Chapter 8Functional Imaging with Mitochondrial Flavoprotein Autofluorescence: Theory, Practice, and Applications

Husson TR, Issa NP.

Publication Details

8.1. INTRODUCTION

While great strides have been made in the development of novel optical imaging techniques in the past two decades, these methods still either measure coarsely or require relatively invasive procedures. The ideal optical measure would faithfully represent local neural activity, both spiking and nonspiking, at subcellular resolution, with persistent, reliable responses without the use of exogenous agents. Practically, there are several opportunities to measure endogenous signals that could serve as indirect measures of neural activity: most commonly, metabolic processes which take advantage of the tight coupling between neural activity and ATP production. Unfortunately, the most robust metabolic signals (such as blood deoxygenation) are only weakly linked to neural activity, and can be spatially diffuse, putting strict limitations on the interpretation of these data.

Recently, the endogenous fluorescence of mitochondrial flavoproteins (FA, flavoprotein autofluorescence) has been exploited for functional imaging in a variety of preparations, both in vivo and in vitro. These flavoproteins are oxidized during aerobic metabolism, which is closely coupled to neuronal signaling, and their fluorescence has been shown to follow neural activity in response to pharmacological, electrical, and sensory stimulation. They have been used to confirm functional maps in several cortical areas, and are supported by a growing body of in vitro work. As these signals (1) are highly spatially localized, (2) can be measured by a variety of optical techniques both in vitro and in vivo, and (3) require no exogenous dyes, they offer substantial improvement in our ability to map functional activity and trace neural circuits. Here, we describe the underlying biochemistry of the flavoprotein signal, the experimental details of using it for functional imaging in vivo, and outline example studies and future directions for this promising new approach to visualizing neural activity patterns.

8.2. SOURCES OF FLAVOPROTEIN AUTOFLUORESCENCE SIGNALS

8.2.1. The Relationship between Neural Activity and Neuronal Metabolism

Neurons, from a metabolic perspective, are exceptionally expensive cells to maintain—so much so that metabolic demand has placed a significant constraint on brain evolution. The massive expansion in gray matter volume during the emergence of the hominids is thought to have occurred after the acquisition of a higher quality diet and the concomitant reduction in gut size (another metabolically expensive tissue) necessary to increase brain volume (the Expensive Tissue Hypothesis [1]). This energetic cost may even constrain neural processing. Some studies suggest that the way in which neurons encode information is designed to minimize energy demand (reviewed in Reference [2]), and that an optimal coding scheme represents a balance between the metabolic cost of an action potential and the fixed cost of maintaining a neuron at resting potential: when this ratio is low the most energy-efficient system requires few neurons that each encode large amounts of information with a high average firing rate, but when it is high (spikes are metabolically expensive) the balance is shifted towards sparse coding schemes. It is clear that the metabolic needs of neurons are enormous and may be an important factor in the brain’s development and function.

Even during a resting state (a period of relatively low-action potential generation), neurons must actively maintain precise ion gradients, preserve complicated morphologies, recycle neurotransmitter, and maintain postsynaptic receptor distribution and function, all of which require large amounts of ATP [3]. In order to meet this demand, neurons are equipped with a high density of mitochondria (17.3% of cellular volume, as compared to 11.0% in astrocytes and 11.3% in oligodendrocytes [4]), highest in regions involved with synaptic action such as dendrites (62% of neuronal mitochondria are found in dendrites) and axons/presynaptic terminals (36%) [3,5]. To keep these mitochondria running, the brain must consume a disproportionately large quantity of metabolic substrates, specifically oxygen and glucose from arterial blood. Despite representing only 2% of body mass, the brain consumes 20% of available oxygen and 25% of the body’s glucose [4,6].

This enormous demand is exacerbated by increased neural signaling. Attwell and Laughlin’s estimates for the rat gray matter suggest that as much as 81% of ATP production in neurons is devoted to electrical activity, most of which is accounted for by postsynaptic effects of glutamate (53%) and a substantial amount is allotted to axonal and presynaptic processes (42%) [3]. They estimate that a 1 Hz increase in the firing rate of an excitatory cortical neuron (with a background rate of 4 Hz) requires a 16% increase in ATP production (though this relationship may not be linear at higher rates). Most of this ATP is consumed by the sodium-potassium ATPase, but many other electrochemical mechanisms consume significant amounts of ATP as well (including glutamate recycling, presynaptic calcium extrusion, etc).

Despite the large variability in metabolic demand, ATP levels in neurons are surprisingly constant in vivo [7]; thus, there must be a rapid, direct upregulation of ATP production in response to neural activity. Within active cerebral cortex, increased neural activity requires large amounts of glucose and oxygen. As a result, oxygen is drawn from the local blood supply, desaturating hemoglobin [8–10] and inducing a subsequent spatially diffuse and long-lasting reperfusion of the active area by local arterial vasodilation [11–13]. If the demand for oxygen following increased activity is not met, perhaps due to damage of the local vasculature as occurs during a stroke, ATP depletion causes failure of Na/K ATPase and loss of transmembrane ion gradients; in cortex, all electrical activity ceases within seconds of ischemic injury [14,15]. Clearly, neuronal signaling is immediately dependent on an adequate supply of oxygen and glucose [14,16–18].

The rapid use of oxygen immediately after neural activity suggests that a substantial fraction of activity-induced metabolism is aerobic and involves activation of the mitochondrial tricarboxylic acid cycle (TCA). The TCA cycle activity is therefore well correlated with the energy demands of active neurons, and could serve as a highly sensitive metabolic reporter of electrical activity. Because several complexes within the TCA cycle and electron-transport chain contain flavoproteins whose fluorescence changes with oxidation state, the metabolic changes induced by activity are readily accessible in the modulation of neuronal autofluorescence. In the next section we review the sources and properties of the FA signal, highlighting its advantages for use in functional imaging.

8.2.2. Origin of the Flavoprotein Autofluorescence Signals

8.2.2.1. Review of Brain Metabolism and Glucose Utilization

Unlike other tissues in the body, the brain is nearly completely reliant on one form of energy production. Stoichiometric relationships between oxygen consumption (160 μmol/100 g/min), and carbon dioxide (identical to oxygen) suggest that the respiratory quotient of the brain is very nearly 1, indicative of pure carbohydrate metabolism [19]. Indeed, its primary energy substrates are glucose (31 μmol/100 g/min) and ketone bodies, and there is very little beta-oxidation of fatty acids [7,20,21]. Ketone bodies are thought to be metabolized only when glucose is unavailable, such as during starvation [22–24], so efforts to understand brain metabolism have focused on glycolytic and glycogenolytic processes [7]. Glycogen, the only stored energy source in the brain, only exists in small amounts and is mostly stored in astrocytes; it is thought to only become available to neurons during increased activity or hypoglycemia [25]. Therefore, the only major energy substrate for neurons is arterial glucose, making the brain almost entirely reliant on the local vasculature to accommodate its large metabolic demands.

Glucose metabolism in the brain involves several stages where fluorescent molecules are oxidized or reduced, leading to putative imaging signals (Figure 8.1). Glucose is first converted to pyruvate in the cytoplasm during glycolysis, a process that reduces two fluorescent nicotinamide adenine dinucleotide (NAD+) molecules to NADH (uses of this fluorescence change are discussed elsewhere, see Reference [26] and Reference [27], and Section 8.2.3 below). This pyruvate is carried into the mitochondria and converted to acetyl coenzyme-A via the pyruvate dehydrogenase complex (PDC), a series of enzymes that includes a flavoprotein adenine dinucleotide (FAD) moiety in the dihydrolipoyl dehydrogenase active site, which when oxidized in the presence of blue visible light emits green fluorescence (characterized below, see Section 8.3.1). The activity of this enzyme involves the reduction of NAD+ and the double-oxidation of FADH2, thus lowering NADH autofluorescence and increasing flavoprotein autofluorescence [28–30]. The resulting acetyl coenzyme-A can then enter the TCA cycle. The TCA cycle converts three more NAD+ molecules to NADH, which then enter the electron transport chain (ETC), donating their electrons to generate the chemiosmotic gradient that drives ATP synthase’s conversion of ADP to ATP; it is this activity of the TCA cycle that directly corresponds to the aerobic production of ATP in cells.

FIGURE 8.1. Mitochondrial sources of fluorescence.

FIGURE 8.1

Mitochondrial sources of fluorescence. Fluorescent imaging signals derive from several stages of aerobic cellular metabolism. In response to increased neural activity, ATP hydrolysis (1) increases and the elevated ADP concentration upregulates the production (more...)

FIGURE 8.2. Cellular sources of flavoprotein autofluorescence.

FIGURE 8.2

Cellular sources of flavoprotein autofluorescence. Flavoprotein autofluorescence signals derive from oxidative phosphorylation and therefore localize primarily to mitochondria in active neuronal tissue. Glucose is removed from arterial blood by the GLUT-1 transporter in astrocytes and the GLUT-3 transporter in neurons. Neuronal glucose undergoes glycolysis and pyruvate enters the mitochondria for oxidative phosphorylation of ADP. Astrocytic glucose can be converted to glycogen for storage or enter glycolysis. Crucially, astrocytic metabolism at synapses is glycolytic, not oxidative, as there are no mitochondria in astrocytic filopodia. Astrocytic glycolysis may produce lactate which could then be shuttled to neurons where it would be converted to pyruvate and fed into the TCA cycle.

The first site of the electron transport chain, complex I (NADH ubiquinone oxidoreductase), is a membrane-bound protein that converts two NADH to NAD+ and donates their electrons to a lipid-soluble carrier (ubiquinone). This protein complex contains another fluorescent flavin prosthetic group, flavin mononucleotide (FMN). Complex II (succinate dehydrogenase) also contains an FAD moiety that oxidizes succinate to fumarate, and becomes fluorescent as a result. It is the combined fluorescence of these two flavoproteins (ETC complexes with flavin prosthetic groups) and the FAD associated with the PDC that underlies the flavoprotein autofluorescence signal used in brain imaging studies (though the FMN contribution is likely negligible [31]).

