Soluble FAS ligand as a biomarker of disease recurrence in differentiated thyroid cancer

Authors

  • Taofeek K. Owonikoko MD, PhD,

    Corresponding author
    1. Department of Hematology and Medical Oncology, Winship Cancer Institute, Emory University School of Medicine, Atlanta, Georgia
    • Department of Hematology and Medical Oncology, Emory University, 1365 Clifton Road, NE, Room C3080, Atlanta, GA 30322

    Search for more papers by this author
    • The first two authors contributed equally to this work and should be considered to be co-first author.

    • Fax: (404) 778-5520

  • Mohammad S. Hossain PhD,

    1. Department of Hematology and Medical Oncology, Winship Cancer Institute, Emory University School of Medicine, Atlanta, Georgia
    Search for more papers by this author
    • The first two authors contributed equally to this work and should be considered to be co-first author.

  • Chandar Bhimani MD,

    1. Department of Hematology and Medical Oncology, Winship Cancer Institute, Emory University School of Medicine, Atlanta, Georgia
    Search for more papers by this author
  • Zhengjia Chen PhD,

    1. Department of Biostatistics and Bioinformatics, Winship Cancer Institute, Emory University School of Medicine, Atlanta, Georgia
    Search for more papers by this author
  • Sungjin Kim MS,

    1. Department of Biostatistics and Bioinformatics, Winship Cancer Institute, Emory University School of Medicine, Atlanta, Georgia
    Search for more papers by this author
  • Suresh S. Ramalingam MD,

    1. Department of Hematology and Medical Oncology, Winship Cancer Institute, Emory University School of Medicine, Atlanta, Georgia
    Search for more papers by this author
  • Shi-Yong Sun PhD,

    1. Department of Hematology and Medical Oncology, Winship Cancer Institute, Emory University School of Medicine, Atlanta, Georgia
    Search for more papers by this author
  • Dong M. Shin MD,

    1. Department of Hematology and Medical Oncology, Winship Cancer Institute, Emory University School of Medicine, Atlanta, Georgia
    Search for more papers by this author
  • Edmund K. Waller MD, PhD,

    1. Department of Hematology and Medical Oncology, Winship Cancer Institute, Emory University School of Medicine, Atlanta, Georgia
    Search for more papers by this author
  • Fadlo R. Khuri MD

    1. Department of Hematology and Medical Oncology, Winship Cancer Institute, Emory University School of Medicine, Atlanta, Georgia
    Search for more papers by this author

Abstract

BACKGROUND:

Reliable predictive biomarkers are required to address the challenge of disease recurrence after thyroid cancer surgery. For this study, the authors assessed the association of cellular-based and serum-based immunologic mediators with thyroid cancer recurrence.

METHODS:

Leukocyte subset counts and immune regulatory cytokine levels were determined in peripheral blood samples using multiparameter flow cytometry and 51-panel, multiplex enzyme-linked immunosorbent assays, respectively. The functional activity of circulating B-lymphocytes, T-lymphocytes, and natural killer lymphocytes was assessed ex vivo. Differences in mean biomarker levels between defined patient groups and correlations between biomarkers and cancer recurrence were assessed using t tests or Wilcoxon tests and by univariate and multivariate analyses with Cox models. Optimal cutoff values of significantly correlated biomarkers that best predicted disease recurrence after surgery were established by receiver operating characteristics and were validated by using an optimal cutpoint determination algorithm.

RESULTS:

In total, 35 patients were enrolled (median age, 49.4 year), including 24 women and 15 patients with recurrent disease; and there were 21 individuals in the control group. Patients without recurrence had higher levels of soluble FAS (tumor necrosis receptor superfamily, member 6) ligand (sFASL), transforming growth factor-β, regulatory T cells, and programmed death 1/ programmed death ligand 1-expressing leukocytes. sFASL (hazard ratio, 0.60; 95% confidence interval, 0.38-0.95; P = .031) and interferon-α (hazard ratio, 1.55; 95% confidence interval, 1.03-2.34; P = .038) were associated significantly with disease recurrence. There was a significant difference in progression-free survival between patient groups stratified by an sFASL optimal cutpoint of 15 pg/mL (log-rank P = .0009).

