MCF7 cells (from ATCC, provided by Central Laser Facility, UKRI) were cultured in minimal essential media (MEM), supplemented with 10% FBS and 1% penicillin/streptomycin. For the experiments, cells were seeded onto individual dishes or multiwell plates and treated with 0.005% v/v DMSO (Control) or 1, 5, 10, and 20 µM CBD (Sigma, United Kingdom) for 24 h prior to assessment at 37°C with 5% CO₂ humidified air.

2P-FLIM was used to image and quantify NAD(P)H lifetime decay as a measure of cellular and mitochondrial metabolism. MCF7 cells were seeded at densities of 1.5 × 10⁵ onto individual glass bottom dishes (MatTek) and treated as described. Before imaging, cells were acclimatized for 15 min to room temperature. The 2P-FLIM setup has been described previously (Botchway et al., 2015). Images were obtained as follows: 750 nm wavelength light for 2-photon excitation of NADH was generated by a mode-locked titanium sapphire laser (Mira F900, Coherent Laser Ltd.), producing 180 fs pulses at 76 MHz. This laser was pumped by a solid-state continuous wave 532 nm laser (Verdi 18, Coherent Lasers Ltd.). Images were collected through a water immersion 60X 1.2 NA objective on a modified Nikon EC2 confocal scanning system attached to a TE2000-U microscope as described previously (Botchway et al., 2015). Emission was collected by the same objective, reflected off a 530 BK-25 filter (Comar Optics, United Kingdom), passed through a BG39 filter and detected with an external hybrid GaAsP (HPM-100–40, Becker and Hickl, Germany), linked to a time correlated single photon counting (TCSPC) module (SPC830, Becker and Hickl, Germany). In this configuration, together with an average laser power of 0.5 mW at the sample, we mostly observe the fluorescence signal from NAD(P)H with negligible contribution from the flavins. This was confirmed with solution phase studies of NAD(P)H and FAD (both purchased from Sigma, United Kingdom, and used without further purification). We have performed a dual channel FLIM setup containing BG39 with or without extra band pass filters (a 440–490 nm band pass filter for NADH and a 520–570 nm band pass filter for FAD and flavins), and we did not detect flavin emission in the NADH channel. Solution concentrations as high as 0.5 mM are observed in the flavin-detecting channel only. The intracellular FAD concentration has been estimated to be around 8 μM (Kimata et al., 2018). For each of the five experimental groups (Control, 1, 5, 10, and 20 μm CBD), five spatially distinct 2P-FLIM images were acquired. Within each image, five nonoverlapping cells with the highest signal-to-noise ratio were chosen, equating to 25 cells per treatment group. As NAD(P)H signal intensity aligns with mitochondrial morphology (See Supplementary Figure S1), pixels corresponding to the mitochondria or the cytosol could be manually selected (See Supplementary Figure S2) (the spatial resolution of the setup was determined to be ∼350 nm) (Alam et al., 2017). Photon counts of at least 1,000 were used for the multiexponential analysis, which also excluded nuclear areas from analysis. The decay curve of each pixel from the 2P-FLIM image was modeled with a biexponential decay curve, according to Eq. 1, where I(t) represents fluorescence intensity at time t after the laser pulse. Goodness of fit was quantified by Chi-squared analysis (See Supplementary Figure S3). From each pixel two observed lifetimes were recorded: τ_Free for free NAD(P)H and τ_Bound for enzyme-bound NAD(P)H and the contributions of each lifetime to the overall decay curve, α_Free and α_Bound, respectively (Sharick et al., 2018; Blignaut et al., 2019). I ( t ) =   α F r e e e x p − t / τ 1 + α B o u n d e x p − t / τ 2   [1]

Mitochondrial ROS and Ca²⁺ levels were quantified using the MitoSOX Red (ThermoFisher, United Kingdom) and Rhod2 (ThermoFisher, United Kingdom) stains, respectively, in octuplicate, as per manufacturer’s instructions. Briefly, MCF7 cells were seeded at densities of 3.0 × 10⁴ cells per well. Following treatment, cells were stained with either 5 µM MitoSOX Red or 1 µM Rhod2 for 30 min. Fluorescence was measured on a FLUOstar Optima microplate reader (BMG, Germany). Excitation/Emission: 510/580 nm for MitoSOX Red and 552/580 nm for Rhod2. Fluorescence intensities were normalized to total cellular protein, measured using the Bradford assay.

