Deep mining of oxysterols and cholestenoic acids in human plasma and cerebrospinal fluid: Quantification using isotope dilution mass spectrometry

Both plasma and cerebrospinal fluid (CSF) are rich in cholesterol and its metabolites. Here we describe in detail a methodology for the identification and quantification of multiple sterols including oxysterols and sterol-acids found in these fluids. The method is translatable to any laboratory with access to liquid chromatography – tandem mass spectrometry. The method exploits isotope-dilution mass spectrometry for absolute quantification of target metabolites. The method is applicable for semi-quantification of other sterols for which isotope labelled surrogates are not available and approximate quantification of partially identified sterols. Values are reported for non-esterified sterols in the absence of saponification and total sterols following saponification. In this way absolute quantification data is reported for 17 sterols in the NIST SRM 1950 plasma along with semi-quantitative data for 8 additional sterols and approximate quantification for one further sterol. In a pooled (CSF) sample used for internal quality control, absolute quantification was performed on 10 sterols, semi-quantification on 9 sterols and approximate quantification on a further three partially identified sterols. The value of the method is illustrated by confirming the sterol phenotype of a patient suffering from ACOX2 deficiency, a rare disorder of bile acid biosynthesis, and in a plasma sample from a patient suffering from cerebrotendinous xanthomatosis, where cholesterol 27-hydroxylase is deficient.


Abstract 23
Both plasma and cerebrospinal fluid (CSF) are rich in cholesterol and its metabolites. Here we 24 describe in detail a methodology for the identification and quantification of multiple sterols 25 including oxysterols and sterol-acids found in these fluids. The method is translatable to any 26 laboratory with access to liquid chromatography -tandem mass spectrometry. The method exploits 27 isotope-dilution mass spectrometry for absolute quantification of target metabolites. The method is 28 applicable for semi-quantification of other sterols for which isotope labelled surrogates are not 29 available and approximate quantification of partially identified sterols. Values are reported for non-30 esterified sterols in the absence of saponification and total sterols following saponification. In this 31 way absolute quantification data is reported for 17 sterols in the NIST SRM 1950 plasma along with 32 semi-quantitative data for 8 additional sterols and approximate quantification for one further sterol. 33 In a pooled (CSF) sample used for internal quality control, absolute quantification was performed on 34 10 sterols, semi-quantification on 9 sterols and approximate quantification on a further three 35 partially identified sterols. The value of the method is illustrated by confirming the sterol phenotype 36 of a patient suffering from ACOX2 deficiency, a rare disorder of bile acid biosynthesis, and in a 37 plasma sample from a patient suffering from cerebrotendinous xanthomatosis, where cholesterol 38 27-hydroxylase is deficient. 39

