27‐hydroxycholesterol promotes oligodendrocyte maturation: Implications for hypercholesterolemia‐associated brain white matter changes

Oxidized cholesterol metabolite 27‐hydroxycholesterol (27‐OH) is a potential link between hypercholesterolemia and neurodegenerative diseases since unlike peripheral cholesterol, 27‐OH is transported across the blood–brain barrier. However, the effects of high 27‐OH levels on oligodendrocyte function remain unexplored. We hypothesize that during hypercholesterolemia 27‐OH may impact oligodendrocytes and myelin and thus contribute to the disconnection of neural networks in neurodegenerative diseases. To test this idea, we first investigated the effects of 27‐OH in cultured oligodendrocytes and found that it induces cell death of immature O4+/GalC+ oligodendrocytes along with stimulating differentiation of PDGFR+ oligodendrocyte progenitors (OPCs). Next, transgenic mice with increased systemic 27‐OH levels (Cyp27Tg) underwent behavioral testing and their brains were immunohistochemically stained and lysed for immunoblotting. Chronic exposure to 27‐OH in mice resulted in increased myelin basic protein (MBP) but not 2′,3′‐cyclic‐nucleotide 3′‐phosphodiesterase (CNPase) or myelin oligodendrocyte glycoprotein (MOG) levels in the corpus callosum and cerebral cortex. Intriguingly, we also found impairment of spatial learning suggesting that subtle changes in myelinated axons of vulnerable areas like the hippocampus caused by 27‐OH may contribute to impaired cognition. Finally, we found that 27‐OH levels in cerebrospinal fluid from memory clinic patients were associated with levels of the myelination regulating CNPase, independently of Alzheimer's disease markers. Thus, 27‐OH promotes OPC differentiation and is toxic to immature oligodendrocytes as well as it subtly alters myelin by targeting oligodendroglia. Taken together, these data indicate that hypercholesterolemia‐derived higher 27‐OH levels change the oligodendrocytic capacity for appropriate myelin remodeling which is a crucial factor in neurodegeneration and aging.


| INTRODUCTION
Factors related to lifestyle play a pivotal role in increasing the risk of developing neurodegenerative disorders (Lisko et al., 2020). Midlife hypercholesterolemia has been linked to both late-and early-onset Alzheimer's disease (AD) (Anstey et al., 2017;Wingo et al., 2019), and hypercholesterolemia is also one of the risk factors for vascular dementia (Appleton et al., 2017). The role of cholesterol metabolism in neurodegeneration is further corroborated by the association between familial hypercholesterolemia and cognitive deficits (Zamb on et al., 2010). Nonetheless, cholesterol is unable to penetrate the blood-brain barrier (BBB) and therefore the brain synthesizes cholesterol de novo (Jurevics & Morell, 1995). Yet there are secondary vascular factors linking high systemic cholesterol to neurodegenerative changes, such as atherosclerosis and lesions of the white matter (de Bem et al., 2021;Tong et al., 2019). A more direct link between high systemic cholesterol and brain changes may be oxysterols, which are metabolites of cholesterol emerging in research as biologically active compounds. Unlike cholesterol, oxysterols can cross the BBB (Björkhem et al., 1998;Heverin et al., 2005).
The peripherally abundant 27-hydroxycholesterol (27-OH) has been associated with AD (reviewed in Björkhem et al., 2006;Loera-Valencia et al., 2019). 27-OH is produced by the enzyme sterol 27-hydroxylase (CYP27A1) that is expressed throughout the body (Heverin et al., 2005;Uhlén et al., 2015) and the majority of 27-OH in the brain has its origin outside the central nervous system (Heverin et al., 2005). 27-OH fluxes into the brain in a concentration-dependent manner where the concentrations are higher in white than gray matter (Heverin et al., 2005).