CONCLUSIONS:

sFASL and IFN-α levels were correlated significantly with thyroid cancer recurrence and may be useful for risk-adapted surveillance strategies in patients with thyroid cancer. Cancer 2013. © 2013 American Cancer Society.

INTRODUCTION

Thyroid cancer is the most common endocrine cancer worldwide and is among the fastest growing malignancies in the United States. Greater understanding of thyroid cancer biology is necessary because of the increasing incidence of this cancer in the last 2 decades.1 Although most patients with thyroid cancer have an excellent prognosis with a 10-year survival rate of 80% to 90%, disease recurrence after potentially curative surgery remains a problem that is difficult to predict.2 Approximately 10% to 20% of patients develop multiple recurrences, which carry a worse overall prognosis, especially when associated with distant metastasis.3 Thyroglobulin is a biomarker secreted by the cancer cells but is inadequate for surveillance before cancer recurrence. This is because of significant difficulty with accurate thyroglobulin measurement, especially in patients with interfering cross-reacting antibodies, as well as4 the poor predictive value of a negative assay, which may be associated with up to 10% disease persistence or recurrence on long-term follow-up.5 A biomarker that does not depend on active disease status will facilitate the identification of patients at risk before they develop disease recurrence, thereby providing an early opportunity to alter the course of the disease.

Previous studies demonstrated a strong correlation between the intensity of tumor-infiltrating lymphocytes, especially the cluster of differentiation 8 (CD8)-positive T-cell subset, with reduced risk of recurrence in thyroid cancer.6-8 Furthermore, patients with hyperactive autoimmune thyroid diseases experienced improved cancer-free survival after surgery.9 These observations suggest an important role for tumor-directed immunity in the outcome of this disease. Elucidation of the role of impaired antitumor immunity in thyroid cancer recurrence may lead to the identification of novel monitoring biomarkers and novel therapeutic approaches.

Inactivation of effector lymphoid cells within the tumor microenvironment and in the peripheral circulation is a potential mechanism of immune evasion by cancer cells.10-12 FAS (tumor necrosis receptor superfamily, member 6) or CD95 is a member of the death receptor family of proteins and a type I transmembrane receptor normally expressed on activated lymphocytes that is also expressed aberrantly in nonlymphoid tissues, including cancer cells.13, 14 It forms a receptor-ligand system with its ligand, FASL or CD95 ligand (CD95L), and its activation by this ligand triggers apoptotic cell death, a central regulatory mechanism of the immune system.15 FASL is a membrane-bound protein that is cleaved through the proteolytic action of metalloproteinase enzyme, leading to the release of a soluble nonmembrane-bound fragment (sFASL), which is also capable of receptor binding and triggering apoptotic death of activated T lymphocytes and natural killer (NK) cells.16 This dynamic interaction between FAS and its ligand, whether membrane-bound or soluble, potently modulates the function of activated T cells and is important for the normal regulation of the body immune system.15 Increased expression of both FAS and FASL was observed on tumor cells that were exposed to cytotoxic agents, indicating a role in cell killing induced by these agents.16

We expect that changes in cellular, immune, and inflammatory cytokine mediators within the tumor microenvironment will induce parallel measurable changes in the peripheral blood circulation. This approach to biomarker identification and application for predictive and prognostic assessment in cancer patients is increasingly being adopted as a discovery tool. In a recent study, a multiplex assay of immunologic markers was used along with mass spectrometry as a screening modality in a patient population at high risk for lung cancer. It is noteworthy that sFAS was one of the identified biomarkers that accurately classified cancer patients and distinguished them from those with chronic lung disease.17 Similarly, a comprehensive multiplex assay has been used to identify a panel of serum-based angiogenic and inflammatory biomarkers that predict the clinical benefit of sorafenib and pazopanib in patients with renal cancer.18, 19 Therefore, we hypothesized that proper characterization of changes in the serum-based immunologic and inflammatory biomarker profile may inform a risk-adapted surveillance strategy in thyroid cancer. The current cross-sectional cohort study was conducted to assess the profile of circulating cellular and cytokine mediators of immunologic and inflammatory function in patients with differentiated thyroid cancer.

