Systems analysis of intracellular pH vulnerabilities for cancer therapy

Abstract

A reverse pH gradient is a hallmark of cancer metabolism, manifested by extracellular acidosis and intracellular alkalization. While consequences of extracellular acidosis are known, the roles of intracellular alkalization are incompletely understood. By reconstructing and integrating enzymatic pH-dependent activity profiles into cell-specific genome-scale metabolic models, we develop a computational methodology that explores how intracellular pH (pHi) can modulate metabolism. We show that in silico, alkaline pHi maximizes cancer cell proliferation coupled to increased glycolysis and adaptation to hypoxia (i.e., the Warburg effect), whereas acidic pHi disables these adaptations and compromises tumor cell growth. We then systematically identify metabolic targets (GAPDH and GPI) with predicted amplified anti-cancer effects at acidic pHi, forming a novel therapeutic strategy. Experimental testing of this strategy in breast cancer cells reveals that it is particularly effective against aggressive phenotypes. Hence, this study suggests essential roles of pHi in cancer metabolism and provides a conceptual and computational framework for exploring pHi roles in other biomedical domains.

Fig. 1

Reconstruction of enzymatic pH-dependent activity profiles. a Six critical pH points, corresponding to the 0%, 50%, and 100% of enzymatic activity at the acidic and basic regimes were extracted from BRENDA from all taxa. Missing data was complemented with existing data from close homologs or were predicted using linear regressors, generating an imputed database of pH-activity profiles, from which one infers the pH-profile of any enzyme (Methods, Supplementary Methods and Supplementary Figures 1–7 for a complete description). b Predicted vs. experimental pH optima, defined as the average of the critical points A100 and B100. The red line depicts linear regression. c Distributions of the pH optima of metabolic enzymes in each cellular compartment. Box widths are proportional to the number of enzymes in each compartment. Each box delineates lower quartile, median, and upper quartile values. Most extreme values (whiskers) are within 1.5 times the inter-quartile range from the ends of the box. Red dots depict the measured physiological pHi range of the compartment. “Including experiments” boxes correspond to the pH optima that were used in the subsequent GSMM modeling. As a validation, we include “Only predictions” boxplots, which are the result of the 10-fold cross-validation (Supplementary Figure 5)

Fig. 2

In silico pH-dependent metabolism of cancer and normal cell models. a Cellular proliferation (biomass yield) as a function of pHi, normalized by the maximal value obtained across all pHi examined, of cancer (circles) and normal healthy (solid) cells, when GAPDH is at physiological levels (black) and when it is inhibited (color), as depicted in the inset. b Uptake/production rates of oxygen, glucose, total ATP, and total NADPH. Uptake rates are conventionally depicted with a negative sign (more negative values denote higher rates). Error bars depict the standard deviation of the mean values across the populations of GSMMs at each pHi. c Anti-proliferative effects of gene inhibition (knockout), showing the classification of knockouts according to their selectivity and pH-specificity scores (top). The predicted targets, ranked by their pH-specificity, with the average selectivity scores superimposed (middle), as well as frequency of scores across all pair comparisons (≥12.5%) are shown (bottom). d The anti-Warburg scores (OCR/ECAR) of knockouts at low and physiological pHi (top), and the changes in the uptake/production rates of key metabolites, relative to the wild type (WT) at low pHi (bottom), are shown for each target. Pathways associated with each target are shown in color code. Results are robust with respect to choice of model parameters (Methods and Supplementary Figures 8–12)

Fig. 3

Experimental proof of concept. a pHi measurements of naïve and acid-adapted (AA) MCF7 breast cancer cells, under different oxygen availability, and following inhibition of MCT1/2. For pHi measurments at least 30 cells were analyzed. b Efficiency of knockdown of GAPDH and GPI mRNA and protein following transfection of MCF7 cells with the respective siRNAs (72 h). PCR was done in three separate biological replicate with three repeats. c Proliferation assays of cells (Methods) transfected with siRNA targeting GAPDH and GPI, across four conditions: normoxia, hypoxia, each at physiological pHe (7.4), or acidic pHe (6.7). Drop in proliferation following inhibition of MCT1/2 is shown as connected lines. Lowest values are obtained when pHi is low (yellow grids). The amplified effect of gene inhibition is seen relative to control (color vs. black). Viability assays is done in three replicates and three reads for each time point. d Viability assays (Methods) demonstrate that when pHi is sufficiently low the predicted strategy is particularly effective against AA cancer cells. The inhibition of GAPDH and GPI achieve efficient killing of cells at low pHi (yellow grids). Note the large slopes obtained for AA cells. The strategy is selective (Supplementary Figure 14), however, fails when sufficiently low pHi is unattainable, as in the case of triple negative breast cancer cells (Supplementary Figure 15). Bars depict the error of the mean over replicas. The experiments were repeated three times with three replicas for each condition

Fig. 4

Validation of systems analyses in predicting pHi-sensitive metabolic vulnerabilities. a pHi measurements of naïve and acid-adapted (AA) breast cancer MCF7 cells, under normoxia and following the inhibition of NHE1, by cariporide treatment. For pHi measurments at least 30 cells were analyzed. b Efficiency of knockdown of indicated targets at the mRNA and protein levels, following reverse transfection of MCF7 cells with the indicated siRNAs. qPCR was repeated at least three times with three replicates. c The effect of gene inhibition in normal and low extracellular pH (pHe) shown for naïve and AA MCF7 breast cancer cells. Similar color code to Fig. 3 is applied. At low pHe where the lowest pHi was obtained there is a large reduction in the viability of cells. In AA cells, only the selective and pH-specific targets (GAPDH, GPI, and ACAT2) achieve amplified anti-proliferative effects following NHE1 inhibition, despite the smaller reduction in the pHi of these cells. PFAS, a selective but not pH-specific target, is similar to control cells following NHE1 inhibition. Knockdown of RPIA had a weak effect in naïve cells and no/opposite effects in AA cells. The viability assay was done three times with four replicates each time. The bars depict the mean and the error bars depict the standard deviation of the mean

Fig. 5

Anti-Warburg effects of lowering pHi and inhibition of pHi-dependent metabolic targets. Seahorse flux experiments were preformed for MCF7 naïve cells under normoxia, with (low pHi, yellow grid) and without (physiological pHi) inhibition of NHE1 and of the indicated metabolic targets. The reduction in pHi (black bars) has an anti-Warburg effect on cancer cell metabolism, as measured by the ratio between the oxygen consumption rate (OCR) and the extracellular acidification rate (ECAR). Both at physiological pHi and at low pHi the additional inhibition of the selected targets (colors), amplifies the anti-Warburg effect on cancer cells. Highest OCR/ECAR ratios are obtained at low pHi (yellow grid). Seahorse experiments were done in six replicas each time and experiments were repeated three times. The bars depict the mean and the error bars depict the standard deviation of the mean