SEARCH

SEARCH BY CITATION

Keywords:

  • melanoma;
  • brain metastases;
  • craniotomy immune infiltrate;
  • gene expression profiling

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SOURCES
  8. CONFLICT OF INTEREST DISCLOSURE
  9. REFERENCES
  10. Supporting Information

BACKGROUND

The prognosis of metastatic melanomas to the brain (MBM) is variable with prolonged survival in a subset. It is unclear whether MBM differ from extracranial metastases (EcM) and primary melanomas (PrM).

METHODS

To study the biology of MBM, histopathologic analysis of tumor blocks from patients' craniotomy samples and whole-genome expression profiling (WGEP) with confirmatory immunohistochemistry were performed.

RESULTS

High mononuclear infiltrate and low intratumoral hemorrhage were associated with prolonged overall survival (OS). Pathway analysis of WGEP data from 29 such craniotomy tumor blocks demonstrated that several immune-related BioCarta gene sets were associated with prolonged OS. WGEP analysis of MBM in comparison with same-patient EcM and PrM showed that MBM and EcM were similar, but both differ significantly from PrM. Immunohistochemical analysis revealed that peritumoral CD3+ and CD8+ cells were associated with prolonged OS.

CONCLUSIONS

MBMs are more similar to EcM compared with PrM. Immune infiltrate is a favorable prognostic factor for MBM. Cancer 2013;119:2737–2746. © 2013 American Cancer Society.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SOURCES
  8. CONFLICT OF INTEREST DISCLOSURE
  9. REFERENCES
  10. Supporting Information

Metastases to the brain are the most frequent intracranial tumors in adults,[1] and melanoma is the most frequent solid cancer to metastasize to the brain.[2] B-Raf inhibitors and ipilimumab are active against metastatic melanomas to the brain (MBM),[3, 4] and mutations in B-Raf and N-Ras proteins are associated with higher incidence of MBM.[5] No association was found between B-RafV600E mutations in MBM and overall survival (OS).[6] MBM exhibit higher levels of phosphoinositide 3-kinase (PI3K) pathway activation compared with that of extracranial metastases (EcM).[7, 8]

The brain microenvironment differs from that of extracranial sites in that it lacks lymphatics and contains glial cells that may influence tumor growth.[9] Blood vessels in brain tumors differ from normal brain vessels.[10] Immune cells have also been described in a handful of metastatic solid tumors to the brain, although their prognostic significance is unknown.[11]

To gain insight into the biology of MBM, we performed histopathologic analysis followed by whole-genome expression profiling (WGEP) and confirmatory immunohistochemistry (IHC) or in situ hybridization (ISH) analysis of craniotomy tumor specimens. Our study proposes new prognostic factors for MBM, and may explain clinical efficacy of immunotherapies in patients with MBM.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SOURCES
  8. CONFLICT OF INTEREST DISCLOSURE
  9. REFERENCES
  10. Supporting Information

Patients and Tumors

Under institutional review board–approved protocols, deidentified cases from patients who underwent craniotomy for MBM at the University of Pittsburgh Medical Center were obtained by performing a CoPath pathology search followed by retrieval of formalin-fixed paraffin-embedded (FFPE) tumor blocks from craniotomy, EcM, and primary melanoma (PrM) specimens.

Histopathologic Assessment of MBMs

Hematoxylin and eosin (H&E)-stained sections were reviewed by a neuropathologist (R.H.) who was blinded to patient history. Immune infiltrate, hemorrhage, gliosis, pigmentation, and necrosis were semiquantitatively scored on a 0 to 3+ scale with 0 to 1+ being absent/low and 2 to 3+ being high. Tumor hemorrhage was defined by the presence of fresh hemorrhage, hematoidin-laden macrophages, organized blood clot, and ruptured vessels or hemorrhage adjacent to necrotic areas. Hemorrhage was quantitated as low with 0 to 2 foci of hematoidin, fresh blood, or organized clot, or as high with blood occupying more than one-third of the specimen (Fig. 1A). Immune infiltrate was quantitated by the presence of mononuclear cells around blood vessels and/or within the tumor parenchyma. Low infiltrate was defined as 0 to 2 perivascular infiltrates, and high infiltrate was defined as > 2 perivascular and/or any infiltrates within the tumor parenchyma (Fig. 1B). Melanin was assessed using a combination of H&E and Gomori's modified iron stain (Fig. 1C). Gliosis was defined as the presence of reactive astrocytes only near the tumor. Necrosis was estimated as a percentage of necrotic tumor.

