Cytochrome P450 27A1 Deficiency and Regional Differences in Brain Sterol Metabolism Cause Preferential Cholestanol Accumulation in the Cerebellum*

Cytochrome P450 27A1 (CYP27A1 or sterol 27-hydroxylase) is a ubiquitous, multifunctional enzyme catalyzing regio- and stereospecific hydroxylation of different sterols. In humans, complete CYP27A1 deficiency leads to cerebrotendinous xanthomatosis or nodule formation in tendons and brain (preferentially in the cerebellum) rich in cholesterol and cholestanol, the 5α-saturated analog of cholesterol. In Cyp27a1−/− mice, xanthomas are not formed, despite a significant cholestanol increase in the brain and cerebellum. The mechanism behind cholestanol production has been clarified, yet little is known about its metabolism, except that CYP27A1 might metabolize cholestanol. It also is unclear why CYP27A1 deficiency results in preferential cholestanol accumulation in the cerebellum. We hypothesized that cholestanol might be metabolized by CYP46A1, the principal cholesterol 24-hydroxylase in the brain. We quantified sterols along with CYP27A1 and CYP46A1 in mouse models (Cyp27a1−/−, Cyp46a1−/−, Cyp27a1−/−Cyp46a1−/−, and two wild type strains) and human brain specimens. In vitro experiments with purified P450s were conducted as well. We demonstrate that CYP46A1 is involved in cholestanol removal from the brain and that several factors contribute to the preferential increase in cholestanol in the cerebellum arising from CYP27A1 deficiency. These factors include (i) low cerebellar abundance of CYP46A1 and high cerebellar abundance of CYP27A1, the lack of which probably selectively increases the cerebellar cholestanol production; (ii) spatial separation in the cerebellum of cholesterol/cholestanol-metabolizing P450s from a pool of metabolically available cholestanol; and (iii) weak cerebellar regulation of cholesterol biosynthesis. We identified a new physiological role of CYP46A1, an important brain enzyme and cytochrome P450 that could be activated pharmacologically.

(iii) weak cerebellar regulation of cholesterol biosynthesis. We identified a new physiological role of CYP46A1, an important brain enzyme and cytochrome P450 that could be activated pharmacologically.
Cholestanol (Fig. 1) is a metabolite of cholesterol, which lacks the double bond at position C5 and represents the 5␣-saturated analog of cholesterol. Normally, cholestanol is present in virtually every mammalian organ at concentrations of only 1/500 to 1/800 of cholesterol (1), yet, in the disease cerebrotendinous xanthomatosis (CTX), tissue cholestanol (as well as cholesterol) is elevated, especially in the tendons and brain, where the two sterols accumulate in xanthomas, nodules rich in lipids (2,3). In the brain, CTX xanthomas are preferentially formed in the cerebellum, in particular its white matter (4), and contain, along with the cerebellum, the highest ratios of cholestanol to cholesterol of all tissues in CTX subjects (2). Besides xanthomas, CTX is manifested by juvenile bilateral cataracts, early atherosclerosis, osteoporosis, and progressive neurological deterioration (3). CTX is an autosomal recessive disease due to deficiency in CYP27A1 2 (5,6), a multifunctional ubiquitous cytochrome P450 enzyme involved in the production of bile acid intermediates in the liver, cholesterol elimination from extrahepatic tissues, and metabolism of vitamin D 3 in the kidney (7,8).
The major pathway for cholestanol production in the CTX brain is established (Fig. 1) and accounts for about 70% of the synthesis of this steroid (9,10). This pathway begins in the liver, where 7␣-hydroxy-4-cholesten-3-one, an efficient precursor to cholestanol in various tissues, cannot be metabolized by CYP27A1 (11,12) and is therefore released in the systemic circulation (13). From the circulation, 7␣-hydroxy-4-cholesten-3one is continuously fluxed across the blood-brain barrier to the brain (4,14,15) for the conversion to cholesta-4,6-dien-3-one, 4-cholesten-3-one, 5␣-cholestan-3-one, and finally cholestanol (10, 14, 16 -19). Remarkably, a flux of only 2% of 7␣-hydroxy-4-cholesten-3-one and other 7␣-hydroxylated sterols into the brain is sufficient to explain cerebral accumulation of cholestanol in CTX (4). However, under normal conditions, the 7␣hydroxy-4-cholesten-3-one-dependent pathway accounts for only about 30% of cholestanol biosynthesis in the brain, and cerebral cholestanol is mainly formed from cholesterol, which is synthesized locally and is oxidized to 4-cholesten-3-one (Fig. 1). The details of this oxidation have not yet been clarified but were suggested to include a microsomal 3␤-hydroxy-⌬ 5 -dehydrogenase reaction (16). A non-enzymatic cholesterol oxidation in the brain was also reported by the amyloid-␤-peptide⅐Cu 2ϩ complex (20) and potentially could be of importance. Amyloid-␤ plaques are already detected in human brain at the age of 30, and their accumulation progresses with age (21).
