ARTICLE pubs.acs.org/ac
A Reproducible and High-Throughput HPLC/MS Method To Separate Sarcosine from r- and β-Alanine and To Quantify Sarcosine in Human Serum and Urine Tamra E. Meyer,*,† Stephen D. Fox,*,‡ Haleem J. Issaq,‡ Xia Xu,‡ Lisa W. Chu,† Timothy D. Veenstra,‡ and Ann W. Hsing† † ‡
Division of Cancer Epidemiology and Genetics, National Cancer Institute, Bethesda, Maryland Laboratory of Proteomics and Analytical Technologies, Advanced Technologies Program, SAIC-Frederick, Inc., National Cancer Institute at Frederick, Frederick, Maryland ABSTRACT: While sarcosine was recently identified as a potential urine biomarker for prostate cancer, further studies have cast doubt on its utility to diagnose this condition. The inconsistent results may be due to the fact that alanine and sarcosine coelute on an HPLC reversed-phase column and the mass spectrometer cannot differentiate between the two isomers, since the same parent/product ions are generally used to measure them. In this study, we developed a high-throughput liquid chromatography mass spectrometry (LC MS) method that resolves sarcosine from alanine isomers, allowing its accurate quantification in human serum and urine. Assay reproducibility was determined using the coefficient of variation (CV) and intraclass correlation coefficient (ICC) in serum aliquots from 10 subjects and urine aliquots from 20 subjects across multiple analytic runs. Paired serum/urine samples from 42 subjects were used to evaluate sarcosine serum/urine correlation. Both urine and serum assays gave high sensitivity (limit of quantitation of 5 ng/mL) and reproducibility (serum assay, intra- and interassay CVs < 3% and ICCs > 99%; urine assay, intra-assay CV = 7.7% and ICC = 98.2% and interassay CV = 12.3% and ICC = 94.2%). In conclusion, this high-throughput LC MS method is able to resolve sarcosine from R- and β-alanine and is useful for quantifying sarcosine in serum and urine samples.
R
ecently, sarcosine, a methylated derivative of the amino acid glycine, was identified as a potential marker for prostate cancer aggressiveness in a small metabolomic study.1 Sarcosine showed a statistically significant progressive elevation in concentration levels from benign adjacent to localized and from localized to metastatic prostate cancer tissue that was confirmed in independent tissue samples. In a subsequent targeted investigation of urinary sarcosine levels, concentration levels were significantly higher in prostate cancer patients than in biopsynegative subjects, and urinary sarcosine levels performed better than serum prostate specific antigen (PSA) levels at predicting risk when PSA levels were slightly to moderately elevated (2 10 ng/mL).1 However, two subsequent studies found no significant difference in urine or serum sarcosine levels2,3 between prostate cancer cases and healthy subjects. While not completely obvious, possible reasons for the discrepant results in these studies include the small sample sizes as well as the differences in the study subjects selected, sample collection, storage, preservation additives, and sample preparation and analysis methods. These conflicting results illustrate the need to meticulously scrutinize each step in the study design when attempting to discover biomarkers for diseases such as cancer.4 Owing to the controversy surrounding the use of sarcosine as a potential biomarker for prostate cancer risk and aggressiveness, r 2011 American Chemical Society
it is critical that the analytical method utilized can resolve and accurately quantify sarcosine. Two of the three major studies to date used gas chromatography mass spectrometry (GC MS) to quantify sarcosine levels in urine.1,2 GC MS methods using single-quadrupole selected ion monitoring (SIM) often lack specificity for the analysis of complex biological matrices, such as human serum and urine. The lack of specificity is usually compensated for by using labor-intensive sample preparation.5 For example, Jentzmik et al.2 used a GC MS method that required solid-phase extraction of urine followed by derivatization, liquid liquid extraction, and solvent exchange. In contrast, LC MS with triple-quadrupole selected reaction monitoring (SRM) can yield similar sensitivity with far greater specificity for target analytes in complex biologic matrices, allowing for less intensive sample preparation, shorter analysis times, and much higher throughput.6,7 Quantification of sarcosine is further complicated by the identical molecular weights of sarcosine and alanine (89.0932 Da), which makes it difficult to distinguish between them by MS without resolving them chromatographically. In a recent report of an LC MS method used to quantify Received: April 19, 2011 Accepted: June 2, 2011 Published: June 02, 2011 5735
dx.doi.org/10.1021/ac201003r | Anal. Chem. 2011, 83, 5735–5740
Analytical Chemistry sarcosine, Jiang et al.8 were able to resolve sarcosine from proline, kynurenine, uracil, glycerol-3-phophate, and creatinine in urine by using a phenyl-hexyl column and a step gradient in 10 min. However, based on the methods used in this article, the authors may have been quantifying a mixture of alanine and sarcosine rather than sarcosine alone. The challenges of developing a method that appropriately separates sarcosine from alanine have also been noted by others9 and is a critical step to accurately quantify sarcosine for the purpose of future studies. The objectives of the present study were (a) to develop an LC MS method that can chromatographically baseline resolve sarcosine from R- and β-alanine in serum and urine; (b) to develop an assay that is sensitive, specific, quantitative, and reproducible; (c) to measure assay parameters such as intraand interassay reproducibility; (d) to study the correlation between serum and urine concentrations in paired samples from the same subject; and (e) to maximize the throughput of the method to make it amenable to future population-based epidemiological studies with large numbers of samples.
