Different effects of CYP27A1 and CYP7B1 on cognitive function: Two mouse models in comparison

The oxysterol 27-hydroxycholesterol (27OHC) is produced by the enzyme sterol 27-hydroxylase (Cyp27A1) and is mainly catabolized to 7 α -Hydroxy-3-oxo-4-cholestenoic acid (7-HOCA) by the enzyme cytochrome P-450 oxysterol 7 α -hydroxylase (Cyp7B1). 27OHC is mostly produced in the liver and can reach the brain by crossing the blood-brain barrier. A large body of evidence shows that CYP27A1 overexpression and high levels of 27OHC have a detrimental effect on the brain, causing cognitive and synaptic dysfunction together with a decrease in glucose uptake in mice. In this work, we analyzed two mouse models with high levels of 27OHC: Cyp7B1 knock-out mice and CYP27A1 overexpressing mice. Despite the accumulation of 27OHC in both models, Cyp7B1 knock-out mice maintained intact learning and memory capacities, neuronal morphology, and brain glucose uptake over time. Neurons treated with the Cyp7B1 metabolite 7-HOCA did not show changes in synaptic genes and 27OHC-treated Cyp7B1 knock-out neurons could not counteract 27OHC detrimental effects. This suggests that 7-HOCA and Cyp7B1 deletion in neurons do not mediate the neuroprotective effects observed in Cyp7B1 knock-out animals. RNA-seq of neuronal nuclei sorted from Cyp7B1 knock-out brains revealed upregulation of genes likely to confer neuroprotection to these animals. Differently from Cyp7B1 knock-out mice, transcriptomic data from CYP27A1 overexpressing neurons showed significant downregulation of genes associated with synaptic function and several metabolic processes. Our results suggest that the differences observed in the two models may be mediated by the higher levels of Cyp7B1 substrates such as 25-hydroxycholesterol and 3 β -Adiol in the knock-out mice and that CYP27A1 overexpressing mice may be a more suitable model for studying 27-OHC-specific signaling. We believe that future studies on Cyp7B1 and Cyp27A1 will contribute to a better understanding of the pathogenic mechanisms of neurodegenerative diseases like Alzheimer ’ s disease and may lead to potential new therapeutic approaches.

Despite the evidence demonstrating the impact of blood cholesterol in the development of AD, both cholesterol and the lipoproteins responsible for its transport are unable to cross the blood-brain barrier (BBB) [6].This implies that brain cholesterol levels are independent of peripheral cholesterol, which suggests that there are other factors implicated in this interplay [7,8].Unlike cholesterol, oxidized cholesterol metabolites known as oxysterols, are early intermediates in the metabolism of cholesterol to bile acids and can cross the BBB from both directions [9].27-hydroxycholesterol (27-OHC) is an oxysterol produced from cholesterol by the activity of the enzyme sterol 27-hydroxylase (Cyp27A1) [10].Higher levels of 27-OHC are found in AD patients' brains and cerebrospinal fluid (CSF) [11,12].Mice overexpressing human CYP27A1 (Cyp27Tg) produce approximately 6-fold higher levels of 27-OHC than control mice in serum and the brain [13,14].Previous work from our lab has shown that Cyp27Tg mice have reduced brain glucose uptake and impaired spatial memory [14], accompanied by diminished dendritic arborization and spine density.In the hippocampus, they also show lower activity-regulated cytoskeleton-associated protein (Arc), and postsynaptic density 95 (PSD95) [15] as well as increased levels of S100A8 alarmin and its receptor RAGE, involved in inflammatory responses [16].
Given the similarities between the CYP27A1 overexpressing and the Cyp7B1KO mice with regards to 27-OHC accumulation in the brain, in this study, we evaluate the effects of Cyp7B1 deletion on cognitive functions and gene expression in vivo and in vitro.To investigate the differences between Cyp27Tg and CYP7B1KO mice we compare the transcriptomic profiles of neurons sorted from cortical regions.By using these two mouse models we ultimately aim to advance our knowledge of the role of brain cholesterol metabolism in health and disease.