The present work was initiated by the ophthalmic characterization of Cyp27a1 Ϫ/Ϫ Cyp46a1 Ϫ/Ϫ mice, whose cholestanol was found to be elevated in the retina (22), a neural tissue in the back of the eye and a part of the CNS. We hypothesized that this sterol increase could be due to cholestanol metabolism by CYP46A1 and decided to investigate whether cholestanol is also increased in the brain of Cyp27a1 Ϫ/Ϫ Cyp46a1 Ϫ/Ϫ mice. CYP46A1 is a CNS-specific enzyme, which catalyzes cholesterol 24-hydroxylation, the major mechanism for cholesterol elimination from the brain (23,24). Unlike cholesterol, which cannot cross the blood-brain barrier, 24(S)-hydroxycholesterol rapidly diffuses to the systemic circulation and is delivered to the liver for further degradation to bile acids (23). CYP46A1 is more abundant in the brain than ubiquitous CYP27A1 (25). Therefore, the regional brain concentrations of 27-hydroxycholesterol, the primary CYP27A1 product, are much lower than those of 24-hydroxycholesterol. They represent only ϳ10 -20% of the brain 24-hydroxycholesterol content (23), despite there being substantial uptake by the brain of extracerebral 27-hydroxycholesterol from the systemic circulation (26). Once in the brain, 27-hydroxycholesterol, blood-borne or synthesized locally, is further metabolized to 7␣-hydroxy-3-oxo-4-cholestenoic acid (Fig. 1), which is then effluxed into the systemic circulation (27).
Similar to CTX, cholestanol is elevated in the plasma, tendons, and brain of Cyp27a1 Ϫ/Ϫ mice and probably has the same mechanism of tissue accumulation (10), yet cholestanol increases are lower in Cyp27a1 Ϫ/Ϫ mice as compared with CTX-affected individuals, and these animals do not develop xanthomas (13,28,29). Nevertheless, Cyp27a1 Ϫ/Ϫ mice provide valuable mechanistic insights into the etiology of CTX (10). Accordingly, these as well as other genetically manipulated animals were used in the present work along with the brain tissues from human donors. We identified several factors that probably underlie the preferential cholestanol accumulation in CYP27A1 deficiency. Proposed pathways of cholestanol biosynthesis and elimination from the brain. The 7␣-hydroxy-4-cholesten-3-one-dependent pathway is initiated in the liver and is indicated with blue arrows; cholestanol precursors and metabolites in this pathway are shown in gray. The cholesterol-dependent pathway is initiated in the brain and is indicated with magenta arrows. In both pathways, the known enzymes are indicated; boldface arrows represent the major mechanism for cholesterol or cholestanol elimination from the brain.

Cholesterol and Cholestanol Concentrations in Mouse and
Human Brain-The specimens characterized were from two background wild type strains (C57BL/6J and C57BL/6J;129S6/ SvEv), three knock-out lines (Cyp27a1 Ϫ/Ϫ , Cyp46a1 Ϫ/Ϫ , and Cyp27a1 Ϫ/Ϫ Cyp46a1 Ϫ/Ϫ ), and two human donors (one female and one male). In the whole mouse brain ( Fig. 2A), the levels of cholesterol were unchanged in all genotypes. In contrast, the levels of the whole brain cholestanol varied and were increased in the knock-out strains. In Cyp27a1 Ϫ/Ϫ mice, the cholestanol increases were 9.9-fold in females and 7.6-fold in males, a pattern consistent with previous work (10). In the Cyp46a1 Ϫ/Ϫ genotype, cholestanol was not changed in females and was increased 4.5-fold in males with the gender difference being statistically significant. Finally, Cyp27a1 Ϫ/Ϫ Cyp46a1 Ϫ/Ϫ mice had the highest increase in the whole brain cholestanol, 11.6-fold in females and 8.3-fold in males. Thus, in Cyp27a1 Ϫ/Ϫ Cyp46a1 Ϫ/Ϫ mice, not only the retina but also the whole brain had an increase in cholestanol (22). Furthermore, in the wild type strains, the whole brain cholestanol represented 0.06 -0.07% of cholesterol, whereas in the knockout strains, the cholestanol to cholesterol percentages were higher: 0.51-0.68% in Cyp27a1 Ϫ/Ϫ mice, 0.11-0.33% in Cyp46a1 Ϫ/Ϫ mice, and 0.6 -0.7% in Cyp27a1 Ϫ/Ϫ Cyp46a1 Ϫ/Ϫ mice.
Next, we carried out the measurements in mouse and human cerebella. In the wild type strains, the basal cerebellar levels of cholesterol and cholestanol were higher than those in the whole brain: 376 -544 versus 278 -360 nmol/mg of protein for cholesterol and 983-1,581 versus 175-237 pmol/mg of protein for cholestanol (Fig. 2B). The content of cholestanol relative to cholesterol was also higher than that in the whole brain (0.25-0.31% versus 0.06 -0.07%). Similarly, the cerebellar cholestanol to cholesterol percentages in the knock-out strains were higher than those in the whole brain: 2.1-2.3% in Cyp27a1 Ϫ/Ϫ mice, 0.3-0.4% in Cyp46a1 Ϫ/Ϫ mice, and 2.7-2.9% in Cyp27a1 Ϫ/Ϫ Cyp46a1 Ϫ/Ϫ mice. These percentages were also higher relative to those in the cerebellum of the corresponding wild types.
