International descriptive and interventional survey for oxycholesterol determination by gas- and liquid-chromatographic methods

Increasing numbers of laboratories develop new methods based on gas-liquid and high-performance liquid chromatography to determine serum concentrations of oxygenated cholesterol metabolites such as 7α-, 24(S)-, and 27-hydroxycholesterol. We initiated a first international descriptive oxycholesterol (OCS) survey in 2013 and a second interventional survey 2014 in order to compare levels of OCS reported by different laboratories and to define possible sources of analytical errors. In 2013 a set of two lyophilized serum pools (A and B) was sent to nine laboratories in different countries for OCS measurement utilizing their own standard stock solutions. In 2014 eleven laboratories were requested to determine OCS concentrations in lyophilized pooled sera (C and D) utilizing the same provided standard stock solutions of OCS. The participating laboratories submitted results obtained after capillary gas-liquid chromatography-mass selective detection with either epicoprostanol or deuterium labelled sterols as internal standards and high-performance liquid chromatography with mass selective detection and deuterated OCS as internal standard. Each participant received a clear overview of the results in form of Youden-Plots and basic statistical evaluation in its used unit. The coefficients of variation of the concentrations obtained by all laboratories using their individual methods were 58.5-73.3% (survey 1), 56.8-60.3% (survey 2); 36.2-35.8% (survey 1), 56.6-59.8, (survey 2); 61.1-197.7% (survey 1), 47.2-74.2% (survey 2) for 24(S)-, 27-, and 7α-hydroxycholesterol, respectively. We are surprised by the very great differences between the laboratories, even under conditions when the same standards were used. The values of OCS´s must be evaluated in relation to the analytical technique used, the efficiency of the ample separation and the nature of the internal standard used. Quantification of the calibration solution and inappropriate internal standards could be identified as major causes for the high variance in the reported results from the different laboratories. A harmonisation of analytical standard methods is highly needed.


Introduction
Ring and side-chain oxidized cholesterols or oxycholesterols (OCS) are essential transport forms for reverse cholesterol transport from the periphery [1,2], for cholesterol homeostasis of the human brain [3] and serve as precursor molecules for bile acid and steroid hormone synthesis [4]. Analysis of serum OCS concentrations as surrogate markers for cholesterol metabolism, degradation and excretion of cholesterol, allows evaluation of cholesterol homeostasis in the human physiology [5][6][7]. Detection of abnormal serum concentrations of OCS, combined with other indicators enables the diagnosis of certain inherited lipid storage diseases i.e. Nieman-Pick diseases [8][9][10], cerebrotendinous xanthomatosis [11] and a subtype of hereditary spastic paresis [12]. Therefore, reliable determination and quantification of OCS is essential for diagnosis and adequate patient care. During the last decades, fundamentally important understanding of the influence of OCS has been achieved with respect to development and progress of atherosclerosis [13], cardiovascular diseases [14] as well as neurodegenerative diseases i.e. Alzheimer disease [15]. Knowing that oxysterols may function as signalling molecules in cholesterol metabolism [16], makes them attractive candidates for a wide field of physiological actions. The ratio of the amount of plasma or CSF 24(S)hydroxycholesterol to 27-hydroxycholesterol is suggested to assess cholesterol metabolism in the brain, integrity of blood-brain barrieror to characterize pathological conditions in neurological disorders [17,18]  The present international OCS survey was designed with two parts taking place in the year 2013 and 2014. The first survey provides information about the variation of reported results from the different laboratories involved and was designed as "descriptive". The second survey was designed as "interventional".

Materials and Methods
The participants submitted results from three different separation and detection methods for OCS determination: capillary gas-liquid chromatography-mass selective detection using epicoprostanol (GC-MS-epi) or deuterated OCS (GC-MS-deuterium) as internal standard and high-performance-liquid-chromatography with mass selective detection (LC-MS-deuterium) using deuterated OCS as internal standard ( Table 1). The explicit values and statistical data are additionally listed in Table 2. Note, 27-hydroxycholesterol is also called (25R)26hydroxycholesterol, taking into account stereochemistry at C-25. However, 27hydroxycholesterol is the more common name and will be used here. The lyophilized probes A-D were sent by ordinary mail with no temperature or air pressure control. This includes the possibility of autoxidation of cholesterol into oxysterols, which are not exclusively formed by enzymatic cytochrome P450 supported oxidation. One example for this could be 7αhydroxycholesterol, which is formed from cholesterol by both, enzymatic and radical oxidation. In knowledge of this fact, we limited the analysis of oxysterols to 24(S)-, 27-and 7α-hydroxycholesterol, the first two of which are almost exclusively formed enzymatically [24]. [22]. The work-up procedures and analytical settings are strikingly different for each participant since each laboratory was requested to use its own specific routine method (Table   1). However, each laboratory used alkaline hydrolysis in order to deconjugate fatty acid esterified oxysterols and thus analyzing total oxysterol serum concentrations after derivatisation of the free hydroxyl groups. It should be emphasized that the individual sample work-up procedures or detection specifications cannot be reported within this article because of the anonymization of the participants. Inconsistent values are defined within the survey as values obtained from sample A and B concentrations which are impaired by different systematic errors. For example, the value for point a of sample A in Figure 1 A is slightly to high reported but the value of sample B is much too low determined. The participant failed to obtain results with the same, reproducible error. Therefore, in both samples A and B two different errors occurred and these inconsistent values must be regarded as failed. The ratios of 24(S)-hydroxycholesterol to 27-hydroxycholesterol are listed in Table 3.

