Effect of Hypoxia and Hyperoxia on Cerebral Blood Flow,Blood Oxygenation,and Oxidative Metabolism

Abstract

Characterizing the effect of oxygen (O₂) modulation on the brain may provide a better understanding of several clinically relevant problems, including acute mountain sickness and hyperoxic therapy in patients with traumatic brain injury or ischemia. Quantifying the O₂ effects on brain metabolism is also critical when using this physiologic maneuver to calibrate functional magnetic resonance imaging (fMRI) signals. Although intuitively crucial, the question of whether the brain's metabolic rate depends on the amount of O₂ available has not been addressed in detail previously. This can be largely attributed to the scarcity and complexity of measurement techniques. Recently, we have developed an MR method that provides a noninvasive (devoid of exogenous agents), rapid (<5 minutes), and reliable (coefficient of variant, CoV <3%) measurement of the global cerebral metabolic rate of O₂ (CMRO₂). In the present study, we evaluated metabolic and vascular responses to manipulation of the fraction of inspired O₂ (FiO₂). Hypoxia with 14% FiO₂ was found to increase both CMRO₂ (5.0±2.0%, N=16, P=0.02) and cerebral blood flow (CBF) (9.8±2.3%, P<0.001). However, hyperoxia decreased CMRO₂ by 10.3±1.5% (P<0.001) and 16.9±2.7% (P<0.001) for FiO₂ of 50% and 98%, respectively. The CBF showed minimal changes with hyperoxia. Our results suggest that modulation of inspired O₂ alters brain metabolism in a dose-dependent manner.

Introduction

A tight control of cerebral perfusion and oxygen (O₂) supply is critical for brain function and health, especially because oxidative metabolism is the primary means of energy production in the brain (Magistretti and Pellerin, 1999). Characterization of the influence of O₂ availability on the brain is therefore an essential step toward a better understanding of brain energy homeostasis and also has important clinical implications. For example, acclimatization to high altitudes is triggered by hypoxia and >80% of people experience a certain degree of discomfort, including headache and shortness of breath, with a fraction of them developing acute mountain sickness that includes nausea, vomiting, fatigue, dizziness, and difficulty sleeping (Imray et al, 2011). However, hyperoxia has been used as a therapeutic intervention in patients with traumatic brain injury and focal ischemia (Thom, 2009). Although increased blood O₂ content is expected to enhance O₂ delivery to tissues, clinical trials of O₂ therapy have yielded mixed results, with some reporting excellent efficacy and others showing marginal benefit (Magnoni et al, 2003; Rockswold et al, 2010). One possible reason is that, while hyperoxia may benefit ischemic tissue, it may concomitantly cause oxidative stress with a potential to damage healthy brain regions. Additionally, under the assumption that changes in O₂ content do not alter brain metabolism, a hyperoxia challenge has recently been used to calibrate the Blood-Oxygenation-Level-Dependent signal in functional magnetic resonance imaging (MRI) studies (Chiarelli et al, 2007). Therefore, there has been a growing interest in understanding the effect of O₂ modulation on brain physiology.

It is well established that hyperoxia increases arterial blood O₂ content, with hypoxia having the opposite effect. Some investigators have reported a cerebral blood flow (CBF) reduction during hyperoxia, but this trend is not always observed (Bulte et al, 2007; Nishimura et al, 2007; Sicard and Duong, 2005). Even more controversial is the question whether O₂ gas modulation alters tissue metabolic rate. This uncertainty is largely due to the limited availability and sensitivity of suitable techniques to measure the cerebral metabolic rate of O₂ (CMRO₂) in vivo. Existing CMRO₂ techniques based on the Kety-Schmidt method or ¹⁵O PET require dynamic sampling of arterial blood, which is not only invasive, but also reduces the reproducibility of the measurement (Kety and Schmidt, 1948; Mintun et al, 1984). Recently, our laboratory has developed an MR-based technique for the quantification of global CMRO₂ (Xu et al, 2009). This method separately measures CBF, arterial and venous oxygenation, and uses the Fick principle (i.e., calculates arteriovenous difference) to estimate CMRO₂. The procedure is noninvasive (i.e., no blood sampling or exogenous agent), fast (<5 minutes), and reproducible (i.e., coefficient of variation, CoV, <3%) (Liu et al, 2012; Xu et al, 2009). These desirable features of the technique may allow a better detection of potential effects of O₂ modulation on CMRO₂.

In this study, we used the techniques described above to examine potential changes in CBF, arterial O₂ content, venous O₂ content, and CMRO₂ due to O₂ gas modulation. The convenience of the method allows us to investigate hypoxia (14% fraction of inspired O₂, FiO₂) and hyperoxia (administrated at two levels: 50% and 98% FiO₂) in the same study, which is different from previous studies that have focused on one challenge only. Our data suggest that both vascular and metabolic parameters of the brain are strongly dependent on the O₂ level in the inspired air.

