New Insights into Central Roles of Cerebral Oxygen Metabolism in the Resting and Stimulus-Evoked Brain

Abstract

The possible role of oxygen metabolism in supporting brain activation remains elusive. We have used a newly developed neuroimaging approach based on high-field in vivo ¹⁷O magnetic resonance spectroscopic (MRS) imaging to noninvasively image cerebral metabolic rate of oxygen (CMRO₂) consumption in cats at rest and during visual stimulation. It was found that CMRO₂ increases significantly (32.3% ± 10.8%, n = 6) in the activated visual cortical region as depicted in blood oxygenation level dependence functional maps; this increase is also accompanied by a CMRO₂ decrease in surrounding cortical regions, resulting a smaller increase (9.7% ± 1.9%) of total CMRO₂ change over a larger cortical region displaying either a positive or negative CMRO₂ alteration. Moreover, a negative correlation between stimulus-evoked percent CMRO₂ increase and resting CMRO₂ was observed, indicating an essential impact of resting brain metabolic activity level on stimulus-evoked percent CMRO₂ change and neuroimaging signals. These findings provide new insights into the critical roles of oxidative metabolism in supporting brain activation and function. They also suggest that in vivo ¹⁷O MRS imaging should provide a sensitive neuroimaging modality for mapping CMRO₂ and its change induced by brain physiology and/or pathologic alteration.

Introduction

Normal brain function requires large amounts of energy. This high energy demand is met mainly through the chemical process of oxidative phosphorylation in the brain mitochondria to produce adequate quantities of adenosine triphosphate (ATP; Attwell and Laughlin, 2001; Du et al, 2008; Hyder et al, 2006; Lei et al, 2003; Raichle, 1987; Sokoloff, 1991). Under the resting condition, this process is responsible for the consumption of most glucose used by the brain, generating ~ 36 ATPs from each glucose molecule. It is also tightly coupled to cerebral oxygen utilization and is responsible for ~ 20% of the total oxygen consumption in the human body even though the brain constitutes only ~ 2% of the body weight. Therefore, the cerebral oxidative metabolism is essential for supplying brain energy and supporting brain function. Consequently, cerebral metabolic rate of oxygen (CMRO₂) consumption provides an important physiologic parameter that can be used to evaluate cerebral bioenergetics in healthy and diseased brains at rest; as well as to investigate the functional neurometabolic relation during brain stimulation and/or task performance. However, the ability to directly and reliably image resting CMRO₂ and CMRO₂ alteration during brain activation has been a major challenge for several decades. Positron emission tomography (PET) combined with ¹⁵O radiotracers has been perhaps the only established in vivo imaging method able to directly image CMRO₂ (Ibaraki et al, 2008; Mintun et al, 1984; Mintun et al, 2002). Despite its relatively coarse spatial resolution, its requirement for complex mathematical modeling, and the involvement of multiple measurement procedures (Mintun et al, 1984), the ¹⁵O-PET technique has been applied to investigate oxygen metabolism in the human brain under various conditions. In one prominent PET application, regional stimulus-evoked CMRO₂ change in the human visual cortex was determined to be ~ 5% whereas alterations in cerebral blood flow (CBF) and cerebral metabolic rate of glucose (CMR_glc) consumption were significantly higher (~ 50%; Fox et al, 1988). These metabolic rate changes only corresponded to ~ 7% increase in the stimulus-evoked ATP production rate, a surprising finding implying that energy cost of stimulus-evoked neuronal activity is relatively small and it might be met partially through the glycolysis process in the awake human brain (Barinaga, 1997; Fox et al, 1988; Raichle and Mintun, 2006). In contrast, the cerebral oxidative glucose consumption rate, determined through the in vivo ¹³C magnetic resonance spectroscopic (MRS) measurement, was reported to increase substantially up to 80%-100% in the anesthetized rat somatosensory cortex during forepaw electric stimulation (Hyder et al, 2001). One possible source of this sizable discrepancy has been speculated to be the difference in baseline CMRO₂ levels between the awake human and the anesthetized rat brains, highlighting the importance of quantitative CMRO₂ measurements under both baseline and stimulated brain states.

