Vous êtes sur la page 1sur 10

Serum Amyloid A As a Prognostic Marker in Melanoma

Identied by Proteomic Proling


Peter Findeisen, Marc Zapatka, Teresa Peccerella, Heike Matzk, Michael Neumaier, Dirk Schadendorf,
and Selma Ugurel
From the Institute for Clinical Chemis-
try, Medical Faculty Mannheim of the
University of Heidelberg; Department of
Theoretical Bioinformatics, German
Cancer Research Center; Skin Cancer
Unit, German Cancer Research Center
Heidelberg, Heidelberg; and the Depart-
ment of Dermatology, University Hospi-
tal of Mannheim, Mannheim, Germany.
Submitted May 14, 2008; accepted
December 2, 2008; published online
ahead of print at www.jco.org on
March 23, 2009.
Supported by a grant from the German
Cancer Aid/Deutsche Krebshilfe
(proposal No. 106856) and by the
German Federal Ministry of Research
and Education through research grant
No. 01 GR 0450 in the National
Genome Research Network.
*P.F. and M.Z. contributed equally to
this work.
Terms in blue are dened in the glos-
sary, found at the end of this article
and online at www.jco.org.
Authors disclosures of potential con-
icts of interest and author contribu-
tions are found at the end of this
article.
Corresponding author: Selma Ugurel,
MD, Department of Dermatology,
Julius-Maximilians-University, Josef-
Schneider-Strasse 2, 97080 Wuerzburg,
Germany; e-mail: ugurel_s@klinik
.uni-wuerzburg.de.
The Appendix is included in the
full-text version of this article,
available online at www.jco.org.
It is not included in the PDF version
(via Adobe Reader).
2009 by American Society of Clinical
Oncology
0732-183X/09/2713-2199/$20.00
DOI: 10.1200/JCO.2008.18.0554
A B S T R A C T
Purpose
Currently known prognostic serum biomarkers of melanoma are powerful in metastatic disease,
but weak in early-stage patients. This study was aimed to identify new prognostic biomarkers of
melanoma by serum mass spectrometry (MS) proteomic proling, and to validate candidates
compared with established markers.
Patients and Methods
Two independent sets of serum samples from 596 melanoma patients were investigated. The rst
set (stage I 102; stage IV 95) was analyzed by matrix assisted laser desorption and ionization
time of ight (MALDI TOF) MS for biomarkers differentiating between stage I and IV. In the second
set (stage I 98; stage II 91; stage III 87; stage IV 103), the serum concentrations of the
candidate marker serum amyloid A (SAA) and the known biomarkers S100B, lactate dehydroge-
nase, and C reactive protein (CRP) were measured using immunoassays.
Results
MALDI TOF MS revealed a peak at m/z 11.680 differentiating between stage I and IV, which could
be identied as SAA. High peak intensities at m/z 11.680 correlated with poor survival. In univariate
analysis, SAA was a strong prognostic marker in stage I to III (P .043) and stage IV (P .000083)
patients. Combination of SAA and CRP increased the prognostic impact to P .011 in early-stage
(I to III) patients. Multivariate analysis revealed sex, stage, tumor load, S100B, SAA, and CRP as
independent prognostic factors, with an interaction between SAA and CRP. In stage I to III
patients, SAA combined with CRP was superior to S100B in predicting patients progression-free
and overall survival.
Conclusion
SAA combined with CRP might be used as prognostic serological biomarkers in early-stage melanoma
patients, helping to discriminate low-risk patients from high-risk patients needing adjuvant treatment.
J Clin Oncol 27:2199-2208. 2009 by American Society of Clinical Oncology
INTRODUCTION
Melanoma is a cutaneous neoplasm known for its
high aggressiveness, its early dissemination of me-
tastases, and its poor prognosis once metastasized.
The early prediction of the risk of metastasis there-
fore is one of the major issues in the clinical care of
melanoma patients. Numerous serological biomar-
kers do exist, which help to determine the prognosis
of patients with distant metastases (stage IV follow-
ing the criteria of the AmericanJoint Committee on
Cancer [AJCC]
1
); the ones withthe widest distribu-
tioninclinical use are S100Band lactate dehydroge-
nase (LDH).
2
However, most of these serummarker
proteins fail to predict the risk of disease progres-
sion, relapse, and metastasis in early-stage mela-
noma patients (AJCC stage I to III), showing only
low or even no tumor burden at the time of clinical
presentation. These patients urgently need simple
andvalidserological tools topredict prognosis, since
up to now the only biomarkers with a prognostic
impact in stage I to III are histomorphological fea-
tures of the primary tumor, like ulceration and
depthof invasion, as well as aninvasivestagingof the
regional lymph node basin done by sentinel lymph
node dissection.
3
Since currently known serum biomarkers are
of little use in the prediction of prognosis of early-
stage melanoma patients, we focused on the identi-
cation of new serum marker proteins, which have
not been investigated in melanoma patients so far.
For this purpose we recently screened the whole
serumproteome of stage I (low-riskprimarytumor)
and stage IV (distant metastases) melanoma pa-
tients for differentially expressed proteins using a
high-throughput methodology, surface-enhanced
JOURNAL OF CLINICAL ONCOLOGY
O R I G I N A L R E P O R T
VOLUME 27 NUMBER 13 MAY 1 2009
2009 by American Society of Clinical Oncology 2199
2010 from 200.18.33.229
Information downloaded from jco.ascopubs.org and provided by at Universidade Federal da Santa Maria on October 17,
Copyright 2009 American Society of Clinical Oncology. All rights reserved.
laser desorption and ionization (SELDI) time of ight (TOF) mass
spectrometry (MS) combined with surface-functionalized protein
chips (H4 chip; Ciphergen Biosystems, Fremont, CA).
4
In the result-
ing spectra, we found a peak at m/z 11.700 as a highly signicant
discriminator between stage I and stage IV. This study was aimed at
the identication of the protein corresponding to this peak, and its
validation as a potential prognostic marker in comparison to known
serumbiomarkers of melanoma. For this purpose, we rst conrmed
the previously described discriminatory peak in serum specimens of
stage I and stage IVpatients different fromthose used in our previous
study
4
with matrix-assisted laser desorption and ionization (MALDI)
TOF MS combined with surface-functionalized C18 magnetic beads.
We thereafter identied the peak as serum amyloid A (SAA), and
validated its prognostic impact in comparison to S100B, LDH, and C
reactive protein (CRP) in an independent serum set comprising all
clinical stages of melanoma.
PATIENTS AND METHODS
Patients and Sera
Serumsamples were selected froma deep-frozen serumbank hosted by
the Skin Cancer Unit. All samples were obtained and processed following a
standardized protocol. Briey, venous blood was drawn into gel-coated
serum tubes (Sarstedt, Nuembrecht, Germany), clotted at room tempera-
ture for 30 to 60 minutes, and thereafter centrifuged at 2,500 g for 10
S
e
r
a

