ABSTRACT
The diffusion dynamics of water within the cerebral cortex regulate solute transport and are fundamental to overall brain function. However, cellular‐level microscopic water dynamics within the brain parenchyma remain poorly understood due to a lack of direct, high‐resolution measurement techniques. Here, we present an optical imaging framework to characterize water dynamics at cellular resolution by quantitatively tracking the diffusion patterns of deuterated water using stimulated Raman scattering (SRS) microscopy. High‐resolution imaging of mouse cortical slices reveals heterogeneous water dynamics: free isotropic diffusion within the parenchyma and restricted diffusion around blood vessels. Detailed analysis enabled the simultaneous determination of apparent diffusion coefficients (ADCs) for water and Alexa Fluor 488 (AF488) within the parenchyma, revealing that the water ADC is ≈4.5‐fold larger than that of AF488. Using an ischemic slice model, we demonstrate that the water ADC is reduced by 40%, concomitant with AF488. Furthermore, we directly resolve differences in water dynamics inside and outside the cells, both of which are retarded during acute phase of ischemia. These quantitative findings facilitate a fundamental understanding of water and solute dynamics in the brain under both physiological and pathological conditions, and provide a microscopic basis for interpreting MRI signals.
Keywords: brain parenchyma, diffusion, ischemia, stimulated Raman scattering microscopy, water
A multimodal multiphoton imaging framework enables high‐resolution quantification of water diffusion in brain tissue. By tracking deuterated water via stimulated Raman scattering (SRS) microscopy, this study resolves the microscopic heterogeneity of water diffusion at cellular resolution and determines apparent diffusion coefficients in health and disease. These quantitative findings facilitate a fundamental understanding of water dynamics in the brain.

1. Introduction
Water constitutes approximately 80% of total brain weight and serves as the primary solvent for the biochemical reactions underlying complex neural functions. The clinical importance of water is well‐established; even subtle alterations in brain water content can severely impair mental function [1]. Beyond net volume, the dynamics of water molecules within brain tissue under physiological and pathological conditions have garnered increasing attention. For example, alterations in water diffusion detected by diffusion‐weighted magnetic resonance imaging (DWI) serve as sensitive indicators of neurological disorders, particularly ischemia [2], making DWI an essential diagnostic modality in neurology. In addition to its pathophysiological relevance, water dynamics has gained prominence in physiology following the introduction of the glymphatic system hypothesis. This hypothesis proposes the existence of bulk water flow within brain tissue that facilitates the clearance of parenchymal metabolites, functioning analogously to a specialized lymphatic system [3]. Within the parenchyma, diffusion is considered the primary determinant of water dynamics, with potential contributions from advection [4, 5, 6, 7]. As the principal solvent, water diffusion influences the transport dynamics of all solutes, including ions, neurotransmitters, pathological molecules, and therapeutic agents. Therefore, precise characterization of water dynamics within the brain parenchyma is crucial for understanding brain physiology, pathophysiology, and pharmacology.
Despite its importance, water dynamics in the parenchyma remain poorly understood owing to the lack of appropriate investigative tools. Magnetic resonance imaging (MRI) is widely used to monitor water diffusion [8]; however, its limited spatial resolution precludes assessment of water dynamics at the cellular level. Despite extensive clinical application, a substantial gap persists between macroscopic MRI measurements and the underlying microscopic diffusion processes [9, 10]. MRI is performed at the millimeter scale, whereas the measured signals are considered to reflect subcellular diffusion processes. Therefore, cellular‐level water dynamics, which are critical for understanding solute transport in physiological and pathophysiological contexts, remain poorly characterized. To date, cellular‐level water dynamics have been inferred indirectly, using fluorescent solute tracers. However, solutes with differing physicochemical properties exhibit distinct diffusion behaviors [11] and differ fundamentally from water [12, 13], underscoring the need for direct measurement. Thus, direct experimental characterization of water dynamics within the parenchyma at cellular resolution and determination of fundamental diffusion parameters are essential for advancing understanding of brain physiology, pathophysiology, and pharmacology. Among these parameters, the apparent diffusion coefficient (ADC) is a fundamental metric describing water diffusion and is indispensable for characterizing its transport properties in the parenchyma. Such data are also vital for realistic numerical simulations of water dynamics in brain tissue, which have been increasingly employed to elucidate parenchymal transport mechanisms [14, 15].
Raman scattering imaging, together with Raman tags, enables visualization of small molecules that cannot be labeled with conventional fluorophores [16, 17]. Stimulated Raman scattering (SRS) microscopy, which utilizes multiphoton excitation with near‐infrared light, permits high‐sensitivity deep tissue imaging. Using this approach, we previously demonstrated simultaneous imaging of influxes of water and fluorescent dyes into brain tissues using SRS‐based multimodal multiphoton microscopy [12]. However, because Raman‐tagged water (D2O) and fluorescent dyes were delivered through a perfusion system requiring tens of seconds, the spatiotemporal resolution of quantitative analyses was severely limited. Herein, we aimed to directly image and quantitatively characterize the diffusion dynamics of water within the murine cortical parenchyma using a pulsative point‐source ejection system and high‐temporal‐resolution imaging. This approach enables characterization of water diffusion at cellular resolution, determination of its diffusion parameters within the parenchyma alongside solutes, and optical validation of diffusion MRI measurements.
2. Results
2.1. Characterization of Deuterated Water and Alexa Fluor 488 (AF488) Diffusion Patterns in Brain Tissue
To analyze water diffusion at a high spatiotemporal resolution, we ejected deuterated water‐containing solutions into brain slices using a fine glass pipette. Coronal brain slices (300‐µm thick) of the cerebral cortex were prepared from 2–3‐week‐old mice and stained with sulforhodamine 101 (SR101) to visualize astrocyte structures [18]. The brain slices were transferred to the heated chamber with a perfusion system to keep the bath solution oxygenated at approximately 37°C. A fine glass pipette filled with deuterated water (50% D2O) and AF488 carboxylic acid was visualized using two‐photon excitation fluorescence and carefully inserted near the middle (approximately 150 µm from the surface) of the cortical slices. Deuterated water and AF488 were pressure‐ejected from the pipette for 100 ms and their diffusion was captured via time‐lapse imaging using a multimodal multiphoton microscope at a frame rate of 8 Hz (Figure 1A). Most of the water molecules in the pipette presumably existed in the form of DOH owing to the fast H–D isotopic exchange in the 50% D2O solution [19]. In support of this idea, the Raman spectrum of the 50% D2O solution exhibited a pattern distinct from the linear sum of the spectra of 100% D2O and 100% H2O (SI Appendix, Figure S1).
