Purpose: To validate the prognostic impact of combined expression levels of three markers (SPP1, RGS1, and NCOA3) in melanoma specimens from patients enrolled in the E1690 clinical trial of high-dose or low-dose IFNα-2b versus observation.

Experimental Design: Tissue was available from 248 patients. Marker expression was determined by digital imaging of immunohistochemically stained slides. The prognostic impact of each marker was first assessed by recording its expression value relative to the median. A multimarker index was then developed to combine marker expression levels by counting for each patient the number of markers with high expression. The impact of the multimarker index on relapse-free survival (RFS) and overall survival (OS) was assessed using Kaplan–Meier analysis, and both univariate and multivariate Cox regression analyses.

Results: By Kaplan–Meier analysis, high multimarker expression scores were significantly predictive of RFS (P < 0.001) and OS (P < 0.001). Stepwise multivariate Cox regression analysis with backward elimination that included routine clinical and histologic prognostic factors revealed high multimarker expression scores and tumor thickness as the only factors significantly and independently predicting RFS and OS. Stepwise multivariate Cox regression analyses that also included treatment type and number of positive lymph nodes generated identical results for both RFS and OS. In the molecularly defined low-risk subgroup, patients treated with high-dose IFN had a significantly improved RFS compared with patients in the other two subgroups (P < 0.05).

Conclusions: These results validate the independent impact of combined expression levels of SPP1, RGS1, and NCOA3 on survival of melanoma in a prospectively collected cohort. Clin Cancer Res; 23(22); 6888–92. ©2017 AACR.

Translational Relevance

This article examines and confirms the prognostic significance of combined expression levels of three markers (SPP1, NCOA3, and RGS1) on outcome associated with melanoma in a prospectively collected cohort, a first in the setting of a prospectively defined, multicenter cohort of resected, high-risk melanoma. In addition, this study has identified a potential predictive role for marker testing in identifying a patient cohort that may benefit from adjuvant IFN therapy.

The unpredictable clinical behavior of melanoma has prompted numerous investigations into the impact of various prognostic factors to better predict melanoma outcome (reviewed in ref. 1). In 2001, the American Joint Committee on Cancer (AJCC) melanoma staging committee amassed a large dataset of melanoma patients and validated the prognostic impact of tumor thickness and ulceration in primary melanoma, and of nodal status in patients with localized disease (2). In an updated analysis in 2009, mitotic rate replaced Clark level in the staging classification for patients with thin (≤1 mm) primary tumors (3), only to be removed in the most recent edition of the staging classification (4).

Given the limitations of existing histopathologic markers for melanoma prognosis, molecular factors represent the next frontier for prognostic factor development (reviewed in refs. 5–7). Assessment of molecular prognostic factors for melanoma has been hampered by the lack of large, well-annotated tissue sets in which to conduct these analyses, and by the lack of validation of promising markers in independent patient cohorts. Importantly, to date, no tissue-based prognostic markers have been validated in a prospective cohort, ideally collected as part of a therapeutic clinical trial with predetermined eligibility criteria. As a result, no molecular markers are routinely assessed in the prognostic assessment of melanoma patients, and none are incorporated into the AJCC staging classification for melanoma.

Previously, we reported the independent prognostic impact of an IHC assay for the combined expression levels of three markers (SPP1, RGS1, and NCOA3) derived from gene expression profiling in two retrospective cohorts: one based on a training cohort of 395 patients from the United States, and the other based on a validation cohort of 141 patients drawn from two German centers (Heidelberg–Kiel cohort; ref. 8). Multimarker scores were independently predictive of survival and emerged as the top factor predicting survival in each cohort. On the basis of these results, we were granted access to tissues collected as part of the intergroup E1690 adjuvant therapy trial to assess the survival impact of multimarker expression levels in this prospective cohort.

Study population

The eligible population for E1690 included patients with a primary melanoma greater than 4 mm thick, or patients with melanoma lymph node metastasis (9). At least one unstained slide was available from 248 patients for this analysis. The demographic composition of this 248-patient sample is provided in Table 1. Written informed consent was obtained from patients for the tissue collection undertaken during the E1690 clinical trial. This study was approved by the Institutional Review Board at California Pacific Medical Center and was conducted in accordance with recognized ethical guidelines (e.g., Declaration of Helsinki).