S
t
a
g
e

I
V

(
n

=

8
8
)
S
e
r
a

S
t
a
g
e

I

(
n

=

8
9
)
11,680
11,680
m/z
176
174
172
170
168
166
164
162
160
157
155
98
96
93
91
89
87
85
83
81
79
76
74
72
70
68
65
63
61
59
38
31
29
24
22
19
17
15
13
9
9
5
3
1
159
152
150
148
146
144
142
140
138
136
134
132
130
128
126
124
122
120
118
116
114
112
110
108
106
104
102
100
95
67
57
55
53
51
49
47
45
43
41
37
35
33
27
20
10
4,000 6,000 8,000 10,000 12,000 14,000
Fig 1. Pseudogelview of 177 high-quality mass spectrometry spectra from
matrix assisted laser desorption and ionization time of ight (MALDI-TOF) proling of
serum set 1. The arrows indicate the discriminative mass at m/z 11.680.
Table 1. Patient Demographics and Clinical Characteristics
Characteristic
Serum Set 1
Marker
Identication
Serum Set 2
Marker
Evaluation
No. % No. %
No. of patients 197 100.0 379 100.0
Sex
Male 97 49.2 208 54.9
Female 100 50.8 171 45.1
Age at diagnosis, year 55.6 54.7
Range 14.9-90.6 6.8-87.8
Localization of primary
Skin 157 79.7 321 84.7
Mucosa 2 1.0 15 4.0
Uvea 7 3.6 11 2.9
Occult 11 5.6 13 3.4
Unknown 20 10.1 19 5.0
Histologic type of primary
Supercial spreading 79 40.1 132 34.8
Nodular 25 12.7 108 28.5
Acrolentiginous 10 5.1 25 6.6
Lentigo maligna 8 4.1 27 7.1
Amelanotic 3 1.5 9 2.4
Uvea 7 3.6 11 2.9
Other 2 1.0 8 2.1
Not classiable 4 2.0 5 1.3
Occult 11 5.6 13 3.4
NA 48 24.3 41 10.9
Tumor staging

Stage I 102 51.8 98 25.9


Tumor 5 2.5 2 0.5
Tumor 97 49.2 96 25.3
IA 74 37.6 74 19.5
IB 28 14.2 24 6.4
Stage II 91 24.0
Tumor 6 1.6
Tumor 85 22.4
IIA 57 15.0
IIB 32 8.5
IIC 2 0.5
Stage III 87 22.9
Tumor 12 3.2
Tumor 75 19.8
IIIA 44 11.6
IIIB 14 3.7
IIIC 29 7.7
Stage IV 95 48.2 103 27.2
Tumor 73 37.1 79 20.8
Tumor 22 11.1 24 6.3
M1a 20 10.1 21 5.6
M1b 22 11.2 27 7.1
M1c 53 26.9 55 14.5
NOTE. Patient characteristics at the time point of blood withdrawal for
serum analysis.
Abbreviations: NA, not available; tumor, patient with detectable tumor load;
tumor, patient tumor free.

Tumor staging was performed according to the staging system of the


American Joint Committee on Cancer.
1
Findeisen et al
2200 2009 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY
2010 from 200.18.33.229
Information downloaded from jco.ascopubs.org and provided by at Universidade Federal da Santa Maria on October 17,
Copyright 2009 American Society of Clinical Oncology. All rights reserved.
minutes. Serum was harvested and immediately frozen at 20C. Thereaf-
ter, all samples underwent one additional freeze-thaw-cycle before the nal
thawing for analysis.
For the current serummarker study, our aimwas to select sera reecting
the composition of patients routinely seen in the clinical setting (ie, a high
proportion of tumor-free patients in stages I to III; patients after surgical
procedure being in regular follow-up), as well as a high proportion of tumor
positive patients in stage IV. Selection criteria were histologically conrmed
melanoma; complete documentationof medical history, primary tumor char-
acteristics, course of the disease and follow-up; and no systemic treatment for
at least 6 weeks before blood withdrawal (to minimize confounding serum
factors). The collection of sera and documentation of clinical data were per-
formed after patients informed consent with institutional review board ap-
proval. Disease staging was done according to the criteria of the AJCC.
1
Twoindependent sets of serumsamples were usedinthis study. The rst
set (set 1; marker identication set) was selected in order to conrm the
previously described
4
mass peak at m/z 11.700 discriminating between
early-stage, low-risk (stage I), and advanced-stage, distant-metastasis
(stage IV) patients. The second set (set 2; marker evaluation set) was
selected for the evaluation of the prognostic value of candidate marker
proteins identied fromset 1 in comparison to known serological biomar-
kers of melanoma, S100B, LDH, and CRP, over the whole range of disease
stages (stage I to IV). Detailed patient characteristics of both sets of sera are
presented in Table 1.
Serum Proteomic Proling
Sera were processed using magnetic beads functionalized with a C18
hydrophobic interaction chromatography surface (MB-HIC18; Bruker Dal-
tonics, Leipzig, Germany). The C18 surface was chosen after a previous com-
parativetestingof C8, C18, andWCXsurfaces inasmaller sampleset, revealing
the C18 surface with the highest number and quality of peaks. Briey, 5 L
serumwere mixedwith5Lmagnetic beadsuspensioninastandardthin-wall
polymerase chain reaction tube (Biozym Scientic, Oldendorf, Germany).
The tubes were placed in a magnetic bead separator (Bruker Daltonics), and
the supernatant with unbound proteins was removed in three washing proce-
dures. Bound proteins were eluted using 10 L of 50% aqueous acetonitrile
(v/v). The eluate was diluted 1:10 in alpha-cyano-4-hydroxycinnamic acid
(HCCA), and 1 L of the resulting mixture was spotted in duplicates onto a
SCOUT 600 m prestructured sample support (AnchorChip target;
Bruker Daltonics).
MALDI TOFMSwas performedusinganAutoexII (Bruker Daltonics)
equipped with a SCOUT ion source, operating in positive linear mode. Ions
formed by a pulsed UV laser beam (nitrogen laser, 337 nm, 50 Hz) were
accelerated to 20 kV. Further instrumental parameters were ion source 2
potential, 18.6 kV; focusing lens voltage, 7.8 kV; pulsed ion extraction, 210 ns.
Matrix solution was prepared by dissolving 0.3 mg/mL of HCCA in etha-
nol/acetone (2:1, v/v). External mass calibration was performed using a
mixture containing angiotensin II, bombesin, insulin, and ubiquitin I
(Bruker Daltonics). MS spectra for peaks in the range of 1 to 15 kDa were
generated by summarizing 600 laser shots (50 shots each at 12 different spot
positions) at a xed laser power of 43%. In order to increase detection sensi-
tivity, excess matrix was removed with 10 shots at a laser power of 70%before
data acquisition.
RESULTS
Mass Peak at m/z 11.680 Discriminates Between
Clinical Stages
Serumsamples of 102 stage I and 95 stage IVmelanoma patients
(set 1, marker identication set) were analyzed using hydrophobic
C18 surfaced magnetic beads and subsequent MALDI TOF MS. Se-
rum samples from 13 stage I and 7 stage IV patients resulted in low
quality spectra and thus were omitted from further analysis. The
remaining proteomic spectra were randomly assigned 10 10 train-
ing and test sets. Algorithms based on support vector machines
were used to discriminate spectra of stage I from those of stage IV
patients. An m/z of 11.680 was detected as the most informative
peak (Fig 1), conrming our previous results obtained in a differ-
ent set of serum samples.
4
Discriminatory Peak at m/z 11.680 Has
Prognostic Impact
In 177 high-quality MS spectra obtained from 89 stage I and 88
stage IV sera of serum set 1, the m/z 11.680 peak was quantied by
peak intensity (area under the curve), and the obtained values were
correlated with the patients overall survival starting from the time
point of blood withdrawal. The most relevant binary split for peak
intensity was determined by successive testing using an appropriate
A
0
S
u
r
v
i
v
a
l