While many cellular enzymes contain flavin cofactors, it is well established that the autofluorescence of ETC flavoproteins and α-lipoamide dehydrogenase (from the PDC) are the primary sources of cellular autofluorescence [28–30,32,33]. The excitation and emission spectra typically used in functional brain imaging studies (Figure 8.3A,B) closely coincide with those of alpha-lipoamide dehydrogenase and ETC FAD fluorescence; although, based on spectrophotometry of flavin fluorescence signals, ETC flavoprotein fluorescence may not be well represented in the brain [31] (Figure 8.3C). When Shibuki and colleagues [34] showed that the FA signal is indeed a result of flavoproteins (discussed below, see Section 8.4.3), they could not distinguish between ETC or PDC flavoproteins because they used a nonspecific flavoprotein inhibitor (diphenyleneiodonium). Further study is needed to fully differentiate among these signals in brain in vivo; but, importantly, whether or not the flavoprotein signal is derived from ETC or PDC, it is restricted to the mitochondrial matrix and measures oxidative phosphorylation only, not glycolysis or any other activity-dependent cellular functions.

FIGURE 8.3. Excitation and emission spectra of flavoprotein autofluorescence.

FIGURE 8.3

Excitation and emission spectra of flavoprotein autofluorescence. (A) The excitation spectra of electron transport chain flavins (FAD) and lipoamide dehydrogenase (from the PDC) are very similar, showing two peaks including a maximum at approximately (more...)

8.2.2.2. Localization of the Signal in Cells and Tissue

ATP diffuses slowly through the cytoplasm, so a general feature of cellular organization is that mitochondria are localized to regions of high ATP demand [35,36]. The distribution of mitochondria in neocortex is, then, unsurprising, given the high metabolic demand of electrochemical signaling: in primate visual cortex, 62% of mitochondria are localized to dendrites, 21% to presynaptic terminals, and 12% to axons (3% are in neuronal soma, leaving a mere 2% to glia) [5,6,37]. This suggests that metabolic processes in neural tissue are mostly devoted to the maintenance of ionic gradients and the generation and propagation of electrical potentials. Attwell and Laughlin’s energy budget for excitatory neurons estimated that 47% of ATP was consumed in processes related to the generation of action potentials, 34% to postsynaptic potentials, 13% to the resting membrane potential, and 3% each to glutamate recycling and presynaptic calcium flux [3]. These calculations strongly suggest that mitochondrial ATP production is spatially correlated with electrochemical activity and is primarily localized to synaptic sites.

Unfortunately, there is still no consensus on the pre- or post-synaptic source of the flavoprotein signal. Shibuki et al. [34] blocked postsynaptic potentials in cortical slices with CNQX and only partially abolished the FA signal, suggesting that presynaptic and postsynaptic activity contributed approximately equally to the overall signal. Reinert et al. [38] used CNQX in the cerebellum in vivo and found a 92% decrease in the amplitude of the light phase response, suggesting an almost entirely postsynaptic origin. This mismatch may reflect differences in glutamate receptor distribution in the cerebrum and cerebellum, or could be a result of metabolic differences in vivo versus in vitro. Both groups found that using a calcium-free bath reduced FA responses almost completely, which is consistent with a greater dependence on synaptic events than on action potentials, per se. It is therefore unclear whether the FA signal reflects presynaptic or postsynaptic activity or a combination of both.

Despite the much greater mitochondrial content of neurons, glial cells might also contribute to brain energy metabolism. There is evidence suggesting that the metabolic demands of sustained activity eventually outstrip a neuron’s ability to generate ATP, and that astrocytes can be recruited to help sustain neural activity. While it is known that ~95% of ATP in the brain is produced by mitochondrial aerobic metabolism and only ~1–5% by glycolysis [37], positron emission tomography (PET) studies by Fox et al. [39] noted a mismatch among signals interpreted as metabolic rate of oxygen consumption, glucose consumption, and blood flow, a phenomenon referred to as metabolic uncoupling. This suggested that, at rest, neurons are at or near their oxidative limit, so stimulus-evoked activity must very quickly recruit other sources of substrates for the TCA cycle. It has been proposed that astrocytes, which can detect increased activity through glutamate uptake at excitatory synapses, help sustain neural activity by generating excess lactate through glycolysis that is then “shuttled” to neurons, converted into pyruvate via lactate dehydrogenase, and fed into the TCA cycle [19,40,41]. While neural activity does lead to increased lactate production, whether this lactate can be taken up or used by neurons remains contentious [42–45]. This debate has important consequences for neurometabolic coupling, because it means that the relationship between the use of metabolic substrates and electrical activity may depend critically on glia. Astrocytes are now known to participate in functional properties of synaptic transmission, and there is evidence that their metabolic response to sensory stimulation has been greatly underestimated (reviewed in Reference [42]). For example, it is known that glycogen stored in astrocytes is made available to neurons during periods of hypoglycemia or increased activity, and could be transported to neurons either as lactate or glucose [46–48]. Similarly, it is now well established that astrocytes respond to local neuronal activity by increasing aerobic glycolysis—a process that has been both measured chemically and imaged [8,40]. In general, it seems that astrocytes metabolically support neuronal activity by providing substrates for activity-dependent increases in oxidative metabolism.

The precise contribution from glia in metabolic optical signals, however, is not yet clear. As mentioned above, only 2% of mitochondria in the brain are glial, suggesting their contribution to mitochondrial signals is small, and these mitochondria are mostly restricted to cell bodies due to the narrow width of astrocytic filopodia [42,49]. Several studies indicate that increased neural activity induces increased glycolysis in astrocytes, not oxidative phosphorylation, though astrocytes certainly have oxidative capacity [42]. Because FA is a measure of oxidative phosphorylation only, it should not modulate in astrocytes during changes in neural activity; conversely, NADH autofluorescence will decrease because glycolysis, astrocytic or neuronal, results in the reduction of some NAD+ to NADH. FA will also increase during any subsequent oxidative phosphorylation of “shuttled” lactate, and be directly related to any aerobic generation of ATP involved with the generation of action potentials (interestingly, Shibuki et al. [34] found that adding 50 mM L-lactate to the agar well used in vivo nearly doubled the FA response amplitude). Indeed, Kashiske et al. found that astrocytes emit high levels of background NADH autofluorescence and responded to neural activity with increased glycolysis, not oxidative phosphorylation [8]. In addition, it has been shown that in fibroblasts the NADH signal is not restricted to mitochondria, but show significant signal from other structures, especially lysosomes [26,50] (Figure 8.5). Thus, flavoprotein autofluorescence is a more direct indicator of neuronal ATP production, whereas NADH autofluorescence may represent a combination of both neuronal and glial metabolism.

FIGURE 8.5. Multiphoton imaging of flavoprotein and NAD(P)H autofluorescence in cardiomyocytes.

FIGURE 8.5

Multiphoton imaging of flavoprotein and NAD(P)H autofluorescence in cardiomyocytes. (A) NAD(P)H autofluorescence multiphoton imaging of canine cardiomyocyte (410–490 nm emission). (B) Flavoprotein autofluorescence multiphoton imaging of same cardiomyocyte (more...)

Converging evidence suggests that there is a very close relationship between metabolic activity and neural signaling, and that neurons must allocate energy resources very carefully in order to participate in functional activity. Also, because the energy demands of firing action potentials are immediate, increases in neuronal firing must be followed by proportional increases in ATP production in quick succession, or else the underlying homeostatic mechanisms required to keep neurons alive and functional will quickly fail. The question for physiologists then becomes, what are these underlying metabolic processes, and how can they be measured? As we will see, there are several opportunities for measuring local metabolic activity in the brain, which because of this tight neurometabolic coupling can be used as indirect measures of neural activity.

8.2.3. Imaging Opportunities from Neuronal Metabolism

Given the tight relationship between neural firing rate and energy metabolism, it has become increasingly popular to use the latter as a measure of the former. Ubiquitous well-established imaging techniques such as functional magnetic resonance imaging (fMRI) and positron-emission tomography (PET) rely on metabolic signals as measures of functional activity, but are limited in spatiotemporal resolution and have difficulty resolving precise neural activity patterns. There are several opportunities to measure the consumption of metabolic substrates in vivo using optical techniques; these offer much greater spatial and temporal resolution and can be used on both large (whole cortical area) and small (cellular circuit) levels. These techniques are all nondestructive, in that they preserve in vivo connections, do not require the use of exogenous agents, and can be used simultaneously with each other and in combination with standard electrophysiological techniques. Each has particular advantages and disadvantages, but either alone or in concert offers ways of examining neural activity on both large and small spatial scales with relatively little disruption of the underlying system.

Increased metabolic action requires a continuous and adequate supply of oxygen, and thus oxygenated hemoglobin. The rate at which tissue consumes oxygen can be measured optically due to the difference in absorption spectra between oxygenated and deoxygenated hemoglobin (the “intrinsic signal” [51,52]). However, the relationship between blood flow and neural activity is highly nonlinear. Directly after neural stimulation, oxygen tension in the stimulated area drops considerably, followed by a large, diffuse increase corresponding to wide-scale vasodilation [11,12,53]. This gives rise to the familiar “initial dip” seen in blood oxygen-level dependent (BOLD) imaging techniques such as intrinsic signal imaging (ISI) and fMRI, and the subsequent long, nonlocalized increase in signal that necessitates a great deal of spatial filtering and signal averaging [54]. Also, because local tissue perfusion takes place in the extra-cellular space, the theoretical limit to the spatial resolution of ISI is much larger than single cells. The size of the hemodynamic signal can be appreciated by considering spatial correlations in unfiltered intrinsic signal images; using this metric, we found that the hemodynamic signal encompasses an area approximately 30% larger than that delineated by flavoprotein autofluorescence [54]. There is also substantial variation in small-scale vasculature between cortical areas [55], which could represent a problem for comparing functional activation between areas. However, ISI signals are very robust and require less illumination with high-energy light [54], suggesting that this technique may be best suited for large-scale mapping of cortical regions in which illumination is limited or precise spatiotemporal information is not critical.

Another opportunity to use metabolic signals as a proxy for neural activity is the fluorescence of an electron-carrying molecule found in all cells, NADH. NAD+ is reduced in both glycolysis and oxidative phosphorylation (see Section 8.2.2.1 above), causing small decreases in fluorescence that can be measured as a percent change from baseline. Importantly, this signal has been used to obtain high resolution (single-cell) images of functional activation in vivo by the use of two-photon microscopy, which was combined with both local field potential and oxygen tension measurements to produce a time-resolved view of spreading depression in hippocampal slices [8]. This study utilized at once many of the advantages offered by optical imaging—namely, simultaneous physiological measurements, single-cell resolution, and resolved temporal information. Unfortunately, we are not aware of any in vivo experiments using NADH autofluorescence with two-photon microscopy, so its usefulness for functional mapping has yet to be established. Also, as mentioned above (see Section 8.2.2.1), NADH fluorescence contains a significant glycolytic component, which may be useful for differentiating cell types on small scales, but which may overrepresent the metabolism of astrocytes over neurons over larger areas or in thicker tissue. This sentiment is echoed by Kann and Kovacs [56], who argue that NADH signals must be interpreted with care due to astrocytic and extramitochondrial contamination.

