Lactate-dependent chaperone-mediated autophagy induces oscillatory HIF-1α activity promoting proliferation of hypoxic cells

Abstract

Response to hypoxia is a highly regulated process, but little is known about single-cell responses to hypoxic conditions. Using fluorescent reporters of hypoxia response factor-1α (HIF-1α) activity in various cancer cell lines and patient-derived cancer cells, we show that hypoxic responses in individual cancer cells can be highly dynamic and variable. These responses fall into three classes, including oscillatory activity. We identify a molecular mechanism that can account for all three response classes, implicating reactive-oxygen-species-dependent chaperone-mediated autophagy of HIF-1α in a subset of cells. Furthermore, we show that oscillatory response is modulated by the abundance of extracellular lactate in a quorum-sensing-like mechanism. We show that oscillatory HIF-1α activity rescues hypoxia-mediated inhibition of cell division and causes broad suppression of genes downregulated in cancers and activation of genes upregulated in many cancers, suggesting a mechanism for aggressive growth in a subset of hypoxic tumor cells.

Keywords: HIF; ROS; Warburg and reverse Warburg effect; cancer microenvironment; chaperone-mediated autophagy; hypoxia; hypoxic oscillations; lactate; quorum sensing; reactive oxygen species.

Fig 1.. Single cell responses to hypoxia display complex and varied dynamic profiles.

(A) Time course analysis of HRE-GFP transfected HEK293 cells exposed to hypoxia (2% O₂) for 12 hours, followed by reoxygenation for 6 hours; n > 500 cells from 4 experiments (exp); (B) Representative time course images of HRE-GFP in HEK293 cells in sustained hypoxia show a subset of cells exhibit reduction in HRE-GFP levels (arrowheads); Scale bar = 25μm. (C) Schematic showing putative dynamic diversity of HRE responses, with a subset of cells switching from HRE-GFP^hi to HRE-GFP^lo state. (D) Representative HRE-GFP intensity traces in HEK293 cells in 12 hours of hypoxia, showing cells that remain in HRE-GFP^lo state (red), those that monotonically increase and maintain their HRE-GFP levels (green), and those that exhibit oscillations between HRE-GFP^hi to HRE-GFP^lo (orange); (E) Live cell imaging analysis of HRE-GFP transitions in sustained hypoxia (12 hours) followed by reoxygenation (6 hours); Shown is percentage of HEK293 cells in 6 hours segments belonging to subclasses in D; statistical significance shown for each subclass (colored) vs previous 6h time segment; n = 7 exp with more than 500 cells/exp. (F) Live metabolic analysis by Seahorse of HeLa cells shows oxygen consumption rate (OCR) decreases following 5 mM DMOG treatment vs control, and rate of acid efflux (ECAR) levels increase with DMOG treatment similar to oligomycin treatment; n = 6 exp; (G) Representative traces of HRE-GFP intensity in HEK293 cells in low density for 18 hours in hypoxia with 10 mM lactate (colors are defined as in (D)). (H) Fractions of HEK293 cells displaying HRE-GFP oscillation during 12 hours of hypoxia in 25 mM glucose, 10 mM lactate (without or with 100 nM AR-C155858, an inhibitor against MCT-1 transporter); n = 5 exp; (I) Imaging of HRE-GFP patient-derived breast cancer cells in the presence of 10mM lactate; colored arrows point to bounded cells representative of stably high (green), stably low (pink), and oscillating (orange) response; Phase-contrast images shown in Fig S1L; (J) Percentage of oscillating patient-derived cells with lactate, without or with AR-155858; n = 6 exp (>300 cells/exp). Error bars: s.e.m.; *: p < 0.05, **: p < 0.01, ***: p < 0.001.

Fig 2.. ROS-induced CMA controls HIF-1α dynamics in hypoxic cells.

