The BOLD response and vascular reactivity during visual stimulation in the presence of hypoxic hypoxia

www.elsevier.com/locate/ynimg NeuroImage 41 (2008) 179 – 188

The BOLD response and vascular reactivity during visual stimulation in the presence of hypoxic hypoxia Yi-Ching L. Ho,a,b Rishma Vidyasagar,c Yuji Shen,d George M. Balanos,d Xavier Golay,a,e and Risto A. Kauppinend,⁎ a

Neuroradiology, National Neuroscience Institute, Singapore, Singapore Centre of Functionally Integrative Neuroscience, University of Aarhus, Aarhus, Denmark c Birmingham University Imaging Centre, School of Psychology, University of Birmingham, Birmingham, UK d School of Sport and Exercise Sciences, University of Birmingham, Birmingham, UK e Laboratory of Molecular Imaging, Singapore Bioimaging Consortium, Singapore, Singapore b

Received 26 October 2007; revised 19 February 2008; accepted 28 February 2008 Available online 6 March 2008

A disproportionate increase in cerebral blood flow (CBF) relative to the cerebral metabolic rate of oxygen (CMRO2), in response to neuronal activation, results in a decreased oxygen extraction fraction (OEF) and hence local ‘hyperoxygenation’. The mismatch is the key ‘physiological substrate’ for blood oxygenation level dependent (BOLD) fMRI. The mismatch may reflect inefficient O2 diffusion in the brain tissue, a factor requiring maintenance of a steep [O2] gradient between capillary bed and neural cell mitochondria. The aim of this study was to assess vascular responsiveness to reduced blood oxygen saturation, using both BOLD fMRI and the CBV-weighted vascular space occupancy (VASO)dependent fMRI technique, during visual activation in hypoxic hypoxia. Our fMRI results show decreased amplitude and absence of initial sharp overshoot in the BOLD response, while VASO signal was not influenced by decreasing oxygen saturation down to 0.85. The results suggest that the OEF during visual activation may be different in hypoxia relative to normoxia, due to a more efficient oxygen extraction under compromised oxygen availability. The data also indicate that vascular reactivity to brain activation is not affected by mild hypoxia. © 2008 Elsevier Inc. All rights reserved. Keywords: fMRI; BOLD; VASO; Vascular reactivity; Hypoxia; Visual; Oxygen limitation

Introduction The blood oxygen level dependent (BOLD) response is one of the most widely used functional imaging contrasts for brain activation studies. The BOLD effect provides a natural contrast for ⁎ Corresponding author. Department of Radiology, Dartmouth College, HB 7795, 706 Vail Hanover, NH 03755, USA. Fax +1 603 650 1717. E-mail address: [email protected] (R.A. Kauppinen). Available online on ScienceDirect (www.sciencedirect.com). 1053-8119/$ - see front matter © 2008 Elsevier Inc. All rights reserved. doi:10.1016/j.neuroimage.2008.02.048

functional magnetic resonance imaging (fMRI) due to the paramagnetic effect of deoxyhaemoglobin (Hb) on T2 (Brindle et al., 1979; Thulborn et al., 1982) and T2⁎ (Ogawa et al., 1990). As a result of venous ‘hyperoxygenation’ (i.e. decrease in Hb/HbO2 ratio) due to a decrease in the oxygen extraction fraction (OEF), the T2(⁎)weighted MR signal increases locally. Several fMRI studies have shown that the increase in CBF-to-CMRO2 ratios during brain activation ranges from 2 to 6 (Davis et al., 1998; Fox and Raichle, 1986; Hoge et al., 1999; Kim et al., 1999; Mandeville et al., 1999; Marrett and Gjedde, 1997). Physiological underpinnings linking the BOLD signal to brain activation are not fully understood, and the fundamental physiological mechanism underlying the mismatch between CBF and CMRO2 remains to be elucidated. According to the O2 limitation model (Buxton and Frank, 1997), a large CBF to CMRO2 ratio during brain activation is required to maintain a steep O2 gradient between the capillary space and site of tissue mitochondria, facilitating oxygen diffusion in the tissue. The results showing a consistent CBF-to-CMRO2 ratio, derived from additive BOLD responses to graded visual stimulation during elevated CBF baseline, have been used to support this key claim of the O2 limitation theory (Corfield et al., 2001; Hoge et al., 1999). Conversely, the O2 limitation model implies that a drop in arterial O2 tension should result in augmented CBF response and thus vasodilation in order to sustain a low OEF during brain activation. Assuming unidirectional O2 transport from capillaries and close to zero tissue O2 tension at mitochondrial sites, the physiological parameters can be related to each other by: CMRO2 ¼ OEF  CBF  Ca

ð1Þ

where CMRO2 is the cerebral metabolic rate of oxygen, Ca is the arterial oxygen content (Buxton and Frank, 1997). Despite the crucial need for O2 for brain function, however, mild to moderate hypoxic hypoxia appears to be well tolerated. Studies have shown that haemodynamics, metabolism, and higher brain functions

180

Y.-C.L. Ho et al. / NeuroImage 41 (2008) 179–188

are not disrupted even down to Ya of 0.8 (Mintun et al., 2001; Rostrup et al., 2005; Shimojyo et al., 1968; Siesjo, 1978; Tuunanen et al., 2006a). Stimulus-evoked CBF, as measured by PET and MRI techniques in conscious humans (Mintun et al., 2001; Rostrup et al., 2005; Tuunanen et al., 2006a; Tuunanen and Kauppinen, 2006) and anaesthetized animals (Sicard and Duong, 2005), proceeds at the same level at Ya of 0.8 as at 1 (full saturation). Several studies have reported that baseline CMRO2 does not change during transient moderate hypoxia (Emoto et al., 1988; Kety and Schmidt, 1948; Sicard and Duong, 2005). Tuunanen et al. (2006a) reported that visual evoked potentials were not affected by hypoxia down to Ya of 0.8, indicating that visual processing in the primary visual cortex is sustained. On the other hand, laser-doppler flowmetry has indicated increase in blood velocity in large arteries by 13% upon a decrease of Ya by 10% in humans (Meadows et al., 2004). Earlier papers on animals have shown increased CBF at oxygen saturation below 0.6 (Ekstrom-Jodal et al., 1979; Koehler et al., 1984; Siesjo, 1978). The effects of limited O2 availability on metabolic and vascular responses during brain activation are still unclear, and in this present study our aim was to determine the effects of hypoxic hypoxia, during visual stimulation, on the BOLD response and vascular responsiveness, as assessed using the CBV-weighted (Donahue et al., 2006) vascular space occupancy (VASO)-dependent technique (Lu et al., 2003). The benefit of the VASO technique is a high contrast-to-noise ratio that affords spatial and temporal resolution comparable to that obtained with BOLD fMRI. If O2 limitation at the tissue level applies, increased vascular reactivity, i.e. larger signal change is expected to be revealed by the VASO technique during visual stimulation in hypoxic hypoxia than for normoxia. Materials and methods Subjects and experimental design The protocol was approved by the Committee on the Ethics of Research on Human Beings of the University of Birmingham. Eight healthy subjects (5 males, 3 females; ages ranging from 24 to 51 years) gave written, informed consent prior to participation. One subject was subsequently excluded from the study due to discomfort during hypoxic exposure. The design of the study was to present visual stimulation in each of the two oxygenation states, with two sets of functional MRI scans (VASO and BOLD) for each of these conditions. Inspired oxygen tension (FIO2) was either 21% (room air) or 12% (O2 balanced with N2 in a non-rebreathing circuitry, delivered through a valve with a mouth piece by a device from Hans Rudolph Inc., Kansas City, KS, USA). Arterial oxygen saturation (Y used for saturation values from pulse oximeter) and pulse rate were monitored from a finger on the left hand with a Pulse Oximeter (System 4500 MRI, In Vivo Research, Inc., Orlando, USA). After an adaptation of about 5–7 min to 12% FIO2, during which a stable (variation ± 0.02) hypoxic Y level was reached, the MRI protocol was started. For visual stimulation, subjects were asked to fixate on the centre of a screen on which a black and white contrast-reversing checkerboard at 8 Hz frequency was displayed. Each run consisted of 5 blocks that alternated between baseline and stimulation (starting and ending with baseline) of 45 s per block. MR parameters MRI was performed using a Philips Achieva 3.0 T clinical imager (Philips Medical Systems, Best, The Netherlands) using

