Me

La

b

a

AA

KSUMPBCs

1

itiaaiPcslbdmltbc

0h

Journal of Chromatography A, 1267 (2012) 162– 169

Contents lists available at SciVerse ScienceDirect

Journal of Chromatography A

j our na l ho me p ag e: www.elsev ier .com/ locate /chroma

onitoring potential prostate cancer biomarkers in urine by capillarylectrophoresis–tandem mass spectrometry

aiel C. Solimana, Yu Huia, Amitha K. Hewavitharanab, David D.Y. Chena,∗

Department of Chemistry, University of British Columbia, Vancouver, BC V6T 1Z1, CanadaSchool of Pharmacy, The University of Queensland, Brisbane, QLD 4072, Australia

r t i c l e i n f o

rticle history:vailable online 14 July 2012

eywords:arcosinerineultiple markers

rostate canceriomarker

a b s t r a c t

Current prostate cancer (PCa) diagnosis based on prostate-specific antigen (PSA) has been graduallylosing its credibility over the last decade due to contradictory results in published literature and clini-cal practice. Recently, a group of potential PCa biomarkers in urine, particularly sarcosine, was found toincrease significantly as the cancer progressed to metastasis. We report a simple, robust, and reproducibleCE–ESI-MS/MS method for the determination of sarcosine and other representative potential biomark-ers in pooled urine. The pooled urine was obtained from 20 healthy adult volunteers between the agesof 23–30 years old. A solid phase extraction (SPE) technique was optimized for maximum recovery ofsarcosine. With no derivatization step, excellent resolution between sarcosine and its isomers (�-alanine

apillary electrophoresis–tandem masspectrometry

and �-alanine) was achieved. A separate non-SPE method was also developed for quantitative determi-nation of highly concentrated urinary metabolites. CE separation was performed on a positively-charged,polyethyleneimine (PEI)-coated capillary using 0.4–2% formic acid in 50% methanol. Precision for intra-and inter-day standard addition calibration of sarcosine were found to be within 15%, whereas intra-dayprecisions for the rest of the metabolites varied from 0.03 to 13.4%. Acceptable intra-day and inter-dayaccuracies, ranging from 80 to 124%, were obtained for sarcosine and the other metabolites.

© 2012 Elsevier B.V. All rights reserved.

. Introduction

In 2011, a recorded number of 266,390 men in North Amer-ca were diagnosed of prostate cancer (PCa), and 37,820 died fromhe disease [1,2]. This year, the number of diagnoses and deathss projected to be greater than last year and will steadily increaseccording to SEER Cancer Statistics Review [1]. Prostate-specificntigen (PSA), initially believed to be a protein specific to prostate,s the first PCa biomarker approved by the FDA for diagnosingCa along with a digital-rectal examination (DRE) [3]. However,ontroversies surrounding cancer-free patients who suffered con-equences of unnecessary treatment due to their elevated PSAevels have stimulated active studies disputing the validity of PSAlood tests for PCa screening [4]. Particularly, its weakness iniscriminating between benign prostatic hypertrophy (BPH) andalignant PCa in patients with PSA levels of 4–10 ng mL−1 has

ed to several false-positive and false-negative results. Therefore,

he search for effective, non-invasive diagnostic tools and newiomarkers for PCa is of paramount importance for the wholeancer research community [5,6].

∗ Corresponding author. Tel.: +1 604 822 0878; fax: +1 604 822 2847.E-mail address: chen@chem.ubc.ca (D.D.Y. Chen).

021-9673/$ – see front matter © 2012 Elsevier B.V. All rights reserved.ttp://dx.doi.org/10.1016/j.chroma.2012.07.021

Cancer progression, accompanied by highly complex molecu-lar processes in the human body, can alter the whole metabolomeof an organ. One reason for the controversies surrounding well-recognized single-molecule biomarkers, such as PSA for PCa, is thatoften an accurate diagnosis based on a single biomarker is limitedby its inherent sensitivity and specificity. Metabolomic strategiesin studying diseases have increasingly become popular in the lastfew years [7,8]. By generating metabolomic profiles for patientsat different stages of the disease, a similar pattern for a groupof molecules can be recognized for patients in the same stage ofcancer.

Following this rationale, Sreekumar et al. [9] profiled 1126metabolites in tissue, plasma, and urine of PCa patients in orderto screen for possible biomarkers. Their multi-marker approachnarrowed down their list of potential biomarkers to 124 metabo-lites that can help to distinguish between benign, localized, andmetastatic PCa. Sarcosine, an N-methyl derivative of glycine, tookthe spotlight by displaying dramatic increase in concentration frombenign to metastatic PCa in tissue and urine. Their study wasthen immediately followed by a multitude of others who either

questioned or supported their findings [10–17]. Furthermore, sev-eral communication reviews and letters to the editor poured in[18–20] to scrutinize the contradictory results from published lit-erature. Many of these letters have criticized the ambiguity in

L.C. Soliman et al. / J. Chromatog

miaf

bcaewilpf

suOoriu

prtudptstf

2

2

pd

Fig. 1. Structures of metabolites including internal standards used in this study.

ethodology, sample collection, data analysis and lack of, orncomplete, validation process. At the time of writing our paper, thenswer to the question of whether sarcosine is or is not a biomarkeror PCa is yet to be established.