A more direct measure of activity-dependent neuronal metabolism would only measure oxidative phosphorylation, which increases in neurons but not astrocytes during neural activity. FA represents the best opportunity in this regard, and its signal properties have been well established both in vitro and in vivo. While its signal strength is relatively small compared with ISI and NADH [26,27,57], and appears to be weaker in larger mammals than in rodents [54,58], it is measurable using standard optical imaging techniques in both mouse and cat. FA functional maps have been confirmed with ISI, and in fact show some superiority in spatial resolution even at large scales, and have been used to confirm maps of spatial frequency first measured with ISI but thought to be contaminated by hemodynamic artifacts (discussed below, see Section 8.4.2). The precise spatial resolution of FA signals has not been measured in neurons, but in prepared cardiomyocytes FA was able to resolve single mitochondria [26]. There is an immediate need for the characterization of the flavoprotein signal in neurons with two-photon microscopy, ideally in direct comparison with NADH autofluorescence. Evidence to date suggests that FA has the particular advantages of being restricted to subcellular compartments, representing neuronal (not glial) metabolism, and having the ability to measure both large-scale functional maps and small-scale neural circuits with the same optical signal.

It was initially thought that flavoprotein and NADH signals were exactly the same except opposite in sign, indicating that they both measured redox state with a strongly mitochondrial source [59]. While the flavoprotein time to peak and dark-phase temporally match the dip and prolonged light-phase of NADH autofluorescence, more precise measurements of these signals have revealed important differences [34,38,60]: First, the FA signal is more dependent on tissue oxygen availability than NADH signals, providing further evidence that flavoprotein signals represent oxidative and not glycolytic activity. Second, NADH and FA are differentially affected by metabolic uncouplers (which separate the ETC from the phosphorylation of ATP), suggesting that the signals can be pharmacologically isolated. Third, while both signals show increased time to peak and eliminated delayed phases in response to blocked Ca++ entry, simply decreasing extracellular calcium preferentially inhibits the flavoprotein signal. Finally, as mentioned above (see Section 8.2.2.2), the NADH signal shows a large contribution from both astrocytes and nonmitochondrial cellular elements. Thus. while the two signals measure similar metabolic processes, they should not be considered substitutes, but rather strategic complements.

In general, endogenous metabolic signals are an underutilized tool in visualizing the dynamics and spatial organization of neural activity. Many opportunities exist both in vitro and in vivo to use these complementary techniques, especially together, to better understand neurometabolic coupling, the relationship between neurons and glia, local and distributed networks, functional aspects of neural circuits, and pathological conditions (discussed in Section 8.5.1 below).

8.3. METHODS AND SIGNAL PROPERTIES

8.3.1. Equipment and Methodology

FA is distinguished from other endogenous optical signals (such as hemoglobin and NADH) by its particular excitation and emission spectra (see Figure 8.3). Loosely, flavoproteins emit green fluorescence in blue illumination; specifically, their excitation wavelengths are mainly between 420–490 nm with a maximum at 450 nm [26,56,61] and emit photons over a wide range of wavelengths above 500 nm [31,32,62,63]. Thus, for optical imaging purposes, excitation filter bandwidths typically emphasize the 450–490 nm range while emissions filters are typically long-pass above 500 nm (Takahashi et al. [64] use a 470–490 excitation filter and 500–550 nm band-pass filter, presumably to reduce the possibility of hemodynamic artifacts at longer wavelengths). It should be noted that some autofluorescence is emitted at lower than 500 nm, and is usually not captured with these filters. Both macroscopic tandem-lens arrangements (described in Reference [54]) or standard epifluorescence microscopes (as used in Reference [34]) have been used in vivo with similar results.

As with other optical imaging techniques, the animal must be secured in a stereotaxic apparatus, and every effort must be made to reduce mechanical disturbances. Typically, we stabilize the exposed brain area with 3% agarose in sterile saline and seal it with a glass coverslip to reduce mechanical disturbances and ensure a level optical plane. Alternatively, a metal or acrylic chamber can be placed around the exposed area and filled with carbogenated Ringer’s solution (as in Reference [38]). Some groups use a paralytic agent and artificial ventilation in mouse experiments to further reduce motion artifacts, but we have found this only to be necessary in larger animals.

The apparatus for FA imaging is remarkably similar to those used for ISI or dye-based optical imaging (Figure 8.4); practically, the only substantial difference is in the excitation and emission filter characteristics [54]. However, several properties of the FA signal require modifications in order to optimize detectability. First, illumination intensity must be relatively high, usually higher than that achieved by standard 12W halogen light sources. In our lab, we simply added an additional halogen light source and achieved suitable illumination, but other groups have reported using 75 W xenon and 100 W mercury–xenon lamps [38,65]. Second, due to the use of a long-pass emission filter and the low levels of fluorescence, light from external sources (such as a stimulus monitor used for vision studies) is a much larger contaminant than in ISI; the chamber surrounding the preparation must be completely opaque and every effort must be made to minimize ambient light. Finally, because of FA’s relatively low signal strength, it is often useful to have long integration times (hundreds of ms), and therefore low frame rates (5–10 fps), as well as long interstimulus intervals to ensure the late-phase reduction in fluorescence returns to baseline. The degree to which this is necessary will, like the previous two issues, depend greatly on the details of the preparation and are much more of a nuisance in vivo than in vitro, where the signal to noise ratio is much higher.

FIGURE 8.4. Imaging rig for FA.

FIGURE 8.4

Imaging rig for FA. Standard tandem-lens combination for macroscopic imaging. (A) CCD camera attached to stereotaxic apparatus. (B) Top lens. (D) Bottom lens (typically, a 50 mm lens). The ratio of focal lengths of the top and bottom lenses sets the magnification (more...)

8.3.2. Properties of the FA Signal

8.3.2.1. Signal Strength

Signal strength also varies both by brain region and species. Electrical stimulation of the parallel fiber system in the cerebellum elicited strong responses (2% F/F) and prominent inhibitory bands (see Section 8.3.2.5 below); responses in visual cortex are larger (5%) and easily detectable in response to sensory stimulation (0.25%). Initial reports suggested that the FA signal was extremely weak in large mammals and that functional imaging could only realistically be performed in rodents [58]. In fact, our lab has been able to generate and confirm strong FA responses in the cat, including maps of orientation and spatial frequency preference [54,66]. It is clear, however, that the signal strength is much weaker in the cat than in the mouse (0.5% F/F as opposed to 5% F/F in the mouse in response to electrical stimulation).

Because FA signals are relatively weak in vivo, most stimulation protocols must be repeated several times in order to improve signal-to-noise ratio with signal averaging, similar to hemodynamic imaging techniques. By applying the phase-encoding Fourier imaging technique of Kalatsky and Stryker [67,68] we are able to obtain reasonable FA retinotopic maps in the mouse with as few as 20 repetitions, comparable to that for ISI. However, our estimates of the signal-to-noise ratio for mapping cat spatial frequency maps suggests that the ISI signal-to-noise is approximately 1.3 times that of FA, which could be a limiting factor in these data [66]. While it is clear that the signal strength varies significantly by preparation, its responses are comparable and often superior to those of other imaging techniques.

8.3.2.2. Spatial Resolution

The question of the spatiotemporal resolution of FA is a complex one. On the spatial side, there is reason to believe that FA responses are restricted to single cells, or even single mitochondria: two-photon studies of cardiomyocytes reveal distinct mitochondrial flavoprotein autofluorescence [26] (Figure 8.5), matching our understanding of the localization of the signal to inner mitochondrial membranes (see Section 8.2.2.1). However, macroscopic studies are limited by optical scatter through neural tissue and the optics of the imaging system, and measurements of the optical point-spread function in vivo are exceptionally difficult (what constitutes a “point” of activity, and how can one be elicited without exciting surrounding tissue, is not at all clear). Our lab has found that iso-orientation domain maps generated with FA have an approximately 30% smaller autocorrelation peak than ISI, suggesting that FA has improved spatial resolution on the scale of orientation domains or smaller (~0.3 mm) (Figure 8.6D,E). Weber et al. [69] demonstrate that hemodynamic maps of barrel cortex show an approximately 2.5 times larger response area than autofluorescence maps, suggesting that FA has comparatively fine spatial resolution. Given our understanding of the biochemistry of FA, and accepting in vitro studies as an indicator of signal source, we believe FA to offer a substantial improvement in spatial resolution over previous techniques for in vivo functional imaging studies.

FIGURE 8.6. Comparison of FA and ISI signals over active cortex.

FIGURE 8.6

Comparison of FA and ISI signals over active cortex. (A) Image of cat primary visual cortex under green (530–550 nm) illumination to highlight vascular pattern. (B) Standard deviation of the FA signal over an orientation mapping session—the (more...)

8.3.2.3. Time Course and Temporal Resolution

Several groups have measured the optical impulse response of FA in a variety of in vivo preparations, summarized in Figure 8.7. In general, there is a large positive peak in fluorescence followed by a longer negative dip; the extent and duration of these “light” and “dark” phases depends greatly on the particular preparation and stimulus protocol. In primary visual cortex of anesthetized mice, one second of electrical stimulation evokes a strong light phase response that peaks 1 second from the onset of the stimulus, and a dark phase with a minimum at approximately 5 seconds. Ten seconds of electrical stimulation of the cerebellar cortex evokes a similarly quick light phase but longer dark phase. Visual stimulation with a 200 ms flash of light produces similar results, but with a reduced signal amplitude and thus SNR (Figure 8.7A). In the cat, visual stimulation of an orientation domain with its preferred drifting grating showed unimodal FA responses with fast rise and decay times, in sharp contrast with ISI responses typically used to generate orientation preference maps. Electrical stimulation in cat cortex showed a very similar response time as seen in mice (Figure 8.7A).

FIGURE 8.7. Time course of FA signal.