MATERIALS AND METHODS

Patient Recruitment and Blood Sample Collection

This study was conducted under an institutional review board-approved protocol. Patients who previously underwent curative surgery for papillary or follicular thyroid cancer and had suffered disease relapse or were on surveillance follow-up were enrolled between September 2010 and September 2011. Healthy individuals without prior cancer history were recruited using advertisement fliers. Clinical and demographic data were collected from patients but not from healthy volunteers based on the scope of institutional review board approval. Single-time-point peripheral blood samples collected from each participant were analyzed by flow cytometry within 4 hours of collection. Simultaneously collected samples for cytokine assay were processed by centrifugation at ×1500g, and the resultant plasma and serum were aliquoted and stored at −80°F until ready for analysis in a single batch.

Complete Blood Count and Multiparameter Flow Cytometry

Flow cytometry and immune function assays were performed in the Flow Cytometry Core Laboratory of the Winship Cancer Institute of Emory University (Atlanta, Ga). Complete blood count enumeration was performed immediately (generally within 4 hours of sample collection) using the Coulter AC*T diff (Beckman Coulter, Miami, Fl). The total white blood cell count per mL of blood was used to estimate the absolute numbers of cells in the immune cell subsets with florescence-activated cell-sorting analysis.

Flow cytometry was performed using 200 to 300 μL of erythrocyte-lysed samples. These samples were stained with antibodies against CD3, CD4, CD8, CD69, programmed death-1 (PD-1), cytotoxic T-lymphocyte antigen 4 (CTLA-4), CD64, programmed death ligand-1 (PD-L1), and PD-L2 to enumerate various subsets of B lymphocytes, T lymphocytes, and antigen-presenting cells (APCs). B lymphocytes were further characterized by staining with antibodies against CD3, CD19, CD38, and CD27. NK cells were characterized by a panel of antibodies against CD3, CD4, CD94, NK group 2A (NKG2A), CD16, and CD56; whereas regulatory T cells (T-regs) were identified with a panel of antibodies against CD3, CD4, CD25, and forkhead box P3 (Foxp3) along with isotype control after fixation and permeabilization using commercial antihuman Foxp3 Kit (eBioscience, San Diego, Calif). A complete list of the circulating leukocyte subsets used for this study is provided in Table 1.

Table 1. Cellular Subsets and Functional Assaya
Cell Subsets
  Functional Assay
Standard FACSBasal ExpressionStimulated Expression
  • Abbreviations: +, positive; −, negative; CD3, cluster of differentiation 3 (T-cell coreceptor); CD4, cluster of differentiation 4 (cell-surface glycoprotein); CD8, cluster of differentiation 8 (transmembrane glycoprotein); CD25, cluster of differentiation 25 (type 1 transmembrane protein; α chain of the interleukin-2 receptor); CD40, cluster of differentiation 40 (costimulatory protein present on antigen presenting cells); CD64, cluster of differentiation 64 (integral membrane glycoprotein; Fc receptor); CD94, cluster of differentiation 94 (killer cell lectin-like receptor subfamily D, member 1); CTLA, cytotoxic T-lymphocyte antigen; ENA-78, neutrophil-activating peptide 78; FACS, fluorescence-activated cell sorting; FGF, fibroblast growth factor; Foxp3, forkhead box P3; G-CSF, granulocyte-colony–stimulating factor; GM-CSF, granulocyte-macrophage colony-stimulating factor; Gran-B, granzyme-B; GRO, chemokine (C-X-C motif) ligand 1 (melanoma growth-stimulating activity, α); HGF, hepatocyte growth factor; ICAM-1, intracellular adhesion molecule 1; IFN, interferon; IL, interleukin; IL-12P40, interleukin 12B (natural killer cell stimulatory factor 2, cytotoxic lymphocyte maturation factor 2, p40); IL-12P70, interleukin 12B (natural killer cell stimulatory factor 2, cytotoxic lymphocyte maturation factor 2, p70); IL-1RA, interleukin-1 receptor antagonist; IP-10, interferon γ-inducible protein 10; LIF, leukemia-inhibitory factor; MCP, membrane cofactor protein; M-CSF, macrophage-colony–stimulating factor; MIG, monokine induced by interferon γ; MIP-1, macrophage inflammatory protein-1; NCO, tumor necrosis factor β; NGF, nerve growth factor; NK, natural killer cells; PAI-1, plasminogen activator inhibitor-1; PD-1, programmed death 1; PDGF-BB, platelet-derived growth factor-BB; PD-L1, programmed death ligand 1; RANTES, regulated upon activation, normally T-expressed, and presumably secreted (C-C motif chemokine 5); SCF, stem cell factor; sFASL, soluble FAS (tumor necrosis receptor superfamily, member 6) ligand; TGF, transforming growth factor; TNF, tumor necrosis factor; TRAIL, tumor necrosis factor (ligand) superfamily, member 10; VCAM-1, vascular cell adhesion molecule 1; VEGF, vascular endothelial growth factor; WBC, white blood cells.

