Coupling-dependent metabolic ultradian rhythms in confluent cells

Abstract

Ultradian rhythms in metabolism and physiology have been described previously in mammals. However, the underlying mechanisms for these rhythms are still elusive. Here, we report the discovery of temperature-sensitive ultradian rhythms in mammalian fibroblasts that are independent of both the cell cycle and the circadian clock. The period in each culture is stable over time but varies in different cultures (ranging from 3 to 24 h). We show that transient, single-cell metabolic pulses are synchronized into stable ultradian rhythms across contacting cells in culture by gap junction-mediated coupling. Coordinated rhythms are also apparent for other metabolic and physiological measures, including plasma membrane potential (Δψ_p), intracellular glutamine, α-ketoglutarate, intracellular adenosine triphosphate (ATP), cytosolic pH, and intracellular calcium. Moreover, these ultradian rhythms require extracellular glutamine, several different ion channels, and the suppression of mitochondrial ATP synthase by α-ketoglutarate, which provides a key feedback mechanism. We hypothesize that cellular coupling and metabolic feedback can be used by cells to balance energy demands for survival.

Keywords: cellular metabolism; gap junctions; ion channels; membrane potential; ultradian rhythms.

Fig. 1.

An oxidative phosphorylation–dependent ultradian rhythm in confluent mouse fibroblast cells. (A) Upper panel is a representative trace of bioluminescence signals detected via Lumicycle in one clone of ear fibroblasts derived from a wild-type (WT) Per2::Luc knockin mouse. Lower panel shows the Wavelet analysis of the time series in the upper panel, with color scale indicating the wavelet power. Shown are a circadian period of ∼24 h and another rhythm with a period of ∼3.5 h that emerged after day 2. (B) Upper panel is a representative trace of bioluminescence signals for one clone of MEFs derived from a Cry1/Cry2 double-knockout (CDKO) mouse stably expressing luciferase under the Per2 promoter. Lower panel is the Wavelet analysis showing an ultradian rhythm emerged after day 2 in the absence of circadian rhythms. (C) Simultaneous imaging of luminescence (Lum) and cytosolic pH (pHrodo Red). Shown are relative intensities normalized across the time series. High pHrodo signal indicates lower pH. (D) Intracellular ATP levels in cells at the peak or trough of the Lum rhythms. Data were normalized to the total protein. (E) Ultradian rhythms in UFs persisted when the glycolysis inhibitor 2-DG was administered (6 mM; as indicated by the gray bar). (F) Ultradian rhythms in UFs were disrupted when the oxidative phosphorylation inhibitor FCCP was administered (200 nM; as indicated by the gray bar). (G) Simultaneous imaging of Lum and mitochondrial membrane potential (TMRM). Shown are relative intensities (R.I.) normalized across the time series. High TMRM intensity corresponded with more-negative ΔΨ_m (hyperpolarization) and was antiphase to luminescence rhythms.

Fig. 2.

Glutaminolysis is required for the generation of ultradian rhythms. (A) LumiCycle bioluminescence recordings of UF/CMV-Luc cells with complete recording medium lacking the indicated components. Only medium lacking glutamine diminished ultradian cycles. (B) Analogous to A but with a minimal medium supplemented with the components indicated. Only medium containing glutamine sustained ultradian rhythms. (C–G) LumiCycle bioluminescence recordings of confluent UF/CMV-Luc cells in complete recording medium (timing of drug administration indicated with gray bars). (C) The B0AT1 type glutamine transporter blocker, nimesulide (100 µM); (D) the ASCT2 type glutamine transporter blocker, L-γ-glutamyl-p-nitroanilide (GPNA) (1 mM); (E) the glutaminase inhibitor, BPTES (30 µM); (F) simultaneous inhibition of glutamate dehydrogenase by EGCG (20 µM) and aminotransferases by AOA (2 mM); and (G) supplementation with DmKG (4 mM); all reversibly inhibited ultradian rhythms. (H) Treatment with DmKG (4 mM) for 24 h reduced intracellular ATP in confluent UF cells (***P < 0.001, t test). (I) ATP synthase activity in mitochondria isolated from UF/CMV-Luc cells. Plotted are mean relative activities (from n = 3 replicates) after subtracting activity in the presence of 10 µM oligomycin A (OligA). K_i = 17.3 mM was calculated by fitting the data to a noncompetitive inhibition model. (J) Fluorescence imaging of ΔΨ_m sensor TMRM (10 nM loaded 30 min before imaging) in UFs. DmKG (2 mM), OligA (2 µM), FCCP (2 µM), or vehicle (DMSO, 0.1%) was added at the time indicated (black arrow), and images were taken at 5-s intervals. Fluorescence intensities were normalized to baseline averages. High TMRM intensity corresponds with a hyperpolarization in ΔΨ_m. (K and L) Metabolomics analysis of cell lysates collected at the trough or peak of the luminescence rhythms detected changes in levels of KG (K, *P = 0.047, t test) and glutamine (L, **P = 0.015, t test).