Statistics
Data are given as mean ± SD. Coefficient of variation (CV), minimum and maximum values are also given in Table 2 for the first and second survey. The calculated statistics from all participants are greatly influenced by presence of single values more than 50% different from the calculated mean value of the submitted OCS concentrations. Such individual values were regarded as outliers (see Figure 1). For 7α-hydroxycholesterol selected parameters of the above-mentioned data were additionally calculated without the outlier values and reported within the manuscript ( Table 2). The statistical evaluation of the data sets was weakened by the fact that only 13 laboratories participated, and that only eight out of 13 laboratories participated in both surveys.

First survey
Nine

Second survey
Eleven laboratories participated in this second survey including four laboratories that were not included the first survey. One group that participated in the first survey was not able to participate in the second survey. Two different lyophilized serum pools (sample C and D) were analysed by each participant. In contrast to the first part of the OCS survey the second part was designed as interventional, focusing on the influence of the calibration solutions used. Each participant was thus requested to use allocated OCS stock solutions (containing 7α-, 24(S)-, 27-hydroxycholesterol, each 10 µg/mL) for the quantification of the lyophilized sample material. Data submission and notification of individual results were performed as described above.

24(S)-Hydroxycholesterol
In and values for all subgroups are listed in detail Table 2 and 3.

7α-Hydroxycholesterol
In the first survey seven different laboratories participated. In the second survey six laboratories used GC-MS-deuterium methods. Two laboratories provided results obtained by LC-MS-deuterium methods.
The results of 7α-hydroxycholesterol from the first and second survey are shown in Figure 1 E and Figure 1 F, respectively, and values for all subgroups are listed in detail in Table 2.

Ratio 24(S)-hydroxycholesterol to 27-hydroxycholesterol
Since the ratio of 24(S)-hydroxycholesterol to 27-hydroxycholesterol has reached importance in clinical diagnosis for the integrity of the blood-brain barrier and description of the state of neurological diseases we calculated this ratio for sample A to D for all participants. We accepted a difference of +/-50% in each sample. Four participants in both survey did not submit values. The average value of all participants and the identified outliers (bold) are listed in Table 3.

Discussion
The survey can be regarded as an attempt to explain the marked differences The ratio of plasma or csf 24S-to 27-hydroxycholesterol is often used as marker of bloodbrain barrier integrity or a marker for the state of neurological diseases [17,18]. We here present the absolute values of 24S-and 27-hydroxycholesterin in Table 1 A-D and the ratios of both in Table 3. Interestingly, in one participant (#11 , Table 3), we do receive unconspicous ratios in comparison to the ratios from other participants. However, in Figure 1 the absolute values of this participant appear as outliers. The reason for this could be that the levels from the standard curves for both 24S-and 27-hydroxycholesterol are equally increased or decreased. Thus, it is of importance for further publication to show the basic individual oxysterol levels before calculating ratios. In case that only the data from a standard curve for 24S-hydroxycholesterol are too high and those for 27-hydroxycholesterol are normal or too low, the ratio 24S-to 27-hydroxycholesterol could give us a wrong impression of the real state of the blood-brain barrier or a neurological disease.
Since the samples were not specifically prepared to prevent cholesterol autoxidation during the lyophilisation process, storage nor shipping, the levels of 7α-hydroxycholesterol were widely distributed in the sample. Therefore, a clear statement about the 7α-hydroxycholesterol cannot be made.
However, it must be noted that also here the use of a common standard appears to reduce the variation of the reported results and brought values together. Therefore, we conclude that even when using individual sample work-up procedures and different analytical methods similar results could be obtained when common quantified stock solutions are used.
Despite this it is evident that some laboratories do have problems with measurement consistence and/or absolute quantification. For LC-MS methods it is crucial to use internal standard compounds of the same chemical constitution co-eluting with the analyte. Therefore, corresponding deuterated oxysterols are necessary for a precise absolute quantification of OCS. In most instances, the participants in this survey did not use the isotope labelled internal standard. The results of both surveys show how extremely the reported concentrations could vary between individual laboratories.

Conclusions and outlook
Comparing the results obtained with different analytical methods and with use of common standard solutions we conclude that the present state for the quantification of oxysterols from serum or plasma samples is quite unsatisfactory. We acknowledge that our pre-analytical protocol was not optimal and that variation in quantification could equally well be a consequence of hydrolysis methods used, independent of LC-of GC-MS measurement.
We do need further attempts to harmonize our different analytical methods in order to compare OCS levels in individual subjects.    Table 3 Ratios of 24(S) to 27-hydroxycholesterol from individual laboratories as measured in the first (A, B) and second (C,D) trial.