FIGURE 1.

High‐resolution multimodal multiphoton imaging reveals locally different diffusion dynamics of deuterated water. A. Schematic diagram of the experiment. A probe solution containing 50% D2O and 500 µM Alexa Fluor 488 (AF488) carboxylic acid was ejected from a pipette tip inserted into an acutely prepared murine cortical slice loaded with sulforhodamine 101 (SR101). B. Representative diffusion patterns at the parenchyma. Imaging was performed in the middle of the parenchyma based on the SR101 image (left panel). The diffusion of the ejected AF488 and deuterated water is shown. Scale bars: 50 µm. C. Representative diffusion patterns around the blood vessels. Blood vessels were identified by continuous patterns of SR101 of the endfeet wrapping around the blood vessels (left). The diffusion of AF488 and deuterated water ejected directly adjacent to the endfeet is shown. Scale bars: 50 µm. D. Analysis of the diffusing deuterated water signals. The spatial patterns of the ejected deuterated water were evaluated using the aspect ratio after fitting them to an elliptical shape. The aspect ratio is significantly higher around blood vessels than in the parenchyma (n = 10).
We first characterized the diffusion patterns of deuterated water and AF488 inside the brain tissues by ejecting the probes at different locations. Subsequently, when the probes were ejected at a position inside the parenchyma, where astrocytes were distributed homogeneously, time‐dependent diffusion of AF488 from a point source was observed (Figure 1B, Figure S2A, B). Diffusing AF488 revealed the positions of neurons that were also distributed homogeneously in the field of view as shadows, owing to their impermeability to the plasma membrane, as shown previously [12, 20]. Simultaneous SRS images successfully captured the diffusion of deuterated water in the same field of view (Figure 1B and Figure S2A, B). The diffusion of deuterated water was faster than that of AF488. However, the diffusion of deuterated water and AF488 showed similar spatial patterns, expanding in all directions from the point source. In sharp contrast to the parenchyma, the spatial patterns of diffusion of deuterated water around the blood vessels differed. We identified blood vessels based on SR101‐stained endfeet that wrapped around them (endfeet‐to‐endfeet distance: 16.7 ± 6.6 µm, mean ± standard deviation) [21]. When deuterated water and AF488 were ejected adjacent to the blood vessels, they appeared to be repelled from the blood vessels and diffused along the structures (Figure 1C, Figure S2C,D). This is consistent with the notion that blood vessels have limited water permeability [22].
To quantitively analyze these differences, diffusion patterns were analyzed by calculating the “aspect ratio” (major axis length/minor axis length of the ellipse‐fitted SRS pattern). The diffusion of deuterated water in the parenchyma was nearly isotropic, with an aspect ratio close to 1, whereas the value was significantly higher when it was ejected close to the blood vessels, showing skewed diffusion (Figure 1D, p = 2.5 × 10−4). Therefore, although the diffusion patterns of deuterated water vary among the observation areas at the tissue level, high‐resolution imaging together with morphological markers can capture the isotropic diffusion of deuterated water in the parenchyma. Theoretical and experimental evidence has established that numerical analyses of isotropically diffusing probe molecules from a point source allow detailed characterization of molecular diffusion and determination of the critical parameter, the ADC [23]. This implies that the ADC of water molecules in the parenchyma can be determined at the cellular level through direct imaging of diffusion at a cellular resolution, in the same manner as that of fluorescent solutes.
2.2. ADC of AF488
We analyzed the diffusion patterns of AF488 to obtain the ADC. We followed well‐characterized analytical methods [23] by fitting the spatial pattern of diffusion at different time points (Figure 2A, Figure S3A, Movie S1), plotting the γ2/4 values at each time point against time, and obtaining the ADC from the linear regression slope (Figure 2B, Figure S3C). To better resolve the time‐dependent changes in signal distribution and thereby increase the precision of linear regression, scanning was performed at a higher temporal resolution of 32.8 Hz (30‐ms interval). The resulting plot showed two phases: frames 1–3 appeared to show steeper slopes, whereas the following points (corresponding to frames 5–16) completely fit a single slope, with frame 4 placed in the middle. Because the timing of frames 1–3 matched that of probe ejection (100 ms) and the slope of the curve corresponded to the speed of the probe's movement, we concluded that the dynamics of probes at frames 1–3 are formed by a combination of diffusion and pressure ejection, whereas those of later frames are governed solely by diffusion. Therefore, linear regressions were performed using 5–16 frames corresponding to 50–380 ms after the end of probe ejection to obtain the ADC (Figure 2B, Figure S3C). From this analysis, we calculated the ADC of AF488 to be 23.17 ± 9.40 × 10−7 cm2s−1 (mean ± standard deviation; Figure 2C). This ADC is similar to and consistent with that previously reported by Xiao and Hrabětová (20.09 ± 1.55 ×10−7cm2s−1) in the mouse cortex [24] despite an order of differences in the measured diffusion space and time (100 µm in 0.5 s in this study, compared with 1 mm in tens of seconds in the study by Xiao and Hrabětová). This consistency indicates the robustness and validity of the analytical method and that this analysis can determine the ADC values at the cellular level with high spatiotemporal resolution.
FIGURE 2.