Cerebellar specimens from only two human donors, one female and one male, were analyzed. Hence, the data obtained do not allow generalizations and are interpreted with extreme caution. Nevertheless, these data provide valuable insight into interspecies similarities and difference for cerebellar sterol and P450 content. In human cerebellum (Fig. 2C), cholesterol content in the gray matter seemed to be Ͼ1.6-fold lower than that in the white matter (392-400 versus 654 -667 nmol/mg of protein) and appeared to be comparable with the cholesterol content in the cerebellum of wild type mice (376 -540 nmol/mg of protein). In contrast, the cerebellar gray matter cholestanol seemed to be Ͼ1.7-fold higher than that in the white matter (900 -924 and 500 -542 nmol/mg of protein, respectively) but appeared to be similar to cholestanol content in the whole mouse cerebellum (980 -1580 pmol/mg of protein). In our two donors, cholestanol represented 0.22-0.23% of cholesterol in the cerebellar gray matter and 0.08% of cholesterol in the cerebellar white matter, thus revealing that the sterol profile of the human cerebellar gray matter could be more similar to that of the whole mouse cerebellum than the white matter.
Cholestanol Is the Substrate for CYP46A1 in Vitro-Cholestanol, which is structurally very similar to cholesterol, was previously found to undergo 24-hydroxylation when incubated with CYP46A1-transfected HEK293 cells (30). However, the yield of the product was low, precluding its detailed characterization by mass spectrometry. In the present work, we used the in vitro reconstituted system and incubated cholestanol with purified recombinant CYP46A1. The incubation times varied and were 30 and 60 min; the product formation was assessed by gas chromatography-mass spectrometry (GC-MS). Only one peak, eluted at 29.46 min during GC, showed the expected intensity increase with the incubation time (Fig. 4A). This peak was not seen in the control incubations lacking NADPH and had a fragmentation pattern consistent with that of 24-hydroxycholestanol with the characteristic m/z values of 533 (M-15), 415 (M-90-43), and 325 (M-90-90-43) (Fig. 4B). Thus, cholestanol is metabolized by CYP46A1 in vitro. Cholestanol was next characterized in the spectral assay and compared with cholesterol for binding to CYP46A1. Both sterols had nanomolar apparent spectral K d values (Table 1), revealing tight P450 binding. The apparent K d of cholestanol was ϳ70 nM, suggesting that in our enzyme assay, 50 M cholestanol was probably saturating for CYP46A1. Hence, we could estimate the turnover number of cholestanol in the 30-min incubation (0.04 min Ϫ1 ) and found that it was comparable with that of cholesterol (0.11 min Ϫ1 ) (31).
When a 1:1 mixture of cholesterol and cholestanol was used in the in vitro reconstituted system, purified recombinant CYP27A1 hydroxylated cholestanol as well (32). Herein, we used only one sterol, cholestanol, and the CYP27A1 in vitro enzyme assay was run for 5 and 15 min. As in incubations with CYP46A1, only one product peak, eluted at 30.96 min, FIGURE 2. Brain levels of CYP27A1, CYP46A1, and their substrates and metabolites. For sterol quantifications, the results are the mean Ϯ S.D. of the independent measurements in five mice per genotype and gender (A), three mice per genotype and gender (B), or duplicate measurements in the specimens from one female and one male donor (C). ND, not detected (the limit of detection is 1 pmol/mg of protein). MRM data from 3-4 transitions/peptide (3 peptides for CYP46A1 and 2 peptides for CYP27A1) were combined to determine the amount of each P450. Data represent mean Ϯ S.D. (error bars) of the independent P450 measurements in three mice or one human donor. NM, not measured. Pink asterisks, significant changes between females of the knock-out strains versus females of the corresponding wild type strain. Blue asterisks, significant changes between males of the knock-out strains and males of the corresponding wild type strain. Gray asterisks, significant changes between female and male mice of the same strain. Black asterisks, significant changes between the genotypes when data were collapsed across genders. *, p Յ 0.05; **, p Յ 0.01; ***, p Յ 0.001; ****, p Յ 0.0001 by two-way analysis of variance followed by Bonferroni correction as described under "Experimental Procedures."  . Cholestanol is an endogenous substrate for CYP46A1 in mice. A, total ion chromatogram of sterol extract (after trimethylsilylation) from the in vitro incubations of cholestanol with purified recombinant CYP46A1. The peak at 29.46 min corresponding to the CYP46A1 product is circled and shown as an inset. B, mass spectrum of the peak at 29.46 min is consistent with the fragmentation pattern of 24-hydroxycholestanol. C, single ion monitoring chromatogram of the oxysterol fraction isolated from the whole brain of wild type (Cyp46a1 ϩ/ϩ ) and Cyp46a1 Ϫ/Ϫ mice. increased with reaction time (Fig. 5A), and this peak had a fragmentation pattern (Fig. 5B) consistent with that of 27-hydroxycholestanol reported previously (32). The turnover number for a 5-min incubation was estimated (3.5 min Ϫ1 ) and appeared to be similar to that of cholesterol (3.0 min Ϫ1 ) (33). The spectral parameters of cholestanol and cholesterol binding to CYP27A1 were also similar, with apparent spectral K d values of 5.0 and 6.3 M, respectively (Table 1). Thus, under the optimal conditions of the in vitro assays, cholestanol binding to CYP46A1 and CYP27A1 and subsequent metabolism by the P450s were comparable with those of cholesterol.
