(C) PLOS One
This story was originally published by PLOS One and is unaltered.
. . . . . . . . . .



Primary and metastatic tumors exhibit systems-level differences in dependence on mitochondrial respiratory function [1]

['Neal K. Bennett', 'Gladstone Institute Of Neurological Disease', 'Gladstone Institutes', 'San Francisco', 'California', 'United States Of America', 'Hiroki J. Nakaoka', 'Department Of Radiation Oncology', 'University Of California', 'Danny Laurent']

Date: 2022-10

The Warburg effect, aerobic glycolysis, is a hallmark feature of cancer cells grown in culture. However, the relative roles of glycolysis and respiratory metabolism in supporting in vivo tumor growth and processes such as tumor dissemination and metastatic growth remain poorly understood, particularly on a systems level. Using a CRISPRi mini-library enriched for mitochondrial ribosomal protein and respiratory chain genes in multiple human lung cancer cell lines, we analyzed in vivo metabolic requirements in xenograft tumors grown in distinct anatomic contexts. While knockdown of mitochondrial ribosomal protein and respiratory chain genes (mito-respiratory genes) has little impact on growth in vitro, tumor cells depend heavily on these genes when grown in vivo as either flank or primary orthotopic lung tumor xenografts. In contrast, respiratory function is comparatively dispensable for metastatic tumor growth. RNA-Seq and metabolomics analysis of tumor cells expressing individual sgRNAs against mito-respiratory genes indicate overexpression of glycolytic genes and increased sensitivity of glycolytic inhibition compared to control when grown in vitro, but when grown in vivo as primary tumors these cells down-regulate glycolytic mechanisms. These studies demonstrate that discrete perturbations of mitochondrial respiratory chain function impact in vivo tumor growth in a context-specific manner with differential impacts on primary and metastatic tumors.

Funding: KN and NKB were supported by the Joan and David Traitel Family Trust to KN. The NIH award R01AG065428 funded KN and JLN. JLN and HN received support from the NIH/NCI award U54CA196519 and the University of California Cancer Coordinating Committee Award to JLN. HN was supported by a Japan Society for the Promotion of Science Fellowship ( https://www.jsps.go.jp/english/ ). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

To our knowledge, a systems-level analysis of genetic modulators of ATP combined with in vivo functional readouts has not been performed in human cancers. By evaluating tumor growth in multiple in vivo contexts, our findings illustrate critical differences in the mitochondrial respiratory chain requirements between primary and metastatic tumor growth, pointing to a functional role for respiratory chain function that is specific to regional growth and distinguishes primary tumors from distant metastases.

Whether and how respiration and glycolysis-derived ATP production differentially impact in vivo growth, especially when multiple anatomically separate tumor sites have developed, is not known on a systems level. To evaluate this, we developed a custom CRISPRi mini-library composed of sgRNAs against genes modulating cellular ATP levels (identified from the ATPome [ 6 ]), to test whether ATP levels correlate with in vivo growth of tumors in multiple preclinical mouse models. In vivo xenograft models revealed differential growth effects of ATP-modulating genes within discrete anatomic sites, where suppression of genes involved in mitochondrial-derived ATP was associated with reduced growth of primary tumors but not their associated metastases. To evaluate the molecular basis for this reduced growth, we performed RNA Seq and metabolomics analysis in isogenic lung cancer cells engineered with discrete silencing of mito-respiratory genes; these studies indicated profound adaptive metabolism, particularly involving the glycolytic metabolic profile, which was highly dependent on the tumor growth context. Targeting genes that modulate ATP level in parallel models identified discrete systems-level requirements for mitochondrial and respiratory chain function in primary versus metastatic tumors, illustrating asymmetric dependence on glycolysis or respiratory metabolism, indicative of metabolic heterogeneity.

In prior work, we developed a high-throughput screening paradigm to identify genetic regulators of ATP, combining FACS and CRISPR with an ATP-FRET sensor capable of monitoring real-time changes in ATP concentrations within individual living cells [ 5 ]. Using this sensor, we performed genome-wide CRISPRi and CRISPRa screens to define an “ATPome” of genes and pathways that regulate ATP levels through energy substrate-specific pathways (respiration or glycolysis) [ 6 ]. A key finding from this work is that many genes and pathways that preserve or reduce ATP exert these effects only under specific metabolic conditions defined by substrate availability. Glycolysis and respiration demonstrate cross-optimization on a systems level, that is, suppression of 1 metabolic mechanism (via members of specific gene classes) optimizes the alternative mechanism. Silencing genes required for respiratory-derived ATP modestly suppressed in vitro growth under respiratory conditions but conversely increased in vitro tumor cell growth under glycolytic conditions. This evidence of cross-optimization points adaptive metabolism that is measurable in cellular ATP; it also indicates a broader repertoire of mechanisms available to cells for optimizing their metabolic function.

The dysregulation of cellular energy metabolism is an early fundamental event in tumorigenesis and a hallmark of cancer. Cancer cells modulate their metabolism as they proliferate, outpace normal cells in growth, and establish disease in diverse and often nutrient-restricted environments. The Warburg effect observed in cancer cells refers to the preferential use of aerobic glycolysis, which produces less ATP than aerobic respiration while favoring biosynthetic functions necessary for tumor growth [ 1 ]. The specific roles of ATP-modulating mechanisms in supporting tumor growth are poorly understood, particularly in vivo. Most studies investigating respiration and the Warburg effect are performed in cultured cells. While these have established that mitochondria are necessary for tumorigenesis [ 2 ] and have led to cancer therapies that target oxidative phosphorylation [ 3 ], precisely how mitochondria participate in cancer metabolic programs is still poorly understood, especially as it pertains to in vivo tumor growth in different anatomic and microenvironmental contexts. In contrast to in vitro tumor modeling, tumor growth in vivo tests specific physiologic contexts in which cancer cells must metabolically adapt to thrive. Metabolomics analyses of lung cancers growing in vivo indicate increased glucose-derived carbon-labeling of TCA intermediates compared to normal lung [ 4 ], supporting the hypothesis that tumor metabolism and fuel utilization are distinct from those of normal tissues. However, how tumors and metastases grow in diverse anatomic locations, and the identity of the metabolic programs underpinning this capacity, are poorly understood.

