Cellular acidosis triggers MondoA transcriptional activity by driving mitochondrial ATP production

MondoA and its transcriptional target thioredoxin-interacting protein (TXNIP) constitute a regulatory loop that senses glycolytic flux and controls glucose availability. Cellular stress also triggers MondoA activity and TXNIP expression. To understand how MondoA integrates glucose and stress signals, we studied its activation by acidosis. We found that acidosis drives mitochondrial ATP (mtATP) synthesis. The subsequent export of mtATP from mitochondria via adenine-nucleotide transporter and voltage-dependent anion channel, and the enzymatic activity of mitochondria-bound hexokinase results in the production of glucose-6-phosphate (G6P), a known activator of MondoA transcriptional activity. MondoA localizes to the outer-mitochondrial membrane (OMM), and in response to G6P, shuttles to the nucleus and activates transcription. Our data suggests that MondoA is a required feature of a glucose- and mtATP-dependent, OMM-localized signaling center. We propose MondoA functions as a coincidence detector and its ability to sense glucose and cellular stress is coupled to the concerted production of G6P.

localized signaling center. We propose MondoA functions as a coincidence detector and its ability 23 to sense glucose and cellular stress is coupled to the concerted production of G6P. 24

INTRODUCTION 26
Glucose is a major source of carbons for the production of ATP and biosynthetic 27 intermediates. Dysregulation of glucose uptake and metabolism underlies many diseases including 28 cancer and diabetes (Petersen et al., 2017, Hay, 2016. Thus, it is important to understand the 29 precise molecular mechanisms that regulate glucose homeostasis in normal and pathological  (breast cancer), MCT1 (lung cancer) and NHE1 (brain cancer). All expression data was collected from TCGA. Spearman and Pearson correlation statistics are reported as r and ρ, respectively. (B) An acidosis gene signature was determined for the 2016 METABRIC breast cancer dataset. These scores were compared to TXNIP expression from the dataset and correlation statistics were performed.  acidosis-driven transcriptional response. We next used regression analysis to look for genes that 156 are affected by both DMEM Acidic treatment and genotype. Loss of MondoA prevented the 157 induction/suppression of several acidosis-regulated genes; however, only two genes, TXNIP and 158 one of its paralogues, ARRDC4, were entirely dependent on MondoA ( Figure 3A). 159 We next performed pathway analysis on genes differentially regulated in HeLa and 160 HeLa:MondoA-KO cells treated with DMEM Acidic . Consistent with the results above, TXNIP and 161 ARRDC4 were the most highly MondoA-dependent genes, with log2(fold-changes) of 7.9 and 5.1, 162 respectively ( Figure 3B). We identified 157 other differentially regulated genes in HeLa:MondoA-163 KO cells (adjusted p-value ≤ 1E-10). Pathways that were upregulated in HeLa:MondoA-KO cells 164 were enriched for fatty acid metabolism, sterol biosynthesis, ion homeostasis, ROS metabolism, 165 and pyridine metabolism pathways, whereas cell death and proliferation pathways were 166 Together these results suggest that cytosolic and/or mitochondrial proton gradients, but not pH-198 Cytosolic and mitochondrial protons contribute to ETC function, which cooperatively 201 builds and consumes a proton gradient to synthesize ATP. We therefore sought to evaluate how 202 staining. (C) Total cellular ATP levels were determined using luciferase-based assay. (D) Mit-ATEAM, a mitochondrial-targeted ATP-biosensor, was used to determine how DMEM Acidic affects mitochondrial ATP. Widefield microscopy was used to capture images in the FRET and CFP channels. After images were obtained, mitochondria were analyzed for FRET and CFP signal. FRET signal was normalized using CFP. (E) ATP5I mRNA level in HeLa cells expressing scrambled (siSCRM, n=1) or ATP5I-specific siRNA (siATP5I, n=2). (F) Mit-ATEAM was used to determine how DMEM Acidic affects mitochondrial ATP production in the context of siSCRM or siATP5I. (G) TXNIP mRNA level following DMEM Acidic treatment of HeLa cells expressing scrambled or ATP5I-specific siRNA. ** p<0.01; ****p<0.0001; ns -not significant the ETC contributes to acidosis-driven MondoA activity. We used 143B⍴ 0 osteosarcoma cells, 203 which lack mitochondrial DNA (mtDNA) and are respiration deficient (King and Attardi, 1989). 204 TXNIP was induced in parental 143B cells treated with DMEM Acidic (Figure 4A), yet the induction 205 of TXNIP was blunted in 143B⍴ 0 cells ( Figure 4B). TXNIP induction was rescued in 143B⍴ 0 cells 206 that had been repopulated with wild type mitochondria (143B⍴ 0 :WT-cybrid cells; Figure 4C). 207 These genetic experiments confirm previous inhibitor studies that implicated a functional ETC in 208 MondoA transcriptional activity (Yu et al., 2010, Han and. 209 Given the predominant role of the ETC in ATP synthesis, we determined whether 210 mitochondrial ATP (mtATP) synthesis is required to trigger the MondoA/TXNIP axis. We used 211 ETC complexes I-IV build a proton gradient by pumping protons from the mitochondrial 222 matrix to the inner membrane space. Given that the outer mitochondrial membrane is freely 223 permeable to protons (Cooper, 2000), we hypothesized that acidosis leads to intracellular 224 acidification, hyperpolarization of the inner-mitochondrial membrane and ATP synthesis. Using 225 the pH-sensitive dye BCECF-AM, we determined that DMEM Acidic treatment shifted intracellular 226 pH from 7.2 to 6.5 ( Figure 5A). The drop in pH was accompanied by an increase in mitochondrial 227 membrane potential as measured by the dye JC1 ( Figure 5B) and an increase in total cellular ATP 228 levels ( Figure 5C). Collectively these data show that treating cells with low pH medium increases 229 total cellular ATP levels. We next sought to determine whether the accumulation of mtATP resulted from increased 240 synthesis or decreased mitochondrial export. We blunted expression of ATP5I, an essential 241 component of the ATP synthase, using siRNA-mediated knockdown ( Figure 5E). Consistent with 242 our working model, ATP5I knockdown decreased not only the steady state level of mtATP, but 243 also the low pH-driven increase in mtATP ( Figure 5F). Furthermore, ATP5I knockdown prevented 244 TXNIP induction in response to DMEM Acidic treatment ( Figure 5G). Together these data show that 245 acidosis drives mtATP production through ATP synthase and that mtATP synthesis is required for 246 low pH-driven MondoA transcriptional activity. contrast, most glycolytic intermediates were decreased in response to DMEM Acidic ; however, G6P 260 levels were increased 3-fold ( Figure 6C). 261 Second, we tested the contribution of the channel, comprised of the adenine-nucleotide 262 transporter (ANT) in the inner-mitochondrial membrane and voltage-dependent anion channel 263 (VDAC) in the outer-mitochondrial membrane, that exports mtATP from the mitochondria. 264 Consistent with our working model, which states that mtATP must be exported from the matrix, 265 siRNA-mediated knockdown of ANT2 prevented TXNIP induction in response to low pH medium 266 ( Figure 6D). This finding suggests that mtATP functions outside the mitochondria to trigger 267 MondoA transcriptional activity, rather than by an indirect signaling-based mechanism. 268 Third, we used several approaches to test the contribution of HK2 to low pH-driven 269 MondoA activity. siRNA pools against HK2 blocked TXNIP induction in response to low pH 270 treatment ( Figure 6E