image

Figure 1. Tumor sections of metastatic melanomas to the brain are shown. (A) Hematoxylin and eosin (H&E)-stained sections that were scored high and low for hemorrhage (black open arrows) are shown (B) H&E-stained sections that were scored high and low for immune infiltrate (yellow open arrows) are shown (C) If H&E-stained sections showed pigmentation, adjacent sections were stained with potassium ferrocyanide and counterstained with nuclear fast red (Kernechtrot). Hemosiderin is stained blue (red open arrows) and melanin remains brown (yellow arrows). Magnifications are shown in right-hand panels.

Download figure to PowerPoint

WGEP Analysis

Using a blade under the guidance of H&E-staining of every 10th adjacent 5-μm section, tissues corresponding to non-necrotic tumor devoid of immune infiltrate, hemorrhage, and glial cells were microdissected. Deparaffinization of tissue pellets, RNA extraction, purification, and incubation with Human Ref-8, version 3, BeadChips (Illumina, San Diego, Calif) followed by scanning on the Illumina BeadStation GX were performed in the University of Pittsburgh Cancer Institute's Cancer Biomarkers Illumina Platform Facility.[12]

Sources of Antibodies

The following primary antibodies were used: CD3 (Dako, Carpinteria, Calif), CD4 (Vector Laboratories, Burlingame, Calif), CD8 (Dako), CD14 (Vector), CD19 (Leica Microsystems, Buffalo Grove, Ill), Forkhead box P3 (FoxP3; eBiosciences, San Diego, Calif), CD247 (Sigma-Aldrich, St. Louis, Mo), transforming growth factor beta (TGFβ; Abcam, Cambridge, Mass), and glial fibrillary acidic protein (GFAP; Dako). The following antibodies were generated in Dr. Ferrone's laboratory: human leukocyte antigen (HLA) class I (HC-10/HC-A2 clones), HLA class II (LGII-612.14), and tapasin (TO-3).

IHC, ISH, and Scoring Definitions

For IHC, FFPE tumor sections were probed with antibodies and stained with Vulcan Fast Red (Biocare Medical, Concord, Calif). Mononuclear cells positive for CD3, CD4, CD8, CD14, CD19, and CD247 were scored separately (R.H. and S.M.) for peritumoral and intratumoral compartments, as described (Fig. 2).[13] Expression of TGFβ, HLA class I/II, and tapasin by melanoma cells were assessed using the H-score whose median was used as the cutoff value between high versus low expression. ISH analyses were performed (B.F.J. and T.R.) using human-specific sequences for detection of chemokine (CXCL13, CCL19, CCL21) messenger RNAs. Autoradiographic exposure times of [35S]-labeled riboprobes were 7 to 10 days.[14]

Statistical Analysis

Cox proportional hazards modeling was used to identify clinicopathologic variables associated with OS defined as time-to-death from first craniotomy, by fitting a univariate regression model for each variable. Kaplan–Meier survival analysis using the log-rank significance test was also performed for dichotomous variables. Patients who were still alive between the first craniotomy and March 12, 2012, were censored at the date of last follow-up. Curves were constructed using IBM SPSS Statistics software, release 19.0.0 (IBM, Armonk, NY). Variables that were significantly associated with OS in univariate analysis were subjected to multivariate Cox regression analysis.

Bioinformatics Analysis

Bioinformatics analysis of WGEP data was performed using the Biometric Research Branch-Array (BRB) Tools software (http://linus.nci.nih.gov/BRB-ArrayTools.html). Data were quantile-normalized using the lumi R package. Probe sets were excluded from further analysis if < 20% of gene expression data values had ≥ 1.5-fold change in either direction of a probe set's median value and the percentage of data missing exceeded 50%. To assess whether specific cellular pathways were associated with OS, we performed survival pathway analysis on the MBM data set using the survival Gene Set Analysis (GSA) tool. The Efron-Tibshirani “maxmean” test was applied to identify gene sets at a P = .05 significance level.

