ATM deficiency alters cellular metabolism and enhances Warburg effect in CRPC cells

Cellular metabolic reprogramming has been shown to support their rapid proliferation and sustainable growth of cancer cells (Pavlova and Thompson 2016). To test whether ATM deficiency-induced CRPC progression is caused by alteration of cellular metabolism, we conducted a metabolite profiling for C4–2/Ctrl and C4–2/ATM-KO cells. Notably, ATM deficiency resulted in a significant change of global metabolites in C4–2 cells (Fig.2A, Supplementary table 1). This observation is supported by a principal component analysis in which C4–2/ATM-KO cells were segregated from C4–2/Ctrl cells on metabolite profiles (Fig.2B). We further analyzed the metabolic pathways associated with the changed metabolites and found that ATM deficiency significantly altered a large number of metabolism pathways, especially glucose-related energy pathways (Fig.2C). It is well known that cancer cells produce energy to support their growth through aerobic glycolysis instead of mitochondria-mediated OXPHOS (Warburg effect). To investigate whether ATM deficiency affects Warburg effect in CRPC cells, we compared the levels of intermediate metabolites during the aerobic glycolysis pathway from the profiling data. All glycolytic intermediates, except pyruvate, were shown significantly higher levels in C4–2/ATM-KO cells compared to that in C4–2/Ctrl cells (Fig.2D). In agreement with the profiling data, ATM deficiency led to an increase in both glucose consumption and lactate production in C4–2 cells (Fig.2E &F). A similar result of increased glucose consumption and lactate production was observed in another ATM-deficient CRPC CWRR1 cells (Fig.2 E&F).Therefore ATM-deficient CRPC cells have an altered metabolism profile with enhanced aerobic glycolysis, which suggests that it is highly likely that ATM deficiency promotes CRPC progression through enhanced aerobic glycolysis.

Open in a separate window

Figure 2

ATM deficiency leads to an enhanced glycolytic utilization in CRPC cells. (A) Heat map to show levels of global metabolites in C4–2/Ctrl and C4–2/ATM-KO cells (n = 3 biological replicates). (B) Principal component analysis (PCA) of metabolic distributions in the indicated cell lines. (C) Enriched glucose-related metabolite pathways in C4–2/ATM-KO cells. Statistically significant differences were assessed by MetaboAnalyst software as indicated in the Materials and Methods. (D) LC-MS analysis of cell extractions to compare the levels of glycolytic intermediates between C4–2/Ctrl and C4–2/ATM-KO cells (n = 3). G6P, glucose 6-phosphate; F6P, fructose 6-phosphate; F1,6P, fructose 1,6-bisphosphate; 1,3BPG, 1,3-bisphosphoglycerate; 3PG, 3-phosphoglycerate; pPy, phosphoenolpyruvate; Py, pyruvate; Lac, lactate. (E & F) Glucose consumption and lactate production in paired C4–2/Ctrl and C4–2/ATM-KO, CWRR1/Ctrl and CWRR1/ATM-KO cells. Glucose consumption was calculated as indicated in the Materials and Methods. Results are normalized to the control group (n = 3). All data are depicted as mean ± s.d. * p < 0.05 and ** p < 0.01 by two-tailed Student’s t-test.