standard body coil transmission and SENSE head coil reception. Using a sagittal localiser, a coronal 5 mm slice was aligned to the calcarine sulcus to cover the visual cortex. T2⁎-weighted, dual-echo BOLD images were acquired with single shot, gradient echo (GRE), echo-planar imaging (EPI): TR = 3 s, TEs = 5 and 40 ms, flip angle = 90°, matrix = 128 × 128, FOV = 224 × 224 mm with partial Fourier = 0.67 and SENSE factor = 2.5. VASO images were acquired with an inversion-recovery pulse for nulling blood (Lu et al., 2003) at TI = 889 ms; other parameters were similar to those of BOLD fMRI, except the pair of TEs = 10.3 and 56 ms. Each BOLD and VASO time series consisted of 225 volumes. Data analyses Processing and statistical analyses of the fMRI scans were performed using in-house software written using IDL (ITT Visual Information Solutions, Boulder, CO, USA). Realignment of the data was done to correct for motion (AIR algorithm). T2⁎ was calculated from the double echoes in the GRE EPI images according to: T2⁎ ¼ ðTE2  TE1 Þ=ðln S1 =S2 Þ

ð2Þ

where TE1 and TE2 are the echo times, and S1, and S2 are the respective signal intensities at those echoes. The calculated T2⁎ data was then used for all subsequent analyses of the BOLD effect. All BOLD results shown refer to the absolute T2⁎ analyses, and the absolute ΔR2⁎( = 1/ΔT2⁎) values shown in conjunction with the BOLD results. T-tests were used to detect task-related T2⁎ (and R2⁎) changes in the fMRI time series on a voxel-by-voxel basis. Areas of significant activation were delineated by using a voxel threshold of t N 3.24, p b 0.001 (uncorrected) plus a cluster threshold of three voxels. The first five volumes of the baseline blocks and first volume of the stimulation blocks were not included in the t-tests to ensure steady-state signal intensity. Using the thresholded areas, fractional signal time courses were obtained from each scan and averaged across all stimulation cycles and subjects. This group of analyses was termed “All Voxels”. In addition to assessing thresholded BOLD and VASO responses specific to normoxia and hypoxia, we also determined the response amplitude in the thresholded areas common in both normoxia and hypoxia, i.e. in this case only voxels that were activated during both normoxia and hypoxia for each particular scan set were considered. This enables direct comparison of BOLD and VASO signal changes in the same parenchymal structures by excluding responses from areas that would be subthresholded in either of the two conditions. This data analysis procedure was termed “Common Voxels”. Using the baseline corrected data sets for “All Voxels” and “Common Voxels”, the differences in the shapes of the signal response curves in normoxia and hypoxia were analysed using the slopes of the response curves from the peak of T2⁎ response to the end of the stimulation block. The data were linearly regressed to determine estimates for the slopes. For analyses of mean response amplitudes, undershoots were analyzed for amplitudes by comparing the corresponding sections of time courses after the onset and termination of visual stimulation, respectively. The steady-state signal responses for baseline and stimulation conditions were calculated by averaging across each condition, but excluding the first five volumes from baseline data sets and first three volumes for the stimulation conditions, in order for steady-state comparisons. In addition, the BOLD and VASO signal changes in hypoxia

Y.-C.L. Ho et al. / NeuroImage 41 (2008) 179–188

Fig. 1. BOLD (T2⁎) and VASO maps overlaid onto the respective T2⁎-weighted and inversion-recovery MR images for a typical volunteer. (a) BOLD in normoxia (Y= 0.99); (b) BOLD in hypoxia (Y= 0.84); (c) VASO in normoxia (Y= 1.0); (d) VASO in hypoxia (Y= 0.93).

were normalized to the normoxic baseline, dividing both baseline and activation signal intensities by the mean normoxic baseline value. Vascular reactivity graphs were created by plotting individual subject BOLD and VASO data points as a function of Y levels. The non-parametric Spearman's ρ test was applied to assess significant correlations (p ≤ 0.05, two-tailed) between the Y levels and both the BOLD and VASO data. All other statistical analyses were done using paired t-tests with the same significance levels. Signal-tonoise ratios (SNR) were calculated for all BOLD (TE = 40 ms) and VASO scans in normoxia and hypoxia. SNR for each scan was taken as the mean divided by the standard deviation (SD) of the baseline values for the same gray matter region in the visual cortex in both normoxia and hypoxia. Values given are mean± SD.

