JO U R N A L OF PR O TE O MI CS 98 ( 20 1 4 ) 4 4 –58

Available online at www.sciencedirect.com

ScienceDirect www.elsevier.com/locate/jprot

Discovery and validation of urinary biomarkers for detection of renal cell carcinoma Maria Frantzia,b,⁎,1 , Jochen Metzgera,1 , Rosamonde E. Banksc , Holger Husid , Julie Kleina , Mohammed Daknaa , William Mullend , Jonathan J. Cartledgee , Joost P. Schanstraf,g , Korbinian Brandh , Markus A. Kuczyki , Harald Mischaka,d , Antonia Vlahoub , Dan Theodorescu j,k , Axel S. Merseburgeri a

Mosaiques diagnostics GmbH, Hannover, Germany Biomedical Research Foundation, Academy of Athens, Biotechnology Division, Athens, Greece c St James's University Hospital, Cancer Research UK Clinical Centre, Clinical and Biomedical Proteomics Group, Leeds, United Kingdom d University of Glasgow, Institute of Cardiovascular and Medical Sciences, Glasgow, United Kingdom e St James's University Hospital, Department of Urology, Leeds, United Kingdom f Inserm, U858/I2MR, Department of Renal and Cardiac Remodeling, Team #5, 1 Avenue Jean Poulhès, BP 84225, 31432 Toulouse Cedex 4, France g Université Toulouse III Paul Sabatier, Institut de Médecine Moléculaire de Rangueil, Toulouse F-31000, France h Hannover Medical School, Institute of Clinical Chemistry, Hannover, Germany i Hannover Medical School, Department of Urology and Urological Oncology, Hannover, Germany j University of Colorado, Department of Surgery and Pharmacology, Aurora, CO, USA k University of Colorado Comprehensive Cancer Center, Aurora, CO, USA b

AR TIC LE I N FO

ABS TR ACT

Article history:

IntroductionRenal cell carcinoma (RCC) is often accompanied by non-specific symptoms. The

Received 17 October 2013

increase of incidentally discovered small renal masses also presents a diagnostic dilemma. This

Accepted 14 December 2013

study investigates whether RCC-specific peptides with diagnostic potential can be detected in

Available online 25 December 2013

urine and whether a combination of such peptides could form a urinary screening tool.

Keywords:

to analyze urine samples from patients with RCC (N = 40) compared to non-diseased controls

Renal cell carcinoma

(N = 68).

Urinary peptide markers

Results and discussion86 peptides were found to be specifically associated to RCC, of which

Diagnosis

sequence could be obtained for 40. A classifier based on these peptides was evaluated in an

Proteomics

independent set of 76 samples, resulting in 80% sensitivity and 87% specificity. The

Protease prediction

specificity of the marker panel was further validated in a historical dataset of 1077 samples

Materials and methodsFor the discovery of RCC-specific biomarkers, we have employed CE–MS

including age-matched controls (N = 218), patients with related cancer types and renal diseases (N = 859). In silico protease prediction based on the cleavage sites of differentially excreted peptides, suggested modified activity of certain proteases including cathepsins, ADAMTS and kallikreins some of which were previously found to be associated to RCC.

Abbreviations: ADAMTS, a disintegrin and metalloproteinase with thrombospondin motifs; BMI, body mass index; DN, diabetic nephropathy; ECM, extracellular matrix, FDR, false discovery rate; LOOCV, leave one out cross-validation; MRI, magnetic resonance imaging; RCC, renal cell carcinoma; ROC, receiver operating characteristic; ST14, suppressor of tumorigenicity 14 protein; SVM, support vector machine; TCC, transitional cell carcinoma; TNM, tumor nodes metastases. ⁎ Corresponding author at: Mosaiques diagnostics GmbH, Mellendorfer Strasse 7–9, D-30625 Hannover, Germany. Tel.: + 49 511 55 47 44 28. E-mail address: [email protected] (M. Frantzi). 1 MF and JM contributed equally to this work. 1874-3919/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.jprot.2013.12.010

45

JO U R N A L OF P ROTE O MI CS 9 8 ( 20 1 4 ) 4 4–5 8

ConclusionsRCC can be detected with high accuracy based on specific urinary peptides. Biological significance Clear cell renal cell carcinoma (RCC) has the highest incidence among the renal malignancies, often presenting non-specific or no symptoms at all. Moreover, with no diagnostic marker being available so far, almost 30% of the patients are diagnosed with metastatic disease and 30–40% of the patients initially diagnosed with localized tumor relapse. These facts introduce the clinical need of early diagnosis. This study is focused on the investigation of a marker model based on urinary peptides, as a tool for the detection of RCC in selected patients at risk. Upon evaluation of the marker model in an independent blinded set of 76 samples, 80% sensitivity and 87% specificity were reported. An additional dataset of 1077 samples was subsequently employed for further evaluation of the specificity of the classifier. © 2013 Elsevier B.V. All rights reserved.

1. Introduction Clear cell renal cell carcinoma (RCC) is the most common malignancy of the kidney. The conventional clear cell histological type is thought to arise from the proximal tubules and accounts for ~80% of RCC cases. Approximately 210,000 new cases of renal cancer are diagnosed each year worldwide, with over 100,000 deaths annually [1]. The standard treatment for locally advanced RCC is nephron-sparing or radical nephrectomy. In patients with small renal tumors and/or significant co-morbidity who are unfit for surgery, an ablative approach, e.g. cryotherapy and radiofrequency ablation remains an option [2,3]. To date no adjuvant therapy has been recommended. However several trials are currently investigating the efficacy of adjuvant tyrosine kinase inhibitors including sunitinib, pazopanib, axitinib and sorafenib in locally advanced disease [4]. Unfortunately in the metastatic situation this cancer is resistant to conventional chemotherapy or radiotherapy. In this case of poor prognosis, tyrosine kinase and mTOR inhibitors are approved as a second line option, increasing the sequential overall survival to 30 months [5–7]. The increase in disease rates, together with the fact that no diagnostic marker is available, has high socio-economic effects and underpins the demand, as recognized by the American Cancer Institute [8]. With either relatively non-specific or absent symptoms, about 30% of patients are diagnosed with distant or local metastatic disease. In addition, 30–40% of patients with initially clinically localized disease relapse during the follow-up period [9]. The 5-year survival for patients with metastatic disease is 30, 12 subjects with > 1 year of hypertension and 12 subjects with > 4 pack-yr smoking history. The rationale for the study concept is described in Supplement Text. We identified 86 peptides displaying statistically significant differences in their distribution between RCC and the non-diseased reference groups after multiple testing correction. All peptides are listed in Table 4. We obtained sequences for 40 out of the 86 (Table 4). By adaptive machine learning a multidimensional classifier was established using all 86 RCC peptide marker candidates. Leave one out cross-validation (LOOCV) of classification scores for the training cohort with this RCC-specific peptide marker model resulted in an AUC value of 0.98 and a 95% confidence

48

Table 4 – Characteristics of the 86 peptides included in the urine peptide marker model for differentiation of RCC patients (N = 40) from normal controls (N = 68). a Peptide-ID b

CE–MS characteristics

Amino acid sequence information Sequence c

CE migration time [min]

4976

902.41

20.85

DpGKNGDKG

9288 12264 14047 15474 16320 16453 16773 17968

956.44 992.42 1032.45 1063.44 1075.49 1077.48 1080.48 1099.49

20.36 20.40 25.90 20.28 20.61 20.42 27.77 28.24

n.i. n.i. n.i. n.i. n.i. IVKWDRDm DRGEpGPpGPA DGGGSPKGDVDP

18216

1104.50

25.81

19773

1128.39

33.59

19791 21338 23224 23360 23409 23467 23518 23697 25003 26163 26326 26866 28022 33716 35027 36331 36819 38510