Cholestanol Is the Substrate for CYP46A1 in Mice-Because cholestanol was metabolized by CYP46A1 in vitro, mouse and human brains were analyzed for 24-hydroxycholestanol, the P450 enzymatic product. The ion peaks characteristic of 24-hydroxycholestanol were indeed detected in the whole brain of wild type mice but not in Cyp46a1 Ϫ/Ϫ mice (Fig. 4C), providing evidence that cholestanol is the substrate for CYP46A1 in vivo. The ion peaks characteristic of 27-hydroxycholestanol (Fig. 5C) were found in the whole brain of wild type mice as well, yet they were not observed in Cyp27a1 Ϫ/Ϫ mice, further support that cholestanol is metabolized by CYP27A1 in vivo (32). 24-and 27-hydroxycholestanols were then quantified in mouse and human brains and compared with the levels of 24-and 27-hydroxycholesterols, which were also measured. In wild type mice, the content of 24-hydroxycholestanol was much lower than that of 24-hydoxycholesterol, by ϳ8-fold in the whole brain and Ͼ100-fold in the cerebellum, yet the percentage ratios of 24-hydroxycholestanol to cholestanol and 24-hydroxycholesterol to cholesterol were comparable (0.1-0.4%), FIGURE 5. Cholestanol is an endogenous substrate for CYP27A1 in mice. A, total ion chromatogram of sterol extract (after trimethylsilylation) from the in vitro incubations of cholestanol with purified recombinant CYP27A1. The peak at 30.96 min corresponding to the CYP27A1 product is circled and shown as an inset. B, mass spectrum of the peak at 30.96 min is consistent with the fragmentation pattern of 27-hydroxycholestanol. C, single ion monitoring chromatogram of the oxysterol fraction isolated from the whole brain of wild type (Cyp27a1 ϩ/ϩ ) and Cyp27a1 Ϫ/Ϫ mice. except 58 -68% of 24-hydroxycholestanol relative to cholestanol in the whole brain. 27-Hydroxycholestanol was detected in the whole brain of wild type mice (17-26 pmol/mg of protein) but not in the cerebellum and was at levels similar to those of the whole brain 27-hydroxycholesterol (18 -32 pmol/mg of protein). The content of the whole brain 27-hydroxycholestanol was 10 -15% of that of cholestanol, which is much lower than the percentage content of 24-hydroxycholestanol to cholestanol (58 -68%). Thus, the cerebellum does not appear to be very efficient in the metabolism of cholestanol, which is mostly hydroxylated in the other brain regions, more by CYP46A1 than by CYP27A1. Neither 24-nor 27-hydroxycholestanol seemed to be detected in human cerebellar gray and white matter (Fig. 2C), yet 24-and 27-hydroxycholesterols were present, at much higher concentrations in the gray matter than in the white matter, the pattern of 27-hydroxycholesterol distribution different from that found in the previous work (26). This could be due to the post-mortem sterol washout (34) and small amounts of tissue 27-hydroxycholesterol. Remarkably, of all of the brain specimens analyzed, human or mouse, the levels of 24-hydroxycholesterol and 27-hydroxycholesterol appeared to be the highest in the human cerebellar gray matter (1487-1521 and 44 -46 pmol/mg of protein, respectively). Hence, CYP46A1 and CYP27A1 expression was measured, which could contribute to tissue differences in the amounts of hydroxylated sterols.
P450 Expression in the Brain-The highest CYP46A1 expression was in the mouse whole brain (860 -980 fmol/mg of protein; Fig. 2A); the lowest seemed to be in human cerebellum (40 -70 fmol/mg of protein; Fig. 2C), with mouse cerebellum having intermediate P450 expression (283-383 fmol/mg of protein; Fig. 2B). Notably, in human cerebellum, CYP46A1, normally a neuron-specific enzyme (35), appeared to be detected in both gray and white matter. This could be a reflection of either white matter contamination with gray matter or CYP46A1 expression in the axons of Purkinje cells and Golgi I neurons that extend deep into the cerebellum. However, the latter interpretation is not supported by the previous work on CYP46A1 immunolocalization showing no CYP46A1 expression in the axons of Purkinje cells and Golgi I neurons (35).
Unlike CYP46A1 expression, the expression of CYP27A1 seemed to be in the same range in all brain specimens except in those with enzyme deficiency (Fig. 2). In the whole mouse brain, CYP27A1 was at ϳ50-fold lower levels than CYP46A1, in agreement with CYP46A1 being the major cerebral cholesterol hydroxylase (23,24). Similarly, CYP27A1 was at lower levels than CYP46A1 in mouse cerebellum. Only in human cerebellum, CYP27A1 expression appeared to be higher than that of CYP46A1 both in gray matter and white matter.
