Novel insights into vascular dysfunction in cuprizone-induced demyelination through functional ultrasound imaging

Multiple sclerosis (MS) is an autoimmune neurodegenerative disorder of the central nervous system. It is the most common cause of neurological disability in young adults, affecting approximately 2.8 million worldwide (Ramagopalan et al., 2010). Chronic inflammation, demyelination, gliosis, and neuronal loss are hallmarks of the pathology, producing plaque formation and tissue destruction, in both white and grey matter (Garg & Smith, 2015; Lassmann, 2013). Magnetic Resonance Imaging (MRI) is largely used for MS diagnosis and follow up, but to date, it is not able to distinguish among MS disease courses. Preclinical studies are crucial for understanding these mechanisms and developing new therapeutics. Models of demyelination, such as cuprizone, have been pivotal in validating novel imaging techniques, including magnetization transfer ratio (Tagge et al., 2016), diffusion tensor imaging (Hübner et al., 2017; Sun et al., 2006), and macromolecular proton fraction mapping (M. Khodanovich et al., 2019; M. Y. Khodanovich et al., 2017). Several studies have even utilized multimodal neuroimaging in this model to evaluate the sensitivity and specificity of these techniques, with electron microscopy serving as the reference for anatomical changes (Ding et al., 2021; Friesen et al., 2024; Wood et al., 2016). Among emerging imaging approaches, hyperpolarized 13C magnetic resonance spectroscopic imaging shows promise for assessing inflammatory lesions (Guglielmetti et al., 2017), while TSPO-[18F]DPA-714 PET enables the imaging of proinflammatory cells, such as astrocytes and microglia, across multiple brain regions (Zinnhardt et al., 2019). Furthermore, NIRS-MRI facilitates the quantification of the reduced cerebral metabolic rate (Hashem et al., 2022).

Despite these advances, few studies have examined the functional consequences of demyelination within neural networks (Cerina et al., 2017; Narayanan et al., 2018). While fMRI studies in MS patients have demonstrated abnormal evoked hemodynamic responses, no preclinical study has yet explored this critical aspect.

Ultrafast ultrasound imaging is a relatively new imaging technology that detects changes of cerebral blood volume with a great sensitivity (Deffieux et al., 2021; Mace et al., 2013), allowing functional studies in both anesthetized (Claron et al., 2021; Mace et al., 2011) and awake animals (Bimbard et al., 2018; Matei et al., 2022), the study of functional connectivity (Osmanski et al., 2014; Rahal et al., 2020), of microvascular flux at a microscopic scale with a large field of view (Errico et al., 2015; Réaux-Le-Goazigo et al., 2022), and finally functional studies at a microscopic scale (Renaudin et al., 2022).

Given the scarcity of preclinical studies investigating functional hyperemia in animal models of CNS demyelination, this study aimed to investigate alterations in cortical hemodynamic response to sensory stimulation in a widely used animal model of CNS demyelination. Cuprizone (bis-cyclohexanone-oxalyldihydrazone, CPZ), a copper chelator, induces chronic (non-immune) demyelination of the brain by targeting oligodendrocytes (Matsushima & Morell, 2001; Praet et al., 2014; Vega-Riquer et al., 2019). Its withdrawal facilitates the study of natural remyelination and its enhancement via pharmacological interventions. Given the extensive sensory innervation of the whisker pad in rodents (Petersen, 2019), we chose the barrel cortex as a model system. Our study demonstrates long-lasting alterations of the hemodynamic response in cuprizone-induced demyelinated animals, and three descriptors of this evoked response correlate with the level of Myelin Basic Protein in several key brain areas involved in sensory processing. Our results also suggest disrupted vascular responsiveness, paralleling clinical observations from the early phase of the disease.

