cholesterol is an isotope-dilution ... a different internal standard, and newer column technology in the ...... (isotopically labeled) internal standards are needed to.
CLIN.CHEM.37/12, 2053-2061 (1991)
Factors Influencing the Accuracy of the National Reference System Total Cholesterol Reference Method T. Bernert, Sampson John
Jr., James
R. Akins,
Gerald
R. Cooper, Abraham K. Poulose, Gary L. Myers,
Previous comparisons between the Reference and Definiserum cholesterol have demonstrated a small but persistent positive bias in the Reference Method, averaging about + 1.6%. Here we clescnbe the results of further investigationsdesigned to better charactenze the nature of this bias. Analysis of a well-characterized model serum sample (SRM 909) suggests that more than half of the difference in cholesterol values determined by the two methods is the result of small contributions from cholesterol precursor sterols and phytosterols, which are also measured for the Reference Method. An additional significant contribution may be from cholesterol oxidation products, particularly 7-hydroxycholesterol isomers, which are active in the Liebermann-Burchard reaction. The 7-hydroxycholesterol in SAM 909, most of which appeared to be already present in the serum rather than formed dunng saponification, may account for as much as 20% of the observed difference between the methods. Contributions from other possible sources, induding impurities in the cholesterol standard and incomplete saponification of cholesteryl esters, are very small. Because the observed bias is both quite small and consistent among samples, the cholesterol Reference Method continues to meet all of the requirements generally expected for a dependable and effective Reference Method. tive Methods for measuring
AddItIonal Keyphrases: variation, source of gaschromatography/mass spectromet,y
isotope-dilution
Above-normal concentrations of serum cholesterol are a recognized risk factor in atherosclerosis and coronary heart disease. Recent emphasis on the identification, classification, and proper treatment of people with high concentrations of cholesterol in blood has increased the demand for accuracy and precision in cholesterol measurements. One of the goals of the National Cholesterol Education Program of the National Heart, Lung, and Blood Institute is to improve the accuracy of cholesterol measurements, and this has led to the establishment of the National Reference System for Cholesterol (NRS/ CHOL).1 The NRSfCHOL (1) represents a consensus stanDivision of Environmental Health Laboratory Sciences, Center for Environmental Health and Injury Control, Centers for Disease Control, U.S. Department of Health andHuman Services, Atlanta, GA 30333. ‘Nonstandard abbreviations: ALBK, Abell-Levy-Brodie--KendalI; CDC, Centers for Disease Control; GLC, gas-liquid chromatography, ID-GC/MS, isotope-dilution gas chromatography/mass spectrometry; L-B, Liebermann-Burchard; NRS/CHOL, National Reference System for Cholesterol; NIST, National Institute of Standards and Technology; SRM, Standard Reference Material; PLC, thin-layer chromatography; and TMS, trimethylsilane. ReceivedMay 30, 1991; accepted October 10, 1991.
and Eric J.
dard based on a hierarchy of reference methods and materials with an accuracy base founded on the National Institute of Standards and Technology (NIST) Definitive Method for cholesterol measurement. In this scheme, field applications of cholesterol analysis are calibrated by the Reference Method, which is itself validated relative to the Definitive Method. The currently accepted Definitive Method for serum cholesterol is an isotope-dilution gas chromatography/ mass spectrometry (ID-GC/MS) method developed and maintained by NIST (2, 3), whereas the Reference Method is the maximum extraction Abell-Levy-Brothe-Kendall (ALBK) procedure, as established and maintained by the Centers for Disease Control (CDC) (4). Traceability between the methods is maintained by using a common cholesterol primary standard for calibration [currently Standard Reference Material (SRM) 911b] and, periodically, common standard materials to compare the results obtained by the two methods. In 1977, analyses of several frozen serum pools by both the CDC Reference Method and the NIST Definitive Method resulted in values that were in close agreement. More recently, however, comparisons between the two methods with use of common materials have suggested that the Reference Method has a mean positive bias of about 1.6% relative to the Definitive Method (5). The bias is observed only in analyses of samples, because