Increased Cortical Oxidative Metabolism Due to Sensory Stimulation: Implications for Functional Brain Imaging

See allHide authors and affiliations

Science  19 Nov 1999:
Vol. 286, Issue 5444, pp. 1555-1558
DOI: 10.1126/science.286.5444.1555


Modern functional brain mapping relies on interactions of neuronal electrical activity with the cortical microcirculation. The existence of a highly localized, stimulus-evoked initial deoxygenation has remained a controversy. Here, the activity-dependent oxygen tension changes in the microcirculation were measured directly, using oxygen-dependent phosphorescence quenching of an exogenous indicator. The first event after sensory stimulation was an increase in oxygen consumption, followed by an increase in blood flow. Because oxygen consumption and neuronal activity are colocalized but the delayed blood flow is not, functional magnetic resonance imaging focused on this initial phase will yield much higher spatial resolution, ultimately enabling the noninvasive visualization of fundamental processing modules in the human brain.

Understanding of human cognitive brain function has greatly improved with the emergence of noninvasive methodologies for high-resolution functional neuroimaging, such as positron emission tomography (PET) and functional magnetic resonance imaging (fMRI). These techniques have allowed the rapid identification of functionally distinct cortical areas (1,2). Moreover, the study of the layout of individual cortical modules (3) within a given brain area and their geometrical relationships (4) has become feasible using the much higher spatial resolution of optical imaging based on intrinsic signals. These three methods monitor regional changes in cerebral blood flow and blood oxygenation level, relying on the coupling between local electrical activity and cerebral microcirculation (2, 5,6).

Previous optical imaging experiments have argued that sensory-evoked electrical activity gives rise to a highly localized initial increase in oxygen consumption. This leads to an increase in the concentration of deoxyhemoglobin, followed by a deoxyhemoglobin decrease caused by a delayed blood flow increase, which is not well regulated at the level of cortical functional domains (7). This interpretation has been further supported by optical imaging spectroscopy (8). The initial increase in deoxyhemoglobin concentration, referred to as the initial dip, was recently confirmed by high-field fMRI measurements at 4 to 9.4 T (9,10). However, the large majority of low-field BOLD (blood oxygenation level–dependent) fMRI measurements, as well as other methodologies, have not detected the initial dip (11,12) or have found only a minimal increase in oxygen consumption (6, 13).

This apparent contradiction raises the question of what is actually being measured by the BOLD effect or by optical imaging spectroscopy. Mayhew and colleagues (12) have argued that a problem exists with the interpretation of previous imaging spectroscopy data (8), claiming that the linear model used was oversimplified and may have introduced large errors in the calculated deoxyhemoglobin time course. We share the opinion that the selected model, as well as assumptions about some of its parameters, are indeed critical (14). In addition, our imaging spectroscopy measurements had a time resolution of 500 ms, thus providing timing precision of ∼1 s. This time resolution is not sufficient to monitor the sequence of very early changes in oxy- or deoxyhemoglobin concentration and blood volume (8) and their temporal relationship to blood flow changes (7). Therefore, to resolve this controversy, we decided to bypass imaging spectroscopy by measuring microvascular oxygen concentration changes directly, at a time resolution improved by a factor of 5.

To this end, we measured the phosphorescence decay of Oxyphor R2 [Pd-meso-tetra(4-carboxyphenyl) porphyrin dendrimer] in the cerebral microcirculation. Oxyphor R2 is an oxygen tension–sensitive phosphorescent probe that, once injected intravenously, associates with albumin and thus does not leak out of the microcirculation. This method, introduced by Wilson and colleagues (15–17), is based on the fact that phosphorescence can be quenched by molecular oxygen. Thus, phosphorescence lifetime can be used to directly measure rapid changes in the oxygen concentration in the probe's immediate environment.

To detect the oxygen concentration changes, we first injected anesthetized cats with the oxygen-sensitive phosphorescent probe intravenously at 30 mg/kg of body weight (18). The exposed cortex was implanted with a cranial window and illuminated by brief flashes at 10 Hz (width ∼2 μs at half-height) using standard epi-illumination via the objective lens of the macroscope (19). The emitted phosphorescence was detected by a photomultiplier (Fig. 1A). The known drop in oxygen concentration during barbiturate anesthesia (20) causes a considerably longer lifetime (Fig. 1B). This result indicates that the measurement can be done in the living brain. Remarkably, the small changes in decay time in response to sensory stimulation were reliably detected as well (Fig. 1C). The emergent order in the set of decay curves, each obtained at a different time with respect to the visual stimulus, proves that changes in the phosphorescence decay times can clearly be detected in response to the sensory stimulation. The phosphorescence decay time becomes longer 1.4 s after stimulus onset (Fig. 1D), which suggests that after sensory activation of the cortex, there is a rapid decrease in the microvascular free oxygen concentration. Several alternative interpretations for the observed change in decay time, critically discussed in previous in vivo and in vitro applications of this technique (16, 17), have been ruled out here (21). Therefore, we conclude that the slower decay curve observed after 1.4 s was indeed due to the activity-induced decrease in the microvascular oxygen tension, resulting from an increased oxygen consumption.

