In vivo tissue-wide synchronization of mitochondrial metabolic oscillations

Summary

Little is known about the spatio-temporal coordination of mitochondrial metabolism in multicellular organisms in situ. Using intravital microscopy in live animals, we here report that mitochondrial metabolism undergoes rapid and periodic oscillations under basal conditions. Notably, mitochondria in vivo behave as a network of functionally coupled oscillators, which maintain a high level of coordination throughout the tissue via the activity of gap junctions. These findings reveal a unique aspect of the relationship between tissue architecture and self-organization of mitochondrial metabolism in vivo.

Introduction

Mitochondria provide the energy required to sustain biological processes by producing ATP through oxidative phosphorylation. The metabolic activity of mitochondria within a cell is the result of the coordination of several highly dynamic processes characterized by complex temporal patterns, which can display dynamic instability (Berridge and Rapp, 1979; Iotti et al., 2010). For example, studies in fermenting yeast, isolated cardiac cells, pancreatic islets, and exocrine glands acini have shown that oxidative phosphorylation exhibits oscillatory behaviors with periods spanning from seconds to minutes upon stimulation (Aon et al., 2003; Bruce et al., 2004; Dano et al., 1999; Kopach et al., 2011; Lloyd et al., 2012; Luciani et al., 2006; Voronina et al., 2002). Metabolic oscillations have been proposed to be either intrinsic to mitochondria (Vergun et al., 2003) or secondary to oscillations in other cellular processes such as, glycolysis (Dano et al., 1999), intracellular Ca²⁺ levels (Voronina et al., 2002), or plasma membrane potential (Berridge, 2008; Bertram et al., 2007). In other systems they have been linked to increased levels of reactive oxygen species (ROS) produced in the mitochondrial matrix (Aon et al., 2003). Moreover, it was recently proposed that mitochondria within a cell behave as a network of coupled oscillators under pathological conditions, and suggested that the communication among mitochondria is mediated by diffusible cytosolic messengers, such as ROS released from the mitochondrial matrix, which synchronize the energy status of the whole population of mitochondria (Aon et al., 2004, 2006; Kurz et al., 2010).

However, whether mitochondrial metabolic oscillations occur in live mammalian tissues in situ under physiological conditions, how they are regulated, and how they are spatially and temporally coordinated on a tissue scale, have not been investigated. Here, by using intravital two-photon (2P) microscopy we show that metabolic oscillations occur in the salivary gland epithelium of living rats under basal conditions, and that their characteristics and response to stimulation differ from what previously reported for ex vivo model systems. Indeed, the oscillations were not altered by manipulating cellular cytoplasmic Ca²⁺, but rather by interfering with mitochondrial ROS production, suggesting that they are driven by metabolism rather than receptor-mediated signal transduction. Notably, we find that the ability of mitochondria to self organize extends beyond individual cells and encompasses larger multicellular units within the tissue. Finally, we show that gap junction activity is important to maintain the synchronization of the mitochondrial metabolic activity.

Results

In order to measure the metabolic activity of the mitochondria in vivo we used intravital 2P microscopy that enables imaging of biological processes at cellular and subcellular levels in live animals (Weigert et al., 2013). As a model organ, we used the salivary glands (SGs) in live rats (Figure 1A), which are ideal to perform in vivo imaging since the motion artifacts due to heartbeat and respiration can be significantly reduced without compromising organ physiology (Masedunskas et al., 2011; Masedunskas and Weigert, 2008). As a reporter of the metabolic activity, we used NADH, a substrate of complex I that has the advantage of emitting upon 2P excitation without the need for exogenous labeling (Kasischke et al., 2004; Vishwasrao et al., 2005; Zipfel et al., 2003) (Figure 1A). The SGs were imaged 20–30 µm from the surface of the organ. At this depth, the tissue is composed primarily of acini, the main secretory units of the glands (Figure 1A), and NADH can be efficiently excited using minimal laser power, thus minimizing tissue photo-damage (Figure S1). Although NADH is present both in the cytoplasm and in the mitochondrial matrix, most of the detected endogenous emissions originated from the latter, as shown by the overlap with mitochondrial markers (Figure S1). When the SGs epithelium was imaged in time-lapse mode by using fast scanning speed, we measured periodic oscillations of the NADH emissions acquired from areas of the tissue encompassing 80–100 acini (Figure 1B and Movie S1). The oscillations were also observed deeper in the tissue (≈100 µm), where the large ducts are located (Figure S1), and were not due to motion artifacts (Figure S2, and Movie S2). They could be recorded over long periods of time, although the prolonged laser illumination resulted in photobleaching with a slight loss of signal (Figure 1B). Finally, the fluorescence intensity of the NADH emissions varied broadly among animals (maximum integrated fluorescence intensity 2.2*10⁷ ± 1.5*10⁷, Av. ± S.D., N=19), whereas the amplitude of the oscillations did not (% of the baseline level 15.2 ± 7.3 %, Av. ± S.D., N=19, Figure S2).