Cell lines 386
A list of cell lines used is provided in the Key Resources Table. All cells were maintained in 387 DMEM +10% FBS (Gibco), 100 units/mL penicillin (Gibco) and 100 units/mL streptomycin 388 (Gibco). 143Br 0 and cybrids were cultured with 1 mM sodium pyruvate and 50 µg/mL uridine. 389 Cells were passaged and treated in an incubator set at 37 °C and 5% CO2. Three days prior to fractionation siRNAs were transfected using Lipofectamine 3000 (Thermo 503 Fischer). Cells were washed with cold PBS and dislodged from plate by scraping. Cells were 504 pelleted by centrifugation and resuspended in 1 mL of fractionation buffer (40 mM HEPES pH 505 7.9, 137 mM NaCl, 2.7 mM KCl, 1.5 mM MgCl2, 0.34 M sucrose, 10% glycerol, 1 mM DTT, 506 0.5% NP40, protease and phosphatase inhibitors). Cells were incubated on ice for 10 minutes then 507 pelleted by centrifugation at 1000 rcf for 5 minutes. The supernatant was kept (cytoplasm) and the 508 pellet (nuclei) was washed three times with 0.5 mL fractionation buffer. 509

GC-MS 520
Following treatment, cells were collected into a 1.5 mL microcentrifuge tube then snap frozen 521 using liquid nitrogen. Cells were kept at -80°C until metabolite extraction was performed. 450 µL 522 of cold 90% methanol and internal standards were added to cells and incubated at -20°C for 1 hour. 523 Tubes were then centrifuged at -20,000×g for 5 minutes at 4°C. Supernatants were dried using a 524 speed-vac. 525 526 Samples were converted into volatile derivatives amenable to GC-MS. Briefly, dried samples were 527 resuspended in O-methoxylamine hydrochloride (40 mg/mL) then mixed with 40 µL N-methyl-528 N-trimethylsilyltrifluoracetamide and mixed at 37°C. After incubation, 3 µL fatty acid methyl ester 529 standard solution was added. 1 µL of this final solution was injected into gas chromatograph with 530 an inlet temperature of 250°C. A 10:1 split ratio was used. Three temperatures were ramped with 531 a final temperature of 350°C and a final 3-minute incubation. A 30 m Phenomex ZB5-5 MSi 532 column was used. Helium was used as carrier gas at 1 mL/minute. Samples were analyzed again 533 with a 10-fold dilution. 534 535 Data was collected using MassLynx 4.1 software (Waters). Metabolites were identified and peak 536 area was determined using QuanLynx. Data was normalized using Metaboanalyst 3.6 537 (http://www.metaboanalyst.ca/). Quantile normalization, log transformation and Pareto scaling 538 were used. Normal distribution of values was used to determine fold changes. 539

RNA-sequencing library construction and analysis 541
Total RNA was extracted from cells using a Quick RNA Miniprep Kit (Zymo Research) according 542 to manufacturer's recommendations. mRNA was isolated and library production performed using 543 a Stranded mRNA-Seq Kit with mRNA Capture Beads (Kapa). Library quality was analyzed using 544 an Agilent High Sensitivity D1000 ScreenTape. Single-end sequencing for 50 cycles was 545 performed using an Illumina HiSeq. The resulting FASTQ files were aligned to the human genome 546 (hg38) using STAR. DESeq2 was used to quantify transcript abundance, differential expression, 547 FPKM values, and interaction terms (genotype:treatment combinatorial statistic). Data is presented as mean ± standard deviation. One-way ANOVA was used to account for 571 variation and significance was determined using a two-tailed Student's t-test. Unless otherwise 572 indicated, at least three biological replicates were used for each analysis.