Short Communication: Subcellular Motion Compensation for Minimally Invasive Microscopy, In Vivo
Evidence for Oxygen Gradients in Resting Muscle
Rationale: In vivo microscopy seeks to observe dynamic subcellular processes in a physiologically relevant context. A primary limitation of optical microscopy in vivo is tissue motion, which prevents physiological time course observations or image averaging.
Objective: To develop and demonstrate motion compensation methods that can automatically track image planes within biological tissues, including the tissue displacements associated with large changes in blood flow, and to evaluate the effect of global hypoxia on the regional kinetics and steady state levels of mitochondrial NAD(P)H.
Methods and Results: A dynamic optical microscope, with real-time prospective tracking and retrospective image processing, was used collect high-resolution images through cellular responses to various perturbations. The subcellular metabolic response to hypoxia was examined in vivo. Mitochondria closest to the capillaries were significantly more oxidized at rest (67±3%) than the intrafibrillar mitochondria (83±3%; P<0.0001) in the same cell.
Conclusions: These data are consistent with the hypothesis that a significant oxygen gradient from capillary to muscle core exists at rest, thereby reducing the oxidative load on the muscle cell.
- two-photon excitation microscopy
- perivascular mitochondria
- NADH fluorescence
- mouse skeletal muscle
- redox state
Recently developed fluorescent probes for gene expression, cellular milieu, and enzyme activities can be coupled with multi-photon microscopy,1 with its improved penetration, sensitivity, and information content,2 to provide a unique window into cell physiology. Most in vivo microscopy preparations are mechanically isolated thin (<1 mm) tissues or tissues constrained using a cover glass. Without such constraints, physiological and perturbation-related motion prevents temporal signal averaging and complicates serial measurements of a given region.1,3 While imaging restrained, translucent specimens has yielded many insights, the ability to observe thick tissues would allow observation of processes in a broader range of experimental models and may simplify preparations.
In several tissues, including skeletal muscle, liver, kidney, and brain, we find drift movement significantly exacerbated by physiological perturbation (±100 μm/min). These displacements may be caused by changes in blood volume, flow, or cellular metabolism, resulting in a gradual redistribution of fluid between plasma, extracellular space, and intracellular space. Movement is amplified in larger tissues, where small percent changes lead to larger microscopic displacements.
We have addressed translational movement prospectively, using a dynamic microscope, and corrected for residual nontranslational movement retrospectively. Our objective was to follow tissue motions on the order of 100 μm, while retaining an in-plane resolution of <2 μm. This was accomplished by real-time tracking of 3D tissue using a dynamic stage, focusing unit, and computer system, adaptable to most microscopes. The utility of this approach was demonstrated by acquiring high-resolution images of the mouse skeletal muscle, vasculature, and mitochondrial distribution, in vivo, to monitor the redox state of the different intracellular pools of mitochondria under resting and hypoxic conditions.
The determination of vascular structure and flow relative to cellular morphology and metabolism may be critical to the understanding of microvascular flow regulation. Mitochondrial NAD(P)H fluorescence is an intrinsic probe for monitoring the redox effects of hypoxia or ischemia in vivo4,5 as well as in vitro.6 The [NAD(P)H]/[NAD(P)] ratio reflects the redox state of the mitochondria, and requires the determination of the fluorescence associated with both the fully reduced and fully oxidized NAD(P)H pool. Limited information is available on the subcellular metabolic responses of different cellular pools of mitochondria under any perturbation, in vivo, and such information could provide evidence for topological differences in cellular metabolism. We selected the use of hypoxia as an experimental perturbation because it is physiologically significant, results in substantial motion, and provides a fully reduced state for determining [NAD(P)H]/[NAD(P)] ratio in muscle.
Several technical and physiological questions were addressed in this study. (1) Can the tracking scheme compensate for the large volume shifts occurring during hypoxia to monitor the vascular structure and mitochondria redox response? (2) Do all of the fiber types in a given field of view follow the same time course of increased mitochondrial [NAD(P)H] in hypoxia? (3) Are the cellular pools of mitochondria at the same NAD(P)H redox state in resting muscle, reflecting regional differences in metabolism or oxygen delivery?
A commercial multiphoton excitation microscopy setup for in vivo imaging7 was coupled to a real-time computer (Figure 1). Prospective motion tracking was accomplished by sequentially acquiring orthogonal image planes, calculating tissue displacements from the images and offsetting motion to maintain a fixed relationship between the objective and the area of interest.
Image acquisition sequentially alternated between the 2 orthogonal imaging planes. An x-y image was acquired using the microscope galvo-mirrors. An x-z image was acquired by holding the y galvo-mirror stationary while line scanning in x and ramping the focusing motor in z. Both the x-y and x-z image planes were recorded by the real-time computer in parallel to the normal acquisition of the microscope system.