’ EXPERIMENTAL SECTION Chemicals. All solvents used in this study were analytical grade purchased from commercial suppliers. Sarcosine, [methyl-2H3]sarcosine, and R- and β-alanine were purchased from Aldrich, (Milwaukee, WI). Instrumentation. The HPLC instrument used in this study was from Agilent Technologies (Palo Alto, CA) series 1200 capillary HPLC equipped with a binary pump, online degasser, column oven, and refrigerated microautosampler coupled to a Thermo Scientific (San Jose, CA) TSQ Quantum Ultra triple quadrupole mass spectrometer through an electrospray ionization source. Sample Preparation and LC MS Assay Procedure. Serum and urine samples were stored at 80 °C until analysis time, when they were thawed at room temperature. For serum, a 100 μL aliquot was mixed with 5 μL of a 10 μg/mL aqueous solution of the internal standard, [methyl-2H3]sarcosine, in a 500 μL polypropylene microcentrifuge tube. After addition of 20 μL of a 1 N aqueous solution of trichloroacetic acid, the mixture was vortexed followed by centrifugation at 12 000g to pellet the serum protein complement. A 75 μL aliquot of the supernatant was removed to a 250 μL glass autosampler vial and mixed with 50 μL of 0.5 M sodium bicarbonate and 50 μL of a 1 mg/mL solution of dansyl chloride in acetone. Calibration standards were prepared in water instead of serum, as no sarcosine-free serum was available; otherwise treatment of the calibration standards was identical to that of the samples. Vials were capped and heated for 30 min at 65 °C to form the sarcosine-dansyl derivative. The resulting solution containing the dansylate derivative of sarcosine and [methyl-2H3]sarcosine was directly analyzed by LC MS. Chromatographic separation was accomplished using a 100 mm long by 1 mm i.d. microcolumn packed with 5 μm Aquasil (Thermo) C18 stationary phase operated at 45 °C. The mobile phase composition used was 0.1% formic acid in water (mobile phase A) and 0.1% formic acid in acetonitrile (mobile phase B) delivered at a flow rate of 65 μL/min as a linear gradient from 5% B for 2 min and then increased to 60% over 13 min for a total runtime of 15 min. Autosampler parameters included an injection volume of 5 μL and a sample tray temperature of 4 °C. The MS was operated in positive ion mode using electrospray ionization with the following general
ARTICLE
conditions: spray voltage, 3500; sheath gas pressure, 17 psi; capillary temperature, 300 °C; tube lens offset, 113; and collision gas, argon at 1.5 mTorr. Data acquisition was performed on a Dell computer running Xcaliber software (Thermo Finnigan) in selected reaction monitoring mode (SRM) recording ion currents for the following transitions: 323 m/z to 170 m/z for dansyl sarcosine and 326 m/z to 170 m/z for dansyl-sarcosine-d3. The analytical procedure for the quantification of sarcosine in urine was identical to that of serum sarcosine with the exception of the absence of 1 N trichloroacetic acid addition. Analytes were identified by comparing the retention times of the standard compound to that of the sample in the ion chromatograms from the MS/MS signal. The retention time observed for sarcosine was well resolved from its amino acid isomers R-alanine and β-alanine (Figure 1). Quantification was accomplished using the internal standard method by applying the dansyl-sarcosine/dansyl-sarcosine-d3 peak area ratio for the sample to the function obtained from a linear regression calibration plot generated from standard solution concentrations of 5 500 ng/mL versus peak area ratio. The lower limit of quantitation (LLOQ) and upper limit of quantitation (ULOQ) were 5 ng/mL and 100 mg/mL, respectively, with a limit of detection (LOD) of 0.5 