All experiments performed in this study complied with the French and European Community Council Directive of September 22 (2010/63/UE). They were also approved by the local Institutional Animal Care and Ethics Committees (#59, ‘Paris Centre et Sud’ project #2019-41). In adherence to these guidelines, the number of animals in our study was kept to the minimum necessary. Using previously published data and preliminary data on the effect of cuprizone treatment, we performed a G-Power analysis (https://www.psychologie.hhu.de/arbeitsgruppen/allgemeine-psychologie-und-arbeitspsychologie/gpower) and determined that N = 8 mice per group were sufficient to detect statistical differences in our imaging experiments.

Our study examined male mice only because male animals exhibited less variability in phenotype. It is unknown whether the findings are relevant for female mice. Therefore, this is one of the limitations of this study.

Our experimental design adhered to the ARRIVE guidelines and included 32 adult mice. After a 1-week acclimation period, the animals were randomly assigned to four equal groups (N = 8 each) and underwent baseline imaging (D0). Following this imaging session, each group was treated as follows (Fig. 1A):

• -

  Control group: Received a regular diet throughout the 7-week study. These animals were imaged at 3 W, 5 W, and 7 W post-baseline.

• -

  Early demyelination group: Fed a 0.3% cuprizone diet for 3 weeks and imaged at 3 W before perfusion.

• -

  Progressive demyelination group: Fed a 0.3% cuprizone diet for 5 weeks. Imaging was performed at 3 W and 5 W to follow and monitor the progression of evoked hemodynamic response during demyelination.

• -

  Remyelination group: Fed a 0.3% cuprizone diet for 5 weeks, followed by a return to regular diet for 2 weeks. These animals were imaged at 5 W and 7 W (Fig. 1A).

At the end of the final imaging session, all animals were perfused intracardially to collect brain tissue for Myelin Basic Protein (MBP) immunofluorescent staining (detailed below). No animals were excluded from the study. The experimenter conducting the functional ultrasound (fUS) imaging was blinded to treatment groups.

The experiments used 32 male adult mice (C57BL/6 Rj, 2–3 months old, 20–30 g, from Janvier Labs, France). Mice were housed 8 per cage (also per group) in very large cages (dimensions: 300 mm x 234 mm x 412 mm, Floor space: 916 cm²), under controlled conditions (22 ± 1°C, 60 ± 10% relative humidity, 12/12 h light/dark cycle) with food and water available ad libitum. Animals acclimated to housing conditions for 1 week prior to the experiment.

Demyelination was induced in the three treatment groups by feeding the animals with a diet containing 0.3% cuprizone (bis-cyclohexanone oxaldihydrazone; Sigma-Aldrich Inc., St. Louis, MO, USA) for 3 or 5 weeks. Following recommendations (Kipp, 2024; Zhan et al., 2020), 0.3 g of fresh Cuprizone powder was mixed every day with 100 g of ground food (SAFE, France). Mice were weighed twice weekly to monitor weight loss, and all cages were devoid of enrichment.

At the end of the imaging session, while the animals were still anesthetized, they received an intraperitoneal (IP) injection of 0.05 mL of Euthasol (Dechra Pharmaceuticals, 364.60 mg/mL pentobarbital solution). A thoracotomy was performed, and the right atrium was incised. Animals were perfused transcardially with 10 mL of saline solution (0.9% NaCl), followed immediately by 40 mL of 4% paraformaldehyde. Brains were extracted, fixed overnight in 4% paraformaldehyde at 4°C, and cryoprotected in a 30% sucrose solution for 2 days. Brains were frozen in an OCT (Optimal Cutting Temperature) matrix using chilled isopentane (−40°C) on dry ice and stored at -80°C. Brain sections (12 μm thick) were prepared using a cryostat (Leica CM 3050S, Wetzlar, Germany), mounted on Superfrost slides (Thermofisher Scientific, Waltham, MA, USA), and stored at -20°C in cryoprotectant until processing for all animals was complete.

Two slides per animal (each one containing eight sections) were washed three times in 0.1 M phosphate-buffered saline (PBS) with 0.9% NaCl. Sections were incubated overnight at room temperature with the primary antibody (mouse anti-MBP, clone 12, Chemicon, Avantor-VWR, #MAB384, dilution: 1:150, diluted in 0.3% triton X-100). After the three additional PBS washes, sections were incubated for 2 h with the secondary antibody (Alexa Fluor 488-conjugated donkey anti-mouse antibody, dilution: 1:1000; Invitrogen). Finally, sections were cover slipped using Fluoromount (Sigma-Aldrich). All staining was performed simultaneously for all animals to ensure consistency.