P
r
o
b
a
b
i
l
i
t
y
Time (months)
1.0
0.8
0.6
0.4
0.2
10 20 30 40 50 60
B
0
S
u
r
v
i
v
a
l

P
r
o
b
a
b
i
l
i
t
y
Time (months)
1.0
0.8
0.6
0.4
0.2
10 20 30 40 50 60
Stages I + IV
P < .0000005
Stage IV
P = .00041
Fig 2. Prognostic impact of discriminatory peak at m/z 11.680. Kaplan-Meier survival curves of melanoma patients of serum set 1, grouped by the area under the curve
of the serum peak at m/z 11.680 using a cut point of 33.8. (A) Patients in stage I and IV, respectively, n 177; yellow line, peak intensity 33.8, n 135; blue line,
peak intensity more than 33.8, n 42; (B) Patients in stage IV, n 88; yellow line, peak intensity 33.8, n 61; blue line, peak intensity more than 33.8, n 27;
statistical differences calculated by log-rank test.
Serum Amyloid A As a Prognostic Marker in Melanoma
www.jco.org 2009 by American Society of Clinical Oncology 2201
2010 from 200.18.33.229
Information downloaded from jco.ascopubs.org and provided by at Universidade Federal da Santa Maria on October 17,
Copyright 2009 American Society of Clinical Oncology. All rights reserved.
two sample test statistic to measure the separation of the two sub-
groups. Toaccount for multiple testing the Pvalue was adjustedbased
on Hothorn and Lausen.
7
A cutoff value of 33.8 was optimal for the
separation of stage I and stage IV patients. Patients presenting high
levels of the m/z 11.680 peak showed a poorer survival compared to
patients with lower values of this protein mass; this observation was
made in the whole patient population comprising stage I and stage IV
(n 177; Fig 2A; P .0000005, as well as in stage IV patients alone
(n 88; Fig 2B; P .00041).
Identication of m/z 11.680 As SAA
Of the tryptic digest of the C18 magnetic bead eluates from ve
pooledspecimens withhighsignal intensities for m/z 11.680, we could
identify three peptides as fragments of serum amyloid A (SAA) by
subsequent MS/MS analysis. These peptides were namely m/z 1550.7;
m/z 1640.8; and m/z 2178.3, which are covering amino acid positions
20to33, 86to105, and109to122, respectively, of the SAAaminoacid
sequence (P02735). These ndings are in line with previous identi-
cations of m/z 11.680 as SAA.
9,10
The signal intensities of the m/z
11.680 peak correlated with the SAA protein concentration deter-
mined by immunoassay in corresponding serum samples (data
not shown).
SAA As a Prognostic Marker
To conrm the prognostic impact of SAA in serum from mela-
noma patients, we measured the concentrationof SAAinserumset 2,
which was chosen from patients completely different from serum set
1, using an immunonephelometric assay. Serumset 2 was chosen as a
broad sample set comprising sera fromall clinical stages of melanoma
including 98 stage I, 91 stage II, 87 stage III, and 103 stage IVpatients.
Using univariate Cox hazard analysis with a cut point of 10 mg/L, we
found patients with low serum levels of SAA showing a favorable
A
0
S
u
r
v
i
v
a
l

P
r
o
b
a
b
i
l
i
t
y

(
S
t
a
g
e

I
-
I
V
)
Time Since Blood Withdrawal (months)
1.0
0.8
0.6
0.4
0.2
20 40 60 80 0
S
u
r
v
i
v
a
l

P
r
o
b
a
b
i
l
i
t
y

(
S
t
a
g
e

I
-
I
V
)
Time Since Blood Withdrawal (months)
1.0
0.8
0.6
0.4
0.2
20 40 60 80
B
C
0
S
u
r
v
i
v
a
l