181

66 ± 26% in VASO activated areas (Figs. 1 and 2). Even with the outliers included for the statistical analyses, however, a positive correlation was evident for the plot of BOLD voxel count against Y (p ≤ 0.01) while the correlation for VASO voxel counts was not significant (p = 0.09, two-tailed; p ≤ 0.05, one-tailed). The signal time courses for the T2⁎-BOLD signals in normoxia showed stimulus response overshoots and post-stimulus undershoots (Fig. 3a). However, the shapes of the normoxic and hypoxic overshoots differed quantitatively: The normoxic overshoot was very sharp at the initial portion and tapered down to a plateau during the rest of the stimulation period (mean slope of −0.05, amplitudes significant at p b 0.01). In contrast, no overshoot was evident during hypoxic stimulation, but rather there was a simple response plateau (mean slope of 0.00, p N 0.05). For each subject, the slopes were more negative in normoxia than hypoxia (p ≤ 0.01). Both the normoxic and hypoxic BOLD responses showed post-stimulus undershoots (p b 0.05, Fig. 3a). The post-stimulus undershoot amplitude determined in normoxia was not significantly different from that seen in hypoxia (p N 0.05). The baseline T2⁎ decreased in hypoxia by 11.7 ± 6.4% (p b 0.01), reflecting a decline in Y. Parenchymal baseline R2⁎ was 22.1 ± 0.4 s− 1 in normoxia. For accurate comparison of fractional responses with the normoxic condition, we normalized responses to the normoxic baseline as done previously (Sicard and Duong, 2005). After normalization, the calculated signal changes showed that the BOLD-related T2⁎ increase was larger in normoxia (4.2 ± 0.7%, ΔR2⁎ = −0.90 ± 0.2 s− 1) than in hypoxia (2.5 ± 0.4%, ΔR2⁎ = −0.77 ± 0.1 s− 1) (p b 0.01) (Fig. 4a). The BOLD response size was positively correlated with oxygen saturation (p b 0.001) as illustrated in Fig. 5a. VASO signal time courses were similar in shape both in normoxia and hypoxia (Fig. 3b). The baseline VASO signal was not different between normoxia and hypoxia (p N 0.05). The normalized taskrelated VASO signal change was −6.1± 1.3% in normoxia, whereas the response in hypoxia was −6.1 ± 1.8% (Fig. 4b). There was no discernable trend in the VASO responses at any Y studied (p N 0.05) (Fig. 5b).

Results Physiological parameters Hypoxic hypoxia increased the heart rate from 62 ± 9 to 74 ± 11 beats per minute (p b 0.05). Oxygen saturation decreased by 13% from 0.99 ± 0.01 to 0.86 ± 0.05 (p b 0.001). All activated voxels in BOLD and VASO scans BOLD, as indicated by the T2⁎ change, and VASO task-related activations were seen in both normoxia and hypoxia (VASO ‘activations’ are negative in sign). Areas showing activation for both BOLD and VASO fMRI were smaller in hypoxia than in normoxia for all subjects, except for one person in each fMRI scan. Excluding both outliers, there was a reduction by 55 ± 18% in BOLD and by

Fig. 2. Activated voxel counts as a function of Y for BOLD (squares) and VASO (triangles) fMRI. Voxel counts in hypoxia are normalized to normoxia for each subject. Only three data points are presented for normoxia (Y N 0.97) due to overlap. There was an outlier for the VASO dataset (value of 5.5) and for scaling purposes, it has been excluded from the Figure, yet all values were included in the statistical analyses. A significant decrease in active voxels was detected with lowering Y for BOLD (p ≤ 0.05), while the correlation for VASO voxel was not significant (p = 0.09).

182

Y.-C.L. Ho et al. / NeuroImage 41 (2008) 179–188

Fig. 3. BOLD (a, c) and VASO (b, d) signal time courses in normoxia and hypoxia for all volunteers. Y was 0.86 ± 0.05 in hypoxia. The gray horizontal bars indicate stimulation duration. Upper panels: “All Voxels”. Lower panels: “Common Voxels”.

Common (overlapping) voxels in BOLD and VASO scans Analysis of the BOLD signal trace (Fig. 3c) showed that similar to the “All Voxels”, there was a sharp stimulus response overshoot in the early part of activation in normoxia (mean slope of −0.06, p b 0.01) that was absent from the hypoxic activation (mean slope of 0.00). The slopes were different in normoxia and hypoxia (p ≤ 0.01). The poststimulus undershoot was present under both oxygenation conditions (p b 0.05), yet the undershoot amplitude was significantly smaller in hypoxia than in normoxia (p b 0.05, Fig. 3c.). Similar to “All Voxels”, the baseline parenchymal T2⁎ value in hypoxia was lower by 11.1 ± 5.4% relative to normoxia (p ≤ 0.01). Baseline R2⁎ was 22.4 ± 0.6 s− 1 in normoxia. Normalized task-related T2⁎-BOLD signal changes were significantly reduced in hypoxia (2.6 ± 0.5%%; ΔR2⁎ = − 0.79 ± 0.2 s− 1) relative to normoxia (5.4 ± 1.8%, ΔR2⁎ = −1.17 ± 0.3 s− 1, p ≤ 0.05) (Fig. 4c). The relationship is further illustrated by a positive correlation of BOLD response size with Y (p ≤ 0.001) and reactivity graph (Fig. 5c), which shows that the BOLD response decreased with lowering oxygen saturation. These results for BOLD data were similar to those in the “All Voxels” analysis; the main difference was that in the overlapping voxel hypoxia had a stronger effect on BOLD response size than in ‘All voxels’. VASO signal time courses for overlapping voxels in both normoxia and hypoxia were similar in shape (Fig. 3d). The VASO baseline signal was not statistically different between the oxygenation states (p N 0.05). Normalized task-related VASO signal change

in normoxia was − 5.8 ± 1.8% and − 6.2 ± 2.1% (p N 0.05) in hypoxia (Fig. 4d). The reactivity graph (Fig. 5d) illustrates the lack of effect on the VASO responses by decreasing Y. Overall, the overlapping voxel results for VASO data were similar to those found the in “All Voxels”. SNRs for BOLD and VASO scans were not significantly different (p N 0.05) between normoxia and hypoxia, therefore comparisons of signal changes can be made between the two oxygenation conditions without SNR bias. Discussion These results show that relative to normoxia, hypoxic hypoxia caused (i) a decrease in activation areas of both BOLD (T2⁎) and VASO signals, (ii) a decrease in the BOLD response size, (iii) a loss of the initial overshoot and a decrease in the size of the post-stimulus undershoot from BOLD response, and (iv) no effect on the VASO response characteristics. Thus, curtailed arterial O2 content influenced oxygenation changes associated with brain activation, while the vascular response, probed by VASO fMRI, remained unaffected. VASO fMRI as a CBV-weighted measure VASO fMRI provides an advantage over both PET and ASL techniques, both of which need relatively low spatial resolution for

Y.-C.L. Ho et al. / NeuroImage 41 (2008) 179–188

183

Fig. 4. Normalized signal values for BOLD (a, c) and VASO (b, d) fMRI in normoxia and hypoxia during baseline and activation. Upper panels: Results from “All Voxels”. Lower panels: Results from “Common Voxels”. The baseline signal BOLD decreased in hypoxia for both voxel groups (p ≤ 0.01). However for VASO, the baselines in hypoxia and normoxia were not significantly different. The BOLD task-related response was lower (p ≤ 0.05) in hypoxia than in normoxia for both types of voxels, while those for VASO did not differ from each other.