the results from the Reference Method and the Definitive Method are identical when solutions of punfled cholesterol (SRM 911b) are analyzed as unknowns (5). Although the Definitive Method was modified between the first and second comparisons to incorporate improvements in the mass spectrometer, a different internal standard, and newer column technology in the gas-chromatographic separation step (3), there is currently no evidence that the changes introduced in the Definitive Method could account for the present bias between it and the Reference Method. A small part of the difference between the two methods might be attributed to a standard impurity correction term used in the Definitive Method-but not in the Reference Method-when calculating the results. In both methods, SRM 911b cholesterol, certified by NIST as 99.8% pure, is used for calibration; however, the Reference Method does not incorporate a correction for the residual 0.2% impurities in this standard, whereas the Definitive Method does (2). Because the impurities probably include other sterols that are active in the Liebermann.-Burchard (L-B) reaction (SRM 911b Certificate of Analysis, NIST), a correction for the total CLINICALCHEMISTRY,Vol.37, No. 12, 1991 2053
0.2% in the Reference Method is probably not indicated. Moreover, the Reference Method would still retain a mean positive bias of about 1.4% even if the full 0.2% of impurities were taken into account. A recent joint CDC-NIST evaluation of the difference between the two methods (5) concluded that the changes in the Definitive Method were not likely to be the source of the bias, but the cause remained elusive. In an attempt to identif’ and evaluate the source(s) of this residual difference between the two methods, we have undertaken further studies by using an ID-GCIMS method established at CDC in comparison with the CDC ALBK Reference Method. Materials and Methods Standards and reagents. The cholesterol primary standard (SRM 911b) and human serum reference material (SRM 909) were obtained from NIST, Gaithersburg, MD. The remaining sterols and other chromatographic standards were from Sigma Chemical Co., St. Louis, Mo; Research Plus, Bayonne, NJ; or Steraloids Inc., Wilton, NH. All of the sterols were examined for purity by several chromatographic methods, and the (3-sitosterol standard, which contained significant impurities, was further purified by high-performance liquid chromatography (HPLC) before use. [25,26,27-13C3]Cholesterol was from MSD Isotopes, St. Louis, MO; [4-’4Cjcholesterol and [4-’4C]cholesteryl oleate were from New England Nuclear, Boston, MA; and the derivatization reagents were from Pierce Chemical Co., Rockford, IL. Synthesis of [4-’4C]cholesteiylpalmitate. We prepared this compound from [4-14C]cholesterol and palmitic acid by a dicyclohexylcarbodiimide coupling procedure based on the method of Sripada (6). The labeled cholesteryl ester was purified by preparative thin-layer chromatography (TLC) on silica plates, with hexane:diethyl ether: acetic acid (90:10:1, by vol) as the eluent. Radiochemical purity of this compound was >98% as judged by analytical TLC and autoradiography. Preparation of [4-’4C]cholesteiiyl oleate-labeled lipoprotein. A liposomal suspension labeled with [4-14C}cholesteryl oleate was prepared and incubated with fresh serum, essentially according to the method of Morton and Zilversmit (7), and the labeled lipoprotein fractions were then separated by density-gradient ultracentnifugation and saponified by the standard procedure. Aliquota were also extracted directly by the Folch procedure (8) without saponification and analyzed by TLC followed by autoradiography to confirm that the label remained predominately in the cholesteryl ester region. Sample preparation for ID-GC/MS. To facilitate comparisons, we used sample preparation procedures quite similar to the methods previously established at NIST for the total serum cholesterol Definitive Method (2,3). We used NIST SRM 911b cholesterol as the primary standard material and [25,26,27-13C3]cholesterol as the internal standard. Lyophilized samples (SRM 909) were reconstituted by weight according to procedure “A” defined in the SRM insert. A 500-pL aliquot of the 2054 CLINICAL CHEMISTRY, Vol. 37, No. 12, 1991