Results

Primary in vivo tumor growth requires mitochondrial-derived ATP Lung cancer is the most common cause of death due to cancer in the United States and is characterized by primary solid tumors that can metastasize widely to diverse organs. HCC827 and H1975 human lung cancer cells are model cell lines for epidermal growth factor receptor (EGFR)-mutant nonsmall cell lung cancer [7,8]. To determine whether ATP-modulating genes impact the in vivo tumor growth of lung cancers, we transduced HCC827 cells expressing dCas9-KRAB with a custom CRISPRi mini-library enriched with respiratory and glycolytic hits that most significantly influenced ATP levels (high or low, depending on the substrate conditions) in a previous screen in K562 and HCC827 cells [6]. This mini-library contains over 400 sgRNAs (1/5 of which are nontargeting sgRNAs included as negative controls), with multiple sgRNAs against each gene target (typically 2 to 4 unique sgRNAs/gene) [6]. We postulated that similar to in vitro growth, individual gene silencing in vivo could confer relative growth advantages or disadvantages that could be estimated on the basis of sgRNA representation. After antibiotic selection, HCC827 cells were injected into the flanks of nude mice. Injected mice developed subcutaneous flank tumors that were allowed to grow for 28 days, then were analyzed by targeted sequencing as described previously [6]. We compared the normalized sgRNA representation among control nontargeting sgRNAs, sgRNAs targeting glycolytic and glycolysis-promoting genes, and sgRNAs targeting mito-ribosomal and respiratory chain (termed mito-respiratory) genes (Fig 1A). As a group, sgRNAs targeting mito-respiratory genes were significantly depleted compared to nontargeting control sgRNAs and sgRNAs targeting glycolytic genes. In fact, among all sgRNAs in the library, some individual sgRNAs targeting mito-respiratory genes, namely MALSU1, HSD17β10, c14orf2, andTMEM261, were the most significantly depleted individual sgRNAs (Fig 1A). Each of these genes is associated with mitochondrial function, and they were previously found to be critical in maintaining mitochondrial-derived ATP levels, although none have known roles in tumorigenesis. MALSU1 (mitochondrial assembly of ribosomal large subunit 1) encodes a mitochondrial protein that is thought to be involved in mitochondrial translation [9]. HSD17β10 encodes Hydroxysteroid 17-Beta dehydrogenase 10, which localizes to the mitochondria and is involved in protein synthesis [10]. TMEM261 (also known as DMAC1) encodes transmembrane protein 261, an electron transport chain component [11]. c14orf2 (chromosome 14 open reading frame 2, gene ATP5MPL) encodes ATP synthase membrane subunit j [12]. In a prior CRISPRi screen in HCC827 cells [6], we found that suppression of c14orf2, HSD17β10, TMEM261, or MALSU1 was associated with low ATP levels in cells grown under respiratory conditions, and with high ATP levels in cells grown in glycolytic conditions (documented in Supplementary Data 4 of [6]), indicating similar ATP effects among all these genes in HCC827 cells. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 1. Human lung cancer cells demonstrate differential requirements for mitochondrial and respiratory genes when grown in vivo in flank or orthotopic lung models. HCC827 and H1975 human EGFR-mutant lung cancer cells were transduced with the mini-CRISPRi library and then injected into either the subcutaneous space in flanks (HCC827) or the left lower lung lobes (H1975) of nude mice and grown for 28 days. DNA from each tumor was sequenced and read counts for each sgRNA quantified. The read count for each sgRNA was normalized to the sum of reads for negative control sgRNAs and the ratio of each sgRNA’s frequency in the tumor model relative to its frequency in vitro (immediately preinjection). (A) Read count frequency in the HCC827 cell model. Two independent replicate experiments were performed, with n = 6 for each experiment (for n = 12 total). Each dot represents a single sgRNA and indicates its average normalized representation (computed from 2 independent replicate experiments), organized and displayed as control sgRNAs and sgRNAs targeting ATP-modulating genes classified as glycolytic or mito-respiratory (termed Mito-Resp). Mito-Resp sgRNAs were among the most severely depleted in vivo (HSD17β10, TMEM261, MALSU1, c14orf2) (mean and SEM shown) (mean % representation of nontargeting, glycolytic and mito-respiratory sgRNAs as groups are 99.9%, 62.6%, and 42.2%, respectively. One-way ANOVA of all 3 groups of sgRNAs demonstrate p-value < 0.0001, with Tukey’s multiple comparisons test between glycolytic and control p < 0.02, mito-resp and control p < 0.0001, and between glycolytic and mito-resp ns). (B) H1975 cells grown orthotopically in the lungs of nude mice (n = 6) were dissected and analyzed using the same approach as for flank tumors (A) (mean % representation of nontargeting, glycolytic, and mito-respiratory sgRNAs are 85.4%, 61.6%, and 40.9%, respectively. One-way ANOVA of all 3 groups of sgRNAs demonstrate p-value = 0.006, with Tukey’s multiple comparisons test between glycolytic and control ns, mito-resp and control p-0.005, and between glycolytic and mito-resp ns). (C) Bioluminescent imaging of tumor growth and metastasis in the H1975 model (2 representative mice shown), demonstrating tumor progression in the primary site and metastatic spread to mediastinum and contralateral lung by day 28. In addition, an extrathoracic distant metastasis (left femoral metastasis) developed (blue signal in the leg of the animal on the right). All metastases were confirmed at necropsy and analyzed by sequencing. (D) Mean % representation of nontargeting, glycolytic, and mito-respiratory sgRNAs in H1975 metastases (7 individual metastases from 4 mice, normalized to the matched primary tumor) are plotted on the y-axis (mean and SEM shown, mean % representation of nontargeting, glycolytic, and mito-respiratory sgRNAs are 183%, 192%, and 177%, respectively, one-way ANOVA of all 3 groups of sgRNAs with Tukey’s multiple comparisons test demonstrate ns). (E-G) Validation of individual sgRNAs indicates that silencing mito-translational genes suppresses in vivo tumor growth. sgRNAs targeting top mito-respiratory hits identified from the mini-CRISPRi library screen—c14orf2, MALSU1, and TMEM261—were transduced into HCC827 cells, cells were selected with antibiotics, then injected into the flanks of nude mice. Tumors were allowed to grow for 28 days. (E) qPCR analysis of transduced HCC827 cells assessing level of silencing achieved with single sgRNAs (mean and SEM shown, t test **p < 0.01). (F, G) Tumor weight (F) and tumor volume (G) of HCC827 cells expressing each of the sgRNAs, mean and SEM shown (Student t test, *p < 0.05, **p < 0.01, ***p < 0.001). Underlying data can be found in S1 Data. https://doi.org/10.1371/journal.pbio.3001753.g001 The depletion of individual sgRNAs in flank tumors after 28 days of in vivo growth may reflect cellular loss in response to metabolic pressures developing not only postinjection but at any time during the tumor growth period. We assessed early in vivo tumors for changes in sgRNA representation in a separate experiment in which HCC827 cells expressing the mini-library were grown as flank tumors for either 4 or 7 days (Panel A of Fig A in S1 File). Although control sgRNAs and glycolytic gene-targeting sgRNAs (approximately 95%) were comparably expressed across these time points, mito-translation and respiratory chain sgRNAs, including the most significantly depleted sgRNAs noted above, demonstrated significantly increased expression at day 4 (Panel A of Fig A in S1 File), before becoming significantly reduced in representation at day 7, preceding the near-complete depletion measured at day 28 (Fig 1A). As a group, mito-respiratory sgRNAs demonstrated greater representation at day 4 relative to day 0 compared to control sgRNAs and glycolytic sgNAs (Panel B of Fig A in S1 File), neither of which significantly changed in this initial timeframe. These measured fluctuations in mito-respiratory sgRNAs at early time points in tumor establishment suggest shifting metabolic requirements as tumors grow. The initial increased sgRNA representation (day 4) followed by reduction at day 7 may suggest an increasing reliance on mitochondrial function as tumors grow, perhaps due to an increased requirement for mitochondrial derived ATP as the tumor establishes itself. To assess an alternate human tumor line and in vivo tumor model, we transduced the same mini-CRISPRi sgRNA library into H1975 human lung cancer cells expressing luciferase, which grow in both flank and orthotopic lung cancer models. In separate experiments, H1975 were injected in the flanks of nude mice or into the left lower lung lobes of C.B-17 SCID mice. Tumors in both models were grown for 28 days before dissection and analysis by sequencing. Analysis of sgRNA representation in H1975 tumors from the flank model demonstrated depletion of mito-respiratory sgRNA compared to control sgRNAs (Fig B in S1 File), similar to our findings in HCC827 cells, although the difference in mean representation of the mito-respiratory sgRNAs and the level of significance were reduced compared to HCC827 cells grown in the flank. As an alternative site of tumor growth, H1975 tumors were also injected into an orthotopic lung tumor model. Similar to flank tumors, primary orthotopic lung tumors (Fig 1B) also demonstrated significant reduction of mito-respiratory sgRNAs as a group and severe depletion of individual MALSU1, HSD17β10, c14orf2, TMEM261 sgRNAs. Orthotopic tumors in general develop in anatomically and physiologically distinct contexts that differ from subcutaneously grown flank tumors. However, these data show that primary tumors across cell lines and anatomic sites exhibit a strong requirement for mitochondrial function. Not only is this finding contrary to expectations, these data also specifically support a functionally significant need for mitochondrial-derived ATP, sharply contrasting with our findings in culture where respiratory chain genes were dispensable for growth [6].