ATM deficiency detours glucose flux into aerobic glycolysis in CRPC cells

To further validate the enhanced Warburg effect in ATM-deficient CRPC cells, we used uniformly ¹³C-labeled glucose ([U-¹³C]-Glc) to trace the glucose influx in C4–2/Ctrl and C4–2/ATM-KO cells. Three different glucose-derived metabolite pathways were followed: 1) ¹³C₃-lactate (m+3 form) that represents the endpoint of aerobic glycolysis; 2) intermediates of TCA (tricarboxylic acid) cycle (from ¹³C₂-citrate to ¹³C₂-malate) that represents the mitochondrial OXPHOS pathway; and 3) ¹³C₃-Alanine and ¹³C₃-Aspartate that indicate other glucose-derived pathways. As shown in Fig.3, the level of glucose-derived ¹³C₃-lactate was significantly raised in ATM-KO C4–2 cells whereas the levels of glucose-derived intermediates of TCA cycle (¹³C₂-citrate/isocitrate, ¹³C₂-α-ketoglutarate, ¹³C₂-succinate, ¹³C₂-fumarate and ¹³C₂-malate) were all reduced in ATM-KO C4–2 cells, suggesting that ATM deficiency shunts the glucose flux to glycolysis from OXPHOS. The augment of the entry of glucose to glycolysis is further supported by the observation that the production of glucose-derived ¹³C₃-alanine and ¹³C₃-aspartate was attenuated in C4–2/ATM-KO cells (Fig.3). A similar result was found in CWRR1 cells as well wherein the ¹³C₃-lactate level was increased in CWRR1/ATM-KO cells while the concentrations of other representative glucose-derived metabolites were decreased in CWRR1/ATM-KO cells (Supplementary Fig.1). Taken together, the glucose tracing results suggest that ATM deficiency further diverts the glucose flux to aerobic glycolytic pathways, i.e. enhances Warburg effect, in CRPC cells.

Open in a separate window

Figure 3

ATM deficiency detours the glucose flux in CRPC cells. C4–2/Ctrl and C4–2/ATM-KO cells were grown in uniformly ¹³C-labeled glucose medium. The incorporation of ¹³C atoms from ¹³C₆-Glc in to citrate/isocitrate, α-ketoglutarate (α-KG), succinate, fumarate, malate, lactate, alanine and aspartate are denoted as m+n, where n is the number of ¹³C atoms. Blue, red, purple and green circles all indicate that ¹³C atoms are derived from ¹³C₆-Glc but through different pathways. A scheme describing glucose flows is shown in the middle of the figure. LDHA, lactate dehydrogenase A; GPT2, glutamate-pyruvate transaminase; PC, pyruvate carboxylase; CO₂ indicates where carbon dioxide is released. All data are depicted as mean ± s.d. * p < 0.05 and ** p < 0.01 by two-tailed Student’s t-test.

Upregulation of LDHA contributes to ATM deficiency-induced CRPC progression

The enhanced Warburg effect as well as the increased production of lactate in ATM-deficient CRPC cells prompted us to determine which enzyme in the glycolysis pathway is regulated by ATM deficiency. We first compared the level of metabolites along the glycolysis pathway from the tracing data and found that only glucose-derived lactate, not other intermediate metabolites, was significantly increased in C4–2/ATM-KO cells (Fig.4A), suggesting that the enzyme LDHA, that is responsible for converting pyruvate to lactate, might be upregulated by ATM deficiency. Actually, we examined the mRNA level of LDHA as well as other enzymes along the glycolysis pathway by RT-qPCR in the two cell lines. Only LDHA was dramatically upregulated in C4–2/ATM-KO cells (Fig.4B) and this upregulation is further supported by the western blot showing ~3 fold increase of LDHA protein expression in ATM-KO cells (Fig.4C). To further confirm the upregulation of LDHA by ATM loss, we treated C4–2 cells with ATM inhibitor KU-60019 and found that the expression of LDHA increased after 72 hours treatment of cells with KU-60019 (Supplementary Fig.2).