Measurement of the apparent diffusion coefficient (ADC) of Alexa Fluor 488 (AF488). A. Representative time‐lapse images and spatial profiles of AF488 ejected inside the parenchyma. Two‐photon excited fluorescence images of AF488 in each frame are shown. Intensity profiles of AF488 in each frame were calculated and plotted along the horizontal axis. The gray circles and red curves represent the original points and fitted curves, respectively. Scale bars: 50 µm. B. Plot to calculate the ADC of AF488. The γ values are obtained from each curve fitting and plotted against time. Linear regression was performed using 5–16 frames, and the resulting slope shows the ADC. Time corresponding to dye ejection (0–0.1 s) are shown in gray. C. Summary of the ADC values from all recordings. The box plot shows individual points (circles) and the 25th, 50th, and 75th percentiles (n = 19).
2.3. Validation of Experimental ADC Assessments
Although the high spatiotemporal resolution of our diffusion measurements revealed locally different diffusion patterns (e.g., parenchyma vs. paravascular sites) and allowed ADC measurements inside the parenchyma without complications originating from these locally different diffusion patterns, this may render the system susceptible to subtle experimental factors. Therefore, to examine the potential effects of experimental factors on ADC determination, numerical simulations were performed (Figure S4) using realistic physical parameters. Particles were introduced into the three‐dimensional (3D) space for 100 ms and allowed to diffuse in random directions to simulate Brownian motion. Particles existing in the observation XY plane at each frame time point were analyzed to simulate the experimental results (Figure S4A,B, Movie S2). This simulation reproduced the experimental data and provided ADC values consistent with those used to simulate particle motion (Figure S4C). To examine the potential effect of the region of interest (ROI) chosen for diffusion analyses, ROI widths were systematically altered. The ADC values obtained from ROIs of different sizes showed no significant differences, suggesting that the analyses were not affected by ROI width (Figure S5A). To examine the effects of the systematic movements of dyes potentially imposed by the flow in the chamber, the initial pressure for ejecting dyes, and other unknown factors, a constant movement of particles was introduced in the x‐ and y‐directions. However, the flow did not influence the calculated ADC, even at high speed (10 µm/s) (Figure S5B). Finally, the potential effects of jitter on ejection and image acquisition timing were assessed. The results suggested that jitter had no significant effect on the calculated ADC (Figure S5C). These results support the robustness and validity of our experimental analyses, which are further supported by the consistency of the obtained values with those of a previous report [24].
2.4. ADC of Deuterated Water
After establishing a multimodal multiphoton microscopy‐based high‐resolution diffusion analysis system, we analyzed the diffusion patterns of deuterated water in the same manner as AF488. SRS time‐lapse imaging successfully captured the diffusion of deuterated water ejected from the pipette into the surrounding area of brain tissue (Figure 3A, Figure S3B, Movie S3). Following the consideration of AF488, only data points after the cessation of probe ejection (frame 5 and later) were used to analyze the diffusion of deuterated water to exclude the complications of pressure ejection. Although fast diffusion of water made curve fitting of the signal distribution at later frames more difficult, 7–12 data points (R2 value of curve fitting > 0.5) were confidently obtained for linear regression of the γ2/4‐t plots to acquire the ADC value of water (Figure 3A, B, Figure S3D). This analysis yielded an ADC of deuterated water of 102.42 ± 28.02 × 10−7 cm2s−1 (Figure 3C, mean ± standard deviation), revealing that the ADC of deuterated water is approximately 4.5 times larger than that of AF488 (Figure 2C).
FIGURE 3.

Measurement of the apparent diffusion coefficient (ADC) of deuterated water. A. Representative time‐lapse images and spatial profiles of the deuterated water ejected inside the parenchyma. The stimulated Raman scattering (SRS) images of deuterated water in each frame are shown. The intensity profile of deuterated water in each frame was calculated and plotted along the horizontal axis. The gray circles and red curves represent the original points and fitted curves, respectively. Scale bars: 50 µm. B. Plot to calculate the ADC of deuterated water. The γ values were obtained from curve fitting and plotted against time. Linear regression was performed using frames 5–16, and the resulting slope shows the ADC. Time corresponding to deuterated water ejection are shown in gray. C. Summary of all recordings. The box plot shows individual points (circles) and the 25th, 50th, and 75th percentiles (n = 19).
2.5. Diffusion Properties of Deuterated Water in a Slice Model of Ischemia
To investigate water dynamics in the parenchyma under pathological conditions, analyses were performed using a brain slice model of ischemia, oxygen–glucose deprivation (OGD). To verify the pathophysiological state during OGD, we recorded intracellular calcium dynamics in astrocytes using Fluo‐4 AM and SR101 (Figure S6A). After OGD exposure, astrocytes displayed a distinct, synchronized calcium surge around 10 min as shown previously [25]. This calcium overload is a hallmark of tissue‐wide ionic collapse and anoxic neuronal depolarization, confirming that this protocol effectively models the clinically relevant acute onset of the ischemic core. Diffusion was assessed at the same site before and after the OGD exposure. After 10 min of OGD, prominent changes were observed in AF488 diffusion; diffusion was restricted in space (Figure 4A, B). Indeed, the calculated ADC significantly decreased under OGD (Figure 4C, p = 3.1 × 10−2). This observation is consistent with the collapse of neuronal physiology at this time point [25] and the accompanying swelling, which reduces the extracellular space (ECS) in the parenchyma where the dyes can diffuse [26]. When the diffusion of water was analyzed from the same sets of experiments, the ADC of deuterated water decreased after OGD to nearly half (OGD/control ratio; 0.61 ± 0.38), demonstrating that water diffusion is hindered within the parenchyma under an ischemic condition (Figure 4D, p = 3.4 × 10−3). These data revealed that the reduction in the ADC of water occurs locally in the parenchyma at the cellular level under ischemic conditions.
FIGURE 4.

The apparent diffusion coefficient (ADC) of Alexa Fluor 488 (AF488) and deuterated water under oxygen–glucose deprivation (OGD). A. Representative time‐lapse images of AF488 and deuterated water ejected into the parenchyma before OGD. Two‐photon excited images of AF488 and SRS images of deuterated water in each frame are shown, with the sulforhodamine 101 (SR101) image on the left side. Scale bars: 50 µm. B. Time‐lapse images of AF488 and deuterated water ejected into the parenchyma at the same location as (A) 10 min after OGD. C. Comparison of the ADC values of AF488 before and after OGD (n = 7). D. Comparison of the ADC values of deuterated water before and after OGD (n = 12).