Introduction 40
Plasma/serum and cerebrospinal fluid (CSF) represent body fluids widely studied with an ultimate 41 goal of revealing biomarkers of disease [1][2][3][4][5][6][7]. Plasma/serum analysis by mass spectrometry (MS) can 42 prove particularly fruitful to reveal inborn errors of metabolism, especially those related to 43 cholesterol biosynthesis and metabolism [8][9][10][11][12], while analysis of CSF may have value to monitor 44 neurodegeneration [7,[13][14][15]. However, comparing data across different laboratories can prove 45 treacherous as a consequence of multiple different platforms and methods used, and differences in 46 the use of standards for quantification [4,16,17]. 47 Isotope-dilution (ID)-MS represents the most reliable methodology for the quantitative 48 measurement of lipids, including sterols, in biological samples [18,19]. Despite this, large differences 49 in inter-laboratory measurements may still occur even when using ID-MS [16,17]. These differences, 50 should, however, be minimised by the use of common isotope-labelled standards accurately 51 prepared in a suitable solvent for distribution to laboratories world-wide. In the era of "omic" 52 science, there is a drive for the quantification of multiple analytes in a single sample and this has led 53 to the development of commercial mixtures of accurately aliquoted combinations of different 54 isotope-labelled standards to allow the quantification of multiple lipids in a single analysis [1,20]. At 55 the more targeted level, a commercial kit containing a mixture of twenty different isotope-labelled 56 bile acids is now available [21]. A second challenge for the inter-laboratory comparison of 57 quantitative data is provided by the variation in the exact nature of the samples analysed and 58 compared. This problem can be overcome by the use of well documented Standard Reference 59 Materials (SRMs). 60 Here, we report the absolute quantification of 17 sterols, including oxysterols and cholestenoic acids 61 in an SRM plasma sample (NIST SRM 1950 [22, 23]) using isotope-labelled cholesterol and a recently 62 commercialised mixture of other isotope-labelled sterols. We have quantified the oxysterols as non-63 esterified free molecules and, where possible, following saponification of esters. In addition, the 64 mixture of isotope-labelled standards has been used to for the semi-quantification of 8 other sterols 65 including oxysterols and sterol-acids in plasma where authentic, but not isotope-labelled standards, 66 were available. Approximate quantification of one further sterol was made in the absence of an 67 available authentic standard. Seven other sterols were identified but not quantified, while 8 further 68 sterols were partially identified in the absence of authentic standards and were not quantified. Note, 69 here we explicitly use the terms: absolute quantification to define quantification performed against 70 an isotope-labelled surrogate of otherwise exactly the same structure, e.g. ( an isotope labelled surrogate, but in the absence of an authentic standard of the sterol to be 76 quantified i.e. 7α-hydroxy-27-norcholest-4-ene-3,24-dione (7αH,27-nor-C-3,24-diO) against 77 [ 2 H 6 ](25R)26-HC. The equivalent numbers of quantified/identified sterols in an internal quality 78 control (QC) CSF sample were: absolute quantification of 10 sterols, semi-quantification of 9 sterols 79 and approximate quantification of 3 sterols. In addition, 5 other sterols were presumptively 80 identified in the absence of authentic standards but not quantified. It should be noted, that besides 81 the sterols reported here in the SRM plasma and QC CSF, a very large number of additional 82 oxysterols and sterol-acids have been detected in samples from patients suffering from inborn errors 83 of sterol metabolism, which are quantitatively minor in samples from healthy individuals [24][25][26][27][28]. We 84 demonstrate the value of the analytical method employed by confirming the sterol phenotype of 85 two such inborn errors of metabolism i.e. ACOX2 (acyl-CoA oxidase 2) deficiency and 86 cerebrotendinous xanthomatosis (CTX), two rare disorder of bile acid biosynthesis [10,[29][30][31]. These 87 disorders highlight the value of the methodology to discriminate between diastereomers with 88 asymmetric carbons at C-24 e.g. 24S-hydroxycholesterol (24S-HC) and 24R-HC, and at C-25 e.g. 7α-89 hydroxy-3-oxocholest-4-en-(25R)26-oic acid [7αH,3O-CA(25R)] and 7αH,3O-CA(25S). Note, 90 Supplemental Table S1 provides a list of systematic names, common names and abbreviations. 91 Herts, UK). The column, SPE1, was first washed with absolute ethanol (4 mL), then conditioned with 147 70% ethanol (6 mL). The sterol extract from above in 70% alcohol (1.5 mL) was applied to the column 148 and allowed to flow at a rate of 0.25 mL/min. If necessary, flow was assisted by negative pressure at 149 the column outlet. The column flow-through was collected and combined with a column wash of 150 70% ethanol (5.5 mL). Oxysterols and sterol-acids elute in this fraction SPE1-Fr1 (7 mL, 70% alcohol). 151

Experimental
The column was washed further with 70% ethanol (4 mL) to give SPE1-Fr2. and incubated at room temperature in the dark for 2 hr, after which it was neutralised by addition of 212 800 µL of water containing 41.6 of µL glacial acetic acid (7.25 x 10 -4 mole). The mixture was then 213 ultrasonicated for 5 min and then centrifuged at 2,400 x g at 4 o C to remove any precipitated matter. 214 The solution (3 mL, 70% alcohol) was then applied to SPE1 and processed as in 2.2.3. The procedure 215 was repeated with the CSF volume increased to 200 µL and the volume of water adjusted to give a 216 final volume of 3 mL, 70% alcohol. In a further experiment, OxysterolSPLASH was replaced by 217 [ 2 H 6 ]24R/S-HC (2 ng) in ethanol. 218