Figure 1

In vivo oxygen tension measurement by analysis of phosphorescence decay kinetics. (A) The setup. Drifting gratings are presented in front of the cat's eyes. Brief pulses are flashed onto the exposed cortex (18) through an interference filter (peak transmission 510 nm; bandwidth 10 nm) and are reflected by a dichroic mirror (cutoff at 590 nm) onto a selected region of 2 mm by 2 mm in area 18, devoid of arteries or veins with a diameter of >80 to 100 μm. Flashing rate, 10 Hz. Emitted phosphorescence is collected by a photomultiplier after filtering the light by a 610-nm long-pass filter. The photomultiplier current is used to measure the decay kinetics of the phosphorescence. (B) The flash (black line) and two phosphorescence decay curves. The faster one (red) was recorded when the animal was lightly anesthetized; the slower one (blue), showingreduced blood oxygenation, was recorded after injecting additional anesthetics (until the electroencephalogram was nearly flat). (C) Decay curves (average of 96 trials) obtained before (dark blue), during (light blue-green), and after (yellow-red) the stimulus. Inset: Detail of the decay curves. (D) Expanded display of each of the three phases seen in (C). Dark blue curves: several prestimulus “baseline” decay curves showing the baseline stability. Decay curves above this area (green shading) imply less oxygen, and those below (pink shading) imply more oxygen. The blue curve is a slower decay curve recorded 1.4 s after stimulus onset, showing a decreased oxygen concentration. The red curve was taken 6 s after stimulus onset. It is faster than the baseline decay, showing the delayed hyperoxygenation phase.

To obtain the time course of relative changes in oxygen concentration, we used signal averaging and calculated the decay times for every flash (22). The results (Fig. 2A) show a biphasic curve: Immediately after the stimulus onset, the phosphorescence decay time becomes longer, and then it begins to decline and undershoots after ∼2.5 s. As a control, we measured the changes in the phosphorescence decay time in the absence of a visual stimulus. The flat line thus obtained (Fig. 2A, thin line) underscores the reliability of these decay time measurements.

Figure 2

Activity-dependent phosphorescence decay time and oxygen tension in the cortical microcirculation. Each trial consisted of 95 decay curves, collected every 0.1 s before, during, and after the visual stimulus. The decay constants were obtained by monoexponential fitting (22) at the time range of 100 to 500 μs after the flash (16). In seven independent experiments on four cats, we obtained 〈τ〉 = 181 μs, σ〈τ〉 = 15 μs. (A) Thick line: decay time constants of the phosphorescence, shown in Fig. 1C, plotted as a function of time with respect to stimulus onset. Thin line: control decay time constants for the decay curves obtained for 9.5 s with same procedure, but without stimulus. Shaded area indicates stimulus time in all panels. (B) Oxygen tension as calculated by Eq. 1 from the thick curve in (A). Error bars denote 1 standard deviation of the mean obtained from six independent groups of 16 trials each. Inset: Onset of the activity-dependent blood volume increase is delayed relative to the fast decrease in oxygen concentration. Blood volume changes (thick trace) were measured with optical imaging of the changes in reflected light at an isosbestic wavelength (570 nm), sensitive primarily to blood volume changes. This measurement was performed simultaneously with the phosphorescence measurements (thin trace) from precisely the same cortical area, devoid of any large blood vessels. These two curves were normalized to their peak amplitudes (100%; not shown in the inset) occurring after 6 s in both cases.

The kinetics of the oxygen concentration changes can be derived from these decay times using the Stern-Volmer equation:Embedded Image(1)where τ and τ0 are the measured and zero-oxygen phosphorescence lifetimes, PO2 is the oxygen tension, and K q is the second-order rate constant for quenching of phosphorescence.

Figure 2B shows the time course of the activity-dependent oxygen tension changes thus obtained (23). The small standard deviations indicate that the changes in oxygen tension in response to a sensory stimulus are reliably detected by this method. The oxygen concentration within the microvasculature starts to decline almost as soon as cortical electrical activity is evoked by the stimulus. It reaches its minimum after ∼1.5 s and then starts to increase quickly, crossing the baseline after ∼2.5 s. The delayed free oxygen increase peaks after 6 s and then begins to decline again. To prove that the onset of the early deoxygenation precedes any blood volume changes, we measured blood volume changes simultaneously with the phosphorescence decay times. The onset of blood volume changes is delayed by ∼0.5 s with respect to the onset of changes in oxygen tension (Fig. 2B, inset). This delay rules out the existence of a fast, vein/venule-specific, blood volume increase similar to that predicted by the balloon model at this early time (24) that might otherwise provide an alternative explanation for the observed deoxygenation (21). These results suggest once again that the first fast event is increased oxygen consumption by the electrically active neurons, followed by a large hyperoxygenation, most probably due to the delayed blood flow increase. The ratio of negative to positive peak is ∼4.5 (Fig. 2B), which indicates a large mismatch between oxygen consumption and supply. Having seen similar results in seven independent experiments performed on four cats (Fig. 3C), we conclude that in the anesthetized cat, the first event after a sensory stimulus is a decrease in oxygen saturation due to an increase in oxygen consumption, caused in turn by an increase in oxidative metabolism in electrically active neurons or adjacent glia.