Materials and methods

Theory for the Measurement of Global Cerebral Metabolic Rate of Oxygen

Similarly to many existing techniques, our approach to estimate CMRO₂ was based on the Fick principle (Kety and Schmidt, 1948). The difference between the present technique and the previous studies is that our method can measure the relevant parameters in a noninvasive manner with short and simple procedures. Brain O₂ metabolic rate can be written as:

$${\mathrm{C}\mathrm{M}\mathrm{R}\mathrm{O}}_2=\mathrm{C}\mathrm{B}\mathrm{F}\cdot ([{\mathrm{O}}_2{]}_{\mathrm{a}}-[{\mathrm{O}}_2{]}_{\mathrm{v}})$$

(1)

where CMRO₂ is in units of μmol/min, CBF is the cerebral blood flow in mL/min, [O₂]_a and [O₂]_v (in μmol O₂/mL blood) are O₂ contents in arterial and venous blood, respectively. Some researchers further define a term called oxygen extraction fraction, OEF=([O₂]_a−[O₂]_v)/[O₂]_a. Note that, under most circumstances (e.g., room-air breathing), considerations of [O₂]_a only need to focus on hemoglobin-bound O₂, as the amount dissolved in plasma is negligible (∼1.8% of that bound to hemoglobin). For the hyperoxia challenge used in the present study, however, the amount of dissolved O₂ is significant and should be accounted for in the calculation. Therefore, the O₂ content in the blood was calculated as:

$$[{\mathrm{O}}_2{]}_{\mathrm{a}}=Y_{\mathrm{a}}\cdot C_{\mathrm{h}}+{\mathrm{p}}_{\mathrm{a}}{\mathrm{O}}_2\cdot C_{\mathrm{d}}$$

(2)

and

$$[{\mathrm{O}}_2{]}_{\mathrm{v}}=Y_{\mathrm{v}}\cdot C_{\mathrm{h}}+{\mathrm{p}}_{\mathrm{v}}{\mathrm{O}}_2\cdot C_{\mathrm{d}}$$

(3)

where C_h (8.97 μmol O₂/mL blood for an Hct of 0.44, but was adjusted for each subject based on individual Hct) and C_d (0.00138 μmol O₂/mL blood/mm Hg O₂ tension) are constants associated with hemoglobin O₂-carrying capacity and blood O₂-dissolving capacity, respectively (Guyton and Hall, 2005). Y_a and Y_v are arterial and venous O₂ saturation fractions of hemoglobin (expressed in %). p_aO₂ and p_vO₂ are the O₂ tensions of the arterial and venous blood, respectively, in units of mm Hg.

Among the parameters needed to compute CMRO₂, the most challenging task has been the measurement of Y_v. We have recently developed and validated a technique, T₂-Relaxation-Under-Spin-Tagging (TRUST) MRI, to estimate global Y_v in the sagittal sinus (Lu and Ge, 2008; Lu et al, 2012). This technique was found to be highly reproducible and can be conducted within 4 minutes. Therefore, TRUST MRI was used to estimate Y_v in the present study. Once Y_v is known, p_vO₂ can also be estimated from the O₂ dissociation curve, although the dissolved O₂ in venous blood is virtually negligible compared with the hemoglobin-bound O₂. Phase-contrast (PC) MRI was used to measure CBF (Xu et al, 2011). Consistent with TRUST MRI, PC MRI was also performed in the sagittal sinus. Thus, the estimated CMRO₂ reflects the total O₂ consumed by brain tissues that are drained by the sagittal sinus. The sagittal sinus was used in our study as opposed to using jugular vein or internal carotid artery because this vessel is subject to minimal blood pulsation, which would negatively impact the image quality for both TRUST and PC MRI. Although the flow in the sagittal sinus does not provide a true whole-brain measure, the comparison of parameters across different O₂ conditions is still expected to be valid, as long as the draining path of venous blood is not altered by O₂ modulation.

For the estimation of arterial O₂ content (equation (2)), Y_a was measured with Pulse Oximeter on the index finger. For p_aO₂, some studies have assumed the value to be the same as alveolar O₂ pressure (p_AO₂), which can be determined by end-tidal (Et) O₂ measurement. In the present study, we have considered the alveolar-arterial O₂ pressure gradient (p_{A−a gradient}) and its dependence on p_AO₂ and age. Specifically, arterial O₂ pressure was estimated by p_aO₂ = p_AO₂ − p_{A−a gradient}, where p_{A−a gradient} is assumed to be a linear function of p_AO₂ and age (Ayres et al, 1964; Stein et al, 1995) (see Supplementary text).

Participants

The study was approved by the Institutional Review Board of the University of Texas Southwestern Medical Center. Sixteen subjects (26±4 years old, nine males and seven females) were recruited. The participants did not report pulmonary, respiratory, neurologic, or psychiatric disorders according to self-completed questionnaires. None of the participants were smokers or had asthma. The subjects gave informed written consent before participating in the study.

Oxygen Modulation

Modulation of O₂ content in the inspired air was achieved using a custom-made breathing apparatus (Xu et al, 2011; Yezhuvath et al, 2009). Briefly, after lying on the magnet table, the subject was fitted with a nose clip and a mouthpiece so that s/he could breathe through the mouth only. The mouthpiece was connected to a three-way valve, which delivers either room air (21% FiO₂) or a special gas mixture contained in a Douglas bag. Three Douglas bags containing different gas mixtures were prepared: (1) 14% O₂ and 76% N₂ (hypoxia); (2) 50% O₂, 1% CO₂, and 49% N₂ (50% hyperoxia); and (3) 98% O₂ and 2% CO₂ (98% hyperoxia). These bags sequentially connected to the valve, thereby switching the inspired air. Note that a small amount of CO₂ was intentionally added to the hyperoxia gas mixture because hyperoxia tends to cause the subject to hyperventilate, resulting in an unwanted physiologic change of reduced arterial CO₂ level, which may have additional effects on CBF and CMRO₂ (Cohen et al, 2002). We therefore added CO₂ to partially offset this side effect. Similar strategies have been used in previous reports in the literature (see also Discussion) (Floyd et al, 2003).