1128.49 1154.48 1178.39 1179.52 1180.52 1181.48 1182.55 1186.53 1210.55 1226.53 1229.48 1238.47 1255.63 1350.57 1372.36 1396.45 1406.64 1433.67

25.65 36.40 20.71 27.11 35.70 37.00 28.27 22.39 20.87 21.02 36.57 33.65 25.13 21.27 40.05 37.59 28.16 28.09

41434

1466.65

28.52

42404

1487.65

29.62

46928

1568.71

29.89

46961 48008 53957

1569.65 1577.69 1669.69

39.31 40.19 21.46

54976 56011

1689.74 1705.73

40.60 40.44

Protein name

Collagen α-2(I) chain

β-2-microglobulin Collagen α -1(I) chain Na+/K+-transporting ATPase subunit γ ISRLEPEDF Ig κ chain V-III region NG9 DFDDFNLED CD99 antigen-like protein 2 ApGEAGRDGNpG Collagen α-2(I) chain n.i. QDGRpGpPGPpG Collagen α-1(I) chain pGDRGEpGPpGP Collagen α-1(I) chain n.i. n.i. n.i. DDGEAGKpGRpG Collagen α-1(I) chain n.i. n.i. n.i. n.i. n.i. n.i. n.i. n.i. n.i. GLpGPpGPp Collagen α-2(V) chain GEGGKpG GpSGPpGPD Collagen α-2(I) chain GNKGEpG GLSMDGGG Na+/K+-transporting ATPase subunit γ SPKGDVDP EGSPGHPGQ Collagen α-1(III) chain pGPpGPpG n.i. n.i. DEAGSEADH Fibrinogen EGTHSTK α chain n.i. n.i.

Mean amp (SD)

Frequency

Mean amp (SD)

Frequency

Fold change in mean amp towards normal

3.6E−02

312 (593)

80

712 (1084)

93

0.44

B2MG_HUMAN CO1A1_HUMAN ATNG_HUMAN

4.8E−04 6.2E−04 1.4E−04 7.0E−04 2.6E−03 4.9E−03 1.8E−06 6.3E−05

0.70 0.67 0.71 0.66 0.67 0.63 0.73 0.74

1.6E−02 1.7E−02 8.3E−03 1.7E−02 3.5E−02 4.7E−02 1.2E−03 4.9E−03

85 14 45 5 100 8 1 348

48 15 35 10 30 13 3 90

282 63 189 96 325 101 112 853

(740) (144) (230) (185) (716) (311) (217) (665)

77 48 67 38 60 37 47 95

0.30 0.22 0.24 0.05 0.31 0.08 0.01 0.41

80–88

KV303_HUMAN

1.8E−03 0.64

2.9E−02

9 (33)

10

55 (102)

38

0.16

26–34

C99L2_HUMAN

5.4E−05 0.74

4.5E−03

1037 (1000)

75

2567 (2609)

100

0.40

920–931

CO1A2_HUMAN

561–572 799–810

CO1A1_HUMAN CO1A1_HUMAN

231–242

CO1A1_HUMAN

678–693

CO5A2_HUMAN

4.8E−03 2.5E−03 1.3E−04 3.9E−05 5.5E−03 5.6E−04 2.5E−04 1.8E−03 2.5E−03 1.6E−03 5.3E−04 5.0E−04 2.1E−03 1.2E−03 7.2E−06 8.2E−04 4.5E−05 6.1E−04

4.7E−02 3.5E−02 7.9E−03 4.5E−03 5.0E−02 1.6E−02 1.1E−02 2.9E−02 3.5E−02 2.7E−02 1.6E−02 1.6E−02 3.1E−02 2.2E−02 2.4E−03 1.9E−02 4.5E−03 1.7E−02

610–625

CO1A2_HUMAN

3–18

ATNG_HUMAN

AA d

SwissProt/ TrEMBLE name

503–511

CO1A2_HUMAN

2.7E−03 0.68

112–119 801–811 7–18

1175–1191 CO3A1_HUMAN

605–620

FIBA_HUMAN

Pe

AUC FDR-adj. Pf

0.66 0.66 0.71 0.74 0.65 0.66 0.69 0.68 0.67 0.69 0.68 0.67 0.65 0.67 0.67 0.68 0.73 0.65

Distribution in RCC case (N = 40)

55 27 30 426 190 13 18 449 76 224 36 25 83 19 63 41 89 73

(188) (43) (84) (16) (256) (26) (4) (365)

Distribution in non-RCC controls (N = 68)

(74) (61) (66) (398) (454) (42) (43) (490) (193) (317) (136) (75) (137) (59) (113) (103) (153) (102)

53 30 25 75 23 10 20 80 30 60 15 15 48 20 38 25 40 43

228 127 206 838 452 130 116 913 214 596 187 265 25 117 3 437 395 25

(287) (288) (309) (483) (840) (224) (169) (873) (330) (681) (353) (568) (70) (171) (19) (953) (455) (89)

65 55 57 97 53 40 53 85 58 80 48 47 18 47 3 53 72 12

0.24 0.21 0.15 0.51 0.42 0.10 0.16 0.49 0.36 0.38 0.19 0.09 3.32 0.16 21.00 0.09 0.23 2.92

3.9E−05 0.73 4.5E−03

77 (223)

28

380 (450)

63

0.20

3.0E−03 0.67 3.7E−02

168 (191)

55

483 (524)

70

0.35

9.3E−05 0.68 5.9E−03

3 (18)

5

91 (194)

40

0.03

5.8E−04 0.65 1.7E−02 2.4E−03 0.66 3.4E−02 1.0E−03 0.69 2.1E−02

3 (16) 190 (585) 1092 (2156)

5 25 85

154 (292) 1399 (2525) 283 (505)

33 50 70

0.02 0.14 3.86

4.2E−03 0.64 4.3E−02 9.8E−04 0.66 2.0E−02

119 (426) 14 (43)

20 13

1634 (2864) 194 (354)

43 40

0.07 0.07

JO U R N A L OF PR O TE O MI CS 98 ( 20 1 4 ) 4 4 –58

Mass [Da]

Statistical analysis

characteristics

case (N = 40) c

migration time [min]

Sequence

56662 58880

1720.69 1764.68

19.67 19.91

59745 60126

1782.84 1793.88

25.91 32.37

60352 60871 61087 61984 62135 64621

1798.72 1809.88 1814.72 1835.71 1838.82 1889.87

36.95 32.30 37.17 19.91 27.06 33.07

69080 69979

1976.88 1996.79

32.38 20.98

75128

2093.92

33.78

83458 84850 91342

2248.99 2272.23 2403.19

25.99 23.88 24.94

97506

2545.12

28.20

2554.19

34.45

n.i. KmHEGDEG PGHHHKPG n.i. EEAPSLRPAP PPISGGGY n.i. n.i. n.i. n.i. n.i. TTDERGpPG EQGPpGPpGP n.i. EETEDANEEA PLRDRSH PGppGDQGPP GPDGPR GApGPpG n.i. n.i. DDILASPPR LPEPQPYPGAPHH GPpGAD GQpGAKG EpGDAG AKGDA GPpGP pGLPGNP GRDGD VGLpGDp GLPGqPGL

(53) 98720

7 2565.14

8.73 23.74

99746

2583.20

28.31

100344

2599.19

28.28

108832

2770.25

29.39

109301

2779.30

35.14

97861

REQGHQ KERNQE mEEG GEEEH AGPpGApGA PGApGP VGPAGKS GDRGE TGP AGpPG ApGApGA PGpVGPA GKSGD RGETGP GPpGADG QpGAKG EpGDAG AKGDAG pPGP AGP n.i.