Figure 1. NADH and mitochondrial potential undergo fast oscillations in vivo.

(A) Diagram of experimental set up used for imaging the submandibular SGs in anesthetized rats. NADH was excited by 2P microscopy (excitation wavelength 750 nm) approximately 20–30 µm below the surface of the gland where the epithelium is primarily composed of acini, myoepithelial and stromal cells. NADH emissions provide the contrast for revealing the acini (arrowhead and inset). (B) Time-lapse imaging of the NADH emission in the SGs. The exposed glands were imaged in time-lapse modality for various times, as indicated (Movie S1), and the levels of NADH measured as described in Experimental Procedures. (C) NADH oscillations reports mitochondrial metabolism. The exposed SGs were bathed with either rotenone or the ROS scavenger TMPyP, and DMSO or PBS, respectively as controls (N=3 rats per condition), and then imaged by time-lapse 2P microscopy to measure the NADH levels. (D) Analysis of mitochondrial potential and NADH oscillations in vivo. Exposed SGs were bathed with TMRM and then imaged by 2P microscopy. The plots of both NADH (black) and TMRM (red) fluorescence intensity were fitted with a sine function α*sin(ωt+ϕ). Both, the periods and the phases were calculated (black symbols NADH, red symbols TMRM, N=4 rats, same symbols correspond to the same animal). Average (broken lines) and S.D. (blue lines) are shown. (E) NADH oscillations are unaffected by Muscarinic agonist and antagonist stimulation. NADH oscillations were monitored during stimulation with 5µg/kg, 15 µg/kg Carbachol or 1mg/kg Atropine. Stimulation with 5µg/kg or 15 µg/kg Carbachol led to water secretion measured in µl/g/min (bar graph) and expulsion of fluorescent dextran from the ducts monitored by time-lapse imaging.

Although NADH oscillations were observed in ex vivo model systems, such as pancreatic islets (Luciani et al., 2006), isolated cardiomyocytes (Zhou et al., 2011), perfused liver (Gaspers et al., 2012) and isolated acini from exocrine glands (Bruce et al., 2004; Voronina et al., 2002), their average period was much slower than what we observed in vivo (12.1 ± 3.1 sec, Av. ± S.D., N=19 vs. tens of seconds to minutes). No correlation was observed among the period of the oscillations and heartbeat, blood pressure, or blood glucose levels (Figure S2). Moreover, whereas in these systems oscillations were triggered by specific stimuli such as glucose (Merrins et al., 2010), laser-induced ROS (Aon et al., 2003), or receptor stimulation (Bruce et al., 2004; Gaspers and Thomas, 2008; Voronina et al., 2002), in vivo they were spontaneous and constitutive, and were impaired by high laser power (>25 mW) (Figure S3). The in vivo NADH oscillations were not due to ischemia, as reported for the intact brain (Mayevsky and Ziv, 1991; Mironov and Richter, 2001), since they occurred when blood flow and perfusion were within physiological levels (Baveja et al., 2002; Oyanagi-Tanaka et al., 2001), and they were completely abolished upon interruption of the circulation (Figure S3). However, as shown for other systems (Aon et al., 2003; Cortassa et al., 2004; Zhou et al., 2011), in vivo rapid NADH oscillations were linked to mitochondrial oxidative phosphorylation and ROS production, since they were abolished by either inhibiting complex I or scavenging of ROS (Figure 1C). In addition, mitochondrial membrane potential measured by loading the SG epithelium with reporters such as Rhodamine123 and TMRM (Lemasters and Ramshesh, 2007) oscillated in phase with NADH, as shown by fitting analysis (Figure 1D, Movie S3, and Experimental Procedures).

In acinar preparations from exocrine glands mitochondrial oscillations are elicited and modulated in a dose-dependent manner by agonists that increase intracellular Ca²⁺ levels (Bruce et al., 2004; Foskett and Wong, 1994; Gray, 1988; Voronina et al., 2002). Therefore, we sought to determine whether this was the case in the SGs in vivo. Unexpectedly, while stimulation of the muscarinic receptors with 5–15 µg/Kg carbachol robustly activated the Ca²⁺-mediated water secretion, it had no effect on the NADH oscillations (Figure 1E). Moreover, inhibition of the muscarinic receptor by atropine did not alter NADH oscillations. (Figure 1E). Thus, muscarinic receptor stimulation is not involved in the NADH oscillations in vivo, suggesting they are regulated by mechanisms other than those reported for ex-vivo acinar preparations.

