A Comparison of Different Algorithms of Sarcosine in Urine

Home , Sarcosine

Efforts to resolve the contradictions in early diagnosis of prostate cancer: a comparison of different algorithms of sarcosine in urine

D-L Cao1,2, D-W Ye1,2, Y Zhu1,2, H-L Zhang1,2, Y-X Wang3 and X-D Yao1,2 1Department of Urology, Fudan University Shanghai Cancer Center, Shanghai, China; 2Department of Oncology, Shanghai Medical College, Fudan University, Shanghai, China and 3Phase I clinical laboratory, Fudan University Shanghai Cancer Center, Shanghai, China

Controversial data on sarcosine as a promising biomarker for prostate cancer (PCa) detection are present. The objective was to clarify these discrepancies and reevaluate the potential value of sarcosine in PCa. Sarcosine algorithms (supernatant and sediment sarcosine/creatinine, supernatant and sediment log2 (sarcosine/alanine)) in urine samples from 71 untreated patients with PCa, 39 patients with no evidence of malignancy (NEM) and 20 healthy women and men were quantified by liquid chromatography/tandem mass spectrometry. Although any sarcosine algorithms were significantly higher in PCa patients than in NEM patients (all Po0.05), comparable sarcosine values were measured in healthy women and men. Additionally, neither biopsy Gleason score nor clinical T-stage were correlated with sarcosine algorithms (all P40.05), and receiver operating characteristic curve analysis indicated that the diagnostic power of any of sarcosine algorithms was nonsignificantly higher than that of serum and urine PSA, but nonsignificantly lower than prostate cancer antigen 3 (PCA3) and the percent-free PSA (%fPSA). Improved diagnostic performances were observed when any of sarcosine algorithms was combined with PCA3 or %fPSA. In conclusion, the predictive power of sarcosine in PCa is modest compared with PCA3 and %fPSA. Sarcosine, which awaits more validation before it reaches the clinic, could be included into the list of candidate PCa biomarkers. Prostate Cancer and Prostatic Diseases (2011) 14, 166–172; doi:10.1038/pcan.2011.2; published online 15 February 2011

Keywords: prostate cancer detection; prostate cancer antigen 3; PSA; sarcosine in urine

Introduction Notably, Sreekumar et al.8 demonstrated that sarcosine can be used as a biomarker for PCa and was non-invasively Prostate cancer (PCa) is one of the most commonly detected in urine, which led to great interest in the scientific diagnosed malignancies, and is the second leading cause community and among potential PCa patients. However, of cancer-related deaths in the Western male population.1 the first independent validation study9 showed that In China, the detection rate of PCa is increasing rapidly sarcosine cannot be used as a biomarker to diagnose because of the extension of life expectancy, the change of PCa and predict its aggression. Following these observa- lifestyles and the improvement of clinical skills.2 Subse- tions, letters and replies10–16 were published to discuss quently, conquering this stubborn disease has significant potential reasons for the contradictions between these two effect on the improvement of the health of the global studies.8,9 Although many possibilities (for example, the male population. On the basis of individual metabolites analytical method, the biomarker assay, the comparators representing end points of the molecular pathways and the study cohort) were discussed; no data from perturbed by events in the genome, transcriptome or complete article has been reported. proteome, newly emerging metabolite profile or meta- In this report, we sought to use liquid chromatogra- bolomics shows potential to monitor metabolite changes phy/tandem mass spectrometry, compare log2 (sarco- that characterize abnormal processes occurring in the sine/alanine) and sarcosine/creatinine in both urinary progression of disease and might provide clues for the supernatant and sediment and use prostate cancer treatment and diagnosis of disease.3,4 To the best of our antigen 3 (PCA3) and the percent-free PSA (%fPSA) as knowledge, metabolomic analysis has already been used comparators to determine whether sarcosine has the for identifying metabolic markers in tumors.5–7 potential role in PCa diagnostics and prognostics.

