctctcctgctttggctccatgaacggtggtgtgtttgctgtctccaggttattctatgttgcgtctcgagagggtcaccttccagaaatcctctccatgattcatgtccgcaagcacactcctctaccagctgttattgttttgcaccctttgacaatgataatgctcttctctggagacctcgacagtcttttgaatttcctcagttttgccaggtggctttttattgggctggcagttgctgggctgatttatcttcgatacaaatgcccagatatgcatcgtcctttcaaggtgccactgttcatcccagctttgttttccttcacatgcctcttcatggttgccctttccctctattcggacccatttagtacagggattggcttcgtcatcactctgactggagtccctgcgtattatctctttattatatgggacaagaaacccaggtggtttagaataatgtcagagaaaataaccagaacattacaaataatactggaagttgtaccagaagaagataagttatatccatatgatgttccagattatgcttaatagtga

PFKM

atgacccatgaagagcaccatgcagccaaaaccctggggattggcaaagccattgctgtcttaacctctggtggagatgcccaaggtatgaatgctgctgtcagggctgtggttcgagttggtatcttcaccggtgcccgtgtcttctttgtccatgagggttatcaaggcctggtggatggtggagatcacatcaaggaagccacctgggagagcgtttcgatgatgcttcagctgggaggcacggtgattggaagtgcccggtgcaaggactttcgggaacgagaaggacgactccgagctgcctacaacctggtgaagcgtgggatcaccaatctctgtgtcattgggggtgatggcagcctcactggggctgacaccttccgttctgagtggagtgacttgttgagtgacctccagaaagcaggtaagatcacagatgaggaggctacgaagtccagctacctgaacattgtgggcctggttgggtcaattgacaatgacttctgtggcaccgatatgaccattggcactgactctgccctgcatcggatcatggaaattgtagatgccatcactaccactgcccagagccaccagaggacatttgtgttagaagtaatgggccgccactgtggatacctggcccttgtcacctctctgtcctgtggggccgactgggtttttattcctgaatgtccaccagatgacgactgggaggaacacctttgtcgccgactcagcgagacaaggacccgtggttctcgtctcaacatcatcattgtggctgagggtgcaattgacaagaatggaaaaccaatcacctcagaagacatcaagaatctggtggttaagcgtctgggatatgacacccgggttactgtcttggggcatgtgcagaggggtgggacgccatcagcctttgacagaattctgggcagcaggatgggtgtggaagcagtgatggcacttttggaggggaccccagataccccagcctgtgtagtgagcctctctggtaaccaggctgtgcgcctgcccctcatggaatgtgtccaggtgaccaaagatgtgaccaaggccatggatgagaagaaatttgacgaagccctgaagctgagaggccggagcttcatgaacaactgggaggtgtacaagcttctagctcatgtcagacccccggtatctaagagtggttcgcacacagtggctgtgatgaacgtgggggctccggctgcaggcatgaatgctgctgttcgctccactgtgaggattggccttatccagggcaaccgagtgctcgttgtccatgatggtttcgagggcctggccaaggggcagatagaggaagctggctggagctatgttgggggctggactggccaaggtggctctaaacttgggactaaaaggactctacccaagaagagctttgaacagatcagtgccaatataactaagtttaacattcagggccttgtcatcattgggggctttgaggcttacacagggggcctggaactgatggagggcaggaagcagtttgatgagctctgcatcccatttgtggtcattcctgctacagtctccaacaatgtccctggctcagacttcagcgttggggctgacacagcactcaatactatctgcacaacctgtgaccgcatcaagcagtcagcagctggcaccaagcgtcgggtgtttatcattgagactatgggtggctactgtggctacctggctaccatggctggactggcagctggggccgatgctgcctacatttttgaggagcccttcaccattcgagacctgcaggcaaatgttgaacatctggtgcaaaagatgaaaacaactgtgaaaaggggcttggtgttaaggaatgaaaagtgcaatgagaactataccactgacttcattttcaacctgtactctgaggaggggaagggcatcttcgacagcaggaagaatgtgcttggtcacatgcagcagggtgggagcccaaccccatttgataggaattttgccactaagatgggcgccaaggctatgaactggatgtctgggaaaatcaaagagagttaccgtaatgggcggatctttgccaatactccagattcgggctgtgttctggggatgcgtaagagggctctggtcttccaaccagtggctgagctgaaggaccagacagattttgagcatcgaatccccaaggaacagtggtggctgaaactgaggcccatcctcaaaatcctagccaagtacgagattgacttggacacttcagaccatgcccacctggagcacatcacccggaagcggtccggggaagcggccgtctaatagtga