Since the metabolic oscillations occurred throughout large areas of the tissue (Movie S4), we expected a certain degree of metabolic coordination among acini of the SG epithelium. Indeed, NADH oscillations measured in individual acini within the same animal exhibited the same period (Figure 2A). Although their phases were distributed over a wide range (Figure 2A), in adjacent acini the oscillations were in phase, indicating that the SG epithelium is organized in metabolically synchronized syncytia (Figure 2B). This finding prompted us to investigate whether metabolic oscillations in mitochondria located in adjacent acini were synchronized as well. To this end, we labeled the SGs with cationic dyes, which enabled the use of high magnification objectives while maintaining low laser power and high sampling frequency, thus making possible to resolve clusters of mitochondria (Figure S1). Strikingly, both the mitochondrial membrane potential and NADH oscillated with same periods and phases (Figure 3 and Movie S5), indicating that mitochondria located in different cells or acini behave as a network of coupled oscillators.

Figure 2. Spatial coordination of metabolic oscillations in the SGs epithelium.

(A) Measurement of NADH oscillations in individual acini in vivo. The boundaries of the acini were highlighted by bathing the exposed glands with 70 kDa Texas Red-dextran (upper panels, Ac, red broken line). A gland per animal was imaged for at least 5 minutes by time-lapse 2P microscopy to measure NADH oscillations (40–70 acini per field of view). The distribution of both periods and phases for each acinus were plotted for each animal (5 rats, N represents the number of acini per animal) and the variance was reported for each group. Average (broken lines) and S.D. (blue lines) are shown. (B) Correlation between the spatial organization of the acini within the tissue and the phase of the NADH oscillations. Acini exhibiting oscillations with the same phase (variance<0.05) were grouped (left panel) and color-coded (right panel). The data shown represent the phases measured in animal #1.

Figure 3. Coordination of mitochondrial oscillations among adjacent acini.

SGs were bathed with TMRM and 2P time-lapse imaging was performed using a 60X objective to reveal metabolic oscillations in mitochondrial clusters (upper panels, arrowheads) within individual acini (upper panels, red broken line). NADH and TMRM oscillations were measured for each mitochondrial cluster in adjacent acini (N represent the number of mitochondrial clusters per acinus, 4 acini per animal, 3 rats) and fitted. The distribution of both periods and phases for each cluster were plotted and the variance was reported for each group. Average (black broken line) and S.D. (blue lines) are shown.

How is the synchronization maintained among acini? “Independent” biological oscillators become synchronized when they are functionally coupled through feedback loops that may involve various signaling molecules (Kim et al., 2010). We hypothesized that such molecules should be transported through gap junctions that in SGs are localized both in the intercalated ducts connecting adjacent acini, and among acinar cells within the same acinus (Figure S4). Notably, when the gap junctions were blocked by carbenoxolone (Davidson et al., 1986), NADH oscillations measured in individual acini persisted became asynchronous (Figure 4A, Figure S4, and Movie S6). To further highlight the loss of synchronization, we calculated the times for the oscillations to reach the first maximum and correlated them with the spatial coordinates of the acini. Under control conditions acini clustered in metabolically coordinated groups (Figure 2B and Figure 4B) whereas carbenoxolone treatment decoupled both the oscillations among adjacent acini (Figure 4A and Figure S4) and the oscillations within the same acinus (Figure 4C). In addition, carbenoxolone reduced the amplitudes of the oscillations, indicating that the activity of the gap junctions plays a key role in maintaining the oscillations, as well (Figure S4).

Figure 4. Role of gap junctions on the coordination of the mitochondrial oscillations.

(A) Effect of carbenoxolone on NADH oscillations. The exposed glands were bathed either in 1 mM carbenoxolone or saline. NADH oscillations were recorded for each individual acinus. The periods of the NADH oscillations were calculated and plotted for each animal (N=3 rats per condition). (B) Correlation between the spatial organization of the acini within the tissue and the phase of the NADH oscillations in the presence of carbenoxolone. The time to reach the first maximum (First peak) was calculated for each acinus, and acini exhibiting oscillations with the same first peak (variance<0.5) were grouped and color-coded. (C). Effect of carbenoxolone on NADH oscillations within the same acinus. Acini were ideally divided in two halves and NADH oscillations were measured in each half. Representative NADH oscillations in a single acinus (half1, black lines, half2 red lines). (D) Model of in vivo synchronization of the metabolic oscillations. The gap junctions among the acini and the intercalated ducts (yellow cylinders) transport a molecule(s) that maintains the synchronization of the mitochondrial oscillations in vivo, possibly, by feedback loops.

