Concentrations of bile acid precursors in cerebrospinal fluid of Alzheimer's disease patients

1. Introduction

Cholesterol has been linked to the aetiology of Alzheimer's disease (AD) for decades with the ε4 allele of apolipoprotein E gene (APOE) being the most robust genetic risk factor for sporadic AD [1], [2]. More recently cholesterol metabolism-related genes including ABCA7 (ATP binding cassette subfamily A member 7), ABCG1 (ATP binding cassette subfamily G member 1), CLU (apolipoprotein J) and SORL1 (LDLR relative with 11 ligand-binding repeats) have been classified among susceptibility loci by large genome-wide association studies (GWAS) [3], [4]. Furthermore, in 2018 Picard et al. reported that polymorphism rs2269657 of the SREBF2 gene, which codes for the protein sterol regulatory element-binding protein-2 (SREBP-2), the master transcription factor regulating cholesterol biosynthesis, showed significant dual association with late-onset AD pathological biomarkers and gene expression levels [5]. Expression levels of the rs2269657 allele of SERBF2 in frontal cortex from late-onset AD brain inversely correlated with plaque density and with age at death [5].

Cholesterol is abundant in brain (2% wet weight) with about 25% of the total body cholesterol being found in brain [6]. Cholesterol cannot cross the blood brain barrier (BBB), hence after parturition essentially all cholesterol in brain is synthesised in situ from acetyl CoA, and all cholesterol export is via metabolism [6]. The first step of cholesterol metabolism is formation of an oxysterol, an oxidised form of cholesterol, and subsequent metabolism leads to steroid hormones and bile acids [7], i.e. in the central nervous system (CNS) to neurosteroids and C₂₇ bile acids [8], [9]. In brain the dominant oxysterol is 24S-hydroxycholesterol [10], formed in neurons by oxidation of cholesterol by CYP46A1 (cytochrome P450 family 46 subfamily A member 1) [11], this can be metabolised further in CNS [12] or exported as the intact molecule over the BBB [13]. Björkhem and colleagues have suggested the balance between 24S-hydroxycholesterol and its positional isomer (25 R)26-hydroxycholesterol (common name 27-hydroxycholesterol) in brain may affect the production of beta-amyloid in brain [14]. (25R)26-Hydroxycholesterol may be formed via CYP27A1 (cytochrome P450 family 27 subfamily A member 1) mediated oxidation of cholesterol in brain or imported into brain from extracerebral sources [14], [15], [16]. (25R)26-Hydroxycholesterol is elevated in AD brain [17] and Björkhem et al. have suggested (25R)26-hydroxycholesterol may provide a missing link between hypercholesterolemia and AD [14].

While it is difficult to investigate brain from living subjects, cerebrospinal fluid (CSF), the fluid which bathes the brain, is available. Papassotiropoulos et al. and Schönknecht et al. found elevated CSF concentrations of 24S-hydroxycholesterol in AD patients, which they explained by increased cholesterol turnover during neurodegeneration [18], [19]. Surprisingly, this elevation is not evident in AD plasma [14]. CYP46A1 is normally expressed in neurons [11], but in AD is also expressed in glia cells [20]. 24S-Hydroxycholesterol reduces expression of enzymes of the cholesterol biosynthesis pathway in mouse neurons and glia, presumably by inhibiting the processing of SREBP-2 to its active form as a transcription factor, but up-regulates expression of APOE [21], [22]. ApoE is a target gene of liver X receptors (LXRα and β, NR1H3 and NR1H2), both of which are expressed in mouse brain [23], and known to be activated by oxysterols, including 24S-hydroxycholesterol [24]. (25 R)26-Hydroxycholesterol has also been found to be elevated in CSF from AD patients [25], [26].

The end products of cholesterol metabolism include bile acids and steroid hormones. Intermediates in the bile acid biosynthesis pathways have been observed in human and rodent CSF [8], [9], [12], [27], [28] and brain [29], [30], while bile acids have been found in rodent [29], [31], [32] and human brain [33]. Pan et al. found levels of taurocholic acid were reduced in AD brain [33]. When plasma was analysed they found cholic acid to be reduced in AD patients [33]. Others have found lithocholic acid to be increased in plasma from AD patients [34], while in an untargeted metabolomics study the glycine conjugates of cholic acid, deoxycholic acid and chenodeoxycholic acid have been found to be elevated in AD plasma [35].

In an effort to understand further the relationship between cholesterol metabolism and AD we have profiled the oxysterol content of CSF from AD patients, those with vascular dementia (VD) and other neurodegenerative disease (OND), and normal controls with no evidence of “normal pressure hydrocephalus” (NPH). We have extended the profile to include cholesterol and its immediate precursors and intermediates in the acidic-, neutral-, 24S-hydroxylase- and 25-hydroxylase-pathways of bile acid biosynthesis. Of the 30 sterols routinely identified in CSF, between the different patient/control groups, only one was changed in abundance in CSF, whether normalised to the cholesterol content (ng/µg-cholesterol) or measured in ng/mL, this was the bile acid precursor 7α,25-dihydroxy-3-oxocholest-4-en-26-oic acid which was reduced in concentration in CSF from the AD group.