CBF imaging, a factor that hampers anatomical comparison of perfusion images with BOLD fMRI. Furthermore, ASL techniques in general require the subtraction of control and labeled scans, possibly violating the assumption of steady state during transients (Petersen et al., 2006). It must be noted, however, that the VASO signal, which is obtained by nulling of blood magnetization with a non-selective inversion pulse, requires extensive procedures and carefully chosen pulsing conditions for absolute quantification of CBV (Donahue et al., 2006). Firstly, the use of a blood nulling, inversion-recovery method depends on the efficiency of the non-selective inversion pulse and accurate knowledge of blood T1 (Donahue et al., 2006). Efficiency of inversion could be potentially reduced by the relatively high velocity of blood, particularly during neuronal activation. As for the blood T1, the value at 3 T is measured in vitro (Lu et al., 2004), and it varies as a function of hematocrit and blood oxygenation. According to Lu et al. (2004), the oxygenation effect is insignificant at 3 T, but may be a significant factor at higher fields. Nevertheless, this issue of blood oxygenation is explored further in the Appendix A, to estimate possible effects of blood T1 and T2 in hypoxia on the VASO signal change. Furthermore, inflow effects from non-inverted blood flowing into the slice in between excitations due to limited body coil length will result in an overall shorter apparent blood T1. This inflow effect may lead to a mismatch between theoretical and chosen bloodnulling points; at short TR (≤ 3 s at 3 T) inflow of fresh blood (uninverted spins) will artificially increase the amplitude of the VASO signal change. Compartment sizes for fresh, uninverted and steady-state blood (possibly capillary and venular) are unknown under the present experimental conditions and thus it is difficult

to calculate the specific contributions from either of these to the VASO signal. Besides CBF, there are CSF volume contributions (by 7–13%) that enhance the negative VASO signal changes, particularly at long TR and TE and at high spatial resolution. On the other hand, the CSF fraction appears to decrease during activation (by 5–6%) (Donahue et al., 2006), and is dependent on the local brain anatomy (Scouten and Constable, 2007), resulting in a reduced size of the VASO signal change. We assume little or no effect of CSF on the VASO signal at different oxygenation levels; with the TR, TE and spatial resolution used in our study, CSF and white matter effects are likely to be small, but the contribution from CBF and inflow may form significant contributions to the VASO signal as indicated by the study of Donahue et al. (2006). While these effects make it difficult to quantify the VASO data in absolute CBV, we can use the VASO signal changes as measures of relative changes in CBVand thus as an indicator of vascular reactivity. Under physiological conditions CBV and CBF are tightly coupled in a steady-state relationship expressed as CBFα = CBV (Grubb et al., 1974), where α ranges from 0.38 to 0.5 (Grubb et al., 1974; Leenders et al., 1990; van Zijl et al., 1998). Studies of hypoxia have shown either increase or no change in CBV [Unchanged: (Bandettini and Wong, 1997); Increased: (Bereczki et al., 1993; Julien-Dolbec et al., 2002)] and CBF [Unchanged: (Mintun et al., 2001; Tuunanen et al., 2006a; Tuunanen and Kauppinen, 2006); Increased: (Bereczki et al., 1993; Buck et al., 1998; Kety and Schmidt, 1948; Shapiro et al., 1970)] but, importantly, no study has yet shown decrease in both CBV and CBF to hypoxia. Thus we may expect concerted contributions of CBVand CBF to the VASO signals acquired in our experiments with hypoxic hypoxia.

184

Y.-C.L. Ho et al. / NeuroImage 41 (2008) 179–188

Fig. 5. Reactivity versus Y for normalized BOLD (squares) and VASO (triangles) signal changes. Each data point represents individual mean value normalized to the respective normoxic baseline. Linear trend lines were fitted to each dataset. “All Voxels”: (a) BOLD% signal changes decreased with decreasing Y (p ≤ 0.001). (b) No significant correlation was found between Y and normalized VASO% signal changes (p N 0.05). “Common Voxels”: (c) Normalized (p ≤ 0.001) BOLD% signal changes decreased significantly with decreasing Y. (d) Normalized VASO% signal changes (p N 0.05) were not significantly modulated by Y.

Hemodynamic baseline values and sensitivity of fMRI techniques Hypoxia induced by reduced FIO2 is likely to result in mild hypocapnia (Mintun et al., 2001; Sicard and Duong, 2005; Siesjo, 1978), which could inhibit vascular responses (Brian, 1998) and thus the relative magnitude of the hemodynamic signal change. Mintun et al. (2001) observed a decrease in PaCO2 by 2 to 3 mmHg in response to hypoxia induced with FIO2 12%. In humans (Poulin et al., 2002) and in spontaneously breathing rats (Duong et al., 2001; Sicard and Duong, 2005; Weiss et al., 1983), decreased baseline CBF was observed after hyperventilation (PaCO2 decreased by ~10–15%). Such baseline CBF reductions, particularly with a larger reaction from CBF than CBV, could contribute to BOLD response amplitude during hypocapnia (Cohen et al., 2002), because the BOLD response is positively dependent on CBF and negatively on CBV (Ogawa et al., 1990). Small hypocapnia superimposed with hypoxia might be a significant factor to vascular responsiveness. Indeed, Noth et al. (in press) have recently shown that isocapnic hypoxia at Y of 0.86 increases gray matter CBF by 5.6 ml/100 g/min (or by ~10%). For the present study, no probe was available to measure PaCO2 directly. Hypoxia results in shortening of T1 and T2⁎ in the blood. The latter effect could influence the sensitivity of the T2⁎-weighted scans to the BOLD effect (Bandettini and Wong, 1997) through the shortening of blood T2⁎ at rest by 21.3 ms (arterial) and 6.4 ms

(venous), and upon activation by 21.3 ms (arterial) and 9.5 ms (venous), as calculated for 3 T with Hct = 0.44 (Silvennoinen et al., 2003; Zhao et al., 2007) (Appendix A). However, we used two echoes acquired for the BOLD scans to calculate T2⁎ signals and as such avoided the possibility of decreased sensitivity. This could have been a problem at long, single TE acquisitions due to faster signal decay in hypoxia. On the other hand, the additional shortening of the longitudinal blood relaxation time is a pertinent issue for the VASO scans, especially since the VASO technique relies on the precise nulling of blood. In moderate hypoxia (Y ~ 0.8), the longitudinal relaxation of blood is shortened by 73 ms (arterial) and 64 ms (venous) at 3 T for Hct = 0.44 (Lu et al., 2004); thus it is likely that blood was no longer nulled as efficiently in hypoxia as in normoxia with the fixed TI used and it is therefore necessary to calculate the potential signal error for VASO changes in hypoxia. We determined that in hypoxia with Y at 0.8, the VASO signal change is underestimated by about 8 ± 6%. (see Appendix A). Given our mean Y = 0.86 ± 0.05 in the hypoxic hypoxia condition, this would yield conservative averaged and corrected signal changes − 6.6 ± 2.0% for “All Voxels” and −6.7 ± 2.3% for “Common Voxels” analyses) in hypoxia as compared to normoxia (−6.1 ± 1.3% for “All Voxels” and − 5.8 ± 1.8% for “Common Voxels” analyses). However, the corrected hypoxic VASO signals at all levels of the simulations were not statistically different from those determined in normoxia.