2.6. Cellular Resolution Analyses
The ADC was determined based on the gross spatial patterns of deuterated water diffused in the parenchyma; more detailed information on water diffusion might be extracted from time‐lapse images. As simultaneously obtained images of AF488 provided information regarding the location of cellular somata, it raises the possibility that the diffusion dynamics of water can be further analyzed in the parenchyma at a higher resolution. Thus, ROI were set at the somatic region (soma) and juxtaposed region (extrasoma) based solely on AF488 images, and time‐dependent changes in SRS signals were analyzed from these ROIs (Figure 5A). Note that an extrasoma is not composed exclusively of ECS but is a mixture of ECS and small‐diameter neuronal and glial processes. The signals showed an immediate increase in intensity upon the ejection of deuterated water, followed by a decrease that could be readily fitted by a single exponential decay curve (Figure 5B). The height and time constants of the decay curve fit can be used as measures of the maximum concentration after ejection, as well as the following diffusion speeds of deuterated water at each ROI. First, the values were compared under normal conditions. This revealed a higher intensity (Figure 5C, p = 1.9 × 10−2) and smaller time constant (Figure 5D, p = 3.2 × 10−4) in the extrasoma than in the soma. These results are consistent with the idea that the cellular membrane imposes a barrier to the deuterated water ejected into the ECS. The same barrier then slows the diffusion of deuterated water inside the soma compared with that in the extrasoma, in the following diffusion phase. Therefore, this detailed analysis revealed the water dynamics inside and outside the cells, as well as the diffusion barrier function of the plasma membrane.
FIGURE 5.

Detailed characterization of deuterated water diffusion. A. Representative images of sulforhodamine 101 (SR101), deuterated water and Alexa Fluor 488 (AF488) right after the end of ejection of the probes. The somatic (black circle) and extrasomatic (red circle) regions were set based on shadow images of AF488. Scale bar: 50 µm. B. Time course of stimulated Raman scattering (SRS) intensities in somatic and extrasomatic regions shown in (A). Decay of signals were fitted with single exponential decay function to estimate decay time constant (τ) for both regions. C. Comparison of the SRS peak intensities between adjacent somatic and extrasomatic regions (n = 11). D. Comparison of decay time constant of SRS signals between adjacent somatic and extrasomatic regions (n = 11).
We then extended this detailed analysis to the OGD data (Figure 6). Comparisons at the same ROI before and after OGD demonstrated that the decay became slower after OGD in the extrasoma (Figure 6A, B, p = 2.0 × 10−3). Furthermore, the retardation of water clearance was also observed inside the cells (Figure 6A, B, p = 2.0 × 10−2). Considering that the clearance of water at the extrasoma is normally faster than that at the soma, this result indicates that somatic water efflux is hindered as a result of retarded diffusion of extrasomatic water under OGD. This scenario predicts that the distribution of water in the extrasomatic region upon ejection is hindered; therefore, the unequal distribution between the soma and extrasoma may become smaller. Indeed, detailed spatial pattern analysis of deuterated water revealed that the solidity (concavity) of the SRS signal at frame 5 became larger and closer to 1 after OGD (Figure 6C, p = 1.6 × 10−2), further supporting the idea of severely affected extracellular water dynamics and a subsequent reduction in intracellular water diffusion under OGD. These data support the above‐described observation of a lower ADC of water under OGD from a different aspect and provide further evidence that the retardation of water dynamics occurs both inside and outside cells.
FIGURE 6.

Detailed characterization of deuterated water diffusion under oxygen–glucose deprivation (OGD). A. Representative time courses of stimulated Raman scattering (SRS) signals in the extrasomatic (left) and somatic (right) regions before (black) and 10 min after OGD (red). The decay of signals was fitted for quantifications. B. Comparisons of decay time constant before and 10 min after OGD at the same sites in the extrasomatic (left) and somatic (right) regions (n = 10). C. SRS signal patterns right after ejection before and after OGD (left) were analyzed for solidity and compared (right; n = 11). D. Ratios of height (left) and decay time constant (right) of adjacent extrasomatic and somatic regions are calculated and compared before and after OGD (n = 10).
Finally, the ratios of the height and time constant of the extrasoma and soma were calculated and compared before and after OGD. This analysis revealed no significant differences between the two groups (Figure 6D, height; p = 1.6 × 10−1, time constant; p = 5.6 × 10−1). The lack of a significant change in the somatic‐to‐extrasomatic signal height ratio and time constants suggests that the initial barriers to water exchange between compartments remain relatively stable during early OGD, rather than undergoing a catastrophic loss of membrane integrity. This physical retention of compartment boundaries was further corroborated by extracellular shadow imaging, which demonstrated that the cell membrane remained structurally intact and excluded the extracellular fluorophore at 10 min of OGD, prior to severe membrane breakdown observed at 14 min (Figure S6B). It should be noted, however, that because this metric does not directly measure absolute membrane permeability, these findings should be interpreted as an indication of preserved compartment partitioning rather than a direct proof of unchanged membrane permeability. Taken together, our analyses revealed that water dynamics inside and outside the cells were retarded as early as 10 min under ischemic conditions.
3. Discussion
Diffusion of molecules in the brain is gaining increasing attention. Pioneering studies have provided the diffusion parameters for different molecules [11]. The molecular diffusion patterns inside the brain tissue are revealed by two major techniques: electrochemical analysis of ions and optical analysis of fluorescent molecules, termed integrative optical imaging (IOI). Both methods are efficient for analyzing molecular dynamics, but have limitations. Electrochemical analyses are performed using a single electrode and therefore cannot provide spatial information. IOI overcomes this limitation, but requires fluorophores for imaging; thus, it cannot be applied to small molecules whose physical properties are severely affected by fluorophore tagging. Water is arguably the best example of a molecule that cannot be analyzed using these methods, despite the increasing interest in its dynamics inside brain tissues. To circumvent this problem, numerical simulations have been performed to understand the nature of water diffusion inside the brain tissues, including the highly influential glymphatic system hypothesis [14, 27]. In these studies, experimentally obtained diffusion parameters are indispensable [15, 27]. Thus, our study provides a critical diffusion parameter for water molecules for use in cell‐/tissue‐level simulation studies. Importantly, changes in the ADC of water under ischemia observed in this study raise the possibility that other factors may change water dynamics, and thereby, the dynamics of solutes. Considering that sleep/wake states change the ECS of the cortex through changes in astrocyte volume [28, 29], it is tempting to speculate that such a state of the brain may change water dynamics in the parenchyma, which then affects the physiology of neurons and astrocytes.