Mitochondrial morphology was analyzed as previously described (Valente et al., 2017). Briefly, MCF7 cells were seeded at densities of 1.0 × 10⁵ cells per well and treated with DMSO (Control) or CBD for 24 h as described above. Following treatment, cells were stained with MitoTracker Deep Red and NucBlue (ThermoFisher, United Kingdom) and imaged using the ThermoFisher EVOS FL2 Auto using a 40X 0.65 NA plan fluorite objective. Five representative images were selected per treatment group and within each image, the mitochondria of three spatially distinct cells were selected for analysis. To quantify CBD-induced changes in morphology, the cells were analyzed with the MiNA plugin for ImageJ. This script, following a preprocessing stage of unsharp masking, local contrast enhancement (CLAHE), and median filtering, binarized and then skeletonized the mitochondrial image to produce an output of the mitochondrial footprint (the total area of the binary image), mean length of mitochondrial networks, mean branch length per network (the mean of the sum length of individual branches on separate networks) and mean number of networks. Three cells were analyzed per image for the five images. This experiment was carried out in triplicate.

Data are presented as box plots, the box shows median with 25th and 75th percentiles, and the whiskers indicate minimum and maximum values. Statistical significance was calculated by one-way ANOVA with Dunnett’s multiple comparison using GraphPad Prism version 8.0 (GraphPad Software, La Jolla, California, United States). An adjusted p < 0.05 was considered statistically significant.

A paradox in the understanding of the function of plant products, including CBD, is how they can enhance survival of normal cells, yet kill cancer cells. The Warburg shift, or aerobic glycolysis (Kroemer, 2006; Wallace, 2012), may be key, as it is overall antiapoptotic and probably key in the earlier stages of carcinogenesis, as more advanced cancers often upregulate mitochondrial function. This has led to the so called “Inverse Warburg Hypothesis”, which, although not fully understood, suggests that high levels of mitochondrial metabolism underlie the observed inverse relationship between degenerative diseases and cancer. Mitochondria are pivotal in the detection of and adaptation to stress, sending out redox signals in response to stress, which may be an important mechanism explaining how plant secondary metabolites work as medicines via a process called mitohormesis (Tapia, 2006). Critically, evidence is accumulating that major plant stress signaling compounds, such as jasmonates, can modulate mitochondria in both plant and animal cells (Bömer et al., 2020). Moreover, plant compounds are known to interfere with glycolytic pathways and major controllers of metabolism, such as VDAC. Thus, they appear to inhibit carcinogenesis, but if a tumor does arise, they can alter its mitochondrial phenotype to restore cell death (Stevens et al., 2018).

In cancer cells, reduced endoplasmic reticulum–mitochondria Ca²⁺ exchange contributes to apoptotic resistance; increased intracellular Ca²⁺ contributes to enhanced cell migration, facilitating metastasis (Marchi and Pinton, 2016). This is matched by changes in mitochondrial morphology, making cells more resistant to apoptosis (Senft and Ronai, 2016). Critically, mitochondrial function is tightly controlled by calcium, ranging from stimulation to enhance respiration, to overload to induce cell death (Rossi et al., 2019)—and in cancer, calcium flow is directed away from the mitochondrion (Danese et al., 2017).