FIGURE 8.7

Time course of FA signal. (A) Impulse response function of FA in response to electrical stimulation (200 uA, 200 Hz, 200 ms) in mouse (black trace) and cat (grey trace) cortex. Also shown is the response in mouse to a brief (100 ms) flash of light (dotted (more...)

Extrapolating temporal resolution from the impulse-response function requires an assumption of linearity (summation of optical signals), an assumption that may or may not hold in different parts of the CNS. Therefore, despite a well-characterized impulse-response function, it is not clear exactly how quickly oxidative phosphorylation can follow neural activity. In our hands we have seen FA responses modulate with a periodic visual stimulus up to 0.3 Hz, but due to low light levels at small integration times, faster changes may be visible with stronger illumination systems. It has been shown that background FA levels decrease with decreasing interstimulus intervals of electrical stimulation, in line with a more reduced redox potential over a longer time course [70]. It is therefore difficult to calculate a specific temporal resolution; but, given known comparisons between FA and ISI time courses, we can safely assume it can follow changes at least as quickly as hemodynamic responses.

8.3.2.4. Vascular Artifacts

One of the most significant problems with hemodynamic imaging techniques is their susceptibility to vascular artifacts [10,51,55,71]. Such arguments have been made against maps of spatial frequency preference in primary visual cortex, suggesting that these responses do not represent the organization of spatial frequency preference but are artifacts of the underlying vasculature [72]. Usually, functional maps must be subjected to extensive templating (exclusion of pixels over blood vessels), but because hemodynamic signals modulate over blood vessels they often report functional selectivity to stimuli despite being nonneural tissue, making templating very difficult and the possibility of inaccurate mapping very real. This represents a major constraint on the use and interpretation of functional maps generated with BOLD intrinsic signals.

Flavoprotein autofluorescence, on the other hand, measures nonhemodynamic signals. The appearance of blood vessels does not modulate in FA images, and therefore blood vessels do not appear tuned for an arbitrary stimulus. Indeed, we have shown that the variation in the autofluorescence signal is greatest over neural tissue [54], whereas the modulation in ISI signals is strongest over blood vessels (Figure 8.6B,C). This strongly suggests that FA functional maps do not represent the local vascular pattern but rather the response of cortical neurons. We have recently used this reduction in vascular artifacts to confirm the structure of spatial frequency maps in cat primary visual cortex [66].

In mice, the lateral spread of hemodynamic signals can lead to significant challenges in functional mapping studies. It has been noted by several studies [54,65,69] that the retinotopic maps of visual cortex in mice are smaller with FA than with ISI, suggesting that the large nonspecific blood flow response is misrepresenting the detailed structure of this area. In studies of ocular dominance plasticity, it is crucial to make accurate measurements of the size of both monocular and binocular zones of visual cortex [73]; with ISI, these structures may appear much larger and more overlapped than they are. FA’s spatially restricted response profile (Figure 8.7c,d) may be much better suited to make these fine spatial measurements.

8.3.2.5. Linearity

Estimates of the metabolic cost of increased firing rate suggest a linear relationship, with an approximate production of 3–5 × 108 ATP molecules per action potential [3]. Thus, as a direct measure of ATP production, we would expect FA to also show a linear increase in fluorescence in response to increased neural activity. To assess this, we have used electrical stimulation to increase the amount of neural activity over an imaged region of cortex and found that this relationship is approximately linear for reasonable stimulus ranges (under 100 uA), and plateaus at very high amplitudes (data not shown). Reinert et al. [38] also find that FA is approximately linear with stimulus amplitude in the cerebellum, and Shibuki et al. [34] find that FA response amplitudes (but not size of responsive area) is linear with stimulation frequency. These findings suggest that FA can linearly follow increases in neural activity over a wide range of activity levels.

Moreover, autofluorescence can also decrease when inhibitory circuits are activated. Gao et al. [74] studied autofluorescence in the cerebellum, in which Purkinje cells have very high baseline firing rates that can be modulated both upward by parallel fiber input and downward by inhibitory inputs from GABAergic molecular layer interneurons. With electrical stimulation of these parallel fibers at 50–200 μA two regions change their fluorescence: there is a strong increase in fluorescence along the parallel fiber input, and a pronounced decrease in fluorescence in parasagittal zones that would be inhibited by molecular layer interneurons. The reduction in fluorescence in parasagittal zones was abolished by a GABAA receptor blockade, suggesting that it depends on inhibitory neurotransmission. Immunostaining of the fixed cerebellar tissue collected after autofluorescence imaging showed that the parasagittal regions in which fluorescence decreased corresponded to parasagittal zebrin-II compartments [75], suggesting that these neighboring functional zones respond oppositely to parallel fiber inputs. This was the first demonstration that FA could be used to image not only excitatory but also inhibitory activity.

8.3.2.6. Analysis Techniques

Sensory-driven response amplitudes are small compared to baseline fluorescence, so some sort of normalization is needed to visualize fluorescence responses. Many of the normalization techniques in standard use with other imaging methods, like intrinsic signal imaging, can be easily adapted to FA [67,76]. For example, we use the same analysis tools for processing FA and ISI images from cat and mouse visual cortex, with the obvious caveat that one or the other signal must be inverted if the two signals are to be compared directly.

However, because FA does not modulate over blood vessels, templating procedures are much more straightforward than with ISI, but care must be taken in the construction of stimulus preference maps. With FA, the response over blood vessels in mean-subtracted images is generally negative for any stimulus that activates neural tissue [66]. The maximal response for pixels over blood vessels, therefore, appears as the condition that stimulates neural tissue the least, and is therefore least affected by high-pass spatial filtering. Thus, if blood vessels are not excluded from analysis, the pixels over them could be mistakenly assigned to low-responsive conditions. Fortunately, in highly responsive conditions, blood vessels appear as dark regions, clearly distinguished from neural tissue. Using these conditions to construct templates (e.g., by a threshold) prevents the mischaracterization of pixels over blood vessels as can occur in hemodynamic preference maps.

Because the autofluorescence signal has been known to reversibly bleach [77], for repetitive stimulation protocols it may be advantageous to extrapolate a “bleaching curve” and subtract it from the optical trace [60]. This can be accomplished by measuring baseline responses during each interstimulus interval and using the changes in these blanks to fit an exponential decay. In addition, for direct comparison between flavoprotein and either absorption (ISI) or NADH signals, the signals must be inverted; in the case of a periodic representation (such as that for retinotopic maps in the mouse), this can be accomplished by phase-shifting the response map by 180° (as performed in Reference [78]). Some studies using NADH autofluorescence in vivo [79] have attempted to eliminate hemodynamic artifacts by subtracting fluorescence at an isobestic wavelength of oxy/deoxyhemoglobin, but we have not found this to be necessary even in cats.

8.3.2.7. Possible Pitfalls

As a relatively new technique for functional imaging, FA still requires some care in properly obtaining and interpreting data. Several obstacles that have been encountered remain unexplained, and further study is needed to both understand the details of the FA signal and to appreciate its applicability to various preparations.

Until recently there appeared to be a strong discrepancy between in vivo and in vitro measures of neuronal redox state with optical techniques. Specifically, in vivo signals seemed to follow sustained activity much more accurately than in vitro measures, which consisted of a much more profound dark phase recovery period. Turner et al. [27] reviewed this issue at length and concluded that differences in tissue oxygenation between intact brains (which have local vasculature) and slices (which rely entirely on perfusion by the bath media) explain these differences and bring the two signals in line with one another. Interestingly, the Ebner group has found that inhibiting astrocytic TCA activity with fluoroacetate caused a stronger attenuation of the dark phase than the light phase [80], suggesting that the dark phase is mostly nonneuronal. Given the exciting work done in slice preparations and the outstanding spatial resolution of FA in prepared tissue (discussed below, see Section 8.4.3), this reconciliation between in vitro and in vivo signals means they can be used as complementary methods in the same model system.

A more practical concern has also arisen due to the nature of flavin fluorescence. Lower excitation wavelengths (blue), corresponding to higher energy light, can cause photodamage in neural tissue at much lower intensities than other wavelengths such as those used for ISI. High energy light generates singlet oxygen, a strong oxidizing agent, which can lead to damage of lipid membranes and cell death [81,82]. While no reports exist of photodamage to neural tissue in macroscopic applications of FA, and responses persist even after several hours of imaging, in these and microscopic studies (including two-photon microscopy) the intense illumination requirements of FA need to be balanced against the risk of photodamage and photobleaching [83–86]. Given that FA time courses in response to electrical stimulation require a long dark phase recovery period (see Figure 8.7A,B), there seems to be some indication that bleaching of the flavoprotein signal can occur; also, a recent study has used this illumination light to focally inactivate neural tissue [77]. It thus seems advisable to control the excitation light source such that it only illuminates the tissue while imaging and at the lowest possible intensity.

Another potential problem with FA is that by using a long-pass emission filter hemodynamic signals could contaminate images. It is known, for example, that hemoglobin can absorb excitation light and quench emitted fluorescence [87]. Our group has done a systematic comparison between hemodynamic and autofluorescence time courses in vivo (Figure 8.8) and found that, as expected, in response to electrical stimulation the hemodynamic (intrinsic) signals are opposite in sign, corresponding to light absorption rather than fluorescence, and are slightly slower in time-to-peak than FA. The responses are similar when stimulating visually with a bright flash, though much lower in intensity. Also, Reinert et al. blocked hemodynamic signals by inhibiting nitric oxide synthase with L-NAME and showed that FA responses were preserved [38]. This suggests that the FA signals are not due to hemodynamic artifacts, and that well-separated filters can adequately restrict the signal to flavoprotein autofluorescence.

FIGURE 8.8. Autofluorescence and intrinsic signal time courses.

FIGURE 8.8

Autofluorescence and intrinsic signal time courses. FA and absorption signal time courses averaged over 22 cycles of electrical stimulation in mouse. Each plot is the response of one pixel measured using (A) the FA signal (420–490 nm excitation (more...)

While FA has been reliably elicited in anesthetized mice and cats, it is not clear what effect anesthesia has on these responses. It is our experience that urethane or ketamine/urethane mixtures are very reliable in mice [54,64,65], but both ketamine/xylazine and pentobarbital have also been successfully used [38,65]. Our lab has only rarely imaged responses under isoflurane anesthesia, and have had little success with alpha-cholorolase induction protocols, though it is not clear what effect these substances would have on mitochondrial activity. In cats, we used sodium thiopental regularly, and were able to see activation with isoflurane in ferrets. A systematic study of the effects of anesthesia protocols on FA responses would not only better inform researcher’s choice for in vivo studies, but especially in combination with other metabolic imaging techniques could enlighten the investigation of these anesethetics’ mechanisms of action, as many common choices are known to be vasoactive or metabolic depressants (for example, barbiturates inhibit the oxidation of NADH, see Reference [88] and references therein).