Correspondence: Dr X-D Yao, Department of Urology, Fudan Materials and methods University Shanghai Cancer Center, No. 270, Dongan Road; Xuhui District, Shanghai 200032, China. E-mail: [email protected] Specimen source, collection and processing Received 4 November 2010; revised 4 January 2011; accepted 6 This study was carried out according to the institutional January 2011; published online 15 February 2011 review board-approved study protocol. Written in- Efforts to solve controversy in sarcosine D-L Cao et al 167 Table 1 Clinicopathological characteristics of patients

Characteristics PSA: o20 ng mlÀ1 PSA: not limited

PCa (n ¼ 59) NEM (n ¼ 38) P-value PCa (n ¼ 71) NEM (n ¼ 39) P-value

Age (year) 0.717a 0.614a Range 53–79 50–79 53–82 50–79 Mean±s.e. 68.14±0.70 67.68±1.10 68.21±0.69 67.59±1.07

Serum PSA (ng mlÀ1) 0.545b 0.079b Range 1.67–17.50 1.27–19.27 1.67–219.52 1.27–28.31 Mean±s.e. 7.57±0.44 7.40±0.60 14.90±3.26 7.94±0.80

%fPSA 0.000b 0.000b Range 2.69–30.60 3.68–39.75 1.76–30.60 3.68–39.75 Mean±s.e. 13.00±0.57 16.86±1.09 13.05±0.54 17.10±1.10

Prostate volume (cm3) 0.294b 0.415b Range 19.31–55.62 22.55–40.31 19.31–55.62 22.55–40.31 Mean±s.e. 31.07±0.78 29.82±0.72 31.00±0.70 29.84±0.70

Biopsy Gleason score PSAo4 PSAX4 — 1.000c PSAo4 PSAX4 — 1.000c p6 3 28 3 30 ¼ 7 2 22 2 28 X804 08 cT stage PSAo4 PSAX4 — 0.495c PSAo4 PSAX4 — 0.592c pcT2a 1 15 1 17 ¼ cT2b 4 25 4 33 XcT2c 0 14 0 16

Abbreviations: %fPSA, the percent ratio of free to total PSA; cT stage, clinical T stage; NEM, no evidence of malignancy; PCa, prostate cancer. aIndependent sample t-test. bMann–Whitney’s U-test. cFisher’s exact test was calculated with PCa patients classified as cases with PSAo4 and PSAX4.

formed consent for each study participant was also Shimadzu LC-20AD UFLC system (Shimadzu, Kyoto, obtained. Between February and August 2010, a total of Japan) in conjunction with an 3200 QTrap mass spectro- 110 consecutive patients with PSA44ngmlÀ1 or abnor- metry (Applied Biosystems, Carlsbad, CA, USA) as mal digital rectal examination, out of which 71 patients previously described.18 Urinary creatinine concentra- with PCa and 39 patients with no evidence of malig- tions were measured by a standard assay on the Hitachi nancy (NEM), as confirmed by 10 core prostate biopsy 7020 analyzer (Hitachi, Tokyo, Japan). These results samples at Shanghai Cancer Center (Shanghai, China), obtained were then used to calculate four kinds of were enrolled. At the time of enrollment and at 1 week sarcosine algorithms (supernatant and sediment sarco- later, the first-void urines from these 110 patients were sine/creatinine, supernatant and sediment log2 (sarco- collected, respectively, after standardized digital rectal sine/alanine)). The expression levels of PSA mRNA and examination17 in about 5 min. The second-morning void PCA3 were determined as previously described.18 urines from 20 healthy men and women were also obtained. In addition, clinicopathological features of all patients investigated were recorded and presented Statistical analysis in Table 1. Statistical analyses were performed using the R software Following collection, 40–50 ml urine sample was cooled (The University of Auckland, Auckland, New Zealand) immediately on ice and centrifuged at 4 1C and at 700 g for and SPSS 17.0 (SPSS, Chicago, IL, USA). As only subjects 10minwithin1h.Uponcentrifugation, the first urinary with PSA o20 ng mlÀ1 were included in the study by sediment collected at the time of enrollment was mixed with Jentzmik et al.,9 patients in this study were divided into a TRIzol reagent (Invitrogen, Carlsbad, CA, USA) to measure subgroup of all patients and a subgroup of patients with PSA mRNA and PCA3; the second urinary sediment PSA o20 ng mlÀ1 to facilitate comparison. collected at 1 week later was added to 5 ml of 1 M boric Independent sample t-test or Mann–Whitney’s U-test acid to analyze sarcosine, and urinary supernatant was also was used to compare differences between two sets of collected to analyze sarcosine. After treatment, these samples continuous variables, and Fisher’s exact test for the were stored at À80 1C until analysis. analysis of n Â n tables. Spearman’s or Pearson’s rank correlation and linear regression analysis were performed to determine the relationships between clinico- Determination of sarcosine, alanine, creatinine and other pathological features and any of sarcosine algorithms. analytes Logistic regression analysis was used to analyze correla- Sarcosine and alanine were derivatized and isolated tion between PCa diagnostic status and tested marker. using propyl chloroformate, and measured using a Diagnostic performance of investigated marker was

