Sage Journals: Discover world-class research

Abstract

Both glycolysis and mitochondrial energetics are targets of interest for developing antiproliferative cancer therapeutics. We developed a novel multiple-read assay based on long-term expression in K562 cells of a genetically encoded intramolecular Förster resonance energy transfer sensor for adenosine triphosphate (ATP). The assay, conducted in a fluorescent plate reader, can identify compounds that inhibit oxidative phosphorylation-dependent ATP production, glycolysis, or both after short-term treatment. We screened a National Cancer Institute (NCI) compound library, identifying inhibitors of oxidative phosphorylation-dependent ATP production and glycolysis. Three glycolysis inhibitors blocked hexokinase activity, demonstrating that our assay can serve as the initial step in a workflow to identify compounds that inhibit glycolysis via a defined desired mechanism. Finally, upon reviewing the literature, we found surprisingly little evidence that inhibiting glycolysis with small molecules is antiproliferative. Using NCI data on proliferation of K562 cells, we found that inhibitors of oxidative phosphorylation-dependent ATP production were no more antiproliferative than the overall library, whereas all glycolysis inhibitors were in the top third of most effective antiproliferative compounds. Our results thus present a powerful new way to screen for compounds that affect cellular metabolism and also provide important support for the idea that blocking glycosis is antiproliferative.

Keywords

metabolism glycolysis oxidative phosphorylation cancer therapeutics cell proliferation fluorescence resonance energy transfer (FRET)

Introduction

Attention has recently returned to cancer cell metabolism as a target for development of small-molecule antiproliferatives.¹ Many cancer cells obtain most of their adenosine triphosphate (ATP) from glycolysis even when oxygen is available, a phenomenon known as the Warburg effect. There is considerable interest in this phenomenon as a potential target for antiproliferative small-molecule therapeutic development.^1–4 At the same time, there is significant interest in targeting oxidative phosphorylation, either alone or as an adjunct to glycolysis inhibition.^5,6

While in vitro assays using purified glycolytic enzymes or transporters can be performed to identify novel inhibitors (e.g., ref. 7), a powerful alternate strategy that has emerged recently to screen for inhibitors of glucose transporter (GLUT) transporters uses cell-based ATP measurements as a primary screen.^8,9 In these efforts, a combinatorial approach is taken: oxidative phosphorylation is blocked with inhibitors, and compounds that suppress remaining ATP production are considered to be putative GLUT inhibitors and tested for this activity. Luciferase, which is most commonly used as a viability/proliferation measure, has been used to measure ATP levels in these efforts. The main advantage of cell-based approaches to glycolysis inhibitor discovery is that compounds identified are essentially guaranteed to be active in the cellular context. This is not the case for compounds discovered using cell-free biochemical approaches, which must be assessed in follow-up experiments for cellular efficacy.

The past several decades have seen an explosion in the use of genetically encoded sensors based on Förster resonance energy transfer (FRET) in basic cell biology research (see ref. 10 for a recent review). These sensors have seen relatively little use in screening, likely because there is a misperception that the signals they generate are too small to be useful. Imamura et al.¹¹ developed a sensor for intracellular ATP levels based on cyan fluorescent protein (CFP) and yellow fluorescent protein (YFP) fused at the N and C termini, respectively, of the epsilon subunit of a Bacillus subtilis F₀F₁-ATP synthase that reports ATP levels in the physiological range. We expressed ATeam in long-term K562 transfectants and used the cells to develop a novel 96-well multiple-read assay that can identify compounds that after short-term treatment inhibit oxidative phosphorylation-dependent ATP production, glycolysis, or both. We tested the effects of glycolysis inhibitors on hexokinase (HK) activity, identifying three novel inhibitors, which demonstrates that our assay can serve as the initial step in a screening campaign to identify compounds that inhibit specific glycolytic enzymes or transporters of interest. We believe that our assay will be of use to researchers who want to find either glycolysis inhibitors or inhibitors of oxidative phosphorylation and wish to exclude the other activity during the initial screen or, alternatively, researchers who want to find compounds that affect metabolism by either mechanism. Finally, we used National Cancer Institute (NCI) data to examine effects of metabolically active compounds on cellular proliferation. We found that inhibitors of oxidative phosphorylation-dependent ATP production were no more antiproliferative than the overall library, whereas all glycolysis inhibitors were in the top third of most effective antiproliferative compounds. This provides important support that is currently lacking for the idea that glycolysis inhibition is an effective means of blocking proliferation.

Materials and Methods

Cells, Constructs, and Reagents

K562 cells were obtained from ATCC (Manassas, VA) and cultured in Iscove’s medium containing 10% newborn calf serum, 1% penicillin-streptavidin, and 1% L-glutamine. Additional cell lines tested in Supplemental Figure 1 were provided by Drs. Charles Giardina and Kenneth Campellone of the University of Connecticut. CDNAs encoding the ATeam FRET sensor¹¹ and the FLII¹²pGLU-600µdelta6 glucose sensor¹² were obtained from Addgene (Cambridge, MA). Cells were transfected using nucleofection and sorted based on YFP fluorescence as described previously.¹³ Flow cytometric measurement of FRET signals were described previously.¹³ Experiments with ATeam-expressing cells were conducted in Ringer’s solution with or without added glucose as indicated. Glucose-free Ringer’s contained (in mM) 145 NaCl, 4.5 KCl, 1 MgCl₂, 2 CaCl₂, and 5 HEPES, with pH adjusted to 7.4 with NaOH. Where indicated, bovine serum albumin (BSA; Sigma, St. Louis, MO) was added at a final concentration of 0.12% to simulate the protein level in cell culture medium. Sodium azide and glucose were from Fisher (Fair Lawn, NJ). The Cell Titer-Glo 2.0 kit used for luciferase confirmation of effects on intracellular ATP levels was purchased from Promega (Madison, WI). The colorimetic HK assay kit was purchased from BioVision (Milpitas, CA). Both kits were used according to the manufacturers’ recommended protocols. Mechanistic Set III and powder resupply of hit compounds was obtained from the NCI’s Developmental Therapeutics Program (Bethesda, MD).