As this model induces a widespread but heterogenous demyelination of the brain, we chose to quantify MBP immunofluorescent staining in several areas involved in sensory processing, including the primary sensory cortex (barrel field), the thalamus (VPM, VPL), the internal capsule, the medial corpus callosum, and the hippocampus.

All stained sections were digitized using a nanozoomer (Zeiss, facility of the Vision Institute, Paris) uniform acquisition parameters. The resulting files were opened with the manufacturer’s software, and two sections per animal, corresponding to coordinates: Bregma -1.34 and -1.70 mm were exported in PNG format. These files were then opened with the Fiji software (ImageJ.net) and transformed into 256- level grayscale images. Two types of quantification were performed:

• i)

  Measurement of the mean gray intensity

  Since our immunostainings were performed using indirect immunofluorescence, with all sections stained simultaneously, the measurement of mean gray intensity serves as a direct and linear indicator of the amount of MBP present in the tissue. For each image, the mean gray value of the background (outside the brain; see Supplementary Fig. 2a) and the MBP staining in various brain regions (Supplementary Fig. 2b–f) were quantified on both the right and left sides. After subtracting the background level for each section and averaging the values from the right and left sides, the data from the two sections were further averaged. The analysis was conducted by an experimenter blinded to the treatment groups. Results were expressed as mean gray intensity ± SEM.

• ii)

  Measurement of the percentage of area covered by the staining

• iii)

  Using the same 8-bit black-and-white images, we performed an additional quantification of the percentage of the area covered by staining. A common threshold value (64) was applied to all sections using ImageJ. The area occupied by the staining was then measured for both the right and left sides and averaged across two sections per animal. Results (presented in Supplementary Fig. 5) were expressed as the mean percentage area ± SEM. This quantification method yielded results consistent with those obtained using mean gray intensity, with a strong linear regression coefficient in the S1BF region for all animals (R² = 0.701).

The mice were initially anesthetized with 2% Isoflurane and an intramuscular (IM) injection of medetomidine (Domitor, 0.08 mg/kg). Anesthesia was maintained using a continuous IM injection of medetomidine (0.08 mg/kg/h) delivered via a syringe pump, in combination with 0.5% isoflurane (0% O₂, 100% air as carrier gas). Anesthesia depth was monitored throughout the imaging sessions using the cardiac and respiratory frequency measurements (Labchart). To prevent hypothermia, body temperature was maintained at 37°C with a rectal probe connected to a heating pad. To protect the mice’s eyes, an ophthalmic ointment (Ocry-gel, TVM, UK) was applied.

Skin alterations, such as scars or local inflammation of the skin above the skull, can reduce the fUS signal, due to ultrasonic wave reflections. To remove such potential artifacts, skin above the imaging site was surgically prepared at the start of each imaging session under anesthesia. Following a local subcutaneous injection of Lidocaine. the skin was incised and opened below the ultrasound probe. Sterile ultrasound gel (1 mL) was applied to the exposed skull. At the end of the imaging session, the skin was sutured with 6.0 sutures (Ethilon, Phymep, France).

To enhance reproducibility in probe placement, the imaging plane (Bregma -1.60 mm) was determined using Icostudio software (Iconeus, Paris, France). A linear scan (6 mm span along the anterior-posterior axis, 0.2 mm step size) was conducted at the beginning of each session, co-registered to an average scan, and aligned with a standard Doppler reference template pre-aligned to the Allen Mouse Brain Atlas common coordinate framework (Nouhoum et al., 2021). This method allowed reproducible placement of the probe above the same imaging plane (coronal plane of the S1BF) with a reproducibility of 200 μm along both X and Y axes (Nouhoum et al., 2021).