Prostate Cancer and Prostatic Diseases Efforts to solve controversy in sarcosine D-L Cao et al

168 evaluated using the receiver operating characteristic above for the subgroup of all patients (data not shown). curve analysis, and the overfitting bias for the area Not only in the cohort of all patients but also in the under curve (AUC) calculations was estimated using cohort of patients with PSA o20 ng mlÀ1, there was no special bootstrap software. A two-tailed P-value o0.05 significant difference between supernatant sarcosine/ was considered statistically significant. creatinine and sediment sarcosine/creatinine in the subgroup of NEM patients, PCa patients, Gleason score o7 patients, Gleason scoreX7 patients, cT stage o2b Results patients or cT stageX2b patients (all P40.05); the same is true for the comparison between supernatant and As shown in Table 1, equalities regarding clinicopatho- sediment log2 (sarcosine/alanine) levels (all P40.05). logical characteristics of patients were estimated and no The diagnostic performances of tested variables were significant differences were found except for %fPSA, summarized in Table 2. Numerically, the AUCs for four both in all patients and patients with PSA o20 ng mlÀ1. kinds of sarcosine algorithms were nonsignificantly higher With regard to sarcosine/creatinine and log2 (sarco- than those for serum PSA and urine PSA mRNA, but sine/alanine) in urinary supernatant from all patients, nonsignificantly lower than those for PCA3 and %fPSA in there were no differences among healthy female, healthy both cohorts. Specifically, the diagnostic power of sediment male and NEM patients (Figures 1a and d). In the cohort log2 (sarcosine/alanine), with the highest AUC among the of all patients, any of sarcosine algorithms was signifi- four kinds of sarcosine values, was not significantly better cantly higher in PCa than in NEM patients (Figures 1a than that of serum PSA, both in all patients (P ¼ 0.219) and and d), but no significant differences in both creatinine in patients with PSA o20 ng mlÀ1 (P ¼ 0.063). Never- and alanine levels in urinary supernatant and sediment theless, including any of sarcosine algorithms in multi- were found between patients with PCa and NEM variate logistic regression models with PCA3 or %fPSA (Supplementary Figure 1), indicating that they would could improve the AUC relative to the single biomarkers in not influence the proper evaluation about the role of both groups (Table 2). After correcting the potential over-fit sarcosine in PCa. Our study also showed that neither for initial AUCs for investigated variables, using specific biopsy Gleason score nor clinical T-stage (cT stage) were bootstrap software, which consequently generated correlated with any of sarcosine algorithms in all patients matched unbiased AUCs, differences between initial AUCs (Figures 1b–c, e–f). It is to be noted that patients with and matched unbiased AUCs were estimated to be small PSA o20 ng mlÀ1 held similar features as mentioned and can be neglected (Table 2).

a Sarcosine/creatinine in urine in the study groups bcBiopsy Gleason score and sarcosine/creatinine Clinical T stage and sarcosine/creatinine in urine in urine Sediment S/C 3500.00 Supernatant S/C Supernatant S/C 3500.00 Sediment S/C P=0.008 Supernatant S/C P=0.014 3500.00 Supernatant S/C Sediment S/C P=0.590 P=0.547 P=0.745 P=0.968 P=0.786 3000.00 127 3000.00

422 3000.00 mol/mol) mol/mol) mol/mol) μ μ μ 2500.00 2500.00 2500.00

2000.00 2000.00 2000.00

1500.00 111 1500.00 1500.00 128 1000.00 1000.00 1000.00 63 64 Sarcosine/creatinine ratio ( Sarcosine/creatinine ratio Sarcosine/creatinine ratio ( Sarcosine/creatinine ratio

500.00 ( Sarcosine/creatinine ratio 500.00 500.00

delog2(sarcosine/alanine) in urine in the study groups Biopsy Gleason score and log2(sarcosine/creatinine) f Clinical T stage and log2(sarcosine/creatinine) in urine in urine 6.00 Supernatant log2(S/A) 6.00 Sediment log2(S/A) 6.00 Supernatant log2(S/A) Sediment log2(S/A) Supernatant log2(S/A) P=0.001 Supernatant log2(S/A) Sediment log2(S/A) P=0.001 P=0.553 P=0.633 4.00 P =0.674 P=0.881 P=0.932 4.00 4.00 79 229 2.00 2.00 2.00