Hierarchical clustering analysis was used to assess the degree of association between PrM, EcM, and MBM using either all genes that passed the filtering criteria or highly discriminant genes only. Highly discriminating genes were selected as differentially expressed between 2 same-patient different sites (MBM–EcM, MBM–PrM, and EcM–PrM) using paired t test. Analysis of variance (ANOVA) using all genes that passed the filtering criteria was used to assess whether MBMs are significantly different from EcM and/or PrM. Enrichment analysis of differentially expressed probe sets to determine their biological annotation to specific cellular pathways was performed for the same patient EcM–MBM tumor samples using the DAVID (Database for Annotation, Visualization, and Integrated Discovery) database (http://david.abcc.ncifcrf.gov). Enrichment P value was set to .05 after Benjamini-Hochberg multiple comparison correction testing.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SOURCES
  8. CONFLICT OF INTEREST DISCLOSURE
  9. REFERENCES
  10. Supporting Information

Clinical Characteristics of Patients With MBM

A total of 115 patients underwent craniotomy at the University of Pittsburgh Medical Center between November 1995 and July 2011 (Table 1), and 26% and 30% of patients originally presented either with unknown primary[15] or thin (stage I according to American Joint Committee on Cancer guidelines) melanoma, respectively. No patients had received vemurafenib or radiation therapy prior to craniotomy, and only 2 of 9 patients with B-RafV600 mutation received vemurafenib following craniotomy.

Table 1. Clinicopathologic Characteristics of Patients Who Underwent Craniotomya
ParameterValue
  1. a

    Patients were considered to have “partial” criteria for unknown primary[15] if they had no knowledge of excision of a pigmented lesion.

  2. b

    Stage based on AJCC Cancer Staging Manual, 7th edition (2009).

Age, y, median (range)51 (14-83)
Sex 
Male74
Female41
Melanoma origin 
Cutaneous79
Noncutaneous2
Unknown primary (all criteria)22
Unknown primary (partial criteria)8
Unknown4
B-Raf mutation status 
Known19
B-RafV600E6
B-RafV600K3
B-Raf wild type10
Unknown96
Stageb 
I34
II23
III11
IV30
Unknown17

High Immune Infiltrate and Low Hemorrhage in MBMs Are Associated With Prolonged OS

Because of tumor block availability (n = 106), absence of viable tumor cells (n = 3), or cytology specimens (n = 2), our final analysis was based on 101 of the cases. Immune infiltrate (P = .006), hemorrhage (P = .04), recursive partitioning analysis class (P < .0001), Eastern Cooperative Oncology Group performance status (P = .024), and local therapy after craniotomy (P = .019) were significantly associated with OS (Table 2). No significant correlation between B-Raf mutation status and immune infiltrate or hemorrhage was noted (Pearson's chi-square test P = .57 and P = .49, respectively). Figure 3 shows the Kaplan-Meier OS curves partitioned by immune infiltrate and hemorrhage. A significant interaction between the “immune infiltrate” and “hemorrhage” variables was noted (Cox regression test, P = .0031). In particular, patients with high immune infiltrate and low/absent hemorrhage had prolonged OS compared with all remaining patients (543 versus 164 days, log-rank P < .001). Cox multiple regression analysis showed that immune infiltrate (P = .008), Eastern Cooperative Oncology Group performance status (P = .043), recursive partitioning analysis (P < .0001), and local therapy (P = .0005) remained significant predictors of OS.

Table 2. Clinicopathologic Factors in Relation to Overall Survival in Patients With Metastatic Melanomas to the Braina
VariableNo. of PatientsMedian Survival (days)Log-Rank PKaplan-Meier Estimator, HR (95% CI)
  1. Abbreviations: CI, confidence interval; ECOG, Eastern Cooperative Oncology Group; HR, hazard ratio.

  2. a

    Presented P values are from univariate analysis.