Open in a separate window

Figure 4

ATM deficiency upregulates LDHA to promote CRPC progression. (A) ¹³C-isotopomer tracing analysis of glucose-derived glycolytic metabolites in C4–2/Ctrl and C4–2/ATM-KO cells (n = 3). G6P, glucose 6-phosphate; F6P, fructose 6-phosphate; F1,6 P, fructose 1,6-bisphosphate; 1,3BPG, 1,3-bisphosphoglycerate; 3PG, 3-phosphoglycerate; pPy, phosphoenolpyruvate; Py, pyruvate; Lac, lactate. (B) RT-qPCR to determine the mRNA levels of glycolytic enzymes. Data are normalized to the control group (n = 3 independent experiments). HK, hexokinase; PFK, phosphofructokinase; PGK, phosphoglycerate kinase; PK, pyruvate kinase; LDHA, lactate dehydrogenase A. (C) Western blot to determine LDHA protein expression. GAPDH was used as a loading control. (D) Relative level of intracellular ROS. Results are normalized to control group (n = 3 independent experiments). (E) C4–2/Ctrl and C4–2/ATM-KO cells were transfected with siRNAs targeting either a scramble sequence (siCtrl) or LDHA (siLDHA). Western blot confirmed knockdown of LDHA expression by using siRNAs. GAPDH was used as a loading control. (F) Intracellular ROS production was detected with DCFDA fluorescence and monitored by flow cytometry at 24 hours post-transfection with siCtrl or siLDHA. Left, median fluorescent intensity. Right, the representative image of histograms of ROS (n = 3 independent experiments). (G) Cell viability of C4–2/Ctrl and C4–2/ATM-KO cells transfected with siCtrl and siLDHA was determined by MTS assay at the indicated time points (n = 3 replicates). All data are depicted as mean ± s.d. * p < 0.05 and ** p < 0.01 by two-tailed Student’s t-test.

The upregulation of LDHA by ATM deficiency facilitated the conversion of glucose-derived pyruvate to lactate (Fig.4A), which is also expected to decrease the entry of pyruvate to mitochondria for OXPHOS and thereby decrease the production of mitochondrial ROS. We therefore determined the ROS level in C4–2/Ctrl and C4–2/ATM-KO cells by DCFDA (5-(and-6)-carboxy-2’, 7’-dichlorodihydrofluorescein diacetate) fluorescence flow cytometry. As expected, we found that C4–2/ATM-KO cells produced much less ROS than C4–2/Ctrl cells (Fig.4D), indicating that the cellular oxidative stress is much less in C4–2/ATM-KO cells and consistent with its increased proliferation capacity. Reduction of LDHA is then expected to divert the pyruvate to mitochondria, increase the production of ROS and thereby induce the cell death (Le, et al. 2010). We used siRNA to knock down LDHA expression in both C4–2/Ctrl and C4–2/ATM-KO cells as well as CWRR1/Ctrl and CWRR1/ATM-KO cells (Fig.4E and Supplementary Fig.3A) and found that reduction of LDHA by siRNA significantly induced more ROS in C4–2/ATM-KO and CWRR1/ATM-KO cells than C4–2/Ctrl and CWRR1/Ctrl cells (Fig.4F and Supplementary Fig.3B) and as a consequence remarkably inhibited cell growth of C4–2/ATM-KO and CWRR1/ATM-KO cells but only caused a marginal inhibition on C4–2/Ctrl and CWRR1/Ctrl cell growth (Fig.4G and Supplementary Fig.3C).