Table 1.

Characteristics of E1690 tissue cohort (N = 248)

  
  
Male gender 167 (67.3%) 
Age over 50 (years) 88 (35.5%) 
Ulceration present 89 (35.9%) 
T stage 
  T1 34 (13.7%) 
  T2 56 (22.6%) 
  T3 62 (25.0%) 
  T4 88 (35.5%) 
  Unknown 8 (3.2%) 
Tumor site 
  Trunk 113 (45.6%) 
  Extremity 105 (42.3%) 
  Head and neck 27 (10.9%) 
  Unknown 3 (1.2%) 
Clark level 
  II/III 73 (29.4%) 
  IV/V 153 (61.7%) 
  Unknown 22 (8.9%) 
Number of positive lymph nodes 
  0 23 (9.3%) 
  1 73 (29.4%) 
  2–3 46 (18.6%) 
  >3 40 (16.1%) 
  Unknown 66 (26.6%) 
  
  
Male gender 167 (67.3%) 
Age over 50 (years) 88 (35.5%) 
Ulceration present 89 (35.9%) 
T stage 
  T1 34 (13.7%) 
  T2 56 (22.6%) 
  T3 62 (25.0%) 
  T4 88 (35.5%) 
  Unknown 8 (3.2%) 
Tumor site 
  Trunk 113 (45.6%) 
  Extremity 105 (42.3%) 
  Head and neck 27 (10.9%) 
  Unknown 3 (1.2%) 
Clark level 
  II/III 73 (29.4%) 
  IV/V 153 (61.7%) 
  Unknown 22 (8.9%) 
Number of positive lymph nodes 
  0 23 (9.3%) 
  1 73 (29.4%) 
  2–3 46 (18.6%) 
  >3 40 (16.1%) 
  Unknown 66 (26.6%) 

IHC analysis

The tissue sections were stained with mAb targeting SPP1, RGS1, and NCOA3 using a Ventana autostainer (BenchMark Ultra). The final staining protocols for the three antibodies appear below and were the results of a rigorous optimization process, accounting for variables such as antigen retrieval, detection kits selection, primary antibody dilution factor, and incubation time, as well as signal amplification. The entire process, including the finalized staining protocol, was reviewed and approved by the study pathologist. In the case of NCOA3 (using Santa Cruz Biotechnology antibody #9119; clone G2711; 1:100 dilution), antigen retrieval, uView Detection Kit, and amplification were utilized. In the case of SPP1 (Abcam ab8448; lot GR52573, 1:200 dilution), antigen retrieval and uView Detection Kit were utilized. In the case of RGS1 (Genetex GT112803; lot 40828; 1:100 dilution), antigen retrieval and iView Detection Kit were utilized.

Evaluation of marker expression

The whole slide image of tissue sections was scanned using the Mirax MIDI whole slide high resolution scanning system (Carl Zeiss MicroImaging). The digitization process was controlled via software running on a Microsoft Windows XP using an Allied Vision Marlin CCD Camera (Allied Vision Technologies GmbH) with a Zeiss Plan-Apopchromat 20×/0.8 NA objective (Carl Zeiss Optronics GmbH) to generate images at a resolution of 0.32 μm/pixel. Regions with an identifiable melanocytic lesion were selected for evaluation. Characterization of the IHC staining was calculated by applying a segmentation feature with two different phase measurement masks recognizing nuclei (hematoxylin-stained), and cytoplasm (brown-immunostained). Image analysis was performed by computer-assisted color segmentation (using IAE-NearCYTE software) by building a range of hue, saturation, and value coordinates for user-selected color points. The defined color model was then applied to the selected area to determine the percentage of positive color-expressing pixels. For each positive pixel, the intensity (defined as the average of red, green, and blue color values) and the lightness (defined as the average of the largest and smallest color components) were calculated. Marker expression was determined in the entire cellular compartment, as well as in the nuclear and cytoplasmic compartments. The final number of cases with an expression score for each marker was significantly reduced due to several factors and was as follows for each marker: 130 cases for SPP1, 125 for RGS1, and 108 for NCOA3. For example, in the case of SPP1, there were 245 slides stained, in which data were missing for the following reasons: 58 cases with wash off of tissue following the staining reaction, 34 cases with lack of tumor on the slide, and 23 cases with a failure in scanning due to tissue folding.