Imaging was performed using the Iconeus One scanner (Iconeus, Paris, France) designed for small animals. The linear probe, consisting of 128 piezoelectric transducers with a 15 MHz central frequency, acquired ultrafast images at 500 Hz using the coherent summation of 11 compounded tilted plane waves emitted at a 5,500 Hz pulse repetition frequency.

To standardize whisker stimulation, we used an Arduino Uno-controlled (https://store.arduino.cc/; Bertolo et al., 2021). This setup synchronized stimulation with data acquisition. The servomotor moved a cotton bud perpendicular to the whiskers at 4 Hz. After a 20-sec baseline period, stimulation was applied for 20 sec, followed by 20 sec of rest, repeated 10 times. Total acquisition time was 420 sec.

Blood signal was separated from tissue signal using a singular value decomposition (SVD) clutter filter. Each power Doppler frame represented the average of 200 filtered ultrafast frames; resulting in a 2.5 Hz frame rate. Following established methods (Demené et al., 2015), a clutter filter removed the tissue signal by eliminating the 60 first singular vectors that correspond mainly to the tissue space. The filtered frames were integrated to produce a Power Doppler image every 400 msec. No motion correction was applied.

A detrending step removed slow signal drift during acquisition. Activation maps displaying significantly active areas (Z scores with p < 0.0000006 after Bonferroni correction) were computed using a general linear model (GLM) analysis in Matlab (Fig.1E, bottom panel). The spatial extent of functional hyperemia was quantified by calculating the number of significantly active pixels. Temporal signal data from active pixels were fitted using MATLAB’s “stepinfo” function, which computed parameters such as the rise time (Supplementary Fig. 1).

All statistical analyses were performed using GraphPad Prism 10.

Data normality was assessed using Shapiro-Wilk’s test, and homoscedasticity was evaluated with Barlett’s test.

Statistical comparisons of mean gray density for MBP immunostaining across brain regions were conducted using:

1. ANOVA with Tukey’s multiple-comparison test: For normally distributed data with homoscedasticity.

2. Brown-Forsythe and Welch ANOVA with Dunnett’s T3 test: For normally distributed but heteroscedastic data.

Hemodynamic descriptors (number of active pixels, rise time, and ∆CBV) were analyzed using repeated-measures one-way ANOVA, followed by Tukey’s post- test.

As the development of new pharmacological treatments requires a longitudinal measure of the level of myelin loss, this study aimed at investigating the use of longitudinal measurement of the altered hemodynamic response as a surrogate for the level of myelin. To do so, we investigated the potential ability of the three descriptors of the hemodynamic response measured by fUS imaging (number of active pixels, steady-state CBV variation, and the rise time) to predict the level of myelin (quantified by MBP immunohistochemistry at the end of the experiment). To do so, we trained a multiple linear regression model (consisting in a successive and automatic removal of the less important predictors) with the three descriptors as inputs, using the Matlab function ‘stepwise’. To assess the overall performance of the model, we conducted an F-test to compare the model with all its predictors to a constant model (analysis of variance test). Furthermore, we assessed the contribution of each predictor by testing whether the corresponding parameter significantly differs from zero using the classic t-test.

The funding source had no role in data collection, analysis, interpretation, the writing of the manuscript, or to submit the manuscript for publication.

While in control animals the average weekly weight gain remained relatively constant at 0.61 g ± 0.13, in contrast, cuprizone treatment significantly impacted weight gain, though this effect varied over time.

From days 3 to 15, animals on cuprizone consistently lost weight, with an average loss of -0.88 g ± 0.19 (Fig. 1B, C). Between days 15 and 21, there was heterogeneity in weight changes: while some animals continued to lose weight, others began to gain weight. On average, the change in weight during this period was negligible (0.0 g ± 0.36 g over N = 24, Fig. 1C). During the final week of cuprizone treatment (days 28–35), animals’ weekly weight gain returned to levels comparable to controls, with an average gain of +0.79 g ± 0.16 (Fig. 1C). Notably, reintroducing standard food after cuprizone treatment resulted in a markedly elevated and statistically significant rate of weight gain (+3.5 g ± 0.23). These results are consistent with prior studies (Leopold et al., 2019) and confirm that the cuprizone model impacts the animals’ natural ability to gain weight.