Animals
The generation and breeding of Cyp7B1KO and Cyp27Tg mice were previously described in [25] and [26] respectively.[27]All the mice used in this study were males, with a C57BL/6 J strain background.Wild-type littermates were used as controls for every experiment using genetically modified mice.All mice were kept under controlled conditions of temperature (21 ± 1 • C) and humidity (55 ± 5%) on a 12 h light-dark cycle.Food and water were provided ad libitum.Mice were sacrificed by decapitation and their brains were divided by the two hemispheres.The right hemisphere cortex and hippocampus from Cyp7B1KO and control mice were dissected, frozen, and stored at − 80 • C, while the left was directly placed in a solution for Golgi staining.All experiments were run in compliance with the local national animal care and guidelines and approved by the local committee of Karolinska Institutet and the Swedish Board of Agriculture.All possible efforts were made to reduce any suffering or distress to the animals.
Neuronal cultures were then treated with 1 μM 27-OHC or vehicle control daily from DIV 5 to DIV 8 and collected for analysis on DIV 10.The knockdown efficiency was evaluated by measuring Cyp7B1 mRNA levels.

RNA isolation and Real-time quantitative PCR
RNA isolation as well as the reverse transcription and quantitative PCR (RT-qPCR) was performed using the TaqMan Fast Advanced Cell-to-Ct kit (#A35377, Thermo Fisher) following the manufacturer's instructions.Lysates were reverse transcribed in a SimpliAmp Thermal cycler (Life Technologies).Selected genes were quantified using specific primers for Snap25, Dlg4, Cyp7B1, and Gapdh (Life Technologies) in a 7500 Fast Real-Time PCR system (Applied Biosystem, Life Technologies).Gapdh was used as the housekeeping gene.

Mouse 18 F-FDG PET imaging
7 months-old Cyp7B1KO and wild-type mice were anesthetized as previously described [14].Briefly, mice were initially anesthetized by inhalation of 5% isofluorane in 100% oxygen.Once the anesthesia was induced, the isofluorane concentration was reduced to 1.5-2% in a 50:50 air/oxygen mixture.Under these conditions, the 18 F-FDG tracer was injected through a cannula from the tail vein.The injected dose of 18 F-FDG was based on the respective mice weights.The mice were under anesthesia for 30 min.After injection, the mice were allowed to regain consciousness and were free to move for 45 min in a heated cage.Thereafter, animals were re-anesthetized to proceed with the 40 min-long PET scan.
The PET scan was performed in a nanoScan PET/MRI (Mediso Ltd.), exhibiting a 700-μm PET and 100-μm MRI in-plane spatial resolution.Details on the reconstruction method have been previously reported [14,29,30].40-minute list mode data from the PET scans were reconstructed into a one-time frame and corrected for attenuation, scatter, and random coincidences.The image reconstruction protocol consisted of a 400-600-keV energy window and a 5-ns time window, using a 20-iteration maximum-likelihood expectation-maximization algorithm and a 10 − 4 regularization parameter, with a final isotropic resolution of 0.3 × 0.3 × 0.3 mm voxel size.
MRI images were sequentially acquired after the PET scans and were spatially transformed to match the reconstructed PET images.The specific MRI scan parameters were previously reported (Ismail., 2017).In short, the parameters used were FSR 3D mouse (20-min scan), with 2000/80 ms (TR/TE, respectively) and 130/130 frequency/phase. 18F-FDG uptake was expressed as a percentage of SUV (%SUV).The analysis of all reconstructed images was done using PMOD Software (PMOD Technologies) and power analysis using G*Power (3.1.7).PET experiments were performed following the approval and consent of the local ethical committee of Karolinska institutet, Stockholm.

