Dynamics of isoflurane-induced vasodilation and blood flow of cerebral vasculature revealed by multi-exposure speckle imaging

Background Anesthetized animal models are used extensively during neurophysiological and behavioral studies despite systemic effects from anesthesia that undermine both accurate interpretation and translation to awake human physiology. The majority of work examining the impact of anesthesia on cerebral blood flow (CBF) has been restricted to before and after measurements with limited spatial resolution. New Method We used multi-exposure speckle imaging (MESI), an advanced form of laser speckle contrast imaging (LSCI), to characterize the dynamics of isoflurane anesthesia induction on cerebral vasculature and blood flow in the mouse brain. Results The large anatomical changes caused by isoflurane are depicted with wide-field imagery and video highlighting the induction of general anesthesia. Within minutes of exposure, both vessel diameter and blood flow increased drastically compared to the awake state and remained elevated for the duration of imaging. An examination of the dynamics of anesthesia induction reveals that blood flow increased faster in arteries than in veins or parenchyma regions. Comparison with Existing Methods MESI offers robust hemodynamic measurements across large fields-of-view and high temporal resolutions sufficient for continuous visualization of cerebrovascular events featuring major changes in blood flow. Conclusion The large alterations caused by isoflurane anesthesia to the cortical vasculature and CBF are readily characterized using MESI. These changes are unrepresentative of normal physiology and provide further evidence that neuroscience experiments would benefit from transitioning to un-anesthetized awake animal models.


Introduction
The use of general anesthesia during neuroimaging is ubiquitous across many animal models despite systemic effects on neurophysiological state and cardiovascular function [1]. Volatile inhalation anesthetics, such as the halogenated ether isoflurane [2], are commonly utilized to immobilize animals during imaging while allowing for fine control over the depth of anesthesia and consciousness. However, isoflurane has been shown to reduce neuronal activity [3] and functional connectivity [4], suppress the magnitude and speed of neurovascular coupling [5,6,7], and induce significant vasodilation [8,9]. Isoflurane also conveys potential neuroprotective effects that reduce and delay the severity of cerebral ischemia [10,11,12,13,14]. These effects can mask the benefits of prospective neuroprotective therapeutics or interventions and confound the outcomes of long-term studies [15,16]. For these reasons and more [17], there is a growing effort within the neuroscience community to transition to un-anesthetized awake animal models in order to more accurately interpret neurophysiological and behavioral experiments and more readily translate findings to awake human physiology.
There have been numerous imaging modalities used to examine the effects of isoflurane on cerebral hemodynamics. Functional magnetic resonance imaging (fMRI), laser Doppler flowmetry, and intrinsic signal optical imaging have all established that isoflurane increases basal cerebral blood flow (CBF) while attenuating and delaying the hemodynamic response to local neural activity via neurovascular coupling [18,3,6,7]. Two-photon phosphorescence lifetime microscopy showed that tissue oxygenation is twice as high under isoflurane anesthesia compared to the awake state and exhibited large layer-specific differences within the cortical vasculature [19]. Calcium signal imaging following ischemic stroke found that the magnitude of spreading depolarizations were significantly smaller in awake mice [20] while laser speckle contrast imaging (LSCI) of photothrombotic stroke observed larger infarct sizes in awake rats compared to their isoflurane-anesthetized counterparts [14]. To our knowledge, no prior studies have directly imaged the induction of general anesthesia and its acute effects upon the cerebral vasculature.
In this paper, we present multi-exposure speckle imaging (MESI) of cerebral blood flow in awake mice during the induction of general anesthesia with isoflurane. The MESI technique [21] provides a more robust estimate of large changes in flow compared to traditional single-exposure LSCI [22] and enables the chronic tracking of CBF [23]. We examine the dynamics of the large anatomical and physiological changes to cortical vasculature in response to isoflurane with wide-field imagery across multiple imaging sessions and animals. Because the anesthetized state is representative of an abnormal physiology, we argue that awake animal models should be utilized whenever possible for neuroscience studies.

