Distinct effects of isoflurane on basal BOLD signals in tissue/vascular microstructures in rats

Isoflurane is a well-known volatile anesthetic. However, it remains equivocal whether its effects on BOLD signal differ depending on the types of intracranial structures, such as capillaries and large blood vessels. We compared dose-dependent effect of isoflurane on the basal BOLD signals in distinct cerebral structures (tissue structure or large vessels) using high resolution T2*-images at 9.4 T MRI system in rat somatosensory cortex. The local field potential (LFP) in the somatosensory cortex and mean arterial pressure (MAP) were also investigated. Isoflurane induced inverted U-shaped dose-dependent change in BOLD signal in large vessels and tissue regions: BOLD signal under 2.0% and 2.5% isoflurane significantly increased from the maintenance dose (1.5%) and that under 3.0% was similar to maintenance dose. Remarkably, BOLD signal increase in tissue regions under 2.5% was significantly smaller than that in large vessels. The MAP decreased monotonically due to the dose of isoflurane and the LFP was strongly suppressed under high dose (2.5% and 3.0%). These results indicate that isoflurane-induced alteration of MAP and neuronal activity affected BOLD signal and, especially, BOLD signal in the tissue regions was more affected by the neuronal activity.


Results
Dose dependent effect of isoflurane on BOLD signal was vascular structure-dependent. BOLD images were acquired in the same rats under 1.5% (maintenance concentration), 2.0%, 2.5% and 3.0% (supra-anesthetic doses) isoflurane (Fig. 1a). Then, regions of interests (ROIs) of tissue (ROI tissue ) and large vessels (ROI vessel ) were automatically discriminated from ROI of the somatosensory cortex (ROI whole ) by means of Otsu's analysis method (Fig. 1b-e). BOLD signals in the somatosensory cortex (BOLD whole ), in the tissue (BOLD tissue ) and in the vessels (BOLD vessel ) were compared.

Discussion
The results presented here with a use of non-invasive high-resolution imaging with UHF MRI successfully demonstrate that basal BOLD signal in tissue structure was less influenced by isoflurane than large vessels. Further, the BOLD signal changes differed from MAP changes because BOLD signal changes are not monotonic (Fig. 2). This discrepancy between BOLD signals and MAP could be explained by stronger suppression of neuronal activity under 2.5% and 3.0% isoflurane than 1.5% and 2.0% isoflurane. The basal BOLD signal is a key factor for "BOLD response" to physiological stimulation (e.g., somatosensory stimulation and visual stimulation) and "resting state functional connectivity". If basal BOLD signals are altered by anesthetics or vasoactive drugs, such as vasodilator or vasoconstrictor, observed BOLD response would be also altered compared to normal state 11 . Therefore, it is important to investigate the effect of vasoactive drugs on basal BOLD signals.
One of the important novel findings of this study is that the effects of isoflurane on the basal level of BOLD signals depend on the type of blood vessels in the somatosensory cortex. Such vascular-structure dependent BOLD responses are in agreement with the previous UHF MRI study in the Whisker-Barrel cortex of medetomidine-sedated rats 12 . Also, region-dependent differences in the BOLD responses have been demonstrated using UHF MRI in the auditory cortex of cats 13 and visual cortex of human subjects 14 . These studies, including ours, took advantage of UHF MRI that can provide improved spatial resolution and T2*-contrast through decreased signal to noise ratio (SNR) and shorter T2*-values. Therefore, on the basis of the present results, it is proposed that 1) high resolution fMRI analysis is the key for distinguishing BOLD signal components in the vascular structure and 2) anesthetics frequently used in fMRI studies would potentially affect the BOLD signals in a type-, dose-, and the target structures-dependent manners.
In the present study, because of the thickness of the images (400 μ m), there would be the partial volume effect: a pixel contains both tissue and large vessels. This might have resulted in a blurred boundary between vessels and tissues. We used Otsu's thresholding method, which was widely used and was robust to this problem 15,16 , and we could successfully classify the large vessels and tissues. However, for further accurate analysis, ultra-higher resolution T2*-images that are powerful enough to observe only large vessel or tissue region in a pixel should be performed in the future.
Previous studies have revealed that isoflurane dose-dependently alters the rCBF, rCBV and CMRO 2 in the brain. Higher dose (1-2 minimum alveolar concentration (MAC)) of isoflurane not only increases the rCBF and rCBV 17,18 but also decreases the CMRO 2 in the cerebral cortex 19 dose-dependently. The MAP is also decreased by isoflurane dose-dependently, but the partial pressure of oxygen (pO 2 ), partial pressure of carbon dioxide (pCO 2 ) and pH are constant 17,20 , consistent with our result. Therefore, from the view of vasoactive effects of isoflurane with biophysical BOLD model 21 , basal BOLD signals seem to increase monotonically due to isoflurane dose; however, highest signal increase was observed under 2.0% and then it decreased from the peak under 2.5% and 3.0% (Fig. 2). It is difficult to explain this phenomenon only with "vasoactive" effect of isoflurane.
Another factor to regulate the BOLD signals is the interaction among neurons, astrocytes and blood vessels, called as neurovascular coupling 22 . BOLD signal is positively correlated with the neuronal activity through the neurovascular coupling. The isoflurane has potent effects on inhibitory and excitatory synaptic transmission [23][24][25] and postsynaptic excitability [26][27][28] , which comprise the core effects underlying their "anesthetic" pharmacology. The isoflurane is also known to suppress astrocyte activity even with maintenance dose. The 1.2% and 1.5% isoflurane suppresses the astrocyte activity, but this range of isoflurane does not suppress the neuronal activity 29 . Our results also show that synaptic activity was significantly suppressed from 2.5% (Fig. 3). Together, the BOLD signal increase under 1.5% and 2.0% could be mainly due to the vasodilation effects of isoflurane with remaining neuronal activity, but BOLD signal decrease under 2.5% and 3.0% compared to 2.0% could be due to the mixture of vasodilation and strong suppression of neuronal activity.
Remarkably, BOLD signal changes in tissue regions under 2.5% were smaller than those in large vessels. Together with above discussion, this indicates that BOLD signals in capillaries are more affected by neuronal activity than large vessels. Importantly, CMRO 2 , rCBF and rCBV changes coupled to neuronal activity are different in the vessel type. For instance, when neuronal activity increases, the artery increases the rCBV more than capillary, but oxygen saturation of hemoglobin extremely increases in the capillary 30 . Furthermore, neurovascular structure is different among the large vessels and capillary. Capillary receives more direct regulation from neurons and astrocytes than artery and the vein 31 .