Qualitatively, the finding that the ratio between the stimulus-evoked percent CMRO₂ change versus that of CBF or CMR_glc is much smaller than 1 (Fox et al, 1988) is consistent with and underlies the interpretation of the blood oxygenation level dependence (BOLD) contrast (Ogawa et al, 1990; Raichle and Mintun, 2006) detected by functional magnetic resonance imaging (fMRI), which has become the most prominent neuroimaging modality for mapping brain activation (e.g., Bandettini et al, 1992; Kwong et al, 1992; Ogawa et al, 1992). However, fMRI is unable to directly detect neuronal activity; instead, it relies on a complex interplay among CBF, cerebral blood volume (CBV), and CMRO₂ changes induced by altered brain activity (Ogawa et al, 1993). Precise interpretation of fMRI results requires a better understanding of the quantitative relationship between the fMRI BOLD contrast and the underlying neurophysiology—in particular the stimulus-evoked CMRO₂ change.

Therefore, the ability to image absolute CMRO₂ with relatively high spatial resolution and temporal resolution is crucial towards understanding cerebral metabolic events that occur during brain activation, and their implication on quantifying neuroimaging signals. Towards this goal, we recently established the three-dimensional (3D) ¹⁷O magnetic resonance spectroscopy imaging (MRSI) method at high/ultrahigh magnetic fields (Zhu et al, 2002; Zhu et al, 2007) as a noninvasive neuroimaging modality for imaging CMRO₂ in vivo; and have rigorously evaluated this method using an animal model (Zhang et al, 2004; Zhu et al, 2007). This CMRO₂ imaging method relies on detecting the dynamic change of ¹⁷O-labeled metabolic water (H₂¹⁷O) produced by oxygen metabolism in the brain mitochondria during a short (2 to 3 mins) inhalation of ¹⁷O-isotope-labeled oxygen gas (Arai et al, 1990; Mateescu et al, 1989; Pekar et al, 1991; Zhu et al, 2005; Zhu et al, 2002; Zhu et al, 2007). In the present study, we exploit this newly developed high-field in vivo ¹⁷O MRSI approach to study the possible roles of cerebral oxygen metabolism in brain function and stimulus-evoked activation by imaging CMRO₂ in the cat primary visual cortex (V1; cortical areas 17 and 18) at rest (defined as baseline condition) and during visual stimulation, ultimately generating functional metabolic activation maps of relative CMRO₂ change (i.e., ACMRO₂/CMRO₂ maps) and comparing them with the BOLD-based fMRI mapping results. The overall results provide new insights into quantitative relationship of CMRO₂ in a resting visual cortex and its change evoked by visual stimulation; they also provide vital evidence to potentially coincide several long-standing and unsolved issues in the literature regarding the quantitative relationships among CMRO₂, brain energy expenditure, cerebral bioenergetics, and brain activation.

Materials and methods

Animal Preparation

Five female adolescent cats were used to conduct this study. Cats were initially anesthetized with a mixture of ketamine (15.0 mg/kg) and xylazine (2.5 mg/kg). After oral intubation, mechanical ventilation (30 to 33 stokes/mins) was applied and anesthesia was switched to 0.9% to 1.2% isoflurane (~ 1.5 MAC) in a N₂O/O₂ mixture of 70:30 volume ratio throughout the experiment. The pupils of the cat were dilated with atropine sulfate solution; corrective contact lenses were placed to focus the eyes on the visual stimulus by refracting and locating the fovea of the cat retina with the aid of a fundus camera (Zeiss, Jena, Germany). The visual stimulus was a binocular high-contrast square-wave moving and rotating gratings (0.3 cycle/deg, 2 cycles/sec, and 16° rotation for every 4 secs) to achieve optimal visual stimulation of the neurons with different orientation preferences in the cat primary visual cortex. The cats were placed in a cradle with head position restrained by the mouth and ear bars. The animal physiologic condition was continuously monitored and maintained during the entire experiment. All animal surgical procedures and experimental protocol were approved by the Institutional Animal Care and Use Committee of the University of Minnesota.

Magnetic Resonance Spectroscopy Imaging and Magnetic Resonance Imaging Measurements

The ¹⁷O MRSI and ¹H MRI measurements were conducted on a 9.4 T horizontal animal magnet (Magnex Scientific, Oxford, UK) interfaced with a Varian INOVA console (Varian Inc., Palo Alto, CA, USA). A radiofrequency (RF) probe consisted of a ¹⁷O surface coil covering the cat V1 region for acquiring ¹⁷O MRSI and a larger ¹H coil for brain anatomic images and fMRI was used in this study.