FRET Measurements and Screening

A Molecular Devices SpectraMax i3x plate reader (Sunnyvale, CA) was used to measure the fluorescence of cells expressing ATeam and FLII¹²pGLU-600µ in three channels: CFP (450/9 nm excitation, 485/15 nm emission), FRET (450/9 nm excitation, 535/15 nm emission), and YFP (485/9 nm excitation, 535/15 nm emission). For screening with ATeam, cells were washed in glucose-free Ringer’s and added to flat-bottomed Greiner half-area 96-well plates, with ~100,000 cells in 100 µL per well. One microliter of DMSO or 1 µL test compounds at 1 mM was added to wells, producing a final compound concentration of 10 µM. After addition of cells, plates were vortexed and incubated for 30 min at 37 °C in a humidified CO₂ incubator.

The following equation was used to calculate FRET ratios

FRET ratio = \frac{FRETc - FRETb}{CFPc - CFPb},

where FRETc and CFPc are the fluorescence intensities of cells and FRETb and CFPb are the background fluorescence intensities of wells containing 100 µL Ringer’s solution.

The FRET ratio was determined from cells treated with glucose but not Na azide (P), treated with compounds (C), or treated with Na azide but not glucose (N), and percentage inhibition is defined as

100 * \frac{P - C}{P - N} .

Percentage inhibition of oxidative phosphorylation-dependent ATP level was determined from read 1, whereas percentage inhibition of glycolysis-dependent ATP level was determined from read 3.

Generating Responses Representing Four Compound Classes

No effect on ATP production. Cells were washed with 0 glucose solution and then treated with sodium azide, then glucose.

Inhibition of oxidative phosphorylation-dependent ATP production. Cells were treated with 5 mM oligomycin, then sodium azide, then glucose.

Glycolysis inhibition. Cells were washed with 0 glucose solution, then treated with sodium azide.

Inhibition of both glycolysis and oxidative phosphorylation-dependent ATP production. Cells were treated with 5 mM oligomycin, then sodium azide.

Hit Confirmation by Cell-Titer Glo

Cells were washed in glucose-free Ringer’s and incubated with DMSO or test compounds for 30 min at 37 °C. Fifty microliters of Titer Glo reagent was added to 50 µL of cells (10,000 cells per well), mixed, and incubated for 10 min. Luminescence was measured on the Spectramax i3x plate reader.

HK Activity Assay

K562 cells (50,000 cells per condition) were lysed with supplied lysis buffer, and 25 µL of cell lysate was incubated with DMSO or test compounds for 10 min at room temperature. Twenty-five microliters of reaction mix (freshly prepared according to the manufacturer’s protocol) was added, and absorbance at 450 nm was measured on the plate reader in kinetic mode. A standard curve of NADH was used to convert deltaOD₄₅₀ to HK activity.

Dose-Response Measurements

Compounds that were confirmed active in a repeat of the primary assay were resupplied from NCI in powder and dissolved in DMSO. Concentrations between 0.1 and 10 µM were tested at twofold dilutions on cells. Percentage inhibition data were fitted to the Hill equation as described previously.

Statistical Analysis

Effects of compounds on proliferation were analyzed using nonparametric analysis of variance (ANOVA; Kruskal-Wallis test). Data were not paired, and Dunn’s correction for multiple comparisons was applied. Prism 7 software (Graphpad Software, Inc., La Jolla, CA) was used to perform the test.

Results

Assessing the Metabolic Phenotype of K562 Cells Using a Genetically Encoded FRET Sensor for ATP

We generated long-term transfectant K562 cells expressing ATeam, a genetically encoded FRET sensor for ATP,¹¹ by transfecting cells and then sorting them based on YFP expression ~2 wk and ~4 wk posttransfection. Although we did not clone cells, this procedure gives rise to cells that express the sensor in a temporally stable fashion. Figure 1 shows a characterization of the long-term transfectants using flow cytometry. ATeam is expressed at high levels, as demonstrated by the YFP fluorescence histogram ( Fig. 1B ). To assess signal amplitude, we compared two conditions. Cells were incubated in the presence of 20 mM glucose to generate a high-ATP signal ( Fig. 1D ) or washed in glucose-free solution and treated with Na-azide to deplete ATP and generate a low-ATP signal ( Fig. 1C ). FRET ratios were higher in cells when glucose was present, and the capacity for oxidative phosphorylation was more intact than in the absence of glucose and the presence of azide.

Figure 1.

Förster resonance energy transfer (FRET) measurements with a genetically-encoded adenosine triphosphate (ATP) sensor in K562 cells using flow cytometry. (A) Plot of forward versus side scatter for ATeam-expressing K562 cells. The live-cell gate is indicated. (B) Histogram of yellow fluorescent protein (YFP) fluorescence for long-term transfectants expressing ATeam sensor. The blue trace represents the YFP fluorescence of untransfected K562 cells. (C, D) Plots of FRET ratio versus YFP for cells washed in glucose-free Ringer’s solution and treated with sodium azide (C) or in the presence of glucose (D). FRET ratio was calculated as described in the Materials and Methods section.

Although we had previously used flow cytometry to measure intramolecular FRET signals for screening,¹³ expression levels of those sensors were not particularly high, and the FRET ratio changes were fairly small. The high levels of expression in the ATeam cells together with the sensor’s relatively large dynamic ratio encouraged us to explore whether signals could be measured using a standard fluorescence plate reader. We found that using ~10⁵ cells/well, we could obtain signals in both the FRET and CFP channels that were ~10-fold higher than background ( Fig. 2Ai and ii ). Although there was variation in the signal level between wells that likely reflects variation in the number of cells, when we computed the FRET ratio (FRET/CFP; Fig. 2Aiii ), variability decreased markedly, as expected.

Figure 2.

Measuring ATeam Förster resonance energy transfer (FRET) using a plate reader. (A) Plots of cyan fluorescent protein (CFP) fluorescence (i), FRET fluorescence (ii), and FRET ratio (iii) for ATeam cells in a 96-well plate treated with indicated conditions. (B) K562 cells can generate adenosine triphosphate (ATP) in the absence of glucose or functional mitochondria but not both. (i) Average FRET ratios for cells from (A). Data are the mean ± SD from 24 wells per condition. Similar results were obtained in multiple independent experiments. See also Supplementary Figure S1 . (ii) Time dependence of changes in ATeam FRET ratio after cells were washed in glucose-free Ringer’s solution (filled circles), washed in glucose-free solution and treated with sodium azide (open squares), treated with sodium azide and glucose (filled squares), or glucose (open circles). Sodium azide and glucose were used at final concentrations of 5 mM and 20 mM, respectively.