P
r
o
b
a
b
i
l
i
t
y

(
S
t
a
g
e

I
-
I
V
)
Time Since Blood Withdrawal (months)
1.0
0.8
0.6
0.4
0.2
20 40 60 80 0
S
u
r
v
i
v
a
l

P
r
o
b
a
b
i
l
i
t
y

(
S
t
a
g
e

I
-
I
V
)
Time Since Blood Withdrawal (months)
1.0
0.8
0.6
0.4
0.2
20 40 60 80
D
P = .00067
+
+
SAA 10 mg/L; n = 176
SAA > 10 mg/L; n = 203
P < .0000005
+
+
CRP 10 mg/L; n = 318
CRP > 10 mg/L; n = 61
P < .0000005
+
+
S100 0.12 g/L; n = 302
S100 > 0.12 g/L; n = 77
P < .0000005
+
+
LDH 248 IU/L; n = 366
LDH > 248 IU/L; n = 13
Fig 3. Prognostic impact of serum marker proteins in 379 stage I-IV melanoma patients. Kaplan-Meier survival curves of patients of serum set 2, grouped by serum
concentrations of marker proteins: (A) serum amyloid A (SAA); (B) C reactive protein (CRP); (C) S100B; (D) lactate dehydrogenase (LDH). Statistical differences
calculated by log-rank test.
Findeisen et al
2202 2009 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY
2010 from 200.18.33.229
Information downloaded from jco.ascopubs.org and provided by at Universidade Federal da Santa Maria on October 17,
Copyright 2009 American Society of Clinical Oncology. All rights reserved.
A
0
S
u
r
v
i
v
a
l

P
r
o
b
a
b
i
l
i
t
y

(
S
t
a
g
e

I
-
I
I
I
)
Time Since Blood Withdrawal (months)
1.0
0.8
0.6
0.4
0.2
40 20 60 80
P = .043
+
+
SAA 10 mg/L; n = 132
SAA > 10 mg/L; n = 144
0
S
u
r
v
i
v
a
l

P
r
o
b
a
b
i
l
i
t
y

(
S
t
a
g
e

I
-
I
I
I
)
Time Since Blood Withdrawal (months)
1.0
0.8
0.6
0.4
0.2
40 20 60 80
P = .0055
+
+
CRP 10 mg/L; n = 254
CRP > 10 mg/L; n = 22
0
S
u
r
v
i
v
a
l

P
r
o
b
a
b
i
l
i
t
y

(
S
t
a
g
e

I
-
I
I
I
)
Time Since Blood Withdrawal (months)
1.0
0.8
0.6
0.4
0.2
40 20 60 80
P = .034
+
+
S100B 0.12 g/L; n = 251
S100B > 0.12 g/L; n = 25
0
S
u
r
v
i
v
a
l

P
r
o
b
a
b
i
l
i
t
y

(
S
t
a
g
e

I
-
I
I
I
)
Time Since Blood Withdrawal (months)
1.0
0.8
0.6
0.4
0.2
40 20 60 80
P = .93
+
+
LDH 248 IU/L; n = 275
LDH > 248 IU/L; n = 1
B
0
S
u
r
v
i
v
a
l

P
r
o
b
a
b
i
l
i
t
y

(
S
t
a
g
e

I
V
)
Time Since Blood Withdrawal (months)
1.0
0.8
0.6
0.4
0.2
20 10 30 40
P = .000083
0
S
u
r
v
i
v
a
l

P
r
o
b
a
b
i
l
i
t
y

(
S
t
a
g
e

I
V
)
Time Since Blood Withdrawal (months)
1.0
0.8
0.6
0.4
0.2
20 10 30 40
P < .0000005
0
S
u
r
v
i
v
a
l

P
r
o
b
a
b
i
l
i
t
y

(
S
t
a
g
e

I
V
)
Time Since Blood Withdrawal (months)
1.0
0.8
0.6
0.4
0.2
20 10 30 40
P = .0010
0
S
u
r
v
i
v
a
l

P
r
o
b
a
b
i
l
i
t
y

(
S
t
a
g
e

I
V
)
Time Since Blood Withdrawal (months)
1.0
0.8
0.6
0.4
0.2
20 10 30 40
P < .0000005
+
+
SAA 10 mg/L; n = 44
SAA > 10 mg/L; n = 59
+
+
CRP 10 mg/L; n = 64
CRP > 10 mg/L; n = 39
+
+
S100B 0.12 g/L; n = 51
S100B > 0.12 g/L; n = 52
+
+
LDH 248 IU/L; n = 91
LDH > 248 IU/L; n = 12
Fig 4. Prognostic impact of serum
marker proteins in (A) 276 stage I-III mel-
anoma patients, and (B) 103 stage IV
melanoma patients. Kaplan-Meier sur-
vival curves of patients of serum set 2,
grouped by serum concentration of
marker proteins; statistical differences cal-
culated by log-rank test. SAA, serum amy-
loid A; CRP, C reactive protein; LDH,
lactate dehydrogenase.
Serum Amyloid A As a Prognostic Marker in Melanoma
www.jco.org 2009 by American Society of Clinical Oncology 2203
2010 from 200.18.33.229
Information downloaded from jco.ascopubs.org and provided by at Universidade Federal da Santa Maria on October 17,
Copyright 2009 American Society of Clinical Oncology. All rights reserved.
survival compared to patients with high serum levels of this protein
(n 379; P .00067, Fig 3). Analyzing patients separately in stage
I-III and stage IV, respectively, serum SAA appeared as a prognostic
marker inbothgroups (stage I-III, n276, P.043, Fig 4A; stage IV,
n103, P.000083, Fig4B). Measurement of SAAduringthecourse
of disease in sera from seven melanoma patients collected over time
showed disease progression associated with an increase in SAAin two
patients, who turned out to be therapy responders and long-term
survivors (Fig 5).
Comparison of SAA With Other Serum Markers
of Melanoma
To evaluate the prognostic impact of SAA compared to known
serum markers of melanoma, we additionally measured S100B and
LDHinall samples of serumset 2. Moreover, since SAAbelongs tothe
family of acute phase proteins, we additionally analyzed the sera for
CRP, the main representative of this protein family. The cutoff values
used for prognostic classications were S100B, 0.12 g/L; LDH, 248
U/L; SAA, 10 mg/L; and CRP, 10 mg/L. Utilizing univariate Cox
hazard analysis, S100B (P .0000005), LDH (P .0000005), CRP
(P .0000005), and SAA (P .0014) were signicant prognostic
indicators in the whole patient population comprising all disease
stages I to IV (n 379, Fig 3), as well as in the subgroup of stage IV
patients (n 103; LDH, P .0000005; CRP, P .0000005; SAA,
P.000083; S100B, P.0010; Fig4B), withhighlevels of serummarker
proteins associatedwitha reducedoverall survival. Looking at patients in
early disease stages (stage I toIII, n276), only CRP(P.0055), S100B
(P .034), and SAA (P .043), but not LDH (P .93) resulted as
prognostic markers (Fig 4A). Notably, SAAwas the only marker separat-
ingearlystagepatientsintotwoprognosticgroupsof comparablesize(144
v 132 patients), whereas the other three markers detected only small
numbers of patients inthe respective groupof poor prognosis (CRP, 22v
254; S100B, 25 v 251; LDH, 1 v 275; Fig 4A). Multivariate analysis using
Cox proportional hazards model including the well-known prognostic
markers sex, stage, and tumor load, and our markers of interest SAA,
S100B, CRP, and LDH using the previously mentioned cutoff values
revealed sex, stage, and tumor load, as well as S100B, CRP, and SAA as
prognosticmarkers instageI-IVpatients together withasignicant inter-
action between CRP and SAA. These ndings were conrmed by the
highlysignicant prognosticdiscriminationdetectedforthecombination
of these two markers inthe whole patient population(P.0000005; Fig
6A), aswell asinthesubgroupsofstageI-III(P.011; Fig6B)andstageIV
(P.0000005; Fig 6C) by Mantel-Haenszel test.
11
S
A
A