LDHA inhibitor FX11 suppresses ATM-deficient CRPC cells growth

Since inhibition of LDHA by siRNA suppressed the C4–2/ATM-KO cell growth by inducing oxidative stress, we next sought to determine whether LDHA inhibitor has the same inhibitory effect on ATM-deficient CRPC cells that may document a new therapeutic target for this subset of CRPC patients. We selected a small-molecule FX11, which has been shown to preferentially inhibit LDHA and suppress tumor progression (Le et al. 2010). First, we examined the IC₅₀ of FX11 on the two C4–2 cell lines and found that C4–2/ATM-KO cells were more sensitive to the inhibitor than C4–2/Ctrl cells (IC₅₀ = 2.52 μM for C4–2/ATM-KO versus IC₅₀ = 13.87 μM for C4–2/Ctrl) (Fig.5A). Then, we examined the lactate production after cells being treated with FX11. An obvious reduction of lactate secretion was observed for both C4–2/Ctrl and C4–2/ATM-KO cells but FX11 treatment resulted in a further significant reduction of lactate production in C4–2/ATM-KO cells (Fig.5B), manifesting the efficacy of this LDHA inhibitor on ATM-deficient CRPC cells. To determine the effect of FX11 on cell biological functions, we measured the ROS level and cell growth after cells being treated with FX11. Similar to the effect of siLDHA, FX11 treatment led to more ROS production in C4–2/ATM-KO cells and consequently induced more cell death on C4–2/ATM-KO cells (Fig.5 C&D). Additionally, we employed FX11 in CWRR1 cells as well and found the similar results shown in C4–2 cells (Supplementary Fig.4). To confirm that FX11 also has the efficacy on the ATM-deficient xenograft tumors in vivo, we established the C4–2 xenograft tumors in SCID/NSG mice and intraperitoneally injected either FX11 or vehicle every other day after tumors reached the size of 200 mm³. As shown in Fig.5E & F, FX11 profoundly inhibited tumor growth of C4–2/ATM-KO xenografts but had a marginal insignificant inhibition on C4–2/Ctrl xenograft tumors, in a similar manner to the in vitro results (Fig.5D).

Open in a separate window

Figure 5

LDHA inhibitor FX11 suppresses ATM-deficient CRPC cell growth. (A) IC₅₀ values of FX11 were measured on C4–2/Ctrl and C4–2/ATM-KO cells treated with a series concentration of FX11 for 72 hours. (B) Lactate production of C4–2/Ctrl and C4–2/ATM-KO cells treated with either DMSO or FX11 for 72 hours. Results are normalized to DMSO group and set as “1”. (C) Intracellular ROS production was detected with DCFDA fluorescence and monitored by flow cytometry at 72 hours post-treatment with either DMSO or 3 μM FX11. Left, median fluorescent intensity. Right, the representative image of histograms of ROS (n = 3 independent experiments). (D) Cell viability of C4–2/Ctrl and C4–2/ATM-KO cells treated with either DMSO or FX11 was determined by MTS assay at the indicated time points (n = 3 replicates). (E & F) Tumor growth of C4–2/Ctrl and C4–2/ATM-KO xenografts treated intraperitoneally with either DMSO or FX11 when tumor reached ~200mm³. Tumor volumes were measured every four days. Top, Representative tumor images in each group. Bottom, The growth curve of xenograft tumors after treatment of vehicle (DMSO) or FX11. All data are depicted as mean ± s.d. * p < 0.05 and ** p < 0.01 by two-tailed Student’s t-test.

Colony formation assay

1×10⁴ Cells were seeded into 6-cm dishes and grown for up to 14 days for colony formation. Cell culture media was refreshed every three days and colonies were fixed with cold methanol and then stained with 1% crystal violet. The number of colonies was counted and imaged with a VersaDoc Imaging System (BioRad).

CRISPR/Cas9-mediated ATM knockout

To establish ATM knockout stable cell line, we used the CRIPR/Cas9 technology. Single guided RNA (sgRNA) sequences were generated by CRISPR design tool (crispr.mit.edu). sgRNA sequences for ATM are available upon request. Annealed double stranded sgRNA oligos were ligated to the lentiCRISPR vector 2 (deposited by Dr. Feng Zhang to Addgene, Camridge, MA) at a ribonucleoprotein complex which expressed both Cas9 and sgRNA. To produce infectious transgenic lentivirus, the transfer plasmid was transfected into 293T cells together with packaging plasmid and envelope plasmid. After medium changed and a brief incubation period, supernatant containing the virus was collected and centrifuged to concentrate virus. Subsequently, cells were selected by puromycin for more than 14 days. Population that showed no target protein expression was confirmed by western blot.

RNA interference

siRNAs targeting human LDHA (ON-TARGETplus SMARTpool: L-008201–00-0005) and ON-TARGETplus Nontargeting pool were purchased from Dharmacon. siRNAs targeting human c-Myc (sc-29226) were purchased from Santa Cruz. 10⁵/well C4–2/Ctrl (CWRR1/Ctrl) or C4–2/ATM-KO (CWRR1/ATM-KO) cells were seeded in 6-well plate. After 24 hours, 30 pmol siLDHA/sic-Myc or siCtrl was transfected into cells using Lipofectamine RNAiMAX (Invitrogen) in Opti-MEM reduced serum medium (Gibco) for 24h.