Lathosterol and Desmosterol Concentrations in Mouse Whole Brain and Cerebellum-Lathosterol and desmosterol are cholesterol precursors in the pathways of cholesterol biosynthesis and reflect the rate of cholesterol biosynthetic process in neurons (lathosterol) and astrocytes (desmosterol) (36). The rate of cholesterol biosynthesis is known to be increased in the cerebrum of Cyp27a1 Ϫ/Ϫ mice, as indicated by increased lathosterol levels (10), and decreased in the whole brain of Cyp46a1 Ϫ/Ϫ mice, as indicated by the balance studies (37). We investigated whether the levels of lathosterol and desmosterol were altered in the cerebellum of Cyp27a1 Ϫ/Ϫ , Cyp46a1 Ϫ/Ϫ , and Cyp27a1 Ϫ/Ϫ Cyp46a1 Ϫ/Ϫ mice, and if so, how these changes were comparable with those in mouse whole brain. In Cyp27a1 Ϫ/Ϫ mice, lathosterol was increased in both cerebellum and whole brain as compared with the corresponding wild type strain (Fig. 6A), and these increases were higher in the cerebellum than in the whole brain (2.5-2.7-fold versus 1.5-1.7-fold). In contrast, Cyp46a1 Ϫ/Ϫ mice had a decrease in lathosterol levels in both cerebellum and whole brain, and these decreases were lower in the cerebellum than in the whole brain (2.1-fold versus 4.4 -11-fold). In Cyp27a1 Ϫ/Ϫ Cyp46a1 Ϫ/Ϫ mice, lathosterol levels were decreased in both cerebellum and whole brain, and these decreases were because of the changes in males. Desmosterol was unchanged in the cerebellum of the knock-out strains and decreased in the whole brain in Cyp46a1 Ϫ/Ϫ and Cyp27a1 Ϫ/Ϫ Cyp46a1 Ϫ/Ϫ mice (Fig. 6B). Thus, genetic deficiencies in CYP27A1 and CYP46A1 mainly affect cholesterol biosynthesis in brain neurons, with changes in astrocytes being pronounced only in CYP46A1-deficient strains in the whole brain. Astrocytes are believed to produce the majority of brain cholesterol (38) and, along with brain neurons, apparently have a compensatory down-regulation of cholesterol biosynthesis in the whole brain in response to a lack of cholesterol 24-hydroxylation, the major pathway of cholesterol elimination in the brain.

Discussion
The present work led to two major findings that, along with general biochemical knowledge, helped us to interpret the measurements of the sterol content and P450 expression in different brain specimens (Fig. 2). Our first finding was that not only cholesterol but also cholestanol is an endogenous substrate for CYP46A1 in the brain (Table 1 and Fig. 4). The second finding was that in mouse cerebellum, the metabolism of cholestanol is low and mainly occurs via 24-hydroxylation, whereas in the other brain regions in mice, the metabolism of cholestanol is  MARCH 24, 2017 • VOLUME 292 • NUMBER 12 high and occurs via both 27-and 24-hydroxylations. In human cerebellum, cholestanol metabolism seems to be non-existent based on studies in two donors (Fig. 2C). It must be kept in mind that our measurements in two human samples, one from a male and one from a female, are at the moment relevant only for these two samples. While analyzing the data, we also kept in mind that enzyme and substrate concentrations as well as substrate availability and the ratio of the substrates that compete for the enzyme active site are all important biochemical factors determining the amount of the enzymatic product formed.

Cholestanol in the Cerebellum
We established that under normal conditions (i.e. in wild type mice) and as compared with the whole brain and human specimens, mouse cerebellum had the highest cholestanol content (928 -1582 pmol/mg of protein), the highest percentage of cholestanol relative to cholesterol (0.25-0.31%), the highest CYP27A1 expression (111-115 fmol/mg of protein), and CYP46A1 expression (286 -383 fmol/mg of protein) only ϳ3-fold lower than that in the whole brain. Nevertheless, the wild type mouse cerebellum had the lowest 24-hydroxycholestanol concentrations (only 3-5 pmol/mg of protein) and the lowest percentage ratio of 24-hydroxycholestanol to cholestanol (0.3-0.5% for wild type mice). Conversely, the whole brain of wild type mice had the lowest cholestanol content (179 -237 pmol/mg of protein) and the lowest percentage of cholestanol relative to cholesterol (0.06%-0.07%). Nevertheless, its 24-hydroxycholesterol and 24-hydroxycholestanol concentrations were the highest (112-150 and 17-26 pmol/mg of protein, respectively) as were the percentage of 24-hydroxycholestanol and 27-hydroxycholestanol to cholestanol (58 -68 and 10 -15%, respectively). The relatively high ratio of 24-hydroxycholestanol to cholestanol could be due to a lower rate of diffusion of 24-hydroxycholestanol from the brain into the circulation. The sterol profile of the human cerebellar gray matter appeared to be similar to that of the mouse cerebellum, except that CYP46A1 expression seemed to be much lower (only 70 fmol/mg of protein) and hydroxycholestanols were not detected, either because they were not formed or they had been washed out post-mortem. Thus, our measurements are consistent with the possibility that it is not the CYP46A1 and CYP27A1 abundance, local amounts, and ratios of cholestanol to cholesterol but rather cholestanol availability for hydroxylation that is most important for cholestanol metabolism in the brain. In addition, human CYP46A1 may have a higher catalytic efficiency of cholesterol 24-hydroxylation than the mouse ortholog. Human and mouse CYP46A1 share 95% amino acid sequence identity (24), yet many of their primary sequence differences are at the entrance or inside the CYP46A1 active site (31), the functionally important regions that could affect enzyme activity and efficiency of catalysis.