Enzyme-assisted derivatisation for sterol analysis (EADSA) 219
To enhance the signal for sterol, oxysterol and sterol-acid analysis by liquid chromatography (LC)-MS 220 derivatisation strategies are often used [32][33][34][35]. Here to enhance the signal in LC -electrospray 221 ionisation (ESI)-MS we have adopted EADSA technology described in Figure 1  To remove excess derivatisation reagent the reaction mixture was subjected to a second SPE step, 234 i.e. SPE2. An Oasis HLB column (60 mg, Waters Inc) was washed with 100% methanol (6 mL), 10% 235 methanol (6 mL) and conditioned with 70% methanol (4 mL). The reaction mixture from above (3.25 236 mL, 69% organic) was loaded onto the column followed by 70% methanol (1 mL), used to rinse the 237 reaction vial. The combined eluent was diluted with water (4 mL) to 35% methanol. The column was 238 equilibrated with 35% methanol (1 mL) which was added to the diluted eluent to give 9 mL of 35% 239 methanol. This solution was re-applied to the column and the eluent diluted with water (9 mL) to 240 give 18 mL of 17.5% methanol. The column was equilibrated with 17.5% methanol (1 mL) and 241 eluents combined. The resultant solution (19 mL 17.5% methanol) was applied to the column and 242 the effluent discarded. At this point all GP-derivatised sterols including oxysterols and sterol-acids 243 are retained on the column. The column was finally washed with 10% methanol (6 mL) and GP-244 derivatives eluted in 3 x 1 mL of methanol followed by 1 mL of ethanol. Oxysterols and cholestenoic 245 acids elute in the first two 1 mL fractions (SPE2-Fr1+Fr2), and cholesterol elutes across the first three 246 1 mL fractions (SPE2-Fr1+Fr2+Fr3). For oxysterol and sterol-acid analysis equal volumes of SPE2-247 Fr1+Fr2 derived from fraction-A and from fraction-B were then combined diluted to 60% methanol 248 and analysed by LC-MS. Similarly, for cholesterol analysis, equal volumes of SPE2-Fr1+Fr2+Fr3 249 derived originally from SPE1-Fr3A and from SPE1-Fr3B were combined and diluted to 60% methanol, 250 followed by dilution by a factor of up to 1000 in 60% methanol and analysed by LC-MS. Note, in 251 100% methanol the derivatives are stable for several months when stored at -20 o C [26, 31]. 252

LC-MS with multistage fragmentation (MS n ) 253
Analysis was performed on either an Orbitrap Elite mass spectrometer equipped with an ESI probe 254 (Thermo Fisher Scientific, Hemel Hempstead, UK) with prior chromatographic separations on an 255 Ultimate 3000 LC system (Dionex, now Thermo Fisher Scientific), essentially as described previously 256 [28, 31] or on an Orbitrap IDX Tribrid mass spectrometer similarly equipped with an ESI probe and 257 linked to an Ultimate 3000 LC system. The column used was Hypersil Gold C 18 (50 x 2.1 mm, 1.9 µm, 258 Thermo Fisher Scientific). Two chromatographic gradients were employed, a 17 min gradient and a 259 35 min gradient described in [28,31]. On the Orbitrap Elite instrument three to five scan events 260 were performed: one high resolution (120,000, FWHM at m/z 400) MS scan event in the Orbitrap 261 analyser in parallel with two to four multi-stage fragmentation (MS n ) scan events in the linear ion 262 trap (LIT). Similar scan parameters were utilised on the IDX instrument. One scan event was 263 performed in the Orbitrap analyser (120,000 FWHM at m/z 400) in parallel to five scan events in the 264 ion trap. One difference between MS n scans on the Orbitrap Elite and IDX is that with the Elite all 265 m/z selection is in the LIT, while on the IDX the first m/z selection was by the quadropole mass filter. 266 Quantification was performed by stable isotope dilution or using isotope labelled structurally similar 267 compounds. 268   Figure S2A). This distorts the RIC peak of [ 2 H 7 ]24R-HC as indicated by the green arrow 299