Using functional ultrasound imaging, we quantified potential alterations of the hemodynamic response associated with the progression of demyelination and remyelination (5 weeks cuprizone followed by 2 weeks of recovery). To achieve this, we analyzed three descriptors of the brain’s hemodynamic response to stop mechanical whisker stimulation: the number of active pixels, the steady state CBV variation, and the rise time (Fig. 1D, G; Supplementary Fig. 1).

In the control cohort, we evaluated the stability and reproducibility of these measurements over time in naïve untreated mice. Both steady-state CBV variation and rise time remained consistent across all time points (Fig. 2A, B). However, the number of active pixels was stable for the three first time points (DO, 3 W, 5 W), but significantly decreased at the final time point (7 weeks, Fig. 2A, B). This decline in response over time in healthy mice has been reported previously by other researchers (https://doi.org/10.1101/2023.12.21.572595).

To investigate alterations in functional hyperemia during demyelination, we imaged a second cohort (‘progressive demyelination’) at baseline (D0), at early time point of demyelination (3 W), and at maximal demyelination (5 W). Our findings revealed a progressive increase in the number of active pixels (p = 0.03, Fig. 2C, D), indicating a more widespread activation, along with a gradual rise in steady-state CBV within these pixels (Fig. 2C, D). These effects were statistically significant at 5 weeks of cuprizone treatment. Additionally, the rise time was statistically increased at 3 weeks of treatment, suggesting early alterations in neurovascular coupling dynamics.

The effects of remyelinating on the hemodynamic response were investigated on a third cohort of animals (‘Remyelination’), imaged at baseline, at the peak of demyelination (5 W), and after 2 weeks of spontaneous remyelination (following a return to normal food for 2 weeks, ‘5 + 2W’, Fig. 2E). The effects observed after 5 weeks of cuprizone treatment were consistent with those seen in the previous cohort, confirming the reproducibility of our findings across batches of cuprizone-treated animals. Following remyelination, there was a modest reduction in both the number of active pixels and delta CBV (Fig. 2F). However, this effect showed considerable variability between animals, resulting in more dispersed measures (Fig. 2F). As discussed later, this variability in hemodynamic responses is likely attributable to inter-individual differences in the extent of spontaneous remyelination, a phenomenon previously highlighted by Stidworthy et al. (2003).

To determine whether the observed changes were specific to the cortical level, we investigated whether similar effects occurred in the thalamic relay, specifically in the ventro-posterior-lateral and -posterior-medial thalamic nuclei (VPL, VPM). Notably, changes were evident only during the early demyelination period (3 W), when both the average ∆CBV and rise time increased (Fig 3). By 5 weeks of cuprizone intoxication, the hemodynamic response in the thalamus remained unaffected. This suggests that the enhanced cortical response observed at this stage, coinciding with complete demyelination, may represent a unique adaptation or dysfunction specific to cortical circuits.

Finally, to determine whether small demyelinating lesions in the grey or white matter would induce similar alterations, we performed stereotaxic injections of lysolecithin (LPC) either in the S1BF or the internal capsule, with saline injections in the internal capsule serving as controls. Surprisingly, despite the strong and localized demyelination induced by LPC injections (Supplementary Fig. 3A), the evoked hemodynamic response remained unchanged compared to saline-injected animals (Supplementary Fig. 3C).

The level of demyelination was assessed using MBP immunofluorescent staining in several areas of the brain. Two approaches of MBP quantification were compared, yielding consistent results (Fig. 4; Supplementary Fig. 4). After 3 weeks of cuprizone treatment, a stage widely described in the literature as an early phase of demyelination process in this model (Kipp et al., 2017; Matsushima & Morell, 2006; Vega-Riquer et al., 2019), we observed a strong and statistically robust demyelination in both the primary sensory cortex barrel field region (S1BF) and the hippocampus (Figs. 4A, B, E and 4J). Within the cortex, the lamina I exhibited the most pronounced demyelination of all layers (Fig. 4F). Interestingly, at this time point, an unexpected increased MBP immunostaining was observed in the corpus callosum (Fig. 4I).