Why is cholestanol availability for CYP46A1/CYP27A1 limited (or nonexistent) in the cerebellum? There is only one cell type highly expressing CYP46A1 in the cerebellum: the somas and dendritic trees of Purkinje cells (24,35). These cells form a single layer in the cerebellar gray matter and account for ϳ0.5 and 0.25% of the total neuronal content in mouse cerebellum and whole brain, respectively (39). Purkinje cells are one of the largest neurons in the brain, and CYP46A1 expression in these neurons is probably very high, as suggested by the measurable cerebellar CYP46A1 content (Fig. 2, B and C) but very low Purkinje cell content. Probably, cholestanol is either not formed/ delivered to the Purkinje cells or is spatially separated in these large cells from CYP46A1, which resides in the endoplasmic reticulum. Cholestanol may also be separated from CYP27A1, a mitochondrial enzyme expressed in many cell types (40) and abundant in the cerebellum (111-115 fmol/mg of protein; Fig.  2B). Apparently, in the cerebellum, neither endoplasmic reticulum and mitochondria of Purkinje cells nor mitochondria of other cell types have a pool of metabolically available cholestanol. This is in contrast to cholesterol, which is available and is hydroxylated by CYP46A1 and CYP27A1 (Fig. 2, B and C). In mice, a minor cerebellar metabolism of cholestanol (24-hydroxylation) possibly takes place in Golgi cells, the granule cell layer interneurons, some of which express CYP46A1 (35). Similarly, cholestanol is probably available for CYP46A1 and CYP27A1 in the P450-containing cells of the other brain regions: in cortical and hippocampal pyramidal neurons along with hippocampal interneurons that express CYP46A1 (35) and in pyramidal neurons of the cortex as well as some cortical oligodendrocytes that express CYP27A1 (40). We suggest that there is a spatial separation of cholestanol from CYP46A1 and CYP27A1 that determines, at least in part, cholestanol accumulation in the cerebellum under the conditions of increased cholestanol load.
Other factors probably contribute to the cerebellar cholestanol accumulation in CYP27A1 deficiency. One of them could be CYP27A1 abundance, comparable with that of CYP46A1 only in the cerebellum, apparently the major site of the brain CYP27A1 expression, at least in mice (Fig. 2). Accordingly, CYP27A1 deficiency is probably affecting the cerebellum more than the other brain regions, where CYP27A1 is much less abundant because it increases the cerebellar cholestanol production. Indeed, the cerebellar 7␣-hydroxy-4-cholesten-3-one, the cholestanol precursor (11,12), cannot be metabolized in this condition by CYP27A1 and is used instead for cholestanol biosynthesis (Fig. 1).
The regulation of cholesterol biosynthesis and overproduction of cholesterol, a substrate for cholestanol biosynthesis, could be another factor that underlies cerebellar cholestanol accumulation in CYP27A1 deficiency. The possibility has been discussed that the accumulation of cholesterol in cholestanolcontaining xanthomas is due to increased synthesis of cholesterol as a consequence of a lower capacity of cholestanol (as compared with cholesterol) to suppress 3-hydroxy-3-methylglutaryl-CoA reductase (HMGCR), the rate-limiting enzyme in the pathway of cholesterol biosynthesis (14). This may be part of the explanation for the increased levels of cholesterol in the cerebellum of the Cyp27a1 Ϫ/Ϫ mice and Cyp27a1 Ϫ/Ϫ Cyp46a1 Ϫ/Ϫ mice (Fig. 2B). In addition, a lack of 27-hydoxycholesterol could contribute as well. Previously, we found a 12-fold increase in protein, but not mRNA, levels of HMGCR in the Cyp27a1 Ϫ/Ϫ retina as compared with the wild type retina and provided an explanation (41). This increase could be a result of the 27-hydoxycholesterol-dependent HMGCR degradation because 27-hydoxycholesterol was shown to bind to the protein called INSIG (insulin-induced gene, which plays an important role in the regulation of cholesterol biosynthesis (42)) and to accelerate through this binding the degradation of HMGCR (43). It is possible that a similar mechanism of posttranslational HMGCR regulation exists in the Cyp27a1 Ϫ/Ϫ brain and is stronger in the cerebellum than in the whole brain because of higher CYP27A1 abundance and levels of 27-hydoxycholesterol.
Cyp46a1 Ϫ/Ϫ mice have unchanged levels of the whole brain cholesterol ( Fig. 2A) because of a compensatory down-regulation of the whole brain cholesterol biosynthesis (37). Consistent with this down-regulation, lathosterol and desmosterol were significantly decreased in the whole brain of Cyp46a1 Ϫ/Ϫ mice (Fig. 6). Lathosterol but not desmosterol was also decreased in the Cyp46a1 Ϫ/Ϫ cerebellum but to a lower extent than in the whole brain (2.1-fold versus 4.4 -11-fold; Fig. 6A). Perhaps this lower decrease was not sufficient to compensate for a lack of 24-hydroxylation; hence, the levels of cholesterol were significantly increased in the cerebellum in Cyp46a1 Ϫ/Ϫ mice (Fig.  2B). We suggest that there may be insufficient coupling between cholesterol biosynthesis and cholesterol 24-hydroxylation in the Cyp46a1 Ϫ/Ϫ cerebellum. Of interest is that in Cyp46a1 Ϫ/Ϫ mice, the cerebellar cholestanol levels (1502-2146 pmol/mg of protein) were approximately equal to the sum of the wild type cerebellar cholestanol (982-1483 pmol/mg of protein) and wild type cerebellar 24-hydroxycholesterol (315-455 pmol/mg of protein). This suggests that in Cyp46a1 Ϫ/Ϫ mice, a part of the cerebellar cholesterol excess formed as a result of cholesterol biosynthesis-cholesterol metabolism uncoupling was converted to cholestanol.