(A-D) ROS increases oscillations in HIF-1α activity; (A) Average DCF-DA intensities in HeLa cells cultured with 25 mM glucose or pyruvate, before and after treatment with Rotenone for 30 minutes; n = 3 exp; (B) Fraction of HRE-GFP cells displaying oscillation in the presence of 5 μM H₂O₂ with or without 1 mM LNAC; n = 10 exp; (C) Fraction of oscillatory cells in subpopulations sorted for high CellROX and low CellROX, imaged after 6 hours subsequent to sorting; n =12 exp. (D) Microarray analysis of gene transcripts related to redox, and chaperone-mediated autophagy (CMA) pathways in HypoxCR-HEK293 cells cultured in hypoxia for 12 hours, and sorted for GFP^hi and GFP^lo levels. (E-J) ROS induced CMA mediates HIF-1α oscillations; (E) Pearson’s coefficient of correlation between ROS levels and LAMP-2A levels in individual HeLa cells shown on the right; n > 28 cells; (F) Percentage of oscillating cells in the presence of activators (50 mM 6-AN, and 200 nM Digoxin), and inhibitors of lysosomal degradation (10nM Bafilomycin, and 50 μM Chloroquin). (G-H) Representative images showing HIF-1α, along with lysosomal Hsc70 13D3 (G), and LAMP-2A (H) in HeLa cells cultured for 12 hours in 1% O₂, 10 mM lactate, lactate with LNAC, as well as 25 μm H₂O₂ without or with LNAC, or CMA inhibitor Leupeptin (50 μM); (I-J) Pearson’s coefficient for colocalization of lysosomal Hsc70, or LAMP-2A and HIF-1α; n = 30 cells; (K) Live cell imaging analysis of GFP oscillations in HIF-1α-GFP transfected HeLa cells cultured in hypoxia (control), or with Leupeptin (50 μM), or after siRNA mediated knockdown of STUB1 and LAMP-2A; n = 12 cells; (L) Schematic showing plasmid maps for HIF-1αWT and HIF-1α^ΔΔ with mutation in the KFERQ-like motif in the HIF-1α encoding sequence. (M) Immunoblot showing abundance of V5 tagged-HIF-1α, and total HIF-1α in HIF-1αWT and HIF-1α^ΔΔ cells maintained in 5% O₂ and treated with doxycline for 8 hours and then withdrawn for 2 hours; GAPDH shown as loading control; (N) Pearson’s coefficient of correlation between HIF-1α and LAMP-2A in HeLa cells stably expressing HIF-1αWT and HIF-1α^ΔΔ; n > 25 cells; (O) Live cell imaging analysis of GFP oscillations in HeLa cells transduced with wild-type HIF-1α^WT and mutated HIF-1α^ΔΔ; n > 20 samples. Error bars: s.e.m.; *: p < 0.05, **: p < 0.01, ***: p < 0.001.

Fig 3.. HIF-1α dynamics is regulated by lactate through control of intracellular ROS and pH.

(A) Oxygen consumption rate (OCR) in HeLa cells with 10 mM glucose, or 10 mM lactate + 5 mM glucose, or oligomycin; n = 6 exp; (B) Intracellular pyruvate levels in HeLa cells to extracellular lactate after 12 hours; n = 6 exp; (C) NBDG uptake in HEK293 cells in normoxia or hypoxia, with10 mM lactate in 12 hours; n = 30 cells; (D) Average DCF intensity in HeLa cells cultured in medium containing a combination of glucose, lactate, H₂O₂, and LNAC after 12 hours; n > 25 cells. (E-G) Lactate mediated acidosis contributes to variation in HRE-GFP levels; (E) Intracellular pH (pHi, measured by Snarf-1 dye) of HeLa cells in the presence of glucose, lactate, H₂O₂, and LNAC; n = 30 cells; (F) Correlation of pHi and HIF-1α-GFP levels in HeLa cells cultured in glucose, lactate, or H₂O₂; (G) Spread in HIF-1α-GFP intensities attributed to pH-mediated degradation; Total noise is the observed GFP intensity values minus the mean; pH corrected noise is the residual deviation remaining after regressing to the pH; error bars denote one standard deviation on each side; n = 10 samples. (H) Representative immunostained confocal images showing colocalization of LAMP-2A and HIF-1α-GFP in HeLa cells treated with 10nM Bafilomycin, and cultured for 12 hours in the presence of glucose, lactate, and H₂O₂ without, or in the presence of LNAC; green: LAMP-2A, red: HIF-1α-GFP; Scale bar = 2 μm; Pearson’s coefficient of spatial correlation shown in (I); Concentrations where applicable: 25 mM glucose; or 2 mM glucose added to 10 mM lactate, 25 μM H₂O₂, 10mM LNAC; n = 30 cells. Error bars: s.e.m.; *: p < 0.05, **: p < 0.01, ***: p < 0.001.