  • a

    This is a list of the assays from FACS analysis and cytokines that were included in the Luminex assay (Luminex Corp., Austin, Tex).

WBCPD-1+/CD4+IFN-γ+/CD4+ T cellsIFN-γ+/CD4+ T cells
CD3+PD-1+/CD8+Gran-B+/CD4+ T cellsGran-B+/CD4+ T cells
CD4+PD-L1+/CD64+IFN-γ+/CD8+ T cellsIFN-γ+/CD8+ T cells
CD8+CD3−/CD64+Gran-B+/CD8+ T cellsGran-B+/CD8+ T cells
B cellsNK+/CD94+IFN-γ+ NK cellsGran-B+ NK cells
Plasma cellsCTLA-4+/CD4Gran-B+ NK cells 
NK+CTLA-4+/CD8  
 CD4+/CD25+/Foxp3+  
Cytokines
CD40 ligandIL-1 αIL-15PAI-1
ENA-78IL-1βIL-17APDGF-BB
EotaxinIL-1RAIL-17FRANTES
Basis FGFIL-2IL-20Resistin
G-CSFIL-4IP-10SCF
GM-CSFIL-5LIFsFASL ligand
GROαIL-6LeptinTGF-α
HGFIL-7MCP-1TGF-β
ICAM-1IL-8MCP-3TNF-α
IFN-αIL-10M-CSFTNF-β
IFN-βIL-12P40MIGTRAIL
IFN-γIL-12P70MIP-1αVEGF-A
NGFIL-13MIP-1βVCAM-1

Immunologic Function Assays

Functional activity of B-cell, T-cell, and NK-cell subpopulations was assessed by granzyme-B and interferon-γ (IFN-γ) production with or without phorbol 12-myrisate 13-acetate (PMA)/ionomycin stimulation in vitro using a standard assay method. Briefly, peripheral blood mononuclear cells (PBMCs) isolated by Ficoll gradient centrifugation were stimulated with LeuAct cocktail (PMA, Ca, ionomycin, and Golgi plug; BD Pharmingen, Boston, Mass) for 4 hours at 37°C under 5% CO2 conditions, as recommended by the manufacturer. Thereafter, cells were stained with antibodies against CD3, CD4, CD8, IFN-γ, and granzyme-B. PBMCs treated only with Golgi plug served as an unstimulated control. Appropriate isotype antibodies were included as controls to detect nonspecific staining. Functional NK (CD3-negative/CD16-positive/CD56-positive), CD4-positive, and CD8-positive T cells were estimated by intracellular florescence-activated cell-sorting analysis for the proportion of cells that produced IFN-γ or granzyme-B using Cytofix/Cytoperm Solution Kit (BD Biosciences, San Diego, Calif).