Quantification GP-derivatised sterols including oxysterols and cholestenoic acids targeted by 373
OxysterolSPLASH 374 24S-HC, 25-HC, (25R)26-HC, 7β-HC and 5α,6β-diHC do not have a natural 3-oxo analogue and their 375 GP-derivatives are only found in fraction-A (Table 1). Cholesterol oxidase is required for their GP-376 derivatisation. 7-OC is derivatised in the absence of cholesterol oxidase, hence its quantity in plasma 377 is determined using data from fraction-B alone. 7α-HC, 7α,25-diHC, 7α,(25R/S)26-diHC, 3β,7α-378 diHCA(25R/S) and their analogous 3-oxo compounds may be present in plasma and following GP-379 derivatisation both 3β-hydroxy and 3-oxo entities are found in fraction-A but only the 3-ones are 380 present in fraction-B. Note, hydroxysteroid dehydrogenase (HSD) 3B7, the dominant enzyme that 381 converts sterols with a 3β-hydroxy-5-ene structure to a 3-oxo-4-ene in the bile acid biosynthesis 382 pathways analytes gave %CVs < 20% and all metabolites gave accuracy >70%, accuracy being is defined as the 412 agreement between actual measured concentration and that derived from eq.1. 413 As 7α,(25R/S)26-diHC is of interest in the current study, further data analysis was confined to 414 OxysterolSPLASH quantities of 1 and 0.5 units (100 µL and 50 µL) with 100 µL of plasma. The data set 415 was expanded by deconvoluting endogenous 3β-hydroxy compounds from their 3-oxo analogues by 416 simply subtracting quantities measured in fraction-B from those measured in fraction-A. This 417 provides data sets for 7α-HC, 7α,25-diHC, 7α,(25R/S)26-diHC, 3β,7α-diHCA(25R) and 3β,7α-418 diHCA(25S) separately from 7α-HCO, 7α,25-diHCO, 7α,(25R/S)26-diHCO, 7αH,3O-CA(25R) and 419 7αH,3O-CA(25S), respectively ( Table 1). The agreement in data obtained with 1 unit and 0.5 units of 420 OxysterolSPLASH was good (>80%) except for 3β,7α-diHCA(25R), where the agreement was 421 acceptable at 77%, and for two of the oxysterols that can also be formed by ex vivo autoxidation of 422 cholesterol i.e. 7-OC and 5α,6β-diHC (both 60%). 423 Standard additions. To further confirm the validity of eq. 1, a standard additions approach was 424 followed in which known amounts of unlabelled standard compounds were added, over a 5-fold 425 range, to 100 µL of plasma prior to quantification with 50 µL (0.5 units) of OxysterolSPLASH. This 426 confirmed the validity of eq.1, as in all cases R 2 >0.99 and at each concentration accuracy, as 427 determined as the % difference between the measured concentration at each level and that 428 determined by solving equation 1, was >90% (Supplemental Table S3A). The standard additions 429 experiment also allowed calculation of "apparent" extraction efficiency which is given by the 430 efficiency of extraction of the added un-labelled standard. In all cases this was >90%. 431 It is not possible to extend the calibration line to concentrations lower than those that are present 432 endogenously in an unadulterated matrix. Instead we have exploited technical dilutions of prepared 433 samples to estimate a lower limit of quantification as the lowest concentration at which the 434 measured concentration of analyte differs from the calculated concentration by less than 30% 435 (Supplemental Table S3 presumptively based on exact mass, MS 3 spectrum and retention time ( Table 1). The isotope-460 labelled standards used for each analyte were chosen based on structural similarity and are colour 461 coded in Table 1. 462 A further 7 oxysterols and sterol-acids were identified but not quantified, while 8 further sterols 463 were partially identified in the absence of authentic standards and were not quantified (see 464 Supplemental Table S1). 465

Esterified oxysterols in plasma 466
Oxysterols are found in plasma in both the non-esterified (free) and esterified forms, where a 467 hydroxy group is esterified to a fatty acyl group in a reaction predominantly catalysed by lecithin-468 cholesterol acyl transferase (LCAT (data not shown). As these latter analytes are of interest, further data analysis was restricted to 487 experiments with 100 µL of plasma and 1 or 0.5 units of OxysterolSPLASH. The agreement in analyte 488 concentrations at these two levels of standard was >80% in all cases, except for 5α,6β-diHC (66%), 489 which can be formed by ex vivo autoxidation of cholesterol during sample handling (Table 1). 490 In summary, 100 µL of plasma with either 1 or 0.5 units of OxysterolSPLASH generates reproducible 491 data for the target oxysterols and cholestenoic acids. 492