Behavioral studies
7 months old CYP7B1KO male mice and their WT littermates were assessed for Morris Water Maze (MWM) test.Afterward, 14 months old CYP7B1KO male mice and WT littermates underwent Elevated Plus Maze (EPM), Y-maze and Barnes Maze tests.The order of the tests started from the least stressful test to the most stressful one as follows: EPM, Y Maze, Barnes Maze [32].Mice were habituated to the experimental room for 45 min before each test.A detailed description of the Y-maze EPM apparatus and experimental protocols is reported in our previous study [33], while the Morris Water Maze settings are described in [34].In older mice, spatial learning and memory were assessed in a modified protocol of the Barnes maze [35].The Barnes maze consisted of a white, circular platform (90 cm in diameter) elevated 90 cm above the floor (Ugo Basile, Italy).20 evenly spaced circular holes (5 cm in diameter) were located 2.5 cm from the edge, and each hole could be connected either to a false hole (a black lid, 5 cm × 5 cm; diameter × depth) or a target hole (a white L-shaped escape tunnel, 5 cm × 5 cm × 7 cm; diameter × depth × length).Each session began by placing the mouse inside a start cylinder that was placed at the center of the maze.After five seconds, the trial started by lifting the start cylinder and allowing the mouse to freely explore the maze, and the trial ended when the mouse entered the target hole or after 2 min had elapsed.Once the mouse climbed into the target hole, the mouse was returned to its home cage lid immediately and rested there for 30 s.If a mouse did not find or enter the target hole within 2 min it was manually turned on and the mouse was gently pushed to the target hole by the experimenter.After each trial, the maze and the target tunnel were cleaned with ethanol (70%) to remove odor cues.During the acquisition phase (days 1-4), mice were given four trials per day with an inter-trial interval of 1 min for 4 consecutive days.A retention probe test, consisting of a single trial for 2 min without any target hole, was performed 24 h after the last acquisition trial.In all tests, data were acquired with a camera installed above the apparatus, connected to the video-tracking software Ethovision XT 14 (Noldus Information Technology, The Netherlands).

Golgi Staining and dendritic spine analysis
After behavior, left brain hemispheres from 14-month-old CYP7B1KO mice were randomly selected and stained with the Golgi-Cox solution using the FD Rapid Golgi Stain™ Kit (Neuro-Technologies, US) for Golgi analysis.The kit was used according to the manufacturer's instructions and the staining was performed on 150 µm thick coronal sections.Images were acquired with a light microscope (Nikon Eclipse E800) with 100x objective, oil immersion (NA: 1.30).Eight slices per brain containing the whole hippocampus were selected for the analysis.Images were collected from the stratum radiatum in the CA1 hippocampal region.We selected collateral dendrites (stratum radiatum) from the pyramidal neurons and analyzed dendrites from two-three orders for a total dendritic length of 300 µm/animal.Cells for analyses were selected based on the following criteria: a) neurons relatively complete (3 orders or greater of dendrites were visible); b) neurons fully impregnated with staining; and c) minimal or no overlap with other labeled neurons.The analyses were conducted blinded to genotype and performed using ImageJ software (NIH, US).

Neuronal nuclei isolation and sorting
The nuclei isolation protocol was performed as previously reported [36] with certain modifications.Mouse brain tissue covered in aluminum foil was kept on dry ice and smashed several times until the tissue was powdered.Powdered tissue was then transferred into a tube containing 5 ml of lysis buffer (10X-PBS-hypotonic stock solution (10.9 g/l Na 2 HPO 4 (dibasic), 6.1 g/l NaH 2 PO 4 -H 2 O, 2.4 g/l KH 2 PO 4 , 2 g/l KCl, 3 g/l NaCl in nuclease-free water), 1.5 μl/ml of 2 M MgCl2, 2.5 μl/ml of 10% igepal, 40 μl/ml of 20 mg/ml BSA and 5 μl/ml of Rnasin Ribonuclease Inhibitor (#N2111, Promega, WI, USA) in nuclease-free water).The mixture was incubated for 10 min on ice, passed through a 40 µm cell strainer (#7340002, VWR, PE, USA) and transferred to a new tube.The remaining pieces of tissue were coaxed with a disposable pestle (#7343820, VWR, PE, USA).RnaseAlert (#11-02-01-02, Integrated DNA technologies, IO, USA) was used to examine the remaining RNAse activity on the lysate and ensure minimal RNA degradation in the samples before continuing with the protocol.The lysate was spined down (500 x g, 3 min at 4 • C) and the pellet was resuspended in 1 ml 0.3 M SPBSTM buffer (114 g/l sucrose, 10X PBS, nuclease-free water, 10 ml/l 10% triton-100, 1.5 ml/l 2 M MgCl 2 ).The lysate was spun down again and washed 2 times in 1 ml PBS with 1% BSA.Finally, the pellet was resuspended and blocked for 20 min on ice in 1 ml PBS with 1% BSA supplemented with Rnasin Ribonuclease.We then stained the nuclei with and Alexa405-NeuN antibody (#NBP1-92693AF405, Novus Biologicals, UK) for 30 min on ice.The labeled nuclei were then sorted on a BD influx (BD biosciences) cell sorter and pooled for further analysis.