serum sample was added to a tared vial, and the sample weight was determined on a five-decimalplace balance. We measured sample densities by using a calibrated 2-mL volumetric flask. The internal standard in ethanol (100 ML) was added at a final concentration previously estimated to correspond to the expected weight of native cholesterol in the sample within ± 5%. High and low bracketing standards were prepared by weight, with use of the same internal standard solution and aliquots of a stock solution of SRM 911b in ethanol. We prepared a fresh set of internal standard stock solution and bracketing standards mixtures for each series of analyses. We added 600 pL of 8.9 mol/L KOH and 4.9 mL of ethanol to the samples, mixed, and incubated them at 37#{176}C for 3 h. We then added 5 mL of water and 10 mL of hexane to the samples, vortex-mixed them for 1 mm, and placed them on a mechanical shaker for 10 mm. After the samples were centrifuged at 560 x g for 5 mm, the hexane layer was recovered and dried over Na2504. The hexane layer was evaporated in a clean tube under a stream of nitrogen and the residue was dissolved in 1 mL of ethanol. Appropriate aliquots (generally 25-100 iL) of sample extracts or standards were dried under nitrogen, and derivatized in 100 ML of Tri-Sil BSA in pyridine. The vials were sealed, mixed, and left overnight at room temperature before analysis. Analysis by GC/MS. We carried out all cholesterol analyses on a Hewlett-Packard 5890/5970B bench-top GCIMS system with a 0.2 mm x 25 m methyl silicone column (Ultra-i; Hewlett-Packard, Avondale, PA) coated with a 0.33-Mm-thick film. The column was interfaced directly to the mass spectrometer. Injector and capillary interface temperatures were 260 and 265 #{176}C, respectively, and analyses were made isothermally at 260 #{176}C. We used a head pressure of 1000 kPa to maintain a helium flow rate of -0.6 mL/min, and the split ratio was set to 50:1. Selected-ion monitoring analyses were made by using the molecular ions of the native and labeled trimethylsilane (TMS)-ethers at m/z 458.5 and 461.4, respectively, at 2.3 Hz and an electron multiplier voltage of 1800-2000 V. To minimize the variation in source temperature resulting from filament heating, we placed the filament on/off cycle under program control and regulated the timing of injections. Under our conditions of analysis, cholesterol was eluted at about 28 mm with a (baseline) peak width of about 36 s, which allowed for >80 scans through the peak. Samples were analyzed in duplicate in sets of low standard, sample, and high standard, in that order; the set was then repeated with the high standard analyzed first, and the results were averaged. No data were discarded from this series of analyses. well-mixed
Analysis of (intact) cholesteyl esters by capillary gasliquid chromatography (GLC). Six 500-ML serum samples were saponified and extracted with hexane as described above. We pooled the hexane extracts (total
volume -60 mL), added cholesteryl heptadecanoate as an internal standard, and dried the samples on a rotary
evaporator at 35 #{176}C. Corresponding blank (60 mL of hexane) and control samples (5 Mg each of cholesteryl palmitate and cholesteryl oleate in 60 mL of hexane) were included with each analysis. The residue from each sample was dissolved in 2 mL of hexane and applied to a 3-mL silica column (J. T. Baker Inc., Marietta, GA), which was successively eluted with hexane and then diethyl ether, 50 mLfL, in hexane. The ether-hexane fraction was dried, recovered in a minimum volume of hexane, and analyzed by GLC. For these analyses we used a short methyl silicone capillary column (0.2 mm x 3.9 rn) at a helium flow rate of about 1 mL/rnin in a Hewlett-Packard 5840 GLC equipped with a flameionization detector. The oven was held at 290#{176}C, and the injector and detector temperatures were 300 #{176}C. Cholesterol oxidation prod ucts. We saponified aliquots of the serum and then prefractionated the pooledhexane extracts to remove most of the cholesterol by using either TLC with heptane:ethyl acetate (50:50, by vol) as the eluent (9) or the silica column method of Yamashita et al. (10). After adding stigmasterol as an internal standard, we dried the final eluate under nitrogen and either benzoylated the sample for HPLC or derivatized it with Tri-Sil BSA for GC/MS analysis. We analyzed the benzoylated derivatives with a 4.6 x 250 mm column of 5-Mm particles of Microsorb-ODS (Rainin Instrument Co., Woburn, MA) that was mounted in a Hewlett-Packard 1090M chromatograph equipped with a diode-array detector. Components were eluted isocratically with acetonitrile/isopropanol (60/40, by vol) at a flow rate of 1 mL/min and monitored at 230 nm. GC/MS analyses of the TMS ethers were made with a 0.2 mm x 12 m column of methyl silicone with splitless injection and selected-ion monitoring. The initial column temperature of 100 #{176}C was