As anticipated, 5 weeks of cuprizone treatment led to extensive demyelination across several brain regions, including the S1BF (Fig. 4C, E), the medial corpus callosum (Figs 3E and 4C), the hippocampus (Fig. 4C, J), and the thalamus (Fig. 4K). In the cortex, the laminae II-III and V showed severe demyelination (Fig. 4G, H). Transitioning to normal food for 2 weeks induced remyelination in the four aforementioned areas (Fig. 4D–K). Similar results were obtained using an alternative method of MBP quantification (Supplementary Fig. 5).

Interestingly, despite some inter-animal variability of MBP expression across the experimental groups, we could not establish demyelination or remyelination in the internal capsule (Fig. 4L).

To investigate potential direct linear correlation between changes of hemodynamic response and the level of demyelination in the brain of cuprizone-treated animals, we studied and first analyzed the statistical correlations between the three descriptors of the hemodynamic response and the MBP immunostaining. Individual values from all animals included in the study were used. Results are presented in Figure 5A.

In the correlation matrix, positive correlation (yellow/red color coded) and anticorrelation between these metrics (green/blue color code) are shown in the bottom left, while the p-values from the Spearman correlation test (corrected for multiple test comparisons) are displayed on the top right of the matrix. Statistically significant correlations (p-value < 0.05) are indicated in red.

Interestingly, the number of active pixels was strongly and positively correlated with the mean CBV plateau (p-value = 6 10^-3, Fig. 5A). Additionally, the extent of MBP loss in various brain areas demonstrated a strong positive correlation between the MBP expression level in the hippocampus and the thalamus. Similarly, MBP levels in the cortex were strongly and significantly correlated with those in the thalamus (p = 10^-4) and hippocampus (p = 3 10^-7, Fig. 5A).

Overall, changes in MBP expression were anticorrelated with the descriptors of the hemodynamic response, but these anticorrelations were not statistically significant.

To assess the predictive capacity of the hemodynamic response descriptors for determining myelin levels in five brain regions, we used linear regression models (Fig. 5B). For each brain region where MBP was quantified, a linear model was trained using the three descriptors selected through stepwise regression. This process involved the successive removal of the least significant predictors, performed automatically using the MATLAB ‘stepwise’ function.

As shown in Figure 5C, the models indicate associations between the MBP levels and hemodynamic descriptors in all regions except the hippocampus. To identify the most significant predictors, we examined the estimates and corresponding p-value for each predictor in the model (bottom row in the table, Fig. 5B).

This analysis revealed the following: i) The number of active pixels and the rise time together correlated with the MBP content in the S1BF and the thalamus, with the effect being more pronounced in the S1BF and primarily driven by the rise time. ii) The rise time alone correlated with the MBP immunofluorescent level of MBP in the internal capsule. iii) The number of active pixels alone was linked to MBP levels in the corpus callosum.

Notably, the rise estimate was consistently negative across all models (-2.26 for the internal capsule, -1.38 for the thalamus, and -1.57 for the S1BF). This was expected, as myelin loss delays the neural signal, resulting in an increased rise time.

Given the urgent need for non-invasive methods to measure pathophysiological changes in preclinical models of demyelination and the scarcity of studies investigating evoked cerebral hemodynamic responses—often altered in clinical observations—our study leveraged the high sensitivity of fUS imaging to study whisker stimulus-evoked hyperemia in the cuprizone model in mice. Our study not only underscores the potential of using altered hemodynamic responses as a correlate for myelin loss in this model, but also reveals the importance of vascular health in the progression of demyelinating diseases.