Figure 3

Calculation of the time course of changes in oxygen tension from imaging spectroscopy. (A) Time course of activity-dependent changes in oxyhemoglobin (red) and deoxyhemoglobin (blue) concentration obtained with imaging spectroscopy. Curves were normalized for easier comparison of timing relations. The inset shows the observed curves without normalization. Arrows: stimulus onset in all panels. (B) Activity-dependent changes in oxygen tension estimated from imaging spectroscopy. See text and (25) (n = 8; error bars: SD of the mean). (C) Oxygen tension observed with phosphorescence; means and SD are from seven independent experiments on four cats. The timing of the deoxygenation phase peak measured by the two methods was similar but not identical, presumably because of the approximations used in the analysis of imaging spectroscopy (14, 25), its low time resolution, and the variability observed in different experiments.

We wondered how the time course of activity-dependent oxygen concentration changes obtained by this direct measurement relates to the indirect measurements obtained by imaging spectroscopy and blood flow measurements (7, 8). Under some assumptions (25), these changes in the concentration of free oxygen in the microvascular system can be calculated from the relative changes in the concentrations of oxyhemoglobin and deoxyhemoglobin as follows:Embedded Image Embedded Image(2)where [Oxy]0 and [Deoxy]0 are the oxy- and deoxyhemoglobin concentrations at rest. We analyzed the imaging spectroscopy data at a wavelength range of 530 to 650 nm. The kinetics of activity-dependent changes in the concentrations of oxyhemoglobin and deoxyhemoglobin thus obtained are shown in Fig. 3A. From these results, we estimated the time course of oxygen concentration changes with the use of Eq. 2. Remarkably, the result (Fig. 3B) is qualitatively similar to the time course obtained with our direct measurements (Fig. 3C): Both methods yielded an initial decrease in oxygen, followed by a large hyperoxygenation. Thus, the existence of the initial dip is supported by these two independent measurements.

The present direct measurements of oxygen tension in the cortical microvasculature indicate that the initial dip reflects a very fast increase in oxidative metabolism in response to sensory stimulation. These results are consistent with the fast decrease in NADH (reduced form of nicotinamide adenine dinucleotide) fluorescence reported in slices and in vivo after electrical stimulation (26). Recent PET measurements of the human brain have both supported (27) and refuted (13) this initial increase in oxidative metabolism, as have many previous fMRI measurements (11). The present results, rather than reflecting a species difference between feline and primate or the state of anesthesia (14), are very relevant to fMRI or PET imaging of the human brain. Indeed, our conclusions have recently been confirmed by 4.7 T fMRI measurements on monkeys (10) and by 4 T, 9.4 T, and 4.7 T fMRI measurements on humans and on anesthetized cats (9). It has been argued, on theoretical (24) and experimental (28) bases, that the amplitude of the initial dip seen with fMRI depends on the second power of the field strength. Furthermore, Logothetis has recently shown that the initial dip detected with high-field fMRI in the parenchyma is not detected in the large vessels, whereas the late BOLD component is detected in both compartments [figure 5 in (10)]. Why the initial dip has not been observed by most low-field BOLD fMRI measurements is an important question (28). It seems that low-field BOLD measurements are not sufficiently sensitive to changes in deoxyhemoglobin concentration in the cortical capillaries. The spatial resolution of such fMRI measurements is therefore not quite clear. Robust ocular dominance columns have been imaged by fMRI only by relying on either differential imaging or signal amplitude (29), yet this is not always practical. There is a better solution: It has been shown that the delayed blood flow and blood volume changes are not colocalized with activated cortical columns (7, 8). Here, we show that the initial dip is a reflection of fast changes in oxidative metabolism. Because changes in oxidative metabolism and in electrical activity are colocalized, it appears that a much better spatial resolution—approaching that obtained with optical imaging—will be achieved using high-field fMRI measurements during the initial dip period. Such a resolution is required to visualize fundamental processing modules performing higher cognitive functions in the human brain.

Note added in proof: The recent visualization of the cat orientation columns by Kim et al. (9) supports this conclusion.


View Abstract

Navigate This Article