held for 1 mm; then the oven was ramped to 260#{176}C at 20 #{176}C/min. Injector and interface temperatures were 260 and 270 #{176}C, respectively. In these assays we monitored ions at mlz 456 (7-hydroxycholesterol), 472 (7-ketocholesterol), and 394 (stigmasterol). Additional analyses. Cholesteryl ester fatty acid proifies were determined as their methyl esters by capillary GLC, by procedures we have previously described (11). For saponification rate assays, we added the cholesteryl heptadecanoate internal standard (after neutralization) to the hexane extract. Cholesterol analyses of standards, fractions, and serum samples were based on the L-B reaction, with the CDC maximum-extraction ALBK
method
(4). Assays of sterols other than
choles-
terol were carried out by GC/MS in the same system as was used for cholesterol assays, with either a 12- or 25-rn methyl silicone column (Ultra-i), and by GLC with flame-ionization detection and either the same type of column or a 30-rn phenyl methyl silicone column (DB5; J&W, Folsom, CA) mounted in a Hewlett-Packard 5880 gas chromatograph. We used the TMS derivatives for all GLC analyses. In some cases, sterols were prefractionated by HPLC before derivatization and analysis by GLC; we made these separations with a 4.6 x 250 mm column of 5-Mm particles of octadecylsilane,
with acetonitrile/isopropanol (50/50, by vol) as the eluent at a flow rate of 1.2 mLlmin, and detection at 200 nm. Resufts To evaluate a possiblebias between the Definitive and Reference Methods for cholesterol, we first had to establish an ID-CC/MS method for comparisons as described in Materials and Methods. The chromatographic separation of cholesterol from several related sterols (all as their TMS ethers) under the conditions of our assay is shown in Figure 1. We originally used [3,4-’3C2]cholesterol as the internal standard in these assays; consequently, conditions were initially selected to provide for a baseline chromatographic separation of cholesterol and cholestanol, because the molecular ion of the latter cannot be distinguished from that of [3,4-13C2Jcholesterol at a mass resolution of less than -29000. This separation is not as essential in our current assays, however, becausewe subsequently changed the internal standard to [25,26,27-’3C3jcholesterol, which is relatively insensitive to cholestanol interference. To calibrate these assays, we used bracketing standards (12) prepared from SRM 911b for each set of analyses as described above. As an additional check on the standards preparation, we submitted aliquots of the low and high standards as unknowns for analysis by the CDC Reference Method. Results such as those given in Table 1 indicated closeagreement between the expected (gravimetric) and the observed values. These comparisons also served to establish a direct link between the calibration of the ID-GCIMS assays and the Reference Method, which facilitated comparisons between them. After implementing the procedure, we selected the well-characterized SRM 909 as a model sample for comparative evaluations. Four vials of the Reference Material were analyzed over a two-month period by our ID-CC/MS method, with the results summarized in Table 2. The overall mean (and SD) for this set of
I -/----
25
30
35
40
45
Mn Fig.1. Sterol-TMSstandardsanalyzed by capillary gas-chromatography/(fullscan)massspectrometty The identity of the numbered peaks Is as follows: (1) coprostanol (2) epicholesterol,( cholesterol,( cholestanol,(5) desmosterol, (5) lathosterol, (7) campesterol, (5) campestanol
CLINICALCHEMISTRY,Vol.37, No. 12, 1991 2055
Table 1. AnalysIs of Gravimetrically Prepared GC/MS Standards by the Reference Method chols.tsrol concn, mmol/I. Standard Expsctsd’ 0.1736 High 0.2002
Low
S dlffsrsncsb 0.12% -0.10%
Observed Dtffsrenc. 0.1738±0.0021#{176} 0.0002 0.2000 ± 0.0019 -0.0002
‘“Expected” Is the concentration expectedfor the diluted standard based on the concentration of the stock material; ‘observed” is the effective concentrationmeasured by the COC Reference Method. b Calculated as ((observed - expected)/expected)x 100. CMean ± SDof8determlnatlons.
______________________________________ Table 2. AnalysIs of SRM 909 by ID-GC/MS Cholesterol, mmol/L Allquot Wa! A 1 2
3
Rspllcats 1
Rspllcat. 2
Msan
Group avsr.g.
3.6181
3.6776 3.6543
3.6491 3.6491
3.6517
3.6413 3.6543
NA
3.6543
Wa! B 3.6336
3.6103
2 3
3.7345 3.6517
3.6698
4 Wa! C
3.5974
3.6155
3.6465 3.6077
3.6388 3.6310
3.7138 3.6284
3.6776 3.6310
3.6982
3.6957
3.6982
3.6569
3.7060
3.6827
3.6517 3.6181
3.6446
3.7267
3.6776 3.6905 3.6258
3.7060
3.6465
2 3 4
Wa!0 1 2 3 4
.