To assess the extent of myelin loss across brain regions, we quantified Myelin Basic Protein (MBP) via immunofluorescent staining, a surrogate for myelin content. More accurate techniques like electron microscopy or spectral confocal reflectance imaging (Gonsalvez et al., 2019) exist, but unfortunately they could not be used within the framework of this study. MBP staining still yielded valuable insights. Most studies focus on the medial corpus callosum (Gudi et al., 2009; Matsushima & Morell, 2006; Vega-Riquer et al., 2019), but we observed widespread demyelination with variability between regions (Goldberg et al., 2015). The superficial cortical laminae and the hippocampus showed MBP loss after just 3 weeks, while the thalamus and cortical layers II-V were affected only after 5 weeks, consistent with prior research (Goldberg et al., 2015; Gudi et al., 2009; Norkute et al., 2009; Silvestroff et al., 2010; Skripuletz et al., 2008; Yang et al., 2009). Remyelination patterns also varied, with complete recovery in the corpus callosum but incomplete and inconsistent recovery in the cortex, thalamus, and hippocampus.

As sensory afferents from the whiskers go through the trigeminal SpV nucleus to the thalamic relay (Wimmer et al., 2010), and then the internal capsule to finally contact cortical neurons (Petersen, 2007), demyelination at any point along this pathway could underlie the delayed hemodynamic response seen in cuprizone-treated animals. Indeed, our results confirm that the abnormal hemodynamic response in the S1BF linked with the loss of MBP in these three neuroanatomical structures (thalamus, internal capsule, and S1BF). Interestingly, the descriptor significantly linked with the MBP alteration in these structures was the rise time. This increased rise time suggests a delay in the hemodynamic response and relates to the latency of evoked potentials, which is a measure classically used in the diagnosis of MS (Compston & Coles, 2008) and was previously demonstrated to be a preclinical marker of remyelination in pre-clinical models of MS (Cordano et al., 2022; Marenna et al., 2022). Both the increased rise time and evoked potentials latency are due to myelin damage and the subsequent reduction in neuronal conduction.

Notably, cuprizone induced only modest alterations in the thalamic hemodynamic response. One observed change—an increased rise time at the early stage of treatment—mirrored that seen in the cortex. Given the thalamus’s upstream position relative to the cortex, the delayed cortical response may therefore stem from delays originating in the thalamic relay.

The cuprizone model, widely used to study demyelination (Vega-Riquer et al., 2019), induces chronic, non-immune-mediated demyelination of the brain through oligodendrocyte degeneration. To explore changes in evoked response across different phases of demyelination or remyelination, we conducted a longitudinal study where each animal serves as its own control. Our findings reveal that after 5 weeks of cuprizone treatment, the whisker stimulus-evoked hyperemia was consistently increased. This was characterized by a larger activated area (more active pixels) and enhanced activation within this region (greater CBV over time).

These findings are highly relevant to the clinical condition (Rocca et al., 2005), as in MS patients with clinically isolated syndrome (CIS; Rico et al., 2011; Rocca et al., 2003), with non-disabling (Rocca et al., 2002) or mildly disabling relapsing-remitting (RR) MS (Reddy et al., 2002), exhibit increased activation of the primary sensorimotor cortex during motor tasks. These enhanced responses (see for review Rocca et al., 2022) have been considered to play a compensatory role in RRMS, since it was related to maintaining a good task performance despite the presence of widespread structural damage, and are believed to be due to both altered cerebral blood flow and oxygen consumption (Hubbard et al., 2017; West et al., 2021).

This interesting parallel between our results and the clinical condition suggests that even though the cuprizone model does not model all hallmarks of the clinical condition, it modeled well its altered hemodynamic responsiveness.

Several authors have hypothesized that these surprising, enhanced responses might be an adaptive plasticity to compensate for the growing disability of the diseased brain. The cellular and molecular mechanisms behind this plasticity are still largely unknown. However, a large body of literature describes an alteration of the vascular system and of the neurovascular unit in MS (Caprio & Russo, 2016), opening the debate on whether vascular events may be the primary cause of neurological diseases or rather a mere participant recruited from a primary neuronal origin. Our findings suggest that demyelination itself, caused by oligodendrocyte disruption, can promote neurovascular coupling alterations. Importantly, the absence of hemodynamic changes in models with localized gray or white matter demyelination (e.g., LPC-induced lesions) indicates that local neuroinflammation and demyelination alone are insufficient to induce these changes. Given that oligodendrocyte disruption is a hallmark of the cuprizone model, it is plausible that these vascular dysfunctions stem from oligodendrocytic origins. Further research is required to validate this hypothesis.