ng/mL (2.5 pg on-column). Statistical analysis revealed a response linearity of R2 = 0.9998 for the line described by the equation y = 0.000465x. All samples and standards were injected in duplicate. Serum Sarcosine Reproducibility. We included overnight fasting blood samples from 10 Caucasian male volunteers between the ages of 50 and 65 years, who had at least 2 mL of serum available and no history of cancer or heart disease. Blood samples were collected by trained study personnel and processed at the NCI contracting lab within 6 h of collection. Serum derived from the blood was then stored at 70 °C until aliquoting. Six 100 μL serum aliquots were created for each subject for analysis; all serum sample aliquots from each subject were derived from the same blood draw. A total of 60 aliquots were assayed, with two aliquots per subject assayed separately on three different days. Urine Sarcosine Reproducibility. Twenty subjects that contributed samples to the serum urine correlation study had sufficient urine sample volume for multiple aliquots (at least two 100 μL aliquots) and were included in the urine reproducibility assessment. All 20 subjects contributed to the intra-assay reproducibility measures (two aliquots in one analytical run). Nine of the 20 subjects had enough sample volume for four 100 μL aliquots and contributed information to the interassay reproducibility measures (two aliquots per subject analyzed on two different days). Serum Urine Correlation. From a population-based casecontrol study of biliary tract cancers and stones in Shanghai, China,10 we selected 21 men and 21 women who did not have evidence of biliary tract cancer and who had paired serum urine samples (at least 100 μL of each) for determination of the correlation between serum and urine sarcosine levels. Overnight fasting blood samples were collected by trained nurses and sent to a central laboratory for processing within 4 h of collection. Twenty-four-hour urine samples were self-collected by subjects prior to blood collection and refrigerated at 4 °C until transfer to the central laboratory for processing. Serum and urine samples were shipped to the NCI repository by FedEx on dry ice and subsequently stored at 70 °C until thawed for aliquoting. One aliquot of serum and one to four aliquots of urine (100 μL each) 5736
dx.doi.org/10.1021/ac201003r |Anal. Chem. 2011, 83, 5735–5740
Analytical Chemistry
ARTICLE
Table 1. Intra- and Interassay Repeatability of Sarcosine Measures in Serum and Urine of Healthy Men and Womena mean CV (%)
mean ICC (%)
mean ( SD sarcosine concentration (ng/mL)b
subject
intra-assay
interassay
intra-assay
interassay
99
99
98
94e
b,c
serum sarcosine levels overall
77 ( 28
3
3
1
42 ( 2
4
5
2
51 ( 3
5
5
3
53 ( 3
1
5
4
66 ( 2
3
1
5 6
73 ( 2 81 ( 3
3 3
2 3
7
85 ( 5
5
5
8
87 ( 3
2
3
9
88 ( 2
2
1
10
142 ( 3
2
1
102 ( 76
8
12e
urine sarcosine levelsd overall
Samples Used for Intra-Assay Reproducibility f 1 2
25 ( 1 50 ( 3
4 5
NA NA NA
3
51 ( 4
8
4
56 ( 8
14
NA
5
67 ( 10
14
NA
6
71 ( 3
4
NA
7
112 ( 17
15
NA
8
133 ( 14
11
NA
9 10
137 ( 6 151 ( 3
4 2
NA NA
11
258 ( 11
4
NA
12
322 ( 9
3
NA
Samples Used for Intra-and Interassay Reproducibility g 1
32 ( 7
14
21
2
35 ( 3
4
10
3
41 ( 5
12
11
4 5
48 ( 4 92 ( 9
8 7
9 10
6
101 ( 18
3
18
7
119 ( 12
8
10
8
139 ( 15
7
11
a
CV, coefficient of variation; ICC, intra-class correlation coefficient; SD, standard deviation. b For six aliquots across three analytic runs (two aliquots per analytic run). c Subjects are all Caucasian men (N = 10). d Subjects are all Chinese men (N = 10) and women (N = 10). e N = 8. f For two aliquots in one analytic run. g For four aliquots across two analytic runs (two aliquots per analytic run).