The study paradigm consisted of a continuous, 50-minute session in which the subject breathed the above-mentioned gas mixtures in the following order: room air (21% FiO₂) for 8 minutes, 14% FiO₂ for 18 minutes, 50% FiO₂ for 15 minutes, and 98% FiO₂ for 12 minutes (see color coding in Figure 1). The duration of each breathing period was determined based on the time needed to reach a new steady state after switching the gas (calibrated from tests conducted outside the MRI). For example, it takes more time for the 14% FiO₂ condition to reach a steady state, therefore the administration for this gas mixture was the longest.

Figure 1

Illustration of experimental paradigm and time courses of arterial oxygen content, [O₂]_a, and cerebral blood flow (CBF). During the session, the subject breathed 21%, 14%, 50%, and 98% fraction of inspired O₂ (FiO₂) for 8, 18, 15, and 12 minutes, respectively, and these are denoted by different colors in the plot. The duration of each condition was preset based on the time needed to reach a new steady state that was determined from several testing experiments. Magnetic resonance imaging (MRI) data acquisitions were performed throughout the session and the type of pulse sequence performed is denoted as dots or bars at the top of the plot, where each dot indicates a 0.5-minute phase-contrast MRI and each bar indicates a 3.2-minute T₂-Relaxation-Under-Spin-Tagging (TRUST) MRI. Due to the large number and high density of the phase-contrast scans, the data allowed the assessment of time course of CBF changes (blue curve) during the experiment (the gap in the curve is due to the TRUST scan). For comparison, [O₂]_a (accounting for both hemoglobin-bound and dissolved O₂) time course is also display (red curve). The color reproduction of this figure is available at the Journal of Cerebral Blood Flow and Metabolism journal online.

During the experiment, the participant was instructed to keep still and maintain a uniform breathing pattern. To prevent drowsiness, an individually selected movie was shown to the participant via a back-projection video system (typically used for functional MRI fMRI) throughout the session. The rationale for this intervention was to prevent potential effects of sleep and drowsiness on brain CBF and metabolism (Braun et al, 1997; Nofzinger et al, 2002). Continuous physiologic recordings were obtained for end-tidal (Et) CO₂ (Capnograph; Novametrix Medical System, Wallingford, CT, USA), EtO₂ (Analox Sensor Technology, Stokesley, North Yorkshire, UK), Y_a (MEDRAD, Pittsburgh, PA, USA), and heart rate (MEDRAD). After the MRI scans were completed and the subject exited the scanner room, a blood sampling with a potassium ethylenediaminetetraacetic acid-coated 10 mL lavender tube was conducted on the basilic vein of the arm and Hct was measured with a centrifuge (Hemata STAT II, Separation Technology, Inc., Altamonte Springs, FL, USA). Determination of Hct is needed to obtain an accurate blood oxygenation estimation from the MRI measure of T₂ (Lu et al, 2012).

Magnetic Resonance Imaging Experiments

The MRI scans were performed on a 3-Tesla system (Philips Medical Systems, Best, The Netherlands). Two sequences, TRUST MRI and PC MRI, were performed multiple times during the entire session and their timing is shown in Figure 1 (bars and dots, respectively, at the top of the figure). The TRUST MRI (duration 3.2 minutes) was performed four times during the steady states of the four O₂ conditions. The PC MRI (duration 0.5 minutes) was performed during the rest of the time. In principle, only four PC scans, one for each O₂ condition under steady state, were needed for the study. We, however, acquired PC MRI continuously throughout the session (except when TRUST is being acquired) because (1) the PC scan is very short in duration and a continuous acquisition would allow us to evaluate CBF changes in transitional states while FiO₂ changes; (2) multiple PC scans acquired before and after the TRUST scan would allow us to interpolate the CBF values to better match the timing of Y_v values from TRUST MRI.

The TRUST technique uses the spin labeling principle to isolate pure venous blood signals and measures T₂ value of the blood, followed by conversion of T₂ to O₂ saturation fraction with a calibration plot (Lu and Ge, 2008). Sequence parameters of TRUST MRI were single slice intersecting superior sagittal sinus at ∼10 mm above the sinus congruence, voxel size=3.44 × 3.44 × 5 mm³, repetition time=8,000 ms, inversion time=1,200 ms, τ_CPMGT₂ preparation with MLEV16 phase cycling, τ_CPMG=10 ms, four effective echo time (TEs)=0, 40, 80, and 160 ms, scan duration=3.2 minutes.

Quantitative PC MRI measures blood flow using magnetic field gradients (Xu et al, 2009). The phase of the magnetization is associated with the velocity value. The sequence parameters of PC MRI were single slice at the same position as TRUST MRI, voxel size=0.45 × 0.45 × 5 mm³, field-of-view=230 × 230 × 5 mm³, maximum velocity=80 cm/s, number of averages=4, duration=30 seconds.