Fig. 3.

Plasma membrane potential and calcium signaling are involved in ultradian rhythms. (A) Bioluminescence rhythms in confluent UF/CMV-Luc cells recorded in the LumiCycle were suppressed in the presence of 50 mM KCl, but not NaCl. (B) The Na/K-ATPase inhibitor ouabain (20 µM) also disrupted ultradian rhythms. (C) Confluent UFs were loaded with ΔΨ_p sensor DiBAC and ΔΨ_m sensor TMRM, and time-lapse fluorescence imaging intensities were normalized to the mean intensities across recording duration. High DiBAC intensity corresponded with depolarization of ΔΨ_p, and high TMRM intensity corresponded with hyperpolarization of ΔΨ_m. Depolarization of ΔΨ_p is always accompanied with depolarization of ΔΨ_m. (D) Confluent UF/GCaMP cells were loaded with ΔΨ_m sensor TMRM and ΔΨ_p sensor BeRST, and the fluorescence of GCaMP, TMRM, and BeRST was simultaneously monitored with DeltaVision at 15-min intervals. Data were normalized to the mean of the first 10 h. High intensity of GCaMP corresponds with high [Ca²⁺]_I, and high intensity of BeRST corresponds with depolarization of ΔΨ_p. Thus, an increase in [Ca^2+]_i was associated with depolarization of ΔΨ_p and ΔΨ_m. (E) The L-type, voltage-gated calcium channel blocker nimodipine (100 nM) suppressed ultradian rhythms in UF/CMV-Luc cells recorded in Lumicycle. (F) The voltage-gated sodium channel inhibitor carbamazepine (CBZ) (500 µM) inhibited ultradian rhythms.

Fig. 4.

Stochastic ultradian rhythms are synchronized by intercellular coupling. (A) Heat map (Upper) and time series traces (Lower) of representative luminescence intensity of eight cells in the same dish subjected to time-lapse, single-cell bioluminescence imaging. Synchronization occurred before the establishment of regular rhythms. (B) Heat map and traces of single-cell, time-lapse bioluminescence imaging of noncontacted UF/CMV-Luc cells in the same dish. No rhythms or synchronization were detected. (C) Maximal information coefficient (MIC) of each pair of cells (Upper) and representative traces of ΔΨ_m (normalized TMRM fluorescence) in noncontacting (Left), subconfluent (Center), and confluent (Right) UFs. Rhythms were sporadic in noncontacting and subconfluent cells and synchronized in confluent cells. Higher MIC scores indicate more correlation between cells. (D) The gap junction blocker carbenoxolone (50 µM), but not its control drug glycyrrhizin (50 µM), blocked the ultradian luminescence rhythms. (E) UFs cocultured (at ratio of 1:1) with U2OS/CMV-Luc cells, which have one (Cx43) or two (Cx43 & Cx45) gap-junction genes knocked down. Disrupting gap junctions impaired the maintenance or sustainability of the ultradian rhythms. (F) A hypothetical model for the biochemical basis of the ultradian rhythms described in this paper. Gln: glutamine; GlnT: glutamine transporter; Glu: glutamate; KG: α-ketoglutarate; KAs: α-ketoacids; AAs: amino acids; GLS: glutaminase; GDH: glutamate dehydrogenase; TA: transaminase; ATPsyn: ATP synthase; NaKp: Na⁺,K⁺-ATP pump; ΔΨ_p: plasma membrane potential; CaCs: calcium channels; GJs: gap junctions.