27-OH concentrations are relatively low overall, but both brain (Heverin et al., 2004) and cerebrospinal fluid (CSF) (Leoni et al., 2004;Mateos et al., 2011) levels are increased in neurodegenerative diseases compared to healthy controls. In elderly persons at risk of dementia who participated in the lifestyle randomized controlled trial FINGER, higher 27-OH was associated with lower hippocampal volume and cognition (Sandebring-Matton et al., 2021). In preclinical models, the effects of excess brain levels of 27-OH have been linked to reduced dendritic spine density in the hippocampus , decreased neuronal glucose uptake, and impaired memory (Ismail et al., 2017). Further, inflammatory responses by 27-OH including astrocytic activation have been shown (Loera-Valencia et al., 2021;Staurenghi et al., 2021;Testa et al., 2014).
White matter changes are a key factor in lipid dyshomeostasisassociated neurodegenerative diseases and dementia (Bartzokis, 2011;Nasrabady et al., 2018). Changes in white matter integrity emerge in the early stages of neurodegeneration (Lee et al., 2016) which may provide a window of opportunity for new treatment strategies. For instance, in studies conducted on asymptomatic subjects with an increased risk of developing AD, a decline in white matter integrity has been associated with pathological AD biomarker levels (Dean et al., 2017). Despite these discovered associations, the role of myelin and oligodendrocytes in neurodegenerative progress remains relatively poorly studied.
27-OH reduces cholesterol synthesis in vitro and in vivo Meir et al., 2002), which raises concerns regarding proper myelin assembly and maintenance for which adequate cholesterol levels are essential. Still, to our knowledge, how 27-OH affects oligodendrocytes and myelination-related processes remains unknown.
Here, we investigated whether 27-OH influences oligodendrocyte populations and myelin integrity. First, the impact of 27-OH treatment on cell differentiation in cultured oligodendrocytes was studied, whereafter oligodendrocyte gene expression was measured in primary co-cultures. Second, learning and memory of female mice overexpressing sterol 27-hydroxylase (Cyp27Tg) and age-matched nontransgenic littermates (TgÀ) were assessed with a battery of cognitive tests, and brain sections from the mice were stained for markers of immature and mature oligodendrocytes, and myelin content. Lastly, associations between 27-OH and myelination proteins in CSF samples from memory clinic patients were analyzed.

| Primary rat oligodendrocyte cultures
For viability assay, oligodendrocytes were obtained from optic nerves of 12-days-old Sprague Dawley rats, as described previously (Barres et al., 1992). Cells were maintained at 37 C and 5% CO 2 in a chemically
Cyp27Tg mice have normal brain cholesterol levels but a slight upregulation of cholesterol synthesis. Cyp27Tg mice have 5-6 times higher 27-OH levels compared to WT (Ismail et al., 2017). Serum levels of 27-OH in Cyp27Tg mice correspond to a concentration of approximately 0.75 μM , and brain concentration of 27-OH ranges approximately from 1 to 6 ng/mg both in humans and mice Heverin et al., 2004;Testa et al., 2016). For this study, 8-month-old female Cyp27tg mice and their age-matched nontransgenic littermates (TgÀ) were used. The mice were fed normal chow and water was provided ad libitum. For behavioral analyses we studied two separate cohorts: for Morris Water Maze (MWM), we tested 10 transgenic and 10 non-transgenic mice, and for Y-maze and Fear Conditioning (FC) nine transgenic and five non-transgenic mice.
Brain tissues from Cyp27Tg and Tg-were dissected and used for biochemical analyses.