Data Processing

The data processing procedures for TRUST MRI and PC MRI were based on algorithms described previously (Lu and Ge, 2008; Xu et al, 2009; see Supplementary text). For each FiO₂ condition, Y_v and CBF were obtained. The Y_v value was estimated from the corresponding TRUST scan. The CBF was obtained from the average of four PC scans before and four PC scans after TRUST MRI. The averaging improves data stability and also better matches the physiologic states between PC and TRUST acquisition periods (in case there is a physiologic drift during these periods).

Physiologic recordings of Y_a and EtO₂ were used to obtain other parameters needed for CMRO₂ calculation. The recorded values during the corresponding MRI acquisition period were averaged to obtain the final value used in the calculation. The CMRO₂ under each O₂ condition was estimated using equations (1–3). Using room-air condition as a reference, changes in physiologic parameters due to the special gas mixture were calculated.

Statistical Analysis

Statistical analyses were performed to examine whether each of the gas mixtures (i.e., FiO₂=14%, 50%, and 98%, respectively) resulted in an alteration in brain metabolic and vascular parameters. Specifically, a one sample Student's t-test was used to evaluate whether ΔCBF/CBF and ΔCMRO₂/CMRO₂ were significantly different from zero. We used fractional changes as opposed to absolute changes because the original experimental measures are affected by the brain size of the individual and the use of fractional changes provides a normalized measure. A P<0.05 is considered a significant effect.

Because hyperoxia and hypoxia may also change CO₂ content in the blood ([CO₂]_a) due to hyperventilation, the observed physiologic changes may be partially attributed to a CO₂ effect. We therefore corrected for the CO₂ effect and reanalyzed the data using the corrected values. The corrections were made by calculating:

$$\frac{\mathrm{\Delta }{\mathrm{C}\mathrm{M}\mathrm{R}\mathrm{O}}_2}{{\mathrm{C}\mathrm{M}\mathrm{R}\mathrm{O}}_2}{|}_{\mathrm{c}\mathrm{o}\mathrm{r}\mathrm{r}\mathrm{e}\mathrm{c}\mathrm{t}\mathrm{e}\mathrm{d}}=\frac{\mathrm{\Delta }{\mathrm{C}\mathrm{M}\mathrm{R}\mathrm{O}}_2}{{\mathrm{C}\mathrm{M}\mathrm{R}\mathrm{O}}_2}{|}_{\mathrm{u}\mathrm{n}\mathrm{c}\mathrm{o}\mathrm{r}\mathrm{r}\mathrm{e}\mathrm{c}\mathrm{t}\mathrm{e}\mathrm{d}}-f(\mathrm{\Delta }[{\mathrm{C}\mathrm{O}}_2{]}_{\mathrm{a}})$$

(4)

and

$$\frac{\mathrm{\Delta }\mathrm{C}\mathrm{B}\mathrm{F}}{\mathrm{C}\mathrm{B}\mathrm{F}}{|}_{\mathrm{c}\mathrm{o}\mathrm{r}\mathrm{r}\mathrm{e}\mathrm{c}\mathrm{t}\mathrm{e}\mathrm{d}}=\frac{\mathrm{\Delta }\mathrm{C}\mathrm{B}\mathrm{F}}{\mathrm{C}\mathrm{B}\mathrm{F}}{|}_{\mathrm{u}\mathrm{n}\mathrm{c}\mathrm{o}\mathrm{r}\mathrm{r}\mathrm{e}\mathrm{c}\mathrm{t}\mathrm{e}\mathrm{d}}-g(\mathrm{\Delta }[{\mathrm{C}\mathrm{O}}_2{]}_{\mathrm{a}})$$

(5)

where Δ[CO₂]_a is the alteration in CO₂ content during the O₂ challenge, the functions f(·) and g(·) represent the dependence of ΔCMRO₂/CMRO₂ and ΔCBF/CBF on Δ[CO₂]_a, respectively. f(·) and g(·) were assumed to be a second-order polynomial (consistent with the expression for O₂ dependence as described later):

$$f(\mathrm{\Delta }[{\mathrm{C}\mathrm{O}}_2{]}_{\mathrm{a}})=m\cdot \mathrm{\Delta }[{\mathrm{C}\mathrm{O}}_2{]}_{\mathrm{a}}+n\cdot \mathrm{\Delta }[\mathrm{C}{\mathrm{O}}_2{]}_{\mathrm{a}}^2$$

(6)

and

$$g(\mathrm{\Delta }[{\mathrm{C}\mathrm{O}}_2{]}_{\mathrm{a}})=p\cdot \mathrm{\Delta }[{\mathrm{C}\mathrm{O}}_2{]}_{\mathrm{a}}+q\cdot \mathrm{\Delta }[{\mathrm{C}\mathrm{O}}_2{]}_{\mathrm{a}}^2$$

(7)

The coefficients in equations (6) and (7) were estimated by fitting of experimental data from an earlier report (Xu et al, 2011), in which [CO₂]_a was specifically maneuvered to examine its effect on CMRO₂. Note that [CO₂]_a is not only dependent on EtCO₂, but also on arterial O₂ saturation level (Y_a) because of the Haldane effect (Loeppky et al, 1983). In this study, the calculation of [CO₂]_a accounted for both factors (see Supplementary text).

