Adenosine triphosphate (ATP) serves as an energy currency, driving energy metabolism and regulating extracellular signalling. In recent years, ATP has garnered attention for its role as a hydrotrope that enhances the solubility of intracellular proteins1. Fluorescence imaging has been indispensable for unravelling biological processes involving ATP. To date, various fluorescent ATP indicators have been developed, including Förster resonance energy transfer (FRET)2, dual-emission ratiometric3,4, and single fluorescent protein (FP)-based indicators5,6. Despite their extensive availability, intensity-based analyses encounter common difficulties related to indicator concentration, excitation light amplitude, photobleaching, and focus drift, which often hamper quantitative analysis. Unlike the intensity-based approach, FLIM-based indicators offer the potential for quantitative analysis7. In particular, FRET-FLIM utilises FRET-based indicators. However, FRET-FLIM occupies two colour channels, restricting the possibility of multiplex imaging. By contrast, a single FP-based FLIM indicator can potentially circumvent this issue. Notably, several indicators using a single FP, such as mTurquoise or T-Sapphire, have recently emerged for detecting calcium8, glucose9, and lactate10. Moreover, a specific red-coloured FP-based Ca2+ indicator, R-CaMP1h, exhibits the capability for FLIM imaging11. Nevertheless, the availability of FLIM-based indicators compatible with a conventional 488 nm laser remains scarce12.
Herein, we present a single green FP-based ATP indicator using FLIM. We previously reported MaLionG, an intensiometric ATP indicator based on single green FP5. This was identified by the generation of mutants in which an ε subunit of an ATP-binding domain of a bacterial FoF1-ATP synthase was inserted into a variant of single green FP (citrine) via peptide linkers (Fig. 1A). By varying the position of the binding domain inserted into citrine and its peptide linkers, MaLionG was optimised with ‘turn-on’ fluorescence properties. Additionally, we identified another variant in which the fluorescence (λex/λem = 512/525) declined by 65% in the presence of 10 mM ATP (Fig. 1B). Intriguingly, while the fluorescence lifetime of MaLionG increased by only 0.16 ns, it decreased by 1.1 ns under 10 mM ATP, which was designated quantitative monitoring ATP level fluorescence lifetime-based turn-off green (qMaLioffG) (Fig. 1C, Supplementary Fig. 1). The dynamic range (Dt) of qMaLioffG was greater than that of conventional FRET-FLIM indicators (0.1–0.6 ns)13. The changes in the fluorescence intensity and lifetime exhibited ATP dose-dependence (Kd = 2.2 mM) and specificity for ATP (Fig. 1E, Supplementary Figs. 2 and 4). Notably, the influence of pH fluctuation on fluorescence lifetime was observed to be less significant than its effect on fluorescence intensity (Supplementary Fig. 3). Regarding the sensing mechanism, in contrast to MaLionG, qMaLioffG is unlikely to follow the exchange mechanism between the protonated and deprotonated states (Fig. 1D) (discussed later). Subsequently, we evaluated the performance of qMaLioffG in HeLa cells. qMaLioffG expressed in HeLa cells detected intracellular ATP depletion after treatment with sodium fluoride (NaF), an enolase inhibitor, which increased the fluorescence intensity and lifetime (Fig. 1F–H, Supplementary Fig. 5–7). Intracellular ATP was quantified using calibration curves of the fluorescence lifetime of pure qMaLioffG against the ATP concentration in membrane-permeabilised cells, as well as in solution (Fig. 1E).
We examined whether qMaLioffG could differentiate between the ATP levels in different cells with mitochondrial dysfunction or distinct activities in energy metabolism. Additionally, to monitor mitochondrial ATP, we validated the mitochondrial target qMaLioffG through oligomycin inhibition (Supplementary Fig. 8–9). First, qMaLioffG was applied to skin fibroblasts derived from patients with mitochondrial dysfunction. Fibroblasts obtained by skin biopsy from a patient with a mutation in dynamin-1-like protein (DNM1L) were used. Correct fission and fusion are essential for mitochondrial function14. As predicted, mitochondrial ATP levels were significantly lower in the mutant cells than in the control human dermal fibroblasts (Fig. 1I–J). Notably, diseased cells exhibited higher cytoplasmic ATP levels than control cells, indicating a potential compensatory mechanism for ATP production through glycolysis to counteract ATP depletion. We next investigated cytoplasmic and mitochondrial ATP in mouse embryonic stem cells (mESCs). mESCs can maintain a naïve state in culture by adding 2i (MEK and GSK3 inhibitors) and leukemia inhibitory factor (LIF)15. After the removal of 2i and LIF (hereafter 2iLIF), the expression of Nanog, an essential factor for sustaining the naïve state16, was entirely lost, whereas slight expression of Oct3/4 was observed, which served as a pluripotency marker (Supplementary Fig. 10). Subsequent quantitative ATP analysis of mESCs showed that cytoplasmic ATP levels were higher in the presence of LIF than in its absence, whereas no significant difference was observed in mitochondrial ATP levels (Fig. 2A). According to previous studies, pluripotency is maintained in the presence of 2iLIF through a combination of glycolysis and oxidative phosphorylation (OXPHOS); these cells are referred to as naïve pluripotent stem cells17. Conversely, in the absence of 2iLIF, ATP production relies solely on glycolysis, without the assistance of OXPHOS. This hypothesis is consistent with the observed decrease in cytoplasmic ATP levels upon 2iLIF removal (Fig. 2A). Additionally, mitochondrial ATP levels remained low due to the rapid efflux of ATP from the mitochondria into the cytoplasm, thereby hindering the detection of a significant difference2.