| Behavioral tests
For MWM, each mouse was tested for four trials per day, for five consecutive days. Reference memory was evaluated on the sixth day (probe test). MWM, Y-maze, and the FC tests were performed as described previously (Eroli et al., 2020;Maioli et al., 2013) except that for FC the conditional stimulus of sound was on day one where mice were exposed to a 55 dB sound at 5000 Hz for 30 s followed by a mild foot shock (0.3 mA for 2 s). The sound-shock pairing was repeated three times in total with a 50-s interval between each one. Cue FC was performed on day three, where mice were let to explore the new surrounding for 2 min, whereafter the sound (55 dB at 5000 Hz) lasted for 2 min continuously. The rectangular-shaped chamber was replaced with a round-shaped chamber (20 cm diameter Â 35 cm high), and the stainless-steel grid floor was replaced with a black flat surface. Instead of wiping the chamber with ethanol, the chambers were cleaned between each mouse with hypochlorous water (50% dil). Freezing behavior was defined as the complete absence of mobility within the same area for 2 s or longer and was measured through TSE Multi Conditioning software. The freezing % recorded during the habituation phase of day 1 (as a measure of baseline freezing) was compared to the freezing % of day 2 to evaluate the context memory. To assess cue memory, we measured the freezing % on day 3 before and during the sound stimulus. Blinded analysis of LFB stained sections was performed using ImageJ 1.52p (Schindelin et al., 2012) as described previously (Khodanovich et al., 2017;Underhill et al., 2011). The mean intensity of each brain section was measured from the corpus callosum and fimbria. Mean    intervertebral space using a 25-gauge needle. CSF was aliquoted in polypropylene tubes and stored at À80 C until further analysis. CSF Aβ 42 , t-Tau, and p-Tau concentrations were measured on fresh samples with commercially available sandwich enzyme-linked immunosorbent assays (Innogenetics, Ghent, Belgium) according to standardized protocols in the clinic. 27-OH levels were quantified in cerebrospinal fluid using isotope dilution mass spectrometry as described previously (Dzeletovic et al., 1995). In all analyses, a p value of <.05 was considered significant. The number of immature oligodendrocytes Olig2 + O4 À MBP À and Olig2 + O4 + MBP À were similar between 27-OH and vehicle treatments (p = .839 and p = .296, respectively).

| 27-OH alters oligodendrocyte gene expression
We next sought to determine the effects on myelin gene expression from 27-OH in mouse co-cultures. Previous data have shown that Olig2 expression is reduced as the oligodendroglial lineage becomes differentiated (Marques et al., 2016). In our cultures in control conditions, Olig2 expression maintains a basal level from DIV3 to DIV5 but decreases by DIV10 (mean difference 0.7 ± 0.2, p = .003, Figure 1d); whereas 27-OH treated cultures show decreased Olig2 expression at DIV3 and DIV5 but not at DIV10 compared to controls (fold change mean differences 0.7 ± 0.2, p = .004, 1.2 ± 0.2 p < .0001, and 0.1 ± 0.2, p = .797, Figure 1d). An opposing expression pattern of the marker of myelinating oligodendrocytes, Cc1, was observed in the same experimental conditions (Figure 1e). Cc1 expression exhibited a difference of 5.0 ± 0.8 in means between 27-OH and vehicle-treated cultures on DIV3 (p < .0001). The difference remained significant at DIV5, yet the magnitude of the difference had decreased to 3.0 ± 0.7 (p = .0005). However, by DIV10 the expression level in 27-OH treated cells had attenuated to a level significantly below the control (mean difference 1.5 ± 0.7, p = .042). The expression of Cc1 was significantly different between DIV5 and 10 in vehicle-treated cultures (1.6 ± 0.7, p = .032). The levels were otherwise stable throughout time points. In 27-OH treated cultures, there was a major reduction in expression from DIV3 to DIV5 (mean difference 2.4 ± 0.8, p = .005), and further from DIV5 to DIV10 (mean difference 2.9 ± 0.7, p = .0007).
Next, we compared the expression of OPC marker Pdgfrα between 27-OH treated and untreated cells. Pdgfrα was reduced by 27-OH treatment at all studied time points (Figure 1f). At DIV3, the reduction in fold change means was 0.8 ± 0.3 (p = .007) on DIV5 the reduction attenuated to 0.6 ± 0.2 (p = .012) yet remained until DIV10 (0.7 ± 0.2, p = .007). Altogether, our results show that 27-OH changes the dynamics of oligodendrocyte maturation towards a more mature state; a decrease in Olig2 and Pdgfrα expression, and an increase in Cc1 expression that promote oligodendrocyte differentiation.