Sensitivity and specificity of a method can be vastly improvedy combining the predictive power of each marker. In this work, weonsidered it redundant to analyze thousands of small moleculesnd regenerate metabolomic profiles for PCa patients as this gru-ling task was already undertaken by Sreekumar et al. [9]. Instead,e selected six metabolites (Fig. 1) from their results includ-

ng sarcosine, l-proline, l-cysteine, l-leucine, l-glutamic acid, and-kynurenine, which showed significant increases during cancerrogression, and showed the feasibility of a multimarker strategyor prostate cancer diagnosis.

We report the first quantitative CE–MS/MS method to analyzearcosine and a group of other potential PCa biomarkers in urinesing a newly developed CE-ESI-MS interface in our group [21].thers have attempted to use CE/MS to profile the metabolomef urine [22,23]; however, here we offer a practical, robust, andeproducible method ready for prospective PCa biomarker screen-ng. This protocol can be modified to accommodate many otherrinary metabolites if necessary.

The method includes an optimized solid phase extraction (SPE)rocedure for low level analytes (e.g., sarcosine) and does notequire a derivatization step unlike existing LC/MS or GC/MS pro-ocols [24,25]. For quantifying more concentrated metabolites inrine, we demonstrate that no extraction is necessary and simpleilution of the sample is sufficient for CE–MS analysis. Clinical sam-les were not used in this study because the scope of this study waso develop and validate a reliable analytical method for the mea-urement of multimarkers for PCa. We intend to use these methodso conduct a clinical investigation in collaboration with clinicians,ollowing ethics approval.

. Experimental

.1. Chemicals and reagents

All metabolite standards with minimum of 98% purity wereurchased from Sigma Chemical Co. (St. Louis, MO). Internal stan-ards, sarcosine-D3·hydrochloride and l-methionine sulfone, were

r. A 1267 (2012) 162– 169 163

obtained from Toronto Research Chemicals Inc. (North York, ON,Canada) and Sigma Chemical Co., respectively. Formic acid (88%),NaOH, HCl, methanol (HPLC grade), and glacial acetic acid werepurchased from Fisher Scientific (Nepean, ON, Canada). Reagentgrade picric acid (98%, 35% water) for creatinine analysis was alsoobtained from Sigma Chemical Co. (St. Louis, MO). PEI coatingreagent of trimethoxysilylpropyl-modified polyethylenimine, 50%in isopropanol was purchased from Gelest Inc. (Morrisville, PA).

2.2. Preparation of standard stock solutions and buffers

Stock solutions of sarcosine (1000 �M), l-alanine (1000 �M),�-alanine (1000 �M), l-proline (1000 �M), l-cysteine (1000 �M),l-glutamic acid (2000 �M), l-kynurenine (200 �M), and l-methionine sulfone (1000 �M) were prepared in deionized 18 M�water. Sarcosine-D3 (7.78 mM) was also prepared in deionized18 M� water, divided into 5 aliquots and stored at −20 ◦C. All stocksolutions and buffers were filtered through 0.22-�m sterile, Nylonsyringe filters and stored at 4 ◦C in a refrigerator unless otherwisestated.

2.3. Urine sampling and optimized SPE extraction procedure

Urine samples were obtained from 20 healthy volunteersbetween the ages of 23 to 30 years. The pooled urine sample wasdivided into 1 mL aliquots and stored at −20 ◦C.

The solid phase extraction (SPE) procedure for urine was opti-mized (Fig. 2) for 100 �L urine sample spiked with 20 �L 0.01 mMsarcosine standard and 20 �L (0.01 mM) IS (sarcosine-D3). The opti-mized SPE procedure was as follows: a 100 �L urine sample wastreated with 350 �L of 0.1 M acetic acid and spiked with the sar-cosine standard and IS prior to extraction. A 30 mg Strata-X strongcation mixed mode cartridge from Phenomenex (Torrance, CA) wasequilibrated and conditioned with 1 mL methanol and 1 mL 0.1 Macetic acid, respectively. The treated urine sample was then loadedinto the cartridge and drained first by gravity followed by positiveair pressure application. Successive washings of 0.5 mL acetic acid(0.01 M) and 0.5 mL methanol were applied under a gentle flowof positive air pressure. The cartridge was then dried with air for1 min before eluting twice with 350 �L of methanol/ammoniumhydroxide (15:85) solution under gravity and positive air pres-sure. Eluted sample (700 �L) was concentrated using EppendorfVacufugeTM and the residue was reconstituted in 20 �L buffer. Thereconstituted sample was filtered through 0.45 �m Costar® Spin-X centrifuge tube filter (Corning, NY) at 13,200 rpm for 30 s priorCE–MS analysis.

2.4. Calibration by standard addition method

2.4.1. SarcosineCalibration standards of 0.50, 1.00, 2.50, 5.00, 10.00, and

20.00 �M sarcosine were prepared by diluting 100 �M sarcosinewith deionized 18 M� water. IS stock solution of sarcosine-D3 wasalso diluted with deionized 18 M� water to give a 10 �M workingsolution which was divided into 1 mL aliquots and stored at −20 ◦C.A calibration curve was generated by adding 20 �L of each calibra-tion standard and 20 �L IS into a 100 �L urine sample prior to SPEextraction. Blank samples were also prepared by substituting thestandard volume of 20 �L with deionized water.