RNA sequencing
The pooled sorted neuronal nuclei were processed for RNA isolation using the Rneasy Plus Micro Kit (Qiagen), according to the manufacturer's instructions.Library preparation was performed at the National Genomics Infrastructure at SciLifeLab using the Takara SMARTer Stranded Total RNA-Seq Kit v3 -Pico kit and sequenced on 1 NovaSeq 600 S4-300 lane, with 2×150bp reads.

Statistical analyses
All behavioral experiments were carried out with at least 6 mice per group and cell culture treatments were performed in triplicate.Statistical analysis was performed using Prism software (GraphPad 9).Student's t-test was used to evaluate the significance of the statistical difference between two groups.The statistical differences among three groups or more was calculated using One-way analysis of variance (ANOVA).For the gene expression data, differential gene expression analysis was performed using DESeq2 applying the standard comparison mode between two experimental groups [27].Fisher's exact test was used to assess the overlap of differentially expressed genes between Cyp27Tg and Cyp7B1KO neurons.For all tests, P < 0.05 was considered statistically significant.

Cyp7B1 deletion does not induce memory deficits in young and old mice
To assess spatial learning and memory, Cyp7b1KO mice and their littermate wild types (WT) were assessed by the MWM test at 7 months of age.Both groups of mice learnt the task and showed a significant reduction in the first occurrence latency (time spent to reach the platform position for the first time) during the 5 days of the acquisition phase (ANOVA repeated measures, the effect of days P < 0001, Fig. 1A).No differences were found in memory retention during the probe test on day 6.The first occurrence latency, crossing frequency on the platform position, and time spent in the quadrant where the platform was located did not differ significantly between groups (Fig. 1B, C, and D).
To further investigate whether the deletion of Cyp7b1 could affect memory function at an older age, we tested 14-month-old Cyp7B1KO mice and their WT littermate controls.We run the EPM test to assess anxiety-like behavior, the Y Maze for spatial working memory and the Barnes Maze for spatial learning and memory.Importantly, the Barnes Maze test requires less stamina than the MWM and it is usually the preferred choice when testing old mice.Our data showed no differences between experimental groups in the percentage of time spent on closed arms in the EPM test (data not shown).In the Y maze, the percentage of correct alternations between the three arms did not differ between transgenic and WT mice (Fig. 1E), indicating that spatial working memory was not affected by Cyp7b1 deletion in old mice.Finally, when testing spatial learning in the Barnes Maze Cyp7B1KO and WT showed similar numbers of primary errors (equivalent to the number of wrong choices before entering the correct target hole, ANOVA repeated measures, the effect of days P < 0001, Fig. 1F) and escape latency to the target hole (time spent to reach the target hole for the first time, ANOVA repeated measures, the effect of days P = 0,0194, Fig. 1G).Both parameters decreased along the days of training indicating that all groups of mice were able to locate the target hole during the acquisition phase.During the probe test on day 5, no significant differences were found in the numbers of primary errors and latency to the target hole between transgenic and WT mice (Fig. 1H and I).Altogether these results suggest that Cyp7b1 deletion does not induce significant changes in spatial learning and memory retention, independently of age.

Cyp7B1KO mice do not exhibit changes in dendritic spines, memoryrelated proteins, and brain glucose uptake
Reduced dendritic spine number and density as well as decreased PSD95, Snap25, and Arc protein levels have been previously reported in the brains of Cyp27Tg mice [15].To assess whether the accumulation of 27-OHC derived from deletion of Cyp7B1 exerts a similar effect in vivo, we examined the morphology and number of dendritic spines from Golgi-stained pyramidal neurons in the CA1 region of 14-month-old Cyp7B1KO male mice (Fig. 2A).Our results showed no significant changes in Cyp7B1KO spine density (measured as the number of spines/μm) (Fig. 2B), mean spine length (μm) (Fig. 2C), spine length frequency (Fig. 2D), mean spine area (μm 2 ) (Fig. 2E) or spinearea frequency (Fig. 2F) compared to sex and aged-matched WT littermate controls (n = 5 animals per group).
To determine potential molecular changes, we measured Arc, Snap25, and PSD95 protein levels by western blot in 14-month-old Cyp7B1KO mice hippocampal and cortical homogenates.No significant differences were observed for any of these proteins between transgenic and wild-type littermates (n = 5 animals per group) in any of the fractions (Fig. 2G-P).
A previous study measuring 18 F-FDG radioligand uptake in mice reported that Cyp27Tg mice had a lower uptake compared with their littermate controls [14].We therefore measured cerebral glucose metabolism in 14-month-old Cyp7B1KO mice by positron emission tomography (PET), using the radioligand 18 F-fluorodeoxyglucose (18 F-FDG) under isoflurane anesthesia.Cyp7B1KO mice showed no radioligand uptake differences compared with their littermate WT counterparts (Fig. 2Q-S).Together, these data suggest that, as opposed to Cyp27Tg mice, Cyp7B1KO mice do not develop synaptic, structural, or metabolic alterations.