Luminex Assay

The Luminex assay (Luminex Corp., Austin, Tex) was performed in a blinded fashion at the Immunology Core Laboratory at Stanford University (Stanford, Calif). Human 51-plex kits were purchased from Affymetrix Inc. (Santa Clara, Calif) and were used according to the manufacturer's recommendations with modifications as described below. Briefly, samples were mixed with antibody-linked polystyrene beads on 96-well filter plates and incubated at room temperature for 2 hours followed by an overnight incubation at 4°C. The plates were vacuum-filtered and washed twice before a 2-hour incubation with biotinylated detection antibody. Samples were then filtered as described above, washed twice, and resuspended in streptavidin-phycoerythrin. After incubation for 40 minutes at room temperature, 2 additional vacuum washes were performed, and the samples were resuspended in reading buffer. All samples were assayed in a single batch, and each sample was measured in duplicate. Plates were read using a Luminex 200 instrument (Luminex Corp.) with a lower bound of 100 beads per sample per cytokine. A complete list of the Affymetrix human 51-plex cytokines is provided in Table 1.

Statistics

The primary outcome measure was progression-free survival (PFS), which was calculated from the date of initial surgery. Depending on whether the data satisfied the assumption of normal distribution, t tests or Wilcoxon rank-sum tests were used to examine differences in mean biomarker levels between defined subgroups based on disease status and sex. Spearman or Pearson correlation coefficients were estimated to measure the relation between biomarkers and clinical covariates. Significant correlations were established by using a Wald test. The univariate association of each biomarker with PFS for patient subgroups was determined by Cox regression analysis. In multivariable survival analyses, best model selection using the backward selection method started with the biomarkers and clinical covariates that achieved statistical significance (P ≤ .10) in the univariate analyses. Survival curves and an estimate of the survival rate associated with each specific biomarker of interest (stratified by >median vs ≤median) were calculated according to the Kaplan-Meier method and were compared using the log-rank test.

Receiver operating characteristic (ROC) analyses were performed for biomarkers that had a significant association with PFS in multivariate analysis. The optimal biomarker level that best discriminated patients' disease recurrence status was determined by measuring the area under the curve (AUC) of ROC curves, and differences in AUCs of ROC curves were tested with chi-square tests. The cutoff values required to obtain 90% sensitivity and 90% specificity levels in the ROC curves were estimated. To obtain the optimal cutpoints that would achieve the maximum sensitivity and specificity for each biomarker with the best discrimination power for disease recurrence, sensitivity and specificity pairs were obtained in the logistic regression under all possible thresholds. Subsequent validation of the cutoff values estimated by ROC analysis was performed using an outcome-oriented optimal cutpoint determination (OCD) algorithm according to the method described by Contal and O'Quigley.20 PFS curves were generated for the 2 groups stratified by the optimal cutoff value estimated with the ROC and OCD methods; significant differences in PFS were assessed by using log-rank tests. The SAS statistical package (version 9.2; SAS Institute, Inc., Cary, NC) was used for data management and analyses.

RESULTS

Clinical Data Analysis

In total, 35 patients with differentiated thyroid cancer and 21 healthy volunteers were enrolled on this study. The median age of patients was 49 years; 24 patients (69%) were women, and 11 patients (31%) were men. There were 6 patients (17%) with follicular cancer, 28 patients (80%) with papillary cancer, and 1 patient (3%) with poorly differentiated thyroid cancer. Fifteen patients (43%) already had disease recurrence at the time of enrollment. The median time to recurrence after initial surgery for all patients was 5.3 years (1.6 years for patients with recurrence). Three patients had died at the time of data cutoff for analysis at a median follow-up after surgery of 6.7 years (6.2 years vs 8.2 years for patients without and with established recurrence, respectively). None of the healthy volunteers reported any current or prior history of cancer.

