Hypoxia increases cerebral blood flow (CBF), but it is unknown whether this increase is uniform across all brain regions. We used H215O positron emission tomography imaging to measure absolute blood flow in 50 regions of interest across the human brain (n = 5) during normoxia and moderate hypoxia. Pco2 was kept constant (∼44 Torr) throughout the study to avoid decreases in CBF associated with the hypocapnia that normally occurs with hypoxia. Breathing was controlled by mechanical ventilation. During hypoxia (inspired Po2 = 70 Torr), mean end-tidal Po2 fell to 45 ± 6.3 Torr (means ± SD). Mean global CBF increased from normoxic levels of 0.39 ± 0.13 to 0.45 ± 0.13 ml/g during hypoxia. Increases in regional CBF were not uniform and ranged from 9.9 ± 8.6% in the occipital lobe to 28.9 ± 10.3% in the nucleus accumbens. Regions of interest that were better perfused during normoxia generally showed a greater regional CBF response. Phylogenetically older regions of the brain tended to show larger vascular responses to hypoxia than evolutionary younger regions, e.g., the putamen, brain stem, thalamus, caudate nucleus, nucleus accumbens, and pallidum received greater than average increases in blood flow, while cortical regions generally received below average increases. The heterogeneous blood flow distribution during hypoxia may serve to protect regions of the brain with essential homeostatic roles. This may be relevant to conditions such as altitude, breath-hold diving, and obstructive sleep apnea, and may have implications for functional brain imaging studies that involve hypoxia.
- obstructive sleep apnea
the global response of the cerebral vasculature to changes in arterial oxygen tension has previously been described (3, 5, 7, 24). Cerebral blood flow (CBF) increases as arterial saturation of oxygen (SaO2) falls (5, 7, 23). While there are numerous papers describing the global response of CBF to hypoxia, there is little information on the regional distribution of CBF during hypoxia; the information that does exist is confounded by simultaneous Pco2 changes. We hypothesize that there would be significant differences among brain regions in the blood flow response to hypoxia in the absence of changes in Pco2.
The most common experimental intervention to produce cerebral hypoxia is exposure to low inspired Po2 without control of other variables. In such experiments, subjects have been allowed to breathe spontaneously during the intervention, and as a consequence the reflex increase in alveolar ventilation also markedly reduced arterial Pco2 (PaCO2). Low PaCO2 reduces CBF, thereby opposing the effect of low Po2 on CBF (10, 38). Outside the laboratory, a similar situation exists at high altitude. In contrast, when cerebral hypoxia is produced by respiratory disease, asphyxia, or ischemia, the Pco2 is generally normal or elevated. Here we held PaCO2 constant during hypoxia and used H215O positron emission tomography (PET) to image the CBF response.
It is not known whether the vascular response to hypoxia is uniform across the brain: some evidence suggests that the response to hypoxia is uniform (33); other evidence using autoradiography (45), video microscopy (17), MRI (36), and PET (4, 20) suggests that some brain regions receive a greater CBF increase. Nonuniform flow responses could be important in situations such as altitude and breath-hold diving and in clinical situations such as sleep apnea; more responsive regions may be better protected during these hypoxic episodes.
The present study used H215O PET to measure regional cerebral blood flow (rCBF) in five healthy men during hypoxia with constant PaCO2. The concentration of radioactivity in arterial blood was measured continuously throughout the PET scans to provide an input function that allows quantitative estimates of absolute blood flow to be obtained. We measured rCBF during periods of moderate hypoxia (SaO2 = 80%) and periods of normoxia (SaO2 = 97%). PaCO2 was maintained constant throughout the study by controlling ventilation and inspired Pco2. Mean regional blood flow responses for 50 functional regions of interest (ROIs) are reported, and the data revealed that there is a nonuniform vascular response to normocapnic hypoxia in humans.
MATERIALS AND METHODS
This study was approved by the Human Subject Committees at Imperial College, London, and at Harvard School of Public Health. Six subjects, all men and all righthanded, gave informed consent. Before the brain imaging study, subjects practiced the protocol in the laboratory so that they were familiar with the procedures, and we could ensure that they were appropriate candidates (i.e., there were no ECG abnormalities during hypoxia). One normal, healthy subject was excluded from the study after cardiac arrhythmias were noted during exposure to hypoxia in a practice session (a normal heart rhythm returned when normoxia was restored). We were unable to replace the excluded subject due to logistical constraints.