2. Methods

3. Results

In the current study we measured the concentrations of unesterified “free” sterols, oxysterols and bile acid precursors in the absence of saponification or solvolysis. This contrasts to most other studies of sterols and oxysterols by other groups where a solvolysis step is included to hydrolyse fatty acid esters [18], [19], [25]. Data for the control sample set has been reported previously in [9]. In the absence of authentic isotope-labelled standards for cholestenoic acids, [²H₇]24(R/S)-hydroxycholesterol was used as the internal standard. This selection has previously been validated by Crick et at for some C₂₇ acids [37] and here specifically for 7α,25-dihydroxy-3-oxocholest-4-en-26-oic acid using the standard additions method where a plot of measured concentration against theoretical concentration gave a straight line with R² > 0.994.

4. Discussion

A limitation of the present study is a lack of isotope-labelled authentic standards for many of the metabolites studied, necessitating the use of structural analogues (See Supplemental Tables S2–S4). Earlier studies by Crick et al. have indicated that this is a valid approach when incorporating GP-derivatisation which equalises for structure-specific variation in ionisation efficiency by incorporating a permanent positive-charge in the analyte [37]. Crick et al. have also shown that in the absence of isotope-labelled standards correction for analyte loss during sample preparation can made by the use of standards with similar hydrophobicity [37]. Here we have confirmed this for 7α,25-dihydroxy-3-oxocholest-4-en-26-oic acid by performing a standard additions experiment, where R² for the plot of measured concentration against theoretical concentration was found to be > 0.994.

We have previously reported the presence of 7α,25-dihydroxy-3-oxocholest-4-en-26-oic acid in human CSF [9], [12], where, as here, the identification was made based on retention time, exact mass and MS³ spectra of the GP-derivative and comparison to the authentic standard which was available as a mixture of 25R- and 25S-epimers. In the present study, and those made earlier, we cannot be sure of the exact stereochemistry at C-25 of the acid found in CSF. The origin of 7α,25-dihydroxy-3-oxocholest-4-en-26-oic acid is possibly via CYP3A4 (cytochrome P450 family 3 subfamily A member 4) catalysed 25-hydroxylation after 7α-hydroxylation and before, or after, C-26-hydroxylation and -carboxylation by CYP27A1 (Fig. 7, see also Supplemental Fig. S7) [12]. This would be consistent with 7α,25-dihydroxy-3-oxocholest-4-en-26-oic acid correlating positively with 7α,25-dihydroxycholest-4-en-3-one. We have demonstrated that CYP3A4 hydroxylates 7α-hydroxycholesterol [42], while Honda et al. have shown that CYP3A4, like CH25H (cholesterol 25-hydroxylase), is a 25-hydroxylase [43]. Alternatively, the start of the biosynthesis pathway for 7α,25-dihydroxy-3-oxocholest-4-en-26-oic acid may be 25-hydroxylation of cholesterol by CH25H [44].

The correlation data presented in Fig. 6 is consistent with 7α,25-dihydroxy-3-oxocholest-4-en-26-oic acid being formed via the enzymes CYP3A4, CYP27A1, HSD3B7 and either CYP7A1 or CYP7B1 (Fig. 7, Supplemental Fig. S7). With the exception of CYP7A1, each of these enzymes is known to be expressed in human brain [8], [45], [46], [47], [48], [49], while 7α-hydroxycholesterol is known to be present in rodent brain, presumably originating in the periphery and crossing the BBB into brain or being derived from non-enzymatic oxidation of cholesterol in brain [42]. CYP27A1 the enzyme required to introduce the carboxylic acid group at C₂₇ is expressed in neurons, oligodendrocytes and some astrocytes in human brain, and in AD brain its expression is reduced in neurons but increased in oligodendrocytes, perhaps reflecting a decrease and increase in their respective cell numbers [46]. CYP3A4 is expressed in neurons, primarily localized in the soma and axonal hillock [45] and CYP7B1 is also expressed in neurons [49]. Yau et al. have shown that in AD brain the per neuron levels of CYP7B1 mRNA are reduced in hippocampul sections, suggesting a selective impairment in ability of AD brain to 7α-hydroxylate oxysterols [49]. Conversely, HSD3B7 is not expressed in neurons, only glia [48]. From the above we suggest that the reduced level of 7α,25-dihydroxy-3-oxocholest-4-en-26-oic acid in AD CSF is a consequence of a reduced synthesis, resulting from a loss of neuron numbers. Of the cholestenoic acids monitored in this work, only 7α,25-dihydroxy-3-oxocholest-4-en-26-oic acid was found to change in abundance in any patient-group, this suggests that the enzyme responsible for 25-hydroxylation, either CH25H or CYP3A4, is critically important for this effect. CH25H is expressed in human cerebral cortex and hippocamps inclusive of pyramidal neurons [48], [50]. CH25H is located on chromosome 10, close to the LIPA (lipase A, lysosomal acid type) gene which codes for a lysosomal protein which hydrolyses cholesterol esters, in a region strongly linked to AD [51]. Haplotypes in the 5′ untranslated region of CH25H are associated with the risk of AD [52], [53]. Whatever the pathway of formation, we can only speculate whether reduced synthesis of 7α,25-dihydroxy-3-oxocholest-4-en-26-oic acid directly contributes to the pathophysiology of AD, as very little is known of the biological activities of this molecule.