7-HOCA does not reduce PSD95 levels and Cyp7B1 silencing does not counteract the effects of 27-OHC in neuronal primary cultures
In the brain, 27-OHC is metabolized by Cyp7B1, and 7-HOCA is its major metabolic product [20].Given the discrepancies regarding cognitive outcomes between the Cyp27A1 overexpression and Cyp7B1 deletion models, we hypothesized that the increase of 7-HOCA in Cyp27Tg mice could play a role in memory impairment.
To elucidate the potential role of 7-HOCA on synaptic function, we analyzed the effect of 7-OHCA treatment on the expression levels of PSD95 (Dlg-4 gene) and Snap25 in neurons in primary cultures.As previously reported (Merino-Serrais et al., 2019), neuronal primary cultures treated daily with 1 μM 27-OHC from day 1-4 days in vitro (DIV), and thereafter left in culture up to 10 DIV showed decreased Dlg-4 expression (Fig. 3A-B).Following this treatment strategy, we treated neuronal cultures with 7-HOCA (1 and 10 μM).No differences were found for Dlg4 expression in 7-HOCA-treated neurons compared to vehicle-treated controls (Fig. 3B).No differences in the expression levels of Snap25 and Cyp7B1 were found for any of the conditions (Fig. 3C-D).These results suggest that high levels of 7-HOCA do not affect synaptic function.
Considering that Cyp7B1KO mice accumulate 27-OHC in the brain [24] but unlike Cyp27Tg mice, they do not display any behavioral impairment or synaptic dysfunction, we speculated that the lack of Cyp7B1 may induce a protective effect on neurons.To test this hypothesis, we silenced the Cyp7B1 gene in neuronal primary cultures (from 1 to 4 DIV), followed by daily 27-OHC treatment (from 4 to 8 DIV) (Fig. 3E), and determined gene expression levels at 10 DIV by RT-qPCR.We observed an 80% reduction in the expression of Cyp7B1 in the siRNA-treated neurons (Fig. 3F).Knocking down Cyp7B1 did not prevent the reduction of Dlg4 expression triggered by 27-OHC (Fig. 3G) and did not have any effect on Snap25 levels (Fig. 3H).These results suggest that the protection against the toxic effects of 27-OHC accumulation observed in the Cyp7B1KO mice is not mediated by a lower neuronal susceptibility to 27-OHC caused by the absence of Cyp7B1.

Neurons from Cyp27Tg and Cyp7B1KO mice display different transcriptional profiles
We finally aimed to explore the molecular mechanisms and pathways involved in the differences observed between Cyp27Tg and Cyp7B1KO neurons.To do so we isolated cortical neuronal nuclei from 14 -monthsold male mice by flow cytometry and subjected the pooled neuronal nuclei to RNA-seq analysis (Fig. 4A).In agreement with our results, hierarchical cluster analysis revealed that the transcriptional profile of Cyp27Tg is different to wild-type mice, while Cyp7B1KO did not separate as clearly (Fig. 4B).