One of the limitations of our study is the lack of investigation of the extent of this neurovascular alteration at later stages (10–12 weeks of cuprizone treatment), which mimic aspects of advanced MS characterized by axonal damage. Examining functional responses at these time points could determine whether, as in MS patients, responses decrease during progressive disease stages. Future work will focus on this critical question.

Interestingly, the amplitude of the response (CBV) alone was not associated with the extent of MBP staining, either independently or in combination with other factors. Given that these changes closely mimic those observed in the early phases of multiple sclerosis (MS) (Rocca et al., 2022), we hypothesize that this measurement is a critical indicator of the vascular dysfunction inherent to the model. As highlighted by various studies (Caprio & Russo, 2016; Cashion et al., 2023; Sweeney et al., 2018), MS patients frequently experience vascular complications, such as blood-brain barrier leakage (Sweeney et al., 2019), microbleeds, and reduced cerebral blood flow. A growing body of evidence suggests that the dysfunction of various cells within the neurovascular unit—such as endothelial cells (Zlokovic, 2008), astrocytes (Correale & Farez, 2015), pericytes (Zhang et al., 2020), and microglial cells (Lassmann et al., 2012)—plays a significant role in MS pathology (Ballabh et al., 2004; Cashion et al., 2023; Sweeney et al., 2019). Therefore, targeting the stabilization of the vasculature by blocking the pathological changes affecting these cells represents a promising therapeutic avenue in MS (see review by Cashion et al., 2023).

In conclusion, our study highlights the importance of vascular health in the progression of demyelinating diseases. Future research should focus on further characterizing these vascular dysfunctions over longer time frames and exploring therapeutic strategies aimed at preserving neurovascular integrity. By advancing our understanding of the interplay between demyelination and vascular dysfunction, we hope to contribute to the development of more effective diagnostic tools and treatments for MS and related neurological conditions.

Source data are available on the repository website zenodo using the following link: https://zenodo.org/records/13136108. Custom codes used for the analysis of fUS data used in this study are protected by INSERM. The rest of this work did not involve any other particular code.

S.P., B.B., M.T., and D.P.B. designed the experimental paradigm. N.I.R. and B.B. performed the ultrasound experiments, and B.B. processed the ultrasound data. T.D. and M.T. supervised the signal processing of the ultrasound data. Y.T.-J. performed the immunohistochemical experiments, while L.D. and A.R.L.G. performed their acquisition and quantifications. B.B. and S.P. wrote the manuscript. B.B. assessed and verified the data, performed the statistical analysis, and wrote the corresponding section in the manuscript. B.B., S.P., and M.T. were involved in the interpretation of the data and wrote some parts of the manuscript. All authors read and corrected the final version of the manuscript.

M.T. and T.D. are co-founders and shareholders of Iconeus company. M.T. is co-inventor of several patents in the field of neurofunctional ultrasound and ultrafast ultrasound. M.T. and T.D. do not have any other financial conflict of interest, nor any non-financial conflict of interests. All the other authors do not have any financial or non-financial conflict of interests.

This work was supported by Biogen (B. Beliard’s PhD fellowship), the ANRT (B. Beliard’s PhD fellowship), the Institut National de la Santé et de la Recherche Médicale, Biogen, and the European Research Council (ERC) Advanced Grant FUSIMAGINE and the ESPCI. The authors would like to thank Mrs L. Antoinette for animal husbandry. Finally, this work was supported by the Inserm ART (Technology Research Accelerator) in “Biomedical Ultrasound”. Several panels of figures were made using Biorender (Biorender.com).

Supplementary material for this article is available with the online version here: https://doi.org/10.1162/imag_a_00534.