Statistical analysis

In the case of duplicate sections for a given case, the highest score in each compartment for each marker was selected for subsequent analyses, consistent with our prior analysis (8). The initial analysis focused on determining the most relevant of the three compartment measures of marker expression in the prediction of survival of the E1690 cohort. We analyzed marker expression in the nucleus, in the cytoplasm, and the total cellular compartment. This analysis identified nuclear staining for each marker as the sole predictor of both relapse-free survival (RFS) and overall survival (OS). Neither the analysis of the cytoplasmic compartment nor of total cellular expression for each marker yielded statistically significant associations with RFS or OS. As a result, all additional analyses focused on nuclear expression of the three markers. For each marker, the cut-off point utilized was the median nuclear expression score for that marker within the sample. Subsequently, to assess the joint effect of the combination of the three biomarkers on melanoma survival, an unweighted prognostic index was developed (with an equal relative contribution of each marker to the index), indicating for each patient a count (0, 1, 2, or 3) of the number of markers whose separate expression scores were high (i.e., fell at or above the sample median for that marker). Use of this index required marker expression to be available for all three factors, resulting in a reduction of the effective sample size to 52 patients. Prognostic analyses were performed initially utilizing the entire 0 to 3 scale. Subsequently, the optimal cut-off point for the index was identified by testing the three different binary cut-off points available, identifying the cut-off point of score of 3 versus <3 as the most significantly predictive of both RFS and OS. The association between increasing multimarker expression scores and RFS and OS of the available tissue sample was assessed using Kaplan–Meier analysis, and both univariate and multivariate Cox regression analyses. All reported P values are two-tailed.

E1690 was a randomized clinical trial examining the utility of adjuvant high-dose versus low-dose IFN versus observation in the setting of resected, high-risk melanoma, whose results were previously described (9). Tissue sections were available on 248 patients enrolled in the trial, with the following treatment assignments: 78 cases from patients on the observation arm; 88 cases from patients on the low-dose IFN arm; and 82 cases from the patients on the high-dose IFN arm. Given that tissues were available on only a subset of the total trial cohort, we initially compared the survival of the subset of patients with available tissue to the overall cohort and found no significant difference in RFS or OS between the two.

The primary goal of the current study was the validation of the prognostic impact of combined expression of SPP1, RGS1, and NCOA3 in the E1690 trial cohort. To this end, a multimarker index was developed to combine marker expression using a 4-point scale (0–3). Initially, the impact of the multimarker index on survival endpoints was evaluated as a 4-point discrete variable, incorporating the entire scale. Increasing multimarker expression scores were significantly predictive of RFS (HR = 1.65, P = 0.016) and OS (HR = 1.63, P = 0.027) by univariate Cox regression analyses. We then assessed the association between molecularly defined high-risk and low-risk groups and survival by analyzing different cut-off points within the index, with the optimal cut-off point identified at a count of 3 versus less than 3. Kaplan–Meier analysis of molecularly defined high-risk versus low-risk groups revealed a significant difference in RFS (HR = 4.25, P < 0.001; Fig. 1) and OS (HR = 5.34, P < 0.0001, Fig. 2). All subsequent analyses of multimarker expression scores were performed using this high-risk versus low-risk dichotomization.

Figure 1.

Kaplan–Meier analysis of RFS in molecularly defined high-risk versus low-risk patients in the E1690 cohort.

Figure 1.

Kaplan–Meier analysis of RFS in molecularly defined high-risk versus low-risk patients in the E1690 cohort.

Close modal
Figure 2.

Kaplan–Meier analysis of OS in molecularly defined high-risk versus low-risk patients in the E1690 cohort.

Figure 2.