Y.-C.L. Ho et al. / NeuroImage 41 (2008) 179–188

BOLD response and vasoreactivity in hypoxia Consistent with recent studies (Rostrup et al., 2005; Tuunanen et al., 2006a,b), our results demonstrate a decline in the number of voxels showing BOLD and VASO responses during visual stimulation in hypoxia. Here, the thresholded areas in hypoxia were on average 55% and 65% smaller for the BOLD and VASO scans, respectively. This was not a result of decreased SNR, because the SNRs in hypoxia were not significantly different from those in normoxia. Neither of these observations resulted from decreased sensitivity of the given fMRI scans to shortened relaxation times, as discussed in the previous section. Baseline CMRO2 has been shown to be unchanged in mild hypoxic hypoxia (Kety and Schmidt, 1948; Shimojyo et al., 1968). Furthermore, amplitudes of short latency visual evoked potentials have been shown to be unaffected at this level of hypoxia (Sicard and Duong, 2005; Tuunanen et al., 2006a), suggesting that the neuronal workload in the visual cortex is independent of level of hypoxia. It is thus reasonable to believe that in the present study, the ‘increased workload’ was processed energetically (i.e. ATP consumption was the same) in the same manner in both normoxia and hypoxia. Tuunanen et al. (2006a) suggested that reduction of BOLD activation area during hypoxic stimulation in the visual cortex, when 1H MRS shows no lactate accumulation, indicate high absolute OEF, resulting from lesser mismatch between CMRO2 and CBF under this oxygenation condition. Decreased BOLD response amplitude in hypoxia agrees with a previous study (Rostrup et al., 2005), which measured BOLD changes in acute hypoxia in subjects adapted to both sea level and high altitudes. In the study by Rostrup et al. (2005), BOLD change decreased by N 50% at FIO2 of 8 or 10% in sea level adapted subjects in the visual stimulation paradigm, but it should be noted that the BOLD data were not corrected for the reduced baseline during hypoxia. In the current study, the decreased baseline for BOLD fMRI was evident in hypoxia both for “All Voxels” (38 ± 16%) and “Common Voxels” (45 ± 23%). The VASO results show no difference in vasoreactivity between normoxia and hypoxia. Bandettini and Wong (1997) assessed blood volume contributions to BOLD task activations during brief hypoxic hypoxia using simulations and empirical data; they did not find significantly increased CBV contributions to BOLD during hypoxia which suggests no additional vasodilation was present in hypoxia, in agreement with our results. According to the variants of O2 limitation model (Buxton and Frank, 1997; Hyder et al., 1998; Vafaee and Gjedde, 2000), any level of hypoxia should result in a large vascular response at low level of OEF to maintain steep O2 gradient from capillaries to mitochondria. During brain activation in hypoxia the vascular response should be even higher than in normoxia. However, we did not observe any difference in vasoreactivity, as assessed by VASO fMRI, between the oxygenation states. This is supported by a recent study (Tuunanen and Kauppinen, 2006) reporting unchanged CBF responses to a motor task (as measured by ASL) both in normoxia and hypoxia, as well as with another study (Tuunanen et al., 2006a) that observed CBF changes of the same magnitude to visual activation in normoxia and hypoxia. Furthermore, in a recent study using spontaneously breathing rats, Sicard et al. (2005) observed similar CBF responses to somatosensory stimulation in hypoxia and normoxia. It is intriguing that against all these CBF observations our BOLD results argue for strictly differing oxygenation responses during visual stimulation in normoxia and hypoxia. A recent paper by Tuunanen et al. (2006a) determined OEF in the parenchymal volumes showing VASO signal change, and found a

185

lower OEF in these parenchymal structures as a function of Y. In hypoxia BOLD response was observed in N 70% of VASO activated parenchymal volume and, taking these two observations together, it was concluded that O2 extraction proceeds at low level in the BOLD active brain parenchyma during hypoxia. Tuunanen et al. (2006a) analysed T2⁎-weighted signal intensity in the parenchyma showing BOLD response in normoxia but not in hypoxia, and found a consistent T2⁎-weighted signal increase that was some 20% of that determined in normoxia. Because ASL showed similar signal change, both in terms of volume and amplitude, in the visual cortex to stimulation during normoxia and hypoxia, this ‘residual’ BOLD signal in visual cortex showing large BOLD in normoxia was assumed to be due to only slightly decreased OEF during hypoxic stimulation. Thus the visual cortex appeared to show heterogeneity in OEF, i.e. regionally varying CMRO2 to CBF ratio, in response to stimulation during hypoxia. Rostrup et al. (2005) concluded in their study involving visual stimulations in hypoxia, that reduced BOLD signal in hypoxia may reflect lesser mismatch between CBF and CMRO2 responses. The current VASO data agree with the conclusions by Rostrup et al. (2005), in that hypoxia does not further augment vascular response to brain activation and so the BOLD observations are likely to be due to differing OEF under the oxygenation states. It is intriguing that in normoxia, the BOLD signal consistently showed an initial, sharp overshoot upon visual activation, while the overshoot was not evident during hypoxia (Fig. 3). Initial, sharp BOLD overshoots have been found in human fMRI studies using visual stimulation block paradigms (Chen et al., 1998). According to BOLD signal modeling by Buxton et al. (2004), an initial, sharp overshoot could be caused by a slowly rising CBV, while both CBF and CMRO2 increase sooner upon stimulation. Assuming this to be the case, a fast rise in CBV during hypoxic activations could eliminate the initial BOLD overshoot. However, in our study, the rising slope of the CBV-weighted VASO signal was practically the same between normoxia and hypoxia, and thus the modulation effect of CBV kinetics is an unlikely explanation for the observation. Similarly, the present data show that the BOLD post-stimulus undershoot is attenuated in overlapping voxels by hypoxia, yet VASO traces indicate similar kinetics for CBV recovery after cessation of visual stimulation. Other factors that could possibly influence these transient characteristics of the BOLD response include OEF and CMRO2 (see Eq. (1)). An unambiguous answer can be obtained by quantifying CMRO2 under these conditions. Nonetheless, these findings suggest once again the heterogeneous pattern of oxygenation changes during brain stimulation both temporally and spatially. To conclude, we have shown that the vascular response during hypoxic hypoxia to visual activation is similar to normoxia. The BOLD response shows reduced amplitude and no initial overshoot during stimulation in hypoxia. These results imply change in oxygen extraction as an adaptive response to curtailed blood O2 content. The results suggest that O2 availability may not be a critical signal for haemodynamic response induced by brain activation. Acknowledgments The authors would like to thank Esben Petersen for the helpful discussions on the study. This study was supported by a UK– Singapore Partners Science Collaboration award, together with the Singapore Millenium Foundation and the grant NMRC/0919/04 from the National Medical Research Council, Singapore.