¹⁷O Magnetic Resonance Spectroscopy Imaging for Imaging Cerebral Metabolic Rate of Oxygen: The spatial localization of ¹⁷O MRSI was achieved by using the 3D Fourier series window chemical shift imaging approach with a total acquisition time of 12.5 secs per MRSI volume (Hendrich et al, 1994; Zhu et al, 2002). Other ¹⁷O acquisition parameters were: 3 3 × 2.5 cm³ field-of-view (FOV); 9 × 9 × 5 phase encodes; 120 μL voxel size, which is corresponding to a nominal voxel size of 43 μL. A 17 × 17 × 9 matrix of ¹⁷O free induced decay signals were generated from the original 9 × 9 × 5 phase encoded raw image data for generating each 3D ¹⁷O MRSI dataset.

Two paired measurements for collecting series of 3D ¹⁷O MRSI before and during 2 to 3 mins inhalation of ¹⁷O₂ gas (up to 89% ¹⁷O enrichment purchased from ISOTEC Inc, Miamisburg, OH, USA) were performed for each cat in the absence and presence of visual stimulation, respectively; and only Cat 5 had repeated twice the paired ¹⁷O MRSI measurements in the same experimental session to test the reproducibility and reliability of the CMRO₂ imaging approach.

The natural abundant ¹⁷O signal of brain water obtained before the ¹⁷O₂ gas inhalation was used as internal reference for quantifying the absolute metabolic H₂¹⁷O concentration in the brain; and the dynamic change of the metabolic H₂¹⁷O concentration obtained during ¹⁷O₂ gas inhalation was applied to determine CMRO₂ using the linear regression model and to generate 3D CMRO₂ images (Zhang et al, 2004; Zhu et al, 2002; Zhu et al, 2007). The functional CMRO₂ activation maps were created based on the value of ACMRO₂/CMRO₂, in which ACMRO₂ represents the difference between the stimulated and control CMRO₂ values.

Two criterions were applied to quantify the CMRO₂ values measured under resting and stimulated conditions, as well as ACMRO₂/CMRO₂. The ¹⁷O RF surface coil used in this study provides optimal sensitivity for detecting the ¹⁷O MRSI signal in the cat V1 and surrounding brain regions. In contrast, the detected ¹⁷O signal intensity reduces substantially in the brain regions, which are distant from the RF coil, resulting in a low signal-to-noise ratio (SNR). Thus, the first criterion was to exclude the ¹⁷O MRSI voxels with low SNR of natural abundance brain water signal from further CMRO₂ quantification. The averaged SNR from the ¹⁷O MRSI voxels, which were included in CMRO₂ quantification was approximately 10:1. The Monte Carlo simulation (unpublished results) has suggested that the CMRO₂ fitting error measured by the high-field ¹⁷O MRSI approach can reach a few percents if the SNR of natural abundance brain water signal is below 10:1. Thus, the second criterion was to set a threshold of −5% > ACMRO₂/CMRO₂ > 5% for calculating the stimulus-evoked percent CMRO₂ change (i.e., ACMRO₂/CMRO₂); and generating the functional metabolic activation maps of ACMRO₂/CMRO₂.

The functional ACMRO₂/CMRO₂ maps were further analyzed in two ways. The first way was to average the baseline and stimulated CMRO₂ value, respectively, only from the same brain regions showing a positive stimulus-evoked CMRO₂ change; and these averaged CMRO₂ values were further used to calculate the averaged value of positive ACMRO₂/CMRO₂. The second way was to calculate the baseline and stimulated CMRO₂ value, respectively, averaged from the brain regions showing either a positive or a negative stimulus-evoked CMRO₂ change; and these CMRO₂ values were used to present the averaged value of net change of ACMRO₂/CMRO₂ in these brain regions.

Functional Magnetic Resonance Imaging Mapping: At first, multislice anatomic images were acquired using a conventional T₁-weighted imaging method. On the basis of these anatomic images, slices covering the cat V1 areas were appropriately selected for acquiring fMRI data using multislice gradient echo planar images (EPI) with the following acquisition parameters: repetition time (TR) = 1.65 sec; echo time (TE) = 19 ms; five adjacent axial EPI slices; FOV = 5 × 5 cm²; 780 μm × 780 μm in-plane spatial resolution; 1 mm slice thickness and 2 to 2.5 mm interval between adjacent slices.