0.00 0.00 0.00

213 135

222 135 log2(sarcosine/alanine) log2(sarcosine/alanine) -2.00 104 -2.00

log2(sarcosine/alanine) -2.00 63 253 64 72 64 103 -4.00 -4.00 -4.00

HF HM NEM PCa NEM PCa Gleason Gleason Gleason Gleason cT cT cT cT (n=20) (n=20) (n=39) (n=71) (n=39) (n=71) score<7 score≥7 score<7 score≥7 stage<2b stage≥2b stage<2b stage≥2b (n=33) (n=38) (n=33) (n=38) (n=18) (n=53) (n=18) (n=53) Figure 1 In the cohort of all patients, four kinds of sarcosine algorithms in urine of patients with prostate cancer (PCa) and with no evidence of malignancy (NEM) after digital rectal examination and healthy female (HF) and healthy male (HM). (a) Sarcosine/creatinine ratio in urine in the study groups. (b) Sarcosine/creatinine ratio in urine of the PCa patients classified as cases with biopsy Gleason score o7 and X7. (c) Sarcosine/creatinine ratio in urine of the PCa patients classified as cases with clinical T (cT) stage o2b and X2b. (d) Log2 (sarcosine/ alanine) in urine in the study groups. (e) Log2 (sarcosine/alanine) in urine of the PCa patients classified as cases with biopsy Gleason score o7 and X7. (f) Log2 (sarcosine/alanine) in urine of the PCa patients classified as cases with cT stage o2b and X2b. Differences were tested by the independent sample t-test and/or Mann–Whitney’s U-test. S/A, sarcosine/alanine ratio; S/C, sarcosine/creatinine ratio.

Prostate Cancer and Prostatic Diseases Efforts to solve controversy in sarcosine D-L Cao et al 169 Table 2 ROC analysis for univariate tested variables and multiplex models in urine for prostate cancer diagnosis

Variable PSA: o 20 ng mlÀ1 PSA: not limited

AUC (95% CI) P-value Bootstrap AUC (95% CI) AUC (95% CI) P-value Bootstrap AUC (95% CI)

Serum PSA 0.537 (0.419–0.654) 0.545 0.537 (0.441–0.633) 0.602 (0.495–0.708) 0.079 0.602 (0.487–0.709) Urine PSA 0.571 (0.453–0.689) 0.237 0.571 (0.454–0.678) 0.590 (0.480–0.701) 0.118 0.590 (0.478–0.680) %fPSA 0.712 (0.600–0.824) 0.000 0.712 (0.607–0.810) 0.706 (0.598–0.814) 0.000 0.706 (0.592–0.804) PCA3 0.703 (0.600–0.806) 0.001 0.703 (0.609–0.796) 0.717 (0.621–0.814) 0.000 0.717 (0.610–0.810) Supernatant log2 (S/A) 0.687 (0.582–0.793) 0.002 0.688 (0.577–0.787) 0.684 (0.583–0.784) 0.001 0.684 (0.570–0.768) Sediment log2 (S/A) 0.698 (0.593–0.803) 0.001 0.697 (0.605–0.785) 0.696 (0.597–0.796) 0.001 0.695 (0.595–0.787) Supernatant (S/C) 0.647 (0.537–0.756) 0.015 0.647 (0.533–0.753) 0.642 (0.539–0.746) 0.014 0.643 (0.537–0.733) Sediment (S/C) 0.660 (0.552–0.768) 0.008 0.660 (0.563–0.763) 0.654 (0.551–0.757) 0.008 0.654 (0.555–0.754) PCA3 + supernatant log2 (S/A) 0.772 (0.678–0.866) 0.000 0.772 (0.671–0.854) 0.772 (0.682–0.862) 0.000 0.773 (0.674–0.857) PCA3 + sediment log2 (S/A) 0.775 (0.681–0.869) 0.000 0.775 (0.699–0.857) 0.777 (0.688–0.866) 0.000 0.777 (0.689–0.858) PCA3 + supernatant (S/C) 0.720 (0.620–0.821) 0.000 0.720 (0.631–0.815) 0.743 (0.651–0.835) 0.000 0.743 (0.661–0.839) PCA3 + sediment (S/C) 0.733 (0.635–0.832) 0.000 0.733 (0.632–0.823) 0.752 (0.662–0.842) 0.000 0.752 (0.665–0.833) %fPSA + supernatant log2 (S/A) 0.753 (0.652–0.854) 0.000 0.753 (0.644–0.855) 0.759 (0.663–0.856) 0.000 0.759 (0.659–0.842) %fPSA + sediment log2 (S/A) 0.760 (0.661–0.859) 0.000 0.760 (0.677–0.863) 0.768 (0.674–0.863) 0.000 0.767 (0.666–0.860) %fPSA + supernatant (S/C) 0.741 (0.634–0.847) 0.000 0.741 (0.624–0.840) 0.739 (0.637–0.842) 0.000 0.739 (0.632–0.828) %fPSA + sediment (S/C) 0.748 (0.642–0.853) 0.000 0.748 (0.627–0.852) 0.746 (0.644–0.847) 0.000 0.746 (0.630–0.842)