Discussion

Here, we showed for the first time that in vivo mitochondrial metabolic oscillations occur under basal conditions, and with rapid periods (10–15 sec). This is in contrast with what has been previously reported in both ex vivo and in vitro model systems where oscillations are triggered by the application of exogenous stimuli (Aon et al., 2003; Bruce et al., 2004; Gaspers and Thomas, 2008; Merrins et al., 2010; Mironov and Richter, 2001; Voronina et al., 2002). Our findings underscore the fact that in the native tissue the complex integration of signals coming from the vasculature, the nervous system, the tissue microenvironment, and the surrounding cells, ensure the proper conditions to maintain mitochondrial metabolism in an oscillatory modality (Kim et al., 2010). A major advantage for cell metabolism to adopt an oscillatory behavior is that oscillations are a more favorable dynamic status to rapidly respond to sudden changes in the environment and energy demand (Goldbeter et al., 2012; Kruse and Julicher, 2005; Richter and Ross, 1981).

The mechanisms triggering and regulating this process in vivo are not well understood and are currently under investigation. Although metabolic oscillations have been observed in isolated mitochondria, thus suggesting that they may be an intrinsic feature of this organelle (Iotti et al., 2010; Vergun et al., 2003), it is more likely that they are secondary to other processes which may target complex I and the Krebs cycle. Such processes might include oscillations in intracellular Ca²⁺ levels (Foskett JK et al. 1994; Bruce JI et al. 2004), plasma membrane potential (Bruce et al., 2004; Foskett and Wong, 1994), glycolysis (Dano et al., 1999), or components of the redox cycle (Aon et al., 2003), as shown in intact cells. Our data suggest that ROS has a role in driving the NADH oscillations, as was described in other systems (Aon et al., 2003; Camello-Almaraz et al., 2006; Iotti et al., 2010; Zhou et al., 2011). On the contrary, cellular Ca²⁺ increase and muscarinic receptor stimulation do not affect NADH oscillations. One possible explanation for this unexpected finding is that in the SGs in vivo, the relationship between mitochondria and the IP₃ receptors in the ER may not be as tight as reported for isolated cells (Csordas and Hajnoczky, 2003; Hajnoczky et al., 1995), thus implying that mitochondria Ca²⁺ levels may not increase in response to muscarinic stimulation. Another possibility is that the highest carbachol concentration used in this study did not increase mitochondrial Ca²⁺ to a level sufficient to affect mitochondrial oscillations.

Notably, we also showed that in live animals mitochondria are functionally linked and form supra-cellular networks, thus extending the ability to self-organize and coordinate their oscillatory behavior beyond the single cell (Aon et al., 2006) (Figure 4D). Specifically, in the SG epithelium groups of acini that are interconnected by the same duct behave as metabolically synchronized syncytia, which maintain coordination through a signaling molecule that is transmitted by gap junctions (Figure 4D). This finding has critical implications as to how tissues regulates their metabolic state, and underscores the advantages of studying mitochondrial metabolism in vivo, where the complexity of tissue architecture is preserved.

Further studies are required to elucidate the mechanisms underlying the metabolic oscillations in vivo. Although we have shown that scavenging ROS inhibits NADH oscillations, this pathway has not been characterized yet. In addition, whether ROS plays a role in maintaining their synchronization across the tissue needs to be elucidated (Aon et al., 2004, 2006). Finally, ongoing studies are focused on cAMP, one of the major second messengers in SGs (Park et al., 2013), which may play an important role in mediating ROS metabolism in vivo, as shown in isolated cardiomyocytes (Anderssson et a., 2011, Perrino et al., 2010).

In conclusion, the ability to measure and characterize metabolic oscillations through NADH emission in vivo, has revealed novel aspects of cellular metabolism that have not been appreciated in reductionist systems, and opens new avenues to study mitochondrial metabolism and tissue energetics at a cellular, subcellular, and tissue scale under both physiological and pathological conditions, such as mitochondrial-associated metabolic disorders.