The fMRI measurements were based on the block paradigm design: three control and two task periods in an interleaved way. Activation maps were generated using a time-shifted cross-correlation method (Bandettini et al, 1992; Xiong et al, 1995). The activated pixels were identified by correlating the fMRI time course with a trapezoidal function (the modified box-car function with the capability of accounting for hemodynamic delays). This procedure generated typical activation maps in V1, which were consistent across different cats. We chose a statistical threshold of P < 0.01 to generate fMRI activation maps.

The paired t-test was performed for statistical analysis and the results are presented as mean ± s.d.

Results, Discussion, and Conclusions

Stimulus-Evoked Cerebral Metabolic Rate of Oxygen Increase During Brain Activation

Figure 1 shows the 3D images of absolute CMRO₂ values obtained at rest and during visual stimulation, functional metabolic activation maps of ACMRO₂/CMRO₂, BOLD-based fMRI maps, and their corresponding brain anatomic images in four adjacent image slices chosen from 3D image data in a representative cat brain. This figure clearly illustrates a significant CMRO₂ increase in the activated brain region in the cat visual cortex (defined as activated V1 region herein) in response to a full-field, binocular grating visual stimulation. The size and location of the activated brain regions depicted in the functional ACMRO₂/CMRO₂ maps largely coincide with those regions determined by fMRI maps showing positive BOLD changes in the same cat brain despite of different spatial resolutions between CMRO₂ and BOLD functional images. Figure 2 shows the absolute CMRO₂ values averaged over these activated brain regions in different image slices from one representative cat, and the summarized results of six measurements from five cats (one cat having two repeated measurements under control and stimulated conditions). The CMRO₂ value averaged spatially from the activated cat V1 region showing a positive change of ACMRO₂/CMRO₂ was 1.26 ± 0.09 μmol/g per min (n = 6) under the visual stimulation condition; and this value is significantly higher than that of 0.97 ± 0.04 μmol/g per min (n = 6) measured at resting condition from the same brain region, indicating a 32.3% ± 10.8% increase of CMRO₂ in the activated V1 region>(P < 0.005, paired t-test). Figures 3A and 3B summarize the intersubject averaged CMRO₂ results measured under resting and stimulated brain states, indicating a substantial CMRO₂ increase ranging approximately from 20% to 50% in the activated V1 region in different cats studied (see Figure 3B). These results suggest that the oxidative metabolic activity is significantly elevated in the anesthetized cat brain to support the stimulus-evoked enhancement in ATP production/consumption associated with the increased neuronal activity in the activated V1 region. Unlike previous studies, which relied on BOLD modeling to assess the relative (or percent) CMRO₂ changes induced by brain stimulation (Davis et al, 1998; Hoge et al, 1999; Kim et al, 1999), the present study directly imaged the absolute CMRO₂ values under both resting and activated brain states, which provide not only a quantitative measure of percent CMRO₂ change elevated by visual stimulation but also crucial information regarding the relationship between the absolute CMRO₂ change and increased brain activity during activation. The finding of significant CMRO₂ increase in the activated brain region observed in the present study reveals a tight neurometabolic coupling during brain activation; suggesting the vital role of oxygen metabolism in supporting the intensified neuronal activity in a working brain.

Figure 1

Significant CMRO₂ increase in activated visual cortex. Three-dimensional CMRO₂ maps obtained at resting (second column from the left) and stimulated (third column) conditions; 3D functional metabolic activation maps showing percent changes of CMRO₂ elevated by visual stimulation (fourth column); BOLD-based fMRI maps (fifth column); and anatomic brain images (first column) in four adjacent image slices from a representative cat brain.

Figure 2

Summary of resting and activated CMRO₂ and percent changes. Summarized absolute CMRO₂ values measured at resting and during visual stimulation (top inserts) and their percent changes (bottom inserts) from four CMRO₂ image slices covering the visual cortex of cat anesthetized with isoflurane. (A) Results from a representative cat. (B) Averaged results from six experiments, for the ¹⁷O MRSI voxels with only positive CMRO₂ change. (C) Averaged results from six experiments for the ¹⁷O MRSI voxels with both positive and negative CMRO₂ changes. Paired t-test indicates a statistically significant difference between the control and activated CMRO₂ values (P < 0.005 using paired t-test for both panels B and C).