Semi-quantification of other sterols including oxysterol and sterol-acids in the absence of 493
isotope-labelled standards 494 In comparison to non-esterified oxysterols and acids, the number of analytes that can be semi-495 quantified is reduced as a consequence of the lability of the 7-hydroxy-5-ene and 7-hydroxy-4-en-3-496 one structures in strongly basic solutions and a lack of authentic isotope-labelled standards available 497 to compensate for this. The data generated is presented in Table 1. 498

Comparison of data for esterified and non-esterified sterols 499
In agreement with earlier reports, about 25% of cholesterol is present in its non-esterified form [3], 500 while levels of non-esterified side-chain hydroxycholesterols varied from about 10 -25% [3, 19] 501 ( Table 1). The % of non-esterified ring-oxidised sterols was higher, ranging from about 30% for 7-OC 502 to 96% for 7α-HCO where there is no 3β-hydroxy group available for esterification. 3β-HCA was 503 found to be essentially all in the free form; this is likely to be true for both epimers of 7αH,3O-CA 504 where the % free form was in excess of 100%. The high % can be explained by the imperfect 505 correction, even with the use of an authentic isotope-labelled standard, to account for loss of 7α-506 hydroxy-4-en-3-one analyte in strong base. 507 In summary, in addition to the 17 free sterols quantified in section 3.1.3.1, 12 sterols, including 508 oxysterols and cholestenoic acids were quantified as "total sterol" representing the sum of non-509 esterified and esterified sterols. Semi-quantitative measurements were made on a further 5 sterols. 510 First, we confirmed the linearity of eq. 1 in CSF (100 µL) using 20 µL of OxysterolSPLASH in a 517 standard addition experiment over a 5-fold concentration range (Supplemental Table S3B). All 518 analytes targeted by OxysterolSPLASH gave R 2 > 0.99, except low abundance 7α,25-diHC (R 2 > 0.98). 519

Sterols including oxysterols and sterol-acids in CSF
This experiment also provided a value for experimental accuracy (>80% in all cases), where accuracy 520 is defined as the agreement between actual measured concentration and that derived from eq.1, 521 and apparent extraction efficiency (99% -122%). Accuracy was least good for 7α,25-diHC and 522 7α,(25R/S)26-diHC where the concentration of internal standard is low and for 7α-HC that can be 523 formed ex vivo from cholesterol by autoxidation. Again, we exploited technical dilutions of prepared 524 samples to estimate a lower limit of quantification as the lowest concentration at which the 525 measured concentration of analyte differs from the calculated concentration by less than 30%. 526 Additional experiments were performed in which the volumes of CSF and OxysterolSPLASH were 527 reduced. There was consequent reduction in signal for both analytes and standard and these 528 experiments were not perused further. 529

Non-esterified sterols including oxysterols and sterol-acids in CSF 530
Using 20 µL of OxysterolSPLASH to provide the isotope-labelled standard, the 25R and 25S epimers 531 of 7αH,3O-CA can be reliably quantified from 100 µL of non-hydrolyzed CSF (%CV < 20%) and by 532 considering data in fraction-A and fraction-B, so can the individual epimers of 3β,7αH-diHCA (%CV < 533 20%, Table 2, Figure 4A & B). Increasing the volume of CSF to 200 µL gave data of similar precision. 534 Cholesterol is likewise measured by reference to added isotope-labelled standard with acceptable 535 precision (%CV <10%). We did not attempt to quantify 7α-HC, 7β-HC, 7-OC or 5α,6β-diHC in CSF, as 536 they can be formed by ex vivo autoxidation of cholesterol. Even a small degree of ex vivo 537 autoxidation will introduce major errors in quantification when the endogenous molecules are of 538 low abundance. 539 In addition to the 4 cholestenoic acids and cholesterol quantified by direct reference to isotope-540 labelled surrogates, we also obtained semi-quantitative data on another 5 sterol-acids and two 541 sterols in the absence of authentic isotope-labelled standards ( Figure 4B & C), and approximate 542 quantification on a further two sterol-acids and one oxysterol, partially identified in the absence of 543 internal standards (Table 2). 544