Protein name

AA

SwissProt/ TrEMBLE name

P

e

AUC FDR-adj. Pf

controls (N = 68)

Mean amp (SD)

Frequency

Mean amp (SD)

change in mean amp towards normal

Protein S100-A9

93–108

S100A9_HUMAN

3.5E−03 0.65 4.0E−02 2.4E−04 0.71 1.1E−02

45 (123) 101 (245)

28 35

261 (468) 584 (1102)

50 63

0.17 0.17

Fibrinogen β chain

54–71

FIBB_HUMAN

9.3E−04 0.69 2.0E−02 2.6E−03 0.68 3.5E−02

618 (468) 585 (507)

78 88

322 (349) 327 (423)

62 65

1.92 1.79

65 239 18 671 152 192

(166) (511) (64) (779) (162) (202)

25 40 10 75 63 60

203 29 95 2049 68 33

(273) (83) (182) (2668) (105) (69)

55 18 35 80 38 23

0.32 8.24 0.19 0.33 2.24 5.82

Collagen α1(IX) chain

264–282

CO9A1_HUMAN

1.5E−03 4.9E−03 4.6E−03 5.5E−03 4.9E−03 9.8E−06

Sarcalumenin

480–496

SRCA_HUMAN

4.8E−03 0.66 4.7E−02 9.1E−04 0.70 2.0E−02

148 (228) 470 (541)

45 70

283 (278) 1403 (1500)

72 82

0.52 0.33

1.2E−03 0.67 2.2E−02

210 (256)

63

103 (264)

30

2.04

19568 (12765) 353 (325) 122 (173)

98 78 55

12738 (11528) 126 (213) 57 (141)

70 45 25

1.54 2.80 2.14

411 (587)

63

114 (227)

38

3.61

Collagen α-4(IV) chain

Collagen α1(XVIII) chain Collagen α-1(I) chain

Collagen α-5(IV) chain

1257–1279 CO4A4_HUMAN

0.67 0.63 0.63 0.66 0.65 0.73

2.5E−02 4.7E−02 4.6E−02 5.0E−02 4.7E−02 2.6E−03

3.8E−03 0.67 4.1E−02 6.7E−05 0.73 4.9E−03 1534–1555 CO18A1_HUMAN 3.6E−03 0.65 4.0E−02 815–843

CO1A1_HUMAN

1.6E−03 0.67 2.7E−02

665–691

CO4A5_HUMAN

2.5E−04 0.65

1.1E−02

96 (170)

35

715–735

RPGR_HUMAN

4.7E−04 0.68 1.6E−02

266 (478)

58

71 (212)

28

3.75

11

X-linked retinitis pigmentosa GTPase regulator, isoform 6 Collagen α-1(I) chain

1042–1071 CO1A1_HUMAN

2.3E−05 0.75 3.9E−03

1205 (843)

95

568 (603)

73

2.12

Collagen α-1(I) chain

1042–1071 CO1A1_HUMAN

5.1E−05 0.74 4.5E−03

1348 (1538)

93

522 (839)

67

2.58

4.1E−05 0.71 4.5E−03

98 (109)

58

22 (52)

22

4.45

2.9E−03 0.64 3.7E−02

95 (164)

45

18 (41)

20

5.28

Collagen α-1(I) chain

815–846

CO1A1_HUMAN

49

(continued on next page)

JO U R N A L OF P ROTE O MI CS 9 8 ( 20 1 4 ) 4 4–5 8

CE

Mass [Da]

d

50

Table 4 (continued) Peptide-ID b

CE–MS characteristics CE migration time [min]

110333

2808.34

24.39

111304 114086

2834.19 2907.35

22.47 35.96

117371

2989.45

24.43

118709

3023.42

35.04

124268

3168.36

24.70

124688 124735 124909 126318

3185.47 3187.36 3194.42 3240.65

25.47 35.62 30.40 25.45

129940

3333.72

23.83

130662

3356.52

25.57

Sequence c

EFTP PVQAAY QKVVAG VANALA HKYH n.i. TGEVGA VGPpGFAG EKGPSG EAGTA GPpGT pGP VMGFPG pKGAAG EPGKAGE RGVpGpp GAVGPAG AVQEWN QNDTYNL YISDTR GIYFTL GEpGRDG VpGGpGM RGmpGSp GGpGSD GKpGPpG n.i. n.i. n.i. TGAkGAA GLpGVA GApGLp GPRGIp GPVGAA GATGARG DVGSYQ EKVDV VLGPIQL QTPPRR EEEPR SGPQ GPGGpP GpKGNS GEPGApG SKGDTGA KGEPGpVG

Protein name

AA d

Statistical analysis SwissProt/ TrEMBLE name

Pe

AUC FDR-adj. Pf

Distribution in RCC case (N = 40)

Distribution in non-RCC controls (N = 68)

Mean amp (SD)

Frequency

Mean amp (SD)

Frequency

Fold change in mean amp towards normal

Hemoglobin subunit β

122–147

HBB_HUMAN

6.5E−06 0.74 2.4E−03

672 (741)

65

154 (418)

23

4.36

Collagen α-2(I) chain

831–863

CO1A2_HUMAN

2.7E−04 0.69 1.1E−02 5.0E−03 0.66 4.8E−02

25 (72) 347 (318)

20 75

149 (244) 192 (293)

53 57

0.17 1.81

Collagen α-1(I) chain

579–611

CO1A1_HUMAN

8.0E−04 0.68 1.8E−02

462 (546)

65

163 (325)

35

2.83

VPS10 domain-containing receptor SorCS3

449–473

SORC3_HUMAN

3.5E−04 0.64 1.4E−02

113 (332)

33

82 (510)

5

1.38

Collagen α-1(III) chain

522–555

CO3A1_HUMAN

7.1E−04 0.70 1.7E−02

784 (804)

85

311 (395)

60

2.52

CO1A2_HUMAN

6.4E−04 7.1E−04 2.9E−03 3.9E−03

187 406 125 255

40 73 38 35

15 154 74 79

(57) (267) (335) (300)

13 48 12 10

12.47 2.64 1.69 3.23

Collagen α2(I) chain

Deleted in malignant brain tumors 1 protein Collagen α-1(I) chain

306–343

2385–2413 DMBT1_HUMAN

418–455

CO1A1_HUMAN

0.65 0.69 0.63 0.62

1.7E−02 1.7E−02 3.7E−02 4.1E−02

(403) (469) (267) (739)

1.6E−04 0.70 8.7E−03

872 (2200)

63

217 (682)

28

4.02

1.3E−03 0.64 2.3E−02

170 (311)

40

32 (135)

13

5.31

JO U R N A L OF PR O TE O MI CS 98 ( 20 1 4 ) 4 4 –58

Mass [Da]

Amino acid sequence information

characteristics Mass [Da]

case (N = 40) migration time [min]

130947 131589 132053

3366.56 3385.55 3401.60

31.00 25.49 25.47

132383

3405.48

25.97

132834

3421.56

25.99

134488 156081

3473.60 4289.93

33.01 28.78

156445

4305.94

28.83

165737 168652 168919 169424 170584 171854

4670.15 4817.16 4833.15 4863.16 4942.53 5039.44

25.86 23.90 23.92 26.74 25.63 25.85

Sequence

n.i. n.i. DPGIpGLD RSGFPGET GSpGIPGHQ GEMGPLGQR ARGNDGA RGSDGQPG PpGppGT AGFpGSpG AKGEVGP GLEGpK GEVGAp GSKGEA GPTGPm GAMGPLG PRGMpG n.i. ARGND GARGSD GQpGppG PPGTAGF PGSpGAK GEVGpAG SpGSNGApG ARGND GARGSDGQ pGpPGP pGTAGFpGS pGAKGEVGpA GSpGSNGApG n.i. n.i. n.i. n.i. n.i. n.i.