from each study participant were used for analysis; all serum and urine aliquots from each subject were derived from the same blood or urine collection. Aliquots were transferred to the Laboratory of Proteomics and Analytical Technologies on dry ice for analysis. Height and weight were measured by study personnel and used to calculate body mass index (BMI, kg/m2). Statistical Analysis. Two measures were used to determine intra- and interassay reproducibility of the sarcosine assays: the coefficient of variation (CV) and the intraclass correlation coefficient (ICC). The CV, expressed as a percent, is calculated as 100 times the ratio of the standard deviation to the mean: CV = 100 σ/μ. Small CVs indicate less dispersion from the mean
and thus better reproducibility. The ICC, expressed as a percent, is calculated as 100 times the ratio of the variance within subjects (σR2) to the sum of variances across all samples which includes within-subject variation (σR2) as well as between-subject variation and residual variance (σε2): ICC = 100 σR2/(σR2 + σε2). In our experience, useful assays have CVs that are less than 20% and ICCs greater than 80%; better assays have ICCs greater than 90%. Pearson’s correlation was used to measure the correlation between the paired serum and urine samples. t tests were used to compare mean sarcosine concentrations by gender. For subjects with replicate serum or urine measurements, the mean of the replicate measurements of serum or urine was used for analysis. 5737
dx.doi.org/10.1021/ac201003r |Anal. Chem. 2011, 83, 5735–5740
Analytical Chemistry
ARTICLE
Figure 1. Reconstructed ion chromatograms of the ESI(+) product ions for sarcosine and alanine separated as (A) dansyl derivatives in urine, (B) native compounds in urine, (C) dansyl derivatives of standards in water, and (D) standards in water. Peak identities are as follows: 1, β-alanine; 2, R-alanine; 3, sarcosine. For parts A and C, the experimental conditions are as described in text, while those of parts B and D are as described in ref 8.
’ RESULTS AND DISCUSSION In this study, we developed a high-throughput LC MS method to quantify sarcosine in both serum and urine samples and evaluated the assay performance properties. Different columns and solvent conditions were evaluated for their ability to separate and quantify sarcosine and alanine in blood and urine specimens by LC MS. Sarcosine and alanine were not resolvable and eluted with the solvent front in most cases. One of the sets of conditions used mimicked (identical column from same manufacturer, mobile phases, and MS parameters) those published in a study by Jiang et al.8 Sarcosine and alanine coeluted using these conditions, as shown in Figure 1B,D. In our study, the compounds were derivatized with dansyl chloride. We found that derivatizing sarcosine and alanine contributed to the resolution of both compounds chromatographically due to steric differences. The advantages of dansyl chloride include ease of derivatization, creation of sterically different isomers that can be resolved by HPLC, and increased sensitivity of MS detection.11 Dansylated R- and β-alanine and sarcosine were easily resolved from each other by reversed-phase HPLC, enabling them to be independently detected and quantified using MS (Figure 1A,C). Using the LC MS method developed in this study, sarcosine levels were quantified in repeated aliquots of serum and urine samples from multiple subjects. The mean and standard deviation for sarcosine concentrations in serum and urine overall and for each subject as well as the intra- and interassay reproducibility for each specimen type are presented in Table 1. Serum sarcosine concentrations were well above the assay limit of quantitation (5 ng/mL) and ranged from 42.1 to 141.5 ng/mL (mean = 76.8 ng/mL,
Table 2. Means, Standard Deviations (SD), and Ranges for Sarcosine Concentrations in Serum and Urine among Healthy Men and Women from Shanghai, China sarcosine (ng/mL) women
men
mean (SD)
range
mean (SD)
range
serum
81 (22)
45 123
102 (36)
53 182
urine
95 (59)
34 290
139 (89)
25 323
SD = 28.1). Reproducibility of the serum assay was excellent; average intra- and interassay CVs were both 2.9% and intra- and interassay ICCs were both greater than 99%. For the urine samples, two aliquots from one individual in one analytic run were below the limit of quantitation, which may be related to freeze/thaw cycle induced precipitation. Therefore, only eight of nine subjects contributed to interassay reproducibility measures. Reproducibility of the sarcosine urine assay was also good, with average intra- and interassay CVs of 7.7 and 12.3%, respectively, and intra- and interassay ICCs both greater than 90%. There was no significant difference in reproducibility of the assay by gender (data not shown). Using information from the 42 subjects with paired urine and serum samples, the correlation between serum and urine sarcosine levels, as well as the correlation of sarcosine levels with age and BMI, was evaluated. For both serum and urine, sarcosine concentrations were higher in men than in women (serum 5738
dx.doi.org/10.1021/ac201003r |Anal. Chem. 2011, 83, 5735–5740
Analytical Chemistry
ARTICLE
Figure 2. Correlation between sarcosine levels in 42 paired serum and urine samples from healthy men and women from Shanghai, China.