3.6388
3.6232 3.7034
3.6465
3.6724
3.6543
3.6698
3.6776 3.6776
Mean ± SD (by aliquot) = 3.6595 ± 0.027 mmol/L; CV NA not analyzed. _________________________________
=
0.74%;
n
=
15.
analyses were 3.660 (0.027) mxnolfL, with a CV of 0.74%. For these analyses, we measured each aliquot twice, using duplicate derivatized samples. When all individual data were eran’ined by a nested analysis of variance, the among-vial and among-aliquot contributions to the total variance (SD = 0.037 mmol/L; n = 29) were 0% and 6.5%, respectively. Most of the variance was apparently derived from the CC/MS part of the assay. Despite this variability observed with our benchtop instrument, we considered the overall precision of these assays to be adequate for our current purposes, and the mean value we obtained was in closeagreement with the value assigned by NIST after analysis with the Definitive Method. Aliquots of these same samples yielded slightly higher values when assayed by the Reference Method (ALBK), as has been noted previously (5). These results thus provide further confirmation of a small difference between GC/MS and Reference Method values for cholesterol, at least for SRM 909, and also suggest that direct comparisons between the results 2056 CLINICALCHEMISTRY,Vol. 37, No. 12, 1991
our CC/MS procedure and the Reference Method could be used to address the apparent bias between the Reference and Definitive Methods. Cholesteryl ester hydrolysis. One important consideration in serum cholesterol measurements is the extent of cholesteryl ester hydrolysis. Previous studies at CDC on the Reference Method, and at NIST during the development of the Definitive Method, indicated that complete saponification (defined as >99.9%) is achieved by both methods. However, saponification in the Definitive Method was examined by using pure, labeled cholesteryl oleate in the saponification medium (2), and we thought it possible that this reaction might be influenced by the presenceof serum. No such effect could be demonstrated, however, when we added [4-14CJcholesteryl oleate to incubations carried out either with or without 500 ML of serum (SRM 909). The mean percentage of total radioactivity recovered in the cholesteryl ester region after TLC and autoradiography was 0.13% (SD 0.036%) (n = 7) for incubations without serum and 0.09% (SD 0.019%) (n = 6) for incubations carried out with serum. Although the results obtained with labeled cholesteryl oleate were consistent with its complete saponification, serum cholesteryl esters are heterogeneous. Because both chemical and enzymatic alcoholysis and (or) hydrolysis rates may vary according to the composition (13, 14), we estimated the rates of hydrolysis of endogenous serum cholesteryl esters in SRM 909 from the initial time course for the reaction, with the results summarized in Table 3. if we assume that (pseudo) first-order conditions were maintained throughout the reaction, these data suggest that all of the esters might be expected to be completely hydrolyzed within 1 h at 37#{176}C. In general, the unsaturated esters were hydrolyzed quite rapidly, whereas saturated esters were hydrolyzed at a slightly slower rate. Although small, this difference suggested that incomplete saponification would be best represented by the saturated esters, particularly cholesteryl pahnitate, which makes up about 10% of the cholesteryl esters in SRM 909 (Table 4). Consequently, we synthesized [4-14C]cholesteryl palmitate and examined it in the same manner as cholesteryl oleate. However, >99.8% saponification of this ester was obtained after 3 h at 37#{176}C, and no significant difference could be detected between the extent of hydrolysis of cholesteryl palmitate or cholesteryl oleate in these analyses. We also considered the possibility that model, labeled cholesteryl esters dissolved in ethanol may be readily from
Table 3. Cholesteryl Ester Saponification
Rates
Cholisturyl ester mln’ t Time to complstlon, mln 16:0 0.2213 3.13 31 18:1 0.2502 2.77 28 18:2 0.2500 2.77 28 20:4 0.3169 2.19 22 ‘Estimated time requiredto achievecompletehydrolysis(defined as 99.9% saponified = 10 half-lives)at 37#{176}C.