We used the ATeam-expressing cells and the plate reader to characterize the metabolic activity of K562 cells ( Fig. 2Bi ). As was seen using flow cytometry, ATP levels were high in cells when glucose was present and oxidative phosphorylation was not inhibited. ATP levels remained high if oxidative phosphorylation was inhibited by adding azide, provided glucose was present. ATP levels fell only when azide was added and glucose was removed. As azide and glucose removal might be seen by some as nonspecific means of altering metabolism, we tested the effects of using oligomycin to inhibit oxidative phosphorylation and 2-deoxyglucose (2-DG) to inhibit glycolysis. For these experiments, extracellular glucose was reduced to 5 mM, as 2-DG competes with glucose for entry via GLUT transporters. Identical results to those observed with azide and glucose-free solution were obtained using oligomycin in combination with 2-DG ( Suppl. Fig. S1 ). Overall, these results suggest that aerobic glycolysis can occur in K562 cells, because in the presence of glucose influx, cells can maintain high ATP even when oxidative phosphorylation is blocked. However, cells were also able to maintain high ATP levels in the absence of both glucose and glutamine, provided the capacity for oxidative phosphorylation was not compromised by drug addition. This indicates that K562 cells, despite being able to generate high levels of ATP via aerobic glycolysis, can also use stored substrate(s) to maintain ATP levels. We used luciferase to measure ATP to confirm these results, yielding essentially identical results in K562 cells and similar results in several other cell lines we tested ( Suppl. Fig. S2 ). Using the long-term ATeam transfectants, we found that K562 cells are able to maintain cellular ATP levels using stored substrates and oxidative phosphorylation for ~3 h, after which ATP declined slowly over the course of an additional ~5 h period ( Fig. 2Bii ).

Identifying Metabolically Active Compounds in the NCI Mechanistic Set

We designed a three-step assay protocol based on the results of Figure 2 and the previous work of Ulanovskaya et al.⁸ and Sibeneicher et al.⁹ to identify compounds that inhibit glycolysis, oxidative-phosphorylation dependent ATP production, or both ( Fig. 3 ). We first washed cells in glucose-free solution and measure FRET ratio (read 1) to identify compounds that inhibit glucose-independent oxidative phosphorylation-dependent ATP production. ATP is high under these conditions unless oxidative phosphorylation is inhibited. We then added azide to block respiratory complex IV and deplete ATP in all samples. Plates are read again (read 2) after ~20 min to reveal compounds that affect FRET ratios via fluorescence artifacts and confirm azide efficacy. Finally, 20 mM glucose is added to initiate glycolysis, and plates are read again after an additional 20 min (read 3). Glycolysis inhibitors will prevent ATP from increasing after glucose addition. Note that using this protocol, we would not be able to identify compounds that inhibit oxidative phosphorylation-dependent ATP production or glycolysis with a delayed mechanism of action. Figure 3A shows a test plate treated using this protocol. Reponses were robust, and there was excellent separation between the signals, consistent with the hypothesis that an assay based on this protocol can be used for screening 96-well plates. When assessed using standard assay performance metrics,¹⁴ we routinely obtained values Z′ > 0.5. For the plate shown, the Z′ calculated from 80 wells comparing incubation in glucose-free saline to incubation in glucose-free saline + azide was 0.79. As described below, these conditions correspond to what is expected for inhibitors of oxidative phosphorylation-dependent ATP production. The Z′ calculated from 80 wells comparing incubation in glucose-free saline + azide to 20 mM glucose + azide was 0.82. As described below, these conditions correspond to what is expected for inhibitors of glycolysis.

Figure 3.

A multiple-read protocol enables detection of inhibitors of both glycolysis and oxidative phosphorylation-dependent adenosine triphosphate (ATP) production. (A) Design and performance of an assay with three reads. ATeam cells were washed in glucose-free solution and put into a 96-well plate. After 30 min of incubation, fluorescence signals were measured and FRET ratio was calculated (1). Sodium azide was added to all but eight control wells in column 12 and incubated for 30 min, before the second measurement was taken (2). Glucose was added to all but eight control wells in column 1 and incubated for 30 min, before the third measurement was taken (3). Sodium azide and glucose were used at final concentrations of 5 mM and 20 mM, respectively. (B) Four potential compound effects and the ATeam response produced in the multiple read protocol. 1, no effect. 2, inhibitor of oxidative phosphorylation-dependent ATP production. 3, glycolysis inhibitor. 4, dual inhibitor. See the Materials and Methods section for details. (C) Heat maps of Förster resonance energy transfer (FRET) ratios of ATeam-expressing cells incubated in a representative plate that contained compounds from the National Cancer Institute (NCI) Mechanistic Set III in columns 2 to 11. Columns 1 and 12 contained control wells. The scaling of heat maps is displayed to the right. One well that contained a fluorescent compound is represented in white. (D) Screening 813 compounds from NCI Mechanistic Set III. Data are presented as percentage inhibition of FRET ratio compared with control wells in each plate. Inhibition of oxidative phosphorylation-dependent ATP production and glycolysis are obtained from readouts 1 and 3, respectively.

We envisioned that compounds with four kinds of effect could be identified using this protocol ( Fig. 3B ). Most compounds would have no effect on either oxidative-phosphorylation or glycolysis ( Fig. 3B1 ). These would display high FRET signals in the first read, low FRET signals in the second read, and high FRET signals in the third read. For compounds that inhibit oxidative phosphorylation-dependent ATP production, FRET ratios would be low in the first and second reads and high in read 3 ( Fig. 3B2 ). Compounds that inhibit glycolysis would have high FRET ratios in read 1 and low FRET ratios in reads 2 and 3 ( Fig. 3B3 ). Finally, compounds that inhibit both oxidative phosphorylation-dependent ATP production and glycolysis would display low FRET ratios in all three reads ( Fig. 3B4 ), as might compounds that, via their fluorescent properties, caused an artifactual decrease in FRET ratio.