Mito-translational genes support in vivo tumor growth in primary sites but are dispensable in metastases The orthotopic lung tumor model also robustly produces intrathoracic and distant metastases (Fig 1C). Tumors form at the primary site (orthotopically) then metastasize to mediastinal lymph nodes, the contralateral lung (Fig 1C) as well as distant sites, recapitulating the lethal pattern of disease spread in patients with lung cancer [13,14]. We then analyzed whether ATP-modulating genes correlate with metastatic tumor spread produced by collecting regional (mediastinal), contralateral lung, and distant metastases from the orthotopic model mice [13] (Fig 1C). Interestingly, while sgRNAs targeting mitochondrial ribosomal and respiratory chain genes were severely depleted in orthotopic tumors (Fig 1B), these sgRNAs were not depleted in metastases from the same animals (Fig 1D). As groups, control, glycolytic and mito-respiratory sgRNAs demonstrated comparable representation in metastases and were not statistically significantly different. These data suggest that ATP-modulating genes have differential effects in primary and metastatic tumors, as knocking down mitochondrial genes that are essential for primary tumor growth was dispensable for metastatic tumor deposits in the same mouse model. Two mice developed metastases in distinctly separate anatomic compartments, and we compared sgRNA representation between these 2 sets of anatomically separated metastases, correlating across functional classes. In the first set of separate site metastases (contralateral lung and bone metastases, Panel B of Fig B in S1 File) the representation of nontargeting sgRNAs and Other Mito sgRNAs correlated (p < 0.01), while in the second set of separate site metastases (contralateral lung and mediastinal metastases) the nontargeting sgRNAs, glycolytic and respiratory chain sgRNAs significantly correlated (p < 0.05). While expanded analyses are clearly needed to fully interrogate the metabolic programs involved in metastatic progression, the correlations in this limited analysis suggest that metastases at distinct anatomic sites share some metabolic programs.