Multi-Exposure Speckle Imaging (MESI)
A schematic of the imaging system is presented in Fig. 1a. MESI was performed using a 685 nm laser diode (50 mW, HL6750MG, Thorlabs, Inc.) intensity modulated with an acousto-optic modulator (AOM, 3100-125, Gooch & Housego, Ltd.) and relayed to illuminate the craniotomy at an oblique angle. The scattered light was imaged by a CMOS camera (acA1920-155um, Basler AG) with 2× magnification and cropped to a field-of-view of 3.6 × 3.0 mm. Camera exposures were temporally synchronized with the modulated laser pulses. Fifteen camera exposures ranging between 50 µs and 80 ms [21,24,23] were recorded for each complete MESI frame in order to broadly sample the speckle dynamics of the specimen [25], resulting in an effective acquisition rate of ∼2.5 MESI frames-per-second. The total amount of light used to capture each exposure was held constant with the AOM in order to minimize the effects of shot noise [21]. The acquisition was controlled using custom software written in C++ along with a multifunction I/O device (USB-6363, National Instruments Corp.) for the generation of camera exposure trigger signals and AOM modulation voltages [26].
The 15 raw intensity images were converted to speckle contrast images (K = σ s /⟨I⟩) with a 7×7-pixel sliding window and used to calculate an estimate of the speckle correlation time (τ c ) at each pixel with the multi-exposure speckle visibility expression [21]: where T is the camera exposure time, x = T /τ c , ρ is the fraction of light that is dynamically scattered, β is a normalization factor that accounts for speckle sampling, and ν represents exposure-independent instrument noise and nonergodic variances. Eq. (1) was fitted with the Levenberg-Marquardt nonlinear least squares algorithm [27] using a custom program written in C.
Because τ c is inversely related to the speed of the moving scatterers [28,29], the inverse correlation time (ICT = 1/τ c ) is frequently used as a metric for quantifying blood flow in vasculature and perfusion in parenchyma [30,31,26].
Recent work to improve the quantitative accuracy of MESI flow measurements in vasculature takes into account the presence of multiple dynamic scattering events instead of assuming only a single dynamic scattering event per photon [32]. This is achieved by scaling the fitted ICT value by the diameter of the vessel in order to obtain an estimate of "vascular flux" that better accounts for variations in vascular volume sampling [32,33]. This paper uses the vascular flux metric (arbitrary units) for all vessel measurements and ICT (s −1 ) for all parenchyma perfusion measurements. The subjects were allowed to recover from surgery and monitored for cranial window integrity and normal behavior for at least four weeks prior to imaging.

Animal Preparation
They were then exposed to head fixation during 20-30 minute sessions over 3-5 days until habituated to locomotion on a linear treadmill awake imaging system [34].

Awake-to-Anesthetized Blood Flow Measurements
Cranial window implanted mice were head constrained on the awake imaging system and allowed to acclimate for 5-10 minutes. The awake hemodynamic baseline was then defined prior to continuous imaging by acquiring 50 MESI frames, averaging the speckle contrast across each exposure time, and computing the corresponding ICT frame using Eq. (1). A custom 3D-printed inhalation nose-cone connected to the anesthesia vaporizer and scavenger was then placed several millimeters away from the subject and used to administer medical air (0.5 L/min) with no isoflurane (0%). Continuous MESI was initiated (t = 0) and used to monitor the awake subject for 10 minutes before increasing isoflurane to 2.0% to induce anesthesia. After one minute of exposure (t = 11 minutes), the nose-cone was fully positioned over the subject to ensure targeted delivery of isoflurane. After two minutes of exposure (t = 12 minutes), a feedback heating pad (55-7030, Harvard Apparatus, Inc.) was placed beneath the subject to maintain a 37 • C body temperature. The now anesthetized subject was continuously monitored with MESI until t = 30 minutes, at which point the anesthetized hemodynamic state was defined by acquiring and averaging an additional 50 MESI frames. The subject was then removed from the imaging system and allowed to recover from anesthesia on a heating pad before being transferred back into its cage. Each subject underwent three imaging sessions with at least one full day of recovery between each session.
One animal (Subject 1) underwent an extended routine where the awake and anesthetized states were each imaged for one hour in order to capture the longer-term hemodynamics. Because of the volume of data being recorded, the acquisition was briefly paused prior to the induction of anesthesia to offload data from the solid-state drive used for writing data, which resulted in a short period of missing imagery. The timing of the anesthesia induction was identical to the protocol described above with the awake and anesthetized states defined at t = 0 and t = 120 minutes, respectively. The four repeated imaging sessions for this subject were spaced at least one week apart to minimize any complications from extended anesthesia exposure.

Data Processing
All data processing was performed using MATLAB (R2021a, MathWorks, Inc.). While head fixation can greatly minimize locomotion-associated brain movement [35], it does not completely eliminate motion artifacts. The induction of general anesthesia also causes major changes in posture that can result in lateral displacements of the brain relative to the skull. In order to account for these shifts, the open source image registration toolbox Elastix [36] was used to rigidly align all speckle contrast frames to the baseline awake imagery prior to fitting Eq. (1). The resulting aligned ICT frames were then smoothed temporally with a central moving average filter (k = 5) and rendered to video at 20 frames-per-second (8× speed). ICT timecourses spanning the entire imaging session were then extracted from the aligned data from both vascular and parenchymal regions of interest (ROIs).

Calculating Vascular Flux
In order to calculate the vascular flux [32], the diameter of vessels (d vessel ) were estimated using the 5 ms exposure speckle contrast frames. Cross-sectional profiles were extracted across each vessel of interest at every timepoint and fit-

Calculating Blood Flow Rise Time
To examine the dynamics of isoflurane-induced flow changes, the rise time in imaging sessions using the previously extracted ROI timecourses.

Statistical Analysis
All statistical analyses were also performed using MATLAB. The differences between arterial, venous, and parenchymal measurements were evaluated with one-way ANOVAs using the Bonferroni method for multiple comparisons. p ≤ 0.05 was considered statistically significant.