Multimodal multiphoton microscopy realizes the simultaneous use of an SRS‐based technique and conventional two‐photon excitation imaging, allowing simultaneous comparison of the diffusion patterns of molecules with different properties. In this study, we determined the ADC of AF488 to be 23.17 ± 9.40 × 10−7 cm2s−1. Xiao and Hrabětová used well‐established IOI and reported the ADC of AF488 as 20.09 ± 1.55 × 10−7cm2s−1 in a mouse cortex [24]. This match occurred despite differences in the spatiotemporal scales measured; diffusion for tens of microns in hundreds of milliseconds in this study compared with hundreds of microns in tens of seconds in that of Xiao and Hrabětová. The consistency between these values suggests the robustness of both measurements as well as the scale‐free nature of diffusion at these spatiotemporal scales. While our current work demonstrates water diffusion in mammalian brain tissue, the robustness of this approach holds strong potential for investigating water dynamics across diverse biological systems, ranging from animal pathology to plant physiology [30]. A technical limitation is that water diffusion was measured using deuterated water rather than pure H2O. The ADC of D2O is reported to be approximately 20% smaller than that of H2O in a free solution [31], which was confirmed by careful side‐by‐side assessments of diffusion coefficient of different D2O/H2O mixed water reported recently [32]. Although 50% D2O may briefly alters local fluid properties upon micro‐ejection, rapid interstitial dilution and H‐D exchange [19] swiftly bring the system into a low‐deuterium regime. Consequently, the perturbation to water mobility remains minor, allowing the derived ADC to closely reflect physiological H2O diffusion in brain tissue.
The determination of the absolute ADC using the approach adopted in this study prompts a comparison with those obtained using other techniques. Analyses of MRI data produced lower but closer ADC values (0.6–1.0 × 10−5 cm2s−1) [33, 34] to that determined in this study (1.02 ± 0.280 × 10−5 cm2s−1). This is notable, considering that the current method and MRI involve completely different mechanisms; our results provide an independent validation of MRI‐derived ADC values from a fundamentally different principle and perspective. In terms of spatial resolution, our system is well located to address the MRI gap; although the voxel resolution is approximately 1 mm3, the ADC of water in MRI is considered to reflect water diffusion within tens of milliseconds. This limited length of time is expected to allow water to diffuse approximately 10 microns inside and outside cells, the contribution of which depends on their relative abundances inside the voxel. Our characterization of water diffusion revealed that, although water diffuses almost freely in the parenchyma, it is repelled from the blood vessels (Figure 1). Because MRI voxels are on the order of millimeters, where numerous wide‐diameter blood vessels exist, the lower ADC values in MRI measurements might result from the average of relatively fast diffusion inside the parenchyma and hindered diffusion around blood vessels, as well as other large objects. Thus, we propose that actual ADC of water inside parenchyma is 1.0–1.2 × 10−5 cm2s−1 (considering underestimation due to deuterium) in the murine cortex. Another intriguing comparison with MRI is the change in water diffusion in the ischemic brain. Accumulated evidence has shown that diffusion‐weighted MRI can detect ischemic sites as having a reduced ADC and it is now utilized in clinics as an important diagnostic tool. Although many theoretical considerations have been made to understand the underlying mechanisms, analyses of other principles at high spatiotemporal resolutions are warranted. Our cellular‐level quantitative direct imaging of water is complementary to MRI and should shed new light on MRI observations from different perspectives.
4. Conclusions
Understanding how water diffuses within the brain is crucial for elucidating brain functions, yet direct measurement of water diffusion at cellular resolution has remained a challenge. Using stimulated Raman scattering microscopy, we directly visualized and quantified water diffusion within the brain parenchyma. We revealed heterogeneous diffusion patterns inside the murine cortex and determined the apparent diffusion coefficient (ADC), the most fundamental parameter of water diffusion. We showed that ischemia severely hinders water diffusion both inside and outside cells, providing new microscopic insights into stroke pathophysiology. Consequently, our study provides new insight into the microscopic water dynamics of brain tissues and a means to bridge microscopic and macroscopic diffusion measurements in physiological and pathological brain states.
5. Methods
5.1. Materials
All chemicals were purchased from Sigma Aldrich Japan (Tokyo, Japan) unless otherwise noted.
5.2. Brain Slice Preparations
The Institutional Animal Care and Use Committee of Keio University approved all procedures related to animal care and treatment. Brain slices were prepared from 2–3‐week‐old C57BL/6JmsSlc mice of either sex (Sankyo Labo Service Corp., Shizuoka, Japan) using a Leica VT1200 microtome (Leica Microsystems, Wetzlar, Germany), following the method described in a previous study [12]. Briefly, the mice were anesthetized using isoflurane and decapitated, and their brains were quickly removed and placed in an ice‐cold cutting solution containing 222 mM sucrose, 27 mM NaHCO3, 2.6 mM KCl, 1.5 mM NaH2PO4, 0.5 mM CaCl2, and 7 mM MgCl2, bubbled with 95% O2/5% CO2 at pH 7.4. Coronal brain slices (300‐µm thick) were prepared using a Leica VT1200S Microtome in ice‐cold cutting solution. Subsequently, the slices were transferred into an aCSF containing 126 mM NaCl, 3 mM KCl, 1.14 mM NaH2PO4, 26 mM NaHCO3, 3 mM CaCl2, 1 mM MgCl2, and 10 mM dextrose pre‐warmed to 37°C and bubbled with 95% O2/5% CO2 at pH 7.4 and supplemented with 200 nM SR101 to label astrocytes. The brain slices were incubated for 30 min for recovery before being cooled to room temperature (approximately 22°C) before use.