Expressing individual sgRNAs against mito-respiratory hits suppresses in vivo tumor growth The most depleted sgRNAs in the HCC827 flank screen and the H1975 orthotopic screen—those targeting c14orf2, MALSU1, and TMEM261—were prioritized for subsequent individual sgRNA experiments. We focused on c14orf2, MALSU1, and TMEM261, which encode mitochondrial proteins that were previously identified as strong low ATP hits under respiratory conditions [6]; these genes are among the strongest growth repressive hits in both the flank and orthotopic tumor growth screen. CRISPRi sgRNAs targeting each of these individual genes were expressed in HCC827 cells, and after silencing was confirmed (Fig 1E), the cells were injected into the flanks of nude mice. Tumors were grown for 28 days, then measured immediately after removal (Fig 1F and 1G). Each of the CRISPRi sgRNAs targeting mito-respiratory hits significantly decreased in vivo tumor growth in the flank compared to pseudogene control sgRNA. c14orf2, MALSU1, and TMEM261 have no known roles in supporting tumor growth; however, these data provide evidence that their functions promote tumor growth and collectively suggest that mitochondrial genes are necessary for in vivo tumor growth. We assessed the tumors for reduced mitochondrial content as a possible consequence of silencing and cause of reduced respiratory function by western blotting for the mitochondrial protein TOMM20. All tumors from each sgRNA-expressing cell line demonstrated comparable TOMM20 protein levels across all sgRNAs (Fig C in S1 File), consistent with maintained mitochondrial content in the context of the silenced mito-respiratory hits.

Suppression of mito-respiratory hits is associated with overexpression of glycolytic genes in vitro and silencing of glycolytic genes in vivo To gain insight into the convergent mechanisms by which these genes impact respiration, we performed RNA-Seq and transcriptional profiling analysis of HCC827 cells expressing sgRNAs against c14orf2, MALSU1, or TMEM261, comparing each of these to control sgRNA-expressing HCC827 cells grown under basal conditions (Fig 2A). Suppression of each of the 3 mito-respiratory sgRNAs was associated with significant overexpression of glycolytic genes (Fig 2A), with PGK1, ENO2, and HK2 being the most significantly overexpressed glycolytic genes in all 3 analyzed lines (Fig 2A). In addition, gene set enrichment analysis identified glycolysis pathway as the most significantly altered among all 3 cell lines (Fig 2B). These concordant data support transcriptional up-regulation of glycolytic genes as a shared compensatory mechanism utilized by cancer cells when mito-respiratory function is handicapped in order to enable increased glycolytic flux. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 2. Transcriptome profiling of human lung cancer cells expressing sgRNAs against mito-respiratory hits reveals overexpression of glycolytic pathway genes in vitro and repression in vivo. RNA-Seq was performed on HCC827 cells expressing individual sgRNA against top hits c14orf2, MALSU1, TMEM261 (or control sgRNA) (n = 4 samples per cell line) to determine changes in gene expression in vitro and in vivo. (A) Averaged expression for each mito-respiratory sgRNA was compared to control cells, fold-change, indicated by log2(FC) plotted on the x-axis, and log2(p-value) on the y-axis. SgRNA-targeted genes are indicated by green triangles, with each cell line demonstrating the expected reduced expression of its targeted gene (upper left quadrant). Glycolytic genes were the most overexpressed genes overall, with ENO2, PGK1, and HK2 being the most overexpressed genes across all 3 cell lines. (B) Gene set enrichment analysis identifies glycolysis and mitochondria-associated pathways and ontologies as enriched in all 3 cell lines. Significance was set at p < 0.05. (C) Volcano plots shown for each of the single sgRNA cell lines. Expected sgRNA-mediated silencing was observed in all 3 cell lines (blue dot). (D) HCC827 cells transduced with either control sgRNA or MALSU1 sgRNA were injected into the flanks of nude mice, then grown for 28 days, after which tumors were removed and analyzed by RNA-Seq. Pathway analysis was performed comparing the MALSU1 sgRNA cells to control sgRNA cells grown in vivo (blue) or in vitro (red), and displayed in a bubble plot indicating normalized enrichment score (NES) and log 10 p-value for significantly altered pathways. Control sgRNA tumors n = 4, sgMALSU1 tumors n = 2. Underlying data can be found in S1 Data. https://doi.org/10.1371/journal.pbio.3001753.g002 Transcriptome profiling indicated that reactive oxygen species (ROS), glutathione, and NADPH were not consistently altered in the context of mito-respiratory gene silencing. In contrast, the expression of genes involved in fatty acid β oxidation was significantly reduced in cells in which c14orf2, MALSU1, or TMEM261 were silenced compared to control sgRNA-expressing HCC827 cells (Fig 2B). Sterol regulatory element-binding proteins (SREBPs)-regulated gene expression, which regulate lipogenesis as well as growth and mitochondrial metabolism in some cancer cells, were also decreased upon silencing of all 3 mitochondrial genes, suggesting cross-talk between mitochondrial function and lipogenesis. Outside of the broad expression changes in major pathways, individual cell lines also demonstrated shared significantly decreased expression of CLDN2, METTL7A, and TXNIP (Fig 2C). CLDN2 encodes claudin-2, a tight junction protein [15]. METTL17 encodes a mitochondrial protein involved in the translation of mitochondrially encoded genes [16]. TXNIP is a thioredoxin-binding protein involved in redox regulation and glucose uptake that functions as a tumor suppressor gene [17,18]. None of these genes are known to interact with each other, and thus their common down-regulation in each of the mito-respiratory silenced cell lines implicates all 3 as participating in a transcriptional response triggered by decreased mitochondrial ATP levels. Other non-ATP functions were transcriptionally altered only in selected cell lines; that is, transcripts for genes involved in glutathione metabolism were significantly reduced in MALSU1-silenced cells. HCC827 cells transduced with either control sgRNA or MALSU1 sgRNA were injected into the flanks of nude mice, then grown for 28 days, after which tumors were removed and analyzed by RNA-Seq. Pathway analysis was performed to compare sgRNA control cells grown in vivo to the same cells grown in vitro (Fig D in S1 File). Tumor cells grown in vivo demonstrated increased expression of genes involved in collagen degradation, collagen chain trimerization, and ECM (Panel A of Fig D in S1 File), consistent with the involvement of these processes in in vivo tumor growth. Comparing pathway enrichment for MALSU1 sgRNA-expressing cells to control sgRNA cells grown in vivo to those grown in vitro demonstrated significant similarities that are likely associated with MALSU1 loss, notably in ribosome, mitochondrial translation, mitochondrial protein import, complex I biogenesis, and cristae formation (Fig 2D and Panel B of Fig D in S1 File). However, this analysis also identified significant differences in transcriptomic response with MALSU1-silencing that were context-dependent. Specifically, expression of glycolysis pathway genes, while increased in vitro, was significantly reduced in MALSU1 sgRNA cells grown in vivo (Fig 2D), supportive of the concept that tumor cells grown in vivo optimize respiratory function [6]. These differences indicate that the in vivo tumor growth context accentuates the significance of some pathways, and given that mitochondrial and respiratory-driven ATP is substrate-dependent, these differences in transcriptome profiles likely reflect tumor responses to substrate restriction.