Conclusion.
In conclusion, high resolution imaging in UHF MRI is a promising tool for investigating the functional responses by distinguishing distinct microstructures of rat brain. Therefore, UHF MRI would play an important role in future preclinical animal study. Because anesthesia is essential for animal experiments to suppress motion artifacts and to reduce stress during the painful, unpleasant stimulation, it is important to clarify the effect of anesthetics on the BOLD signals differentially in the neuronal and vascular structures, such as the arteries, veins and capillaries, which is at this moment only possible with UHF MRI.

Methods
Animal preparation. The  An intubation was made for mechanical ventilation to ensure normoxic breathing during the scanning under 1.5% isoflurane (in air) followed by an initial induction of anesthesia with 3% isoflurane lasting within 1 min. The intratracheal cannula was connected to a mechanical ventilator (SAR-830 Ventilator, CWE Inc., CA) and ventilation was made with following parameters throughout the experiments: respiratory rate = 50/min; inspiratory time: expiratory time = 1; tidal volume = 1.7 ml. The pO 2 , pCO 2 and pH have been confirmed to be kept in normoxic and normocapnic ranges throughout the experiment, and there was no significant difference under 1.5-3.0% isoflurane 9 .
The body temperature was maintained at 36.5-37 °C using an MR-compatible circulating water heating system (CW-05G, JEIO TECH CO., LTD., Seoul, Korea). The ventilation-controlled the waveform and rate of the respiration (50/min) and body temperature were continuously monitored throughout the experiments using an MR-compatible monitoring system (Model 1025, SA Instruments, Stony Brook, NY). FMRI procedure. The MRI experiments were performed using a 9.4 T horizontal bore MRI scanner (BioSpec 94/20 USR, Bruker, Ettlingen, Germany) equipped with a BGA12S gradient system. A volume coil (Bruker, Ettlingen, Germany) was used for transmitting the signal and a 4-channel array coil (Bruker, Ettlingen, Germany) was used for reception of the signal. Following scout scans and magnetic field homogeneity optimization (MAPSHIM), we obtained gradient-echo BOLD images using a T2*-weighted echo planer imaging sequence with the following parameters: time of repetition = 1,500 ms, echo time = 25 ms, segmentation = 2, field of view = 21 mm × 21 mm, acquisition matrix = 210 × 210, slice thickness = 0.4 mm, (spatial resolution = 100 × 100 × 400 μ m), slice gap = 1.0 mm, slice number = 3 and number of average = 20. The positions of the slices were determined using sagittal imaging and were between − 3.0 and + 1.0 mm from the bregma.
Drug effect evaluation of isoflurane. The animals were first anesthetized with 1.5% isoflurane for the fMRI setup and field homogeneity correction. The rats were kept anesthetized with 1.5% isoflurane (defined as "maintenance concentration" which was almost equivalent to its MAC, 1.4% 32 ). BOLD images were acquired in the same rat under 1.5% (maintenance concentration), 2.0%, 2.5% and 3.0% (supra-anesthetic doses) isoflurane. Scanning was started 5 min after changing anesthetic concentration. The sequence of isoflurane dose was set randomly (C1-C3 in Fig. 1a). The time-course of experiment procedure is shown in Fig. 1a. LFP measurement. The electrophysiological recording was performed separately from MRI bore as previously described 9 . The animals, first anesthetized with 1.5% isoflurane, were placed in a stereotaxic frame (David Kopf instrument, CA), and a hole centered at 4.0 mm lateral, 0.8 mm anterior from the Bregma was drilled on the left or right side of the skull. The electrode tip (< 1.0 MΩ, a 1 μ m tip and 0.127-mm shaft diameter, Alpha Omega Engineering, Nazareth, Israel) was positioned at a depth of 1.8-2.3 mm from the cortical surface. LFP signals were acquired at 1 kHz sampling rate using dedicated data acquisition software (Power Lab, AD Instruments, Dunedin, New Zealand). The reference electrode was positioned on the scalp.