Abbreviations: ‘+’, combined with; %fPSA, the percent ratio of free to total PSA; AUC, area under the ROC curve; CI, confidence interval; PCA3, prostate cancer antigen 3; ROC, receiver operating characteristic; S/A, sarcosine/alanine ratio; S/C, sarcosine/creatinine ratio.

By Pearson and Spearman correlation analysis, we Metabolomics, an omics science in systems biology, is observed that significant relationships were detected the globally quantitative assessment of endogenous among four kinds of sarcosine algorithms, both in all metabolites within a biological system and allows for patients and patients with PSA o20 ng mlÀ1, except for a comprehensive evaluation of a cellular state within supernatant sarcosine/creatinine versus supernatant the context of the immediate environment, taking into log2 (sarcosine/alanine), supernatant sarcosine/creati- account genetic regulation, altered kinetic activity of enzy- nine versus sediment log2 (sarcosine/alanine), sediment mes and changes in metabolic reactions.4,19,20 Importantly, sarcosine/creatinine versus supernatant log2 (sarcosine/ the tumor metabolome is generally characterized by alanine) and sediment sarcosine/creatinine versus sedi- increased phospholipid levels, elevated glycolytic capa- ment log2 (sarcosine/alanine) in patients with NEM in city, high glutaminolytic function and overexpression both cohorts (Table 3). Additionally, both in the popula- of the glycolytic isoenzyme (pyruvate kinase type M2, tion of all patients and patients with PSA o20 ng mlÀ1, M2-PK), as previously reviewed.3 With regard to the age, serum PSA, prostate volume, urine PSA mRNA, application of metabolomics in cancer diagnostics, PCa %fPSA and PCA3 did not correlate with any of sarcosine demonstrated a different metabolic profile characterized algorithms in any of the two subgroups (patients with by high glycolytic products, total choline-containing PCa and patients with NEM), with the exception of compounds and phosphocholine levels.21 In addition, PCA3 versus supernatant and sediment log2 (sarcosine/ prostatic fluid from men with PCa, exhibited reduced alanine) in patients with PCa in the first cohort, and citrate and increased spermine levels, which correlated serum PSA versus supernatant and sediment log2 well with Gleason score, and outperformed PSA in cancer (sarcosine/alanine), urine PSA mRNA versus sediment diagnosis.22,23 Not limited to the field of PCa diagnosis, sarcosine/creatinine and PCA3 versus supernatant and metabolomics is also providing essential information sediment log2 (sarcosine/alanine) in patients with PCa about tumorigenesis, cancer diagnosis in other tumors, in the second cohort (Supplementary Table 1). cancer prognosis, therapeutic evaluation and revealing As the range of 2.5–10 ng mlÀ1 is considered to be new therapeutic targets.3,4 These observations clearly the clinically challenging value for PSA, we have showed that metabolomics have a significant role in also evaluated the potential role of any of sarcosine providing a more comprehensive picture of tumor algorithms in the subset of patients having PSA development. 2.5–10 ng mlÀ1 and found that the role of any of sarcosine Sarcosine is found naturally as an intermediate in the algorithms in the subset of patients with PSA between metabolism of choline and methionine,24 and also 2.5 and 10 ng mlÀ1 were similar to those in the subsets of generated by the enzymatic transfer of a methyl group all patients and patients with PSA o20 ng mlÀ1 (data not from S-adenosylmethionine to glycine, and thus shows a shown). substantial role in methyl balance in humans.25 Interest- ingly, a recent report in Nature by Sreekumar et al.8 demonstrated that sarcosine as one of metabolites offers a promising, non-invasive diagnostic and prognostic test Discussion for PCa. However, the first validation study9 did not support that sarcosine has any advantage over serum The present study validated some reasons for the PSA, %fPSA, Gleason score and tumor stage. Obviously, discrepancies on sarcosine as a promising biomarker it will be of great importance to clarify these contra- for PCa and subsequently reevaluated the potential role dictions regarding sarcosine as a promising test for PCa of urinary sarcosine in PCa detection and prognosis. in greater details.