Table 4. Fatty Acid Composition Cholesteryl Esters
of SRM 909
Fatty acid
S of total (by wL)
14:0 16:0 16:1n7 18:0 18:1n9 18:2n6 18:3n3
0.5±0.12 10.6±0.50 2.0 ± 0.20 0.9±0.15 19.6 ± 0.30 57.5 ± 0.25 0.3 ± 0.06 0.7±0.15 7.9 ± 0.42
20:3n6
20:4n6
accessible during saponification, whereas the endogenous cholesteryl esters initially sequestered in the lipophilic core of the lipoprotein particles might be less readily attacked by the reagent. Although the fatty acid proffle results described above suggested that endogenous cholesteryl esters were readily hydrolyzed in the assays, this question was further addressed in three ways. First, we prepared [4-’4C}cholesteryl oleate-labeled high-density and low-density lipoproteins, as doscribed in Materials and Methods. Incubations carried out with individual labeled lipoprotein species alone, with equimolar mixtures of labeled low- and highdensity lipoproteins, or with equivolume mixtures of labeled lipoprotein plus intact serum all yielded complete hydrolysis, as judged by TLC and autoradiography. Second, to pooled serum saponification extracts we added cholesteryl heptadecanoate as an internal standard, cleaned up and concentrated the extracts as described in Materials and Methods, and then analyzed the extracts directly by capillary GLC for the presence of residual intact cholesteryl esters. A blank and a standards mixture were carried through the same procedure along with each sample. Representative chromatograms from one such series of analyses are shown in Figure 2. The mean percentage of unhydrolyzed cholesteryl esters (as cholesterol) estimated in these
A
C
B
J
2
Pin Pin PAIn Fig. 2. Analysisof cholesterylesters by capillarygas-liquidchroma-
tographywith flame-Ionizationdetection A = blank;B = sample; C = control. The identity of thenumberedpeaksIs as follows: (1) cholesteryl palmitate, ( cholesteryl heptadecanoate (internal standard), ( cholesteryloieate
assays was 0.073 (SD 0.056) (n = 4), suggesting that the hydrolysis of endogenous cholesteryl esters was 99.9% complete under these conditions. Third, we fractionated saponified samples of SRM 909 by TLC as described above and recovered the cholesteryl ester region for direct analysis by the ALBK procedure, in which both free and esterified cholesterol react comparably (15, 16). The samples were combined in groups of six to provide sufficient sensitivity for these assays. On the basis of the color development observed in the cholesteryl ester region as a fraction of that recovered in the total sample, we determined that the saponification efficiency of the Definitive Method estimated in this way averaged 99.88 (SD 0.04%) (n = 4). Thus, the results of all of our studies were consistent with an essentially complete hydrolysis of cholesteryl esters in these assays. Additional sterols. Another possible contributing factor to the difference in serum cholesterol concentration obtained by the two methods could be the presence of other, noncholesterol sterols in serum. Several additional sterols could be identified in extracts of SRM 909 when they were examined by capillary GLC with flameionization detection. After confirming that no stigmasterol could be detected in SRM 909 under our conditions of analysis, we used this sterol as an internal standard for quantitative estimates by GLC and by rn-CC/MS in the selected-ion mode. Because cholestanol was not well resolved by GLC from the unusually large excess of cholesterol in these assays,we quantified cholestanol by rn-CC/MS and by HPLC as the benzoylated ester, again with stigmasterol as the internal standard. The identities of lathosterol, campesterol, and j3-sitosterol in SRM 909 extracts were confirmed by their mass spectra; cholestanol was also confirmed by full-scan mass spectrometry after a preliminary separation from cholesterol by reversed-phase HPLC. Desmosterol was present at a lower concentration in SRM 909 and was tentatively confirmed by its relative retention time in conjunction with the presence of two characteristic ions in selected-ion mode analysis. The noncholesterol sterol content of SRM 909 estimated in these analyses is summarized in Table 5, both in terms of observed concentrations and as the “effective” (cholesterol equiv-
Table 5. EstImated Sterol Content of SRM 909 Other than Cholesterol Conan, pmoVL
Sterol Cholestanol
Desmosterol Lathosterol Campesterol -Sitosterol
9.0 ± 2.1 ± 3.6 ± 8.2 ± 7.2 ±
0.57 0.23 0.29 0.30
0.12
Factor’ 0
0.5 2.2 1.0 0.6
Apparent cholesterol,
0 1.1
7.9 8.2 4.3
21.5LmoVL ‘Ratio of color developmentIn the ALBKreactionforthe indicatedsterol vs thatforcholesterol. b Cholesterolequivalents (estimatedfor the ALBK reaction) as the product of the observed sterol concentration times the response factor.
CLINICALCHEMISTRY, Vol. 37, No. 12, 1991 2057
alent) concentration. The “effective” values were calculated from response factors determined from the analy-
sis of pure standards in the L-B reaction. As indicated in Table 5, the major noncholesterol sterols in SRM 909 were present at about 30.1 tmoI/L by weight, representing a cholesterol equivalent contribution of about 21.5 Lxno]/L.