5.3. Multimodal Multiphoton Microscopy
Multimodal multiphoton microscopy was performed as described previously [12]. Briefly, an Emerald Engine picosecond laser (Angewandte Physik & Elektronik GmbH, APE, Berlin, Germany) was used to emit picosecond lasers at 1,032 nm and 516 nm. The 516‐nm laser was fed to a Levante IR picosecond optical parametric oscillator (APE) to emit a pump beam tuned to approximately 823 nm to measure the O–D vibration at 2,450 cm−1. Typical laser intensities at the objective lenses were 50–200 mW for the pump and Stokes beams. Pump and Stokes lasers were introduced into a custom‐made laser scanning multimodal multiphoton microscope based on ThorLabs (Thorlabs, New Jersey, U.S.A.) optical components and run using Scanimage software (Vidrio Technologies, Virginia, U.S.A.) and focused on samples using a 25× objective lens designed for multiphoton microscopy (Evident, NA 1.05, Tokyo, Japan). Subsequently, two‐photon excited fluorescence signals were collected using photomultiplier tubes (Thorlabs) with appropriate optical filters (520–550 nm for green and 580–620 nm for red). The SRS signals were obtained using a custom‐made lock‐in amplifier based on changes in the pump laser collected by the photodiode induced by the modulation of the incoming Stokes laser. To image water diffusion, time‐lapse imaging was performed at 1.6 µs/pixel for 128 × 128 pixels at a 3× zoom, yielding a 32.8‐Hz frame scan (30 ms/frame). Imaging was performed at 37°C using chamber and solution heaters (Warner Instruments, Massachusetts, U.S.A.) with continuous perfusion with oxygenated aCSF (290–310 mOsm/kg, 95% O2 / 5% CO2). Under these standardized physiological conditions, acute brain slices retain an in vivo‐like hydration state and extracellular volume fraction (α = ≈0.2), as validated by TMA+ measurements [11].
5.4. Delivery of Deuterated Water and AF488
Glass pipettes were manufactured using a Sutter P‐1000 micropipette puller (Sutter Instruments, California, U.S.A.). The pipettes were loaded with 50% D2O and 500 µM Alexa Fluor 488 carboxylic acid (AF488; Thermo Fisher Scientific, Massachusetts, U.S.A.) in 4‐(2‐hydroxyethyl)‐1‐piperazineethanesulfonic acid (HEPES)‐based aCSF containing 150 mM NaCl, 3 mM KCl, 10 mM dextrose, and 10 mM HEPES. The pipettes were visualized via two‐photon excitation of AF488 to guide them to the target position in layers II and III of the visual cortex located in the middle of the brain slices using a micromanipulator (Narishige, Tokyo, Japan). Ejections were performed for 100 ms at 3–5 psi, using a programmed ejector (Narishige).
5.5. OGD
OGD was induced as previously described [12, 25]. Briefly, the brain slices were imaged under normal conditions to characterize the diffusion patterns of AF488 and deuterated water as described above. Subsequently, the external solution was changed from a normal solution to an OGD solution containing 10 mM sucrose instead of dextrose, and bubbled with 95% N2/5% CO2 rather than 95% O2/5% CO2. Next, 10 min after the solution was changed, diffusion analyses were performed at the same site to compare the diffusion patterns before and after OGD.
5.6. Calcium Imaging
Calcium imaging of astrocytes under OGD was performed as described previously [25]. Briefly, Fluo‐4 AM (final 10 µM, Thermo Fisher Scientific) was pre‐mixed with 0.02% Pluronic F‐127 (Thermo Fisher Scientific) and introduced to brain slices together with SR101 at 37°C for 1 hr. Brain slices were exposed to OGD condition as above and time‐lapse imaging was performed at 1‐Hz for 10 – 15 min.
5.7. Shadow Imaging
For shadow imaging, SR101‐stained brain slices were imaged under continuous presence of 5 µM Alexa Fluor 488 carboxylic acid (Thermo Fisher Scientific) in aCSF. The same concentration of Alexa Fluor 488 was also included in the OGD solution and images of Alexa Fluor 488 together with SR101 were obtained at different time points after OGD exposure.
5.8. Data Analyses
Intensity profiles were obtained from each frame using Fiji software (NIH, U.S.A.). First, square ROIs with center positions close to the pipette tip were selected. The intensity profiles of the same ROIs along the horizontal axis were obtained from time‐lapse images using the Fiji plot profile function. The background intensity values, calculated by averaging the intensities of the three images before the ejection of deuterated water and dyes, were subtracted from the post‐ejection data for background adjustment. The obtained data were imported into the MATLAB workspace (MathWorks, Massachusetts, U.S.A.) for diffusion analyses following the method of Nicholson and Tao [23]. The intensity profiles of the individual frames were fitted with the following function using the custom‐function curve fitting of the MATLAB curve fitter application:
| (1) |
The γi values were obtained From sequential images and plotted against time as follows:
| (2) |
By fitting the data using linear regression, the ADC D* was obtained using Origin Pro software (Origin Lab,, Massachusetts, U.S.A.). For analyses of shapes of diffused molecules, “Aspect Ratio” and “Solidity” were obtained from “Analyze Particle” functions of Fiji performed on auto‐thresholded images.
5.9. Numerical Simulation
The potential effects of the systematic movement of molecules and other factors were assessed through numerical simulations using the MATLAB software. In this simulation, 2–5 × 104 particles were allowed to move in random directions in 3D space. To mimic ejection from the pipette, the initial positions of the particles were assigned a normal distribution centered at (x, y, z) = (0, 0, 0) with a standard deviation of 10 µm. Subsequently, the particles were released from the initial volume of the ejection site sequentially for 100 ms at various starting times to assess the timing between ejection and image acquisition. Once released, each particle moved in a random direction for 5 × 10−5 s with the speed defined by a given ADC in each cycle until the end of the 0.5 s simulation (1 × 104 cycles). To simulate systematic flow, constant movement in the xy‐direction was added to each cycle's movement. To simulate multiphoton microscopy observations, particles existing −2 to +2 µm in Z‐axis were counted as those observed in the observation plane. ROIs to analyze particles were defined by various ranges of y positions (e.g., −30 to +30 µm) and defined x positions (−100 to +100 µm). At each frame time (1–16 frames with 30‐ms intervals from the beginning of the simulation), the number of particles inside the ROI was counted with the x‐axis divided into 1,000 sections to calculate the intensity profiles along the x‐axis. The obtained particle distributions were subsequently fitted as in the experiments described above to calculate the ADC.