2.4.2. Other metabolitesExtremely high endogenous levels of l-Pro, l-Cys, l-Leu, l-Glu,

and l-Kyn in the pooled urine sample were found to saturate the MSdetector after SPE extraction/pre-concentration step; thus, a sepa-rate method was developed for analytes that were present at highconcentrations in native urine. Calibration standard mixtures of

164 L.C. Soliman et al. / J. Chromatogr. A 1267 (2012) 162– 169

ne sam

0lpsASTbossom

2

aatccap

TP

Fig. 2. SPE optimization flowchart for pooled uri

.625–20.0 �M for l-Pro, 3.125–100 �M for l-Cys, 2.50–80.0 �M for-Leu, 3.125–100 �M for l-Glu, and 0.125–4.00 �M for l-Kyn wererepared by serial dilution in BGE from their corresponding stockolutions (Table 2). The pooled urine sample was filtered throughmicon® Ultra-15 5 kDa Centrifugal Filter Units (Billerica, MA) in aorvall swinging bucket rotor at 4000 G for 20 min at around 15 ◦C.he standard addition calibration curve was constructed by com-ining 191 �L of each standard mixture listed in Table 2 with 5 �Lf IS (200 �M) and 14 �L of filtered urine. These volumes were cho-en to yield a final dilution factor of 15 for the urine sample. Blankamples were also prepared by substituting the standard volumef 191 �L with deionized water. The concentrations of the standardixtures are listed in Table 1.

.4.3. CreatinineConcentrations of endogenous metabolites in urine can fluctu-

te between different individuals depending on water consumptionnd physiological factors. However, the formation of creatinine inhe human body is set at a relatively constant rate [26]. Hence, theoncentration of creatinine is commonly used to normalize the con-

entrations of other metabolites in urine. A Jaffe reaction [27] basedssay on a 96-well plate was employed using a saturated alkalineicrate solution to produce a yellow complex with creatinine. The

able 1reparation of standard mixtures for the five metabolites by serial dilution.

Analyte Standard mixture (�M)

1 2 3 4 5 6

l-Pro 0.625 1.25 2.5 5 10 20l-Cys 3.125 6.25 12.5 25 50 100l-Leu 2.50 5.00 10 20 40 80l-Glu 3.125 6.25 12.5 25 50 100l-Kyn 0.125 0.25 0.5 1 2 4

ple (underlined indicates optimum parameter).

samples were analyzed on a Beckman Coulter DTX880 multimodereader set at 450 nm.

2.5. Instrumentation

2.5.1. CE–MS system and softwareSome optimizations and qualitative analyses were performed

on a P/ACE MDQ capillary electrophoresis system (Beckman Coul-ter, Brea, CA) coupled with a Finnigan LCQ*DUO ion trap massspectrometer (Thermo Scientific, Waltham, MA). All calibrationexperiments were carried out on a PA800 plus capillary elec-trophoresis system (Beckman Coulter, Brea, CA) connected to anAB SCIEX API 4000 triple quadrupole mass spectrometer (AB SCIEX,Framingham, MA). Nitrogen (UHP) was used as curtain and colli-sion gas. A modified capillary cartridge was used for CE–MS andthe details of the ESI interface developed in our group have beenpublished earlier [21]. The needle tip was positioned 1–2 cm fromthe front of the heated capillary inlet or curtain plate orifice. Itis possible to control the temperature of the CE–MS cartridge byemploying a liquid coolant, but it was not used in this particularapplication because the current generated was less than 8 �A. Theheat generated in the CE operation is minimal. All data acquisi-tion, system control, and integration were performed with Analyst®

1.4.2 software (AB SCIEX, Framingham, MA).

2.5.2. Electrophoretic procedureCE separations were carried out on a 50 �m inner diameter

(I.D.) × 365 �m outer diameter (O.D.) × 85 cm length (L) fused sil-ica capillary (Polymicro Technologies, Phoenix, AZ) coated withcationic PEI. A chemical modifier was introduced through a 75 �m(I.D.) × 365 �m (O.D.) × 80 cm (L) bare fused silica capillary set at

200–250 nL min−1 flow rate. The concentrations of formic acid andmethanol, the components of background electrolyte (or separa-tion buffer), were varied methodically to give the best compromisebetween sensitivity, resolution, and analysis time. The modifier was

L.C. Soliman et al. / J. Chromatogr. A 1267 (2012) 162– 169 165

Table 2List of MRM channels and parameters used for the metabolites and their internal standards.

Analyte Q1 mass Q3 mass DP (V) CE (V) CXP (V) Dwell time (ms)

Sarcosine 90 44 40 19 3 30l-Pro 116 70 35 20 10 30l-CysCys (dimer) 241.2 152.2 35 19 10 30l-Leu 132.1 43 40 37 8 30l-Glu 148.1 84.1 45 21 7 30l-Kyn 209.2 94.1 55 19 7 30

4060

assTm

f(fi4mvc

f(−saw

2

itrpumletimcasecdo

cfq

3

3

3

s

*Sarcosine-D3 93 47

*l-Methionine sulfone 182.1 136.2

lso optimized by changing the formic acid and methanol compo-ition. When the separation buffer and modifier solution had theame composition, the most stable and sensitive ESI was achieved.herefore, the same solution was used for separation and flow rateodification throughout all the experiments.For sarcosine analysis, 0.4% formic acid in 50% methanol was

ound to give the fastest analysis time with good baseline resolutionRs > 1.5) from �-Ala and �-Ala. At the start of the day, the modi-er and separation capillaries were rinsed with BGE for 10 min at0 psi. The sample was injected at 0.5 psi for 5 s. The separation andodifier capillaries were then immersed in the buffer vials and a

oltage of −30 kV was applied. Before each injection, the separationapillary was rinsed with buffer for 3 min at 40 psi.