Figure 3

Stimulus-evoked CMRO₂ increase and its relation to resting CMRO₂.(A) Averaged absolute CMRO₂ values for each cat brain at resting and during visual stimulation. (B) The averaged percent CMRO₂ increases elevated by visual stimulation. (C) The negative correlation between the percent CMRO₂ changes and baseline CMRO₂ values. This figure shows that visual stimulation elevates a significant CMRO₂ increase in the activated visual cortex of five cats studied (P < 0.005 using paired t-test for both panels A and B), and the percent CMRO₂ changes are negatively correlated to the corresponding baseline CMRO₂ levels.

Total Brain Energy Expenditure During Brain Activation

Interestingly, we also consistently detected a CMRO₂ decrease in the brain region surrounding the territory showing positive CMRO₂ or BOLD change in the cat V1. This phenomenon was observed in all cats, and the functional metabolic CMRO₂ image results from two representative cats are shown in Figure 4. When both positive and negative CMRO₂ alterations were counted and spatially averaged, the net CMRO₂ increase induced by visual stimulation reduced to 9.7%± 1.9% (n = 6, see Figure 2C) compared with the resting CMRO₂ value measured from the same brain regions showing either a significant positive or negative CMRO₂ alteration during the visual stimulation; the stimulus-evoked CMRO₂ increase was still statistically significant (P < 0.005, paired t-test). This finding reveals that the stimulus-evoked CMRO₂ change varied significantly across the examined brain regions and consistently displayed two distinct response patterns with positive and negative CMRO₂ alterations in the cat brain during visual stimulation.

Figure 4

Spatially compensated CMRO₂ changes elevated by brain stimulation. Functional CMRO₂ maps (show both positive and negative changes) and fMRI BOLD maps during visual stimulation from two representative cats. This figure shows that the CMRO₂ increases in the central activated visual cortex regions are commonly accompanied by CMRO₂ decreases in surrounding brain regions.

Similar spatial characteristics have been shown by other physiologic measurements based on the hemodynamic changes of CBV and CBF or BOLD contrast. For instance, negative CBV changes surrounding the central visual cortical regions showing stimulus-evoked increases in CBV were previously reported in the cat brain (Harel et al, 2002); and negative changes bordering on the territory of positive changes in BOLD and CBF maps were also observed in the human brain during retinotopic visual stimulation (Shmuel et al, 2002). In the latter, BOLD modeling was used to argue that the negative CBF/BOLD changes could correspond to a decrease in CMRO₂ in these brain regions; however, ambiguities in the BOLD modeling was recognized and only a qualitative conclusion was reached (Shmuel et al, 2002).

Our results suggest that CMRO₂ indeed decreases in the brain regions surrounding the primary foci of increased activity, suggesting that the neuronal activity in the regions with negative CMRO₂ change might be suppressed owing to a tight neurometabolic coupling. Consistent with this notion, reduction in electrophysiologic signals of neuronal activity was recently reported in the negative BOLD regions surrounding the activated visual cortex in the primate (Shmuel et al, 2006). Therefore, these collective lines of evidence suggest that there are similar trends among CMRO₂, CBF, CBV, and neuronal activity changes in response to visual stimulation in both activated cortical region and surrounding deactivated cortical region, indicating that a tight coupling among the metabolic and hemodynamic responses to neuronal activity change might qualitatively hold under both activated and deactivated conditions (Raichle and Mintun, 2006).