Kaplan–Meier analysis of OS in molecularly defined high-risk versus low-risk patients in the E1690 cohort.

Close modal

Next, we performed multivariate Cox regression analyses to determine the independent impact of multimarker expression scores on survival. With respect to RFS, we initially performed an analysis that included dichotomized multimarker expression scores, as well as standard prognostic factors for melanoma recorded in the dataset, including tumor thickness, ulceration status, Clark level, patient age, patient gender, and tumor site, to mirror the original prognostic analyses performed by the AJCC melanoma committee (10), as well as our own prior analyses (8). Stepwise multivariate Cox regression analysis with backward elimination of these seven factors revealed tumor thickness (P = 0.005) and high multimarker scores (P = 0.007, HR = 3.77) as the only factors significantly predictive of RFS. We then repeated the analysis, this time including all available factors, including treatment assignment (high-dose IFN vs. other groups) and number of positive lymph nodes. Adding these two additional prognostic factors produced the same results, with a significant impact of tumor thickness and multimarker scores.

Similar analyses were then performed to assess the association between multimarker expression and OS. Stepwise multivariate Cox regression analysis with backward elimination of multimarker expression scores, along with tumor thickness, ulceration status, Clark level, patient age, patient gender, and tumor site revealed high multimarker scores (P = 0.003, HR = 4.08) and tumor thickness (P = 0.04) as the only factors with a significant impact on OS. The additional analysis that also included treatment assignment and number of positive lymph nodes yielded the same two statistically significant factors.

A secondary goal proposed in this study was to assess the potential role of multimarker expression scores to identify patient subsets that may benefit from IFN treatment, in an attempt to identify a predictive role in addition to the demonstrated prognostic role. In the overall trial cohort, high-dose IFN therapy was associated with a significant impact on RFS, without a significant impact on OS (9). In the 248-patient available tissue sample, high-dose IFN therapy had a nonsignificant impact on both RFS and OS by univariate Cox regression analysis. There was no significant survival impact of IFN treatment versus observation in either of the molecularly defined low-risk or high-risk subgroups, analyzed separately. However, in the molecularly defined low-risk group, there was a significantly improved RFS in patients treated with high-dose IFN versus the other two groups combined by Cox regression analysis (P < 0.05, HR = 0.30).

In this study, we assessed the prognostic impact of combined expression of three molecular markers (SPP1, RGS1, and NCOA3) using IHC analysis, in a correlative study of the E1690 trial examining the utility of two doses of IFN versus observation as adjuvant therapies in the setting of resected, high-risk melanoma. High multimarker scores, using an index constructed to combine marker expression, was associated significantly with adverse outcome, as evidenced by significantly reduced RFS and OS. By multivariate analysis, high multimarker scores were significantly and independently predictive of RFS and OS. In the multivariate analysis of OS, multimarker expression scores were the top factor predicting survival, similar to our previously described results in two distinct cohorts (8). To our knowledge, this is the first study to validate the prognostic impact of any tissue-based molecular markers for melanoma in a prospectively collected cohort, and specifically in the setting of a multicenter clinical trial carried out by a cooperative group. The prognostic impact of combined marker expression had been previously demonstrated in two distinct retrospective cohorts of melanoma patients, but has now been validated in this analysis of the E1690 cohort.

It is important to note that there were significant differences between our initial analyses (8) and the current analysis. The initial training set consisted of a tissue microarray of primary melanoma specimens, whereas the validation (Heidelberg–Kiel) set consisted of tissue sections of primary melanoma specimens. The available E1690 tissue sample consisted of tissue sections primarily of lymph node metastases, with a small proportion of primary tumors. In addition, in the initial studies, IHC analysis was performed manually, whereas in the current analysis, IHC was performed using an autostainer routinely used in pathology laboratories. Finally, in the initial analyses, pathologist scoring was used to quantitate marker expression, whereas in the current study, a digital imaging platform was utilized. Despite these differences, combining the expression levels of SPP1, RGS1, and NCOA3 continued to provide significant and independent prognostic information in this E1690 cohort of melanoma patients. However, significant additional work will be required to refine the specific assays and algorithms required to prepare these markers for routine clinical use, as optimal cut-off points for marker expression will likely depend on the stage and type of melanoma tissue analyzed, as well as the scoring system utilized.