Simultaneous separation of other metabolites required a higherormic acid percentage (2%) with the same methanol composition50%). The sample was injected at 1 psi for 10 s and a voltage of30 kV was applied. Between injections, it was crucial to rinse the

eparation capillary at 40 psi with methanol for 5 min, air for 1 min,nd buffer for 5 min as the migration would slow down significantlyithout these steps.

.5.3. Tandem mass spectrometryThe multiple-reaction monitoring (MRM) method, which

s outlined below, was developed manually instead of usinghe instrument’s built-in auto-optimization feature to ensureeproducibility and accuracy. Separation independent MS/MSarameters were optimized by directly infusing 10 �M of individ-al analytes in background electrolyte. A Q1 scan on positive ionode was first performed over a 2 amu width of the particular ana-

yte to optimize the declustering potential (DP) and identify thexact mass of the [M+H]+ ion in the spectrum. The entrance poten-ial (10 kV) was kept at default value. The new DP value was enteredn the software before doing a product ion scan to roughly deter-

ine the two most abundant unique fragments while changing theollision energy (CE). Finally, MRM was used to optimize the CEnd collision exit potential (CXP) for a specific transition. Table 2ummarizes the final MRM method. Other optimized MS param-ters are as follows: collision activated dissociation (CAD) gas: 4;urtain gas (CUR): 10; nebulizer gas (GS1) and auxiliary gas (GS2):isabled (not applicable with the interface); IS: 4500 V; resolutionf Q1 and Q3: unit.

Cysteine readily oxidizes in air to produce a dimer known asystine. [M+H]+ of cysteine (m/z = 122) was barely detectable in theull scan mode, but the dimer (m/z = 241) was observed and used touantify cysteine.

. Results and discussion

.1. Sarcosine

.1.1. Separation from its isomersThe main objective of our study was to provide a practical

creening method for sarcosine and other potential PCa biomarkers.

19 3 3015 10 30

It is probably not widely known that sarcosine has two other iso-mers, namely �- and �-Ala. An LC/MS technique typically takes thederivatization route to separate these three compounds. Thoughunderivatized LC/MS separation has been attempted by Jiang et al.[12], they were unable to distinguish between sarcosine and �-Alasince they have identical transition channels (m/z 90 > 44).

Separation and identification of sarcosine from its isomers iscrucial for quantitative analysis as the small quantity of sarcosineis often obscured by the larger peaks of its isomers in biologicalsamples. The total ion electropherogram and extracted ion elec-tropherograms of native urine sample are shown in Fig. 3a. Thestandards (Fig. 3b) show clearly that �-Ala has a unique transition(m/z 90 > 72) in addition to well-resolved electrophoretic separa-tion from sarcosine and �-Ala. On the other hand, the most sensitivetransition for sarcosine is identical to �-Ala, giving the same m/zfragments. Since endogenous level of �-Ala in urine is much higherthan that of sarcosine’s, the two must be electrophoretically sepa-rated for quantitative purposes.

3.1.2. SPE recoveryTo achieve maximum SPE efficiency, we studied different vari-

ables outlined in the flow chart shown in Fig. 2. A significant partof our method development was spent on sensitivity enhancementin detecting endogenous sarcosine in urine. This depends not onlyon the sensitivity of the MS instrument available to us but also onthe efficiency of the extraction procedure. For this reason, the LCQion trap MS was used for optimizing separation conditions (e.g.,buffer selection) as it was incapable of detecting endogenous sar-cosine due to its poor sensitivity. The final method was eventuallytransferred into the more sensitive triple quadrupole MS for MRManalysis.

In order to calculate the recovery or efficiency of an extractionprocedure, a blank matrix, a urine sample that does not contain theanalyte of interest, is necessary. Because sarcosine is endogenous inurine some laboratories choose simulated or synthetic urine. How-ever, we decided to use real samples and deuterated sarcosine toestimate the recovery of sarcosine. Deuterated sarcosine (methyl-D3) and a second IS (l-methionine sulfone) were spiked into thepooled urine, and calculations were based on the ratio of the areasof extracted and non-extracted D3 over the area of l-methioninesulfone. The recovery of the optimized SPE procedure was found tobe 94.4 ± 7.4% (n = 3).

3.1.3. Precision, accuracy and linearityIntra-day precision and accuracy were evaluated by analyzing

five individually extracted samples of six different concentrationson the same day. The peak area ratio of sarcosine to D3 was plottedagainst the added concentration of sarcosine. The endogenous levelof sarcosine was determined from the x-intercept of the calibration

curve. Concentration of the original 100 �L aliquot was calcu-lated by multiplying the x-intercept to the pre-concentration factorof 5. Conversely, inter-day precision and accuracy were deter-mined by analyzing three individually extracted samples of each

166 L.C. Soliman et al. / J. Chromatogr. A 1267 (2012) 162– 169

Fig. 3. Representative total ion electropherograms (TIE) and extracted ion electropherograms of (a) native urine and (b) 0.01 mM standard mix. BGE and modifier, 0.4% formicacid, 50% methanol; modifier flow rate, 200 nL min−1; sample in buffer injected at 0.5 psi for 5 s; CE inlet, −30 kV; needle, 4.5 kV; MS/MS conditions, in Table 2 and Section2.5.3.