Prostate Cancer and Prostatic Diseases Efforts to solve controversy in sarcosine D-L Cao et al

170 In this study, any of sarcosine algorithms was significantly higher in PCa than in NEM patients not -value) P only in all patients (Figures 1a and d) but also in patients 0.03 0.04 0.05 0.05 À1 À À À À with PSA o20 ng ml (data not shown), supporting the report by Sreekumar et al.8 and simultaneously opposing the study by Jentzmik et al.9 As urinary creatinine concentration of patients with NEM did not differ from that of PCa patients in both groups (Supplementary a a a a -value) NEM r (

P Figure 1), the creatinine normalization could not be the reason for this contradictory result. These inconsistent results might be explained by differences between the study populations or in pathological evaluation of specimens. For example, a fair comparison between urinary sarcosine and serum PSA would be the one in -value) PCa r (

P which asymptomatic men aged 50–70 years were 0.040.04 0.39 0.37 0.06 0.31 -0.05 0.32

À À À screened by each method, followed by prostate biopsy of the men with positive screening results, and calcula- tion of diagnostic sensitivities and specificities. Impor- tantly, comparable sarcosine values were detected in healthy women and men and no significant differences a a a a

-value) NEM r ( among healthy females, males and NEM patients were P

reatinine ratio. found as listed in our results, indicating that sarcosine could not be used as a promising biomarker for PCa. In addition, no significant differences between urinary supernatant and sediment sarcosine levels as described in our results (all Po0.05) illustrated that digital rectal examination did not significantly change urinary sarco-

0.040.04 — 0.990.060.05 — 0.99 — 0.99sine 0.99 — levels, — 0.99 which 0.99 — would 0.99 — generate 0.99 — a concept that 0.99 0.39 0.99 0.32 À À À À sarcosine as a small metabolite is rapidly released into the urinary supernatant. Furthermore, any of sarcosine algorithms was independent of biopsy Gleason score and cT stage, both in all patients (Figures 1b–c, e–f) and in patients with PSA a a a a a a

-value) NEM r (P-value) PCa r ( À1

P o20 ng ml (data not shown), these observations were similar to previous studies.8,9 Comparing with Gleason score and T stage, determined in radical prostatectomy tissue specimens, the fact that PCa is heterogeneous disease and thus results in underestimating biopsy Gleason score and cT stage, might explain the above -value) PCa r (

P mentioned issue. Follow-up data on the final Gleason 0.040.03 0.39 0.37 0.050.05 0.31 0.31 0.99 —0.99 — — 0.39 — 0.31

À À À À score and T stage after radical prostatectomy will clarify in time whether sarcosine excretion has association with

Spearman rank correlation analysis. Gleason score and T stage. Our study also demonstrated a that the AUC for any of sarcosine algorithms was nonsignificantly higher than that for serum PSA and a a a a a a

-value) NEM r ( urine PSA mRNA, but nonsignificantly lower than that P — —— (0.00) — (0.00) (0.00) (0.00) (0.00) (0.84) (0.00) (0.00) (0.76) (0.87) (0.00) (0.79) (0.00) (0.00)(0.00) — (0.00) — — (0.00) — (0.79) (0.00) (0.00) (0.72) (0.82) (0.00) (0.74) (0.00)(0.00) (0.84) (0.87)(0.00) (0.00)(0.00) (0.00) (0.76) (0.79)for (0.79) (0.82) PCA3 (0.00) and — (0.00) (0.00) %fPSA, (0.72) in (0.74)both — (0.00) population — (0.00) (0.00) — (Table — (0.00) 2), (0.00) which — (0.00) — (0.00) — basically was consistent with data obtained from Pca r ( Sreekumar et al.8 and Jentzmik et al.9 It is to be noted that improved AUCs were obtained when any of sarcosine values was combined with PCA3 or %fPSA (Table 2), which could support the speculation that adding sarcosine to a panel has the potential to increase the accuracy of diagnosing PCa. The limitations of the present study have been

Sediment S/C 0.98 Sediment S/C 0.99 addressed as follows. First, cancer patients were en- Supernatant S/C — — 0.99 Supernatant S/C — — 0.98

Sediment log2 (S/A) 0.39 Sediment log2 (S/A) 0.32 riched during sample collection process to generate Supernatant log2 (S/A) 0.39 Supernatant log2 (S/A) 0.32 subsequent robust data analysis, which resulted in the ) Variable Supernatant S/C Sediment S/C Supernatant log2 (S/A) Sediment log2 (S/A)