8.4. EXAMPLE STUDIES

8.4.1. Early Studies

The pioneering studies of fluorescent intracellular molecules and their relationship to energy utilization began almost 50 years ago by Britton Chance, who first used flavin, NADH, and cytochrome fluorescence to trace the process of oxidative metabolism [29,62,63,89–95]. These studies found that both ADP and calcium ions increased the redox activity of these fluorophores, suggesting a strong link between their activity and mitochondrial ATP production. These signals were later used to study both functional activation (mostly in response to electrical stimulation) in cell cultures and slices from the hippocampus and cerebral cortex [59,96,97] and pathological responses to spreading depression and epilepsy [87,98–100]. The development and availability of high quality CCD cameras allowed these relatively small signals to be used in a variety of preparations, beginning with the work of Mironov and Richter [96] in the brainstem, Shuttleworth et al. [57] in the hippocampus, Shibuki et al. [34] in the somatosensory cortex, and Reinert et al. [38] in the cerebellum. These studies not only showed reliable and robust responses in several preparations, but also extended FA’s functional imaging abilities to in vivo preparations.

8.4.2. The Development of FA Imaging In Vivo

Shuttleworth et al. [57] imaged NAD(P)H (both NADH and NADPH) and flavin fluorescence in response to electrical and chemical stimulation in slice preparations of mouse hippocampal area CA1. They characterized their respective time courses, both consisting of an initial, stimulus-linked component and an aftershoot (as in Figure 8.7A,B). These responses were abolished by the ionotropic glutamate receptor blockade, suggesting that the signal was predominantly postsynaptic, and the authors concluded that while these signals were not explicitly coupled to calcium ion flux they were reliable measures of neural activity.

Dr. Katsuei Shibuki and colleagues extended these findings to the rat somatosensory cortex (see the chapter by Shibuki et al. in this book). In slices, they found large increases in FA (27 +/− 2%) in response to electrical stimulation using a highly sensitive cooled CCD camera and wide-view epifluorescence microscope. Taking this robust setup forward, they were the first to show stimulus-evoked activity patterns with FA in anesthetized mice. They used both electrical and vibratory skin stimulation to characterize the FA responses under urethane anesthesia and distinguished them from hemodynamic signals under the same conditions. This was a crucial step in proving the difference between ISI and FA responses and showing the viability of the signal in vivo.

The Ebner Lab at the University of Minnesota followed these results by showing strong in vivo responses in the mouse cerebellum [38]. They used a well-understood neuronal circuit, the parallel fiber/Purkinje cell system, which upon electrical stimulation showed a well-defined band of fluorescence along the parallel fiber tract (Figure 8.9A,B; these responses were confirmed shortly thereafter by Reference [101]). Their signal had the same time course shown previously by Shuttleworth and Shibuki, with an initial bright phase followed by a longer dark phase, and signal strength was linearly related to stimulus amplitude and frequency. They also used pharmacological treatments similar to those used by Shibuki et al. in slice (see Section 8.4.3 below) to show that the in vivo fluorescence was a result of flavoprotein activity, and found that the optimal wavelengths of excitation and emission light matched those previously reported for flavoproteins. In subsequent studies, this group has shown that electrical stimulation of parallel fibers evokes parasagittal inhibitory bands that are abolished by GABAA antagonists (Figure 8.9), reflected in FA imaging as dark bands that run alongside the bright parallel fibers [74]. Their use of FA in this robust and stereotyped system demonstrates the ability of FA to identify both excitatory and inhibitory activity patterns and to trace neural circuits over large areas in vivo.

FIGURE 8.9. Imaging inhibition with FA.

FIGURE 8.9

Imaging inhibition with FA. (A) Image of mouse cerebellar cortex; the stimulating electrode is apparent. (B) Pseudocolor image of FA response to electrical stimulation in cerebellum. Fluorescence increased along the linear band (red), but decreased in (more...)

One of the primary advantages of optical imaging techniques is that they allow the visualization of changes in spatial activity patterns over time. The Shibuki lab quickly followed up their characterization of the FA signal by showing that it can be used to study short-term plasticity in the somatosensory cortex in vivo [102]. They showed that optical responses to somatosensory stimulation were reliably enhanced if they were followed by deep-layer (1.5–2 mm from cortical surface) electrical stimulation in the responsive region; likewise, more superficial (<0.5 mm) electrical stimulation attenuated responses. Also, they could greatly potentiate ipsilateral responses with ipsilateral stimulation, but could also potentiate contralateral responses, suggesting a commissural projection affecting this plasticity. Subsequently, they mapped tono-topic organization of primary auditory cortex and the anterior auditory field, and demonstrated differences in FA response durations between mice raised in sound-deprived versus normal acoustic environments and used the in vitro slice preparation to find specific response differences between the vertical layers of cortex [64]. Because they could measure the same flavoprotein fluorescence signal both in vivo and in vitro, they could use the complementary advantages of each to create a wholistic view of these experience-dependent changes. Finally, they have shown that FA can reliably measure ocular dominance plasticity in the mouse visual cortex [65], one of the most studied assays of experience-dependent changes in neural circuitry. In sum, the Shibuki group has demonstrated that FA can be used to functionally map several areas of the mouse brain, measure experience-dependent plasticity, and link large-scale activity to specific circuits in vitro.

It was thought, however, that the only viable in vivo preparation for FA was the mouse. Indeed, the difference in signal strength between sliced (~20% changes in F/F) and anesthetized mouse cortex (~2%) is enormous, and preliminary reports indicated that the difference between small and large mammals is also large [58]. Our group, however, found that FA could be reliably used to image not only the retinotopic organization of mouse visual cortex (Figure 8.10A–D) but also the classic orientation maps in cat (Figure 8.10E–H) [54]. These maps were very similar to those generated using intrinsic signal imaging, but were more spatially localized and required less spatial filtering of raw data (due to the correlation structure shown in Figure 8.6D–E). Moreover, we have used this greater signal specificity to confirm the structure of spatial frequency maps in the cat [66], which have been called into question due to the possibility of vascular artifacts in ISI maps. Because FA is a nonhemodynamic measure of neural activity, the similarity between FA and ISI spatial frequency maps suggests that the patchy structure of spatial frequency domains observed by functional imaging is a genuine feature of the organization of primary visual cortex.

FIGURE 8.10. Functional imaging of primary visual cortex with FA.

FIGURE 8.10

Functional imaging of primary visual cortex with FA. (A,B) Responses over visual cortex in mouse in response to a drifting visual bar. Regions that responded to the stimulus showed increased brightness with FA imaging, and reduced brightness (increased (more...)

Over the past 5 years, FA has been shown to be a reliable and relatively noninvasive measure of neural activity patterns in several brain regions that is linear with stimulus intensity and sensitive to evoked activity. Experiments are currently under way to bring FA’s spatial specificity and tighter link to neural activity into more complicated systems, such as primate sensory and motor cortex. There is a great deal of hope that this technique will be an improvement on previous techniques, such as hemodynamic and voltage sensitive dye imaging, and that new opportunities will emerge to study the details of neuron–glia and neurometabolic coupling in intact animals.

8.4.3. Functional Imaging In Vitro

One of the fundamental advantages offered by FA is its ability to measure functional activity both in vivo and in vitro with the same underlying signal. This allows an unprecedented degree of interchange between preparations—findings at the level of cortical areas can be “zoomed in” on by further slice preparations, and findings in slices or cultures can be extended to intact brains. Because of this advantage and the extensive work done on FA in slices, we here review several studies that have been performed in vitro, and discuss their relationship with findings in live animals. In general, these techniques greatly improve our understanding of the FA signal and have outstanding potential for revealing the detailed physiology that underlies larger-scale brain activity.

Huang et al. have characterized the autofluorescence of both NAD(P)H and flavoproteins in cardiomyocytes using two-photon laser scanning microscopy (2P) [26]. Their 2P excitation and emission spectra match previously reported characteristics (see Section 8.3.1) and showed appropriate polarity (blocking oxidative phosphorylation via cyanide increases flavoprotein fluorescence and decreases NAD(P)H fluorescence). Importantly, they were able to identify individual mitochondria with limited signal averaging (five trials) and limited photodamage (due to the longer wavelengths used in 2P). At this resolution, they were able to differentiate between cytosolic and mitochondrial NAD(P)H signals, with the latter appearing as distinct “hot spots” amongst weaker background fluorescence; measuring flavoprotein autofluorescence, they saw only these “hot spots” (which were aligned with those seen in NAD(P) H imaging), providing direct evidence that FA is restricted to single mitochondria (Figure 8.5). This differentiation allows for better calculation of FAD/NAD(P)H ratios, which could be done over individual mitochondria (and may be valuable for measuring disease states). While isolated cardiomyocytes are a far cry from in vivo brain preparations, the mitochondrial energy demand of the heart is very large and is comparable to that in brain, so results in neurons may prove similar [6]. More recently, they have employed 2P techniques in rat hippocampal slices to study the metabolic response to hypoxia, but only using NADH autofluorescence [103]. This work provides at least a proof of concept that FA can spatially resolve individual mitochondria.

The Shibuki group has contributed a wide variety of studies using flavoprotein autofluorescence in cortical slices. In order to deflnitively isolate the endogenous signal, they performed pharmacological treatments and compared local field potential recordings to optical traces [34]. Abolishing all activity with tetrodotoxin blocked both the optical and field potential recording. Reduced Ca++ in the bath eliminated postsynaptic activity in the field potential as well as optical responses. Reducing both glucose and oxygen content also restored neural activity but not FA, and finally the flavoprotein blocker diphenyleneiodonium was shown to completely abolish the optical responses while completely preserving electrical activity in the slice (though diphenyleneiodonium is a widely used flavin inhibitor, there is controversy over its precise effects, see Reference [104]). These treatments strongly implicate flavoproteins in the signals obtained measuring green fluorescence with blue excitation light (these findings were subsequently confirmed by Reference [38], except the finding of mixed pre- and postsynaptic origin [see Section 8.2.2.2]). They have since studied intracortical activity patterns in several preparations, and have demonstrated functional, experience-dependent connections between primary and secondary somatosensory cortex [105] and between primary auditory and surrounding areas [106]. Their repeated success in applying FA to several brain regions and many functional assays highlights the robustness of the method.