We used the protocol we developed to screen the 813 compounds in the NCI’s Mechanistic Set III in 96-well plates at 10 µM final concentration ( Fig. 3C , D ). We excluded wells in which fluorescence in either the CFP or FRET channels was more than twofold higher than in the overall plate average as a means of avoiding fluorescent artifacts that compromised FRET measurements. After initial screening, we confirmed the activity of hits by repeating the assay with samples of compounds from the original provided plates and then obtained new powder samples of the compounds, which we also tested using the FRET assay. We then confirmed the activity of powder-resupplied hits using luciferase to measure ATP as an orthogonal assay. For luciferase experiments, we obviously had to test effects on oxidative phosphorylation-dependent ATP production and glycolysis separately, as measuring ATP required cell lysis. Ultimately, we identified 14 compounds that were confirmed in saline solution to inhibit oxidative phosphorylation-dependent ATP production by >75% and 13 compounds that inhibited glycolysis by >80% (see Table 1 and structures in Suppl. Fig. S3 ). Dose-response measurements on confirmed active compounds were performed to assess potency ( Table 1 ).

Table 1.

Metabolically Active Compounds in NCI Mechanistic Set III.^a

	NSC Number	Inhibition of Glycolytic ATP Production (%)^b	Inhibition of Nonglycolytic ATP Production (%)^c	IC₅₀ (µM)^d	Hill Slope^d	G₁₀^e
1	136037	111 ± 3; 113 ± 6	3 ± 2; 4 ± 5	1.5 ± 0.3	2.1 ± 0.8	−4.9
2	105808	107 ± 3; 110 ± 5	8 ± 5; 2 ± 3	3.3 ± 0.2	4.8 ± 1	−34.6
3	111041	110 ± 3; 97 ± 14	8 ± 3; 28 ± 6	4.4 ± 0.4	5.2 ± 2.8	−49.4
4	697923	99 ± 13; 86 ± 22	9 ± 9; 2 ± 2	>10 µM		−24.2
5	328587	108 ± 3; 101 ± 7	13 ± 8; 3 ± 3	2 ± 0.4	1.9 ± 0.5	6.3
6	400944^f	89 ± 11; –10 ± 2	20 ± 7; 3 ± 3	1.7 ± 0.3	5.4 ± 2.9	47.4
7	686349^f	90 ± 9; 90 ± 6	36 ± 24; 7 ± 5	4.2 ± 0.5	2.8 ± 0.8	−36.6
8	30916^f	106 ± 6; 73 ± 13	51 ± 20; 43 ± 11	4 ± 0.4	3.6 ± 1	−35.7
9	635448	98 ± 9; 43 ± 12	71 ± 16; 59 ± 10	5.6 ± 1	4.2 ± 3.1	−41.3
10	659997	118 ±4; 107 ± 6	87 ± 6; 45 ± 10	1.1 ± 0.2	2.2 ± 1	−19
11	631529	102 ± 4; 70 ± 13	99 ± 8; 25 ± 19	>5 µM		−38
12	168597	113 ± 3; 86 ± 11	120 ± 11; 112 ± 3	2.2 ± 0.4	1.4 ± 0.3	−2.2
13	164914	85 ± 8; 45 ± 8	124 ± 11; 113 ± 7	2.8 ± 0.4	1.5 ± 0.3	−54.3
14	377	−3 ± 4; –28 ± 4	110 ± 2; 100 ± 6	0.04 ± 0.02	1.98 ± 0.84	35.8
15	52141	−1 ± 6; –4 ± 3	109 ± 2; 93 ± 3	2.9 ± 2	1.6 ± 1.4	4.5
16	126771	−15 ± 5; –40 ± 8	109 ± 7; –14 ± 27	4.3 ± 0.2	5.6 ± 1.2	15.6
17	658285	−20 ± 8; –13 ± 19	106 ± 8; –14 ± 10	3.8 ± 0.7	3 ± 1.2	90.7
18	379531	3 ± 3; –6 ± 5	102 ± 2; 93 ± 3	0.2 ± 0	2.7 ± 1.2	56.3
19	79451	−5 ± 7; –23 ± 3	101 ± 2; 76 ± 3	1 ± 0.1	2.9 ± 0.6	55
20	667467	−14 ± 8; –10 ± 4	98 ± 7; 52 ± 6	6 ± 0.6	5.3 ± 2.1	72.2
21	5890	−6 ± 4; –11 ± 2	97 ± 3; 80 ± 13	1.4 ±0.2	2.4 ± 0.8	75.8
22	328477	−9 ± 3; –24 ± 1 0	94 ± 5; 70 ± 5	0.4 ± 0	4.6 ± 1.5	46.7
23	79456	−6 ± 8; –27 ± 7	90 ± 5; 65 ± 7	0.7 ± 0.1	10.4 ± 11.4	69.9
24	76455	−9 ± 7; –7 ± 3	81 ± 9; 83 ± 5	<0.1 µM		8.5
25	664329	−13 ±7; –8 ± 4	80 ± 12; 18 ± 4	6.9 ± 0.5	5.5 ± 1	96
26	104129	−2 ± 7; –12 ± 4	79 ± 10; 34 ± 9	5 ± 1.1	1.8 ± 0.5	60.2
27	82025	−9 ± 5; –27 ± 5	79 ± 9; –31 ± 10	>10 µM		67.9

Compounds are numbered in the same order as their structures are displayed in Supplementary Figure S2 . NCI, National Cancer Institute; ATP, adenosine triphosphate.

Measured from read 3, expressed as percentage control. Data are mean ± SEM from four measurements from powder resupply. The first value was obtained in the absence of bovine serum albumin and the second in its presence.

Measured from read 1, expressed as percentage control. Data are mean ± SEM from four measurements from powder resupply. The first value was obtained in the absence of bovine serum albumin and the second in its presence.

From fits to the Hill equation for data from three separate experiments including doses ranging from 0.1 to 10 µM. When dose-response curves could not be fit, a range of inhibitory concentrations is indicated. Performed with resupplied compounds.