Quantitative Proteomics

LUC7L2^WT and LUC7L2^KO cells were grown in n = 3 replicate plates in DMEM media containing 25mM glucose or galactose for 2 weeks. Quantitative proteomics was performed at the Thermofisher for Multiplexed Proteomics (Harvard). In short, cells were harvested and total protein quantification was performed using micro-BCA assay (Pierce). Samples were reduced with DTT and alkylated with iodoacetamide before protein precipitation in methanol/chloroform. Pellets were resuspended in 200 mM EPPS, pH 8.0 and a digestion was performed sequentially using LysC (1:50) and Trypsin (1:100) based on protease to protein ratio. ~50μg peptide per sample was labeled with TMTpro 16 reagents. A small aliquot of each sample was then combined and analyzed by LC-MS3 to verify labeling efficiency and mixing ratios. Samples were combined, desalted, and dried by speedvac. 14 fractions from the total proteome HPRP set were analyzed on an Orbitrap Eclipse mass spectrometer using a 180-minute method MS3 method with real-time search. Peptides were detected (MS1) and quantified (MS3) in the Orbitrap. Peptides were sequenced (MS2) in the ion trap. MS2 spectra were searched using the COMET algorithm against a custom protein database containing only one protein per gene (referenced as the canonical isoform). Peptide spectral matches were filtered to a 1% false discovery rate (FDR) using the target-decoy strategy combined with linear discriminant analysis. The proteins from the 14 runs were filtered to a <1% FDR. Proteins were quantified only from peptides with a summed SN threshold of >100. Only unique peptides were considered for downstream analysis. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (Perez-Riverol et al., 2019) partner repository with the dataset identifier PXD021917 and 10.6019/PXD021917″.

Next-Generation RNA Sequencing

Total RNA from LUC7L2^WT and LUC7L2^KO K562 and HeLa cells (n = 3, replicate plates each) were isolated using a RNeasy kit (QIAGEN). RNA sequencing libraries were prepared by the Genomics Platform at the Broad Institute based on the True-Seq protocol (Illumina), which selects for polyadenylated RNA and preserves strand specificity. Libraries were sequenced using a NovaSeq 6000 instrument, generating 2 × 101bp paired-end reads with a minimum of 40M pairs per sample. The RNA sequencing data have been deposited to the Gene Expression Omnibus (GEO) with the dataset identifier GSE157917.

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical tests and sample sizes are reported in the legends associated with each figure. Methods for quantification and statistical analysis of large-scale datasets are described below.

eCLIP

Raw reads were mapped using STAR (Dobin et al., 2013) to the hg19 genome following standard ENCODE guidelines (https://github.com/alexdobin/STAR/blob/master/doc/STARmanual.pdf, page 9). Read length was relaxed to accommodate the slightly shorter average eCLIP read length (--outFilterMatchNminOverLread 0.33). Duplicate PCR reads were removed from the mapped reads to generate final reads. Mapped reads were then processed into peaks using CLIPper (Lovci et al., 2013) with standard specifications. For each dataset, only peaks shared between two replicates and not appearing in the input controls were considered in subsequent analyses. Reads were then filtered to a robust p-value of P<10⁻⁴ or as indicated in the manuscript. Using these highest-confidence peaks, a metaplot centered on the exon was created, showing an additional 500 bases upstream and 500 bases downstream of the flanking introns. For each region (exonic or intronic), the relative positions of each single base of a CLIP peak were summed and normalized to the mean base coverage in that region.