Age at craniotomy    
<6580189.331.30 (0.75-2.15)
≥6521137  
Sex    
Male63154.101.44 (0.94-2.26)
Female38228  
ECOG performance status    
012481.0241.71 (1.05-2.70)
161191  
224142  
348  
Extracranial disease    
Absent48143.360.82 (0.54-1.26)
Present51191  
Unknown2143  
Recursive partitioning analysis    
126475<.00013.4 (1.97-6.23)
>174151  
Systemic therapy prior to craniotomy  .66 
Immunotherapy31244 0.81 (0.50-1.29)
Chemotherapy8185 1.26 (0.55-2.52)
Sequential or combination therapy4335 0.78 (0.23-1.90)
No systemic therapy58154  
Number of brain lesions    
149181.610.88 (0.52-1.44)
>129212  
Systemic therapy after craniotomy  .13 
Immunotherapy7212 0.43 (0.13-1.06)
Chemotherapy33222 0.95 (0.60-1.48)
Sequential or combination therapy41011 0.35 (0.09-0.98)
No systemic therapy57133  
Local therapy after craniotomy  .019 
Stereotactic radiosurgery only50248 0.34 (0.20-0.58)
Whole-brain irradiation only7253 0.37 (0.14-0.88)
Repeat craniotomy4170 0.76 (0.22-2.00)
Combination local therapy17232 0.41 (0.21-0.78)
No local therapy2357  
Immune infiltrate    
Low56160.0060.54 (0.35-0.84)
High44259  
Nonassessable1   
Hemorrhage    
Low42234.041.58 (1.02-2.48)
High58160  
Nonassessable1   
Necrosis    
Low56156.560.88 (0.57-1.34)
High44202  
Nonassessable1   
Melanin    
Low73181.280.75 (0.44-1.24)
High23191  
Nonassessable5   
Gliosis    
Low9120.250.64 (0.32-1.48)
High45164  
Nonassessable47   

WGEP of MBM Identifies Pathways Associated With Outcome

To gain insight regarding cellular processes associated with survival in MBM, we performed WGEP of RNA obtained from FFPE brain sections. Only 29 patients had tumor blocks that were suitable for microdissection. Survival analysis using the log-rank test of the entire 101-patient cohort compared with the 29-patient subset revealed no significant difference in time-to-death from craniotomy (P = .17; median survival is 177 versus 246 days, hazard ratio = 0.72, 95% confidence interval = 0.49-1.13). We used 15,067 probe sets (of a total of 24,526) that passed the filtering criteria to perform Cox proportional hazards model in order to identify gene sets associated with survival. Table 3 shows 20 (of a total of 284) prognostic BioCarta pathways. The leading gene sets that were associated with prolonged OS are immune-related, with the T-cell receptor (TCR) function pathway being the most significant category. In contrast, pathways associated with shortened OS involved genes associated with hypoxia, the lissencephaly gene (LIS1) in neuronal migration and development, and oxidative stress, among others.

Table 3. Biocarta Pathways That Were Prognostically Significant in Metastatic Melanomas to the Brain
 Pathways Associated With Good PrognosisGSA Test P Value
1CD3 complex<.005
2T helper (Th) surface molecules<.005
3HIV-induced T-cell apoptosis<.005
4B-cell surface molecules<.005
5Th1/Th2 differentiation<.005
6Role of Tob in T-cell activation<.005
7Activation of Csk inhibits signaling through the TCR.005
8Lck and Fyn kinases initiate TCR activation.005
9Cells/molecules involved in local acute inflammatory response.005
10Dendritic cells regulate Th1/Th2 development.005
 Pathways Associated With Adverse Prognosis 
1West Nile Virus<.005
2Y branching of actin filaments.005
3Hypoxia-inducible factor.005
4Regulation of ck1/cdk5 by type 1 glutamate receptors.005
5fMLP-induced chemokine expression in HMC-1 cells.005
6Free radical–induced apoptosis.01
7Repression of pain sensation by DREAM.015
8Transcription regulation by methyltransferase of CARM1.015
9Cadmium induces DNA synthesis and proliferation in macrophages.02
10Lissencephaly gene in neuronal migration and development.025

Validation of WGEP and Histopathologic Data

Because one of the gene set categories associated with prolonged OS was the TCR pathway, we performed IHC analysis of MBM for immune cell subsets. High peritumoral CD3+ and CD8+ mononuclear cells were significantly associated with prolonged OS (Table 4). To assess whether melanoma-infiltrating immune cells were functional within MBMs, we stained tumor sections for CD247, the ζ chain of the TCR that is essential for amplification of TCR signaling and is frequently lost in cancer.[16] Although CD247+ mononuclear cells were detectable, high numbers of CD247+ cells were infrequently observed. In addition, neither peritumoral CD4+ and CD14+ mononuclear cells nor the presence of any intratumoral mononuclear cell population tested were associated with prolonged OS (Table 4).