The experimental protocol was same as fMRI (Fig. 1). After 5 min from the change of dose of isoflurane, the LFP was recorded for 2 min. Then, averaged total power of LFP for 2 min was calculated. We confirmed that LFP signal stabilized after 5 min from the change of isoflurane dose.
MAP measurement and blood component. The measurement of MAP was performed separately from fMRI experiment. The catheters were surgically placed into tail artery and MAP was measured by means of MR-compatible monitoring system (Model 1025, SA Instruments, Stony Brook, NY). The experimental protocol was same as fMRI (Fig. 1). After 5 min from change of dose of isoflurane, the MAP during 2 min was averaged. We confirmed that MAP stabilized after 5 min from the change of isoflurane dose.

Data analysis. Changes in mean BOLD signals within identified ROI.
To compare the changes by anesthetic drugs between different structures, the normalized intensity from defined ROIs were calculated as follows: First, ROI in the somatosensory cortex (ROI whole ) was defined based on the high resolution images and rat brain atlas 33 (Fig. 1b and c). Same ROI whole was used in the same rat under all doses of isoflurane. Then, threshold to separate the ROI of the large blood vessels (ROI vessel ) and ROI of tissue (ROI tissue ) was automatically identified according to discriminant analysis method (Otsu's method) 34 . The typical histogram in the ROI whole has two classes: the class 1 at lower intensity corresponds to large blood vessels and the classes 2 at higher intensity corresponds to tissue regions. The Otsu's method determines the optimum threshold for dividing the histogram into two classes (class 1 and class 2) to maximize the ratio of variance of inter-class and intra-class (F). where σ b 2 and σ w 2 correspond to the variance of intra-class and inter-class respectively. The ω 1 and ω 2 are the number of voxels for class 1 and class 2 respectively. The m 1 , m 2 and m T are the averaged signal intensity for class 1, class 2 and for all voxels within the ROI whole respectively. The class 1 was classified as ROI vessel and the class 2 was classified as ROI tissue (Fig. 1d and e). The ROI vessel and ROI tissue were calculated at each dose of isoflurane because the contrast of vessels might have possibly changed by vasodilation effect of isoflurane. The mean BOLD signals for each of these structures were defined as BOLD vessel and BOLD tissue . The BOLD vessel and BOLD tissue intensity were normalized so that those under maintenance dose were 2000. The volumes of the large blood vessels and tissue regions were calculated by counting the number of voxels of ROI tissue and ROI vessel . The numbers of the pixels of ROI tissue and ROI vessel were normalized so that those under maintenance dose were 100. Statistical analysis. The significance of BOLD signal intensity, MAP and LFP under supra-concentration of isoflurane (2.0-3.0%) compared with maintenance dose (1.5%) was assessed via paired t-test. The significance of BOLD signal intensity between ROI vessel and ROI tissue was assessed via t-test following repeated measures ANOVA.