Compiled from NCI data. In some cases, multiple reported values were averaged.

Hexokinase inhibitors.

Both limbs of our multiple-read assay detected effects on the output of an entire pathway. Compounds that inhibit oxidative phosphorylation-dependent ATP production could act via multiple mechanisms: they could inhibit respiratory complexes; depolarize the mitochondrial membrane potential; inhibit the import of energy-providing substrate(s), adenosine diphosphate, or phosphate into mitochondria; or could block export of ATP. Three of the inhibitors of oxidative phosphorylation-dependent ATP production that we found—nonactin (NSC 52141), usnic acid (NSC 5890), and peliomycin (NSC 76455)—are known inhibitors of oxidative phosphorylation,^15–17 confirming that that limb of the assay can identify compounds with desired mechanism of action. Glycolysis inhibitors could act on one or more of the enzymes or transporters involved in the glycolytic pathway. To our knowledge, none of the compounds we identified have been reported previously to inhibit glycolysis. To assess the mechanism of action of glycolysis inhibitors, we first tested effects on glucose influx ( Fig. 4A ), as K562 cells are known to express glucose transporter 1 (GLUT1) at high levels.¹⁸ We created a long-term K562 cell line expressing a FRET sensor for intracellular glucose¹² (see Suppl. Fig. S4 ) and used it to test whether any of the glycolysis inhibitors block increases in intracellular glucose when glucose is added to cells suspended in glucose-free extracellular solution. The known GLUT inhibitor cytochalasin b¹⁹ blocked FRET ratio increases significantly, but none of our glycolysis inhibitors did. We rescreened the entire library for glucose transport blockers and found one compound, NSC 96932, that we confirmed was active. Examination of the relationship between extracellular glucose or cytochalasin b concentration and intracellular glucose and cellular ATP measured via FRET indicates that glucose influx must be reduced to extremely low levels before cellular ATP is affected ( Suppl. Fig. S3B ), which is likely why this compound was not a hit in the initial ATP screen. NSC 96932 did block increases in ATP in the presence of 2 mM extracellular glucose (data not shown).

Figure 4.

Examining molecular mechanism of action for select glycolysis inhibitors. (A) Measuring effects of compounds on glucose influx. Long-term transfectant lines expressing a Förster resonance energy transfer (FRET) sensor for intracellular glucose were washed in glucose-free solution and incubated with indicated glycolysis inhibitors for 30 min. Glucose was added at a final concentration of 20 mM and plates incubated for an additional 30 min. Fluorescence signals were measured on the plate reader, and the FRET ratio was calculated. Dotted lines indicate control wells and their FRET ratios. See also Supplementary Figure S3 . (B) Measuring effects of compounds on hexokinase activity. (Top) K562 cell lysates were incubated with indicated glycolytic inhibitors in vitro for 10 min before the hexokinase colorimetric assay kit was used to detect hexokinase activity. The negative control received no lysate, and the positive control received cell lysate treated with DMSO. (Bottom) Hexokinase inhibitors were tested for their effects on G6PDH. 100 µM glucose-6-phosphate was added directly to the reaction mix.

We next tested effects on HK activity using a kit that monitors NADH produced from NAD by Aspergillus glucose-6-phosphate dehydrogenase (which is supplied) to measure the conversion of glucose to glucose-6-phosphate by HK activity in cell lysate samples that were treated with compounds or DMSO ( Fig. 4B ). Four of the eight compounds we selected for testing inhibited production of NADH when glucose was added to cell lysate samples, consistent with inhibition of HK activity. However, one compound, NSC 111041, also inhibited NADH production when glucose-6-phosphate was added directly to the reaction mix, indicating that it blocks Aspergillus glucose-6-phosphate dehydrogenase ( Fig. 4B ). Dose-response measurements indicate that the three HK inhibitors we identified exhibit micromolar potency in glucose-containing saline solution ( Suppl. Fig. S5 ). These results indicate that our cell-based assay can serve as the initial screening step for efforts to detect inhibitors of specific glycolytic enzymes.

Glycolysis-Inhibiting Compounds Are Effective at Blocking Proliferation

The effects of NCI library compounds on cell proliferation have been extensively characterized and are available in an online database (described in ref. 20, https://dtp.cancer.gov/databases_tools/default.htm). This allowed us to assess the effects of the hits we found on proliferation ( Fig. 5 ). We had in mind a simple idea: if inhibiting a particular metabolic process is an effective means of blocking proliferation, then all inhibitors of that process should be antiproliferative and thus the ensemble of inhibitors will be significantly more antiproliferative than the library as a whole. In contrast, if inhibiting a given process does not reduce proliferation, then inhibitors will only be as effective as the library because effects on proliferation will be no different than for compounds selected at random from the library. We downloaded the complete dose-response data set for growth measurements on the NCI 60 cell panel, which were assessed using a formula that accounts for cell proliferation during the 2-d course of the assay. Compounds that do not inhibit growth give a percentage growth of 100. Compounds whose actions result in the same number of cells after 2 d compared with at the start of the experiment give a percentage growth of 0. Finally, compounds that reduce the number of cells at the end of 2 d compared with the start of the experiment give negative percentage growth—a complete absence of cells after 2 d would correspond to −100% growth. We extracted the data obtained at 10 µM compound concentration (percentage growth at 10 µM, which we call G₁₀) for the effects of compounds in Mechanistic Set III on K562 cells. We chose this concentration because at 10 µM, all of the inhibitors of glycolysis-dependent and oxidative phosphorylation-dependent ATP production we identified were >75% effective ( Table 1 , see below). Data were available for 714 of the 813 compounds in the set, including all of the metabolically active compounds we identified. We plotted G₁₀ for the compound collection in histogram form ( Fig. 5A ) and then, as a means of assessing the relative fraction of each of these kinds of compounds, computed the cumulative density function for the plated compounds ( Fig. 5B ). The median G₁₀ in the library was 49.

Figure 5.