Silencing mito-respiratory genes suppresses TCA cycle activity Silencing mito-respiratory hits was associated with shifts in gene expression (from respiration to glycolysis and vice versa) that occurred in a context-dependent manner (Fig 2). To determine how mito-respiratory hits c14orf2, MALSU1, and TMEM261 alter metabolite levels and the pathways in which they function, we performed 13C-glucose and 13C-glutamine-based metabolomics analysis of HCC827 cells expressing individual CRISPRi sgRNAs against these genes. Cells expressing individual CRISPRi sgRNAs were grown under either basal or respiratory (10 mM 2DG), or glycolytic (oligomycin) conditions with either 13C-glucose or 13C-glutamine for 18 hours. Cells were then collected, metabolites extracted and analyzed by mass spectroscopy. Metabolomics analysis demonstrated the relative differences in sources of TCA cycle metabolites, with Gln being the predominant source (Fig E in S1 File). Glc labeling resulted in relatively low percent labeling of citrate (ranging from 5% to 30%; seen in Panel A of Fig E in S1 File, right panel), whereas Gln labeling of most TCA cycle metabolites was approximately 80% or greater (Panel C of Fig E in S1 File, right panel). Overall, total amounts of TCA cycle metabolites varied between the 4 cell lines, with Glc and Gln labeling demonstrating similar differences between cell lines under both basal and 2DG conditions (Fig E in S1 File). Basal growth of TMEM261 and MALSU1-silenced cells was associated with significant reduction in the percent labeling of the TCA metabolites citrate and aconitate (Panel A of Fig E in S1 File), while c14orf2-slienced cells resembled control cells. Growth under glycolytic block (2DG) generally decreased 13C incorporated into these TCA metabolites when compared to basal conditions, as expected, but in this condition, c14orf2-silenced cells showed reduced labeling of citrate and aconitate compared to control cells. c14orf2 silencing reduced Glc-labeling of citrate when 2DG was present, similar to silencing TMEM261 and MALSU1, indicating a deficiency shared upon silencing all 3 mito-respiratory-hits (Panel B of Fig E in S1 File, right panel). However, since this effect was only visible with a glycolytic block, c14orf2 may be more dispensable than either TMEM261 and MALSUI when glucose is available. 13C-glutamine-labeling also distinguished metabolite labeling profiles for cells in which mito-respiratory hits were silenced. Compared to basal control conditions, under forced respiration (2DG; Panels C and D of Fig E in S1 File, right panels), 13C incorporation into succinate was the most significantly reduced in all 3 mito-respiratory-deficient lines compared to control.