Using equations (6) and (7) and data in Xu et al (2011), the coefficients, m, n, p, and q, were found to be −0.006, −0.029, 0.068, and 0.098, respectively. To further examine error propagation from these coefficient values to the results of the present study, we conducted Monte Carlo simulations based on covariance matrices of m, n, p, and q (see Supplementary text).

Because CMRO₂ and CBF alterations during O₂ modulation are most likely mediated by O₂ content in the arterial blood and that different individuals may manifest different responses in arterial O₂ even given identical FiO₂, we conducted further analyses to directly examine the dependence of ΔCMRO₂/CMRO₂ and ΔCBF/CBF on Δ[O₂]_a. A mixed effect regression model (R software, Wirtschaftsuniversität Wien Vienna University, Austria) was used in which Δ[O₂]_a was the independent variable and had three observations for each subject, and the respective physiologic parameter was used as the dependent variable. Second-order polynomials, i.e.,

$$\frac{\mathrm{\Delta }{\mathrm{C}\mathrm{M}\mathrm{R}\mathrm{O}}_2}{{\mathrm{C}\mathrm{M}\mathrm{R}\mathrm{O}}_2}=a\cdot \mathrm{\Delta }[{\mathrm{O}}_2{]}_{\mathrm{a}}+b\cdot \mathrm{\Delta }[{\mathrm{O}}_2{]}_{\mathrm{a}}^2$$

(8)

and

$$\frac{\mathrm{\Delta }\mathrm{C}\mathrm{B}\mathrm{F}}{\mathrm{C}\mathrm{B}\mathrm{F}}=c\cdot \mathrm{\Delta }[{\mathrm{O}}_2{]}_a+d\cdot \mathrm{\Delta }[{\mathrm{O}}_2{]}_{\mathrm{a}}^2$$

(9)

were used for the model as the data (see Results and Discussion) suggested that the second-order model provides a better fitting than the linear model, as determined by a hierarchical regression analysis. Error propagation from the CO₂ coefficients to O₂ coefficients was again assessed by Monte Carlo simulations (see Supplementary text).

Results

All subjects were able to complete the breathing tasks without discomfort. Table 1 summarizes the vital signs during each breathing condition. Compared with room air, hyperoxia increased EtO₂, Y_v, [O₂]_a, and [O₂]_v, while hypoxia decreased them (P<0.001 for all tests), as expected. Y_a was reduced by hypoxia (P<0.001), but was minimally affected by hyperoxia as the values were already close to unity at room-air breathing. Both hyperoxia and hypoxia reduced EtCO₂ (P<0.001) (possibly due to hyperventilation), although the absolute values of the EtCO₂ changes were small (1 to 3 mm Hg). Hypoxia appeared to also increase the heart rate (P<0.001), similarly to previous reports (Tuunanen and Kauppinen, 2006).

Table 1

Vital signs under various FiO₂ conditions (N=16, mean±s.e.)

FiO 2	EtO 2 (mm Hg)	Y a (%)	Y v (%)	[O 2 ] a (μmol/mL)	[O 2 ] v (μmol/mL)	[O 2 ] a -[O 2 ] v (μmol/mL)	EtCO 2 (mm Hg)	[CO 2 ] a (μmol/mL)	HR (beats/min)	BR (breaths/min)
21%	121.3±0.8	98±0.1	65.1±1.4	8.6±0.2	5.6±0.2	3.0±0.1	42.4±0.9	19.7±0.2	65±3	13±1
14%	67.4±0.7	87±0.9	54.6±1.1	7.5±0.2	4.7±0.1	2.8±0.1	41.2±0.8	19.7±0.2	75±3	14±1
50%	336.5±0.9	99±0.03	70.5±1.5	8.9±0.2	6.1±0.2	2.8±0.1	41.2±0.8	19.4±0.2	62±3	15±1
98%	692.9±2.1	99±0.05	75.7±1.9	9.3±0.1	6.5±0.3	2.8±0.1	39.2±0.8	18.9±0.2	64±3	15±1

EtO₂/CO₂, end-tidal O₂/CO₂; Y_a/v, arterial/venous blood oxygenation; [O₂]_a/v, arterial/venous blood O₂ content; HR, heart rate; BR, breathing rate; FiO₂, fraction of inspired O₂.

Figure 2 shows representative imaging data for the measurements of venous oxygenation (panels A to C) and CBF (panel D). Figure 2A illustrates control, tagged, and difference images from TRUST MRI. The difference signals were then fitted as a function of effective TE to obtain the decay time constant T₂ (Figure 2B). Blood T₂ values are then converted to oxygenation using a hematocrit-specific calibration plot (Figure 2C). Figure 2D shows flow velocity maps obtained from PC MRI. Darker color indicates faster flow velocity at head-to-foot direction.

Figure 2

Representative magnetic resonance imaging (MRI) data obtained in the experiment. (A) Typical T₂-Relaxation-Under-Spin-Tagging (TRUST) MRI data under different fraction of inspired O₂ (FiO₂) conditions. Under each condition, control and tagged types of images were acquired, the subtraction of which yielded pure blood signal. In this study, the venous signal in the primary draining vein, sagittal sinus (center of the yellow box), was used for quantitative analysis. (B) In the TRUST sequence, the signal was acquired at different effective echo time (TE) values. Thus, the fitting of the signal as a function of effective TE can provide an estimation of blood T₂. The legend shows that T₂ increases with FiO₂ value. (C) Using a calibration plot established previously, the blood T₂ can be converted to blood oxygenation, given the subject's hematocrit value. (D) Representative phase-contrast MR images under different FiO₂ conditions. In these images, head-to-foot flow direction is displayed as black color. Thus, the darker the voxel appears, the higher the flow velocity is. The color reproduction of this figure is available at the Journal of Cerebral Blood Flow and Metabolism journal online.