How 7α,25-dihydroxy-3-oxocholest-4-en-26-oic acid is metabolised is not known. It may fall into the Duane bile acid biosynthesis pathway [54] with formation of 7α,25-dihydroxy-3,24-bisoxocholest-4-en-26-oic acid and elimination of 2-oxopropanoic acid with formation of 7α-hydroxy-3-oxochol-4-en-24-oic acid and ultimately chenodeoxycholic acid or be reduced at Δ⁴ by AKR1D1 (aldo-keto reductase family 1 member D1) then at C-3 by AKR1C4 (aldo-keto reductase family 1 member C4) prior to C-24 carbonylation and elimination of 2-oxopropanoic acid (Fig. 7). Although AKR1D1 and 1C4 are usually regarded as liver specific, Mano et al. have shown that enzymes in the cytosolic fraction prepared from rat brain can convert 7α-hydroxy-3-oxochol-4-en-24-oic acid to chenodeoxycholic acid [55]. Whether these enzymes are present in human brain and can similarly act on C₂₇ acids is unknown. Further studies of CSF targeting additional metabolites are required to learn more about the metabolism of 7α,25-dihydroxy-3-oxocholest-4-en-26-oic acid.

The authentic standard of 3β,7α,25-trihydroxycholest-5-en-26-oic acid, from which 7α,25-dihydroxy-3-oxocholest-4-en-26-oic acid is synthetically derived by bacterial cholesterol oxidase treatment, was only recently custom synthesised by Avanti Polar Lipids Inc. We await synthesis of an isotope-labelled version to fully validate the current findings on a larger cohort of AD patients.

Previous studies have shown that the concentration of 7α,25-dihydroxy-3-oxocholest-4-en-26-oic acid in CSF is higher than in serum [9], and we suggest that in the healthy brain 7α,25-dihydroxy-3-oxocholest-4-en-26-oic acid provides a route for removal of cholesterol via metabolism to more a hydrophilic metabolite. Its reduced concentration in CSF of the AD patient-group suggests that this route is attenuated in this disease state. It cannot be ruled out that, like many other cholesterol-derived molecules, 7α,25-dihydroxy-3-oxocholest-4-en-26-oic acid may also have signalling properties which become diminished in the disease state. In light of the reduced level of 7α,25-dihydroxy-3-oxocholest-4-en-26-oic acid in CSF of the AD patient-group it is tempting to devise routes to enhance its formation. Although perhaps a naïve suggestion, if CYP3A4 is the necessary 25-hydroxylase this could be up-regulated through activation of the pregnane X receptor (PXR, NR1I2) or the constitutive androstane receptor (CAR, NR1I3), as CYP3A4 is a target gene of both PXR and CAR [56], [57]. PXR is activated by multiple agonists including dexamethasone, rifampicin and enzyme-inducing antiepileptics, as well as many constituents of herbal remedies, while CAR is activated by several environmental chemicals and pharmaceuticals [58], [59], [60]. In fact, the effects in human of PXR activation by rifampicin, also known as rifampin, which demonstrates good CNS penetration [61], have recently been assessed in a randomized, open, placebo-controlled crossover trial on an oral glucose tolerance test [62]. The PXR agonists elicited postprandial hyperglycemia, suggesting a detrimental role of PXR activation on glucose tolerance. Although none of the participants reached the clinical criterion of impaired glucose tolerance [62], there are obvious limitations of this approach. Alternatively, cholesterol 25-hydroxylase, CH25H, an interferon-stimulated gene, may be activated via the interferon receptor [63]. Interestingly, Merck KGaA have performed a clinical trial (ClinicalTrials.gov Identifier: NCT01075763) to evaluate interferon β-1a in the treatment of AD.

In summary, we have identified a statistically significant reduction of 7α,25-dihydroxy-3-oxocholest-4-en-26-oic acid in the CSF of the patient-group with AD. Based on LC-MS identification of cholesterol metabolites and correlations of their concentrations, we suggest mechanisms for the formation of 7α,25-dihydroxy-3-oxocholest-4-en-26-oic acid. Interestingly, one of the suggested pathways relies on CH25H, an interferon stimulated gene implicated in AD.

Appendix A. Supplementary material