Comparison between patients with and without disease recurrence

Univariate analysis of associations between PFS and clinical prognostic factors for thyroid cancer (age, tumor size, biochemical recurrence, locoregional metastasis, and distant metastasis) revealed a trend in the expected direction in the study population (Table 2).

Table 2. Univariate Analysis Based on Clinical Variablesa
CovariateHR95% CILog-Rank P
  • Abbreviations: CI, confidence interval; HR, hazard ratio.

  • a

    These are the results from univariate analysis of the impact of clinical covariates on progression-free survival.

Sex   
 Men2.0680.723-5.913.166
 Women 
Stage   
 I0.1780.011-2.893.133
 II0.8410.092-7.651 
 III1.5320.187-12.524 
 IV 
Tumor classification   
 T1.013
 T22.1730.361-13.064 
 T38.0841.627-40.157 
 T40.0000.000- 
Positive lymph nodes   
 No0.5100.170-1.529.221
 Yes 
Metastasis   
 No0.0000.000< .001
 Yes 
I-131   
 No0.0000.000.145
 Yes 
External-beam radiation   
 No0.3550.109-1.154.072
 Yes 
Anatomic recurrence   
 No0.0000.000< .001
 Yes 
Biochemical recurrence   
 No0.0610.013-0.277<.001
 Yes 
Age1.0270.982-1.0740.239

Biomarker Analysis

Comparison of biomarker levels between patients and healthy control

There were significant differences between patients with thyroid cancer and healthy controls in 17 of 51 cytokines tested. Leptin and intracellular adhesion molecule-1 (ICAM1) exhibited the greatest difference with ≥4-fold increased levels in patients compared with healthy controls (Table 3).

Table 3. Comparison of Biomarker Levels Between Patients With Thyroid Cancer and Healthy Individualsa
CovariatePatientsControlsP
  • Abbreviations: +, positive; CD3, cluster of differentiation 3 (T-cell coreceptor); CD4, cluster of differentiation 4 (cell-surface glycoprotein); CD8, cluster of differentiation 8 (transmembrane glycoprotein); CD40, cluster of differentiation 40 (costimulatory protein present on antigen presenting cells); CD64, cluster of differentiation 64 (integral membrane glycoprotein; Fc receptor); FACS, fluorescence-activated cell sorting; GM-CSF, granulocyte-macrophage colony-stimulating factor; ICAM-1, intracellular adhesion molecule 1; IFN-γ, interferon-γ; IL, interleukin; IL-12P40, interleukin 12B (natural killer cell stimulatory factor 2, cytotoxic lymphocyte maturation factor 2, p40); LIF, leukemia-inhibitory factor; MIG, monokine induced by interferon γ; NGF, nerve growth factor; NK, natural killer cells; PAI-1, plasminogen activator inhibitor-1; PD-1, programmed death 1; PDGF-BB, platelet-derived growth factor-BB; PD-L1, programmed death ligand 1; RANTES, regulated upon activation, normally T-expressed, and presumably secreted (C-C motif chemokine 5); TRAIL, tumor necrosis factor (ligand) superfamily, member 10; VCAM-1, vascular cell adhesion molecule 1; T-regs, regulatory T-cells.

  • a

    Mean blood or plasma levels of biomarkers are compared between healthy individuals and patients with thyroid cancer.

  • b

    Cytokine concentrations are measured in pg/mL.

  • c

    Cellular subsets are absolute cell counts ×106 per mL of blood.

Cytokinesb   
 CD40 ligand63.4443.4.004
 GM-CSF26.5915.35< .001
 ICAM-11,233,564.1114,926.16.006
 IFN-γ3.756.002
 IL-100.861.32.018
 IL-12P4019.2513.9.008
 IL-155.5713.14.010
 IL-511.2214.06.014
 IL-742.8231.79< .001
 Leptin1920.88447.58< .001
 LIF1.742.57.008
 MIG52.3735.41.034
 NGF6.3810.39.031
 PDGFBB237.67147.36.039
 RANTES617.79416.69.010
 TRAIL41.6334.09.009
 VCAM-12446.811747.46.006
FACS and functional assayc   
 PD-1+/CD4+0.30.09< .001
 PD-1+/CD8+0.10.06.028
 PD-L1+/CD64+0.120.04< .001
 Absolute T-regs0.030.01< .001
  % T-regs55.4723.23< .001
  Basal granzyme-B+ NK cells0.470.15.002
 Stimulated IFN-γ+/CD4+ T cells0.250.15.017
 Stimulated IFN-γ+/CD8+ T cells0.420.22.004
 Stimulated granzyme-B+/CD8+ T cells0.380.22.012