Silencing mito-respiratory hits shifts cells to greater glycolytic metabolism in vitro Genetically mediated modulation of cellular ATP involves genes that concurrently optimize 1 metabolic pathway while suppressing the alternative pathway [6]. To detect metabolic shifts towards glycolytic or respiratory function, we compared isotopologues of glycolytic and respiratory metabolites among mito-respiratory-deficient cells, grown under basal or forced respiratory (2DG) conditions (Fig F in S1 File). Examining fructose 1,6-bisphosphate (F16BP) as an index glycolytic metabolite, under basal conditions, the unlabeled (0 carbon) F16BP in MALSU1 sgRNA and c14orf2 sgRNA cells was comparable to control and modestly decreased in TMEM261 sgRNA cells, while fully labeled (6 carbon) F16BP was comparable between control and TMEM261- as well as c14orf2-silenced cells and modestly reduced in MALSU1-silenced cells (Panel A of Fig G in S1 File, upper panels). However, adding 2DG revealed that glycolytic metabolism in mito-respiratory-deficient cells diverged from control cells (Panel A of Fig G in S1 File, lower panels). In the presence of 2DG, all 3 mito-respiratory silenced cell lines increased labeling of the downstream glycolytic metabolite F16BP (6 carbon labeled), in contrast to control cells (Panel A of Fig G in S1 File, lower right panel), consistent with increased relative glycolytic capacity compared to control cells. We similarly compared Gln-labeled glutamate among the cell lines under basal and forced respiratory conditions (Panel B of Fig G in S1 File). Under basal conditions, silencing of c14orf2, MALSU1, or TMEM261 resulted in a marked increase in the percentage of unlabeled glutamate compared to control cells (Panel B of Fig G in S1 File). c14orf2-silenced cells had reduced fully labeled glutamate (5 carbons labeled) in contrast to TMEM261 or MALSU1-silenced cells (Panel B of Fig G in S1 File, right panel), which had increased fully labeled glutamate. When grown in 2DG, c14orf2, MALSU1, and TMEM261-silenced cells all demonstrated increased unlabeled glutamate and reduced fully labeled glutamate compared to control cells (Panel B of Fig G in S1 File). These Gln-derived labeling patterns common to the mito-respiratory-deficient lines presumably reflect their discrete defects in respiratory metabolism that are notably worsened with forced respiration. We next used the incorporation of Glc and Gln-labeled metabolites as proxies for the utilization of glycolysis and respiration. The ratios of glycolytic label incorporation (6C/unlabeled F16BP) to the respiratory label incorporation (5C/unlabeled glutamate) for c14orf2, MALSU1, and TMEM261-silenced cells indicated a shared glycolysis-shifted metabolic response (Fig 3A, Panels E and F of Fig E in S1 File, and Fig F in S1 File). PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 3. Mito-respiratory hits have distinct metabolic signatures that differ between in vitro and in vivo contexts. HCC827 cells expressing individual CRISPRi sgRNA (control, or mito-respiratory hits c14orf2, MALSU1,TMEM261 silencing confirmation shown in Fig 2A) were grown in basal media or media with 10 mM 2DG (n = 4 per group, individual data points shown as grey dots) with either [U-13C]glucose or [U-13C]glutamine for 18 hours. Cells were collected, metabolites extracted and analyzed by mass spectrometry. (A) Estimated metabolic flux ratios for cells expressing sgRNAs targeting c14orf2, MALSU1, or TMEM261. Using the measured % of total all labeled and unlabeled glycolytic [[U-13C]glucose→F16BP indicates % total unlabeled (0 carbon) or completely labeled (6 carbon)] and respiratory metabolite values [[U-13C]glutamine→glutamate analysis indicates total unlabeled (0 carbon labeled) on the left and fully labeled (all 5 carbons labeled)] (shown in Panels A and B of Fig F in S1 File), the estimated metabolic flux ratios of glycolytic flux (6C/unlabeled) to the respiratory flux (5C/unlabeled) are shown for 3 mito-respiratory hits (TMEM261, c14orf2, and MALSU1). The ratios distinguish the glycolysis-shifted metabolism apparent in the 3 cell lines in which mito-respiratory hits are silenced. (n = 4 replicates per sample, 1-way ANOVA, Dunnett’s multiple comparisons test, *p < 0.05, ***p < 0.001). (B, C) PCA was applied to the fractional contribution values of the metabolomics data for HCC827 cells expressing individual CRISPRi sgRNA (control, or mito-respiratory hits c14orf2, MALSU1, or TMEM261). 13C glucose-derived labeling of cells grown under either control or 2DG media (D) was compared in (B) control or oligomycin in (C). The absolute change in PC1 and PC2 values of the fractional contribution analysis for each cell line comparing control to 2DG growth (ΔPC1 = PC1 control − PC1 2DG , ΔPC2 = PC2 control − PC2 2DG ) shown in (B) or control to oligomycin shown in (C) are plotted (n = 4 replicates). (D, E) HCC827 cells expressing either control or MALSU1 sgRNA were injected into the flanks of nude mice. After 28 days of growth, mice were injected with 13C glucose to label tumor metabolites, after which tumors were collected and metabolites analyzed. (D) PCA was performed using the fractional labeling values of glucose-derived metabolites comparing MALSU1-deficient and control HCC827 cells grown in vitro or in vivo. (E) PCA was performed on the amount labeled of glucose-derived metabolites in MALSU1-deficient and control HCC827 cells grown in vitro or in vivo. Underlying data can be found in S1 Data. c14orf2, chromosome 14 open reading frame 2; F16BP, fructose 1,6-bisphosphate; MALSU1, mitochondrial assembly of ribosomal large subunit 1; PCA, principal component analysis. https://doi.org/10.1371/journal.pbio.3001753.g003 Thus, while the absolute magnitude of basal Glc and/or Gln-labeling in individual glycolytic and respiratory metabolites varied between each of the mito-respiratory-silenced cell lines, the relative incorporation patterns from combined Glc and Gln-labeling highlight increased glycolytic utilization upon c14orf2, MALSU1, or TMEM261 knockdown.