Protein name

AA

d

SwissProt/ TrEMBLE name

P

e

AUC FDR-adj. Pf

controls (N = 68)

Mean amp (SD)

Frequency

Mean amp (SD)

change in mean amp towards normal

130 (357) 4458 (3161) 2609 (2091)

40 88 90

15 (49) 1948 (2440) 1299 (1192)

12 57 80

8.67 2.29 2.01

Collagen α-3(IV) chain

840–873

CO4A3_HUMAN

6.5E−04 0.65 1.7E−02 3.9E−05 0.74 4.5E−03 1.1E−03 0.69 2.2E−02

Collagen α1(III) chain

319–355

CO3A1_HUMAN

5.4E−03 0.64 4.9E−02

2318 (3153)

48

419 (1152)

28

5.53

Collagen α-2(V) chain

306–342

CO5A2_HUMAN

7.3E−05 0.73 4.9E−03

2407 (1872)

93

1065 (1015)

75

2.26

Collagen α1(III) chain

319–366

CO3A1_HUMAN

5.1E−03 0.62 4.8E−02 3.9E−03 0.67 4.1E−02

96 (203) 4647 (3306)

38 85

12 (43) 2794 (3013)

15 70

8.00 1.66

Collagen α-1(III) chain

319–366

CO3A1_HUMAN

2.9E−03 0.68 3.7E−02

4700 (2366)

98

3274 (1951)

98

1.44

3.1E−03 4.5E−04 2.6E−04 1.7E−04 1.7E−03 2.1E−03

773 1156 1300 604 536 2036

58 50 48 75 40 43

32 17 13 43 15 17

5.26 3.25 3.26 2.52 10.94 8.05

0.66 0.67 0.67 0.71 0.64 0.64

3.8E−02 1.6E−02 1.1E−02 9.2E−03 2.8E−02 3.1E−02

(1739) (3279) (4150) (551) (1034) (8087)

147 356 399 240 49 253

(326) (1487) (1608) (443) (156) (926)

JO U R N A L OF P ROTE O MI CS 9 8 ( 20 1 4 ) 4 4–5 8

CE

c

The table lists the peptide identification number (Peptide-ID), experimental mass (in Da) and CE migration time (in min) for all peptides in the urine RCC classifier. For all sequence-identified peptides, the amino acid sequence, the name of the protein precursor and the amino acid positions within the protein's primary sequence (according to UniProtKB) are presented. For MS sequencing of peptides no fixed modification was selected and oxidation of methionine, lysine and proline were set as variable modifications. Mass error window of 10 ppm and 0.05 Da were allowed for MS and MS/MS, respectively. Further, the frequency and the mean amplitude in the RCC patient and non-diseased control groups of the training cohort, as well as the false discovery rate-adjusted P and AUC value for the peptides in the group comparison is provided. The fold change of the peptide's mean amplitude in the RCC patient group towards the normal reference group was used as an indicator if a peptide is down-regulated (values < 1) or up-regulated (values > 1) during RCC progression. a Abbreviations: AA, amino acid; Amp, amplitude; AUC, area under the curve; CE-MS, capillary electrophoresis mass spectrometry; Da, Dalton; FDR, false discovery rate; min, minutes; N, number of patients; RCC, renal cell carcinoma; SD, standard deviation. b Peptide identification number. c Lower case p, k and m indicate hydroxyproline, hydroxylysine and oxidized methionine. d Amino acid position according to UniProt Knowledge Base numbering. e Wilcoxon P value. f FDR adjustment of P value according to the method of Benjamini and Hochberg [28].

51

52

JO U R N A L OF PR O TE O MI CS 98 ( 20 1 4 ) 4 4 –58

interval (CI) from 0.93 to 1.0 in ROC analysis (Fig. 1, left panel). LOOCV is done by using a single patient (case or control) from the original training samples as the validation data, and the remaining observations as the training data. This is repeated such that each patient in the sample is used once as the validation data. Following the recommendations for biomarker identification and qualification in clinical proteomics [32], the classifier was subsequently assessed in an independent set of 30 RCC patients and 46 controls. The classifier enabled differentiation of RCC from control with an AUC of 0.92 (95% CI, 0.84 to 0.97; P = 0.0001) (Fig. 1, right panel). At a cut-off of 0.38 this resulted in correct classification of 25 of the 30 RCC patients (80% sensitivity [95% CI, 65% to 94%]) and 39 of the 46 normal controls (87% specificity [95% CI, 74% to 95%]). The specificity of RCC classification was further assessed by comparison with previous data from an additional set of normal controls consisting of unselected non-diseased subjects (not age-matched, N = 84), smokers (N = 62), subjects with a BMI >30 (N = 25) or with hypertension (N = 33) (Table 1, specificity analysis, normal controls). Specificity values were evaluated at the sensitivity level of 80% for RCC detection in the validation cohort, while the predetermined cut-off was 0.38. As presented by the Box-and-Whisker representation of classification scores in Fig. 2A for these reference groups, the classifier is highly discriminative and the distance between the lower quartile for the RCC case and the upper quartiles of these reference groups amounts to more than 0.4 classification score units. To evaluate the ability of the classifier to discriminate RCC from closely related cancer types and non-malignant renal and systemic diseases we also investigated large sets of patient samples from the following patient groups: diabetic nephropathy (N = 195), focal segmental glomerulosclerosis (N = 54), membranous glomerulonephritis (N = 65), systemic lupus erythematosus (N = 46), IgA nephropathy (N = 126), vasculitis (N = 121), cardiovascular disease (N = 33) and bladder cancer (N = 219) (Table 1, specificity analysis, patients with non-cancer renal diseases and other urological cancer types). As revealed in Fig. 2B the classifier clearly discriminates between RCC and most of these other diseases. However, a substantial degree of overlap (~25%) was observed for vasculitis (64% specificity) and bladder cancer (76% specificity), possibly indicating vascular damage as origin of some of the RCC peptide markers. A subgroup specificity evaluation was performed for the different stages of the bladder cancer group resulting in specificity values of 84% for non-invasive (pTa and pT1 stages) and 67% for invasive bladder cancers (pT2+ and pTis). In the post-hoc rank sum test, the difference of the classification scores between the two bladder cancer sub-groups is significant (Fig. 2C).

3.2. Peptide sequencing reveals differences in protein degradation between RCC and controls From the 40 sequenced peptides, 27 were fragments of collagen chains. This is not surprising as collagen fragments represent the major component of the low molecular weight urinary proteome and cancer-related processes such as invasion potentially result in collagen breakdown. Additional prominent peptides were from fibrinogen chains and Na/K-transporting ATPase subunit γ.

We subsequently examined an RCC-specific classifier when using only the 40 sequenced peptides instead of the full RCC classifier composed of both the 86 sequenced and nonsequenced RCC peptides described above. Combining these 40 sequenced biomarkers enabled classification of the validation set with a sensitivity of 77%, and a specificity of 85%; the AUC being 0.87 (95% CI, 0.77 to 0.94; P = 0.0001). In comparison to the marker model including all 86 statistically significant RCC peptide markers, this result shows that additional, yet unidentified biomarkers enhance the performance of the classification. However, an AUC of 0.87, as observed here, clearly indicates relevance and validity of the sequenced biomarkers. Omitting the collagen fragments in the biomarker classifier reduced the specificity to 76%, which means that 11 out of the group of 46 normal controls were classified as false positives. In contrast, the sensitivity was not influenced and remained at the level of 77%. Sequencing of the identified peptides both from the discovery and validation phase patients revealed for the majority of collagen fragments > 1.4 kDa (70%) an increased urinary abundance in renal cancer, whereas smaller collagen fragments are decreased. This observation is supported by a correlation coefficient r of 0.48 for the relation of the fold change of mean amplitude in RCC towards normal and the molecular mass of the collagen fragments (P < 0.0001) (Fig. 3). The extended list of statistically significant sequence-identified peptides for the differentiation of patients with RCC (N = 70) and non-diseased controls (N = 114) as the basis for this analysis is presented in the Supplementary Table 1.