Image analysis compared each incoming image with a previously acquired image. At the start of motion tracking, one image from each acquisition plane was stored as a reference. Each subsequent image was compared to the reference image using normalized 2D cross-correlation to determine displacement (Figure 2). The stage was automatically adjusted based on the calculated offset to compensate for the motion.
The imaging setup simultaneously acquired 2 channels (expanded Methods section, available in the Online Data Supplement at http://circres.ahajournals.org). An emission centered on 460 nm originated from NAD(P)H.4 An emission centered on 590 nm detected di-8-ANEPPS dye (di-8-[butyl] amino-naphthyl-ethylene-pyridinium-propyl-sulfonate) in vascular endothelial cells, and this channel was used for tracking. Dye in the vasculature provided visualization of flow by negative contrast of moving red blood cells.
To correct for small in-plane deformations, a retrospective nonrigid body compensation was applied using deformable registration.8 A motion map was calculated for each frame by iteratively maximizing the local cross-correlation image subsets at different resolutions. The motion map calculation used the spatial information in the vascular/dye channel and then applied to both channels. To optimally filter the registered time course images, a principal component analysis was conducted (Online Data Supplement).
Free-breathing mice were anesthetized via a mask and the lower leg was immobilized by clamping just above the foot without obstructing proximal blood flow. The skin and surface fascia over the muscle were removed. A clear optical gel7 coupled the objective to the tissue. Alternatively, buffered saline, flowing from the objective to the muscle was used to deliver compounds (Online Data Supplement). A field-of-view showing vasculature and blood flow was selected ≈100 μm below the surface.
Protocols to validate resolution during motion tracking and to determine fully oxidized NAD(P)H fluorescence signal after uncoupler was performed (Online Data Supplement). In the hypoxia protocol, 6 animals underwent a systemic hypoxia, through death, to generate the fully reduced NAD(P)H signal. After collecting a control period breathing room air, the ventilated gas was switched to nitrogen to induce global hypoxia. Blood oxygenation of the animal was monitored with a pulse oximeter in 50% of the studies to confirm systemic hypoxia. The time course response was monitored in both image channels, until the vascular images showed a cessation or reversal of red cell flow and mitochondrial NAD(P)H fluorescence reached a plateau, several minutes after death.
Motion compensation during the hypoxia experiments averaged 28±5 μm per minute for up to 10 minutes (n=6, SE). Tracking was successful in maintaining the plane of interest within <2 μm. Image data were significantly improved using the deformable registration algorithm (Online Video 1). During hypoxia, the NAD(P)H signal increased (50% half-maximal [NAD(P)H] in 35±5 seconds, n=6, SE), visible in the filtered data and the calculated NAD(P)H/NAD(P) ratio images (Online Video 2). Two principle components of the hypoxia time courses reached significance in the experiments (expanded Results section in the Online Data Supplement). The primary and secondary components were linear combinations of each other (R2=0.73±0.07, n=6, SE), so within the temporal and spatial resolution of our experiments, no multimodal distribution of kinetics between cells or mitochondrial pools occurred within muscle. Thus, the kinetic response of NAD(P)H fluorescence to hypoxia was statistically identical in all fiber types and intracellular pools.
The percentage change in NAD(P)H fluorescence was significantly higher in the perivascular regions than the core or intrafibrillar regions. This indicates that the outer regions of cells were significantly more oxidized than the core under control conditions (Figure 3). NAD(P)H/NAD(P) images were calculated using regional fluorescence as [resting level]/[hypoxic level]. Mid-cell cross-sections were identified by fiber dimension and the absence of capillaries on the surface. Mid-cell NAD(P)H/NAD(P) profiles were generated for all datasets (expanded Results section in the Online Data Supplement). The NAD(P)H ratio at rest averaged 67±3% reduced in perivascular regions and 83±3% reduced in intrafibrillar regions (n=8 cells from 6 animals; P<0.0001, paired t test; SE).
The motion compensation methods described are capable of tracking regions within a cell, in vivo, during physiological perturbations. Using prospective tracking and retrospective image correction, we demonstrated improved average image signal-to-noise ratios with minimal degradation of spatial resolution. This approach is useful in producing in vivo images with high signal-to-noise ratios and enables in vivo fluorescence methods with low signal-to-noise, such as spectral imaging, that require extensive signal averaging. It also allows continuous measurement of a set of cells through time during a physiological perturbation, in vivo.
Given the extensive motion associated with hypoxia, it would have been extremely difficult to observe intracellular events through this perturbation without the tracking system. The topology of the fluorescence was consistent with mitochondrial NAD(P)H, and fluorescence intensity responded appropriately to uncoupler and hypoxia. The mean increase in NAD(P)H fluorescence during hypoxia over the entire field-of-view was approximately 50%, consistent with whole muscle measurements made with single photon excitation averaged over large regions of muscle3,7,9 or in arteriole versus venule regions of capillaries. However, these whole muscle studies average NAD(P)H redox state over multiple cells, whereas we demonstrate that the redox state is dependent on position within a cell.