Nevertheless, these results suggest that SPP1, RGS1, and NCOA3 play important roles in melanoma biology. There is ample literature regarding the role of SPP1 (also known as osteopontin) in melanoma progression, including studies describing its relevance as a prognostic marker (11–13) and as a potential serum marker (14) in melanoma. In addition, SPP1 has demonstrated roles in melanoma progression by virtue of its participation in the NF-κB signaling pathway and of its activation of integrin αvβ3 (15, 16). In contrast, little is known regarding the roles played by RGS1 and NCOA3 in melanoma progression. Intriguingly, both genes reside on chromosomal loci shown to be gained in melanoma, including 1q31 (in the case of RGS1) and 20q12 (in the case of NCOA3; ref. 17). Thus, RGS1 and NCOA3, which were initially identified by their differential expression at the RNA level using cDNA microarray analysis (18), appear to be selected for at the DNA level in melanoma progression, ultimately resulting in their overexpression at the protein level as assessed in this and previous studies. However, additional studies will be required to define the functional roles played by RGS1 and NCOA3 in melanoma progression.

A secondary aim of this study was to identify a potential predictive role for our multimarker assay in the setting of IFN therapy. We were unable to identify a molecularly defined subset of patients that had a statistically significant survival benefit from high-dose IFN therapy when compared with the observation group. This task was made difficult by the small sample size available for testing, by the dilution of the treated sample into the high-dose and low-dose IFN groups, and by the lack of survival benefit to high-dose IFN in the tissue set available. As a result, additional studies will be required to determine whether marker expression can be used as a guide to select patients for therapy with IFN or ipilimumab, which has been recently shown to prolong OS in the adjuvant setting (19). However, the observation that the molecularly defined low-risk subgroup had a significantly improved RFS when compared with the other two groups combined is intriguing and deserves confirmation in a larger patient cohort. It would be especially interesting to assess multimarker expression levels in recently completed adjuvant trial cohorts for further validation of the prognostic role and to assess a putative predictive role.

Our study highlights some of the challenges encountered in the testing of tissue collected in a cooperative group setting. To begin with, we had access to tissues comprising a subset of the overall trial cohort. In addition, given the time that has elapsed since the trial's initiation and conduct, there was significant degradation of a high proportion of the samples, rendering several of the slides not usable for marker testing, resulting in further sample size losses in the final analysis.

Recently, a 31-gene signature has been shown to have a prognostic signal in a number of retrospective sample cohorts (20, 21) and has been marketed as a clinical prognostic test to melanoma patients and practitioners. However, its widespread use is controversial at the current time, as the assay has not been validated in a prospective, multicenter cohort and is not associated with benefit to a particular treatment, such as adjuvant immunotherapy with IFN or ipilimumab, the two FDA-approved adjuvant therapies for melanoma. In fact, the 2016 National Comprehensive Cancer Network guidelines indicate that currently available gene expression profiling tests should not be used to guide patient care, and not outside of a clinical trial (22). Our experience suggests that, despite the challenges of using tissues collected as part of cooperative group studies, such a validation is possible and indispensable for the further clinical development of any prognostic assay. Of note, SPP1, one of the markers used in our analysis, is also included in the 31-gene signature.

In conclusion, our study identifies a significant and independent prognostic role for combined expression of SPP1, RGS1, and NCOA3, a first in the setting of a prospectively defined, multicenter cohort of resected, high-risk melanoma. Further validation of the prognostic role of these markers in melanoma is warranted, as is its potential predictive role to select patients for active adjuvant immunotherapy regimens.

M. Kashani-Sabet has ownership interests (including patents) at Melanoma Diagnostics and Myriad Genetics, reports receiving speakers bureau honoraria from Cepheid, and reports receiving commercial research grants from Merck. J. Miller has ownership interests (including patents) at MDMS LLC. J. Kirkwood reports receiving speakers bureau honoraria from and is a consultant/advisory board member for Array, Bristol-Myers Squibb, EMD Serono, Genentech, Merck, Novartis and Roche. No potential conflicts of interest were disclosed by the other authors.