There seems to be at least one factor regulating cholestanol homeostasis in mouse brain, as indicated by the decreased levels of 7␣-hydroxy-4-cholesten-3-one in Cyp27a1 Ϫ/Ϫ and Cyp27a1 Ϫ/Ϫ Cyp46a1 Ϫ/Ϫ mice (Fig. 3, C and D). These decreases may be considered as a compensatory up-regulation of 7␣-hydroxy-4-cholesten-3-one metabolism in response to the markedly increased sterol flux from the systemic circulation in the brain. It is possible that this increased metabolism reflects an induction of the dehydratase catalyzing the first and rate-limiting step in the conversion of 7␣-hydroxy-4-cholesten-3-one into cholestanol. Another possibility is that there is an induction of a hitherto unknown hydroxylase active toward 7␣-hydroxy-4-cholesten-3-one. It should be emphasized that high levels of 7␣-hydroxy-4-cholesten-3-one are cytotoxic (14). Hence, metabolism of 7␣-hydroxy-4-cholesten-3-one, including its conversion into cholestanol, can be regarded as a detoxification.
The standard treatment for CTX is replacement therapy with chenodeoxycholic acid either alone or in combination with one of the statins, drugs that inhibit cholesterol biosynthesis (44). These pharmacologic interventions usually improve clinical symptoms and lead to stabilization of CTX, which is a slowly progressive disease. However, not all of the patients are responsive to the chenodeoxycholic acid with/without statin treatment, and some of them have progressive neurological deterioration (44). Previously, we established that CYP46A1 can be activated allosterically both in vitro and in mice by a very small dose of the anti-HIV drug efavirenz (45,46). The clinical trial now is in preparation to investigate whether efavirenz activates CYP46A1 in the human brain and affects people with normal cognition and mild cognitive impairment. If successful, this trial will identify CYP46A1 as a new drug target and will suggest, along with the data of the current investigation, that pharmacologic activation of CYP46A1 should be considered not only for the treatment of people with mild cognitive impairment but also for the treatment of patients with CTX who do not respond to the treatment with chenodeoxycholic acid.
In summary, we conducted a comprehensive quantitative characterization of mouse and human brain specimens for sterol and P450 levels and revealed several factors that probably underlie the preferential cerebellar cholestanol accumulation in CYP27A1 deficiency. These factors include the ability of CYP46A1 to metabolize cholestanol, low cerebellar abundance of CYP46A1 and high cerebellar abundance of CYP27A1, spatial separation in the cerebellum of cholesterol/cholestanolmetabolizing P450s from a pool of metabolically available cholestanol, and weak cerebellar regulation of cholesterol biosynthesis. A new physiological role of CYP46A1, an important brain enzyme, was established and suggests that CYP46A1 could be not only a potential anti-Alzheimer's disease target but also a target for the treatment of CTX.
Human Tissues-Brain specimens were obtained during autopsy 9 -11 h after death from de-identified donors following informed consent of the respective families. The brain tissue was rinsed in cold 0.9% NaCl (w/v), blotted, and immediately flash-frozen in liquid nitrogen. Samples were stored at Ϫ80°C until further analysis. Demographic information and available medical history on donors 3 (male) and 4 (female), whose specimens were used in the present work, are described elsewhere (34). Human tissue use conformed to the Declaration of Helsinki and was qualified as "not human subjects" research by the institutional review board at the University of Texas Medical Branch.