Table 4. Prognostic Significance of Immunohistochemistry and In Situ Hybridization Variables Assessed in Tissues of Metastatic Melanomas to the Braina
VariableMedian Survival (Days)Proportional Hazard PKaplan–Meier Estimator HR (95%CI)VariableMedian Survival (Days)Proportional Hazard PKaplan–Meier Estimator HR (95%CI)
  1. Abbreviations: CI, confidence interval; HR, hazard ratio.

  2. a

    Presented P values are from univariate analysis. All but chemokines (CCL19, CCL21, and CXCL13) were assessed using immunohistochemistry. Regarding CD19+ immune cells, values could not be estimated because none of the tissues contained high numbers (N/A indicates not applicable).

CD3+ (n = 40)   CD8+ (n = 37)   
Peritumoral   Peritumoral   
High (n = 7)688.0060.24High (n = 10)540.0420.43
Low (n = 33)156 (0.07-0.63)Low (n = 27)164 (0.18-0.95)
Intratumoral   Intratumoral   
High (n = 10)   High (n = 12)370.120.53
Low (n = 29)191.190.57Low (n = 24)171 (0.22-1.15)
N/A (n= 1)156 (0.22-1.27)N/A (n = 1)   
CD4+ (n = 36)   CD14+ (n = 38)   
Peritumoral   Peritumoral   
High (n = 6)370.240.57High (n = 11)281.510.76
Low (n = 29)164 (0.19-1.36)Low (n = 20)160 (0.32-1.67)
N/A (n = 1)   N/A (n = 7)   
Intratumoral   Intratumoral   
High (n = 3)51.132.5High (n = 13)156.541.25
Low (n = 32)191 (0.59-7.30)Low (n = 25)250 (0.58-2.53)
N/A (n = 1)       
CD19+ (n = 16)   FoxP3+ (n = 38)   
High (n = 0)N/AN/AN/AHigh (n = 19)246.870.94
Low (n = 16)293  Low (n = 19)178 (0.46-1.93)
CD247+ (n = 32)       
Peritumoral   HLA class I (n = 20)   
High (n = 5)156.720.82High (n = 10)219.900.94
Low (n = 27)246 (0.24-2.18)Low (n = 10)236 (0.37-2.35)
Intratumoral       
High (n = 4)536.840.88    
Low (n = 28)222 (0.20-2.61)    
HLA class II (n = 18)   Tapasin (n = 19)   
High (n = 9)148.601.29High (n = 10)281.410.67
Low (n = 9)246 (0.48-3.41)Low (n = 9)212 (0.24-1.75)
TGFβ (n = 21)   CCL19 (n = 31)   
High (n = 11)246.670.72Present (n = 9)273.680.84
Low (n = 10)370 (0.26-1.89)Absent (n = 22)184 (0.35-1.87)
CXCL13 (n = 30)   CCL21 (n = 27)   
Present (n = 17)293.790.9Present (n = 3)536.240.43
Absent (n = 13)143 (0.41-2.07)Absent (n = 24)260 (0.07-1.48)

We then investigated whether the higher immune cell infiltrate is secondary to either higher expression of molecules involved in the antigen presentation machinery by melanoma cells (eg, HLA class I, class II, tapasin) or to lower expression of the immunosuppressive cytokine TGFβ by melanoma cells or to the expression of the chemokines CCL19, CCL21, and CXCL13 by mononuclear cells[17] or to lower abundance of naturally occurring T regulatory cells (eg, FoxP3+). The expression and/or abundance of none of these molecules by melanoma cells and/or immune cell subsets was associated with prolonged OS (Table 4).

MBM Are Biologically Closer to EcM Than PrM

To assess whether MBM differ from other EcM or PrM in their WGEP, we performed WGEP of 72 samples. Same-patient MBM-EcM, MBM-PrM, and EcM-PrM pairs were available from 26, 12, and 12 patients, respectively. ANOVA using all probe sets showed that MBM and EcM cluster significantly differently from PrM (P < .0001). A dendrogram of the 72 samples was constructed using all 15,067 probe sets (Fig. 2). To measure the degree of similarity between MBM, EcM, and PrM, we used the Euclidean distance metric and average linkage and calculated the distances between same-patient MBM-EcM, MBM-PrM, and EcM-PrM. The distance between MBM-EcM was significantly smaller, compared to distances calculated between the other 2 groups (MBM-PrM and EcM-PrM) when clustering was performed on all 72 samples using all 15,067 probe sets (mean = 124 versus 169 and 162, respectively; ANOVA P = .09, t test to compare the distance between MBM-EcM and MBM-PrM as well as between MBM-EcM and EcM-PrM have a P value of .04 for both).