The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. government. The opinions expressed in this article are those of the authors and do not necessarily represent those of Merck Sharp & Dohme Corp.

Conception and design: M. Kashani-Sabet, J.R. Miller III, S.J. Lee, J.M. Kirkwood

Development of methodology: M. Kashani-Sabet, M. Nosrati, J.R. Miller III

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Nosrati, S.P.L. Leong, S. Tong, J.M. Kirkwood

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Kashani-Sabet, M. Nosrati, J.R. Miller III, A. Lesniak, S. Tong, S.J. Lee, J.M. Kirkwood

Writing, review, and/or revision of the manuscript: M. Kashani-Sabet, M. Nosrati, J.R. Miller III, R.W. Sagebiel, S.P.L. Leong, A. Lesniak, S. Tong, S.J. Lee, J.M. Kirkwood

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Nosrati, S. Tong

Study supervision: M. Kashani-Sabet, M. Nosrati

Other [IHC antibody optimization, Kaplan–Meier survival analysis, digital pathology, staff/resource management, and graphic arts (figure preparation/production)]: M. Nosrati

This study was coordinated by the ECOG-ACRIN Cancer Research Group (Robert L. Comis, MD and Mitchell D. Schnall, MD, PhD, Group Co-Chairs).

This work was funded in part by a grant from the NIH (R01CA114337) and a research grant from Merck Sharp & Dohme Corp. This work was also supported by the NCI of the NIH under the following award numbers CA180820, CA180794, and CA180844.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Zettersten
E
,
Shaikh
L
,
Ramirez
R
,
Kashani-Sabet
M
. 
Prognostic factors in primary cutaneous melanoma
.
Surg Clin North Am
2003
;
83
:
61
75
.
2.
Balch
CM
,
Buzaid
AC
,
Soong
SJ
,
Atkins
MB
,
Cascinelli
N
,
Coit
DG
, et al
Final version of the American Joint Committee on Cancer staging system for cutaneous melanoma
.
J Clin Oncol
2001
;
19
:
3635
48
.
3.
Balch
CM
,
Gershenwald
JE
,
Soong
SJ
,
Thompson
JF
,
Atkins
MB
,
Byrd
DR
, et al
Final version of 2009 AJCC melanoma staging and classification
.
J Clin Oncol
2009
;
27
:
6199
206
.
4.
Gershenwald
JE
,
Scolyer
RA
,
Hess
KR
,
Thompson
JF
,
Long
GV
,
Ross
MI
, et al
:
Melanoma of the skin, in: AJCC Cancer Staging Manual (ed 8)
.
Switzerland
:
Springer
; 
2017
.
pp
563
89
.
5.
Kashani-Sabet
M
. 
Molecular markers in melanoma
.
Br J Dermatol
2014
;
170
:
31
5
.
6.
Weiss
SA
,
Hanniford
D
,
Hernando
E
,
Osman
I
. 
Revisiting determinants of prognosis in cutaneous melanoma
.
Cancer
2015
;
121
:
4108
23
.
7.
Foth
M
,
Wouters
J
,
de Chaumont
C
,
Dynoodt
P
,
Gallagher
WM
. 
Prognostic and predictive biomarkers in melanoma: an update
.
Expert Rev Mol Diagn
2016
;
16
:
223
37
.
8.
Kashani-Sabet
M
,
Venna
S
,
Rangel
J
,
Sucker
A
,
Egberts
F
,
Simko
J
, et al
A multi-marker prognostic assay for melanoma
.
Clin Cancer Res
2009
;
15
:
6987
92
.
9.
Kirkwood
JM
,
Ibrahim
JG
,
Sondak
VK
,
Richard
J
,
Flaherty
LE
,
Ernstoff
MS
, et al
High- and low-dose interferon α-2b in high risk melanoma: first analysis of Intergroup Trial E1690/S9111/C9190