1 study population that does not represent those in the À everyday clinical practice. As explained by Jentzmik et al.,9 the risk of type II error could be eliminated and a Correlative associations among sarcosine algorithms screening population would probably not change the data. In addition, bias may be generated from the retrospective design, but the recent study mainly used 20: All Table 3 PSA range (ng ml Abbreviations: NEM, no evidence of malignancy; PCa, prostate cancer; r, correlation coefficient; S/A, sarcosine/alanine ratio; S/C, sarcosine/c Correlations were calculated by Pearson correlation analysis except for o one of the retrospective designs, also named as nested

Prostate Cancer and Prostatic Diseases Efforts to solve controversy in sarcosine D-L Cao et al 171 case–control design, by which urine specimens were metabolic correlation network: finding potential biomarkers for collected from patients with PSA44ngmlÀ1 or abnormal breast cancer. Analyst 2009; 134: 2003–2011. digital rectal examination before prostate biopsy and 6 Kim K, Aronov P, Zakharkin SO, Anderson D, Perroud B, later undergone retrospective evaluation, which can help Thompson IM et al. Urine metabolomics analysis for kidney minimize the problem of baseline inequality because the cancer detection and biomarker discovery. Mol Cell Proteomics presence and absence of disease cannot affect selection of 2009; 8: 558–570. 7 Osl M, Dreiseitl S, Pfeifer B, Weinberger K, Klocker H, Bartsch G subjects or handling of specimens. Although metabolites et al. A new rule-based algorithm for identifying metabolic (for example, sarcosine) in metabolomic profile could markers in prostate cancer using tandem mass spectrometry. fluctuate markedly depending on differently physiologi- Bioinformatics 2008; 24: 2908–2914. cal and pathological state, the validity of our data could 8 Sreekumar A, Poisson LM, Rajendiran TM, Khan AP, be guaranteed by the use of nested case–control design Cao Q, Yu J et al. Metabolomic profiles delineate potential role and sophisticated liquid chromatography/tandem mass for sarcosine in prostate cancer progression. Nature 2009; 457: spectrometry technology. Along with sound design of 910–914. study, technical progress, establishment of database and 9 Jentzmik F, Stephan C, Miller K, Schrader M, Erbersdobler A, accumulation of evidences, it can be increasingly Kristiansen G et al. Sarcosine in urine after digital rectal examination believed that effectively controlled bias could not change fails as a marker in prostate cancer detection and identification of the results, and the metabolomics would be a good aggressive tumours. Eur Urol 2010; 58: 12–18, 20-21. 10 Schalken JA. Is urinary sarcosine useful to identify patients with source of biomarkers for diagnostics and prognostics. significant prostate cancer? The trials and tribulations of biomarker development. Eur Urol 2010; 58: 19–20. 11 Stephan C, Jentzmik F, Jung K. Reply from authors re: Jack A Schalken. Is urinary sarcosine useful to identify patients with Conclusion significant prostate cancer? the trials and tribulations of In summary, the present study demonstrated that the biomarker development. Eur Urol 2010; 58: 20–21. 12 Sreekumar A, Poisson LM, Rajendiran TM, Khan AP, Cao Q, Yu J predictive power of sarcosine in PCa is modest com- et al. Re: Florian Jentzmik, Carsten Stephan, Kurt Miller, et al. pared with PCA3 and %fPSA. Although cancer-related Sarcosine in urine after digital rectal examination fails as a metabolomics has a substantial role in providing a more marker in prostate cancer detection and identification of comprehensive picture of tumor development, metabo- aggressive tumours. Eur Urol 2010; 58: e29–e30. lomic profile could fluctuate markedly depending on 13 Jentzmik F, Stephan C, Jung K. Reply to Arun Sreekumar, Laila different status and thus the data for metabolites (for M Poisson, Thekkelnaycke M. Rajendiran, et al.’s Letter to the example, sarcosine) should be interpreted with caution Editor re: Florian Jentzmik, Carsten Stephan, Kurt Miller, et al. before increased understanding of metabolomics are Sarcosine in Urine after Digital Rectal Examination Fails as a obtained. The sarcosine pathway might provide clues Marker in Prostate Cancer Detection and Identification of for PCa diagnosis or progression, while sarcosine that Aggressive Tumours. Eur Urol 2010; 58: e31–e32. awaits more validation, before it contributes to clinical 14 Hewavitharana AK. Re: Florian Jentzmik, Carsten Stephan, Kurt Miller, et al. Sarcosine in urine after digital rectal examination decision making, could be included into the long list of fails as a marker in prostate cancer detection and identification candidate PCa biomarkers. of aggressive tumours. Eur Urol 2010; 58: e39–e40. 