Fig. 4. Extracted ion electropherograms of (a) 100 �L SPE native urine sample and (b) non-SPE 15-fold diluted native urine sample. BGE and modifier, 2.0% formic acid, 50%methanol; modifier flow rate, 200 nL min−1; sample in buffer injected at 1 psi for 10 s; CE inlet, −30 kV; needle, 4.5 kV; MS/MS conditions, in Table 2 and Section 2.5.3.

Table 3Inter-day (n = 5) and intra-day (n = 3) precision and accuracy of peak area ratio and recovery of sarcosine by CE–ESI-MS/MS.

Added concentration (�M) Peak area ratio (RSD %) Intra-day Inter-day

Intra-day Inter-day Found (�M) Accuracy (%) RSD % Found (�M) Accuracy (%) RSD %

0.00 3.58 7.00 – – – – – –0.50 0.43 0.93 0.40 80 5.85 0.62 124 8.431.00 1.92 1.63 0.91 91 12.5 0.96 96 10.12.50 4.78 – 2.51 100 14.3 – – –5.00 4.73 5.23 5.82 116 8.78 5.26 105 10.2

10.0 3.69 4.39 10.5 105 5.44 10.1 101 6.5620.0 1.60 – 19.5 98 2.01 – – –

L.C. Soliman et al. / J. Chromatogr. A 1267 (2012) 162– 169 167

Table 4Summary of results for inter-day (n = 5) and intra-day (n = 3) calibration of sarcosine in pooled urine.

Validation parameters

Regression equationa LinearIntra-day y = (1.72E−02 ± 2.73E−04)x + (8.59E−02 ± 2.23E−03) R2 = 0.9951Inter-day y = 1.81E−02x + 8.19E−02 R2 = 0.9990

Limit of detection (LOD, �M) 0.0176Linear range (�M) 0.50–20.0SPE efficiency (n = 3) 94.4 ± 7.4%Endogenous sarcosine concentration in100 �L urine sample prior SPE(�M)

0.99 ± 0.04b or 0.10 ± 0.01 �mol/mmol of creatinine<0.329; 0.476c �mol/mmol of creatinine

a y and x stand for peak area ratio (sarcosine/IS) and concentration in �M respectively.b Endogenous concentration calculated from the x-intercept of the intra-day standard addition graph; standard deviation of x-intercept calculated by SD =

s√ ∑

2 ts; y

o

ccbasGrlisnteooupss

wtiaotw[

3

3

ttato∼etvirttphs

y/|m| (1/n + (y2/m2 (xi − x) ), where m is the slope; n is the # of data poin

f y.c Normal range for healthy adults (>18 years) based on literature [29–31].

oncentration for three consecutive days. Inter-day results werealculated using the same calibration curve from intra-day cali-ration. Precision and accuracy values for intra-day and inter-dayre presented in Table 3. Variations in inter- and intra-day mea-urements are all less than 15% which conform to the Internationaluideline for Bioanalytical Method Validation [28]. Similarly, accu-acy values are within 15% of the added concentration except for theowest standard, 0.50 �M, found at 80% and 124% for intra-day andnter-day, respectively. The anomaly is explained by the calibrationtarting to deviate from linearity below 0.50 �M. The deviation isot a serious concern to the overall method for two reasons: first,he endogenous concentration of sarcosine in the pre-concentratedxtracted urine is 4.93 ± 0.19 �M which is greater by one orderf magnitude than the lower limit of quantitation 0.50 �M; sec-nd, according to Sreekumar et al. [9], sarcosine concentration inrine was found to increase significantly during prostate cancerrogression. Therefore, the limit of detection and calibration forarcosine are adequate for concentrations expected in a clinicaltudy (Table 4).

Linearity of the method was assessed by spiking pooled urineith 0–20 �M sarcosine standards, spanning the range of 0.1–4

imes the estimated endogenous concentration. The intra-day andnter-day calibration curves have good linear fits with R2 = 0.9951nd 0.9990 respectively. The creatinine normalized concentrationf sarcosine in pooled urine in the original 100 �L sample was foundo be 0.10 ± 0.01 �mol/mmol of creatinine which is in agreementith the normal range reported by the Human Metabolome Database

29–31].

.2. Other metabolites

.2.1. Sample preparationl-Pro, l-Cys, l-Leu, l-Glu, and l-Kyn were all found to saturate

he MS detector after SPE. The SPE procedure has a preconcen-ration factor of 5: from the initial sample volume of 100 �L, to

final reconstitution volume of 20 �L. The sample was then fur-her diluted by 20-fold. The estimated endogenous concentrationsf l-Pro, l-Cys, l-Leu, l-Glu, and l-Kyn in urine are ∼38, ∼416,107, ∼331, and ∼11 �M respectively. This estimation was nec-ssary to roughly determine the amount of each analyte in urine,hus allowing the selection of relevant concentration range for thealidation process. The estimation was done by a single-point cal-bration using a blank and a spiked urine sample. At this point, weealized that it may not be necessary to extract and preconcentratehe analytes as they are already present in very high concentra-

ions in native urine. Although we were initially concerned aboutossible matrix effects when bypassing SPE, we realized that urineas a less complex matrix compared to other biological samplesuch as plasma and tissue. In addition, the rinsing and conditioning