Cholesterol oxidation. We have also addressed the potential influence of cholesterol oxidation on these assays. When measuring the oxidation products, we wanted to adhere as closely as possible to the usual conditions for cholesterol analysis; therefore, we took no unusual precautions to avoid autoxidation during saponification and extraction of the serum. When we examined the pattern of total cholesterol oxidation products in sample extracts, we consistently found 7-ketocholesterol and the isomeric 7-hydroxycholesterols. These are prominent secondary autoxidation products arising from the thermal decomposition of the initial sterol hydroperoxides (17). Samples were independently analyzed by two techniques as described in Materials and Methods: by HPLC of the benzoylated derivatives and by CC/MS selected ion monitoring of the TMS ethers, with good agreement between the results. The benzoylated 7-hydroxycholesterol isomers coeluted on HPLC, but they were well resolved by GLC, which enabled us to establish that they were present in approximately equal amounts in the extracts. Preliminary fractionation of the samples by TLC resulted in slightly higher and more variable results, possibly indicative of additional deterioration during the application or development of the samples. Consequently, we used the column procedure for all quantitative estimates. We were hindered in these assays by the low concentrations and limited stability of the analytes, and also by the fact that we could not add the stigmasterol internal standard until just before derivatization. However, the results of standard addition studies suggested that the recovery of the analytes from the original hexane extracts averaged 65% overall, and we adjusted our sample results accordingly. On this basis, the content of 7-ketochclesterol in the saponified SRM 909 extracts averaged 1.7 jnnol/L, and the mean 7-hydroxycholesterol concentration (the sum of 7a- and 7k-) was about 3.2 pinol/L. When we saponified SRM 909 samples in the presence of [4-.14C]cholesteroland subjected the extracts to autoradiography, we found only a minor increase in (labeled) oxidation products compared with the nonsaponified tracer material, suggesting that most of the 7-keto- and 7-hydroxycholesterol we measured in SRM 909 was already present in the sample. if that was in fact the case, then as much as 8.1 tmol/L (-0.22%) of the cholesterol response in the Reference Method might be attributable to these oxidation products.
Discussion To investigate the bias between the Definitive and Reference Methods for serum cholesterol, we first established an ID-CC/MS method to provide a frame of 2058
CLINICAL CHEMISTRY, Vol. 37, No. 12, 1991
reference for the studies and then selected the NISTcertified SRM 909 lyophilized serum as the model sample. Although the bias is slightly smaller for SRM 909 than the mean difference observed for all pools that have been compared (1.1% vs 1.6%), we chose this material because it is very well characterized, has a documented cholesterol value determined on several occasionsby both the Definitive and Reference Methods, and has been assayed in nearly all recent studies of other proposed reference or definitive methods. SRM 909 was apparently prepared in 1979 from serum derived from fresh, heparinized plasma, which was subsequently filtered and lyophilized by the “omega” process. Although such processed and lyophilized serum may differ from “normal” fresh serum samples, we have no reason to believe that SRM 909 is significantly different from other reference materials or from fresh serum samples in its lipid composition, or in its behavior in cholesterol Reference or Definitive Method analyses. Analysis of SRM 909 by our ID-GC/MS procedure resulted in an estimated cholesterol concentration of 3.66 (SD 0.027) mmoliL, with a CV of 0.74%. The mean concentration we measured was essentially identical to the target value established by the NIST Definitive Method for this material (3.66 mmol/L at the time of our assays), although our CV was substantially greater. This imprecision was probably strongly influenced by differences in the mass spectrometric equipment: the NIST method involves a specialized, high-precision instrument. The results of our ID-CC/MS analyses of SRM 909, which were somewhat lower than those obtained by the Reference Method, were also quite consistent with the previously established bias between the Reference and (CC/MS-based) Definitive Methods. We therefore initiated studies to determine the possible factor(s) contributing to this difference. Hydrolysis of serum cholesteryl esters. One important consideration was whether the extent of saponification was complete. The hydrolysis of serum cholesteryl esters is involved in both the Reference and Definitive Methods, and we examined this step in detail, particularly in the Definitive Method, where its role is critical. Serum cholesterol is approximately 74% esterified, and complete saponification is essential in any chromatographic procedure for total cholesterol analysis, including the Definitive Method. Because the 13C-labeled internal standard is added as the free alcohol, it cannot compensate for variability or incomplete reaction during saponification. By contrast, the ALBK procedure is relatively insensitive to the extent of hydrolysis, because cholesteryl esters are readily extracted by hexane from the saponification medium, and the color development in acetic acid/acetic anhydride (unlike chloroform) is essentially the same for both free and esterified cholesterol (15,16). Because these differences suggested that incomplete saponification would be consistent with the bias observed between the Reference and Definitive Methods, we examined saponification efficiency in several ways.