15 Jentzmik F, Stephan C, Jung K. Reply to Amitha K Hewavithar- ana’s Letter to the Editor re: Florian Jentzmik, Carsten Stephan, Conflict of interest Kurt Miller, et al. Sarcosine in urine after digital examination fails as a marker in prostate cancer detection and identification The authors declare no conflict of interest. of aggressive tumours. Eur Urol 2010; 58: 12–18. Eur Urol 2010; 58: e41–e42. 16 Colleselli D, Stenzl A, Schwentner C. Re: Florian Jentzmik, Acknowledgements Carsten Stephan, Kurt Miller, et al. Sarcosine in urine after digital rectal examination fails as a marker in prostate cancer detection We thank Jun Qian from our Cancer Center (Phase I and identification of aggressive tumours. Eur Urol 2010; 58: 12– clinical laboratory), Yu-Hu Xin and Yue Cao from our 28, 2010; 58: e51. Cancer Center (Department of Laboratory) and Yan 17 Groskopf J, Aubin SM, Deras IL, Blase A, Bodrug S, Clark C et al. Huang from our Cancer Center (Department of Gynecol- APTIMA PCA3 molecular urine test: development of a method ogy) for providing technical assistance. This study was to aid in the diagnosis of prostate cancer. Clin Chem 2006; 52: supported by the Shanghai Charitable Cancer Research 1089–1095. Center; Grant no. Shaz200908. 18 Cao DL, Ye DW, Zhang HL, Zhu Y, Wang YX, Yao XD. A multiplex model of combining gene-based, protein-based and metabolite-based with positive and negative markers in urine References for the early diagnosis of prostate cancer. Prostate 2010; e-pub ahead of print 18 October 2010; doi:10.1002/pros.21286. 1 Jemal A, Siegel R, Ward E, Hao Y, Xu J, Thun MJ. Cancer 19 Mendes P, Kell DB, Westerhoff HV. Why and when channelling statistics, 2009. CA Cancer J Clin 2009; 59: 225–249. can decrease pool size at constant net flux in a simple dynamic 2 Sim HG, Cheng CW. Changing demography of prostate cancer channel. Biochim Biophys Acta 1996; 1289: 175–186. in Asia. Eur J Cancer 2005; 41: 834–845. 20 Mendes P, Kell DB, Westerhoff HV. Channelling can decrease 3 Spratlin JL, Serkova NJ, Eckhardt SG. Clinical applications of pool size. Eur J Biochem 1992; 204: 257–266. metabolomics in oncology: a review. Clin Cancer Res 2009; 15: 21 Swanson MG, Zektzer AS, Tabatabai ZL, Simko J, Jarso S, 431–440. Keshari KR et al. Quantitative analysis of prostate metabolites 4 Griffin JL, Shockcor JP. Metabolic profiles of cancer cells. Nat Rev using 1H HR-MAS spectroscopy. Magn Reson Med 2006; 55: Cancer 2004; 4: 551–561. 1257–1264. 5 Chen Y, Zhang R, Song Y, He J, Sun J, Bai J et al. RRLC-MS/MS- 22 Kline EE, Treat EG, Averna TA, Davis MS, Smith AY, Sillerud LO. based metabonomics combined with in-depth analysis of Citrate concentrations in human seminal fluid and expressed

Prostate Cancer and Prostatic Diseases Efforts to solve controversy in sarcosine D-L Cao et al 172 prostatic fluid determined via 1H nuclear magnetic resonance 24 Mitchell AD, Benevenga NJ. Importance of sarcosine formation spectroscopy outperform prostate specific antigen in prostate in methionine methyl carbon oxidation in the rat. J Nutr 1976; cancer detection. JUrol2006; 176: 2274–2279. 106: 1702–1713. 23 Serkova NJ, Spratlin JL, Eckhardt SG. NMR-based metabolo- 25 Mudd SH, Brosnan JT, Brosnan ME, Jacobs RL, Stabler SP, Allen mics: translational application and treatment of cancer. Curr RH et al. Methyl balance and transmethylation fluxes in humans. Opin Mol Ther 2007; 9: 572–585. Am J Clin Nutr 2007; 85: 19–25.

Supplementary Information accompanies the paper on the Prostate Cancer and Prostatic Diseases website (http:// www.nature.com/pcan)

Prostate Cancer and Prostatic Diseases