Tidal volume (Vt) and respiratory frequency were controlled by mechanical ventilation (Siemens 900B in volume control mode) via a mouthpiece at an initial minute ventilation approximate to 0.16 l·min−1·kg body wt−1, and a respiratory rate of 12 breaths/min. Airway pressure, Vt, and respiratory frequency were recorded from the analog output of the ventilator. Heart rate and SaO2 were measured from the index finger using a pulse oximeter (SpO2) (Biox 3700, Ohmeda, Louisville, CO). Because hypoxia poses a cardiac risk, subjects were monitored using a 12-lead ECG; for safety, maximal hypoxia was limited to SpO2 not less than 70%. End-tidal Pco2 (PetCO2) and Po2 (PetO2) were measured at the mouth (Capnograph, PK Morgan, Haverhill, MA; LB1, Liston Scientific, Irvine, CA, respectively) and corrected to give estimates of arterial levels (22).
PET scans were carried out on a CTI 966 PET scanner (Siemens, Oxford, UK) with a 24-cm-deep field of view. The field of view was positioned to include the whole brain, including cerebellum and brain stem. The mean spatial resolution was 4.8 ± 0.2 mm (transaxial, 1 cm off-axis) and 5.6 ± 0.5 mm (axial, on-axis).
Each subject underwent 15 scans (3 sessions of 5 scans, see below) following a bolus intravenous injection of H215O (0.5 mSi/scan). The concentration of H215O was measured continuously in arterial blood drawn from the radial artery throughout each scan (41).
There were three interventions made: 1) normoxia with baseline Vt (mean = 0.86 ± 0.12 liter); this intervention produced no dyspnea; 2) hypoxia with baseline Vt (mean = 0.89 ± 0.11 liter); although this intervention was intended to cause dyspnea, it was not consistently effective; and 3) hypoxia with large Vt (mean = 1.43 ± 0.13 liter); this condition produced no dyspnea. Each intervention was 8 min long, and an imaging session (5 images) consisted of one of each type presented in a randomized order. Data for this report were generated during a study that was partially designed to investigate the neural activations associated with dyspnea caused by hypoxia. The dyspnea study was not published, because the intervention intended to cause dyspnea failed to cause adequate dyspnea within the hypoxia safety limit. This did not affect our second objective, because the level of hypoxia achieved did produce significant increases in CBF.
During the hypoxia interventions, inspired oxygen was reduced from 20 to 10% (inspired Po2 = 70 Torr), which reduced SpO2 to a mean value of ∼80%. This level of hypoxia is expected to induce a substantial vascular response [threshold for vascular response is 90% (14)]. To avoid changes in CBF associated with changes in PaCO2, we maintained PetCO2 constant. Vt, frequency, and inspired CO2 fraction (FiCO2) were the same during conditions 1 and 2; thus alveolar ventilation and PetCO2 did not differ. Condition 3 employed a higher Vt; thus we increased FiCO2 to keep PetCO2 constant in the face of increased alveolar ventilation. In all conditions, continual fine tuning of FiCO2 was done by an experimenter using feedback from the capnometer trace to keep PetCO2 at the desired value (9).
Quantitative functional images of rCBF were generated as described by Lammertsma et al. (26) using the integrated tissue radioactivity from 30 to 120 s after scan start. The time delay associated with the blood counter was measured by comparison of the blood time activity curve with that of the total count rate of the scanner. A lookup table relating the expected integral tissue activity to blood flow rate was then generated, assuming a dispersion effect in the measured blood with a time constant of 8 s, together with an equilibrium partition coefficient for H215O of 0.86 ml blood/ml brain tissue. The lookup table was then used to convert the integral images to quantitative CBF images on a pixel-by-pixel basis.
Flow images were placed into standard brain space [Montreal Neurological Institute Space (12); Statistical Parametric Mapping 99 (SPM99; Wellcome Dept. of Cognitive Neurology, Institute of Neurology, London; http://www.fil.ion.ucl.ac.uk/spm)]. Images were then spatially smoothed (4 mm full-width half-maximum). Because inclusion of CSF or white matter in any of the brain volumes would likely cause an underestimation of blood flow, white matter and CSF were masked out. To do this, we removed areas with a signal < 80% of the mean signal (SPM99) before further analysis was performed. The remaining gray matter images were then divided into 50 ROI using in-house software (Marsbar, SPM99) (15). (See Table 2 for the ROI designations.) Absolute mean blood flow was calculated for each ROI for periods of hypoxia and normoxia.