CBD modulates TRPV1 and calcium influx into the cell and potentially into the mitochondrion (Ryan et al., 2009; Mato et al., 2010; Olivas-Aguirre et al., 2019). More directly, CBD has been shown to bind to and close VDAC1, a major channel responsible for transporting Ca²⁺, potentially responsible for the immunosuppressive and anticancer effects of CBD (Rimmerman et al., 2013). In leukemic cells, CBD induces dose-related mitochondrial dysfunction, leading to apoptosis, which is associated with mitochondrial calcium overload, loss of mitochondrial membrane potential, and enhanced ROS production (Olivas-Aguirre et al., 2019). This all supports earlier data that CBD application can result in mitochondrial calcium overload and enhanced ROS production (Ryan et al., 2009; Mato et al., 2010).

In response to CBD, we identified a dose-dependent decrease in mitochondrial bound NAD(P)H (α_Bound). This appears to indicate reduced activity of the ETC, as the oxidation of NADH by complex I is the first point where electrons enter the ETC (Chakraborty et al., 2016). Indeed, comparable reductions in NAD(P)H α_Bound have been observed in response to known mitochondrial toxins and inhibitors such as potassium cyanide and rotenone (Bird et al., 2005; Schneckenburger et al., 2004), while in contrast, agents which increase metabolic activity increase mitochondrial α_Bound NAD(P)H (Alam et al., 2017). So, in relation to our data, the increase in unbound NAD(P)H may represent an inhibition of the ETC and/or enhanced TCA leading to increased NADH, possibly coupled to inhibition of glycolysis and enhanced PPP activity, a generalized inhibition of metabolism leading more unbound NAD(P)H, and possibly, inhibition of NADP⁺ binding to G6PDH. The latter effect is a known ability of the polyphenol epigallocatechin gallate (EGCG) (Shin et al., 2008) which also modulates mitochondrial function (Oliveira et al., 2016). Inhibition of G6PD has also been observed for other phenolic compounds (Adem et al., 2014), while many have also been found to dissociate hexokinase from VDAC1, which inhibits glycolysis and can enhance apoptosis (Tewari et al., 2015; Tewari et al., 2017; Goldin et al., 2008). There is thus, perhaps, a generalization that many of these bioactive plant compounds have the capacity to induce metabolic reprogramming, which may also apply to CBD. Whilst the contribution of each species to the decay curve were affected by CBD treatment, there were no significant changes in the lifetimes of each species itself (τ_free and τ_Bound). Changes in τ_Bound are often associated with a redistribution of NAD(P)H binding sites as a result of a shift in metabolic pathways (Skala et al., 2007b). The absence of changes in this study may indicate no such pivot in specific metabolic pathways, although further studies comparing CBD-mediated τ_free and τ_Bound to agents known to perturb mitochondrial metabolism would be required to validate this.

In terms of intracellular localization, 2P-FLIM permits highly localized, label-free, sub-micron determination of the redox state, enabling the differentiation of the mitochondrial redox response from that of the cytosol. The lack of effect on free or bound cytosolic NAD(P)H in our results would seem to indicate that the actions of CBD are mitochondrial specific, which has been proposed based on the highly lipophilic properties of CBD (Huestis, 2007; Ryan et al., 2009). However, the subtle, interconnected relationships between the many cellular pathways maintaining stores of NADH are not fully deconvoluted in the aggregate FLIM signal (Blacker and Duchen, 2016), making full interpretation difficult. For example, while cancer cells shift their preferred energy-generating pathway to aerobic glycolysis (Trigos et al., 2018), the relative rates of glycolysis and OXPHOS do not appear to affect the intracellular NADH fluorescence lifetime (Blacker and Duchen, 2016; Guo et al., 2015).

It should be noted that the 24 h post‐exposure protocol employed here was based on our earlier investigations into the actions of CBD (Kosgodage et al., 2018) and therefore limits our ability to define an accurate timescale for the specific cascade of intracellular events following CBD administration. However, previous studies have indicated that CBD-triggered increases in Ca²⁺ precede any mitochondrial changes in ROS, membrane potential, or morphology (Olivas-Aguirre et al., 2019). This might suggest a sequential modulation of various parts of the cells, starting on the outside and working inwards as this highly lipophilic compound is absorbed by the cell.