(
m
g
/
L
)
Course of Disease Over Time
0
10
20
30
40
50
60
70
80
90
100
110
AJCC I/II
(1)
AJCC I/II
(2)
AJCC III
(1)
AJCC III
(2)
AJCC III
(3)
AJCC IV
(1)
AJCC IV
(2)
AJCC IV
(3)
Pat 1
Pat 2
Pat 3
Pat 4
Pat 5
Pat 6
Pat 7
Fig 5. Serum serum amyloid A (SAA) measured over time in seven melanoma
patients. In patients (pat) 2, 3, 4, 6, and 7, a progression in disease stage is
associated with an increase in SAA serum concentration; these patients subse-
quently died from melanoma. Patients 1 and 5 progressed to stage IV paralleled
by a decrease in SAA; both patients showed a complete remission to rst-line
treatment and revealed long-term survival. The x-axis gives time intervals of
blood withdrawal in different disease stages. AJCC, American Joint Committee
on Cancer.
A
0
S
u
r
v
i
v
a
l

P
r
o
b
a
b
i
l
i
t
y

(
s
t
a
g
e
s

I
-
I
V
)
Time Since Blood Withdrawal (months)
1.0
0.8
0.6
0.4
0.2
20 40 60 80
B
0
S
u
r
v
i
v
a
l

P
r
o
b
a
b
i
l
i
t
y

(
s
t
a
g
e
s

I
-
I
I
I
)
Time Since Blood Withdrawal (months)
1.0
0.8
0.6
0.4
0.2
20 40 60 80
C
0
S
u
r
v
i
v
a
l