186

Y.-C.L. Ho et al. / NeuroImage 41 (2008) 179–188

Appendix A. Calculating VASO signal error in hypoxia due to shortening of blood T1 and T2⁎ Using Eq. (3) from Lu et al. (2003), dealing with spatially nonselective inversion-recovery experiment at given TI with sliceselective excitation and gradient echo detection, MRI signal of parenchyma (Spar) from each voxel is written as: Spar ¼ Sblood þ Stissue
⁎ fðn  Cblood Þ  Mblood ðTI Þ  eTE=T2;blood þ Cpar  n  Cblood

ðA1Þ

TE=T⁎2;tissue

Mtissue ðTI Þ  e

where Sblood (or Stissue) is the MRI signal from the blood (or tissue), ξ is the total microvascular space occupancy, M is the magnetization contribution of the microvessels, C is the water density of microvascular blood (or tissue) [in ml water/ml blood (or tissue)] (Lu et al., 2002), T⁎2,blood and T⁎2,tissue are the effective transverse relaxation times of blood and tissue, respectively. Assuming imperfect blood nulling independently of oxygenation, the term Mblood(TI) ~ M(0) ∙ [1 − 2e− TI / T1 + e− TR / T1] ≠ 0 and it remains in Eq. (A1). As a result Mblood (TI) terms need be included in the equation describing fractional signal change for VASO, where: act rest  Spar DS Spar ¼ rest Spar S

ðA2Þ

in which ‘act’ and ‘rest’ denote the states of activation and rest (baseline), respectively. Both T1 and T2⁎ of blood are influenced by oxygen saturation (Lu et al., 2004; Silvennoinen et al., 2003; Zhao et al., 2007) and as expressed in the VASO signal Eq. (A1), both relaxation times inherently contribute to the VASO signal. From the paper by Lu et al. (2004), it can be calculated using the expressions given that at 3 T and at Hct = 0.44, arterial blood T1 is 1667 ms in normoxia (Ya = 0.98) and 1594 ms in hypoxia (Ya = 0.8), while venous blood T1 is 1517 ms in normoxia (Yv = 0.61) and 1453 ms in hypoxia (Yv = 0.45). During brain activation, oxygen saturation changes are substantial in veins and thus, the T2⁎ of venous blood has to be estimated (Ogawa et al., 1993), particularly with imperfect blood nulling. To calculate venous oxygen saturation (Yv), we used the expression (van Zijl et al., 1998): ð1  Yv Þ ¼ ð1  Ya Þ þ OEFYa

ðA3Þ

where OEF is the O2 extraction fraction. PET (Fox et al., 1988; Leenders et al., 1990) and MRI (Golay et al., 2001; Oja et al., 1999) studies have shown that baseline OEF ranges from 0.38 to 0.4 and it declines to 0.2–0.25 during brain activation in normoxia. Thus from Eq. (A3), using Ya = 0.98, OEF is 0.38 at baseline and 0.25 during activation, while Yv is 0.61 at baseline and 0.74 during activation. During hypoxia (Ya = 0.83), baseline OEF has been reported to be 0.43 (Shimojyo et al., 1968). Assuming a linear relationship between Ya and OEF, and similar decreases (by 35%) in OEF to brain activation, it can be estimated that for Ya of 0.8, baseline OEF is 0.44 and 0.28 during activation. Therefore, from Eq. (A3), Yv is 0.45 at baseline and 0.57 during activation in hypoxia. On one hand, if OEF proceeds at a very low level of 0.15 during hypoxic activation, then the computed Yv would be 0.68. On the other hand, if OEF proceeds at 0.40 during hypoxic activations, Yv would be 0.48. Indeed, Tuunanen et al. (2006a) have reported heterogeneous changes in OEF in the visual cortex during hypoxic stimulations. With these three possible

Table A1 T2⁎ of venous blood during baseline and activation states in hypoxia (Ya = 0.8, Hct = 0.44), at varying OEF and Yv values

Baseline Activation 1 Activation 2 Activation 3

OEF

Yv

Venous blood T2⁎ (ms)

0.44 0.15 0.29 0.40

0.45 0.68 0.57 0.48

13.3 23.7 17.8 14.3

Activation 1: OEF is assumed to decrease by 65% from baseline. Activation 2: OEF is assumed to decrease by 35% from baseline. Activation 3: Assuming a decline in OEF by 10% from baseline.

scenarios, we estimate corresponding values for T2⁎ of venous blood for baseline and activation in normoxia and hypoxia (Table A1), using the data published by Zhao et al. (2007) for a quadratic R2⁎ expression with a linear term. T2⁎ values of arterial blood were estimated using the same data (Zhao et al., 2007) yielding 55 ms in normoxia (Ya = 0.98) and 33 ms in hypoxia (Ya = 0.8). Since the T2⁎ of blood changes in hypoxia, the T2⁎ of tissue (gray matter), which can also be seen as the extravascular T2⁎ of blood, will also change correspondingly (Lu and van Zijl, 2005). Extravascular ΔR2⁎ is about two thirds of the total ΔR2⁎ of blood upon activation (Lu and van Zijl, 2005). Total R2⁎ values were taken from our BOLD scan data for both normoxia and hypoxia. Our total and estimated extravascular R2⁎ values for normoxia were very similar to the literature data (Lu and van Zijl, 2005). For the hypoxic condition, tissue T2⁎ was estimated to be 41 ms at rest and 42 ms during activation in hypoxia. Taking the measured VASO signal change in normoxia (from the “Common Voxels”) assuming perfect blood nulling, we rearranged the expanded Eq. (A2) to find the term ξ during activation. Holding ξ at this fixed value, and inserting T1 and T2⁎ values from hypoxia into Eq. (A2), one can calculate the expected VASO signal changes with imperfect blood nulling, for the given ξ. Blood was assumed to be 0.3 arterial and 0.7 venous. The VASO signal calculations showed that under hypoxic conditions when OEF decrease was assumed to be 35% during activation, error in VASO signal change was − 8%. Either of the extreme OEF changes yielded VASO signal errors of −14% for the large OEF decrease (by 65%), and −3% for the small OEF decrease (by 10%) during activation. References Bandettini, P.A., Wong, E.C., 1997. A hypercapnia-based normalization method for improved spatial localization of human brain activation with fMRI. NMR Biomed. 10, 197–203. Bereczki, D., Wei, L., Otsuka, T., Acuff, V., Pettigrew, K., Patlak, C., Fenstermacher, J., 1993. Hypoxia increases velocity of blood flow through parenchymal microvascular systems in rat brain. J. Cereb. Blood Flow Metab. 13, 475–486. Brian Jr., J.E., 1998. Carbon dioxide and the cerebral circulation. Anesthesiolog. 88, 1385–1386. Brindle, K.M., Brown, F.F., Campbell, I.D., Grathwohl, C., Kuchel, P.W., 1979. Application of spin-echo nuclear magnetic resonance to wholecell systems. Membrane transport. Biochem. J. 180, 37–44. Buck, A., Schirlo, C., Jasinksy, V., Weber, B., Burger, C., von Schulthess, G.K., Koller, E.A., Pavlicek, V., 1998. Changes of cerebral blood flow during short-term exposure to normobaric hypoxia. J. Cereb. Blood Flow Metab. 18, 906–910.