The negative CMRO₂ changes observed during visual stimulation in this study could be driven by oxidative metabolic, other physiologic or neuronal process. Irrespective of the operative mechanism, however, this observed CMRO₂ phenomenon may arise to limit total brain energy expenditure superimposed on an expensive energy budget used at the resting state, for instance, to maintain the resting brain network and neuronal connectivity sustained by spontaneous neuronal activity rhythm (Buzsaki and Draguhn, 2004; Fox et al, 2006) and to provide a significant amount of ‘house-keeping’ ATP energy (Du et al, 2008). This means that a resting brain is not truly at ‘rest’ and it costs sizeable brain energy. Thus, the increased energy demand in the activated brain regions could be partially compensated by the reduction of energy usage in other brain regions, leading to a smaller extent of total (or global) brain CMRO₂ increase during activation. This energy compensation mechanism could provide a satisfactory explanation for the observation in the present study as well as to the surprising finding from a well-documented work published in 1955, showing no global CMRO₂ change in the human brain during an intense mental arithmetic task (Sokoloff et al, 1955). Two speculations were discussed in this early publication to interpret the finding: one was that the total energy requirements of the brain, as reflected by the cerebral oxygen consumption, may be independent of the brain activity change elevated by task performance; the other speculation was that the task performance may lead to a spatial redistribution of neuronal and metabolic activities, that is, brain regions with increased functional and metabolic activities could then be counterbalanced by regions of reduced activities so that the global CMRO₂ is unchanged (Sokoloff et al, 1955). Nevertheless, it was difficult to justify these two possibilities by the global CMRO₂ measurements (Sokoloff et al, 1955) because lack of information regarding the spatial distribution of positive and negative CMRO₂ changes inside the brain. The results from the present study provide vital evidence suggesting that the increased neuronal activity evoked by brain stimulation requires a significantly higher energy demand of oxygen consumption in the activated brain regions than that at rest; however, this increased regional energy demand could be partially compensated by the reduction of energy demand in other brain regions resulting in a smaller increase in CMRO₂ over a larger brain volume. This mechanism could be critical for maintaining normal, sustained brain activation in a working brain.

Although, the present study is unable to examine whether the metabolic activity itself or neuronal activity is the primary origin driving the CMRO₂ suppression in the surrounding brain regions during visual activation, it has been suggested that the decrease of CMRO₂ likely indicates regional depression of synaptic activity (Pasley et al, 2007; Seitz and Roland, 1992; Shmuel et al, 2006).

One important implication of both negative and positive CMRO₂ alterations in the nearby brain regions during stimulation as observed in this study is that the averaged stimulus-evoked CMRO₂ change (in either relative or absolute scale) becomes substantially smaller when the spatial resolution of CMRO₂ imaging is inadequate to differentiate the regions showing negative and positive CMRO₂ alterations; with increased imaging voxel size (i.e., reducing spatial resolution) and increased partial volume effect, the change in CMRO₂ may even become undetectable. This observation could provide key for potentially reconciling the long-standing controversy in the literature regarding the essential questions: if, and how much CMRO₂ changes in response to brain stimulation; and whether the oxygen metabolism is vital in supporting brain work (Barinaga, 1997).

It is interesting to note that although the negative CMRO₂ (this study) and CBV (Harel et al, 2002) changes were observed in the surrounding regions of cat visual cortex, the negative BOLD was not robustly seen in all cats in our study. This observation suggests the complexity of the BOLD signal nature owing to its dependence on the interplay among three stimulus-evoked physiologic parameter (i.e., CMRO₂, CBF, and CBV) changes as well as on the baseline levels of these parameters as discussed in the following section. Thus, a single CMRO₂ parameter may provide a better and more quantitative measure reflecting the brain energy state or the corresponding neuronal activity level if the limited ¹⁷O detection sensitivity is not a major concern; and it should be useful for understanding many neuroscience questions, for example, the relationship between brain energy usage and neuronal excitation versus neuronal inhibition (Buzsaki et al, 2007).

Importance of Baseline Cerebral Metabolic Rate of Oxygen

Another striking observation in the present study is that the positive percent CMRO₂ changes evoked by visual stimulation were negatively correlated with their corresponding baseline CMRO₂ values, as shown in Figure 3C. This finding indicates a strong influence of baseline metabolic activity level on the percent CMRO₂ change in response to brain stimulation (Hyder et al, 2002; Pasley et al, 2007; Shulman et al, 2007); thus, the stimulus-evoked percent CMRO₂ change alone can become less significant for quantitatively interpreting and fully understanding the neurometabolic relationship if the baseline CMRO₂ is not determined.