Fig 4.. Modeling CMA mediated HIF-1α degradation accounts for diverse dynamics in HIF-1α activity mediated by cell density.

(A) Hypothesized molecular circuit controlling the dynamics of HIF-1α responses; CMA: chaperone mediated autophagy; Also listed are the main perturbations used in the study: Red: inhibitors, Green: activators, Blue: reporters. (B) Bifurcation diagram of the mathematical model corresponding to (A), see Methods, demonstrating existence of different dynamic response modes; ‘C’ denotes the critical point of the phase diagram corresponding to the onset of oscillatory responses; (C) Model predictions for dynamic HRE-GFP intensity profiles in hypoxic cells in the presence of low and high levels of lactate; Traces correspond to the sets of parameter values indicated by two rows of circular dots in the bifurcation diagram in (B). (D) Pearson’s coefficient showing correlation of HRE-GFP and CellROX levels; n = 30 cells; (E-G) The effect of cell density on variation in HRE-GFP expression; (E) Laconic FRET ratio in A375 cells; n = 12 cells; (F) HRE-GFP transitions in HEK293 cells cultured at low density, at high density with and without 10 mM lactate, or MCT-1 transporter inhibitor; n > 12 exp; (G) Percentage of HRE-GFP oscillations in A375 in co- culture with BJ5ta fibroblasts at different relative percentages, also with lactate; n = 3 exp; (H) Schematic showing that high density in solid tumors can result in lactate buildup in the core, causing increased oscillations in HIF-1α activity; (I-J) Confocal time lapsed images of HRE-GFP HEK293 spheroids in CoCl₂ and frequency of GFP switches in the core (area shown within dotted line); Scale bar = 100 μm; n = 4 spheroids. (K) Confocal images showing dynamics of HRE-GFP HEK293 cells in a zebrafish embryo, imaged 1 day after microinjection into the yolk sac; white arrows show cells with changes in GFP expression; (L) Coefficient of temporal variation (standard deviation normalized by mean) over 14 hours of observation for detectable cells in 2D cultures, and zebrafish; n > 40 locations. Error bars: s.e.m.; *: p < 0.05, **: p < 0.01, ***: p < 0.001.

Fig 5.. Oscillating hypoxic response can lead to differential gene regulation and allow proliferation under hypoxia.