Gene Expression

The reads were aligned with STAR (Dobin et al., 2013) to the human genome hg19 using default parameters and a two-pass approach. Following a first-pass alignment of each sample, novel splice junctions were pooled across all samples from the same cell type and incorporated into the genome annotation for a second-pass alignment. Second-pass gene counts derived from uniquely mapping pairs with the expected strandedness were output by STAR. Differential gene expression analysis between LUC7L2^WT and LUC7L2^KO samples was performed in R using the package DESeq2 (Love et al., 2014). Differentially-expressed genes were considered significant below FDR<10⁻⁴ and with an absolute fold change value greater than 50%.

Splicing

Splicing events were analyzed with rMATS.4.0.2 (three replicates per condition per cell line) using standard specifications and the hg19 genome. Events were considered significant below FDR<0.1 and with an absolute Δψ value greater than 0.05. Only events in genes with total TPM>1 in the WT cell line of interest as called by kallisto (Bray et al., 2016) according to standard parameters were considered.

Overlap

Overlap analyses were normalized for gene expression as in (Friedman et al., 2008). Briefly, ten bins (percentiles) of gene expression were established for HeLa and K562 cells using the average gene expression from three replicates determined by kallisto (Bray et al., 2016) according to standard parameters. Averaged TPMs were filtered for TPM>1 as above. For each bin, the percentage of genes within that bin appearing in corresponding eCLIP or rMATS was calculated. A 10 × 10 matrix was then created by multiplying these frequencies together for each cell. A second 10 × 10 matrix was then populated by the actual counts of total overlapping genes in each cell (of 100). A third 10 × 10 matrix of expected counts of eCLIP/rMATS was subsequently generated from these two matrices and the total sum of this matrix represented the “expected” value of the overlap based on gene expression alone. Finally, the significance of the observed count of overlapping genes per query was estimated by Poisson (R 1.3.1093, poisson.test).

Correlation

Published protein expression datasets (Geiger et al., 2013, Huttlin et al., 2010, Ping et al., 2018, Ubaida-Mohien et al., 2019) were filtered to exclude genes with null values in >20% of samples and remaining missing values were replaced with half the minimum observed for that protein. Each protein was normalized via Z-score and then Pearson correlations were calculated between LUC7L2 profiles and each protein. Anti-correlation with OXPHOS was assessed via a Wilcoxon Rank Sum test.

Gene Ontology

Gene ontology analysis was performed using GOrilla with default settings and using a ranked gene list as input (Eden et al., 2009). Only GO categories with <500 genes and represented by >2 significant genes were considered. Highlighted genes in figures correspond to the GO categories “RNA splicing” (GO:0008380), “U1 snRNP” (GO:0005685) and “OXPHOS” (manually curated) and GO gene lists are reported in Table S1.

Supplementary Material

table S4

Table S4: LUC7L2 Immunoprecipitation and Mass Spectrometry (IP-MS), Related to Figure 4 and S4.

table S6

Table S6: Differential Gene Expression Analysis in LUC7L2^KO K562 Cells, Related to Figure 4 and S4.

table S5

Table S5: LUC7L2 Enhanced Crosslinking and Immunoprecipitation (eCLIP), Related to Figure 4 and S4.

table S8

Table S8: Global Quantitative Proteomics in LUC7L2^KO And Galactose-Grown K562 Cells Relative to Glucose-Grown LUC7L2^WT Cells, Related to Figure 6 and S6.

table S7

Table S7: Differential Splicing Analysis in LUC7L2^KO K562 Cells, Related to Figure 4 and S4.

table S1

Table S1: Genome-Wide CRISPR/Cas9 Screening Results and Gene Ontology, Related to Figure 1 and S1.

table S2

Table S2: Media Metabolite Profiling in LUC7L2^KO K562 Cells, Related to Figure 3 and S3.

table S3

Table S3: Cellular Metabolite Profiling in LUC7L2^KO K562 Cells, Related to Figure 3 and S3.

HIGHLIGHTS.

• Expression of LUC7L2 and the U1 snRNP represses OXPHOS

• Pre-mRNA splicing and expression of PFKM and SLC7A11 (xCT) requires LUC7L2

• Loss of LUC7L2 and glycolysis promotes respiratory chain (super)complex assembly

• Cross-regulation and role in energy metabolism for the LUC7 family