8.5. CONCLUSIONS AND FUTURE DIRECTIONS

The use of mitochondrial flavoprotein autofluorescence in the last few years has demonstrated substantial advantages for functional imaging, yet its full utilization has yet to begin. There is reason to believe that further improvements in spatiotemporal resolution are possible, and work towards this end can benefit from recent advances in imaging technology. Even with established techniques and preparations, FA opens the door to new lines of investigation that could have widespread implications for fields ranging from neurochemistry and neurometabolic coupling to functional imaging of local and global circuits.

However, several investigations are immediately required to fully understand the source and resolution of the FA signal. Preliminary reports with two-photon microscopy in cardiac muscle suggest subcellular resolution (Figure 8.5), but these results have not been extended to neurons either in vitro or in vivo. In theory, FA could offer an improvement due to its subcellular localization and endogenous source, and thus might be particularly suited to long-term study of synaptic changes (such as in Reference [107]). Though the underlying signal is small, FA may stand to benefit from recent advances in optical imaging technology that allow for whole-field illumination with minimal photobleaching [108]. As the mechanical modifications to traditional two-photon imaging systems for FA are small, and because FA responses have been measured with such techniques in other cell types [50,109–111], these applications may not be far away.

FA could also be used to map functional circuits in vivo. For example, the demonstration that FA can be used to identify fine cortical structures in large animals suggests that FA could be used to visualize functional interactions between columnar structures such as orientation domains in visual cortex with either pharmacological, electrical, or sensory stimulation, or several in combination. The lack of long-range spatial correlations in FA maps [54] makes clear the improvement over ISI for these studies. In addition, FA could also help unravel the highly contentious issue of neurometabolic coupling. As it serves as a faithful indicator of redox state and ISI measures local tissue deoxygenation, a combination of the two techniques could help visualize the changes that occur in neural tissue as a result of activity or insult.

Fortunately, recent advances in our understanding of neurometabolic coupling suggest that metabolic signals may be very useful in studying various types of functional plasticity (reviewed in Reference [112]). It has been shown that neuron–glia interactions are modified with changes in synaptic plasticity, suggesting that FA, with its presumptive restriction to synapses (see Section 8.2.2.2) may be able to indirectly monitor these changes. Specifically, long-term changes in the functional properties of synapses such as LTD and LTP are supported by related changes in the metabolic processes at those synapses, as there is a great density of mitochondria at synapses, FA in combination with two-photon microscopy may be able to build on previous long-term imaging studies of synaptic plasticity in vivo. On larger scales, FA may prove a more subtle measure of developmental plasticity than ISI, which has been used, for example, to investigate the effects of monocular deprivation in developing mice and cats [65,73,113–118].

Perhaps the most promising application of FA is in the realm of translational neuroscience. It is now appreciated that a large number of neurological conditions are associated with mitochondrial dysfunction, including amyotrophic lateral sclerosis, Alzheimer’s, Parkinson’s, and Huntington’s diseases [119–122]. It may be possible to use FA in transgenic mouse models of these diseases to better assess the regional and functional deficits across neural tissue. For example, Parkinson’s disease can be modeled by specific inhibition of complex I of the mitochondrial ETC [123–125]. As FA is a direct measure of ETC activity, these mice can be used both in vivo and in vitro to study local or global alterations in functional activity as well as changes in activity brought about by pharmacological treatments. Also, studies have already exploited the ratio between FAD and NADH autofluorescence to grade precancerous epithelial cells [126]; perhaps similar approaches could be applied to the study of gliomas or neuromas in cortex, which already have acceptable animal models [127,128]. In these and many other ways, FA could serve as a powerful assay of metabolic disease.

The imaging of mitochondrial flavoprotein autofluorescence is a powerful yet underappreciated tool for measuring many aspects of brain function. It adds another way of visualizing the interactions between metabolism and neural activity, and serves as a complement to preexisting techniques that measure other relevant signals. In a perfect world, we would be able to measure every aspect of neural activity simultaneously with precise resolution; while this is still very much a fantasy, we now have a variety of techniques that can be used in combination to understand the complex interactions and changes in the brain. FA brings us one step closer to a direct visualization of these processes.

ACKNOWLEDGMENTS

We would like to thank Dr. Murray Sherman and members of the Sherman lab for helpful discussions, and members of the Issa lab, especially Atul Mallik and Ari Rosenberg, for extensive discussion and review. Work in our lab was supported by grants from the Brain Research Foundation and Mallinckrodt Foundation (NPI).