Periods of hypoxia were compared with periods of normoxia for each subject and for the group; the absolute and relative (percentage) differences in blood flow were calculated for each ROI. Interventions 2 and 3 were grouped as the hypoxia condition (as there was no significant difference between the flow changes seen in these interventions, analysis of covariance). Intervention 1 was the normoxia intervention. All values are means ± SD.
Reduction in inspired O2 levels caused PetO2 to fall significantly (P < 0.05) from 93.8 ± 4.8 to 45.5 ± 6.3 Torr. SpSpO2 fell from 96.6 ± 0.8 to 80.0 ± 7.4%. There was a wide range in the saturation change induced: from 7.8% (subject 2) to 26.1% (subject 5). The difference in PetCO2 between hypoxic and normoxic conditions was calculated for each subject (range, −0.5 to 0.5 Torr). The group mean change in PetCO2 between the two conditions was 0.06 ± 0.43 Torr (T-test, P = 0.77). The absolute mean values for PetCO2 were 43.6 ± 1.5 Torr during normoxia and 43.5 ± 1.4 Torr during hypoxia. Arterial saturation and end-tidal gas data are shown in Fig. 1. Physiological data for each subject are shown in Table 1.
Changes in CBF.
Whole brain measurements showed that hypoxia significantly increased global CBF (gCBF) in four of the five subjects. Mean gCBF was 0.39 ± 0.10 ml·g−1·min−1 during normoxia and significantly increased (T-test, P < 0.05) to 0.45 ± 0.13 ml·g−1·min−1 during hypoxia (a 14.8% rise above normoxic levels). In general, the magnitude of the gCBF response was correlated with the level of desaturation induced (r2 = 0.82, linear regression; see Fig. 2).
The magnitude of the hypoxia-driven blood flow increase to an ROI was significantly related to the ROI's normoxic blood flow (ANOVA, P < 0.05), i.e., the greater the blood flow during normoxia, the greater the hypoxic response. Relative increases in gray matter blood flow were not uniform across the specified ROI. Increases ranged from only 0.04 ml·g−1·min−1 (9% increase) in the right occipital lobe to 0.12 ml·g−1·min−1 (28.9% increase) in the right nucleus accumbens. There was a trend for phylogenically “older” parts of the brain to receive both a greater absolute blood flow and a greater increase in blood flow. For example, the nuclei of the basal ganglia showed consistently higher increases in blood flow during hypoxia than most other brain regions (putamen, 19.7 ± 13.6%; pallidum, 18.5 ± 9.7%; caudate nucleus, 18.6 ± 11.8%), while the cerebral hemispheres generally received a smaller increase in blood flow (e.g., frontal lobe, 11.3 ± 8.8%; temporal lobe, 10.2 ± 11.1%; parietal lobe, 12.2 ± 10.1%). The absolute and relative changes in blood flow are shown for each ROI in Table 2. After designating each ROI as either “old” (archipallium and palleomammalian brain) or “new” (neopallium brain) (Ref. 28; see Table 2), we determined that blood flow response of old regions (16.9 ± 3.6%) was significantly greater (Wilcoxon test, P < 0.05) than the response of new regions (13.23 ± 2.6%). A summary of the rCBF changes for old and new brain is shown in Fig. 3.
The pattern of blood flow distribution was not significantly different among subjects (analysis of covariance, P = 0.34): this included subject 4, who, despite not showing a significant whole brain response, did show the same pattern of changes in the individual ROI as the other subjects.
We found that the increase in CBF during isocapnic hypoxia was not uniform: a number of brain regions received a proportionally greater increase in blood flow. The most prominent increases in blood flow were seen in the nuclei of the basal ganglia, as well as several other phylogenetically old brain regions, specifically the putamen, the thalamus, nucleus accumbens, and the pallidum.
Our study has confirmed previous findings, as well as shown new information. The gCBF increase (14.8%) observed at this level of hypoxia (Sp= 80.0 ± 7.4%) is similar to increases seen in previous studies (4, 30, 37). The present investigation, however, has determined blood flow changes more focused and numerous ROIs. The two earlier studies that measured rCBF during hypoxia used an intervention that produced both hypoxia and hypocapnia. Buck et al. (4) showed only a small decrease in Pco2 during hypoxia (1.5 Torr), but this may have been sufficient to cause a 3–4% change in CBF. However, we also note that the subjects in the study of Buck et al. started with relatively low Pco2, (36 Torr), and so hypoxia periods may have included a hypocapnic contribution to blood flow changes. Hypocapnia reduces gCBF (11, 20, 37, 40) and may have attenuated regional inhomogeneity. Nonetheless, these two studies detected relative increases in two of the regions we observed: basal ganglia (33) and thalamus (4).