For a more comprehensive understanding of proliferation and metastasis in cancer biology, we examined the ATP levels in cancer cells with varying metastatic abilities. Previously, Morita et al. subcloned organoids derived from intestinal metastatic tumours carrying driver mutations, such as ApcD716, KrasG12D, Tgfbr2−/−, and Trp53R270H (AKTP). They then transplanted a single subcloned cell line into mice to investigate the occurrence of liver metastasis18,19. The subclones were categorised into two populations based on high (lines SC3 and SC24) or low (SC4 and SC6) metastatic abilities. We observed that cytoplasmic ATP levels were significantly higher in high metastatic cell lines than in the low metastatic cell lines. No significant differences were observed in mitochondrial ATP levels (Fig. 2B, Supplementary Fig. 11). Further genetic analysis of these cell lines demonstrated the loss of stemness markers in cell lines with low metastatic ability, whereas the genetic background identified by AKTP remained unaltered19. The observed higher ATP levels in metastatic cell lines corresponded well with the observation of elevated cytoplasmic ATP levels in mESCs, which maintain stemness, as shown in Fig. 2A.
To further evaluate qMaLioffG in multicellular systems, it was applied to the Drosophila brain. FLIM imaging of the brain expressing qMaLioffG exhibited higher cytoplasmic ATP levels in the Kenyon cells of the mushroom body compared to the other neurons (P neurons, including pSP and pIP neurons) (Fig. 2C). This observation implies that the mushroom body, known for its pivotal role in learning and memory processes, likely necessitates a higher energy supply. Additionally, although ATP levels varied among cells within the same region, the addition of oligomycin to the biospecimens significantly reduced the variance in ATP levels (Fig. 2C–D). In the future, this cellular heterogeneity could be addressed by integrating other technologies that identify cell types based on genetic information.
Finally, we imaged 3D spheroidal HeLa cells stably expressing qMaLioffG. Cultured spheroids are used for drug screening because their denser multicellular structure is closer to that of tissues. Remarkably, we observed lower cytoplasmic ATP levels within the spheroids than in their surroundings (Fig. 2E–F). Although a systematic study is yet to be conducted, such a gradient in the ATP concentration has rarely been observed in smaller spheroids (data not shown) or in 2D cell cultures. This observation was likely due to the limited nutrient diffusion to the centre of the spheroid. Subsequently, the spheroids were exposed to 2-deoxy glucose (2DG), a glycolytic inhibitor. Over time, the concentration gradient of cytoplasmic ATP gradually disappeared, although heterogeneous ATP levels persisted in comparison with the 2D cells.
For reference, the fluorescence lifetime data obtained through this study was converted to ATP concentration using an in-cell calibration curve, whereupon it was seen to be in good agreement with the literature (Supplementary Table 1)20. Still, the exact sensing mechanism of the indicator underlying the alteration of fluorescence lifetime remains unclear. We hypothesised that the binding of ATP might increase the freedom of rotation of the free-moving bond that connects the phenolic moiety and imidazolidone in the tyrosine-based chromophore. Consequently, while the deprotonated state of the chromophore is maintained to some extent, the energy from the excited state would be dissipated via a non-radiative process owing to the acceleration of chromophore motion. Supporting this hypothesis, we observed a significant change in the non-radiative rate constant (knr) of qMaLioffG after ATP binding, while the change in the radiative rate constant (kf) was minimal (Supplementary Table 2). Consequently, an alteration in the fluorescence lifetime was expected as it is inversely proportional to the combined kinetic rates of radiative and non-radiative processes7,21. However, in the case of MaLionG, the persistent dominance of this anionic chromophore as the emitting species, with or without ATP, accounts for the observed negligible change in fluorescence lifetime. Further elucidation of this mechanism will accelerate the generation of various single FP-based FLIM indicators to provide unique access to subcellular live cell physiology.