3.3. Protease frequency distribution analysis indicates reduced cathepsin activity in RCC This difference in distribution of small versus large urinary collagen fragments (Fig. 3) in RCC prompted us to study in detail the in silico protease activity involved in the generation of the peptide fragments. With this aim a mathematical model for the estimation of protease frequency distribution was applied, according to which a prediction of protease activity can be made based on the fold changes of the obtained RCC peptide marker sequences. The peptide sequence list (Supplementary Table 1) was screened against a matrix of 180 proteases through Proteasix [31]. Significantly reduced activity was predicted for cathepsins S, K and H, while enhanced activity was predicted for endothelin converting enzyme 2 and neutral protease (Table 5).

4. Discussion We have identified and validated a set of urinary peptide markers that enables detection of RCC with good sensitivity and specificity in a large study cohort that contains not only non-diseased controls, but also patients with similar diseases (other cancers), diseases affecting the same organ (kidney), or with systemic manifestations. Several of these diseases are characterized by similar symptoms, i.e. (micro-)hematuria. Several studies have already reported the identification of urinary candidate biomarkers for RCC detection and prognosis by proteomic analysis. However these studies resulted in lower sensitivity and specificity values than obtained for the RCC

JO U R N A L OF P ROTE O MI CS 9 8 ( 20 1 4 ) 4 4–5 8

53

peptide model described in the current study or did not include independent validation studies. Minamida et al. proposed 14-3-3 α/β as a urinary biomarker for RCC, detected by Western Blot analysis as screening method [33]. According to the above study, an AUC value of 0.8813 was reported after Western blot analysis in 89 samples from RCC patients and 76 healthy controls, but independent validation was not performed [33]. Kaya et al. showed that urinary nuclear matrix protein 22 (NMP22)

Fig. 1 – Diagrams representative of the ROC curves for the discovery (training) cohort after total cross-validation (left ROC curve) and the independent validation (test) cohort (right ROC curve). The RCC model was constructed based on 86 RCC specific peptides. Other ROC characteristics, such as area under the curve (AUC), 95% confidence intervals (CI), and P value are indicative for the differentiation of RCC patients from normal subjects, including hypertensive people, smokers and those with BMI > 30.

Fig. 2 – Specificity of RCC detection by the urine peptide marker model composed of the 86 RCC specific peptides. A post-hoc rank-test was performed for average rank differences between the non-RCC reference groups and the RCC case group (each with P < 0.05) after a significant result in the global Kruskal– Wallis test (P < 0.0001). A: Distribution of classification values for unselected normal controls and separately of individuals with smoking history, body mass index > 30 or more than one year of hypertension as well as center specific controls not used in the discovery phase of the study. The range between the lower quartile of the classification scores for the RCC case group and the upper quartiles of the classification scores for non-RCC reference groups is indicated by a gray semitransparent vertical bar visualizing the high degree of discriminative ability of the urine peptide marker model. Sensitivity and specificity values are also reported on the bottom part of the graph. B: Distribution of classification values for patients with bladder cancer and various non-cancer renal and systemic diseases, as obtained by analysis of historical datasets. The following disease etiologies were included in the specificity analysis as reference controls: diabetic nephropathy (DN, N = 195), focal segmental glomerulosclerosis (FSGS, N = 54), membranous glomerulonephritis (MGN, N = 65), systemic lupus erythematosus (SLE, N = 46), IgA nephropathy (IgAN, N = 126), vasculitis (N = 121), cardiovascular disease (CVD, N = 33) and bladder cancer (N = 219). Specificity values were calculated at the predetermined cut-off of 0.38 at the sensitivity for RCC detection in the validation cohort of 80%. C: Distribution analysis for non-invasive and invasive subgroups of patients with bladder cancer. This subgroup analysis reveals a statistically significant interference of RCC detection with invasive bladder cancer.

54

JO U R N A L OF PR O TE O MI CS 98 ( 20 1 4 ) 4 4 –58

Fig. 3 – Scatter diagram demonstrating the correlation of molecular mass to the fold change of mean amplitudes in RCC towards normal controls for all collagen fragments included in the extended list of statistically significant sequence-identified peptides for the differentiation of patients with RCC (N = 70) and non-diseased controls (N = 114) presented in the Supplementary Table 1. As revealed by a correlation coefficient r of 0.48 at a significance level p of 0.0001 a moderate positive correlation exists between the molecular mass of a collagen peptide and increased excretion into the urine. With the exception of one collagen α-2(I) chain derived peptide all collagen peptides < 1.4 kDa detected by CE–MS are decreased (N = 46), whereas 70% of peptides with > 1.4 kDa are increased in their excretion levels compared to normal (N = 111).

concentrations allowed the detection of 23 out of 30 RCC patients (NMP22 urinary levels > 10 U/ml) [34]. Finally, Rogers et al. studied the urinary proteome using SELDI profiling resulting in a range of specificity and sensitivity values between 41.0–76.6% upon the use of two independent sets of validation in 32 and 80 samples, respectively [35]. The current study presents a biomarker model for RCC diagnosis in a large number of urine samples with superior specificity and sensitivity compared to the reported studies so far. In total, 1153 urine samples were analyzed to evaluate the RCC-specific peptide marker model, 76 for initial independent validation and 1077 for cross reactivity evaluation. This led in the independent test set of RCC patients and non-diseased controls to an AUC of 0.92. The data presented in this study indicate that multiple biomarker based models composed of surrogate markers for RCC associated pathological processes may be employed as diagnostic tool. However, the individual biomarkers, that are the basis of such a multivariate biomarker model, must be well defined to provide a link to the corresponding disease processes. Ideally, all of these biomarkers should be sequenced. As outlined recently [10,36], such a goal is almost impossible to reach, mostly due to post-translational modifications and resistance to fragmentation of larger peptides, which hinder MS/MS sequencing.

However, we were able to obtain the sequences from 40 of the 86 peptide markers included in the RCC marker model. Of those sequences that are available to date, most were fragments of collagen chains. Collagen fragments represent the majority of urinary peptides, even in healthy controls [24]. Upon sequencing of the peptides including both patients from the discovery and the validation phase of the study, collagen fragments below 1.4 kDa in size are consistently downregulated in RCC patients. This likely reflects molecular changes in the physiological turnover of extracellular matrix (ECM), resulting in alterations of degradation products excreted into the urine. ECM turnover was found to be reduced in several pathological conditions (e.g. diabetes and DN [11,37]). This leads us to the hypothesis that (i) differential protease activity during RCC progression is responsible for the distinct pattern of differentially identified peptide sequences (especially collagens), and (ii) the urinary peptides of the RCC peptide marker panel are surrogate markers for local changes in the proteolytic environment of the tumorous kidney. Besides peptide fragments of various collagens, peptides in the RCC peptide marker panel potentially originate from serum proteins including fibrinogen chains, immunoglobulin Fc regions and hemoglobin subunits and from proteins most likely expressed in the kidney such as Na/K-transporting ATPase subunit γ, retinitis pigmentosa GTPase regulator, VPS10 domain-containing receptor SorCS3 and the endothelial adhesion molecule CD99 antigen-like protein 2. The observation that for some of these proteins lower levels of peptide fragments were found whereas for others higher quantities were detected in urine during RCC progression argues against the possibility that their differential excretion into urine might be attributed to kidney damage and breakdown of the filtration barrier. Most evident in this context is the finding that peptides from inflammatory and immune proteins are decreased whereas those involved in coagulation and platelet aggregation are increased. A noteworthy aspect in respect to migration and invasion of renal tumor cells is the identification of increased levels of a C-terminal fragment of the deleted in malignant brain tumors 1 (DMBT1) protein. Upon its secretion from epithelial cells, polymerization and ECM assembly, DMBT1 binds to α6-containing integrins, which are important cellular mediators of ECM adhesion, proteolysis and cell movement [38,39]. Among the most prominent findings was an increase of fibrinogen-derived peptide levels in RCC patients. This is in line with a recent study reporting significantly higher plasma fibrinogen levels in RCC than in renal benign tumor patients [40]. Similarly, Du et al. [41] assume an active role of fibrinogen in the RCC disease process. Fibrinogen yields monomers that polymerize into fibrin and it acts (besides vonWillebrand factor and fibronectin) as a bridging protein and cofactor to promote platelet aggregation. The latter ability of fibrinogen provides a way for the interaction of cancer cells with platelets which is an import step in cancer metastasis since platelets facilitate migration of tumor cells, as well as their invasion [42–44] and their arrest within the vasculature [45–48]. During this interplay both the tumor cells and the activated platelets release proteolytic enzymes, i.e. gelatinase, heparanase and matrix metalloproteinases, that degrade the structural components of the vessel basement membrane