Table 2 summarizes the CMRO₂ and CBF values under different O₂ conditions before and after correction of CO₂ effects. The raw data (i.e., before correcting for CO₂ effects) revealed that inhalation of 14% O₂ significantly increased CMRO₂ (N=16, P=0.02) and CBF (P=0.001), while 50% FiO₂ decreased them (P<0.001 and P=0.01 for CMRO₂ and CBF, respectively). Inhalation of 98% O₂ decreased them further. After correction for CO₂ effects, the results became slightly different (Table 2). Figure 3 shows the CMRO₂ and CBF changes due to hypoxia and hyperoxia based on the CO₂ effect corrected data. Hypoxia increased both CMRO₂ (5.0±2.0%, mean±s.e., P=0.02) and CBF (9.8±2.3%, P<0.001). However, hyperoxia did not alter CBF, but decreased CMRO₂ by 10.3±1.5% (P<0.001) and 16.9±2.7% (P<0.001) for FiO₂ of 50% and 98%, respectively.

Table 2

CBF and CMRO₂ values as a function of FiO₂ before and after correction for CO₂ effect

FiO 2	CMRO 2 (μmol/mL)		CBF (mL/min)
	Before correction	After correction	Before correction	After correction
21% O2	1,112.7±22.5	1,112.7±22.5	384.7±17.3	384.7±17.3
14% O2	1,166.4±28.8	1,166.8±29.1	421.8±19.3	421.4±19.0
50% O2	1003.7±27.7	997.8±27.0	365.6±14.3	377.3±15.0
98% O2	940.2±36.0	915.2±33.6	344.6±13.1	387.7±14.9

O₂, oxygen; CBF, cerebral blood flow; CMRO₂, cerebral metabolic rate of O₂; FiO₂, fraction of inspired O₂.

Figure 3

Percent changes in cerebral metabolic rate of oxygen (CMRO₂) and cerebral blood flow (CBF) due to fraction of inspired O₂ (FiO₂) modulation. The changes were calculated based on comparisons between the special gas mixture and room air.

Dynamic acquisitions of CBF MRI allowed us to obtain a time course of this parameter and the group-averaged data (corrected for CO₂ effect) are shown in Figure 1. Consistent with the quantitative results described above, the time course revealed a gradual CBF increase during the hypoxia period, but did not significantly change during hyperoxia.

A scatter plot of corrected ΔCMRO₂/CMRO₂ versus Δ[O₂]_a is shown in Figure 4A. Regression analyses revealed a significant inverse relationship between these parameters (P<0.001). A second-order polynomial was found to provide a better fitting of the data compared with a first-order model (second-order versus first-order comparison P=0.007). The coefficient values and their confidence intervals are listed in Table 3. A third-order model was not better than the second-order model (P=0.89). Figure 4B shows a scatter plot of corrected ΔCBF/CBF versus Δ[O₂]_a. The data were well fitted by a regression model, with a second-order polynomial (Table 3) again providing a better fitting (P=0.025) than a linear model. A third-order model was not better than the second order (P=0.26).

Table 3

Coefficient values and their confidence intervals describing the dependence of CMRO₂ and CBF on Δ[O₂]_a

	$\frac{\mathrm{\Delta }{\text{CMRO}}_2}{{\text{CMRO}}_2}=a\cdot \mathrm{\Delta }{[{\mathrm{O}}_2]}_a+b\cdot \mathrm{\Delta }{[{\mathrm{O}}_2]}_a^2$		$\frac{\mathrm{\Delta }\text{CBF}}{\text{CBF}}=c\cdot \mathrm{\Delta }{[{\mathrm{O}}_2]}_a+d\cdot \mathrm{\Delta }{[{\mathrm{O}}_2]}_a^2$
Coefficient	a	b	c	d
Value	−0.15	−0.056	−0.046	0.038
95% CILL	−0.17	−0.095	−0.073	0.0052
95% CIUL	−0.12	−0.017	−0.019	0.071
P value	<0.001	0.007	0.002	0.025

CILL, confident interval lower limit; CIUL, confident interval upper limit; CBF, cerebral blood flow; CMRO₂, cerebral metabolic rate of O₂.

Figure 4

Scatter plots comparing (A) ΔCMRO₂/CMRO₂ with Δ[O₂]_a and (B) ΔCBF/CBF to Δ[O₂]_a. Even given the same fraction of inspired O₂ (FiO₂) gas type, different individuals manifest slightly different Δ[O₂]_a values. Thus, the use of Δ[O₂]_a allows a more accurate assessment of the dependence of brain physiology on O₂ content. CBF, cerebral blood flow; CMRO₂, cerebral metabolic rate of oxygen.