CYP27A1 and CYP46A1 Quantifications-Half of mouse brain, whole mouse cerebellum, or a portion of human cerebellum was used for sample preparation. In each case, the tissue was added to 1 ml of 25 mM ammonium bicarbonate supplemented with 1:200 (v/v) protease inhibitor mixture (Sigma-Aldrich) and sonicated twice at 10% intensity for 10 s (Sonicator 3000, Misonix Inc.). Total protein was measured using a Pierce BCA protein assay kit (Thermo Fisher Scientific), and 8 mg of each homogenate was sampled for analysis. Samples were centrifuged at 106,000 ϫ g for 1 h at 4°C. The resulting pellet was reconstituted with 25 mM ammonium bicarbonate, 2% SDS (w/v), 20 mM DTT, 7 pmol of human 15 N-labeled CYP46A1, 3 pmol of human 15 N-labeled CYP27A1. After incubation for 1 h at room temperature for reduction of cysteine residues, samples were alkylated by treatment with 55 mM iodoacetamide for 1 h. Sequential protein precipitation with a chloroform/methanol/water mixture (1:4:3, v/v/v) and then with 4 volumes of methanol was performed (49) to separate proteins from salts, lipids, and processing reagents. Precipitated protein samples were reconstituted in 25 mM ammonium bicarbonate, 0.2% sodium cholate (w/v) and digested with a 1:8 trypsin/protein ratio (w/w) overnight at 37°C. Digested sample was quenched with 0.5% trifluoroacetic acid (v/v) and centrifuged at 106,000 ϫ g for 20 min at 4°C. Supernatant was dried using Eppendorf AG Vacufuge, and peptides were stored at Ϫ20°C. For analysis by LC-MS, dried peptides were reconstituted in 3% acetonitrile, 97% water, and 0.1% formic acid (v/v/v). Separation was performed on an Agilent Zorbax Eclipse Plus C18 RRHD column (2.1 ϫ 50 mm, 1.8-m particle) coupled to an Agilent 6490 triple quadrupole mass spectrometer with iFunnel technology for multiple reaction monitoring (MRM) analysis. Peptides were eluted at a flow rate of 200 l/min using the following gradient of solvent B in solvent A: 3% B (v/v) for 3 min, 3-30% B (v/v) in 32 min, 30 -50% B (v/v) in 5 min, and 50 to 3% B (v/v) in 3 min. Solvent A was water containing 0.1% formic acid (v/v), and solvent B was acetonitrile containing 0.1% formic acid (v/v). The acquisition method had the following parameters in positive mode: fragmentor, 380 V; cell accelerator, 4 V; collision energy, 25 V; dwell, 100 ms; electron multiplier, 500 V; and capillary voltage, 3500 V. MRM transitions were predicted using Pinpoint software (Thermo Fisher Scientific) for both unlabeled and 15 N-labeled peptides. Precursors with ϩ2 charge were selected for analysis. Transitions were screened in a digest of 15 N-labeled standard to obtain the three or four most intense transitions, which were further used for quantification (Table 2). Relative signal ratios of transitions for quantification were similar in both the 15 N-labeled standard digest and when standards were added into tissue homogenates, indicating no obvious interference from biological matrices on the quantification using selected transitions. Quantification was performed by calculating the ratio of peak areas for unlabeled biological peptides to labeled standard peptides using MassHunter software (Agilent) multiplying by the ratio of known fmol of standard to mg of total protein. MRM data from 3-4 transitions per peptide (3 peptides for CYP46A1 and 2 peptides for CYP27A1) were combined to determine the amount of each P450. Data represent mean Ϯ S.D. of the independent P450 measurements in three mice or one human donor. Statistical significance of mean differences was calculated using Student's two-tailed t test and was considered significant if p was Յ0.05.

TABLE 2 Peptides and transitions for quantification of CYP46A1 and CYP27A1
Peptide quantifications were carried out as described under "Experimental Procedures". Transitions are listed for both unlabeled, light (L) and fully 15 N-labeled, heavy (H) peptides. All precursor ions were ϩ2 charge, and product ions were ϩ1 charge with y-ion included for reference. h These peptides were used for quantifications in humans. m These peptides were used for quantifications in mice. nase) (54). Cholestanol for both enzyme assays was added from 5 mM stock in 45% aqueous 2-hydroxypropyl-␤-cyclodextrin (w/v) (HPCD). Sterols were extracted by dichloromethane (55) and analyzed by GC-MS (34).
Spectral Binding Assays-Titrations and data analysis were carried out as described (33,54). The truncated, ⌬(2-50), form of CYP46A1 was used, whose binding of substrates is similar to that of full-length enzyme (31). CYP46A1 (0.4 M) was titrated at 24°C in 1 ml of 50 mM KP i (pH 7.2), containing 100 mM NaCl and 40 g of dilauroylglycerol 3-phosphatidylcholine. Cholesterol or cholestanol was added from 0.5 mM stocks in 4.5% (w/v) aqueous HPCD. CYP27A1 (0.4 M) was titrated at 30°C in 1 ml of 50 mM KP i (pH 7.2), containing 20% glycerol (v/v) and 500 mM KCl. Cholesterol or cholestanol were added from 0.25 mM stocks in 4.5% aqueous HPCD (w/v). Apparent binding constants (K d ) were calculated using either of the following equations, in which [E] is the enzyme concentration; ⌬A is the spectral response at different ligand (sterol) concentrations [L], and ⌬A max is the maximal amplitude of the spectral response.
Statistics-Data represent mean Ϯ S.D. Comparisons of sterol measurements were made by first performing a two-way analysis of variance in the Minitab statistical software package to assess for differences based on genotype, gender, and the interaction between genotype and gender. If significant differences were found based on the interaction between genotype and gender, then the individual main effects of genotype and gender were ignored. If the interaction between genotype and gender was not significant, pairwise comparisons were made based on genotype only using Bonferroni correction when data were collapsed across genders. p values for specific pairs were adjusted further by multiplying each by K/N. K is the number of biologically appropriate pairwise comparisons, namely comparisons between (i) knock-out mice of both genders and their corresponding wild type controls; (ii) knock-out mice of one gender and their corresponding wild type control; or (iii) female and male mice of the same genotype. N represents the total number of possible pairwise comparisons and was dependent on the number of genotypes in which each sterol could be detected. Statistical significance was defined as follows: *, p Յ 0.05; **, p Յ 0.01; ***, p Յ 0.001; and ****, p Յ 0.0001.