P
r
o
b
a
b
i
l
i
t
y

(
S
t
a
g
e

I
V
)
Time Since Blood Withdrawal (months)
1.0
0.8
0.6
0.4
0.2
10 20 30 40
CRP 10 mg/L + SAA 1 mg/L; n = 167
CRP 10 mg/L + SAA > 10 mg/L; n = 151
CRP > 10 mg/L + SAA 10 mg/L; n = 8
CRP > 10 mg/L + SAA > 10 mg/L; n = 53
Stages IIV
P < .0000005
CRP 10 mg/L + SAA 10 mg/L; n = 126
CRP 10 mg/L + SAA > 10 mg/L; n = 128
CRP > 10 mg/L + SAA 10 mg/L; n = 6
CRP > 10 mg/L + SAA > 10 mg/L; n = 16
Stages IIII
P = .011
CRP 10 mg/L + SAA 10 mg/L; n = 41
CRP 10 mg/L + SAA > 10 mg/L; n = 23
CRP > 10 mg/L + SAA 10 mg/L; n = 2
CRP > 10 mg/L + SAA > 10 mg/L; n = 37
Stage IV
P < .0000005
+
+
+
+
+
+
+
+
+
+
+
+
Fig 6. Prognostic impact of the combination of serum marker proteins C reactive
protein (CRP) and serumamyloid A(SAA) in (A) 279 stage I-IV melanoma patients, (B)
276 stage I-III melanoma patients, and (C) 103 stage IV melanoma patients.
Kaplan-Meier survival curves of patients of serum set 2, grouped by serum
concentration of marker proteins; statistical differences calculated by log-rank test.
Findeisen et al
2204 2009 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY
2010 from 200.18.33.229
Information downloaded from jco.ascopubs.org and provided by at Universidade Federal da Santa Maria on October 17,
Copyright 2009 American Society of Clinical Oncology. All rights reserved.
Superiority of CRP and SAA Compared to S100B As
Prognostic Markers in Early-Stage Melanoma Patients
Conditional inference trees analyzing all four serum markers re-
vealedCRP(P.001), S100B(P.001), andSAA(P.001)splittingup
patients into signicantly different prognostic groups in 202 patients of
stageIandIV(Fig7Aand7B), whereasCRP(P.001), S100B(P.001),
SAA(P.001), and LDH(P.012) were found as relevant markers in
394 patients of stage I, II, III, and IV (Fig 7C). Receiver operating
characteristic (ROC) curve analysis was used to compare the sen-
sitivity andspecicity of the panel of four serummarker proteins to
predict the prognosis of early-stage melanoma patients. Serum
data of 276 melanoma patients in stages I to III were used to dene
A
C
B
1
18.8
2
3
>
Node 4 (n = 238)
II I III IV
5
Node 6 (n = 75) Node 7 (n = 10) Node 8 (n = 31)
0
0.2
0.4
0.6
0.8
1.0
Node 9 (n = 40)
II I III IV
1
2
> 0.107
Node 3 (n = 123)
0
0.2
0.4
0.6
0.8
1.0
0
0.2
0.4
0.6
0.8
1.0
Node 4 (n = 34) Node 5 (n = 45)
SAA (mg/L)
P < .001
1
Node 2 (n = 168)
0
0.2
0.4
0.6
0.8
1.0
0
0.2
0.4
0.6
0.8
1.0
Node 3 (n = 34)
AJCC I
AJCC IV
0
0.2
0.4
0.6
0.8
1.0
AJCC I
AJCC IV
CRP (mg/L)
P < .001
CRP (mg/L)
P < .001
SAA (mg/L)
P < .001
S100
P < .001
S100
P < .001
LDH (IU/L)
P = .012
II I III IV II I III IV II I III IV
0
0.2
0.4
0.6
0.8
1.0
0
0.2
0.4
0.6
0.8
1.0
0
0.2
0.4
0.6
0.8
1.0
0
0.2
0.4
0.6
0.8
1.0
0.107
> 9.6 9.6
> 56 56
> 28.8 28.8
> 122 122
0.19 > 0.19
> 18.8 18.8 18.8
Fig 7. Conditional inference trees for the serum marker panel consisting of C reactive protein (CRP), serum amyloid A (SAA), S100B, and lactate dehydrogenase (LDH).
For each marker protein, the Bonferroni adjusted P values as well as the optimal cutoff values calculated for each model are given. (A, B) CRP, S100B, and SAA splitting
up patients into signicantly different prognostic groups in 202 patients of stage I and IV. The ratio of patients in stage I and IV, respectively, is displayed as bars for
each terminal node. (C) CRP, S100B, SAA, and LDH splitting up patients into signicantly different prognostic groups in 394 patients of stage I, II, III, and IV. The ratio
of patients in stage I to IV, respectively, is displayed as bars for each terminal node. AJCC, American Joint Committee on Cancer.
Serum Amyloid A As a Prognostic Marker in Melanoma
www.jco.org 2009 by American Society of Clinical Oncology 2205
2010 from 200.18.33.229
Information downloaded from jco.ascopubs.org and provided by at Universidade Federal da Santa Maria on October 17,
Copyright 2009 American Society of Clinical Oncology. All rights reserved.
ROC curves showing that S100B revealed low AUC values (0.54
and 0.58, respectively) to predict a patient to be progression free at
18 months after blood withdrawal and to be alive at 36 months
after blood withdrawal, respectively (Figs 8A and 8B). SAA re-
vealedbetter AUCvalues of 0.58 and0.61, respectively, whereas the
combination of SAA and CRP showed the highest AUC values of
0.60 and 0.63, respectively.
DISCUSSION
MS-based proteomic proling is heralded as a new tool for diagnosis
and prognosis of malignant diseases, and can easily be applied to a
variety of clinical specimens.
12-14
Serum in particular is a difcult
target because of preanalytical variabilities in sample handling
0
T
r
u
e

P
o
s
i
t
i
v
e

R
a
t
e

(
s
e
n
s
i
t
i
v
i
t
y
)
False Positive Rate (1Specificity)
1.0
0.8
0.6
0.4
0.2
0.2 0.4 0.6 0.8 1.0
A
0
T
r
u
e

P
o
s
i
t
i
v
e

R
a
t
e

(
s
e
n
s
i
t
i
v
i
t
y
)
False Positive Rate (1Specificity)
1.0
0.8
0.6
0.4
0.2
0.2 0.4 0.6 0.8 1.0
B
0
T
r
u
e

P
o
s
i
t
i
v
e

R
a
t
e

(
s
e
n
s
i
t
i
v
i
t
y
)
False Positive Rate (1Specificity)
1.0
0.8
0.6
0.4
0.2
0.2 0.4 0.6 0.8 1.0
C
0
T
r
u
e

P
o
s
i
t
i
v
e

R
a
t
e

(
s
e
n
s
i
t
i
v
i
t
y
)
False Positive Rate (1Specificity)
1.0
0.8
0.6
0.4
0.2
0.2 0.4 0.6 0.8 1.0
D
0
T
r
u
e

P
o
s
i
t
i
v
e

R
a
t
e

(
s
e
n
s
i
t
i
v
i
t
y
)
False Positive Rate (1Specificity)
1.0
0.8
0.6
0.4
0.2
0.2 0.4 0.6 0.8 1.0
E
0
T
r
u
e