Assessing effects of metabolically active small molecules on K562 cell proliferation. (A) Histogram of percentage cell growth of National Cancer Institute Mechanistic Set III compounds tested at 10 µM (G₁₀) in K562 cells. (B) Cumulative distribution for compounds computed from (A). Dotted line represents the median activity. G₁₀ values for inhibitors of glycolysis and oxidative phosphorylation-dependent adenosine triphosphate production that were active in the presence of bovine serum albumin are indicated by circles and squares, respectively. The mean and standard deviation for each group are indicated.

When we initially examined the effects on proliferation of our hits, we found that inhibitors of oxidative phosphorylation-dependent ATP production had a wide range of effects ( Table 1 ). Only some reduced proliferation, which is inconsistent with inhibition of oxidative phosphorylation-dependent ATP production representing an effective means of blocking growth. In contrast, all of the glycolysis inhibitors except one, NCI 400944, were fairly effective at inhibiting proliferation. We believe that if even one glycolysis inhibitor that was effective in cells was to fail to block proliferation, it would cast doubt on the idea that inhibiting glycolysis inhibits proliferation. However, in attempting to find experimental reasons why discordant results could be obtained, we noted that the NCI’s experiments effects on proliferation were measured in growth medium that was supplemented with 5% serum, whereas our screening was conducted in serum-free saline (serum is sufficiently fluorescent that it increases background and compromises FRET measurements; not shown). Because serum proteins, primarily BSA, can bind some compounds and reduce their free concentration,²¹ we assessed the effects of compounds on ATP production in the presence of 0.12% BSA, which is close to the amount of BSA present in medium containing 5% serum. Including BSA increased background fluorescence but not so much as to prevent measurement of FRET signals. We found that several metabolically active compounds, including NCI 400944, were not effective in the presence of BSA. We excluded these compounds from our analysis of effects on proliferation. The seven oxidative phosphorylation inhibitors (squares in Fig. 5B ) that were found to inhibit by >70% in the presence of BSA had G₁₀ ranging from 4.5% to 75%, with a mean of 40% ± 26% (mean ± SD), not different from the overall compound set mean of 43 ± 51 (p = 0.98, as assessed using nonparametric one-way ANOVA). By contrast, the 10 compounds that inhibited glycolysis >70% in the presence of BSA (circles in Fig. 5B ) had G₁₀ values ranging from −54% to 6.3%, with a mean of −24% ± 18%, significantly different from the mean of the overall compound set (p < 0.0001). We assessed growth inhibitory effects of the metabolic inhibitors we identified on the five other leukemic lines in the NCI 60 cell line panel ( Suppl. Tables S1 and S2 ). Not all compounds were tested by the NCI in all cell lines, and we did not confirm cellular efficacy of the inhibitors in those lines, but the trend described above for K562 cells held for the other leukemic lines: inhibitors of oxidative phosphorylation were no more antiproliferative than the library as a whole, whereas glycolysis inhibitors were significantly more effective.

Discussion

Our multiple-read FRET sensor-based assay protocol offers the ability to detect multiple kinds of metabolically active compounds. This is a major advantage over the assays that others have used to identify glycolysis-inhibiting GLUT transporter blockers.^8,9 We note that in the original luciferase assays, multiple reads were impossible, as cells needed to be lysed to measure ATP. Subsequent work used stably expressed luciferase and caged luciferin.²² Multiple reads could be performed using that system. Identification of HK inhibitors confirms that with appropriate follow-up assays, our screen could serve as the first step in identifying inhibitors of specific glycolytic enzymes or transporters of interest. Follow-up assays could likely be devised to identify inhibitors of oxidative phosphorylation-dependent ATP production that affect particular steps of interest, although we tested only two of the possible targets. As our assay involves only solution additions, it seems likely that it could adapted to higher-density formats. It is surprising to us that so little use of genetically encoded FRET reporters is made in screening; provided expression levels are high enough and FRET ratio changes are large enough, our results with the ATP sensor and the glucose sensor confirm signals can be obtained in a plate reader that are more than adequate. We note that a multiple-read protocol was used by Allen et al.²³ in 2006 to measure activators and inhibitors of adenylyl cyclase using a FRET reporter measured in a plate reader.

Despite the interest in targeting metabolism as a means of inhibiting cancer cell proliferation, upon review, we found that there were weaknesses with some of the evidence supporting the idea. This is particularly true for glycolysis, which has received relatively more attention. Although it is abundantly clear that a number of glycolytic proteins are up-regulated in cancer cells,^24,25 and knockouts or knockdowns of glycolytic enzymes can inhibit cancer initiation and progression in animal models,^26,27 functions of glycolytic enzymes unrelated to glycolysis can confound results of genetic studies. For example, HK2 binds to the mitochondrial outer membrane and may regulate the permeability transition pore that releases cytochrome C during apoptosis,²⁸ pyruvate kinase M2 can act directly as a transcription factor,²⁹ and glycerlaldehyde-3-phosphate dehydrogenase regulates vesicle transport.³⁰ Knocking these proteins down could thus affect proliferation in ways unrelated to effects on glycolysis.

There are also issues with support from small-molecule inhibitors. Compounds such as 2-DG and 3-bromopyruvate⁴ were among the first to be explored, but each has effects apart from glycolysis inhibition that could inhibit growth. 2-DG inhibits n-linked glycosylation, which may be responsible for inhibiting proliferation under normoxic conditions,³¹ and 3-bromopyruvate is an alkylating agent³² that may have multiple additional affects. ENOBlock, originally reported to be an enolase 1 inhibitor,³³ appears not to block enolase and does not affect proliferation via glycolysis at all,³⁴ whereas STF-31, originally identified in a cytotoxicity screen and subsequently shown to block GLUT1, may affect proliferation via effects on nicontinamide phosphoribosyltransferase.³⁵ GlaxoSmithKline developed a series of HK2-inhibiting glucosamine derivatives, but effects on growth were not reported.⁷ Bayer developed a series of GLUT1 inhibitors,^9,22 but the concentrations of these agents required to inhibit proliferation are orders of magnitude higher than those required to block GLUT1.³⁶ Thus, the link between small-molecule inhibition of glycolysis and antiproliferative effects remains uncertain. This is reflective of a general problem encountered when using small molecules to validate potential drug targets. Compounds available prior to a large-scale screening and chemical optimization effort almost always have multiple effects, so the level of support they provide for a given target can in fact be quite weak.³⁷ In fact, even after optimization, establishing that effects on a particular target produce a phenotype of interest can be quite difficult. This is particularly problematic when the phenotype is, like reduced proliferation, a decrease in fitness (see ref. 38 for a thoughtful discussion of the issue).