Discussion
A large body of evidence supports the neurotoxic effects of CYP27A1 overexpression and high levels of 27-OHC.We have reported that CYP27A1 overexpression in vivo impairs neuronal morphology, reduces hippocampal spine density and PSD95 levels already at 2 months of age [15], and leads to cognitive decline in both adult male and female mice (8-9 months), and a reduction of brain glucose uptake in males [14,47].The detrimental effects of 27-OHC were observed also in vitro where treated neurons show a reduction of neurites and branching points, lower viability, and decreased metabolic activity [15,22].When 27-OHC was injected intracerebrally in wild-type mice these negative effects were recapitulated [15].Finally, high levels of BBB permeable oxysterol 27-OHC caused by high cholesterol diet intake induced memory impairment in both rats and mice [48][49][50].The negative effects of dietary cholesterol were reverted by knocking out CYP27A1 in mice [51].
In this study, we tested Cyp7b1KO mice, with comparable 27-OHC levels to Cyp27Tg mice, at young and old ages.Differently from Cyp27Tg mice, we could not observe a cognitive decline or a reduction in brain glucose uptake and hippocampal dendritic spine density.These findings were further supported by unchanged levels of memory-related proteins in both cortex and hippocampus of the knock-out mice.In contrast with our results, a recent study described the memory performance of Cyp7b1KO mice at 3-4 months in the Morris Water maze test [52].The authors report no differences in the learning phase compared to control mice while a weak impairment was found in the probe test in KO mice when the percentage time spent in the platform quadrant was analyzed against an arbitrary chance level of 25% [52].In the same study, young Cyp7B1KO mice presented lower spine density when compared to control mice.While we cannot explain the reasons for the different findings between this and our study, we show that our results are consistent and confirmed with different methodologies and the use of mice at two different ages.
To understand the mechanisms underlying the absence of cognitive dysfunction despite increased levels of 27-OHC in Cyp7b1KO mice we investigated in vitro the effects of the 27-OHC metabolite, 7-HOCA, as well as the effects of 27-OHC upon Cyp7b1 deletion in primary neurons.Two different high concentrations of 7-HOCA did not affect the levels of synaptic proteins such as PSD95 and SNAP25.7-HOCA is a cholestenoic acid, intermediate of bile acids.In a previous study [53], several cholestenoic acids were shown to regulate survival in motor neurons by interacting with Liver X receptors, but 7-HOCA treatments did not have negative or positive effects, supporting our results.Furthermore, the effects of 27-OHC were not reverted by the silencing of Cyp7b1 in neurons, suggesting that the neuronal-specific Cyp7b1 deletion is not mediating the protective effect observed in Cyp7B1KO mice.
Transcriptomics data of neuronal nuclei from Cyp27Tg and Cyp7B1KO mice reveal two distinct expression profiles where widespread metabolic alterations were observed in Cyp27Tg.These alterations were accompanied by changes in pathways linked to neurodegeneration and Alzheimer's disease.Indeed, Cyp27Tg neurons showed lower levels of synaptic and memory-related genes like Arc, Nrgn, Snap25, and Dlg4, supporting previous findings [14,15].Of these genes, only Arc is downregulated in Cyp7b1KO but to a much lower extent, however, we did not see a significant change at the protein level.Noteworthy, GABAergic signaling is upregulated in Cyp7B1KO mice.This could be due to elevated levels of 3β-Adiol which is a Cyp7B1 substrate and a modulator of GABA receptors [54].Several oxysterols including 27OHC activate NMDARs [50,55] and increased GABA-ergic activity in Cyp7B1KO mice could serve as a protective neuronal mechanism.Moreover, 3β-Adiol is an estrogen receptor agonist [56,57], and target genes of estrogen receptors such as Bdnf [58] are found to be upregulated only in Cyp7B1KO mice, possibly conferring neuroprotection.Moreover, the gene that codes for Basigin, a cell surface molecule of the immunoglobulin superfamily involved in dendrite morphogenesis [59], is also upregulated in Cyp7B1KO mice.
Another main difference between the two mouse models Cyp27Tg and Cyp7B1KO is the accumulation of 25OHC in the latter.Oxysterol 25OHC is a signaling molecule with multiple relevant functions in the periphery and the central nervous system.In the brain, 25OHC is produced in microglial cells by the enzyme cholesterol 25-hydroxylase (Ch25h) [60,61].Regarding neurons, 25OHC treatments have been reported to induce a decrease in neurites, viability, and metabolism [53].
Nevertheless, several pieces of evidence have shown that 25OHC exhibits anti-viral properties and immune-modulatory effects which can be either pro-or anti-inflammatory.It has been shown that 25OHC amplifies IL-1B secretion by microglia in an ApoE-dependent manner [62] while loss of 25OHC promotes interleukin 1 beta (IL-1B) production in macrophages and suppresses IL-1-driven inflammation downstream of type I interferon [63,64].All these findings point to a possible difference in the microglial and inflammatory state between Cyp7B1KO and Cyp27Tg mice.We recently showed an increase of the pro-inflammatory alarmin S100A8 in Cyp27Tg mice suggesting higher inflammation in the brain of these mice [65].Microglial cells are suggested to have an active role in neuronal activity [66]; thus, we could speculate that a possible anti-inflammatory state could be mediated by high levels of 25OHC Cyp7B1KO neurons.The inflammatory profiles of Cyp27Tg and CYP7B1KO mice could provide future mechanistic insights into this hypothesis.