P
o
s
i
t
i
v
e

R
a
t
e

(
s
e
n
s
i
t
i
v
i
t
y
)
False Positive Rate (1Specificity)
1.0
0.8
0.6
0.4
0.2
0.2 0.4 0.6 0.8 1.0
F
AUC = 0.541
S100B
AUC = 0.579
SAA
AUC = 0.602
SAA + CRP
AUC = 0.583
S100B
AUC = 0.612
SAA
AUC = 0.627
SAA + CRP
Fig 8. Receiver operating characteristic curves for the serum marker proteins S100B, serum amyloid A (SAA), and SAA plus C reactive protein (CRP) in 276 early-stage
melanoma patients (stages I-III). (A), (C), (E) Sensitivity and specicity to dene a patient to be progression-free 18 months after blood withdrawal for serum analysis.
(B), (D), (F) Sensitivity and specicity to dene a patient to be alive 36 months after blood withdrawal for serum analysis. Area under the curve (AUC) values are given
in the right lower corner of each plot.
Findeisen et al
2206 2009 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY
2010 from 200.18.33.229
Information downloaded from jco.ascopubs.org and provided by at Universidade Federal da Santa Maria on October 17,
Copyright 2009 American Society of Clinical Oncology. All rights reserved.
and processing that cause substantial changes of MS pro-
les.
15,16
Nevertheless, MS-based serum analysis has great po-
tential to screen for valuable biomarkers which, once identied,
can be quantied by immunoassays.
5,17,18
This study was intended to identify new prognostic biomarkers
of melanoma using proteomic proling. We could conrm our pre-
vious nding of an MS peak at m/z 11.700 (SELDI TOF MS
4
) by
detecting a peak at m/z 11.680 (MALDI TOF MS) differentiating
between stage I and stage IVpatients. Thus, SELDI and MALDI TOF
MS revealed comparable results using different sets of serumsamples,
demonstrating the robustness of these high-throughput proteomic
proling techniques. We could subsequently identify the m/z 11.680
peak as SAA, an acute-phase protein reecting the activity of inam-
matory diseases.
19
Besides, SAA has been reported as a marker of
disease progression in different malignancies including renal cell car-
cinoma,
20
lung,
21
colorectal, pancreatic, and breast cancer,
22
but has
not been associated with melanoma so far.
SAA belongs to the group of high-abundance serum pro-
teins that are predominantly detected by MS-based serum pro-
teomic proling.
23
Another acute-phase serum protein, CRP,
has already been reported as a prognostic marker in melano-
ma.
24,25
Both proteins, CRP and SAA, are synthesized by the liver
following pro-inammatory stimuli, and have been reported to be
comparably upregulated in serumof renal cell cancer patients after
treatment with interleukin (IL)-2.
26
Despite these similarities, the
sensitivity of SAA to detect inammatory changes has been re-
ported to be higher than that of CRP.
19,27
To validate the newly
identied serological marker SAA in melanoma, we therefore de-
cided to not only measure the widely used biomarkers S100B and
LDH, but also CRP in a large set of serum samples including all
stages of the disease. About the signicance of elevated SAA levels
on clinical course of melanoma, we hypothesize that both proteins,
SAA and CRP, are released due to a pro-inammatory state in
melanoma patients who are at a high risk of disease progression.
This hypothesis is strengthened by recent observations of elevated
levels of pro-inammatory cytokines in melanoma patients asso-
ciated with disease progression and poor prognosis. However,
many cytokines are hard to measure because of stability problems
and cost, thus rendering these serum factors not suitable for the
day-by-day clinical routine. In contrast, SAA and CRP are stable,
easy-to-measure serum parameters included in the routine serum
chemistry panel of most clinical laboratories. These markers, if
proving reliability and validity to predict patients prognosis,
would therefore be of high value for the clinical management of
melanoma patients.
We found all four markers tested, S100B, LDH, CRP, and SAA, as
prognostic markers inpatients of all disease stages (I toIV) using univar-
iate data analysis. S100B, LDH, and CRPall showing P.0000005 were
superior to SAA with P .0014. Looking at patients in early disease
stages (stage I to III), only CRP, S100B, and SAA, but not LDH
showed prognostic signicance. Here, SAA was the only marker
separating early-stage patients into two prognostic groups of com-
parable size, indicating a poor prognosis for 52% of patients,
whereas the other three markers classied only approximately 10%
(CRP, S100B) or less (LDH) patients as of poor prognosis. Multi-
variate analysis showed S100B, CRP, and SAA as independent
prognostic markers in melanoma, together with a signicant inter-
action between CRP and SAA. LDH did not show independent
prognostic signicance testedtogether withits covariates, conrm-
ing the results of an earlier study describing the superiority of CRP
compared to LDH in predicting distant metastasis in stage I to III
melanoma patients.
25
The signicant interaction found between CRP and SAA by
multivariate analysis deserves particular notication, as far as we ob-
servedasignicant decrease inthe probabilityof survival startingfrom
patients showing normal values of both markers, followed by patients
showing elevated levels of SAA, then followed by patients showing
elevated levels of CRP, and nally ending with the worst prognosis in
patients showing elevated levels of both SAA and CRP. This observa-
tion could be consistently made in all patients, stage I to IV, as well as
inthe subgroups of stage I to III andstage IV. These results are further
strengthened by our nding of a superiority of CRP and SAA com-
pared to S100B as prognostic markers in stage I to III melanoma
patients, as detected using ROCcurve analysis. Herein we found SAA
combined with CRP showing higher sensitivity and specicity to pre-
dict progression-free as well as overall survival than the current stan-
dard serum marker S100B. These results indicate that SAA and CRP
combined could serve as useful prognostic serumbiomarkers in mel-
anoma, particularly in early-stage patients. In this patient group these
new markers could help to identify patients of high risk of disease
progression in strong need of adjuvant treatment.
AUTHORS DISCLOSURES OF POTENTIAL CONFLICTS
OF INTEREST
The author(s) indicated no potential conicts of interest.
AUTHOR CONTRIBUTIONS
Conception and design: Peter Findeisen, Marc Zapatka, Michael
Neumaier, Dirk Schadendorf, Selma Ugurel
Administrative support: Michael Neumaier, Dirk Schadendorf
Provision of study materials or patients: Dirk Schadendorf,
Selma Ugurel
Collection and assembly of data: Peter Findeisen, Teresa Peccerella,
Heike Matzk, Selma Ugurel
Data analysis and interpretation: Peter Findeisen, Marc Zapatka,
Michael Neumaier, Dirk Schadendorf, Selma Ugurel
Manuscript writing: Peter Findeisen, Marc Zapatka, Selma Ugurel
Final approval of manuscript: Peter Findeisen, Marc Zapatka, Teresa
Peccerella, Heike Matzk, Michael Neumaier, Dirk Schadendorf,
Selma Ugurel
REFERENCES
1. Balch CM, Buzaid AC, Soong SJ, et al: Final
version of the American Joint Committee on Cancer
staging system for cutaneous melanoma. J Clin
Oncol 19:3635-3648, 2001
2. Utikal J, Schadendorf D, Ugurel S: Serologic
and immunohistochemical prognostic biomarkers of
cutaneous malignancies. Arch Dermatol Res 298:
469-477, 2007
3. Balch CM, Soong SJ, Gershenwald JE, et al:
Prognostic factors analysis of 17,600 melanoma
patients: Validation of the American Joint Commit-
tee on Cancer melanoma staging system. J Clin
Oncol 19:3622-3634, 2001
4. Mian S, Ugurel S, Parkinson E, et al: Serum
ngerprinting discriminates between clinical
stages and predicts disease progression in me-
lanoma patients. J Clin Oncol 23:5088-5093,
2005
Serum Amyloid A As a Prognostic Marker in Melanoma
www.jco.org 2009 by American Society of Clinical Oncology 2207
2010 from 200.18.33.229
Information downloaded from jco.ascopubs.org and provided by at Universidade Federal da Santa Maria on October 17,
Copyright 2009 American Society of Clinical Oncology. All rights reserved.
5. Zhang Z, Bast RC Jr, Yu Y, et al: Three
biomarkers identied from serum proteomic analy-
sis for the detection of early stage ovarian cancer.
Cancer Res 64:5882-5890, 2004
6. Reference deleted
7. Hothorn T, Lausen B: On the exact distribu-
tion of maximally selected rank statistics. Computa-
tional Statistics & Data Analysis 43:121-137, 2003
8. Reference deleted
9. Howard BA, Wang MZ, Campa MJ, et al:
Identication and validation of a potential lung can-
cer serum biomarker detected by matrix-assisted
laser desorption/ionization-time of ight spectra
analysis. Proteomics 3:1720-1724, 2003
10. Nedelkov D, Kiernan UA, Niederkoer EE, et
al: Investigating diversity in human plasma proteins.
Proc Natl Acad Sci U S A 102:10852-10857, 2005
11. Mantel N, Haenszel W: Statistical aspects of
the analysis of data from retrospective studies of
disease. J Natl Cancer Inst 22:719-748, 1959
12. Anderson NL, Anderson NG: The human
plasma proteome: History, character, and diagnostic
prospects. Mol Cell Proteomics 1:845-867, 2002
13. Eisenstein M: Protein arrays: Growing pains.
Nature 444:959-962, 2006
14. Petricoin EF, Belluco C, Araujo RP, et al: The
blood peptidome: A higher dimension of information
content for cancer biomarker discovery. Nat Rev
Cancer 6:961-967, 2006
15. Baumann S, Ceglarek U, Fiedler GM, et al:
Standardized approach to proteome proling of human
serumbased on magnetic bead separation and matrix-
assisted laser desorption/ionization time-of-ight mass
spectrometry. Clin Chem 51:973-980, 2005
16. Findeisen P, Sismanidis D, Riedl M, et al:
Preanalytical impact of sample handling on pro-
teome proling experiments with matrix-assisted
laser desorption/ionization time-of-ight mass spec-
trometry. Clin Chem 51:2409-2411, 2005
17. Patz EF Jr, Campa MJ, Gottlin EB, et al: Panel
of serum biomarkers for the diagnosis of lung can-
cer. J Clin Oncol 25:5578-5583, 2007
18. Jackson D, Craven RA, Hutson RC, et al:
Proteomic proling identies afamin as a potential
biomarker for ovarian cancer. Clin Cancer Res 13:
7370-7379, 2007
19. Yamada T: Serum amyloid A (SAA): A concise
review of biology, assay methods and clinical use-
fulness. Clin Chem Lab Med 37:381-388, 1999
20. Kimura M, Tomita Y, Imai T, et al: Signicance
of serum amyloid A on the prognosis in patients
with renal cell carcinoma. Cancer 92:2072-2075,
2001
21. Gao WM, Kuick R, Orchekowski RP, et al:
Distinctive serum protein proles involving abun-
dant proteins in lung cancer patients based upon
antibody microarray analysis. BMC Cancer 5:110,
2005
22. Biran H, Friedman N, Neumann L, et al: Serum
amyloid A (SAA) variations in patients with cancer:
Correlation with disease activity, stage, primary site,
and prognosis. J Clin Pathol 39:794-797, 1986
23. Hortin GL: The MALDI-TOF mass spectromet-
ric view of the plasma proteome and peptidome.
Clin Chem 52:1223-1237, 2006
24. Deichmann M, Benner A, Waldmann V, et al:
Interleukin-6 and its surrogate C-reactive protein are
useful serum markers for monitoring metastasized
malignant melanoma. J Exp Clin Cancer Res 19:301-
307, 2000
25. Deichmann M, Kahle B, Moser K, et al: Diag-
nosing melanoma patients entering American Joint
Committee on Cancer stage IV, C-reactive protein in
serum is superior to lactate dehydrogenase. Br J
Cancer 91:699-702, 2004
26. Rossi L, Martin BM, Hortin GL, et al: Inam-
matory protein prole during systemic high dose
interleukin-2 administration. Proteomics 6:709-720,
2006
27. Maury CP: Comparative study of serum amy-
loid A protein and C-reactive protein in disease. Clin
Sci (Lond) 68:233-238, 1985