REFERENCES

1.
Aiello LC, Wheeler P. The expensive-tissue hypothesis: The brain and the digestive system in human and primate evolution. Curr Anthropol. 1995;36(2):199.
2.
Laughlin SB. Energy as a constraint on the coding and processing of sensory information. Curr Opin Neurobiol. 2001;11(4):475–480. [PubMed: 11502395]
3.
Attwell D, Laughlin SB. An energy budget for signaling in the grey matter of the brain. J Cereb Blood Flow Metab. 2001;21(10):1133–1145. [PubMed: 11598490]
4.
Foster KA, et al. Optical and pharmacological tools to investigate the role of mitochondria during oxidative stress and neurodegeneration. Prog Neurobiol. 2006;79(3):136–171. [PMC free article: PMC1994087] [PubMed: 16920246]
5.
Wong-Riley MT. Cytochrome oxidase: An endogenous metabolic marker for neuronal activity. Trends Neurosci. 1989;12(3):94–101. [PubMed: 2469224]
6.
Rolfe DF, Brown GC. Cellular energy utilization and molecular origin of standard metabolic rate in mammals. Physiol Rev. 1997;77(3):731–758. [PubMed: 9234964]
7.
McKenna MC, et al. Energy metabolism of the brain. In: Siegel GJ, et al., editors. Basic Neurochemistry: Molecular, Cellular, and Medical Aspects. Elsevier; Academic Press; 2006. pp. 531–557.
8.
Kasischke KA, et al. Neural activity triggers neuronal oxidative metabolism followed by astrocytic glycolysis. Science. 2004;305(5680):99–103. [PubMed: 15232110]
9.
Thompson JK, et al. Single-neuron activity and tissue oxygenation in the cerebral cortex. Science. 2003;299(5609):1070–1072. [PubMed: 12586942]
10.
Malonek D, Grinvald A. Interactions between electrical activity and cortical microcirculation revealed by imaging spectroscopy: Implications for functional brain mapping. Science. 1996;272(5261):551–554. [PubMed: 8614805]
11.
Logothetis NK. The neural basis of the blood-oxygen-level-dependent functional magnetic resonance imaging signal. Philos Trans R Soc Lond B Biol Sci. 2002;357(1424):1003–1037. [PMC free article: PMC1693017] [PubMed: 12217171]
12.
Logothetis NK, Wandell BA. Interpreting the BOLD signal. Annu Rev Physiol. 2004;66:735–769. [PubMed: 14977420]
13.
Vanzetta I, Grinvald A. Increased cortical oxidative metabolism due to sensory stimulation: Implications for functional brain imaging. Science. 1999;286(5444):1555–1558. [PubMed: 10567261]
14.
Dugan LL, Kim-Han JS. Hypoxic-Ischemic Brain Injury and Oxidative Stress. In: Siegel GJ, et al., editors. Basic Neurochemistry: Molecular, Cellular, and Medical Aspects. Elsevier; Academic Press; 2006. pp. 559–572.
15.
Kristian T, Siesjo BK. Changes in ionic fluxes during cerebral ischaemia. Int Rev Neurobiol. 1997;40:27–45. [PubMed: 8989615]
16.
Schiff S, Somjen G. The effect of graded hypoxia on the hippocampal slice: An in vitro model of the ischemic penumbra. Stroke. 1987;18(1):30–37. [PubMed: 3027927]
17.
Schurr A, Rigor BM. Cerebral ischemia revisited: New insights as revealed using in vitro brain slice preparations. Cellular and Molecular Life Sciences (CMLS) 1989;45(8):684–695. [PubMed: 2668016]
18.
Schurr A, et al. Brain lactate, not glucose, fuels the recovery of synaptic function from hypoxia upon reoxygenation: An in vitro study. Brain Res. 1997;744(1):105–111. [PubMed: 9030418]
19.
Magistretti PJ, Pellerin L. Cellular bases of brain energy metabolism and their relevance to functional brain imaging: Evidence for a prominent role of astrocytes. Cereb Cortex. 1996;6(1):50–61. [PubMed: 8670638]
20.
Page MA, Williamson DH. Activity of ketone-body utilization pathway in brain of suckling and adult rats. Biochem J. 1971;121(1):16P. [PMC free article: PMC1176526] [PubMed: 5116530]
21.
Yang S, et al. Fatty acid oxidation in rat brain is limited by the low activity of 3- ketoacyl-coenzyme A thiolase. J Biol Chem. 1987;262(27):13027–13032. [PubMed: 3654601]
22.
Nehlig A, Pereira de Vasconcelos A. Glucose and ketone body utilization by the brain of neonatal rats. Progr Neurobiol. 1993;40(2):163–220. [PubMed: 8430212]
23.
Hawkins RA, et al. Regional ketone body utilization by rat brain in starvation and diabetes. Am J Physiol Endocrinol Metab. 1986;250(2):E169–178. [PubMed: 2937307]
24.
Ruderman NB, et al. Regulation of glucose and ketone-body metabolism in brain of anaesthetized rats. Biochem J. 1974;138(1):1–10. [PMC free article: PMC1166169] [PubMed: 4275704]
25.
Oz G, et al. Human brain glycogen content and metabolism: Implications on its role in brain energy metabolism. Am J Physiol Endocrinol Metab. 2007;292(3):E946–951. [PubMed: 17132822]
26.
Huang S, et al. Two-Photon Fluorescence Spectroscopy and Microscopy of NAD(P)H and Flavoprotein. Biophys J. 2002;82(5):2811–2825. [PMC free article: PMC1302068] [PubMed: 11964266]
27.
Turner DA, et al. Differences in O(2) availability resolve the apparent discrepancies in metabolic intrinsic optical signals in vivo and in vitro. Trends Neurosci. 2007;30(8):390–398. [PMC free article: PMC3340602] [PubMed: 17590447]
28.
Kunz WS, Kunz W. Contribution of different enzymes to flavoprotein fluorescence of isolated rat liver mitochondria. Biochim Biophys Acta. 1985;841(3):237–246. [PubMed: 4027266]
29.
Hassinen I, Chance B. Oxidation-reduction properties of the mitochondrial flavoprotein chain. Biochem Biophys Res Commun. 1968;31(6):895–900. [PubMed: 4299232]
30.
Voltti H, Hassinen IE. Oxidation-reduction midpoint potentials of mitochondrial flavoproteins and their intramitochondrial localization. J Bioenerg Biomembr. 1978;10(1):45–58. [PubMed: 555461]
31.
Kunz WS, Gellerich FN. Quantification of the content of fluorescent flavoproteins in mitochondria from liver, kidney cortex, skeletal muscle, and brain. Biochem Med Metab Biol. 1993;50(1):103–110. [PubMed: 8373630]
32.
Chorvat D Jr, et al. Spectral unmixing of flavin autofluorescence components in cardiac myocytes. Biophys J. 2005;89(6):L55–57. [PMC free article: PMC1367005] [PubMed: 16227502]
33.
Aubin JE. Autofluorescence of viable cultured mammalian cells. J Histochem Cytochem. 1979;27(1):36–43. [PubMed: 220325]
34.
Shibuki K, et al. Dynamic imaging of somatosensory cortical activity in the rat visualized by flavoprotein autofluorescence. J Physiol. 2003;549(pt. 3):919–927. [PMC free article: PMC2342977] [PubMed: 12730344]
35.
Ames A. CNS energy metabolism as related to function. Brain Res Rev. 2000;34(1-2):42–68. [PubMed: 11086186]
36.
Jones DP. Intracellular diffusion gradients of O2 and ATP. Am J Physiol Cell Physiol. 1986;250(5):C663–675. [PubMed: 3010727]
37.
Erecinska M, Silver IA. ATP and brain function. J Cereb Blood Flow Metab. 1989;9(1):2–19. [PubMed: 2642915]
38.
Reinert KC, et al. Flavoprotein autofluorescence imaging of neuronal activation in the cerebellar cortex in vivo. J Neurophysiol. 2004;92(1):199–211. [PubMed: 14985415]
39.
Fox P, et al. Nonoxidative glucose consumption during focal physiologic neural activity. Science. 1988;241(4864):462–464. [PubMed: 3260686]
40.
Pellerin L, Magistretti PJ. Glutamate uptake into astrocytes stimulates aerobic glycolysis: A mechanism coupling neuronal activity to glucose utilization. Proc Natl Acad Sci USA. 1994;91(22):10625–10629. [PMC free article: PMC45074] [PubMed: 7938003]
41.
Pellerin L, et al. Evidence Supporting the Existence of an Activity-Dependent Astrocyte-Neuron Lactate Shuttle. Dev Neurosci. 1998;20(4-5):291–299. [PubMed: 9778565]
42.
Hertz L, et al. Energy metabolism in astrocytes: High rate of oxidative metabolism and spatiotemporal dependence on glycolysis/glycogenolysis. J Cereb Blood Flow Metab. 2007;27(2):219–249. [PubMed: 16835632]
43.
Mangia S, et al. The aerobic brain: Lactate decrease at the onset of neural activity. Neuroscience. 2003;118(1):7–10. [PubMed: 12676131]
44.
Mangia S, et al. Issues concerning the construction of a metabolic model for neuronal activation. J Neurosci Res. 2003;71(4):463–467. [PubMed: 12548701]
45.
Caesar K, et al. Glutamate receptor-dependent increments in lactate, glucose and oxygen metabolism evoked in rat cerebellum in vivo. J Physiol. 2008;586(5):1337–1349. [PMC free article: PMC2375663] [PubMed: 18187464]
46.
Brown AM, et al. Energy transfer from astrocytes to axons: The role of CNS glycogen. Neurochem Int: Role of Non-synaptic Communication in Information Processing. 2004;45(4):529–536. [PubMed: 15186919]
47.
Brown AM. Brain glycogen re-awakened. J Neurochem. 2004;89(3):537–552. [PubMed: 15086511]
48.
Dringen R, et al. Glycogen in astrocytes: Possible function as lactate supply for neighboring cells. Brain Res. 1993;623(2):208–214. [PubMed: 8221102]
49.
Wolff JR, Chao TI. Cytoarchitectonics of non-neuronal cells in the central nervous system. Adv Mol Cell Biol. 2004;31:1–52.
50.
Andersson, et al. Autofluorescence of living cells. J Microscopy. 1998;191:1–7. [PubMed: 9723186]
51.
Frostig RD, et al. Cortical functional architecture and local coupling between neuronal activity and the microcirculation revealed by in vivo high-resolution optical imaging of intrinsic signals. Proc Natl Acad Sci USA. 1990;87(16):6082–6086. [PMC free article: PMC54476] [PubMed: 2117272]
52.
Grinvald A, et al. Functional architecture of cortex revealed by optical imaging of intrinsic signals. Nature. 1986;324(6095):361–364. [PubMed: 3785405]
53.
Takano T, et al. Cortical spreading depression causes and coincides with tissue hypoxia. Nat Neurosci. 2007;10(6):754–762. [PubMed: 17468748]
54.
Husson TR, et al. Functional imaging of primary visual cortex using flavoprotein autofluorescence. J Neurosci. 2007;27(32):8665–8675. [PubMed: 17687044]
55.
Harrison RV, et al. Blood capillary distribution correlates with hemodynamic-based functional imaging in cerebral cortex. Cereb Cortex. 2002;12(3):225–233. [PubMed: 11839597]
56.
Kann O, Kovacs R. Mitochondria and neuronal activity. Am J Physiol Cell Physiol. 2007;292(2):C641–657. [PubMed: 17092996]
57.
Shuttleworth CW, et al. NAD(P)H fluorescence imaging of postsynaptic neuronal activation in murine hippocampal Slices. J Neurosci. 2003;23(8):3196–3208. [PubMed: 12716927]
58.
Shibuki K, Hishida R. Physiology Online. Vol. 2005. The Physiological Society; 2004. Flavoprotein autofluorescence.
59.
Duchen MR. Ca(2+)-dependent changes in the mitochondrial energetics in single dissociated mouse sensory neurons. Biochem J. 1992;283(pt. 1):41–50. [PMC free article: PMC1130990] [PubMed: 1373604]
60.
Kosterin P, et al. Changes in FAD and NADH Fluorescence in Neurosecretory Terminals Are Triggered by Calcium Entry and by ADP Production. J Membr Biol. 2005;208(2):113–124. [PubMed: 16645741]
61.
Benson RC, et al. Cellular autofluorescence—is it due to flavins? J Histochem Cytochem. 1979;27(1):44–48. [PubMed: 438504]
62.
Chance B, et al. Intracellular oxidation-reduction states in vivo. Science. 1962;137:499–508. [PubMed: 13878016]
63.
Chance B, et al. Metabolically linked changes in fluorescence emission spectra of cortex of rat brain, kidney and adrenal gland. Nature. 1962;195:1073–1075. [PubMed: 13878020]
64.
Takahashi K, et al. Transcranial fluorescence imaging of auditory cortical plasticity regulated by acoustic environments in mice. Eur J Neurosci. 2006;23(5):1365–1376. [PubMed: 16553797]
65.
Tohmi M, et al. Enduring critical period plasticity visualized by transcranial flavoprotein imaging in mouse primary visual cortex. J Neurosci. 2006;26(45):11775–11785. [PubMed: 17093098]
66.