55

JO U R N A L OF P ROTE O MI CS 9 8 ( 20 1 4 ) 4 4–5 8

Table 5 – Top 10 of significantly altered proteases in the protease frequency distribution analysis. RCC peptide biomarker sequences were investigated based on their potential cleavage sites, for in silico prediction of protease activity. Top 10 significantly altered proteases are presented, as predicted to be with either enhanced or decreased activity. Xscore is a correlation score which represents the predicted protease activity based on urinary abundance of potential protease target peptides in RCC. Top-10 ranked proteases with increased and decreased activity Enhanced activity

Xscore

A disintegrin and metalloproteinase domain-containing protein 8 (ADAM8) Tissue-type plasminogen activator (PLAT) Thermolysin (NPR)

868.00

Kallikrein-1 (KLK1)

610.17

Kallikrein-7 (KLK7)

536.81

Endothelin-converting enzyme 2 (ECE2)

528.00

Transmembrane protease serine 6 (TMPRSS6) Suppressor of tumorigenicity 14 protein (ST14) A disintegrin and metalloproteinase with thrombospondin motifs 1 (ADAMTS1) Vitamin K-dependent protein C (PROC)

500.00

Reported tumor association Prognostic marker of poor survival in hepatocellular [56], and squamous cell carcinomas [55]

745.75 700.78

500.00 434.00

Differentially expressed in various tumors and RCC [61] Differentially expressed in various tumors and RCC [61] Increased expression of the endothelin axis during RCC [62,63] Increased expression of matriptase in RCC [57] Associated with cancer [58–60] Promotes pro-tumorigenic changes, peritumoral remodeling, tumor progression and metastasis [54]

434.00

[49,50] providing the rational to search for proteases involved in RCC progression. Apart from their potential utility as a panel of biomarkers, the identified peptide sequences were therefore used to search for proteases differentially expressed in RCC. The protease activity analysis based on the possible cleavage sites within these sequences indicated a multitude of proteases being potentially involved in RCC progression. This is most probably indicative of a whole range of proteases that need to be orchestrated to promote renal carcinoma development. More specifically, our data confirmed previous observations of dysregulated protease activity in the context of malignancies and RCC for cathepsins [51–53] ADAMTS [54–56], matriptases [57–60], kallikreins [61] and endothelin [62,63]. Among those, elevated ADAMTS1 promotes pro-tumorigenic changes, facilitating peritumoral remodeling, tumor progression and metastasis [54] while elevated ADAM8 expression is a poor prognostic biomarker for survival of patients with hepatocellular [56] or head and neck squamous cell carcinomas [55]. Moreover, matriptase (Suppressor of tumorigenicity 14 protein, ST14) and matriptase-2 (Transmembrane protease serine 6, TMPRSS6) have been associated with cancer [59,60] and increased expression of matriptase was found in RCC [57]. Kallikreins and kallikrein related peptidases (KLKs) such as KLK1, KLK6 or KLK7 are differentially expressed in various tumors and also in RCC. Of special note is the association of decreased KLK6 expression with poor disease outcome in RCC [61]. Finally, the pathological role of endothelin signaling in RCC and the increased expression of the endothelin axis

Decreased activity

Xscore

Reported tumor association

Cathepsin S (CTSS)

−1183.08 Promote neovascularization and tumor growth [68]

Cathepsin H (CTSH) Cathepsin L1 (CTSL1) Myeloblastin (PRTN3) Myelin basic protein (MBP) Cathepsin K (CTSK)

−1103.53

Beta-secretase 1 (BACE1) Cathepsin B (CTSB) Kallikrein-6 (KLK6)

Presenilin-1 (PSEN1)

−964.29 −936.00 −841.41 −661.73 Marker for Xp11 translocation RCC [51,52] −600.00 −581.79 −500.00 Differentially expressed in RCC [61]

−428.57

during the disease should also be mentioned [62,63] thus strengthening our observation of increased endothelinconverting enzyme 2 activity.

5. Conclusions Collectively, our results demonstrate the utility of a multiplemarker approach, especially when attempting to distinguish between several different diseases that present, at least in part, with similar symptoms. We could not identify a single biomarker that enables detection of RCC with sufficient accuracy, especially in a typical clinical condition, when the challenge is not to distinguish between healthy and disease, but to distinguish between a distinct disease (RCC) and several other diseases, presenting with similar symptoms. This is exemplified by small renal masses. Currently, there is no diagnostic test available enabling accurate assessment of small renal masses. Therefore, urine based markers and their combination to a non-invasive test for RCC detection might improve the current clinical status and would greatly augment the diagnostic workup of renal lesions. The clinical utility of the described RCC marker panel would be to differentiate radiological indeterminate lesions, to obtain further information before ablative treatments (radiofrequency ablation, cryoablation) and to select patients with small renal masses for surveillance protocols. Additionally, marker information could provide useful information on

56

JO U R N A L OF PR O TE O MI CS 98 ( 20 1 4 ) 4 4 –58

the decision on targeted pharmacologic therapy in the setting of metastatic disease with the ability to monitor response. The major limitation of the current work is the proof-ofconcept, retrospective character of the study that necessitates further validation. In-depth validation of the classifier must be performed for the defined context of use in an appropriately powered prospective study. If in such a study, which is currently planned, a significant benefit is demonstrated, the multidimensional RCC classifier can then be immediately applied in clinical routine for the diagnosis of small renal masses by CE–MS, an aim which is already being realized for the detection of graft versus host disease [64] and cholangiocarcinoma [20]. In addition, CE–MS is employed in the large, multicenter PRIORITY trial, aimed at early identification of diabetic nephropathy, followed by intervention with spironolacton (www.eu-priority.org). For the use in routine clinical practice, the CE–MS system has proven to meet all analytical requirements as demonstrated in several technical reports and large-scale prospective and/or longitudinal clinical studies [22,64–67]. Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.jprot.2013.12.010.

Statement of competing financial interests Harald Mischak is the founder and co-owner of Mosaiques Diagnostics, who developed the CE–MS technology. Maria Frantzi, Julie Klein, Mohammed Dakna and Jochen Metzger are employed by Mosaiques Diagnostics.

Acknowledgements This work was supported in part by the grant PITN-GA2012-31750 BCMolMed to AV and HM. Further, contributions through the support of ECMC and CRUK Centre/Programme funding to RB is acknowledged. JK acknowledge ERA-EDTA (ALTF 82–2011) and Marie-Curie (ProteasiX, 300582 from FP7-PEOPLE-2011-IEF) for their financial support in the development of Proteasix.