Results
The systematic changes caused by general anesthesia with isoflurane in Subject 1 are shown in Fig. 1b

Discussion
The systemic effects of isoflurane undermine the reliability of neurophysiological and behavioral studies because anesthesia is a fundamentally unnatural physiological state [17]. Imaging blood flow on the cortical surface with MESI during the induction of anesthesia directly visualizes and quantifies these large hemodynamic changes. On average, the vasodilation caused by isoflurane increased vessel diameter by 14.1% across all trials, with arteries increasing by 15.0% and veins increasing by 13.0% (Fig. 3a). Direct microscope measurements in fentanyl-and nitrous oxide-anesthetized rats documented a 17% increase in arteriolar diameter and 6% increase in venule diameter after the topical application of isoflurane [8]. Similar measurements in pentobarbital-anesthetized dogs observed 10-28% increases in the diameter of small arterioles following the inhalation of isoflurane [9]. Both studies found that isoflurane dilated arterioles in a concentration-dependent manner. More recent measurements in mice using optical coherence tomography (OCT) angiography found that isoflurane increased artery diameter by 12-55% and vein diameter by 14-22%, depending on vessel branch order [37].
Much larger changes were measured in surface vessel blood flow, with vascular flux increasing on average by 96.0% across all trials (Fig. 3b). This suggests that increases in blood flow velocity rather than vasodilation are largely responsible for the measured change in flow. Because few imaging modalities are capable of measuring the dynamics of blood flow with sufficient spatial resolution to distinguish individual vessels, there has been limited work on this topic in awake animals. One study using laser Doppler flowmetry found only non-significant 18% increases in both CBF and red blood cell (RBC) velocity in the barrel cortex after mice were anesthetized with isoflurane [6]. However, because large blood vessels were avoided, the laser Doppler measurements only sampled smaller subsurface vasculature that may exhibit different responses to isoflurane than the larger surface vasculature imaged in this study. The vascular flux metric calculated from the speckle correlation time is also not analogous to either the CBF or RBC velocity measured by laser Doppler flowmetry, so the results are not directly comparable [32]. Another more recent study using Doppler OCT measured a 55% increase in volumetric blood flow in mice under isoflurane [37]. Unlike the results presented above, this change was largely driven by vasodilation rather than an increase in blood flow velocity.
Within the parenchyma, tissue perfusion increased on average by 84.7% across all trials after isoflurane exposure (Fig. 3b). The narrower distribution of values compared to the vascular measurements is likely a byproduct of the more homogenous structure of the unresolvable capillaries within the parenchyma. fMRI measurements in isoflurane-anesthetized rats have reported ∼50% increases in global CBF with the cerebral cortex experiencing the smallest regional increase (20%) compared to the awake state [38]. However, these numbers are not directly comparable because fMRI samples a much larger volume of the brain than MESI, which has both a smaller field-of-view and only penetrates several hundred microns into the cortex.
While LSCI has been utilized extensively with awake imaging [39,16,14,20,40], this is one of the first uses of MESI in an awake animal model [34]. The effects of isoflurane fundamentally undermine the imaging of cortical hemodynamics in the normal physiological state. Blood flow imaging techniques such as LSCI, laser Doppler flowmetry, and fMRI require greater sensitivity because of the suppression of the hemodynamic response [6] and measure a delayed neurovascular coupling [7]. Anatomical imaging with two-photon microscopy only captures heavily-dilated vasculature unrepresentative of the awake resting state [19]. The consistency and reliability of all imaging techniques can suffer from the effects of prolonged isoflurane exposure, which can cause continual vasodilation as seen in Fig. 2b. This could cause extended imaging sessions to document disparate anatomies and physiologies between the beginning and conclusion of an experiment.

Limitations
The impact of repeated exposure to isoflurane was not examined in this study. Previous work has found that it can impair synaptic plasticity in the basolateral amygdala [41] and cause persistent motor deficits via structural changes to the corpus callosum [42]. While image registration can help maintain the spatial alignment of data across an experiment, it is unable to decouple the effects of animal motion from the underlying blood flow. Even with the head fully restrained, walking and grooming both caused subtle brain movements that manifested as large changes in speckle contrast, as seen by the abrupt spikes during the awake section of Fig. 2c. While these fluctuations may represent real hemodynamic responses, it is difficult to isolate them from broader animal motion. A simple solution might be the addition of an external sensor to the treadmill to exclude timepoints when the animal is actively moving [35]. However, a more refined awake imaging system that further minimizes brain movement would likely be necessary to directly probe these phenomena with LSCI or MESI.

Conclusion
We have used MESI to continuously image CBF during the induction of general anesthesia with isoflurane in head-restrained mice. The vasodilatory effects of isoflurane caused rapid and large anatomical and physiological changes that were drastically different from the awake state. These increases in vessel size and blood flow were repeatable across multiple imaging sessions and subjects. We also documented disparate response times to the induction of isoflurane anesthesia between arteries, veins, and parenchyma regions. This study demonstrated that MESI can be readily used for chronic awake imaging and allows for direct day-to-day comparisons of blood flow. These results provide further evidence that the anesthetized state is unrepresentative of normal physiology and that neurophysiological and behavioral experiments would benefit immensely from transitioning to un-anesthetized awake animal models.