The areas that received a proportionally greater blood flow during hypoxia are mostly perfused by the lenticulostriate arteries [branches of the middle cerebral artery (MCA)]. The responses of the main trunk of the MCA to hypercapnia (19), hypoxia (37, 39), and hyperoxia (44) have been extensively described, perhaps because of its accessibility to Doppler ultrasound, but no Doppler studies have compared the hypoxic responses of different cerebral vessels. It is unclear whether the MCA or the lenticulostriate arteries have a greater sensitivity to hypoxia than other cerebral vessels. Doppler evidence from the hypoxic human fetal brain suggests that the anterior cerebral artery has a more sustained vasodilation to chronic hypoxia than either the MCAs or posterior cerebral arteries (6), affording the frontal lobes of the fetal brain better protection than the “primitive brain” regions we have described in the adult brain. However, the fetal cerebrovasculature shows a different response to hypercapnia than the adult (18) and may show a different response to acute hypoxia, as used in this study. The response of the rat brain shows similar blood flow distributions to those we observed during hypoxic-ischemia, with the cerebral cortex receiving less protection than other brain regions (43) and hypoxic hypercapnia results in thalamic regions of the piglet brain being better perfused than the cerebral cortex (1). From this and previous studies, it appears that some brain regions are afforded better protection against hypoxia than others.
The cerebrovascular response protects the brain against hypoxia during conditions such as obstructive sleep apnea, suffocation, and drowning (situations in which there is no concurrent fall in Pco2, as during exposure to hypobaria). The nucleus accumbens had the greatest perfusion increase in our study, and this area has been shown to suffer less anoxic damage than other brain regions in humans (16) and in piglets (29). The mechanism by which certain regions of the brain are better protected from arterial gas changes is not clear, but may be related to the present findings. Ito et al. (20) also saw the “older” brain (pons, cerebellum, thalamus, and putamen) receiving proportionally greater increases in blood flow during hypercapnia than the frontal, temporal, occipital, and parietal cortexes. Hypoxia and hypercapnia-induced changes in CBF may share common mediators and/or pathways, although an obvious candidate, the peripheral chemoreceptor, does not appear to be involved with CBF control (31). Regional changes in pH can also influence local blood flow during both hypercapnia (25) and hypoxia (27). Hypoxic and hypercapnic changes in vascular tone may both be mediated by adenosine (35) or nitric oxide (34), although the direct role of the latter is less clear (13). The highest concentrations of the A2 adenosine receptor (that are thought to be involved with vascular tone) are found in the caudate, putamen, and nucleus accumbens (21): three regions receiving high blood flow during hypoxia. The regional distribution of the enzyme nitric oxide synthase in the rat (34) or human brain (2) does not seem to correlate with the distribution of blood flow seen in this study.
Preservations of the older brain regions responsible for homeostatic function (32) may have provided a selective advantage. When sudden and complete anemia is induced in the cat, the survival time of cerebral structures appears related to the regional blood flow measurements seen in this study, with older structures surviving 25–40 s and the cerebral cortex lasting only 14–15 s (42).
The heterogeneity of blood flow distribution during hypoxia complicates studies that use blood flow distribution as an indicator of brain function, e.g., functional MRI and H215O PET. Without compensating for the vascular changes associated with hypoxia per se, redistribution of blood flow could easily be misinterpreted as increases in local neural activity. Changes in arterial CO2 also present a similar problem (46), and, because PaCO2 is more easily influenced by changes in breathing than SpSpO2, extra precautions should be taken.
In summary, the changes in blood flow associated with hypoxia are not equally distributed across the brain. In general, greater flow is directed to the older brain (e.g., nucleus accumbens, putamen, pallidum, caudate, thalamus), possibly to maintain essential homeostatic functions, even at the cost of reduced cognitive function. This heterogeneity should be acknowledged and accounted for during functional brain imaging studies where hypoxia may be present.
This research was supported by National Heart, Lung, and Blood Institute grant HL46690, The Breathlessness Charitable Trust, and the Medical Research Council.
Our sincere thanks go to the subjects who participated in the study. We acknowledge the invaluable contribution of Andrew Blythe, Jasbir Grewal, and the staff at the Hammersmith Cyclotron Unit. Kevin Murphy of Charing Cross Hospital, London, provided superb support for this study, as well as Karl Evans and Claudine Peiffer, who assisted in the original study from which these data were derived.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2008 the American Physiological Society