Discussion

To the best of our knowledge, the present study is the first to concomitantly examine the influence of hypoxia and hyperoxia on brain O₂ metabolism in conscious humans. Our data suggested that hypoxia by inhalation of 14% O₂ increased CMRO₂, while hyperoxia by inhalation of 50% O₂ decreased CMRO₂. Further reduction in CMRO₂ was observed when increasing the FiO₂ to 98%. The present study also revealed that CBF was augmented by hypoxia, but was unchanged during hyperoxia.

Physiological Considerations

The influence of FiO₂ on brain metabolism has previously been investigated by a limited number of studies (the relatively small number is mainly due to technical complexity and availability of CMRO₂ measurement). The results were heterogeneous. A few studies reported findings similar to ours. For example, Richards et al, (2007) measured O₂ metabolism in beagles using a ¹³C NMR method and found that hyperoxia treatment after ischemia reduced O₂ metabolism. Working with a rat model, Harik et al, (1995) measured glucose metabolism with a 2-deoxyglucose method and observed that hypoxia with 10% FiO₂ increased glucose metabolism by 10% to 40%. A preliminary report by Smith et al, (2011) also noted an elevated CMRO₂ during hypoxia in humans, using techniques similar to ours. A few other reports contradict our findings. Comparing high altitude (∼11% FiO₂) to sea level, Moller et al, (2002) found no changes in CMRO₂ or CBF using a ¹³³Xe technique. Rockswold et al, (2010) studied O₂ metabolism in brain injury patients using a nitrous oxide technique and found that hyperoxia increased CMRO₂, but Diringer et al, (2007) found no CMRO₂ changes. Several reasons may have contributed to these discrepancies, including different species, experimental conditions as well as pathological effects (Diringer et al, 2007; Maandag et al, 2007; Moller et al, 2002; Rockswold et al, 2010). Collectively, these factors make a direct comparison virtually impossible. The techniques used in the previous studies required the injection of an exogenous tracer as well as continuous sampling of arterial and venous blood, all of which may present physiologic stress and alteration to the brain (Moller et al, 2002; Rockswold et al, 2010). The present study capitalized on a recently developed global CMRO₂ technique that was fast, reliable, and noninvasive (Xu et al, 2009). The relatively simple procedure afforded by this technique allowed us to apply both hypoxia and hyperoxia in the same session.

If our findings of CMRO₂ alteration reflect a corresponding change in neural activity, what could the possible mechanisms be? Hyperoxia has been long known to increase the generation of reactive oxygen species (Jamieson et al, 1986), which could cause oxidative damage to lipids and proteins (Tatarkova et al, 2011). Reactive oxygen species may in turn decrease enzymatic activities in aerobic metabolism pathways via inhibition of pyruvate dehydrogenase (Bogaert et al, 1994), a critical enzyme in transforming pyruvate into acetyl-CoA for the citric acid cycle. A few reports have also suggested that hyperoxia may decrease gene expression associated with neurotransmitter transport and transmission (Chen et al, 2009). Finally, studies have shown that long-duration hyperoxia (>1 hour) may lead to apoptosis by modification of antiapoptotic proteins (Brutus et al, 2009; Mudduluru et al, 2010). This evidence suggests that hyperoxia may suppress neural activity and therefore CMRO₂ via several O₂ toxicity-related pathways. The effect of hypoxia on CMRO₂ supposedly operates via mechanisms opposite to that of hyperoxia. Previous studies have suggested that hypoxia with 8% FiO₂ can cause a 57% increase in the rate of glycolysis and a threefold increase in the activity of cytochrome oxidase, an enzyme in the mitochondrial electron transport chain (Hamberger and Hyden, 1963). Furthermore, these changes were primarily located in neurons but not in glial cells (Hamberger and Hyden, 1963), suggesting an enhancement in neuronal activity during hypoxia.

Compared with CMRO₂, more literature is available on the effect of FiO₂ on CBF. Our observation of an increased CBF during hypoxia is consistent with earlier reports (Noth et al, 2008). For a hyperoxia challenge, we observed a 11.7±1.3% CBF reduction in the presence of a 3.2±0.4 mm Hg EtCO₂ decrease (for a 98% FiO₂). Given the modulation effect of CO₂ on CBF, it seems that the observed CBF change is predominantly attributed to a CO₂ effect rather than to an O₂ effect. Indeed, after correcting for the CO₂ effect, no apparent effect of O₂ was detected. Similar findings have been reported in the literature. For example, observed a 16% CBF reduction in the context of a 3 mm Hg EtCO₂ decrease comparing 100% FiO₂ with 21% FiO₂. However, a few reports in the literature showed some discrepancy from our results. used Arterial-Spin-Labeling MRI to evaluate the effect of hyperoxia on CBF and noted a 28% reduction in hyperoxia. One possible explanation for this finding is that arterial T₁ is also affected by hyperoxia. Specifically, hyperoxygenation causes a blood T₁ reduction by ∼30% (Silvennoinen et al, 2003), which would result in a lower Arterial-Spin-Labeling signal even if the CBF value did not change. The PC MRI technique used in the present study does not depend on arterial T₁. used a Dynamic-Susceptibility-Contrast MRI technique and observed a regional CBF reduction in some, but not all brain regions. Interestingly, the authors also conducted a large vessel-based measure using transcranial Doppler techniques, and found no CBF reduction, which is similar to our large-vessel measure with PC MRI.