| Elevated levels of 27-OH result in impairment of spatial learning in female Cyp27 transgenic mice
To mimic long-term hypercholesterolemia, we further investigated the chronic effects of 27-OH on myelin and cognition. As a first step, we In Y-maze, there was no significant difference between the percentage in alternation between arms (Cyp27Tg 58.5%, TgÀ 62.5%, F I G U R E 2 Female transgenic mice overexpressing human CYP27A1 (Cyp27Tg) have impaired spatial learning compared to controls. Behavioral testing was performed in TgÀ and Cyp27Tg mice. (f) graph shows the percentage of freezing time measured on day 1 (habituation phase) vs day 2 (context test); (g) graph expresses the freezing percentage measured before vs during the cue stimulus (sound). Statistical evaluation was performed using repeated-measures ANOVA with Bonferroni correction in (a), Mann-Whitney U test in (b) unpaired Student's two-tailed t test in (c-e) and Mixed model with Bonferroni correction in (f,g). All data are presented as mean ± SD, except in (a) where presented as mean ± SEM, *p < .05; **p < .01; ***p < .001. p = 0.613, Figure 2d). Neither was there a significant difference between the number of entries into the different arms between strains (Cyp27Tg 20.1, TgÀ 14.2, p = .063, Figure 2e). On the first day of FC mice were exposed to a high-frequency sound and an electric shock after the sound. On the second day, the floor was kept as on day 1 (context), but mice were not exposed to sound or electrical F I G U R E 3 Cyp27Tg mice have preserved brain myelination and numbers of immature oligodendrocytes but increased mature oligodendrocytes in the cortex. (a) Brain sections were stained with Luxol fast blue, and areas of corpus callosum and fimbria were analyzed for intensity differences between Cyp27Tg and TgÀ mice. Scale bars represent 100 μm. Intensity difference in (b) corpus callosum and (c) fimbria.
(d) Representative images of different brain regions from Cyp27Tg and TgÀ mice. Brains were stained for PDGFRα (platelet-derived growth factor receptor α, green), a marker for immature oligodendrocytes, and NeuN, a marker for neurons (red).
White shock. In both strains, the time freezing increased significantly on the second day (Cyp27Tg 29.9%, TgÀ 41.9%, for both p < .0001, Figure 2f). However, the amount of freezing during the second day was less for Cyp27Tg than for TgÀ mice, yet the difference was not significant (p = .074). On the third day, mice were exposed to the same sound (cue) as on the first day on a new floor (change of context). Both strains increased their freezing time significantly during the sound compared to the time before the sound (Cyp27Tg 47.2% p < .0001, TgÀ 37.3% p = .001, Figure 2g). There was no significant difference between strains in freezing time before sound (p = .103) nor during sound (p = .782).

| Chronic exposure to 27-OH in vivo does not alter myelin overall structure but increases MBP levels
To assess the overall myelin status in the TgÀ and Cyp27Tg mice, In myelin-enriched fractions of brain homogenates, MBP levels were increased in Cyp27Tg mice (p = .030 Figure 5a,b). Conversely, levels of MOG and proteolipid protein (PLP) were not significantly different between lines (p = .059, Figure 5a,c and p = .076, Figure 5a,d).
F I G U R E 5 Western blot analysis of myelin proteins in the mouse brain. (a) Immunoblots of the different myelin proteins in myelin enriched fractions from brain homogenates from Cyp27Tg and TgÀ mice. Quantification of (b) myelin basic protein (MBP), (c) myelin oligodendrocyte glycoprotein (MOG), and (d) proteolipid protein (PLP) levels in myelin enriched fractions. Myelin protein levels were normalized to β-actin. Statistical evaluation was performed using unpaired Student's two-tailed t test in (b,d) and Mann-Whitney U test two-tailed in (c). All data presented as mean ± SD, *p < .05.
T A B L E 1 Baseline characteristics of the memory clinic cohort.  Table 2. We did not observe any statistically significant associations between MBP and 27-OH in any of the models suggesting that myelin is neither degraded nor released from the brain in a 27-OH concentration dependent manner. The myelination regulating enzyme CNPase was instead associated significantly with 27-OH in age and sex-adjusted Model 1 (β = 0.38, p = .022) and the significance remained after further adjustment for the AD-related pathologies: Aβ 42 , t-Tau, and p-Tau levels (Model 2, 3, and 4 respectively). The results did not change even if models were further adjusted for cognition (data not shown). As expected, the CSF levels of myelin proteins MBP and CNPase were associated with each other (Model 1: β = 0.70, p < .0001). The myelin protein levels did not associate with the levels of AD biomarkers, except for MBP which was associated with Aβ 42 (β = 0.32, p = .034). Neither of the myelin proteins was associated with cognition as measured by RAVLT or MMSE when adjusted for age, sex, and education.