is the mean value of y; x is the mean value of x; and sy is the standard deviation

procedure between runs developed for this work is capable of elim-inating the adverse effect from analyzing these samples. With thisrationale, we omitted SPE and filtered the urine samples using 5 kDamolecular cutoff filters before further diluting 15-fold in buffer.Using similar single-point calibration, we roughly estimated theendogenous levels of l-Pro, l-Cys, l-Leu, l-Glu, and l-Kyn to beapproximately ∼48, ∼533, ∼186, ∼256, and ∼11 �M, respectively.These values seem to agree with the previous estimated concen-trations, although the majority of them are higher. However, sincethese values were only estimated from a single-point calibration,they are not as accurate as the endogenous concentrations deter-mined by a full validation summarized in Tables 5 and 6. Theconcentrations in Table 5 are the final concentrations of the ana-lytes after dilution. We are confident that the recovery and accuracyobtained using this method are greater than otherwise acquiredusing SPE, since they did not undergo extraction and were analyzeddirectly.

3.2.2. CE–MS/MS analysisThe only drawback we found for eliminating SPE was that we

had to perform additional rinsing steps in the CE procedure toensure reproducibility of migration times. The PEI coated capil-lary provided fast analysis times (∼15 min), as depicted in Fig. 4,without the need for pressure assistance. The analytes were notcompletely resolved in the total ion electropherograms (TIEs), butindividual MRM channels were checked for “cross talk” betweenanalytes. Complete separation and identification of all peaks werecarried out for channels that contain more than one peak such asl-Leu. The more sensitive but common transition (m/z 132 > 86)was not used for l-Leu to minimize isobaric interferences fromisoleucine and allo-isoleucine. Other isobars, 4-hydroxy-l-prolineand �-aminocaproic acid, were fully resolved from l-Leu in thechannel of m/z 132 > 43, which was used in this method.

The PEI coating minimizes adsorption of positively charged pep-tides and proteins to the capillary wall. Yet we discovered that witha buffer rinse alone, the migration times were delayed considerablyafter each diluted urine sample was analyzed. Various combina-tions of rinsing solvents/solutions were examined. The sequencethat gives the most reproducible results is methanol, air, and bufferrinses. These series of rinses must be performed between each anal-ysis of non-SPE diluted samples to condition the coating and flushunwanted molecules that are adsorbed to the capillary walls.

3.2.3. Method validation: precision and accuracyA standard addition method was employed for calibrating the

five metabolites in diluted urine sample. Endogenous amountsof each analyte were roughly estimated first and the value ofcalibration standards were adjusted so that the endogenous con-centrations are in the range of the chosen standards. A summary of

168L.C.

Soliman

et al.

/ J.

Chromatogr.

A 1267 (2012) 162– 169

Table 5Intra-day precisions (n = 3) and accuracies of five metabolites by standard addition method of diluted urine. Final concentrations are calculated based on dilutions made during sample preparation.

l-Pro l-Cys l-Leu l-Glu l-Kyn

Concentration(�M)

Accuracy(%)

RSD % Concentration(�M)

Accuracy(%)

RSD % Concentration(�M)

Accuracy(%)

RSD % Concentration(�M)

Accuracy(%)

RSD % Concentration(�M)

Accuracy(%)

RSD %

0.57 90 5.72 2.84 111 3.54 2.27 104 5.10 2.84 92 12.4 0.11 112 13.41.14 103 5.31 5.68 115 14.0 4.55 103 3.71 5.68 104 1.50 0.23 98 7.652.27 103 1.91 11.4 100 2.32 9.1 101 0.92 11.4 100 0.66 0.45 98 3.474.55 103 1.50 22.7 101 1.85 18.2 101 2.20 22.7 101 4.61 0.91 103 1.799.10 100 0.27 45.5 95 0.15 36.4 98 0.03 45.5 100 0.29 1.82 100 0.56

18.2 100 0.10 91.0 101 0.39 72.8 100 0.32 91.0 100 0.52 3.64 100 0.52

Table 6Summary of intra-day (n = 3) results for non-SPE sample validation.

Calibration curve R2 LOD (�M) Concentration of pooledurine ± SD (�mol/mmol ofcreatinine)a,b

Literature value (�mol/mmolof creatinine)c

Linear range(�M)d

l-Pro y = (1.67 ± 0.01)x + (0.790 ± 0.045) 0.9998 0.134 0.74 ± 0.06 0.86 0.63–20.0l-CysCys y = (0.098 ± 0.001)x + (0.448 ± 0.030) 0.9989 0.034 7.12 ± 0.69 10.5 3.13–100l-Leu y = (0.125 ± 0.001)x + (0.233 ± 0.012) 0.9998 0.131 2.91 ± 0.24 3.57 2.50–80.0l-Glu y = (0.478 ± 0.002)x + (0.813 ± 0.064) 0.9998 0.116 2.66 ± 0.28 1.39 3.13–100l-Kyn y = (0.492 ± 0.002)x + (0.070 ± 0.003) 0.9997 0.010 0.22 ± 0.02 0.41 0.13–4.00

a Endogenous concentration calculated from the x-intercept of the standard addition graph; standard deviation of x-intercept calculated by SD = sy/|m|√