The results indicated, however, that the extent of hydrolysis of serum cholesteryl esters when we used either the Reference or Definitive Method conditions was consistently >99%. In particular, results of the high-temperature capillary GLC analyses of intact cholesteryl esters provided a sensitive estimate of the residual endogenouscholesteryl ester content, and these assays invariably indicated that hydrolysis was complete. Thus incomplete saponification is unlikely to be
contributing >0.1% to the bias between the Reference and Definitive Methods. Additional sterols. Another possible source of bias is the contribution of sterols other than cholesterol that might be present in serum, and this was also examined. In addition to cholesterol, the sterols we could identifSr in SRM 909 included cholestanol, lathosterol, desmosterol, campesterol, and 13-sitosterol, all at very low concentrations. Cholestanol, a known cholesterol precursor in blood, was detected in SRM 909 at an estimated concentration of about 9.0 p.mol/L. The identity of this compound, and also of lathosterol and the two phytosterols, was confirmed by their mass spectra. However, coprostanol, a significant
cholesterol product in
the gut and a reported trace sterol in blood, was not detectable in SRM 909 in our analyses, probably reflecting the very low concentrations of this sterol in normal serum. Kinter et al. (18), who used lathosterol as an internal standard in their CC/MS analysis of serum cholesterol, reported that they detected no endogenouslathosterol in 10 representative blood samples. Conversely, several authors have found lathosterol in serum from both normal and hypercholesteremic people, in the range of about 2.6-7.8 itmol/L (19,20). Kempen et al. (19) measured total serum lathosterol in 47 healthy volunteers and determined a mean content of 0.96 ± 0.34 pmol of lathosterol per millimole of cholesterol. This would correspond to a predicted lathosterol concentration of about 3.5 .anoI/L in SRM 909, which agrees well with our measured estimate of 3.6 zmoI/L. Phytosterols are also commonly detected in blood (21, 22). However, although plant sterols are significant dietary components, they are poorly absorbed and their normal concentrations in blood remain quite low. Kuksis et al. (21) analyzed plasma lipid profiles from more than 3000 people and found that plant sterols never exceeded 1-2% of total sterols in either normal or hypercholesteremic subjects.The plasma concentrations for phytosterols reported by these authors on the basis of a detailed analysis of 35 normal individuals were 3.4 Mxnolof campesterol per millimole of cholesterol and 3.3 .imol of 13-sitosterolper millimole of cholesterol. Miettinen et al. (22) found much lower concentrations of these sterols in the serum lipids of 63 randomly selected Finnish men, with mean concentrations of campesterol and f3-sitosterol at 1.4 and 1.0 tmollmmol of cholesterol, respectively. The content of these sterols that we measured in SRM 909 was in the same general range as the measurements from those studies: 2.3 mol/mmol of
cholesterol for campesterol and 2.0 zinol/mmol of cholesterol for -sitosterol. Although 13-sitosterol is the predominant plant sterol, a slightly higher concentration of campesterol is characteristically observed in normal serum. Other phytosterols may be present in normal serum samples (including SRM 909) at very low concentrations, but the contributions of phytosterols present at trace amounts would not be expected to significantly affect the cholesterol results. We have also tentatively identified desmosterol in SRM 909 extracts, albeit at a rather low concentration. Desmosterol was detected in human serum in several older studies, but those reports were primarily qualitative. The concentration of desmosterol we estimated in SRM 909 (0.57 mo1/mmol of cholesterol) is nearly identical to the mean (0.56 &moI/mmol cholesterol, n = 63) found in a recent study by Miettinen et a!. (22). 7-Dehydrocholesterol probably is also present in SRM 909, but at a concentration too low for us to quantifr reliably under our analytical conditions. Although rather high concentrations of serum 7-dehydrocholesterol were reported by Koehler and Hill (23) in 1953 and have been widely cited, those results were based on nonspecific methodology. More recent analyses by HPLC either failed to detect any (unesterified) 7-dehydrocholesterol in serum (24) or found only trace amounts (