REFERENCES

[1] Siegel R, Naishadham D, Jemal A. Cancer statistics. CA Cancer J Clin 2013;63:11–30. [2] Zini L, Perrotte P, Jeldres C, Capitanio U, Duclos A, Jolivet-Tremblay M, et al. A population-based comparison of survival after nephrectomy vs nonsurgical management for small renal masses. BJU Int 2009;103:899–904. [3] Van Poppel H, Da Pozzo L, Albrecht W, Matveev V, Bono A, Borkowski A, et al. A prospective, randomised EORTC intergroup phase 3 study comparing the oncologic outcome of elective nephron-sparing surgery and radical nephrectomy for low-stage renal cell carcinoma. Eur Urol 2011;59:543–52. [4] Ljungberg B, Bensalah K, Bex A, Canfield S, Dabestani S, Hofmann F, et al. Guidelines on renal cell carcinoma. Arnhem (The Netherlands): European Association of Urology (EAU); Mar 2013 [56 pp.].

[5] Merseburger AS, Kuczyk MA. Value of targeted therapies for renal cell cancer. Urologe A 2008;47:1303–10. [6] Grunwald V, Merseburger AS. The progression free survival-plateau with vascular endothelial growth factor receptor inhibitors — is there more to come? Eur J Cancer 2013;49:2504–11. [7] Grunwald V, Merseburger AS. Axitinib for the treatment of patients with advanced metastatic renal cell carcinoma (mRCC) after failure of prior systemic treatment. Onco Targets Ther 2012;5:111–7. [8] National Cancer Institute. Report of the kidney/bladder cancers progress review group, August 2002. BiblioGov; 2012 [91 pp.]. [9] Cairns P. Renal cell carcinoma. Cancer Biomark 2011;9:461–73. [10] Fliser D, Novak J, Thongboonkerd V, Argiles A, Jankowski V, Girolami MA, et al. Advances in urinary proteome analysis and biomarker discovery. J Am Soc Nephrol 2007;18:1057–71. [11] Rossing K, Mischak H, Dakna M, Zürbig P, Novak J, Julian BA, et al. Urinary proteomics in diabetes and CKD. J Am Soc Nephrol 2008;19:1283–90. [12] Grossman HB, Soloway M, Messing E, Katz G, Stein B, Kassabian V, et al. Surveillance for recurrent bladder cancer using a point-of-care proteomic assay. JAMA 2006;295:299–305. [13] Theodorescu D, Wittke S, Ross MM, Walden M, Conaway M, Just I, et al. Discovery and validation of new protein biomarkers for urothelial cancer: a prospective analysis. Lancet Oncol 2006;7:230–40. [14] Theodorescu D, Schiffer E, Bauer HW, Douwes F, Eichhorn F, Polley R, et al. Discovery and validation of urinary biomarkers for prostate cancer. Proteomics Clin Appl 2008;2:556–70. [15] Greene FL, editor. AJCC cancer staging manual. 6th ed. New York: Springer; 2002. [16] Haubitz M, Dhaygude A, Woywodt A. Mechanisms and markers of vascular damage in ANCA-associated vasculitis. Autoimmunity 2009;42:605–14. [17] Alkhalaf A, Zürbig P, Bakker SJ, Bilo HJ, Cerna M, Fischer C, et al. Multicentric validation of proteomic biomarkers in urine specific for diabetic nephropathy. PLoS One 2010;5: e13421. [18] Delles C, Schiffer E, von Zur Muhlen C, Peter K, Rossing P, Parving HH, et al. Urinary proteomic diagnosis of coronary artery disease: identification and clinical validation in 623 individuals. J Hypertens 2010;28:2316–22. [19] Schiffer E, Vlahou A, Petrolekas A, Stravodimos K, Tauber R, Geschwend JE, et al. Prediction of muscle-invasive bladder cancer using urinary proteomics. Clin Cancer Res 2009;15:4935–43. [20] Metzger J, Negm AA, Plentz RR, Weismuller TJ, Wedemeyer J, Karlsen TH, et al. Urine proteomic analysis differentiates cholangiocarcinoma from primary sclerosing cholangitis and other benign biliary disorders. Gut 2013;62:122–30. [21] Wittke S, Mischak H, Walden M, Kolch W, Radler T, Wiedemann K. Discovery of biomarkers in human urine and cerebrospinal fluid by capillary electrophoresis coupled to mass spectrometry: towards new diagnostic and therapeutic approaches. Electrophoresis 2005;26:1476–87. [22] Mischak H, Vlahou A, Ioannidis JP. Technical aspects and inter-laboratory variability in native peptide profiling: the CE–MS experience. Clin Biochem 2013;46:432–43. [23] Neuhoff N, Kaiser T, Wittke S, Krebs R, Pitt A, Burchard A, et al. Mass spectrometry for the detection of differentially expressed proteins: a comparison of surface-enhanced laser desorption/ionization and capillary electrophoresis/mass spectrometry. Rapid Commun Mass Spectrom 2004;18:149–56. [24] Coon JJ, Zurbig P, Dakna M, Dominiczak AF, Decramer S, Fliser D, et al. CE–MS analysis of the human urinary proteome for biomarker discovery and disease diagnostics. Proteomics Clin Appl 2008;2:964.

JO U R N A L OF P ROTE O MI CS 9 8 ( 20 1 4 ) 4 4–5 8

[25] Stalmach A, Albalat A, Mullen W, Mischak H. Recent advances in capillary electrophoresis coupled to mass spectrometry for clinical proteomic applications. Electrophoresis 2013;34:1452–64. [26] Decramer S, Wittke S, Mischak H, Zürbig P, Walden M, Bouissou F, et al. Predicting the clinical outcome of congenital unilateral ureteropelvic junction obstruction in newborn by urinary proteome analysis. Nat Med 2006;12:398–400. [27] DeLeo JM. Receiver operating characteristic laboratory (ROCLAB): software for developing decision strategies that account for uncertainty. Second International Symposium on. Uncertainty Modeling and, Analysis; 1993. p. 318–25. [28] Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc Ser B Methodol 1995;57:125–33. [29] Kuznetsova T, Mischak H, Mullen W, Staessen JA. Urinary proteome analysis in hypertensive patients with left ventricular diastolic dysfunction. Eur Heart J 2012;33:2342–50. [30] Zürbig P, Renfrow MB, Schiffer E, Novak J, Walden M, Wittke S, et al. Biomarker discovery by CE–MS enables sequence analysis via MS/MS with platform-independent separation. Electrophoresis 2006;27:2111–25. [31] Klein J, Eales J, Zurbig P, Vlahou A, Mischak H, Stevens R. Proteasix: a tool for automated and large-scale prediction of proteases involved in naturally occurring peptide generation. Proteomics 2013;13:1077–82. [32] Mischak H, Allmaier G, Apweiler R, Attwood T, Baumann M, Benigni A, et al. Recommendations for biomarker identification and qualification in clinical proteomics. Sci Transl Med 2010;2:46ps2. [33] Minamida S, Iwamura M, Kodera Y, Kawashima Y, Tabata K, Matsumoto K, et al. 14-3-3 protein beta/alpha as a urinary biomarker for renal cell carcinoma: proteomic analysis of cyst fluid. Anal Bioanal Chem 2011;401:245–52. [34] Kaya K, Ayan S, Gokce G, Kilicarslan H, Yildiz E, Gultekin EY. Urinary nuclear matrix protein 22 for diagnosis of renal cell carcinoma. Scand J Urol Nephrol 2005;39:25–9. [35] Rogers MA, Clarke P, Noble J, Munro NP, Paul A, Selby PJ, et al. Proteomic profiling of urinary proteins in renal cancer by surface enhanced laser desorption ionization and neural-network analysis: identification of key issues affecting potential clinical utility. Cancer Res 2003;63:6971–83. [36] Good DM, Thongboonkerd V, Novak J, Bascands JL, Schanstra JP, Coon JJ, et al. Body fluid proteomics for biomarker discovery: lessons from the past hold the key to success in the future. J Proteome Res 2007;6:4549–55. [37] Rossing K, Mischak H, Rossing P, Schanstra JP, Wiseman A, Maahs DM. The urinary proteome in diabetes and diabetes-associated complications: new ways to assess disease progression and evaluate therapy. Proteomics Clin Appl 2008;2:997–1007. [38] Pozzi A, Zent R. Integrins, extracellular matrix, and terminal differentiation of renal epithelial cells. J Am Soc Nephrol 2008;19:1043–4. [39] Lochter A, Navre M, Werb Z, Bissell MJ. alpha1 and alpha2 integrins mediate invasive activity of mouse mammary carcinoma cells through regulation of stromelysin-1 expression. Mol Biol Cell 1999;10:271–82. [40] Xiao B, Ma LL, Zhang SD, Xiao CL, Lu J, Hong K, et al. Correlation between coagulation function, tumor stage and metastasis in patients with renal cell carcinoma: a retrospective study. Chin Med J (Engl) 2011;124:1205–8. [41] Du J, Zheng JH, Chen XS, Yang Q, Zhang YH, Zhou L, et al. High preoperative plasma fibrinogen is an independent predictor of distant metastasis and poor prognosis in renal cell carcinoma. Int J Clin Oncol 2013;18:517–23.