Technical Considerations

The CMRO₂ method used in the present study was based on the Fick principle of arteriovenous differences in O₂ contents, in which the contributing parameters were measured individually (Xu et al, 2009). This principle has been known for decades, but its implementation in humans has been challenging. The main technical obstacle was the difficulty of estimating venous oxygenation quantitatively and reliably. With a recent TRUST MRI technique that was developed by our laboratory, the quantification of global venous oxygenation becomes more feasible (Lu and Ge, 2008). The TRUST technique has been validated in humans against a gold-standard pulse oximetry method (Lu et al, 2012). Another parameter in the Fick principle, CBF, was determined with a PC MRI technique. Compared with other CBF quantification techniques such as Arterial-Spin-Labeling MRI, the PC technique is not affected by arterial transit time, blood and tissue T₁ values, and is thus expected to provide a more accurate estimation of global CBF. Phase-contrast MRI has previously been validated under both in-vitro and in-vivo conditions (Bakker et al, 1999; Evans et al, 1993; Zananiri et al, 1991).

Since changes in alertness during the experiment may alter CMRO₂ (in addition to any O₂ effect) (Braun et al, 1997; Nofzinger et al, 2002), the subjects of this study were allowed to view a movie while in the scanner. The postMRI questionnaire confirmed that all subjects remained awake during the entire session. Another physiologic variable that was controlled was the end-tidal CO₂ level. The subjects were instructed to maintain a uniform breathing pattern throughout the experiment. However, due to O₂ effects on chemoreceptors in the brain (Dean et al, 2004), hyperventilation is known to occur during a hyperoxia challenge, which results in a reduced EtCO₂ as documented by a number of studies in the literature (Baddeley et al, 2000; Dean et al, 2004; Guyton and Hall, 2005). To minimize CO₂ changes during the hyperoxia challenge, we added a small amount of CO₂ to the hyperoxic air (1% and 2% CO₂ to FiO₂ 50% and 98%, respectively). This procedure was found to be partially effective, especially for the 50% FiO₂ condition where the EtCO₂ change is now only ∼1 mm Hg (Table 1). To further minimize the influence of CO₂ change on our data interpretation, a correction was made based on the literature reports of CMRO₂ and CBF dependence on [CO₂]_a, as described in the Materials and methods. Because the correction curve itself may contain experimental errors, we conducted an error propagation analysis using Monte Carlo simulations (see Supplementary text). We found that the coefficients associated with the CMRO₂ and O₂ relationship were only modestly affected by the correction curve. Within our simulations (1,000 m and n sets), the coefficients a and b in equation (8) were −0.14±0.019 (mean±s.d.) and −0.055±0.011, respectively, both of which were significantly less than zero (P<0.001).

In the estimation of venous O₂ pressure (thereby the amount of dissolved O₂), we did not account for the pH influence on the O₂ dissociation curve (known as the Bohr Effect). Although changes in [CO₂]_a may alter pH, thereby shifting the curve, the impact of this effect on our CMRO₂ estimation is expected to be minimal as the dissolved O₂ in the venous blood is negligible compared with hemoglobin-bound O₂.

Implications for Calibrated Functional Magnetic Resonance Imaging

It has recently been proposed that hyperoxia can be used as a physiologic challenge in calibrated fMRI (Chiarelli et al, 2007). The previous study has assumed that hyperoxia does not alter CMRO₂. In the context of a CMRO₂ reduction as suggested in the present study, the theoretical framework described in Chiarelli et al remains valid. However, the experimental implementation should be adjusted slightly. Specifically, one could consider adding a TRUST scan during both normoxia and hyperoxia periods (which gives [dHb]_v0 and [dHb]_v) and then estimating M using equation (3b) in Chiarelli et al (2007).

One may also speculate the use of hypoxia for calibrated fMRI. However, since arterial blood also contains deoxyhemoglobin, the model becomes considerably more complicated due to the need to account for arterial oxygenation and arterial blood volume. More work is therefore needed to explore the feasibility of hypoxia for fMRI calibration.

Limitations of the Study

The findings from the present study should be interpreted in view of several limitations. One is that the technique used in the present study could only measure global CMRO₂, but no regional values were obtained. Therefore, while the data suggested a clear effect of O₂ on brain metabolism, it was not possible to evaluate potential regional heterogeneity of this effect or whether certain brain regions are more sensitive to O₂ challenge. Another limitation is that the study has only included young, healthy control subjects, thus the results may not be applicable to elders or patients with traumatic brain injury or brain ischemia. Finally, despite our efforts to control and correct for CO₂ effects, it would have been ideal to use a CO₂ clamping method (Wise et al, 2007) in the experiments. The correction approach used in the present study is based on the literature results on the influence of CO₂ on CMRO₂ and CBF. It should be noted, however, that these literature data were measured under normoxia conditions. Under hyperoxia or hypoxia conditions, the relationship between CMRO₂/CBF and CO₂ may be different (i.e., there could be an interaction term between O₂ and CO₂ effects). There is some evidence suggesting that the interaction term is negligible for CBF (Floyd et al, 2003). However, future studies are needed to examine the interaction effect on CMRO₂.

Conclusions

The present study suggests that, aside from the expected effect on blood O₂ saturation, hypoxia enhanced brain metabolic rate while hyperoxia suppressed it. Additionally, CBF was increased in hypoxia but showed no apparent changes during hyperoxia.