Experimental Procedures

Animals

Sprague–Dawley male rats weighing 150 – 250 g were obtained from Harlan Laboratories Inc. (Frederick, MD). The animals were acclimated for one week before used for the procedures. Water and food were provided ad libitum. All experiments were approved by the National Institute of Dental and Craniofacial Research (NIDCR, National Institute of Health, Bethesda, MD, USA) Animal Care and Use Committee. The animals were anesthetized by an intraperitoneal injection of 100 mg/Kg ketamine (Fort Dodge Animal Health, Fort Dodge, IA) and 20 mg/Kg xylazine (Akorn Inc., Decatur, IL). The SGs were externalized, and the body temperature of the animals was controlled and maintained at 37–38 °C, as previously described (Masedunskas et al., 2013b)

Labeling of the SGs and drug administration

To highlight the acini the exposed glands were bathed in situ in 1mg/ml 70-kDa Texas Red–dextran in saline (Invitrogen) for 10 minutes. To label mitochondria, the exposed glands were bathed in 0.5 µM mitotracker (Invitrogen) for 20 min, and to image the mitochondrial potential they were bathed with either 1µM TMRM or 10µg/ml rhodamine 123 (Invitrogen) for 20 minutes. To block complex I the exposed SGs were bathed for 10 min with 10µM rotenone (Sigma, St. Louis, MO) using 4% DMSO, as control whereas to scavenge ROS they were bathed with 1 mM TMPyP (Aon et al., 2003) To inhibit complex the activity of the gap junctions, the glands were bathed with 1mM carbenoxolone (Sigma, St. Louis, MO) for 30 minutes. Carbachol (5–15 µg/Kg) and Atropine (1 mg/Kg) were injected SC in the neck of the anesthetized rats. Higher concentrations of carbachol (50 µg/Kg) induced significant contractions of the glandular tissue and adverse reactions in the animal, making impossible to reliably measure the oscillations.

Intravital Microscopy

The externalized SGs were accommodated on a coverslip mounted on the stage above the objective. The SGs and the body of the animal were immobilized using custom-made holders as previously shown (Masedunskas et al., 2013a; Masedunskas et al., 2013b). The blood flow was assessed visually by using the eyepiece. Two-photon microscopy was performed as described before using an IX81 inverted confocal microscope equipped with a Fluoview 1000 scanning head (OlympusAmerica) (Masedunskas and Weigert, 2008). Emission were detected using a custom-made non-descanned detector (Masedunskas and Weigert, 2008) and specifically, NADH was detected on the first PMT (dichroic mirror 510 nm, barrier filter 400 nm-480 nm), Rhodamine123 was detected on the second PMT (dichroic mirror 570 nm, barrier filter 505 nm-560 nm) and TMRM, Mitotracker, and Texas Red on the third PMT (barrier filter 590 nm-650 nm). For time-lapse imaging, acquisition speed ranged from 0.2 to 1 s/frame. All the images and time-lapse series were acquired between 20 and 30µm from the surface of the glands using the following objectives from Olympus America:

1. UPLSAPO60×, 60X N.A 1.2 water immersion objective (Figure 3, Suppl Figure 1, and 5, and Movie S5)

2. XLUMPFL20XW, 20X N.A. 0.95 water immersion objective (Figure 1, 2, Suppl Figure 1, and Movie S1, S2, and S3)

3. XLPL25XWMP, 25X N.A. 1.05 (Suppl. Figure 1)

4. UPLSAPO30X, 30X N.A. 1.05 silicon oil objective (Figure 4 and Movie S6)

5. PlanAPON 5X and 2X air objectives, N.A. 0.08 (Movie S4)

Fitting and Statistics

To correct for the bleaching of the NADH signal during the course of imaging and to reduce the noise at high frequency a sliding window approach was applied over 10 frames. To fit the corrected curves the lowest point was defined as baseline and subtracted by each data point. The shifted curves were fitted with a sine function α*sin(ωt+φ), where α represents the amplitude of the oscillations, ω represents their frequency, and φ represents their phase. The period T of the oscillations was calculated as 2π/ω. A Chi-square test was performed and curves with a chi-square value larger than 2 were rejected. Average, S.D. and variances were calculated for the accepted α, ω and φ values. A matlab program was developed to automate the process. During carbonexolone treatment the metabolic oscillations in individual acini exhibited different periods within the same animal. For this reason, the phases could not be calculated and compared among acini. Instead, as a measurement of the lack of synchronization, the time at which the oscillations reached the first maximum was calculated and termed First Peak.

To determine the spatial coordination of the oscillations, the phase φ (or the first peak for the carbonexolone treatment) for each acinus was determined and the Euclidean distance between any pair of φ values (or first peak) was calculated. A binary hierarchical cluster was drawn by grouping acini with the same phase (or first peak). Acini belonging to the same clusters exhibited a variance lower than 0.05, were color-coded, and located on the corresponding acquired image.