(1/n + (y2/m2∑

(xi − x)2).b Standard deviations (SDs) calculated using relative uncertainties of x-intercepts and creatinine concentration.c Values obtained from Refs. [29–31].d Reported linear range corresponds to concentration of added standards before dilution without taking into account the endogenous metabolite concentrations.

atog

ra1es

3

otidagltlr

ftmcM

3

b1bpasR

btsfof

4

stevpsnaSbhdcboob

[

[

[[

[[

[

[

[[[[

[

[

[[

[

[[

[

[

MacInnis, A.M. Weljie, R. Dowlatabadi, F. Bamforth, D. Clive, R. Greiner, L. Li,

L.C. Soliman et al. / J. Chrom

esults is shown in Table 5, which demonstrates excellent precisionnd accuracy values in all cases. Accuracy values range from 90 to12%, while precision variations are <13.38% in which the great-st variation and lowest accuracy occurs at the least concentratedtandard for each analyte.

.2.4. LOD, LLOQ, and linear rangeLOD values reported in Tables 4 and 6 were calculated based

n the concentration that will give a peak height of three timeshe noise level (S/N = 3). S/N calculation is based on an algorithmn the Analyst® software where the adjacent region of the peak isefined as the noise level. The dynamic range was tested on neatqueous solutions and urine sample. The calibration demonstratesood linearity for all metabolites over the validation standards. Theower limit of quantitation (LLOQ), defined as the lowest concen-ration that is within ±20% RSD, is most likely to be lower than theeast concentrated standard for each analyte as indicated by theirespective RSD values (<15%).

The endogenous concentration of each analyte was calculatedrom the calibration curve. They were then normalized over crea-inine concentration as it is the standard practice when expressing

etabolite concentrations in urine. Our experimental endogenousoncentrations agree with existing values obtained from the Humanetabolome Project database [29–31].

.3. Stability

Bench-top stability of SPE processed samples was determinedy analyzing a sarcosine-D3 spiked urine blank over the course of

h, 12 h, and 24 h at room temperature. Sarcosine was found toe stable for 12 h at room temperature with little variation in theeak area ratio. However, after 24 h at room temperature, the signalnd calculated concentration decreased by 25%. For freeze–thawtability, sarcosine was found to be stable over three FT cycles (<7%SD).

Stability of non-SPE diluted samples was assessed differentlyecause they did not go through SPE. Therefore, they have to beested under more rigorous conditions. Freshly diluted spiked urineamples were analyzed within 2 h and kept in the refrigerator at 4 ◦Cor 7 days and analyzed again. No significant variation (<10% RSD)n the peak area ratios of l-Pro, l-Cys, l-Leu, l-Glu, and l-Kyn wereound after 7 days in the refrigerator.

. Conclusions

We have developed a method for quantifying low level analytesuch as sarcosine in urine without derivatization. Excellent resolu-ion was achieved between sarcosine and its isomers, �- and �-Ala,ven with the use of a positively charged PEI-coated capillary with aery fast electroosmotic flow (EOF). The optimized SPE protocol andreconcentration technique yielded very good recovery for sarco-ine (94.4 ± 7.4%). On the other hand, an extraction procedure wasot necessary for analyzing high concentration urinary metabolites,nd the sample must be diluted to avoid MS detector saturation.eparate calibrations for low and high level analytes could indeedecome laborious particularly for clinical study applications. Per-aps a more sensitive MS would allow us to detect sarcosine fromiluted non-SPE samples. Regardless of the circumstances, we areonfident that the method developed is validated for potential PCa

iomarker screening and we look forward to embarking on ourwn clinical study and finding an answer to the intriguing questionf whether sarcosine and other metabolites are indeed potentialiomarkers.

[

r. A 1267 (2012) 162– 169 169

Acknowledgement

This work was supported by a grant from Beckman Coulter Inc.,Brea, CA, USA, and the Natural Sciences and Engineering ResearchCouncil (NSERC) of Canada. LCS was partially supported by anNSERC CGSM Scholarship.

References

[1] N. Howlader, A.M. Noone, M. Krapcho, N. Neyman, R. Aminou, W. Wal-dron, S.F. Altekruse, C.L. Kosary, J. Ruhl, Z. Tatalovich, H. Cho, A. Mariotto,M.P. Eisner, D.R. Lewis, H.S. Chen, E.J. Feuer, K.A. Cronin, in: B.K. Edwards(Ed.), SEER Cancer Statistics Review, 1975–2008, National Cancer Institute.http://seer.cancer.gov/csr/1975 2008/, 2011 (accessed 20.04.12).

[2] Prostate Cancer Canada. http://www.prostatecancer.ca/Prostate-Cancer/Prostate-Cancer/Statistics, 2011 (accessed 20.04.12).

[3] W. Catalona, D. Smith, T. Ratcliff, K. Dodds, D. Coplen, J. Yuan, J. Petros, G.Andriole, N. Engl. J. Med. 324 (1991) 1156.

[4] M.J. Roobol, F.H. Schröder, E.D. Crawford, S.J. Freedland, A.O. Sartor, N. Fleshner,G.L. Andriole, J. Urol. 182 (2009) 2112.

[5] R.A. Clarke, H.J. Schirra, J.W. Catto, M.F. Lavin, R.A. Gardiner, Cancer 2 (2010)1125.

[6] T. Jamaspishvili, M. Kral, I. Khomeriki, V. Student, Z. Kolar, J. Bouchal, ProstateCancer Prostatic Dis. 13 (2010) 12.