According to the trend shown in Figure 3C, we could further speculate that the stimulus-evoked CMRO₂ percentage change in the awake brains would be less pronounced as compared with the anesthetized brains because a higher baseline level of CMRO₂ is expected under the awake state. The visual cortex of the awake human, for instance, has a relatively higher baseline CMRO₂ value of B1.7 μmol/g per min (Fox et al, 1988) compared with the value of 0.97 μmol/g per min in the anesthetized cat visual cortex as measured in the present study. On the basis of Figure 3C and the assumption of similar CMRO₂ change behaviors between the human and cat visual cortices in response to visual stimulation, one could expect that the percent CMRO₂ increase in the human visual cortex during visual stimulation should be significantly less than 32% observed in the anesthetized cat brain in this study, and certainly substantially smaller than the CBF change of 40% to 50% in the human visual cortex (Fox et al, 1988; Zhu et al, 1998). This notion is consistent with the fact that a hyperoxygenation level in the blood is elevated in the activated brain region because the mismatched CMRO₂ and CBF changes, leading to a positive BOLD contrast in fMRI (Ogawa et al, 1990). This estimated range of percent CMRO₂ increase in response to visual stimulation is also qualitatively consistent with the literature results reporting a wide range from 5% to 30% in the human visual cortex (Chen et al, 2001; Davis et al, 1998; Fox et al, 1988; Hoge et al, 1999; Kim et al, 1999). However, ultimately studies under the awake state either with the ¹⁷O CMRO₂ imaging methodology or using this methodology as a calibration for BOLD modeling (Davis et al, 1998; Hoge et al, 1999; Kim et al, 1999; Ogawa et al, 1993) will be necessary to draw definitive conclusions in the human brain.

The possible mechanism to link the negative correlation between the stimulus-evoked percent CMRO₂ change and its baseline CMRO₂ level shown in Figure 3C can be explained by two possible models with distinct assumptions. The first one assumes that CMRO₂ can rise from its baseline level to a higher and constant level in the activated brain region during stimulation (Hyder et al, 2002; Pasley et al, 2007) so that the absolute stimulus-evoked CMRO₂ becomes independent of the baseline CMRO₂ level. In this case, a relatively low baseline CMRO₂ level should result in a larger stimulus-evoked percent increase in CMRO₂; this is consistent with the negative correlation in Figure 3C. The second model is to assume that CMRO₂ can rise from its baseline level to a higher level but with a constant CMRO₂ increment in the activated brain region. On the basis of this model, the stimulus-evoked percent CMRO₂ increase will be larger again if the baseline CMRO₂ level is relatively low; this scenario will also lead to a similar negative correlation as shown in Figure 3C. Owing to the inevitable variation among animals in the CMRO₂ data measured in the present study, these two models are difficult to be justified with statistical confidence; although the first model was slightly favored by our data showing a relatively smaller intersubject variation (quantified by s.d.) in the absolute values of stimulus-evoked CMRO₂ compared with that of stimulus-evoked CMRO₂ increments.

The tight correlation between the percent CMRO₂ change and baseline CMRO₂ level as shown in Figure 3C has several profound impacts on both fMRI methodology and functional brain electrophysiology. First, the percent change of stimulus-evoked CMRO₂ can vary as a function of baseline CMRO₂ level, and the influence of relative (percent) CMRO₂ change on the percent change of BOLD signal measured by fMRI could rely on the baseline CMRO₂ level, which can vary substantially among subjects or between awaked and anesthetized brains. Therefore, it is critical to consider the implication of baseline CMRO₂ level on quantifying the relative changes of stimulus-evoked CMRO₂ and/or BOLD (Pasley et al, 2007; Shulman et al, 2007). Second, the correlation shown in Figure 3C implies that a large portion of brain energy consumption remains in a resting brain. This resting brain energy expenditure should be essential for supporting spontaneous brain activity, coherent neuronal activity rhythm, neurotransmission cycling, and other neuronal signaling processes; and it could be superimposed on the extra stimulus-evoked energy change to perform brain work appropriately. This energy cost could also be crucial for supporting the resting brain network and neuronal connectivity sustained by spontaneous neuronal activity rhythm (Buzsaki and Draguhn, 2004; Fox et al, 2006), which has been linked to the slow frequency BOLD signal fluctuation for obtaining functional connectivity maps in a resting brain (Biswal et al, 1995; Fox and Raichle, 2007).