Mallik AK, et al. The organization of spatial frequency maps measured by cortical flavoprotein autofluorescence. Vision Res. 2008;48:14, 1545–1553. [PMC free article: PMC2543938] [PubMed: 18511098]
67.
Kalatsky VA, Stryker MP. New paradigm for optical imaging: Temporally encoded maps of intrinsic signal. Neuron. 2003;38(4):529–545. [PubMed: 12765606]
68.
Kalatsky VA, et al. Fine functional organization of auditory cortex revealed by Fourier optical imaging. Proc Natl Acad Sci USA. 2005;102:37, 13325–13330. [PMC free article: PMC1201601] [PubMed: 16141342]
69.
Weber B, et al. Optical imaging of the spatiotemporal dynamics of cerebral blood flow and oxidative metabolism in the rat barrel cortex. Eur J Neurosci. 2004;20(10):2664–2670. [PubMed: 15548209]
70.
Brennan AM, et al. Modulation of the amplitude of NAD(P)H fluorescence transients after synaptic stimulation. J Neurosci Res. 2007;85(15):3233–3243. [PubMed: 17497703]
71.
Erinjeri JP, Woolsey TA. Spatial integration of vascular changes with neural activity in mouse cortex. J Cereb Blood Flow Metab. 2002;22(3):353–360. [PubMed: 11891441]
72.
Sirovich L, Uglesich R. The organization of orientation and spatial frequency in primary visual cortex. Proc Natl Acad Sci USA. 2004;101(48):16941–16946. [PMC free article: PMC534713] [PubMed: 15550540]
73.
Cang J, et al. Optical imaging of the intrinsic signal as a measure of cortical plasticity in the mouse. Vis Neurosci. 2005;22(5):685–691. [PMC free article: PMC2553096] [PubMed: 16332279]
74.
Gao W, et al. Cerebellar cortical molecular layer inhibition is organized in parasagittal zones. J Neurosci. 2006;26(32):8377–8387. [PubMed: 16899733]
75.
Sillitoe RV, et al. Zebrin II compartmentation of the cerebellum in a basal insectivore, the Madagascan hedgehog tenrec Echinops telfairi. J Anat. 2003;203(3):283–296. [PMC free article: PMC1571161] [PubMed: 14529046]
76.
Bonhoeffer T, Grinvald A. Optical imaging based on intrinsic signals. In: Toga AW, Mazziotta JC, editors. Brain Mapping: The Methods. Academic Press; 1996. pp. 55–97.
77.
Kubota Y, et al. Transcranial photo-inactivation of neural activities in the mouse auditory cortex. Neurosci Res. 2008;60(4):422–430. [PubMed: 18291543]
78.
Husson TR, et al. Autofluorescence imaging in mouse and cat visual cortex. Soc Neurosci. 2006;32:640–649.
79.
Harbig K, et al. In vivo measurement of pyridine nucleotide fluorescence from cat brain cortex. J Appl Physiol. 1976;41(4):480–488. [PubMed: 186025]
80.
Reinert KC, et al. Flavoprotein autofluorescence imaging in the cerebellar cortex in vivo. J Neurosci Res. 2007;85(15):3221–3232. [PubMed: 17520745]
81.
Kolosov MS, et al. Photodynamic injury of isolated neuron and satellite glial cells: Morphological study. IEEE J. Selected Topics in Quantum Electronics. 2003;9(2):337–342.
82.
Oleinick NL, Evans HH. The Photobiology of Photodynamic Therapy: Cellular Targets and Mechanisms. Radiation Res. 1998;150(5):S146–156. [PubMed: 9806617]
83.
Sacconi L, et al. Overcoming photodamage in second-harmonic generation microscopy: Real-time optical recording of neuronal action potentials. Proc Natl Acad Sci USA. 2006;103(9):3124–3129. [PMC free article: PMC1413939] [PubMed: 16488972]
84.
Hopt A, Neher E. Highly nonlinear photodamage in two-photon fluorescence microscopy. Biophys J. 2001;80(4):2029–2036. [PMC free article: PMC1301392] [PubMed: 11259316]
85.
Patterson GH, Piston DW. Photobleaching in two-photon excitation microscopy. Biophys J. 2000;78(4):2159–2162. [PMC free article: PMC1300807] [PubMed: 10733993]
86.
Schonle A, Hell SW. Heating by absorption in the focus of an objective lens. Opt Lett. 1998;23(5):325–327. [PubMed: 18084500]
87.
Jobsis FF, et al. Intracellular redox changes in functioning cerebral cortex. I. Metabolic effects of epileptiform activity. J Neurophysiol. 1971;34(5):735–749. [PubMed: 4398562]
88.
Choi IY, et al. Effect of deep pentobarbital anesthesia on neurotransmitter metabolism in vivo: On the correlation of total glucose consumption with glutamatergic action. J Cereb Blood Flow Metab. 2002;22(11):1343–1351. [PubMed: 12439292]
89.
Chance B, Williams GR. A method for the localization of sites for oxidative phosphorylation. Nature. 1955;176(4475):250–254. [PubMed: 13244669]
90.
Chance B, et al. Respiratory enzymes in oxidative phosphorylation. V. A mechanism for oxidative phosphorylation. J Biol Chem. 1955;217(1):439–451. [PubMed: 13271406]
91.
Chance B, Williams GR. Respiratory enzymes in oxidative phosphorylation. IV. The respiratory chain. J Biol Chem. 1955;217(1):429–438. [PubMed: 13271405]
92.
Chance B, Williams GR. Respiratory enzymes in oxidative phosphorylation. III. The steady state. J Biol Chem. 1955;217(1):409–427. [PubMed: 13271404]
93.
Chance B, Williams GR. Respiratory enzymes in oxidative phosphorylation. II. Difference spectra. J Biol Chem. 1955;217(1):395–407. [PubMed: 13271403]
94.
Chance B, Williams GR. Respiratory enzymes in oxidative phosphorylation. I. Kinetics of oxygen utilization. J Biol Chem. 1955;217(1):383–393. [PubMed: 13271402]
95.
Chance B, et al. Intracellular oxidation-reduction states in vivo. Science. 1962;137:499–508. [PubMed: 13878016]
96.
Mironov SL, Richter DW. Oscillations and hypoxic changes of mitochondrial variables in neurons of the brainstem respiratory centre of mice. J Physiol. 2001;533(pt. 1):227–236. [PMC free article: PMC2278595] [PubMed: 11351030]
97.
Schuchmann S, et al. Monitoring NAD(P)H autofluorescence to assess mitochondrial metabolic functions in rat hippocampal-entorhinal cortex slices. Brain Res Brain Res Protoc. 2001;7(3):267–276. [PubMed: 11431129]
98.
Hashimoto M, et al. Dynamic changes of NADH fluorescence images and NADH content during spreading depression in the cerebral cortex of gerbils. Brain Res. 2000;872(1-2):294–300. [PubMed: 10924711]
99.
Mayevsky A, Chance B. Repetitive patterns of metabolic changes during cortical spreading depression of the awake rat. Brain Res. 1974;65(3):529–533. [PubMed: 4370335]
100.
O’Connor MJ, et al. Effects of potassium on oxidative metabolism and seizures. Electroencephalogr Clin Neurophysiol. 1973;35(2):205–208. [PubMed: 4124614]
101.
Coutinho V, et al. Functional topology of the mossy fibre-granule cell—Purkinje cell system revealed by imaging of intrinsic fluorescence in mouse cerebellum. Eur J Neurosci. 2004;20(3):740–748. [PubMed: 15255984]
102.
Murakami H, et al. Short-term plasticity visualized with flavoprotein auto-fluorescence in the somatosensory cortex of anaesthetized rats. Eur J Neurosci. 2004;19(5):1352–1360. [PubMed: 15016093]
103.
Vishwasrao HD, et al. Conformational Dependence of Intracellular NADH on Metabolic State Revealed by Associated Fluorescence Anisotropy. J Biol Chem. 2005;280(26):25119–25126. [PubMed: 15863500]
104.
Riganti C, et al. Diphenyleneiodonium inhibits the cell redox metabolism and induces oxidative stress. J Biol Chem. 2004;279(46):47726–47731. [PubMed: 15358777]
105.
Kamatani D, et al. Experience-dependent formation of activity propagation patterns at the somatosensory S1 and S2 boundary in rat cortical slices. Neuroimage. 2007;35(1):47–57. [PubMed: 17234433]
106.
Hishida R, et al. Functional local connections with differential activity-dependence and critical periods surrounding the primary auditory cortex in rat cerebral slices. Neuroimage. 2007;34(2):679–693. [PubMed: 17112744]
107.
Trachtenberg JT, et al. Long-term in vivo imaging of experience-dependent synaptic plasticity in adult cortex. Nature. 2002;420(6917):788–794. [PubMed: 12490942]
108.
Holekamp TF, et al. Fast three-dimensional fluorescence imaging of activity in neural populations by objective-coupled planar illumination microscopy. Neuron. 2008;57(5):661–672. [PubMed: 18341987]
109.
Mulder DJ, et al. Skin autofluorescence, a novel marker for glycemic and oxidative stress-derived advanced glycation endproducts: An overview of current clinical studies, evidence, and limitations. Diabetes Technol Ther. 2006;8(5):523–535. [PubMed: 17037967]
110.
Rocheleau JV, et al. Quantitative NAD(P)H/flavoprotein autofluorescence imaging reveals metabolic mechanisms of pancreatic islet pyruvate response. J Biol Chem. 2004;279(30):31780–31787. [PubMed: 15148320]
111.
Evans ND, et al. Non-invasive glucose monitoring by NAD(P)H autofluorescence spectroscopy in fibroblasts and adipocytes: A model for skin glucose sensing. Diabetes Technol Ther. 2003;5(5):807–816. [PubMed: 14633346]
112.
Magistretti PJ. Neuron-glia metabolic coupling and plasticity. J Exp Biol. 2006;209(pt. 12):2304–2311. [PubMed: 16731806]
113.
Smith SL, Trachtenberg JT. Experience-dependent binocular competition in the visual cortex begins at eye opening. Nat Neurosci. 2007;10(3):370–375. [PubMed: 17293862]
114.
Kaneko M, et al. TrkB kinase is required for recovery, but not loss, of cortical responses following monocular deprivation. Nat Neurosci. 2008;11(4):497–504. [PMC free article: PMC2413329] [PubMed: 18311133]
115.
Crair MC, et al. The role of visual experience in the development of columns in cat visual cortex. Science. 1998;279(5350):566–570. [PMC free article: PMC2453000] [PubMed: 9438851]
116.
Crair MC, et al. Relationship between the ocular dominance and orientation maps in visual cortex of monocularly deprived cats. Neuron. 1997;19(2):307–318. [PubMed: 9292721]
117.
Issa NP, et al. The critical period for ocular dominance plasticity in the Ferret’s visual cortex. J Neurosci. 1999;19(16):6965–6978. [PMC free article: PMC2413141] [PubMed: 10436053]
118.
Frank MG, et al. Sleep enhances plasticity in the developing visual cortex. Neuron. 2001;30(1):275–287. [PubMed: 11343661]
119.
Beal MF. Mitochondrial dysfunction in neurodegenerative diseases. Biochimica et Biophysica Acta (BBA)—Bioenergetics. 1998;1366(1-2):211–223. [PubMed: 9714810]
120.
Rego AC, Oliveira CR. Mitochondrial dysfunction and reactive oxygen species in excitotoxicity and apoptosis: Implications for the pathogenesis of neurodegenerative diseases. Neurochem Res. 2003;28(10):1563–1574. [PubMed: 14570402]
121.
Beal MF. Mitochondrial Dysfunction and Oxidative Damage in Alzheimer’s and Parkinson’s Diseases and Coenzyme Q10 as a Potential Treatment. J Bioenerg Biomembr. 2004;36(4):381–386. [PubMed: 15377876]
122.
Lin MT, Beal MF. Mitochondrial dysfunction and oxidative stress in neurodegenerative diseases. Nature. 2006;443(7113):787–795. [PubMed: 17051205]
123.
Schapira AH, et al. Mitochondrial complex I deficiency in Parkinson’s disease. J Neurochem. 1990;54(3):823–827. [PubMed: 2154550]
124.
Betarbet R, et al. Chronic systemic pesticide exposure reproduces features of Parkinson’s disease. Nat Neurosci. 2000;3(12):1301–1306. [PubMed: 11100151]
125.
Dauer W, Przedborski S. Parkinson’s disease: Mechanisms and models. Neuron. 2003;39(6):889–909. [PubMed: 12971891]
126.
Skala MC, et al. In vivo multiphoton microscopy of NADH and FAD redox states, fluorescence lifetimes, and cellular morphology in precancerous epithelia. Proc Natl Acad Sci USA. 2007;104(49):19494–19499. [PMC free article: PMC2148317] [PubMed: 18042710]
127.
Holland EC. Brain tumor animal models: Importance and progress. Curr Opin Oncol. 2001;13(3):143–147. [PubMed: 11307055]
128.
Gutmann DH, et al. Mouse models of human cancers consortium workshop on nervous system tumors. Cancer Res. 2006;66(1):10–13. [PubMed: 16397207]