One of the probe sets that was significantly upregulated in MBM, as opposed to EcM and PrM, was GFAP, a protein expressed by reactive astrocytes. IHC analysis confirmed that GFAP was detectable only in astrocytes of 21 MBM and absent in all 6 stained EcM. GFAP+ cells were not only surrounding melanoma cells, but were also detectable in clusters or islands within the tumor (Fig. 4B). DAVID analysis of the differentially expressed genes between same-patient EcM-MBM tumor samples failed to reveal cellular processes that were significantly different between EcM and MBM.

image

Figure 2. Assessment of the degree of similarity between metastatic melanomas to the brain (MBM), extracranial metastases (EcM), and primary melanoma (PrM). Dendrogram plot shows the gene expression relatedness of the 72 samples based on the 15,067 probe sets that passed the filtering criteria. Each sample is labeled by its site (B, brain, red; E, extracranial, blue; P, primary, black) and patient number. Brackets below patient samples link same-patient MBM and EcM (black: closely related; gray: distantly related; red: closely related distant metastases, EcM or MBM, with PrM). Numbers shown above nodes (red dots) are Euclidean distances calculated between 2 samples (eg, B1-E1, B14-E14, B15-E15, E9-P9) based on the Euclidean distance scale shown on the left. Euclidean distance is one of the metrics used to quantify the difference between 2 points in space, which in our study are each individual tumor tissue samples. The difference between each tumor tissue samples is based on expression data from genes that are common to the sites. The larger the Euclidean distance, the larger the separation between tumor pairs.

Download figure to PowerPoint

image

Figure 3. Kaplan-Meier curves are shown of prognostic histopathologic factors in metastatic melanomas to the brain: (A) immune infiltrate, (B) intratumoral hemorrhage, and (C) immune infiltrate and intratumoral hemorrhage.

Download figure to PowerPoint

image

Figure 4. Immunohistochemical staining of brain tumor sections is shown. Tumor sections were stained with Vulcan Fast Red and counterstained with hematoxylin. (A) Representative images with CD3 stain (20× magnification) show low and high degrees of intratumoral (upper row) or peritumoral-perivascular infiltrate (lower row). (B) Tumor sections were stained with an antibody against the glial fibrillary acidic protein (GFAP). GFAP-positive cells (blue arrows) are not only present in tumor periphery (left image), but can also be found within the tumor (middle and right image).

Download figure to PowerPoint

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SOURCES
  8. CONFLICT OF INTEREST DISCLOSURE
  9. REFERENCES
  10. Supporting Information

Patients with metastatic melanoma live longer, in the current era of effective new therapies.[18, 19] Clinical trials using these agents in patients with MBM have shown efficacy,[3, 4] although the mechanism for their effects in MBM is unclear. Our IHC analysis and GSA of MBM show that a high degree of immune infiltrate and low intratumoral hemorrhage in MBM tumor tissues are favorable prognostic factors for patients who undergo craniotomy for MBM. Equally important, WGEP of MBM demonstrates similarity to EcM, and both expression profiles differ from PrM. The role of stromal components, such as the immune infiltrate or reactive glial (GFAP+) cells interdigitating with melanoma, along with the similarity between EcM and MBM in gene expression suggests that signals from the stroma, and to a lesser extent the metastatic site-specific biology of melanoma cells, play an important role in determining prognosis. The significance of the tumor microenvironment mediating cross-talk with cancer cells has been shown in several other solid tumors.[13]

The prognostic significance of the high CD3+/CD8+ immune infiltrate in MBM is supported by histopathologic, IHC, and WGEP analyses. A previous study in patients with glioblastoma multiforme showed that glia-associated CD8+ infiltrates correlate with long-term survival.[20] However, peritumoral, but not intratumoral, T cells were associated with prolonged OS, as was described.[21] The lack of high numbers of intratumoral as well as activated (eg, CD247+) mononuclear cells may account for the short OS seen in patients with MBM. However, the presence of immune cells provides an opportunity for immunotherapeutic approaches. The lack of association between other mononuclear populations, or expression of various components of the antigen expression machinery, immunosuppressive cytokines (eg, TGFβ), and chemokines by melanoma cells with OS suggests that other yet-to-be-identified mechanisms may be involved.