The overall findings lead to an important conclusion that both baseline and stimulus-evoked brain energies should be vital for normal brain function in performing specific tasks, and the fraction of these two brain energy expenditures relies on the baseline metabolic activity level, which could be susceptible among subjects, species, and physiologic/pathologic conditions; and the baseline level can also be manipulated readily by varying experimental conditions, for instance, by changing anesthetic depth. The isoflurane anesthetic applied in the present study could significantly low the baseline CMRO₂ activity level in the cat visual cortex, ultimately, lead to a relatively larger increase in the percent CMRO₂ change and a large absolute CMRO₂ increment during brain stimulation as compared with the waked brain state. If the basal CMRO₂ level can be further suppressed by using different anesthetic, for example, α-chloralose (Ueki et al, 1992), the percent CMRO₂ increase could be even larger. Thus, the stimulus-evoked percent CMRO₂ change is sensitive to the baseline CMRO₂ level and can cover a wide range from deeply anesthetized to awaked brains. This notion provides an explanation to the large discrepancy of the CMRO₂ percentage changes reported in the literature (e.g., Fox et al, 1988; Hyder et al, 2001).

In addition, our results suggest that both baseline CMRO₂ level and the stimulus-evoked percent CMRO₂ change are equally important for brain to work properly. Although the baseline CMRO₂ level can be easily modulated, the total brain energy may have to be resumed and integrated for supporting brain activation during stimulation if the involved neurons can fully response to the stimulus.

Advantages and Promise of 17O-Based Cerebral Metabolic Rate of Oxygen Imaging Method

Overall findings in this study offer new insights into the underlying physiology of cerebral oxidative metabolism and its essential role in bioenergetics associated with brain activation; they also provide possible clarifications to a number of long-standing neuroscience questions. Moreover, they also highlight the importance of the neuroimaging capability to noninvasively imaging absolute CMRO₂ values under both resting and activated brain states with adequate spatial resolution, for instance, to differentiate different brain regions with positive and negative CMRO₂ changes in response to visual stimulation as illustrated in this study (Figure 4). This ability is crucial to quantitatively investigate the coupling relationship between the stimulus-evoked neuronal and oxidative metabolic activities. The high-field in vivo ¹⁷O MRSI approach for 3D mapping of absolute CMRO₂ and its change provides a useful neuroimaging modality with satisfactory capability. In this study, we present the first exploration of this promising neuroimaging approach for noninvasively obtaining 3D functional CMRO₂ maps in the cat brain during visual stimulation. The results clearly show the advantages of this approach in several aspects. First, the spatially averaged resting CMRO₂ value (0.97 ± 0.04 μmol/g per min) measured in the isoflurane anesthetized cat V1 region in this study is in agreement with the value of 1.02 μmol/g per min reported in the literature (Todd and Drummond, 1984), which was measured in the cats under a similar anesthesia condition using the ‘golden-standard’ Kety-Schmidt method. Second, paired CMRO₂ imaging measurements performed twice under resting and stimulated conditions in one cat (Cat 5) during the same experimental session show an excellent level of reproducibility for imaging CMRO₂ under both conditions (see Figure 3A). This result suggests that the ¹⁷OMRSI detection sensitivity offered by high field is adequate for obtaining reliable 3D functional metabolic CMRO₂ maps based on only two paired measurements: one under resting and another under stimulation, without averaging from repeated experiments. Third, based only on paired CMRO₂ imaging measurements, functional metabolic activation maps of CMRO₂ change were obtained reliably in all animals studied. Fourth, the spatial resolution of 3D CMRO₂ image achieved in this study was able to provide distinct characteristics of CMRO₂ changes in different brain regions in response to brain stimulation in a relatively small animal model. Another and perhaps the most important one is the merits of noninvasiveness, simplicity, robustness, and fair temporal resolution (a few minutes) offered by the high-field in vivo ¹⁷OMRSI approach, which should hold great promise for numerous biomedical applications beyond healthy subjects and normal brain function research.

Owing to the large differences in spatial resolution and voxel shape between the functional CMRO₂ and BOLD maps in the present study, the investigation for correlating the stimulus-evoked CMRO₂ change and BOLD signal was not pursued in this study. Nevertheless, the superior quality of high-field in vivo ¹⁷O MRSI approach for imaging CMRO₂ change as shown in this study should be readily combined with MR-based functional imaging approaches for mapping BOLD, CBF, and CBV changes in the same animal model. This ability will provide a unique opportunity for further studying the implications of stimulus-evoked CMRO₂ change on neuroimaging signals, in particular, on the BOLD signal; and for improving the reliability of BOLD calibration methods to estimate the percent CMRO₂ change (Buxton et al, 2004; Davis et al, 1998; Hoge et al, 1999; Kim et al, 1999; Ogawa et al, 1993).