Silencing of ATP-modulating mito-respiratory genes is associated with discrete metabolite profiles and metabolic network structure that distinguish in vitro and in vivo growth We then examined metabolic networks on a broader scale by performing principal component analysis (PCA) of the metabolomics data (Fig 3, Figs G and H in S1 File). After computing Principal Component (PC)1 and PC2 scores for control and mito-respiratory silenced HCC827 tumor cells that were grown under either control versus 2DG, or control versus oligomycin (Figs E and G in S1 File), we plotted the ΔPC1:ΔPC2 score between control and 2DG for each cell line (Fig 3B). This depicts the net magnitude and directionality of PC score changes between control and 2DG, which did not indicate a common respiration-driven metabolite signature among the mito-respiratory silenced lines. In contrast, oligomycin treatment (forced glycolysis) was associated with similar net shifts in PC1 and PC2 scores among the 4 cellular genotypes (Fig 3C), suggesting a common glycolytic response to suppression of all 3 mito-respiratory hits (Fig 3B). Considering the differential requirement for mito-respiratory hits in vitro versus in vivo, we next assessed metabolites in vivo. We injected HCC827 cells expressing either control or MALSU1 sgRNA into the flanks of nude mice, grew tumors for 28 days, then injected mice with 13C glucose, after which tumors were collected and metabolites analyzed. We performed PCA based on the fractional contributions (amount of metabolite labeled by 13C glucose divided by the total metabolite pool size) and also the total amounts of measured metabolites (Fig 3D and 3E). This dimensionality reduction technique identifies PCs (PC1, PC2) from linear recombinations of the data that explain a majority of the variance. This approach allows us to compare differences in variance across fractional contributions and total amounts between conditions. These comparative analyses showed clear growth context dependence (in vitro versus in vivo) of metabolites in general, however also indicated MALSU1-specific effects. PC1 describes most of the variation between metabolite pool sizes and fractional labeling (62.06% and 75.52% of the variation, respectively) in vitro versus in vivo, which may result from differences in substrate introduction and metabolite extraction. However, PC2 describes variation in metabolite pool sizes and fractional labeling (16.42% and 7.7% of the variation, respectively) unassociated with in vitro versus in vivo difference, likely attributable to MALSU1 knockdown. On the basis of fractional contributions, samples from MALSU1 knockdown cells and control cells clustered in vivo but separated in vitro (Fig 3D). Conversely, PCA of metabolite amounts separated samples from MALSU1 knockdown cells and samples from control cells in vivo but only marginally in vitro (Fig 3E), suggesting that MALSU1-silencing effects on the metabolic network in vivo are driven by discrete metabolites. We then sought to assess metabolic network structure for each of the mito-respiratory genes by integrating transcriptomic and metabolomic data on metabolism pathway-focused graphs [19–21] (Fig 4). All 3 mito-respiratory-deficient cell lines shared similar activation of nodes in glycolytic metabolism (Fig 4A–4C) and similar reduction in TCA cycle network (Fig I in S1 File); these shared features suggest a conserved metabolic network structure among mito-respiratory functional deficits. Furthermore, similar integrated transcriptomic and metabolomic-based pathway analysis of glucose-labeling by MALSU1-deficient cells grown in vivo revealed a striking reduction in glycolytic metabolite pools as compared to in vitro growth (Fig 4D), consistent with the transcriptional changes summarized in Fig 2D. These data provide further evidence that glycolytic function distinguishes the metabolic networks of mito-respiratory-deficient cells and may underlie the differential substrate-driven requirement of these genes in vivo and in vitro. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 4. Mito-respiratory gene silencing alters metabolic network structures and in vivo tumor growth. (A-D) Pathway analysis of HCC827 cells expressing individual CRISPRi sgRNA (control or mito-respiratory hits c14orf2, MALSU1, or TMEM261) grown in vitro (A-C) or HCC827 cells expressing an individual CRISPRi sgRNA against MALSU1 grown in vivo as flank tumor (D) was performed by integrating transcriptome and metabolomics data and plotted on a KEGG graph depicting pathway components for glycolysis/gluconeogenesis. (E) H1975 cells transduced with either control sgRNA or individual sgRNAs against c14orf2, MALSU1, or TMEM261 were analyzed using Seahorse. OCR shown, compilation of 3 independent experiments, each analyzing n = 4 per cell line, normalized to cell numbers. Mean shown, 2-way ANOVA, **p < 0.01, ***p < 0.001. (F) OCR and ECAR measurements for cells grown under basal or forced glycolysis (oligomycin). Compilation of 3 independent experiments, each analyzing n = 4 replicates per cell line and normalized to cell numbers. (G) ATP levels were measured in different substrate conditions (basal, respiratory (10 mM 2DG), glycolytic (5 μM oligomycin) or depletion (no glucose or pyruvate)) mean and SEM shown, 2-way ANOVA, ***p < 0.001). (H, I) NAD+/NADH ratios and NAD+/NADH pool size. Shown are compilation of 2 independent experiments **p < 0.01, ***p < 0.001. (J) MitoSox assay measuring mitochondrial superoxide. Compilation of 3 independent experiments, each analyzing n = 4 replicates per cell line. Two-way ANOVA, **p < 0.01. (K) Cells expressing either control sgRNA or MALSU1 sgRNA were injected orthotopically into the left lungs of mice, then mice were imaged using BLI 21 days postinjection. (L) Day 21 radiance within the left 1/3, central 1/3, and right 1/3 chest regions in mice injected with control sgRNA tumor cells. N = 4 mice, mean shown. (M) Day 21 radiance within left 1/3, central 1/3, and right 1/3, chest regions in mice injected with MALSU1 sgRNA tumor cells. N = 6 mice, mean shown. (N) Central 1/3 and contralateral right 1/3 lung radiance each normalized to the matched left 1/3 radiance for control and MALSU1 mice, mean shown, t test *p < 0.05. (O) Day 21postinjection radiance within the left 1/3 chest in mice injected with either control or TMEM261-silenced tumor cells (n = 4 mice for control, n = 5 mice for TMEM261, mean shown Mann–Whitney test, *p < 0.05). (P) Day 28 postinjection central 1/3 and contralateral right 1/3 lung radiance, each reading normalized to the matched left 1/3 radiance for control and TMEM261 cells (n = 4 mice and n = 5 mice, respectively) mean shown, Mann–Whitney test, ns. (Q) HE-stained sections of right lungs from mice injected with either control sgRNA tumor cells or TMEM261 sgRNA tumor cells showing micrometastatic tumor deposits. Red arrows indicate intraparenchymal micrometastases; black arrows indicate perivascular metastases. (R) Systems-level testing implicates context-specific growth effects of mitochondrial/respiratory function and ATP levels. ATP-modulating CRISPRi hits grown in lung cancer cells produce specific energy substrate-driven growth effects in vitro and in vivo. In vitro growth correlates with glycolytic ATP, while in vivo primary growth in the subcutaneous and orthotopic lung setting correlate with mitochondrial-derived ATP. Silencing discrete mito-respiratory genes also impacts the growth of regional metastases, which are distinguishable from the primary tumor. Underlying data can be found in S1A–S1D and S2E and S2F and S3G–S3J Data. BLI, bioluminescence imaging; ECAR, extracellular acidification rate; HE, hematoxylin–eosin; OCR, oxygen consumption rate. https://doi.org/10.1371/journal.pbio.3001753.g004