Our results and the NCI’s database allow us to do something that we believe to be novel: assess the effects on proliferation of ensembles of compounds identified through a phenotypic screen designed to isolate specific kinds of actions—in this case, on metabolism—and compare them to the effects of the library as a whole. Critically, as our assay is cell based, we know that the compounds we selected for our analysis were active under the conditions in which proliferation was measured. This minimizes the chances that experimental considerations can skew the comparison. Our results show that glycolysis inhibitors are significantly more effective at blocking proliferation than the library as a whole, whereas inhibitors of oxidative phosphorylation-dependent ATP production are not. Thus, it seems unlikely that inhibiting oxidative phosphorylation-dependent ATP production per se is antiproliferative in K562 cells. If we assume that the antiproliferative properties of the library as a whole define the probability of a randomly selected compound having antiproliferative effects, we can use the binomial distribution to estimate the chances of selecting any number of compounds with a given antiproliferative efficacy from the library at random. For the least effective glycolysis inhibitor we found, NSC 328587, G₁₀ was 6, corresponding to a 94% inhibition of proliferation, a significant antiproliferative effect that we can then use to calculate the likelihood of obtaining compounds at least as antiproliferative from the library purely by chance. 189 of the 714 compounds (26.4%) for which G₁₀ values were available for K562 cells had G₁₀ values ≤6. According to the binomial distribution, the chances of selecting 10 out of 10 compounds at least this effective at random are 1.7 × 10⁻⁶, odds so small they suggest strongly that glycolysis inhibition is an effective means of blocking proliferation in K562 cells.

In summary, we present here a cell-based assay that could serve as the basis for screening efforts to identify compounds that inhibit glycolysis, oxidative phosphorylation, or both. We envision that the assay could be used in a workflow that involves identifying pathway inhibitors in the initial screen and then testing effects on specific mechanisms of action in a second stage. We believe that using such an assay system to expand the repertoire of available compounds could help provide tools that will aid in efforts to probe the links between cellular metabolism and proliferation.

Supplemental Material

Supplemental_material_for_Zhao_et_al._ATP_FRET_sensor – Supplemental material for A Novel Multiple-Read Screen for Metabolically Active Compounds Based on a Genetically Encoded FRET Sensor for ATP

Supplemental material, Supplemental_material_for_Zhao_et_al._ATP_FRET_sensor for A Novel Multiple-Read Screen for Metabolically Active Compounds Based on a Genetically Encoded FRET Sensor for ATP by Ziyan Zhao, Rahul Rajagopalan and Adam Zweifach in SLAS Discovery

Footnotes

Acknowledgements

We acknowledge the help of Dr. Wu He of the University of Connecticut’s COR²E Flow Cytometry Facility. We thank Dr. Ming-hui Chen of the Department of Statistics at the University of Connecticut for advice. We are particularly grateful to Dr. Daniel Zaharevitz of the National Cancer Institute for helpful advice on downloading data from the NCI’s database. Data sets will be provided upon reasonable request. ATeam-expressing K562 cells will be provided if the requestor pays for shipping.

Supplementary material is available online with this article.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: We thank the University of Connecticut for funding.

References

Amoedo

N. D.

Obre

Rossignol

Drug Discovery Strategies in the Field of Tumor Energy Metabolism: Limitations by Metabolic Flexibility and Metabolic Resistance to Chemotherapy. Biochim. Biophys. Acta 2017, 1858, 674–685.

Liberti

M. V.

Locasale

J. W.

The Warburg Effect: How Does It Benefit Cancer Cells?

Trends Biochemi. Sci. 2016, 41, 211–218.

Granchi

Fancelli

Minutolo

An Update on Therapeutic Opportunities Offered by Cancer Glycolytic Metabolism. Bioorg. Med. Chem. Lett. 2014, 24, 4915–4925.

Pelicano

Martin

D. S.

R. H.

et al . Glycolysis Inhibition for Anticancer Treatment. Oncogene 2006, 25, 4633–4646.

Zong

W.-X.

Rabinowitz

J. D.

White

Mitochondria and Cancer. Mol. Cell 2016, 61, 667–676.

Weinberg

S. E.

Chandel

N. S.

Targeting Mitochondria Metabolism for Cancer Therapy. Nat. Chem. Biol. 2015, 11, 9–15.

Lin

Zeng

Xie

et al . Discovery of a Novel 2, 6-Disubstituted Glucosamine Series of Potent and Selective Hexokinase 2 Inhibitors. ACS Med. Chem. Lett. 2016, 7, 217–222.

Ulanovskaya

O. A.

Cui

Kron

S. J.

et al . A Pairwise Chemical Genetic Screen Identifies New Inhibitors of Glucose Transport. Chem. Biol. 2011, 18, 222–230.

Siebeneicher

Bauser

Buchmann

et al . Identification of Novel Glut Inhibitors. Bioorg. Med. Chem. Lett. 2016, 26, 1732–1737.

10.

Sanford

Palmer

Recent Advances in Development of Genetically Encoded Fluorescent Sensors. Methods Enzymol. 2017, 589, 1–49.

11.

Imamura

Nhat

K. P. H.

Togawa

et al . Visualization of Atp Levels inside Single Living Cells with Fluorescence Resonance Energy Transfer-Based Genetically Encoded Indicators. Proc. Nat. Acad. Sci. U.S.A. 2009, 106, 15651–15656.

12.

Takanaga

Chaudhuri

Frommer

W. B.

Glut1 and Glut9 as Major Contributors to Glucose Influx in Hepg2 Cells Identified by a High Sensitivity Intramolecular Fret Glucose Sensor. Biochim. Biophys. Acta 2008, 1778, 1091–1099.

13.

Doucette

Zhao

Geyer

R. J.

et al . Flow Cytometry Enables Multiplexed Measurements of Genetically Encoded Intramolecular Fret Sensors Suitable for Screening. J. Biomol. Screen. 2016, 21, 535–547.