5.10. Spontaneous Raman Scattering Spectroscopy
The Raman spectra of 100% H2O, 50% D2O + 50% H2O, and 100% D2O were obtained using a custom‐made spontaneous Raman scattering spectroscope. The system is composed of a 561.2‐nm laser (Cobolt, Solna, Sweden), IX‐73 microscope (Evident, Tokyo, Japan), PlanApo N 60× Objective lens (N.A. 1.42, Evident), MK‐300 imaging spectrometer (Bunkokeki, Tokyo, Japan), iDus416 CCD camera (Oxford Instruments, Buckinghamshire, U.K.) run on Solis software (Oxford Instruments). Solutions were placed under a microscope, and spectra were obtained with a spectrometer set at a central wavelength of 650 nm and 600 lines/mm grating with 20 accumulations recorded every 5 s with a cosmic ray removal filter.
5.11. Statistical Analysis
Normality of the dataset were examined using Origin Pro (Originlab). To test for statistical significance, non‐parametric Mann–Whitney U tests and analysis of variance were used to compare groups. Wilcoxon signed‐rank tests were used to compare the paired samples. The boxes in the box plots represent the 25th, 50th, and 75th percentiles. Paired data points are connected by lines.
Funding
Japan Agency for Medical Research and Development (AMED) 25ak0101251h0001 (MN); JSPS KAKENHI 24K02217, 24K21952, and 20K20593 (MN); JST‐Mirai Program JPMJMI22G5 (MN).
Conflicts of Interest
The authors declare no conflicts of interest.
Supporting information
Supporting File 1: smll75652‐sup‐0001‐SuppMat.docx.
Supporting File 2: smll75652‐sup‐0002‐MovieS1.avi.
Supporting File 3: smll75652‐sup‐0003‐MovieS2.mp4.
Supporting File 4: smll75652‐sup‐0004‐MovieS3.avi.
Acknowledgements
The authors thank Dr. Yasuyuki Ozeki (University of Tokyo) and Dr. Hideaki Kano (Keio University) for their tremendous help in setting up the SRS and spontaneous Raman systems, respectively, and Dr. Yoshitaka Bito (Hokkaido University) for helpful discussions regarding the data.
Data Availability Statement
All original data included in this study, as well as the MATLAB code used for the simulations, were deposited in Mendeley Data with https://doi.org/10.17632/rdkfnxchfh.3.
References
- 1. Benton D. and Young H. A., “Do small differences in hydration status affect mood and mental performance?,” Nutrition Reviews 73 (2015): 83–96, 10.1093/nutrit/nuv045. [DOI] [PubMed] [Google Scholar]
- 2. Le Bihan D. and Johansen‐Berg H., “Diffusion MRI at 25: Exploring Brain Tissue Structure and Function,” Neuroimage 61 (2012): 324. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Hablitz L. M. and Nedergaard M., “The glymphatic system,” Current Biology 31 (2021): R1371. [DOI] [PubMed] [Google Scholar]
- 4. Jin B. J., Smith A. J., and Verkman A. S., “Spatial model of convective solute transport in brain extracellular space does not support a “glymphatic” mechanism,” Journal of General Physiology 148 (2016): 489–501, 10.1085/jgp.201611684. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Holter K. E., Kehlet B., Devor A., et al., “Interstitial solute transport in 3D reconstructed neuropil occurs by diffusion rather Than bulk flow,” Proceedings of the National Academy of Sciences 114 (2017): 9894–9899, 10.1073/pnas.1706942114. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Bohr T., Hjorth P. G., Holst S. C., et al., “The Glymphatic System: Current Understanding and Modeling,” iScience 25 (2022): 104987. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Hladky S. B. and Barrand M. A., “The glymphatic hypothesis: The theory and the evidence,” Fluids and Barriers of the CNS 19 (2022): 9, 10.1186/s12987-021-00282-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Alshuhri M. S., Gallagher L., Work L. M., and Holmes W. M., “Direct Imaging of Glymphatic Transport Using H217O MRI,” JCI Insight 6 (2021): 141159. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Le Bihan D. and Iima M., “Diffusion Magnetic Resonance Imaging: What Water Tells Us about Biological Tissues,” PLoS Biology 13 (2015): 1002203. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Schaefer P. W., Grant P. E., and Gonzalez R. G., “Diffusion‐weighted MR Imaging of the Brain,” Radiology 217 (2000): 331–345, 10.1148/radiology.217.2.r00nv24331. [DOI] [PubMed] [Google Scholar]
- 11. Syková E. and Nicholson C., “Diffusion in Brain Extracellular Space,” Physiological Reviews 88 (2008): 1277–1340, 10.1152/physrev.00027.2007. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Shinotsuka T., Miyazawa T., Karasawa K., Ozeki Y., Yasui M., and Nuriya M., “Stimulated Raman scattering microscopy reveals a unique and steady nature of brain water dynamics,” Cell Reports Methods 3 (2023): 100519, 10.1016/j.crmeth.2023.100519. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Syková E., Voříšek I., Starčuk Z., et al., “Disruption of Extracellular Matrix and Perineuronal Nets Modulates Extracellular Space Volume and Geometry,” The Journal of Neuroscience 45 (2025): 0517242024, 10.1523/JNEUROSCI.0517-24.2024. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Martinac A. D. and Bilston L. E., “Computational modelling of fluid and solute transport in the brain,” Biomechanics and Modeling in Mechanobiology 19 (2019): 781–800, 10.1007/s10237-019-01253-y. [DOI] [PubMed] [Google Scholar]