Acknowledgment
This work was supported by a grant fromthe German Cancer Aid/Deutsche Krebshilfe (Proposal No. 106856) and by the German Federal
Ministry of Research and Education through a research grant in the National Genome Research Network (01 GR 0450). We wish to thank
Antje Sucker and Diamandula Sismanidis for excellent technical assistance, as well as Jurgen C. Becker for helpful discussions. The mass
spectrometry data can be accessed at https://services.ichip.de.
Glossary Terms
Prognostic (prognostic marker): A marker that predicts
the prognosis of a patient (eg, the likelihood of relapse, progres-
sion, and/or death) independent of future treatment effects. A
factor can be both prognostic and predictive.
Biomarker: A functional biochemical or molecular indicator
of a biologic or disease process that has predictive, diagnostic,
and/or prognostic utility.
Matrix-assisted laser desorption and ionization
(MALDI): Matrix-assisted laser desorption and ionization
(MALDI) is a soft ionization technique used in time-of ight
(TOF) mass spectrometry, allowing the analysis of proteins
and peptides, which fragment when ionized by a laser beam. A matrix is
used to protect the biomolecule from being destroyed by direct laser
beam and to facilitate vaporization and ionization.
Surface-enhanced laser desorption and ionization
(SELDI): Surface-enhanced laser desorption and ionization (SELDI)
is an ionization method in TOF mass spectrometry that is used for the
analysis of protein mixtures in tissue samples, blood, urine, or other
clinical samples. Comparison of protein levels between patients with
and without a disease can be used for biomarker discovery. Compared
to MALDI, SELDI uses a target modied to achieve biochemical afnity
with the analyte compound, e.g. protein chips.
Findeisen et al
2208 2009 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY
2010 from 200.18.33.229
Information downloaded from jco.ascopubs.org and provided by at Universidade Federal da Santa Maria on October 17,
Copyright 2009 American Society of Clinical Oncology. All rights reserved.