Silencing mito-respiratory genes decreases respiration H1975 human lung cancer cells expressing control sgRNA or sgRNA against c14orf2, MALSU1, or TMEM261 were then compared for differences in respiratory chain function. Under basal growth conditions, silencing of all 3 mito-respiratory genes was associated with a reduction in basal and maximal oxygen consumption rate (OCR), with silencing of TMEM261 or c14orf2 producing more profound and significant reduction in respiration than MALSU1 silencing (Fig 4E and 4F). Silencing of all 3 mito-respiratory genes was also associated with a reduction in ATP-linked respiration (shown as reduced shift along the y-axis with oligomycin as compared to control cells; Fig 4F). The glycolytic reserve capacity (shift in extracellular acidification rate (ECAR) along the x-axis with oligomycin) associated with all 3 mito-respiratory hits was similarly reduced versus control. c14orf2 silencing also resulted in elevated baseline glycolysis. We then measured ATP levels in cells across multiple substrates (basal, respiratory, glycolytic) and found that under basal conditions, c14orf2 and TMEM261 silencing are associated with increased ATP (Fig 4G). Under respiratory conditions, c14orf2 and TMEM261 silencing significantly reduced ATP (Fig 4G). Under glycolytic conditions, ATP was similar across lines with the exception of TMEM261, whose silencing was associated with increased ATP (Fig 4G). Among all the mito-respiratory hits, TMEM261 silencing was associated with the largest ATP fluctuations across conditions, suggesting that TMEM261 function may be more substrate-dependent than c14orf2 and MALSU1. Interestingly, MALSU1 silencing was not associated with changes in ATP levels, although OCR was reduced compared to control, presumably reflecting a compensatory reduction in ATP consumption to equilibrate against reduced ATP production. To determine potential effects of mito-respiratory gene silencing on redox function, we then measured NAD+/NADH levels (Fig 4H and 4I). Silencing of either c14orf2 or TMEM261 significantly decreased NAD(H) pool size compared to control sgRNA (that is, NAD+ and NADH), while NAD(H) pool size was unchanged in MALSU1-silenced cells. For TMEM261-silenced cells, the reduction in NAD+/NADH ratio occurred in the context of a significantly increased total NAD(H) pool size, whereas MALSU1-silenced cells demonstrated a slightly reduced NAD(H) pool size compared to control cells. Taken together, these data further support decreased oxidative phosphorylation activity for c14orf2 and TMEM261-silenced cells but indicate that the mito-respiratory genes have distinct effects on NAD+/NADH metabolism, suggesting that changes in NAD+/NADH metabolism do not underlie their effects on tumor growth and survival. The observed changes in NAD+/NADH metabolism may indicate that there are changes in NAD+/NADH synthetic or catabolic processes less directly related to energy metabolism that occur in MALSU1 or TMEM261-silenced cells. In summary, changes in NAD+/NADH metabolism were not consistent between mito-respiratory silenced lines and therefore are unlikely to explain the similar changes in growth. Mitochondrial production of ROS is a by-product of respiration, and excessive ROS production can influence cell growth. To determine whether mito-respiratory gene-silencing results in ROS overproduction, which can be a limiting factor in tumor cell growth, we measured mitochondrial superoxide levels in mito-respiratory gene-silenced cells across multiple substrates (basal, respiratory, and glycolytic). Mitochondrial superoxide levels were only modestly reduced in cells expressing sgRNA against MALSU1, and comparable to control in the other mito-respiratory-deficient cell lines across all 3 substrates (Fig 4J), arguing against an increase in mitochondrial ROS potentially causing toxicity to cells and inhibiting tumor growth by this mechanism.

[END]
---
[1] Url: https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.3001753

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/

You are viewing proxied material from magical.fish. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.