Cell viability assays

The MTS assay (Biovision) was employed to determine cell viability. 1000/well cells were seeded in 96-well plate. 10 μl /well MTS Reagent was added into each well and incubated for 1 hour at each time point. Absorbance was measured at 490nm using SpectraMax M3 reader with SoftMax Pro 6 software for data acquisition and analysis. For cells treated with siRNAs, 5 pmol siLDHA or siCtrl was transfected into each well and cell viability was measured at each time point. For cells treated with FX11 (Millopore), 3 μM FX11 was added in culture medium followed by cell viability measurement. For IC₅₀ determination of FX11, cells were treated by a series of concentration for 72 hours. IC₅₀ value was calculated by GraphPad Prism software.

Animal study

The animal experiment protocol (A092–16-04) was reviewed and approved by the Duke Institutional Animal Care and Use Committee. Immunocompromised NSG mice were purchased from Jackson Laboratories. 10⁷ C4–2/Ctrl or C4–2/ATM-KO cells were resuspended in 100 μl of saline with 50% Matrigel (Corning) and injected subcutaneously into both sides of flank regions of 4–6 weeks old male NSG mice. When tumor volume reached a size of 100 mm³ which was set as Day 0, tumors were measured twice a week and tumor volumes were determined by caliper measurement of tumor length (L) and width (W) according to the formula (LxW²)/2. For FX11 in vivo experiment, when tumor volume reached 200 mm³, C4–2/Ctrl and C4–2/ATM-KO grafted mice were divided into two groups randomly. Each group of mice were injected with either control 2% (vol/vol) DMSO or 50 μg of FX11 intraperitoneally every other day. Tumor volumes were measured every four days.

Immunohistochemical staining and Quick-score system

Immunohistochemical staining (IHC) was performed as described in previous study (Park, et al. 2017). Xenograft tumor sections were deparaffinized and rehydrated. Endogenous peroxidase activity was blocked with 3% hydrogen peroxide in methanol. Heat-induced antigen retrieval was carried out for all sections in 0.01 M citrate buffer, pH 6.0, using a steamer at 95^oC. Primary antibodies were diluted with PBS containing 1% BSA to an optimal concentration and applied to the sections overnight. Slides were then incubated with rabbit anti-mouse or goat anti-rabbit IgG/HRP second antibody for 45 minutes. Diaminobenzidine (DAB) was then applied for 10 minutes. The sections were counterstained with hematoxylin, dehydrated, coverslipped and visualized. Primary antibodies used in this study are ant-CD31 (Abcam, Ab28364, 1:100), anti-Ki67 (ThermoFisher, RM-9106, 1:300) and anti-Snail (Abcam, Ab85936, 1:600). Quick-score was calculated as reported (Park et al. 2017), the intensity of staining is multiplied by the percentage of staining to derive a composite score (a range from 0 to 300).

Metabolite extraction and mass spectrometry

As described previously(Pan, et al. 2016), 10⁵ /well cells were plated in 6-well plate. 24 hours later, medium was replaced freshly as equal volume (2 ml) per well and cells were incubated for another 24 hours. Next, medium was aspirated and 80% (vol/vol) HPLC grade methanol/water was added into each well. Plates were placed on dry ice immediately and then transferred in a −80°C freezer for 15 minutes for further enzyme inactivation. Cells were then scraped into extraction solvent. After centrifuge for 10 minutes, supernatant was transferred into a new Eppendorf tube. A speed vacuum was used to dry the samples at room temperate then the dry pellets were sent to LOCASALE LAB (Duke University) for further liquid chromatography mass spectrometry analysis (LC-MS). Data was normalized to cell numbers.