These results indicate that either the metabolic state in mitochondria in the perivascular region is substantially different than in the intrafibrillar pool, or that a significant oxygen gradient exists from the outer regions of the cell, close to the capillary networks, to the center of the cell. It has been suggested that the perivascular mitochondria function, composition, and susceptibility to flow insults10 may be different from the intrafibrillar mitochondria.10,11 However, none of these observations would predict a resting difference in NAD(P)H redox state.
A likely reason for the NAD(P)H redox gradient would be an oxygen gradient from the capillaries to the center of the cell. The existence of an oxygen gradient across single cells or within the muscle5 has been a controversial subject for many years.12 However, the larger concentration of mitochondria at the vascular interface13 could support steep oxygen gradients because the cellular oxygen consumption capacity is focused in this region.14 Supporting this hypothesis is the calculation that the oxygen consumption capacity of the perivascular regions is ≈4-fold greater than the core of the muscle cell.11 This electron microscopy estimate of the relative distribution of mitochondria could be an underestimate; 3D views of the tissue structure show a large fraction of the mitochondria situated near the capillaries (Online Videos 3 and 4).
The distribution of mitochondria at the periphery of the cell requires energy transfer within the cell to be dependent on the diffusion of ATP, creatine phosphate, and creatine and not the faster diffusion of oxygen. This dependence on metabolite diffusion and associated high concentration of metabolites may contribute to the metabolic homeostasis across the muscle fiber.14 We speculate that the high concentration of mitochondria located at the capillaries results in the generation of significant oxygen gradients across cells in low-flow resting conditions. This oxygen gradient may be beneficial in reducing the concentration of oxygen in the cell volume and preventing oxidative damage, relying on the high concentration of metabolites to maintain the distribution of potential energy throughout the cell. The distribution of mitochondria may reflect a strategy to minimize cellular oxygen tension, reducing the potential toxic effects of oxygen in the cytosol.
Sources of Funding
Supported by the NIH/National Heart, Lung, and Blood Institute Division of Intramural Research (grant HL004610-02).
Rubart M. Two-photon microscopy of cells and tissue. Circ Res. 2004; 95: 1154–1166.
Denk W, Strickler JH, Webb WW. Two-photon laser scanning fluorescence microscopy. Science. 1990; 248: 73–76.
Chance B, Cohen P, Jobsis F, Schoener B. Intracellular oxidation-reduction states in vivo. Science. 1962; 137: 499–508.
Mayevsky A, Rogatsky GG. Mitochondrial function in vivo evaluated by NADH fluorescence: from animal models to human studies. Am J Physiol Cell Physiol. 2007; 292: C615–C640.
Philippi M, Sillau AH. Oxidative capacity distribution in skeletal muscle fibers of the rat. J Exp Biol. 1994; 189: 1–11.
Henneman E, Olson CB. Relations between structure and function in the design of skeletal muscles. J Neurophysiol. 1965; 28: 581–598.
Novelty and Significance
What Is Known?
Visualization of metabolic and signaling events within cells in vivo is required to fully understand many clinically relevant physiological processes.
Tissue motion is a major limitation for micrometer-resolution imaging in vivo.
The existence of subcellular oxygen tension gradients in vivo has been speculated to be important in metabolic signaling and protection from free radical damage.
What New Information Does This Article Contribute?
A strategy is presented to compensate for tissue motion while maintaining micrometer resolution using an image-guided dynamic microscope.
Muscle mitochondria, which use oxygen, are concentrated near vascular capillaries, which supply oxygen.
Mitochondria in close proximity to vessels were shown to be more oxidized than mitochondria in the center of the cell, consistent with the existence of a gradient in oxygen tension across the cell, at rest.
In vivo observations of intracellular signaling and metabolic responses during cellular processes are complicated by the lack of robust methods to compensate for tissue motion associated with physiological perturbations in bulk tissue. We report a strategy to compensate for tissue motion using an image-guided dynamic microscope and retrospective image analysis. These imaging methods expand possible targets for in vivo intracellular observations to new tissue preparations and physiological experiments, while improving image quality. Using these methods, in vivo measurements of the subcellular metabolic response to hypoxia were made in mouse skeletal muscle for the first time. Micrometer-resolution images of mitochondrial NAD(P)H autofluorescence showed that a large fraction of mitochondria were in close proximity to the exogenously dyed capillaries and that resting mitochondrial redox state was proportional to distance from capillaries. These results are consistent with an oxygen gradient, at rest, from the capillaries to the core of the muscle cell. The lower oxygen tension in the core may be an important adaptation to minimize free radical generation in resting muscle and to increase the sensitivity of metabolism to alterations in oxygen delivery.
Original received October 24, 2009; revision received February 8, 2010; accepted February 10, 2010.