.
J Clin Oncol
2000
;
18
:
2444
58
.
10.
Balch
CM
,
Soong
SJ
,
Gershenwald
JE
,
Thompson
JF
,
Reintgen
DS
,
Cascinelli
N
, et al
Prognostic factors analysis of 17,600 melanoma patients: validation of the American Joint Committee on Cancer melanoma staging system
.
J Clin Oncol
2001
;
19
:
3622
34
.
11.
Alonso
SR
,
Tracey
L
,
Ortiz
P
,
Perez-Gomez
B
,
Palacios
J
,
Pollan
M
, et al
A high-throughput study in melanoma identifies epithelial-mesenchymal transition as a major determinant of metastasis
.
Cancer Res
2007
;
67
:
3450
60
.
12.
Rangel
J
,
Nosrati
M
,
Torabian
S
,
Shaikh
L
,
Leong
SP
,
Haqq
C
, et al
Osteopontin as a molecular prognostic marker for melanoma
.
Cancer
2008
;
112
:
144
50
.
13.
Conway
C
,
Mitra
A
,
Jewell
R
,
Randerson-Moor
J
,
Lobo
S
,
Nsengimana
J
, et al
Gene expression profiling of paraffin-embedded primary melanoma using the DASL assay identifies increased osteopontin expression as predictive of reduced relapse-free survival
.
Clin Cancer Res
2009
;
15
:
6939
46
.
14.
Filia
A
,
Elliott
F
,
Wind
T
,
Field
S
,
Davies
J
,
Kukalizch
K
, et al
Plasma osteopontin concentrations in patients with cutaneous melanoma
.
Oncol Rep
2013
;
30
:
1575
8
.
15.
Philip
S
,
Bulbule
A
,
Kundu
GC
. 
Osteopontin stimulates tumor growth and activation of promatrix metalloproteinase-2 through nuclear factor-κB-mediated induction of membrane type 1 matrix metalloproteinase in murine melanoma cells
.
J Biol Chem
2009
;
276
:
44926
35
.
16.
Yin
M
,
Soikkeli
J
,
Jahkola
T
,
Virolainen
S
,
Saksela
O
,
Holtta
E
. 
Osteopontin promotes the invasive growth of melanoma cells by activating integrin αvβ3 and down-regulating tetraspanin CD9
.
Am J Pathol
2014
;
184
:
842
58
.
17.
Curtin
JA
,
Fridlyand
J
,
Kageshita
T
,
Patel
HN
,
Busam
KJ
,
Kutzner
H
, et al
Distinct sets of genetic alterations in melanoma
.
N Engl J Med
2005
;
353
:
2135
47
.
18.
Haqq
C
,
Nosrati
M
,
Sudilovsky
D
,
Crothers
J
,
Khodabaksh
D
,
Pulliam
B
, et al
The gene expression signatures of melanoma progression
.
Proc Natl Acad Sci U S A
2005
;
102
:
6092
7
.
19.
Eggermont
AM
,
Chiarion-Sileni
V
,
Grob
JJ
,
Dummer
R
,
Wolchok
JD
,
Schmidt
H
, et al
Prolonged survival in stage III melanoma with ipilimumab adjuvant therapy
.
N Engl J Med
2016
;
375
:
1845
55
.
20.
Gerami
P
,
Cook
RW
,
Wilkinson
J
,
Russell
MC
,
Dhillon
N
,
Amaria
RN
, et al
Development of a prognostic genetic signature to predict the metastatic risk associated with cutaneous melanoma
.
Clin Cancer Res
2015
;
21
:
175
83
.
21.
Gerami
P
,
Cook
RW
,
Russell
MC
,
Wilkinson
J
,
Amaria
RN
,
Gonzalez
R
, et al
Gene expression profiling for molecular staging of cutaneous melanoma in patients undergoing sentinel lymph node biopsy
.
J Am Acad Dermatol
2015
;
72
:
780
5
.
22.
Coit
DG
,
Thompson
JA
,
Algazi
A
,
Andtbacka
R
,
Bichakjian
CK
,
Carson
WE
, et al
NCCN Guidelines Insights: Melanoma, Version 3.2016
.
J Natl Compr Canc Netw
2016
;
14
:
945
58
.