¹³C isotope tracing assay

10⁵ /well cells were seeded in 6-well plate in 2 ml complete medium. After overnight incubation, medium was removed, and replaced by exact 2 ml fresh glucose-deprived media (Gibco 11966) containing 25 mmol/L uniformly ¹³C-labeled glucose isotope (Cambridge). After 24 hours incubation, metabolites were extracted and sent for further LC-MS analysis.

Glucose consumption assay

Glucose concentration in culture medium was determined by Glucose Colorimetric/Fluorometric Assay Kit (Biovision) following the manufacturer’s protocol. Glucose consumption was calculated as measured value subtracted by original concentration.

Lactate secretion assay

The lactate level in the medium was analyzed by Lactate Colorimetric/Fluorometric Assay Kit (Biovision). 10 μl of sample was added into 40 μl Lactate Assay Buffer, and then 50 μl Reaction Mix was added into each diluted sample. The reaction was incubated for 30 minutes at room temperature while being protected from light. Absorbance at OD_{570 nm} was measured by SpectraMax M3 reader. All assays were performed in triplicate.

RNA extraction and quantitative real-time PCR

Total RNA was extracted using RNeasy mini kits (Qiagen, Valencia, CA). cDNA synthesis was performed by PrimeScript RT Master Mix (Takara). Quantitative real-time PCR (qRT-PCR) was applied on ABI 7500 (Applied Biosystems, Foster City, CA) using SYBR Green PCR Master Mix (Quanta Biosciences). Actin expression assay was used as an internal control. All qPCRs were determined in triplicate. Sequences of primers use in study are available upon request.

Immunoblotting

Total protein was collected by lysing adherent cells with RIPA buffer (Sigma) supplemented with phosphatase and protease inhibitor cocktail (Thermo Fisher). For nuclear protein extraction, nuclear lysates were isolated with an NE-PER Nuclear and Cytoplasmic Extraction Reagents Kit (Thermo) with phosphatase and protease inhibitor cocktail contained. Protein concentration was measured using the Bradford reagent (Bio-rad). Equal amount of protein was loaded for immunoblotting. After electrophoresis, the protein was transferred to polyvinylidene difuoride transfer membrane (PVDF) followed by 5% non-fat milk blocking buffer and incubated overnight with primary antibodies. After washing with TBST (TBS with 0.1% Tween), anti-mouse or anti-rabbit second antibodies conjugated with horseradish peroxidase-conjugated (HRP) were used to probe the membranes. Samples were developed by Chemiluminescent Substrate (Thermo Fisher) and exposed by Odyssey Imaging Systems (LI-COR). Densitometry was performed using ImageJ software (National Institutes of Health) and samples were normalized to internal controls. Primary antibodies used in this study are anti-ATM (Novus, NB100–104), anti-p84 (Thermo Fisher, PA4–27816), anti-LDHA (Cell Signaling, 2012S), anti-c-Myc (Abcam, 32072) and anti-GAPDH (Abcam, Ab9484).

ROS measurement

As previously described(Le et al. 2010), intracellular ROS production was determined by staining cells with 10 μM 5-(and-6)-carboxy-2’,7’-dichlorodihydrofluorescein diacetate (carboxy-H2DCFDA, Thermo Fisher) according to the manufacturer’s instructions. 10⁵ cells/well (6-well plate) were treated with either 3 μM FX11 for 72 hours or 30 pmol siLDHA for 24 hours. Stained cells were analyzed by FACScan flow cytometers at excitation and emission wavelengths of 485 and 520 nm, respectively.

We thank Drs. Yanjing Li, Yu Yin and Yan Chang for their technical help; Dr. Yen-Nien Liu for his comments. We thank Drs. Juan Liu and Jason Locasale for their expertise in LC-MS analysis and tracing.

Funding

This study was supported by NIH grants R01 CA172603 (J. Huang), R01 CA205001 (J. Huang) and National Natural Science Foundation of China (#81630019).