(A) Experimental protocol to entrain intracellular hypoxic response by imposing oscillations of extracellular O₂. (B) Immunoblot showing HIF-1α abundance in HeLa cells after 1% hypoxia for 6 hours, and after re-oxygenation for 10 min, and 1 hour, GAPDH is loading control; (C) RNA-Seq based hierarchical clustering of gene expression in HeLa cells kept in normoxia, and stable or oscillating hypoxia for > 24 h; Genes differentially regulated by oscillatory hypoxia shown in bracket. (See Fig S5). (D) Key Ingenuity pathways upregulated in gene-set differentially regulated by oscillatory hypoxia (referred to in B) vs persistent hypoxia. (E-J) Oscillations in HIF-1α abundance allows cells to continue to proliferate in hypoxia; (E) Immunoblot showing abundance of Cyclin B1 and Cyclin D1 in cells in normoxia, and stable or oscillating hypoxia; GAPDH is loading control. (F) RNA-Seq-based analysis of pairwise fold-changes between cells in above conditions for cell-cycle related genes. (G) Propidium iodide levels in HEK293 cells after conditioning with stably high, low or oscillatory input of 2% oxygen for 18 h; Quantification of gated cycling cells shown in (H); n = 12 exp. (I) Percentage of HypoxCR-GFP transfected HeLa cells positive for Geminin-mCherry after 12 h each of environmental normoxia, hypoxia, and oscillating hypoxia; above panel shows a representative location. (J) Percentage of HypoxCR-GFP transfected HEK293 cells positive for Geminin-mCherry within HRE-GFP+ve, HRE-GFP-ve subpopulations, and cells that spontaneously oscillate in persistent hypoxia; n = 6 exp. (K) Percentage of proliferating cells in HeLa cells stably transfected with Tet-inducible wild-type HIF-1α (HIF-1α^WT) and HIF-1α with mutation in KFERQ-like lysosomal targeting motif (HIF-1α^ΔΔ), with stable or oscillating doxycycline concentrations; n = 6 exp. Error bars: s.e.m.; *: p < 0.05, **: p < 0.01, ***: p < 0.001.

Fig 6.. Oscillating hypoxic response dynamics regulates gene expression congruent to many human cancers.

(A-C) Genes downregulated by oscillating hypoxia show a correlatively downregulated signature in many human cancer types. (A) Heatmap showing genes upregulated by hypoxia, but downregulated by oscillatory hypoxia (Osc^down genes), and (B) their relative expression in cancers vs normal tissue in the TCGA tumor database; colors refer to a TCGA study where the given gene is expressed significantly higher (red), or significantly lower (green) in tumor compared to normal tissue sample, with significance for the whole set of Osc^down genes in (C). (D-F) Genes upregulated by oscillating hypoxia also show a correlatively upregulated signature in many human cancer types: (D) Heatmap showing genes downregulated in hypoxia, but upregulated in oscillatory hypoxia (Osc^up genes), and (E) direction of their expression in TCGA tumors vs normal samples; color code same as B. (F) Significance of the whole set of Osc^up genes shown in tumor vs normal; In C, and F, bars show overall fold change of the geneset, colorbar represents the mean log₂fold change between cancer and normal samples. (G-H) Genes differentially regulated by oscillatory hypoxia are significantly downregulated (G), or upregulated (H) when compared against the directional expression of all other genes in those TCGA tumors vs control samples; In both G, and H, each dot represents a tumor comparison with normal tissue; x-axis refer to log(pvalue) of tumor vs normal comparison; bars show 95% confidence interval for mean vs all genes difference; vertical red line marks threshold of pvalue = 0.05. (I-L) Representative examples of Osc^down and Osc^up gene sets in breast cancer BRCA, and liver hepatocellular carcinoma LIHC: (I) Histogram showing density of Osc^down genes (green), or all genes (pink); Genes are normalized by total gene numbers; BRCA tumor samples = 1085, normal breast samples = 291. (J) GSEA enrichment analysis in Osc^down gene expression in BRCA vs normal samples; all genes are ranked in order of decreasing tumor/normal fold changes on x-axis, with bars representing Osc^down genes. (K) Histogram showing relative distribution of fold change of Osc^down genes (green), or all genes (pink); Genes are normalized by total gene numbers; LIHC tumor samples = 369, normal breast samples = 160. In I, and K, asterisk represents the mean log₂fold change between cancer and normal samples for the corresponding gene set (all or Osc^down /Osc^up). (L) GSEA enrichment analysis in Osc^up gene expression in LIHC cancers. (M) Osc^down genes are significantly downregulated in many TCGA cancer types when compared to all the other genes; bars show a 95% confidence interval for the difference. (N) Many Osc^up genes are significantly upregulated in many TCGA cancer types compared to all the other genes; bars show a 95% confidence interval.