Y.-C.L. Ho et al. / NeuroImage 41 (2008) 179–188 Buxton, R.B., Frank, L.R., 1997. A model for the coupling between cerebral blood flow and oxygen metabolism during neural stimulation. J. Cereb. Blood Flow Metab. 17, 64–72. Buxton, R.B., Uludag, K., Dubowitz, D.J., Liu, T.T., 2004. Modeling the hemodynamic response to brain activation. Neuroimage 23 (Suppl 1), S220–S233. Chen, W., Zhu, X.H., Kato, T., Andersen, P., Ugurbil, K., 1998. Spatial and temporal differentiation of fMRI BOLD response in primary visual cortex of human brain during sustained visual simulation. Magn. Reson. Med. 39, 520–527. Cohen, E.R., Ugurbil, K., Kim, S.G., 2002. Effects of basal conditions on the magnitude and the dynamics of the blood oxygenation leveldependent fMRI response. J. Cereb. Blood Flow Metab. 22, 1042–1053. Corfield, D.R., Murphy, K., Josephs, O., Adams, L., Turner, R., 2001. Does hypercapnia-induced cerebral vasodilation modulate the hemodynamic response to neural activation? Neuroimage 13, 1207–1211. Davis, T.L., Kwong, K.K., Weisskoff, R.M., Rosen, B.R., 1998. Calibrated functional MRI: mapping the dynamics of oxidative metabolism. Proc. Natl. Acad. Sci. U. S. A. 95, 1834–1839. Donahue, M.J., Lu, H., Jones, C.K., Edden, R.A., Pekar, J.J., van Zijl, P.C., 2006. Theoretical and experimental investigation of the VASO contrast mechanism. Magn. Reson. Med. 56, 1261–1273. Duong, T.Q., Iadecola, C., Kim, S.G., 2001. Effect of hyperoxia, hypercapnia, and hypoxia on cerebral interstitial oxygen tension and cerebral blood flow. Magn. Reson. Med. 45, 61–70. Ekstrom-Jodal, B., Elfverson, J., von Essen, C., 1979. Cerebral blood flow, cerebrovascular resistance and cerebral metabolic rate of oxygen in severe arterial hypoxia in dogs. Acta Neurol. Scand. 60, 26–35. Emoto, S.E., Kintner, D., Feyzi, J.M., Gilboe, D.D., 1988. Relating cerebral ischemia and hypoxia to insult intensity. J. Neurochem. 50, 1952–1958. Fox, P.T., Raichle, M.E., 1986. Focal physiological uncoupling of cerebral blood flow and oxidative metabolism during somatosensory stimulation in human subjects. Proc. Natl. Acad. Sci. U. S. A. 83, 1140–1144. Fox, P.T., Raichle, M.E., Mintun, M.A., Dence, C., 1988. Nonoxidative glucose consumption during focal physiologic neural activity. Science 241, 462–464. Golay, X., Silvennoinen, M.J., Zhou, J., Clingman, C.S., Kauppinen, R.A., Pekar, J.J., van Zij, P.C., 2001. Measurement of tissue oxygen extraction ratios from venous blood T2: increased precision and validation of principle. Magn. Reson. Med. 46, 282–291. Grubb Jr., R.L., Raichle, M.E., Eichling, J.O., Ter Pogossian, M.M., 1974. The effects of changes in PaCO2 on cerebral blood volume, blood flow, and vascular mean transit time. Stroke 5, 630–639. Hoge, R.D., Atkinson, J., Gill, B., Crelier, G.R., Marrett, S., Pike, G.B., 1999. Linear coupling between cerebral blood flow and oxygen consumption in activated human cortex. Proc. Natl. Acad. Sci. U. S. A. 96, 9403–9408. Hyder, F., Shulman, R.G., Rothman, D.L., 1998. A model for the regulation of cerebral oxygen delivery. J. Appl. Physiol. 85, 554–564. Julien-Dolbec, C., Tropres, I., Montigon, O., Reutenauer, H., Ziegler, A., Decorps, M., Payen, J.F., 2002. Regional response of cerebral blood volume to graded hypoxic hypoxia in rat brain. Br. J. Anaesth. 89, 287–293. Kety, S.S., Schmidt, C.F., 1948. The effects of altered arterial tensions of carbon dioxide and oxygen on cerebral blood flow and cerebral oxygen consumption of normal young men. J. Clin. Invest. 27, 484–492. Kim, S.G., Rostrup, E., Larsson, H.B., Ogawa, S., Paulson, O.B., 1999. Determination of relative CMRO2 from CBF and BOLD changes: significant increase of oxygen consumption rate during visual stimulation. Magn. Reson. Med. 41, 1152–1161. Koehler, R.C., Traystman, R.J., Zeger, S., Rogers, M.C., Jones Jr., M.D., 1984. Comparison of cerebrovascular response to hypoxic and carbon monoxide hypoxia in newborn and adult sheep. J. Cereb. Blood Flow Metab. 4, 115–122. Leenders, K.L., Perani, D., Lammertsma, A.A., Heather, J.D., Buckingham, P., Healy, M.J., Gibbs, J.M., Wise, R.J., Hatazawa, J., Herold, S., 1990. Cerebral blood flow, blood volume and oxygen utilization. Normal values and effect of age. Brain 113 (Pt 1), 27–47.