ACOX2 560
Mass spectrometry is an ideal method to diagnose inborn errors of cholesterol metabolism [8]. One 561 such disorder is ACOX2 deficiency [10]. ACOX2 is a peroxisomal enzyme involved in the side-chain 562 shortening of C 27 to C 24 acids as part of the bile acid biosynthesis pathways (see [49] for details of 563 metabolic pathways). Its substrates are CoA thioesters of C 27 acids with 25S-steriochemistry, which 564 themselves are derived from the corresponding CoA thioesters with 25R-steriochemistry in a 565 reaction catalysed by alpha-methylacyl-CoA racemase (AMACR) [50]. Plasma analysis of bile acid 566 precursors reveals C 27 acids rather than their CoA thioesters, hence, it is anticipated that 3β,7α-567 diHCA(25S) and 7αH,3O-CA(25S) should be elevated in plasma from patients with ACOX2 deficiency. 568 The availability of the [ 2 H 3 ]-labelled forms 7αH,3O-CA(25S) and 7αH,3O-CA(25R) allows 569 quantification using the EADSA method of these two endogenous acids and also of 3β,7α-diHCA (25S)  570 and 3β,7α-diHCA(25R) ( Table 1 and Figure 5A & 5B). In normal plasma the two 25R-epimers are 571 about three and six times more abundant than the 25S-epimers, but in plasma from the ACOX2 572 deficient patient the 25S-epimers are more abundant, confirming the biochemical phenotype of the 573 patient. It is also noteworthy that the ratio of 3β,7β-diHCA(25R) to 3β,7β-diHCA(25S) in an ACOX2 574 heterozygote is seven, while in the ACOX2 deficient patient only about two (Table 1). This suggests 575 that 3β,7β-diHCA(25S) as the Co-A thioester is a substrate for ACOX2, which provides a route to side-576 chain shortened 7β-hydroxy C 24 bile acids, usually characterised as secondary bile acids [43]. 577 ACOX2 deficiency is one of a number of peroxisomal disorders which present to differing extents 578 with cholestatic liver disease in infants and children [8,51]. It is known that 3β-hydroxy-5-ene and 3-579 oxo-4-ene C 24 acids can inhibiting the bile acid export pump [52] and we speculate that the 580 corresponding C 27 acids may similarly inhibit the export pump and contribute to infantile/childhood 581 cholestasis in peroxisomal disorders. The result is an absence of (25R)26-HC in plasma and an elevation in 7α-HCO [31,53]. This is evident 587 in Figure 5C which shows a RIC for monohydroxycholesterols and monohydroxycholestenones in a 588 plasma sample from a CTX patient. Note the absence of a peak corresponding to (25R)26-HC in the 589 RIC for monohydroxycholesterols and that 24R-HC becomes evident without the need to plot a 590 specific MRM chromatogram targeting 24R/S-HC cf. Figure  human plasma and also their quantification. A similar pattern of monohydroxycholesterols was 593 revealed upon analysis of CSF from CTX patients following hydrolysis. It is also of interest to explore 594 the RIC of 7αH,3O-CA ( Figure 5D). Surprisingly, both 25R and 25S epimers are present in the CTX 595 sample at about equal levels, in stark contrast to the situation in the NIST SRM 1950 sample where 596 the 7αH,3O-CA(25R) epimer is dominant ( Figure 5D). This finding will be discussed in more detail in a 597 future report. 598 In the present study we have "deep mind" the NIST SRM 1950 plasma in terms of sterol 620 identification and quantification (see Table 1 and Supplemental Table S1). We only report absolute 621 quantification for those sterols for which an isotope-labelled authentic standard was included. This 622 gave data for 17 sterols, with another 8 sterols semi-quantified without using an authentic isotope-623 labelled standard, while one further sterol was approximately quantified but only partially identified. 624

Discussion
In addition, 7 other sterols were identified but not quantified while 8 additional sterols were 625 partially identified. our best effort to report the figures of merit of the current methodology in terms of lower limit of 661 quantification, linearity of response, apparent extraction efficiency, accuracy and precision (see 662 Supplemental Table S3 and Table 1) . We also make our data publicly available in a data repository 663 (OFS, Center for Open Science). 664 In summary, we report here the absolute and semi-quantification of sterols, including oxysterols and 665 cholestenoic acids in NIST SRM 1950