57

[42] Sarach MA, Rovasio RA, Eynard AR. Platelet factors induce chemotactic migration of murine mammary adenocarcinoma cells with different metastatic capabilities. Int J Exp Pathol 1993;74:511–7. [43] Poggi A, Vicenzi E, Cioce V, Wasteson A. Platelet contribution to cancer cell growth and migration: the role of platelet growth factors. Haemostasis 1988;18:18–28. [44] Belloc C, Lu H, Soria C, Fridman R, Legrand Y, Menashi S. The effect of platelets on invasiveness and protease production of human mammary tumor cells. Int J Cancer 1995;60:413–7. [45] Lewalle JM, Castronovo V, Goffinet G, Foidart JM. Malignant cell attachment to endothelium of ex vivo perfused human umbilical vein. Modulation by platelets, plasma and fibronectin. Thromb Res 1991;62:287–98. [46] Felding-Habermann B, O'Toole TE, Smith JW, Fransvea E, Ruggeri ZM, Ginsberg MH, et al. Integrin activation controls metastasis in human breast cancer. Proc Natl Acad Sci U S A 2001;98:1853–8. [47] Jones DS, Wallace AC, Fraser EE. Sequence of events in experimental metastases of Walker 256 tumor: light, immunofluorescent, and electron microscopic observations. J Natl Cancer Inst 1971;46:493–504. [48] Labelle M, Hynes RO. The initial hours of metastasis: the importance of cooperative host–tumor cell interactions during hematogenous dissemination. Cancer Discov 2012;2:1091–9. [49] Olas B, Wachowicz B, Mielicki WP. Cancer procoagulant and blood platelet activation. Cancer Lett 2001;169:165–71. [50] Honn KV, Cavanaugh P, Evens C, Taylor JD, Sloane BF. Tumor cell-platelet aggregation: induced by cathepsin B-like proteinase and inhibited by prostacyclin. Science 1982;217:540–2. [51] Zheng G, Martignoni G, Antonescu C, Montgomery E, Eberhart C, Netto G, et al. A broad survey of cathepsin K immunoreactivity in human neoplasms. Am J Clin Pathol 2013;139:151–9. [52] Argani P, Hicks J, De Marzo AM, Albadine R, Illei PB, Ladanyi M, et al. Xp11 translocation renal cell carcinoma (RCC): extended immunohistochemical profile emphasizing novel RCC markers. Am J Surg Pathol 2010;34:1295–303. [53] Merseburger AS, Hennenlotter J, Simon P, Ohneseit PA, Kuehs U, Kruck S, et al. Cathepsin D expression in renal cell cancer—clinical implications. Eur Urol 2005;48:519–26. [54] Tan Ide A, Ricciardelli C, Russell DL. The metalloproteinase ADAMTS1: a comprehensive review of its role in tumorigenic and metastatic pathways. Int J Cancer 2013;133:2263–76. [55] Zielinski V, Brunner M, Heiduschka G, Schneider S, Seemann R, Erovic B, et al. ADAM8 in squamous cell carcinoma of the head and neck: a retrospective study. BMC Cancer 2012;12:76. [56] Jiang C, Zhang Y, Yu HF, Yu XT, Zhou SJ, Tan YF. Expression of ADAM8 and its clinical values in diagnosis and prognosis of hepatocellular carcinoma. Tumour Biol 2012;33:2167–72. [57] Jin JS, Chen A, Hsieh DS, Yao CW, Cheng MF, Lin YF. Expression of serine protease matriptase in renal cell carcinoma: correlation of tissue microarray immunohistochemical expression analysis results with clinicopathological parameters. Int J Surg Pathol 2006;14:65–72. [58] Sanders AJ, Webb SL, Parr C, Mason MD, Jiang WG. The type II transmembrane serine protease, matriptase-2: possible links to cancer? Anticancer Agents Med Chem 2010;10:64–9. [59] Bergum C, Zoratti G, Boerner J, List K. Strong expression association between matriptase and its substrate prostasin in breast cancer. J Cell Physiol 2012;227:1604–9. [60] Wu SR, Cheng TS, Chen WC, Shyu HY, Ko CJ, Huang HP, et al. Matriptase is involved in ErbB-2-induced prostate cancer cell invasion. Am J Pathol 2010;177:3145–58. [61] Gabril M, White NM, Moussa M, Chow TF, Metias SM, Fatoohi E, et al. Immunohistochemical analysis of kallikrein-related peptidases in the normal kidney and renal tumors: potential clinical implications. Biol Chem 2010;391:403–9.

58

JO U R N A L OF PR O TE O MI CS 98 ( 20 1 4 ) 4 4 –58

[62] Douglas ML, Richardson MM, Nicol DL. Endothelin axis expression is markedly different in the two main subtypes of renal cell carcinoma. Cancer 2004;100:2118–24. [63] Pflug BR, Zheng H, Udan MS, D'Antonio JM, Marshall FF, Brooks JD, et al. Endothelin-1 promotes cell survival in renal cell carcinoma through the ET(A) receptor. Cancer Lett 2007;246:139–48. [64] Weissinger EM, Metzger J, Dobbelstein C, Wolff D, Schleuning M, Kuzmina Z, et al. Proteomic peptide profiling for preemptive diagnosis of acute graft-versus-host disease after allogeneic stem cell transplantation. Leukemia 2013 [ahead of print]. [65] Good DM, Zurbig P, Argiles A, Bauer HW, Behrens G, Coon JJ, et al. Naturally occurring human urinary peptides for use in

diagnosis of chronic kidney disease. Mol Cell Proteomics 2010;9:2424–37. [66] Zurbig P, Jerums G, Hovind P, Macisaac RJ, Mischak H, Nielsen SE, et al. Urinary proteomics for early diagnosis in diabetic nephropathy. Diabetes 2012;61:3304–13. [67] Klein J, Lacroix C, Caubet C, Siwy J, Zurbig P, Dakna M, et al. Fetal urinary peptides to predict postnatal outcome of renal disease in fetuses with posterior urethral valves (PUV). Sci Transl Med 2013;5:198ra06. [68] Small DM, Burden RE, Jaworski J, Hegarty SM, Spence S, Burrows JF, et al. Cathepsin S from both tumor and tumor-associated cells promote cancer growth and neovascularization. Int J Cancer 2013;133:2102–12.