Overall, our FLIM data suggest that CBD induces mitochondrial oxidative stress in these cancer cells, which correlate with the dose-related increase in mitochondrial ROS and calcium and evidence of increased mitochondrial fission. Although cancer cells often display abnormal mitochondrial dynamics, mitochondrial fission is often a response to oxidative stress, especially if ROS originates from damaged mitochondria and is associated with a loss of the mitochondrial membrane potential and influx of calcium, leading to mitophagy and renewal or cell death (Sharma et al., 2019). Mitochondrial fission can also form part of a positive feedback cytosolic calcium signaling pathway that can promote autophagy, which also seems to be reliant on increased ROS production (Huang et al., 2019). Mitochondria have a pivotal role in maintaining cellular homeostasis by sequestering and releasing intracellular Ca²⁺. Within the mitochondria, Ca²⁺ biphasically affects energy metabolism via activation/inhibition of mitochondrial enzymes and direct charge alteration of the mitochondrial membrane potential, which in turn regulates the supply of electrons into the respiratory chain and production of ROS (Danese et al., 2017). Ultimately however, prolonged accumulation of mitochondrial calcium, known as calcium overload, is associated with the mitochondrial permeability transition (MPT), a key step in the initiation of apoptosis, and has been identified in T-cells following treatment with CBD (Olivas-Aguirre et al., 2019). Indeed, these data support our previous observations of CBD-induced mitochondrial fission in a different cell line (Nunn et al., 2013) and our unpublished observations in several other cell lines. Overall, this supports the concept that CBD can induce ROS in a range of tissue types, including monocytes (Wu et al., 2018), glioma stem cells (Singer et al., 2015), and breast cancer cells (Kis et al., 2019).

To explain this, we suggest that CBD triggers calcium uptake via TRPV and/or release from endoplasmic stores, which is then taken up by the mitochondrion. In fact, TRPV1 is also expressed intracellularly, and it has been suggested to play a role in calcium signaling to the mitochondrion (Nita et al., 2016). Combined with data that it may also modulate VDAC1 (Rimmerman et al., 2013), we suggest that as CBD is absorbed by the cell, it sequentially modulates pathways that can result in either enhanced cellular protection or, potentially, induction of cell death; key in this is both direct and indirect modulation of mitochondrial function. Support for this comes from research that indicates that CBD can induce autophagy, which could be important in epilepsy (Hosseinzadeh et al., 2016), alcohol-induced liver damage (Yang et al., 2019), and cancer (Shrivastava et al., 2011). Autophagy is pivotal in maintaining overall mitochondrial health and involves VDAC1 and mTOR; a key trigger of the process is mitochondrial dysfunction (Li et al., 2020). Induction of mitochondrial stress could lead to mitophagy, which in some cells restores homeostasis, for instance, ensuring inflammation resolution, but in others, such as cancer cells, it results in death. Figure 4 summarizes a possible mode of action. A key point here is that VDAC1 binds many different proteins, including HK and components of the Bcl2/Bax, so it is also central in controlling cell fate and a target for cancer treatment (Shoshan-Barmatz et al., 2017).

In broader terms, our data suggest that CBD, as its actions are often biphasic, could be mitohormetic (Nunn et al., 2020), inducing sublethal mitochondrial stress that results in an adaptive response via increased ROS production (Ristow and Schmeisser, 2014). In cancer cells, being at higher levels of stress than healthy cells, this tips them into a terminal oxidative stress. This selectivity has been observed following treatment of T-cells with CBD, inducing Ca²⁺ overload and apoptosis in cell lines derived from acute lymphobastic leukemia of T-lineage, but not in healthy T-cells (Olivas-Aguirre et al., 2019). This demonstrates the potential of CBD in anticancer therapy by “priming” cells, through sensitizing the mitochondria to respond better to chemotherapy and hence reducing the severity of treatment and side effects (Henley et al., 2017; Kosgodage et al., 2018).