Conclusions
To conclude, this work shed light on the role of CYP27A1 and CYP7B1 in cognitive functions and on the possible molecular mechanisms behind the differences found between Cyp27Tg and Cyp7B1KO mice.Despite the accumulation of 27OHC in both models, Cyp7B1KO mice do not display cognitive dysfunction at both young and old age.These findings were supported by the transcriptional profile of Cyp7B1KO neurons that revealed upregulation of genes likely to confer higher neuroprotection to the knock-out animals.These neuroprotective effects could be mostly mediated by high levels of 25OHC and other Cyp7B1 substrates such as 3β-Adiol, however, future mechanistic studies will be needed to confirm our findings.Differently from Cyp7B1KO mice, RNAseq data from Cyp27Tg neurons showed significant downregulation of synaptic and metabolic functions strengthening previous findings on the deleterious effects of CYP27A1 overexpression and high levels of 27-OHC in the brain.This suggests that Cyp27Tg is a more suitable mouse model to study 27-OHC-specific signaling than Cyp7B1KO.
CYP27A1 and CYP7B1 are enzymes with key roles spanning from cholesterol metabolism and catabolism to neurosteroid signaling and are of growing importance in neurodegenerative diseases.Advancing our knowledge of their function will be essential to understanding the pathogenic mechanisms of diseases like AD and to discover potential new therapeutic approaches.

Funding
This research was supported by Margaretha af Ugglas Foundation, King Gustav V:s and Queen Victorias Foundation, the regional agree-

Fig. 1 .
Fig. 1.Cyp7B1KO mice do not develop learning and memory impairments.A) Escape latency or time required to reach the hidden platform in 7 months-old wildtype and age-matched Cyp7B1KO mice over 5 days acquisition phase in the Morris Water Maze (MWM) test.B) Frequency spent on the platform (sec), C) latency of first occurrence and D) time spent in the quadrant (sec) where the platform was located during the 24 h MWM probe test.E) Percentage of spontaneous alternations during the Y-maze test performed by 14 months-old wild-type and Cyp7B1KO mice.F) Number of primary errors and G) escape latency to the target hole during the 4 days acquisition phase of the Barnes maze test performed by 14 months-old wild-type and Cyp7B1KO mice.H) Number of primary errors and I) escape latency to target hole probe test of the Barnes maze test after 24 h.N = 6/7mice per group.Data is represented as mean ± SEM.

Fig. 4 .
Fig. 4.. Neurons from Cyp27Tg and Cyp7B1KO mice present different transcriptional profiles.A) Schematic representation of FANS-RNAseq analysis from nuclei isolated from cortical neurons of 14 months old wild-type control, Cyp27Tg and Cyp7B1KO mice (n = 2 mice per group).B) Heatmap of differentially expressed genes identified by RNA-seq (pval<0.05) between wild-type, Cyp27Tg and Cyp7B1KO cortical neurons.C) Volcano plot of the differential gene expression analysis between Cyp27Tg and wild-type neurons.D) Pathway analysis of Cyp27Tg versus wild-type neurons.E) Volcano plot of the differential gene expression analysis between Cyp7B1KO and wild-type neurons.F) Pathway analysis of Cyp7B1KO versus wild-type neurons.G) Venn diagram showing the overlap of downregulated (ocre) and upregulated (blue) DEGs in Cyp27Tg and Cyp7B1KO neurons.N = 2 mice per group.Statistical analysis was performed with a Fisher's exact test (*** pval < 0.0001).