[7] M. Oresic, A. Vidal-Puig, V. Hanninen, Expert Rev. Mol. Diagn. 6 (2006) 575.[8] D. Gruson, S. Bodovitz, Biomarkers 15 (2010) 289.[9] A. Sreekumar, L.M. Poisson, T.M. Rajendiran, A.P. Khan, Q. Cao, J. Yu, B. Laxman,

R. Mehra, R.J. Lonigro, Y. Li, M.K. Nyati, A. Ahsan, S. Kalyana-Sundaram, B. Han,X. Cao, J. Byun, G.S. Omenn, D. Ghosh, S. Pennathur, D.C. Alexander, A. Berger,J.R. Shuster, J.T. Wei, S. Varambally, C. Beecher, A.M. Chinnaiyan, Nature 457(2009) 910.

10] F. Jentzmik, C. Stephan, K. Miller, M. Schrader, A. Erbersdobler, G. Kristiansen,M. Lein, K. Jung, Eur. Urol. 58 (2010) 12.

11] E.A. Struys, A.C. Heijboer, J. van Moorselaar, C. Jakobs, M.A. Blankenstein, Ann.Clin. Biochem. 47 (2010) 282.

12] Y. Jiang, X. Cheng, C. Wang, Y. Ma, Anal. Chem. 82 (2010) 9022.13] F. Jentzmik, C. Stephan, M. Lein, K. Miller, B. Kamlage, B. Bethan, G. Kristiansen,

K. Jung, J. Urol. 185 (2011) 706.14] D. Cao, D. Ye, H. Zhang, Y. Zhu, Y. Wang, X. Yao, Prostate 71 (2011) 700.15] H. Wu, T. Liu, C. Ma, R. Xue, C. Deng, H. Zeng, X. Shen, Anal. Bioanal. Chem. 401

(2011) 635.16] F. Bianchi, S. Dugheri, M. Musci, A. Bonacchi, E. Salvadori, G. Arcangeli, V. Cupelli,

M. Lanciotti, L. Masieri, S. Serni, M. Carini, M. Careri, A. Mangia, Anal. Chim. Acta707 (2011) 197.

17] T.E. Meyer, S.D. Fox, H.J. Issaq, X. Xu, L.W. Chu, T.D. Veenstra, A.W. Hsing, Anal.Chem. 83 (2011) 5735.

18] H.J. Issaq, T.D. Veenstra, J. Sep. Sci. 34 (2011) 3619.19] A.K. Hewavitharana, Eur. Urol. 58 (2010) E39.20] D. Colleselli, A. Stenzl, C. Schwentner, Eur. Urol. 58 (2010) E51.21] E.J. Maxwell, X. Zhong, H. Zhang, N. van Zeijl, D.D.Y. Chen, Electrophoresis 31

(2010) 1130.22] R. Ramautar, O.A. Mayboroda, R.J.E. Derks, C. van Nieuwkoop, J.T. van Dissel,

G.W. Somsen, A.M. Deelder, G.J. de Jong, Electrophoresis 29 (2008) 2714.23] R. Ramautar, J.S. Torano, G.W. Somsen, G.J. de Jong, Electrophoresis 31 (2010)

2319.24] K. Guo, L. Li, Anal. Chem. 81 (2009) 3919.25] B. Cavaliere, B. Macchione, M. Monteleone, A. Naccarato, G. Sindona, A. Tagarelli,

Anal. Bioanal. Chem. 400 (2011) 2903.26] S.B. Heymsfield, C. Arteaga, C. McManus, J. Smith, S. Moffitt, Am. J. Clin. Nutr.

37 (1983) 478.27] J. Clarke, Clin. Chem. 7 (1961) 371.28] S. Kollipara, G. Bende, N. Agarwal, B. Varshney, J. Paliwal, Chromatographia 73

(2011) 201.29] D.S. Wishart, C. Knox, A.C. Guo, R. Eisner, N. Young, B. Gautam, D.D. Hau, N.

Psychogios, E. Dong, S. Bouatra, R. Mandal, I. Sinelnikov, J. Xia, L. Jia, J.A. Cruz,E. Lim, C.A. Sobsey, S. Shrivastava, P. Huang, P. Liu, L. Fang, J. Peng, R. Fradette,D. Cheng, D. Tzur, M. Clements, A. Lewis, A. De Souza, A. Zuniga, M. Dawe, Y.Xiong, D. Clive, R. Greiner, A. Nazyrova, R. Shaykhutdinov, L. Li, H.J. Vogel, I.Forsythe, Nucleic Acids Res. 37 (2009) D603.

30] D.S. Wishart, D. Tzur, C. Knox, R. Eisner, A.C. Guo, N. Young, D. Cheng, K. Jewell, D.Arndt, S. Sawhney, C. Fung, L. Nikolai, M. Lewis, M. Coutouly, I. Forsythe, P. Tang,S. Shrivastava, K. Jeroncic, P. Stothard, G. Amegbey, D. Block, D.D. Hau, J. Wagner,J. Miniaci, M. Clements, M. Gebremedhin, N. Guo, Y. Zhang, G.E. Duggan, G.D.

T. Marrie, B.D. Sykes, H.J. Vogel, L. Querengesser, Nucleic Acids Res. 35 (2007)D521.

31] Human Metabolome Database Version 2.5. www.hmdb.ca, 2009 (accessed18.04.12).