P = 0.03, urine P = 0.07); mean serum concentrations were 102.3 and 80.8 ng/mL and mean urine concentrations were 138.5 and 94.8 ng/mL for men and women, respectively (Table 2). For both men and women, sarcosine levels in urine were higher than those in serum. There was a positive correlation between serum and urine sarcosine levels among women (r = 0.31, P = 0.17) but not among men (r = 0.02, P = 0.93) (Figure 2). Age was positively correlated with both serum (r = 0.40, P = 0.07) and urine (r = 0.32, P = 0.16) sarcosine levels in women but not in men (serum r = 0.02, P = 0.95; urine r = 0.05, P = 0.83). Neither serum nor urine sarcosine levels were correlated with BMI in men or women (data not shown). In our samples, serum sarcosine concentrations were higher in Chinese men than in Caucasian American men. Both serum and urine assays had good reproducibility, but the urine assay had higher intra- and interassay variability than the serum assay. Reasons for the higher variability in urine assays are unclear, but it does not appear to be related to assay sensitivity or sample concentrations, as all of the serum and all but two urine samples tested in the study were at least an order of magnitude higher than the limit of detection and well within the boundaries of the standard curve. It is also unlikely that the observed larger variability in urine is due to sample processing, since a standard protocol of mixing before aliquoting and thoroughly vortexing the urine samples before extraction was followed. Furthermore, the previous study of urine sarcosine described by Sreekumar et al.1 reported that sarcosine concentrations in urine supernatant were similar to that in urine sediment, so combining urine supernatant and sediment by mixing was appropriate. Variability in urine results may be related to the normal variability of urine samples with respect to pH, specific gravity, protein, and medications. In our study of Chinese subjects, sarcosine levels in urine were slightly higher than those in serum in both women and men. However, sarcosine concentrations in urine and serum were only weakly correlated in women and not at all in men. Sarcosine is a natural amino acid that is found in muscles and other body tissues and is ubiquitous in food items, such as egg yolks, turkey, ham, vegetables, and legumes. It has been suggested that urine sarcosine levels are mostly from renal excretion of the metabolite
because of the strong correlation with creatinine levels in the urine.2 Therefore, serum levels of sarcosine may be more related to dietary intake, while urine levels may be related to sarcosine metabolism and excretion. Our correlation results need to be verified in larger studies and other population groups. Furthermore, the correlation between sarcosine levels in commonly collected biospecimens, such as serum and urine, with those found in prostate tissue should be examined, as should determinants of sarcosine levels (e.g., diet) in biologic specimens to better understand the usefulness of sarcosine as a prostate cancer biomarker. Limitations of this study should be noted when interpreting results. We included samples from healthy subjects with large sample volumes that were conveniently available from two of our previous studies. While these samples were appropriate for reproducibility assessment, their distribution, including means and variation, of sarcosine levels may not represent levels in normal or other populations.
’ CONCLUSIONS In this study, we describe an LC MS method that can resolve sarcosine from R- and β-alanine and reliably quantify sarcosine in serum and urine. The method is sensitive (limit of detection = 0.5 ng/mL), reproducible (CV < 3% for serum and