This study provides new insights regarding the biology of MBM in relation to EcM and PrM. Although several genes overexpressed in PrM include among others, keratins, which may imply contamination, several keratins are indeed expressed by melanoma cells.[22] In addition, a significant number of genes that were previously found differentially expressed between PrM and/or human epidermal melanocytes versus metastatic melanomas were also found in our list of 514 genes (eg, KRT6B/14/16/17, LOR, KLK7, GJB6, PITX1, CST, DSC).[23] In contrast, we were surprised to find that only 23 of the 514 genes on the list were differentially expressed between EcM and MBM. The gene expression similarity between MBM and EcM, which remained significant using all 15,067 probe sets for the clustering analysis, is in line with a recent study regarding the concordance of B-Raf and N-Ras mutations in melanoma metastases[24, 25] as well as the concordance of gene copy number variations between same-patient EcM and MBM.[25] This similarity is clinically relevant, because it explains results from ongoing clinical trials using B-Raf inhibitors in patients with MBM, which show that the response rate of MBMs is similar to slightly lower from that of EcM.[3] The biological similarity and interconnectivity between EcM and MBM may explain the clinical observation that the outcome of patients with MBM is not solely dependent on the effective control of MBM, but also relates to control of EcM since only 50% of patients with MBM die from events directly associated with MBM.

Our study was unable to investigate whether melanoma cells within the brain acquire neuronal cell characteristics, as was described.[26] However, proteins, such as LIS1, which is associated with neuronal migration, and its corresponding BioCarta pathway was associated with shortened OS in patients with MBM in our study (Table 3), was expressed by melanoma cells in MBMs as well as melanoma cell lines (data not shown). Despite the similarities between EcM and MBM, our WGEP analysis identified a number of genes that were differentially expressed between MBM and EcM, such as GFAP. Because GFAP is solely expressed by glial cells, our frequent observation of glial cells interspersed within (but not solely surrounding) melanoma lesions should imply an intimate relation with melanoma cells that may have clinical implications, such as, for example, the glial-mediated protection of melanoma cells from chemotherapeutics[27] or the glial-mediated upregulation of the PI3K pathway in neighboring melanoma cells that have metastasized to the brain.[7, 8] This latter finding may imply that targeted therapies against the PI3K pathway may be relevant in patients with MBM, especially those that escape B-Raf inhibition.[8] Therefore, the role of the glial microenvironment needs to be further characterized because it may provide clues regarding inferior responses of traditional cytotoxics or failures to the brain from treatment with B-Raf inhibitors.[8]

In summary, we present a translational analysis of a patient cohort with MBM, EcM, and PrM tumor biopsies which illuminates the biological factors that are important in established MBM. Although the tumor tissues that were selected and analyzed for histopathology, IHC, and WGEP were derived from patients who underwent craniotomy within a single large academic institution, the findings and conclusions have important clinical implications. First, patients with MBM have variable prognosis, and high levels of immune infiltrate and low levels of hemorrhage identify a patient subgroup with better prognosis. Second, the pursuit of the expanding number of systemic immunotherapies in MBM may be indirectly supported by these findings.

FUNDING SOURCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SOURCES
  8. CONFLICT OF INTEREST DISCLOSURE
  9. REFERENCES
  10. Supporting Information

This study is funded by Samuel and Emma Winters Foundation for Cancer Research (to Dr. Moschos), the UPCI Core Grant P30CA047904 (to Dr. Romkes), and National Institutes of Health grants P01 NS040923-09 (to Dr. Hamilton), P01 CA132714 (to Dr. Reinhart), P50 CA121973 (to Dr. Kirkwood), and R01 CA110249 (to Dr. Ferrone).

CONFLICT OF INTEREST DISCLOSURE

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SOURCES
  8. CONFLICT OF INTEREST DISCLOSURE
  9. REFERENCES
  10. Supporting Information

Dr. Kondziolka has been a consultant for Elekta AB. All other authors made no disclosures.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SOURCES
  8. CONFLICT OF INTEREST DISCLOSURE
  9. REFERENCES
  10. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SOURCES
  8. CONFLICT OF INTEREST DISCLOSURE
  9. REFERENCES
  10. Supporting Information

Additional Supporting Information may be found in the online version of this article.

FilenameFormatSizeDescription
cncr28029-sup-0001.pdf944KSupplementary Data

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.