- 15. Alcaide D., Cacheux J., Bancaud A., Muramatsu R., and Matsunaga Y. T., “Solute transport in the brain tissue: What are the key biophysical parameters tying in vivo and in vitro studies together?,” Biomaterials Science 11 (2023): 3450–3460, 10.1039/D3BM00027C. [DOI] [PubMed] [Google Scholar]
- 16. Zhao Z., Shen Y., Hu F., and Min W., “Applications of vibrational tags in biological imaging by Raman microscopy,” The Analyst 142 (2017): 4018–4029, 10.1039/C7AN01001J. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Nuriya M., “Uses of Coherent Raman Scattering Microscopy in Neuroscience,” Frontiers in Neuroscience 19 (2026): 1715954, 10.3389/fnins.2025.1715954. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Nimmerjahn A., Kirchhoff F., Kerr J. N. D., and Helmchen F., “Sulforhodamine 101 as a specific marker of astroglia in the neocortex in vivo,” Nature Methods 1 (2004): 31–37, 10.1038/nmeth706. [DOI] [PubMed] [Google Scholar]
- 19. Maithani S., Maity A., and Pradhan M., “Probing the H–D isotopic exchange reaction in a liquid droplet via surface plasmon resonance,” Journal of Analytical Atomic Spectrometry 37 (2022): 544–550, 10.1039/D1JA00350J. [DOI] [Google Scholar]
- 20. Tønnesen J., Inavalli V. V. G. K., and Nägerl U. V., “Super‐Resolution Imaging of the Extracellular Space in Living Brain Tissue,” Cell 172 (2018): 1108–1121.e15, 10.1016/j.cell.2018.02.007. [DOI] [PubMed] [Google Scholar]
- 21. Nuriya M., Shinotsuka T., and Yasui M., “Diffusion Properties of Molecules at the Blood‐Brain Interface: Potential Contributions of Astrocyte Endfeet to Diffusion Barrier Functions,” Cerebral Cortex 23 (2013): 2118–2126, 10.1093/cercor/bhs198. [DOI] [PubMed] [Google Scholar]
- 22. Dickie B. R., Parker G. J. M., and Parkes L. M., “Measuring water exchange Across the blood‐brain barrier using MRI,” Progress in Nuclear Magnetic Resonance Spectroscopy 116 (2020): 19–39, 10.1016/j.pnmrs.2019.09.002. [DOI] [PubMed] [Google Scholar]
- 23. Nicholson C. and Tao L., “Hindered diffusion of high molecular weight compounds in brain extracellular microenvironment measured With integrative optical imaging,” Biophysical Journal 65 (1993): 2277–2290, 10.1016/S0006-3495(93)81324-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Xiao F. and Hrabětová S., “Enlarged extracellular space of aquaporin‐4–deficient mice does not enhance diffusion of Alexa Fluor 488 or dextran polymers,” Neuroscience 161 (2009): 39–45, 10.1016/j.neuroscience.2009.03.017. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Shinotsuka T., Yasui M., and Nuriya M., “Astrocytic gap junctional networks suppress cellular damage in an in vitro model of ischemia,” Biochemical and Biophysical Research Communications 444 (2014): 171–176, 10.1016/j.bbrc.2014.01.035. [DOI] [PubMed] [Google Scholar]
- 26. Risher W. C., Andrew R. D., and Kirov S. A., “Real‐time passive volume responses of astrocytes to acute osmotic and ischemic stress in cortical slices and in vivo revealed by two‐photon microscopy,” Glia 57 (2009): 207–221, 10.1002/glia.20747. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Schreder H. E., Liu J., Kelley D. H., Thomas J. H., and Boster K. A. S., “A hydraulic resistance model for interstitial fluid flow in the brain,” Journal of The Royal Society Interface 19 (2022): 20210812 , 10.1098/RSIF.2021.0812. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Xie L., Kang H., Xu Q., et al., “Sleep Drives Metabolite Clearance From the Adult Brain,” Science 342 (2013): 373–377, 10.1126/science.1241224. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Sherpa A. D., Xiao F., Joseph N., Aoki C., and Hrabetova S., “Activation of β‐adrenergic receptors in rat visual cortex expands astrocytic processes and reduces extracellular space volume,” Synapse 70 (2016): 307. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Zhu Y., Zhu X., Bi S., et al., “Epidermal hydrodynamics controls water homeostasis of shoot meristems for plant adaptation to terrestrial environments,” Developmental Cell 61 (2026): 59–72.e9, 10.1016/j.devcel.2025.09.007. [DOI] [PubMed] [Google Scholar]
- 31. Mills R., “Self‐diffusion in normal and heavy water in the range 1‐45.deg,” The Journal of Physical Chemistry 77 (1973): 685–688, 10.1021/j100624a025. [DOI] [Google Scholar]
- 32. Bi S., Li Y., Ao J., Liu Z., Weng M., and Ji M., “On‐Chip Stimulated Raman Scattering Imaging and Quantification of Molecular Diffusion in Aqueous Microfluidics,” Analytical Chemistry 97 (2025): 2052–2061, 10.1021/acs.analchem.4c04317. [DOI] [PubMed] [Google Scholar]
- 33. Sener R. N., “Diffusion MRI: Apparent diffusion coefficient (ADC) values in the normal brain and a classification of brain disorders based on ADC values,” Computerized Medical Imaging and Graphics 25 (2001): 299–326, 10.1016/S0895-6111(00)00083-5. [DOI] [PubMed] [Google Scholar]
- 34. Le Bihan D., “Apparent Diffusion Coefficient and Beyond: What Diffusion MR Imaging can Tell us About Tissue Structure,” Radiology 268 (2013): 318. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supporting File 1: smll75652‐sup‐0001‐SuppMat.docx.
Supporting File 2: smll75652‐sup‐0002‐MovieS1.avi.
Supporting File 3: smll75652‐sup‐0003‐MovieS2.mp4.
Supporting File 4: smll75652‐sup‐0004‐MovieS3.avi.
Data Availability Statement
All original data included in this study, as well as the MATLAB code used for the simulations, were deposited in Mendeley Data with https://doi.org/10.17632/rdkfnxchfh.3.