14.

Zhang

J. H.

Chung

T. D.

Oldenburg

K. R.

A Simple Statistical Parameter for Use in Evaluation and Validation of High Throughput Screening Assays. J. Biomol. Screen. 1999, 4, 67–73.

15.

Abo-Khatwa

A. N.

al-Robai

A. A.

al-Jawhari

D. A.

Lichen Acids as Uncouplers of Oxidative Phosphorylation of Mouse-Liver Mitochondria. Nat. Toxins 1996, 4, 96–102.

16.

Graven

S. N.

Lardy

H. A.

Johnson

et al . Antibiotics as Tools for Metabolic Studies. V. Effect of Nonactin, Monactin, Dinactin, and Trinactin on Oxidative Phosphorylation and Adenosine Triphosphatase Induction. Biochemistry 1966, 5, 1729–1735.

17.

Walter

Lardy

H. A.

Johnson

Antibiotics as Tools for Metabolic Studies. X. Inhibition of Phosphoryl Transfer Reactions in Mitochondria by Peliomycin, Ossamycin, and Venturicidin. J. Biol. Chem. 1967, 242, 5014–5018.

18.

Ahmed

Berridge

M. V.

N-Glycosylation of Glucose Transporter-1 (Glut-1) Is Associated with Increased Transporter Affinity for Glucose in Human Leukemic Cells. Leuk. Res. 1999, 23, 395–401.

19.

Kapoor

Finer-Moore

J. S.

Pedersen

B. P.

et al . Mechanism of Inhibition of Human Glucose Transporter Glut1 Is Conserved between Cytochalasin B and Phenylalanine Amides. Proc. Nat. Acad. Sci. U.S.A. 2016, 113, 4711–4716.

20.

Shoemaker

R. H.

The Nci60 Human Tumour Cell Line Anticancer Drug Screen. Nat. Rev. Cancer 2006, 6, 813–823.

21.

Smith

D. A.

Kerns

E. H.

The Effect of Plasma Protein Binding on In Vivo Efficacy: Misconceptions in Drug Discovery. Nat. Rev. Drug Discov. 2010, 9, 929–939.

22.

Siebeneicher

Cleve

Rehwinkel

et al . Identification and Optimization of the First Highly Selective Glut1 Inhibitor Bay 876. ChemMedChem 2016, 11, 2261–2271.

23.

Allen

M. D.

DiPilato

L. M.

Rahdar

et al . Reading Dynamic Kinase Activity in Living Cells for High-Throughput Screening. ACS Chem. Biol. 2006, 1, 371–376.

24.

Giatromanolaki

Sivridis

Arelaki

et al . Expression of Enzymes Related to Glucose Metabolism in Non-Small Cell Lung Cancer and Prognosis. Exp. Lung. Res. 2017, 43, 167–174.

25.

Yongzhi

Chen

et al . The Prognostic Value of Glut1 in Cancers: A Systematic Review and Meta-Analysis. Oncotarget 2017, 8, 43356–43367.

26.

Young

C. D.

Lewis

A. S.

Rudolph

M. C.

et al . Modulation of Glucose Transporter 1 (Glut1) Expression Levels Alters Mouse Mammary Tumor Cell Growth In Vitro and In Vivo. PloS One 2011, 6, e23205.

27.

Patra

K. C.

Wang

Bhaskar

P. T.

et al . Hexokinase 2 Is Required for Tumor Initiation and Maintenance and Its Systemic Deletion Is Therapeutic in Mouse Models of Cancer. Cancer Cell 2013, 24, 213–228.

28.

Rasola

Bernardi

The Mitochondrial Permeability Transition Pore and Its Adaptive Responses in Tumor Cells. Cell Calcium 2014, 56, 437–445.

29.

Dong

Mao

Xia

et al . Pkm2 and Cancer: The Function of Pkm2 beyond Glycolysis (Review). Oncol. Lett. 2016, 11, 1980–1986.

30.

Tisdale

E. J.

Kelly

Artalejo

C. R.

Glyceraldehyde-3-Phosphate Dehydrogenase Interacts with Rab2 and Plays an Essential Role in Endoplasmic Reticulum to Golgi Transport Exclusive of Its Glycolytic Activity. J. Biol. Chem. 2004, 279, 54046–54052.

31.

Kurtoglu

Gao

Shang

et al . Under Normoxia, 2-Deoxy-d-Glucose Elicits Cell Death in Select Tumor Types Not by Inhibition of Glycolysis But by Interfering with N-Linked Glycosylation. Mol. Cancer Ther. 2007, 6, 3049–3058.

32.

Cardaci

Desideri

Ciriolo

M. R.

Targeting Aerobic Glycolysis: 3-Bromopyruvate as a Promising Anticancer Drug. J. Bioenerg. Biomemb. 2012, 44, 17–29.

33.

Jung

D. W.

Kim

W. H.

Park

S. H.

et al . A Unique Small Molecule Inhibitor of Enolase Clarifies Its Role in Fundamental Biological Processes. ACS Chem. Biol. 2013, 8, 1271–1282.

34.

Satani

Lin

Y. H.

Hammoudi

et al . Enoblock Does Not Inhibit the Activity of the Glycolytic Enzyme Enolase. PLoS One 2016, 11, e0168739.

35.

Adams

D. J.

Ito

Rees

M. G.

et al . Nampt Is the Cellular Target of Stf-31-Like Small-Molecule Probes. ACS Chem. Biol. 2014, 9, 2247–2254.

36.

Gruenewald

Rennefahrt

Maldonado

S. G.

et al . Metabolic Responses in Cancer Cells with Differential Susceptibility to Glut1 Inhibition. AACR 106th Annual Meeting 2015. Cancer Res. 2015, 75, Abstract 1164.

37.

Lee

J. A.

Berg

L; Neoclassic Drug Discovery the Case for Lead Generation Using Phenotypic and Functional Approaches. J. Biomol. Screen. 2013, 18, 1143–1155.

38.

Kaelin

W. G.

Common Pitfalls in Preclinical Cancer Target Validation. Nat. Rev. Cancer 2017, 17, 425–440.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

1.82 MB