| DISCUSSION
Cholesterol and its metabolites are fundamental in the physiology of the nervous system and consequently, peripheral hypercholesterolemia is a condition that has been associated with brain disorders and cognitive deficits (Anstey et al., 2008). Changes in the white matter are an early phenomenon in cognitive disorders and are linked to vascular factors including high cholesterol (Murray et al., 2005).  Popko et al., 1987;Saher et al., 2005). Further, altering MBP may affect the intricate interplay between a proper lipid composition and protein profile that is crucial for myelin homeostasis and function (Krugmann et al., 2020). Whether these 27-OH-induced MBP changes are altering overall myelin structure and function remain to be determined, yet our LFB staining did not indicate any major structural changes in the Cyp27Tg mice suggesting that the changes are subtle.
Previous studies show that male Cyp27Tg mice exhibit deficits in memory at the age of 8-9 months along with impaired neuronal morphology and function (Ismail et al., 2017;Merino-Serrais et al., 2019).
In mice modeling cerebrovascular pathology, high-cholesterol-diet induced white matter changes and reduced numbers of immature oligodendrocytes, accompanied by cognitive deficits (Tong et al., 2019).
In our study, female Cyp27Tg mice displayed impairment in learning in MWM at the age of 8 months, but not in Y-maze and FC tests, T A B L E 2 Association between myelin proteins and 27-OH in cerebrospinal fluid. Note: Values are standardized β-coefficients with p values from linear regression models. 27-OH concentration was used as the independent variable. Model 1 was adjusted for age and sex. Model 2 was adjusted for age, sex, and Aβ 42 . Model 3 was adjusted for age, sex, and t-Tau. Model 4 was adjusted for age, sex, and p-Tau. p values are considered significant (bold) if <.05. Abbreviations: 27-OH, 27-hydroxycholesterol; Aβ 42 , amyloid-beta; CNPase, 2 0 , 3 0 -cyclic-nucleotide 3 0phosphodiesterase; MBP, myelin basic protein; p-Tau, phospho-tau; T-tau, total tau. indicating sex differences in spatial learning. The findings of the current study unveil an effect of 27-OH in oligodendroglia and myelin which act in concert with previously described 27-OH-induced changes (Loera-Valencia et al., 2019) ultimately resulting in impaired cognition.
Myelin proteins are released into CSF when the white and gray matter is damaged (Sellebjerg et al., 2017). In post-mortem brains of AD patients, there is a depletion of cholesterol along with a reduction in MBP and CNPase (Roher et al., 2002). Moreover, 27-OH levels in CSF correlate with brain structural damage in diseases including demyelinating polyneuropathy and AD (Leoni et al., 2004;Mateos et al., 2011), emphasizing a plausible connection between this oxysterol in CSF and demyelination and/or neuronal damage. Notably, increased CSF 27-OH levels correlate with white matter hyperintensities in a memory clinic cohort (Besga et al., 2012) and in cases of type 5 spastic paraplegia, a disease that results in elevated 27-OH levels due to mutations in the CYP7B1 gene that is responsible for catabolizing excess 27-OH (Biancheri et al., 2009). Interestingly, in CSF from a memory clinic cohort we found a positive association between CNPase and 27-OH that was independent of clinical AD biomarker levels. Based on the available literature (Raasakka & Kursula, 2014) CNPase is important for the formation of myelin membranes and in maintaining axonal integrity. Whether the associations between 27-OH and CNPase in CSF reflect oligodendrocyte maturation in the human brain should be further investigated in larger cohorts of individuals with elevated 27-OH levels.
In sum, we found that 27-OH is toxic to immature oligodendrocytes while favoring their OPC maturation and modifying protein levels in myelin. These features may underlie part of the effects of hypercholesterolemia on brain function, yet the clinical implications of these findings should be further evaluated.