187

Lu, H., van Zijl, P.C., 2005. Experimental measurement of extravascular parenchymal BOLD effects and tissue oxygen extraction fractions using multi-echo VASO fMRI at 1.5 and 3.0 T. Magn. Reson. Med. 53, 808–816. Lu, H., Golay, X., van Zijl, P.C., 2002. Intervoxel heterogeneity of eventrelated functional magnetic resonance imaging responses as a function of T1 weighting. Neuroimage 17, 943–955. Lu, H., Golay, X., Pekar, J.J., van Zijl, P.C., 2003. Functional magnetic resonance imaging based on changes in vascular space occupancy. Magn. Reson. Med. 50, 263–274. Lu, H., Clingman, C., Golay, X., van Zijl, P.C., 2004. Determining the longitudinal relaxation time T1 of blood at 3.0 Tesla. Magn. Reson. Med. 52, 679–682. Mandeville, J.B., Marota, J.J., Ayata, C., Moskowitz, M.A., Weisskoff, R.M., Rosen, B.R., 1999. MRI measurement of the temporal evolution of relative CMRO2 during rat forepaw stimulation. Magn. Reson. Med. 42, 944–951. Marrett, S., Gjedde, A., 1997. Changes of blood flow and oxygen consumption in visual cortex of living humans. Adv. Exp. Med. Biol. 413, 205–208. Meadows, G.E., O'Driscoll, D.M., Simonds, A.K., Morrell, M.J., Corfield, D.R., 2004. Cerebral blood flow response to isocapnic hypoxia during slow-wave sleep and wakefulness. J. Appl. Physiol. 97, 1343–1348. Mintun, M.A., Lundstrom, B.N., Snyder, A.Z., Vlassenko, A.G., Shulman, G.L., Raichle, M.E., 2001. Blood flow and oxygen delivery to human brain during functional activity: theoretical modeling and experimental data. Proc. Natl. Acad. Sci. U. S. A. 98, 6859–6864. Nöth, U., Kotajima, F., Deichmann, R., Turner, R., Corfield, D.R., in press. Mapping of the cerebral vascular response to hypoxia and hypercapnia using quantitative perfusion MRI at 3T. NMR Biomed. Ogawa, S., Lee, T.M., Kay, A.R., Tank, D.W., 1990. Brain magnetic resonance imaging with contrast dependent on blood oxygenation. Proc. Natl. Acad. Sci. U. S. A. 87, 9868–9872. Ogawa, S., Menon, R.S., Tank, D.W., Kim, S.G., Merkle, H., Ellermann, J.M., Ugurbil, K., 1993. Functional brain mapping by blood oxygenation level-dependent contrast magnetic resonance imaging. A comparison of signal characteristics with a biophysical model. Biophys. J. 64, 803–812. Oja, J.M., Gillen, J.S., Kauppinen, R.A., Kraut, M., van Zijl, P.C., 1999. Determination of oxygen extraction ratios by magnetic resonance imaging. J. Cereb. Blood Flow Metab. 19, 1289–1295. Petersen, E.T., Lim, T., Golay, X., 2006. Model-free arterial spin labeling quantification approach for perfusion MRI. Magn. Reson. Med. 55, 219–232. Poulin, M.J., Fatemian, M., Tansley, J.G., O'Connor, D.F., Robbins, P.A., 2002. Changes in cerebral blood flow during and after 48 h of both isocapnic and poikilocapnic hypoxia in humans. Exp. Physiol 87, 633–642. Rostrup, E., Larsson, H.B., Born, A.P., Knudsen, G.M., Paulson, O.B., 2005. Changes in BOLD and ADC weighted imaging in acute hypoxia during sea-level and altitude adapted states. Neuroimage 28, 947–955. Scouten, A., Constable, R.T., 2007. Applications and limitations of whole-brain MAGIC VASO functional imaging. Magn. Reson. Med. 58, 306–315. Shapiro, W., Wasserman, A.J., Baker, J.P., Patterson Jr., J.L., 1970. Cerebrovascular response to acute hypocapnic and eucapnic hypoxia in normal man. J. Clin. Invest 49, 2362–2368. Shimojyo, S., Scheinberg, P., Kogure, K., Reinmuth, O.M., 1968. The effects of graded hypoxia upon transient cerebral blood flow and oxygen consumption. Neurology 18, 127–133. Sicard, K.M., Duong, T.Q., 2005. Effects of hypoxia, hyperoxia, and hypercapnia on baseline and stimulus-evoked BOLD, CBF, and CMRO2 in spontaneously breathing animals. Neuroimage 25, 850–858. Siesjo, B.K., 1978. Brain energy metabolism and catecholaminergic activity in hypoxia, hypercapnia and ischemia. J. Neural. Transm., Suppl 17–22. Silvennoinen, M.J., Clingman, C.S., Golay, X., Kauppinen, R.A., van Zijl, P.C., 2003. Comparison of the dependence of blood R2 and R2⁎ on oxygen saturation at 1.5 and 4.7 Tesla. Magn. Reson. Med. 49, 47–60.

188

Y.-C.L. Ho et al. / NeuroImage 41 (2008) 179–188

Thulborn, K.R., Waterton, J.C., Matthews, P.M., Radda, G.K., 1982. Oxygenation dependence of the transverse relaxation time of water protons in whole blood at high field. Biochim. Biophys. Acta 714, 265–270. Tuunanen, P.I., Kauppinen, R.A., 2006. Effects of oxygen saturation on BOLD and arterial spin labelling perfusion fMRI signals studied in a motor activation task. Neuroimage 30, 102–109. Tuunanen, P.I., Murray, I.J., Parry, N.R., Kauppinen, R.A., 2006a. Heterogeneous oxygen extraction in the visual cortex during activation in mild hypoxic hypoxia revealed by quantitative functional magnetic resonance imaging. J. Cereb. Blood Flow Metab. 26, 263–273. Tuunanen, P.I., Vidyasagar, R., Kauppinen, R.A., 2006b. Effects of mild hypoxic hypoxia on poststimulus undershoot of blood-oxygenation-level-dependent fMRI signal in the human visual cortex. Magn. Reson. Imaging 24, 993–999.

Vafaee, M.S., Gjedde, A., 2000. Model of blood-brain transfer of oxygen explains nonlinear flow-metabolism coupling during stimulation of visual cortex. J. Cereb. Blood Flow Metab. 20, 747–754. van Zijl, P.C., Eleff, S.M., Ulatowski, J.A., Oja, J.M., Ulug, A.M., Traystman, R.J., Kauppinen, R.A., 1998. Quantitative assessment of blood flow, blood volume and blood oxygenation effects in functional magnetic resonance imaging. Nat. Med. 4, 159–167. Weiss, H.R., Buchweitz, E., Sinha, A.K., 1983. Effect of hypoxic-hypocapnia on cerebral regional oxygen consumption and supply. Microvasc. Res. 25, 194–204. Zhao, J.M., Clingman, C.S., Narvainen, M.J., Kauppinen, R.A., van Zijl, P.C., 2007. Oxygenation and hematocrit dependence of transverse relaxation rates of blood at 3 T. Magn. Reson. Med. 58, 592–597.