A comparison of biomarkers between the 2 patient subgroups revealed higher levels of CD40L, sFASL, and transforming growth factor-β (TGF-β) as well higher numbers of leukocyte subsets with PD-1, PD-L1, or CTLA4 surface expression in patients without recurrence (Table 4). The functional assay reveled a nonsignificant trend toward a higher IFN-γ response to PMA stimulation in patients without disease recurrence (data not shown).

Table 4. Comparison of Biomarkers Between Patients Based on Recurrence Statusa
CovariatebRemissionRecurrenceP
  • Abbreviations: +, positive; CD4, cluster of differentiation 4 (cell-surface glycoprotein); CD8, cluster of differentiation 8 (transmembrane glycoprotein); CD40, cluster of differentiation 40 (costimulatory protein present on antigen presenting cells); CD64, cluster of differentiation 64 (integral membrane glycoprotein; Fc receptor); CTLA-4, cytotoxic T-lymphocyte antigen 4; PD-1, programmed death 1; PD-L1, programmed death ligand 1; sFASL, soluble FAS (tumor necrosis receptor superfamily, member 6) ligand; TGF-β, transforming growth factor-β; T-regs, regulatory T cells.

  • a

    Mean blood or plasma levels of biomarkers were compared between patients with and without thyroid cancer recurrence.

  • b

    Cytokine concentrations are measured in pg/mL; cellular subsets are absolute cell counts ×106 per mL of blood.

CD40 ligand71.4352.78.033
sFASL ligand18.8713.08.011
TGF-β6.614.84.011
PD-1+/CD4+0.470.09< .001
PD-1+/CD8+0.130.06.004
CTLA-4+/CD8+0.020.01.010
PD-L1+/CD64+0.160.07< .001
Absolute T-regs0.030.02.017

Univariate and multivariate analyses of biomarker impact on survival

Granulocyte-colony–stimulating factor, interleukin-12B (NK cell stimulatory factor 2, cytotoxic lymphocyte maturation factor 2, p40) (IL-12P40), plasminogen activator inhibitor-1 (PAI-1), sFASL, TGF-β, PD-1–positive CD4 cells, PD-1-positive CD8 cells, and CD64-positive/PD-L1–positive cell subsets were associated significantly with PFS on univariate analysis; whereas TGF-β, PAI-1, sFASL, IFN-α, and PD-1–positive/CD8-positive cell subsets maintained a significant association with PFS on multivariate analysis. Only sFASL and IFN-α were associated significantly with PFS on multivariate analysis that incorporated both significant biomarkers and clinical variables identified from univariate analyses (Table 2). Consistent with this association, patients who had disease progression, according to PFS measured from the time of original surgery, had a statistically significant lower level of sFASL in circulation than patients without disease recurrence. The differences in median IFN-α levels based on cancer recurrence status of the patients did not reach statistical significance (Table 5, Fig. 1). There was no significant correlation between sFASL and patient age (data not shown).

Figure 1.

This box plot illustrates the differences in levels of soluble FAS (tumor necrosis receptor superfamily, member 6) ligand (sFASL) and interferon-α (IFN-α) between patient subgroups

Table 5. Significant Findings on Univariate and Multivariate Analysesa
CovariateHR95% CILog-Rank P
  • Abbreviations: +, positive; CD4, cluster of differentiation 4 (cell-surface glycoprotein); CD8, cluster of differentiation 8 (transmembrane glycoprotein); CD64, cluster of differentiation 64 (integral membrane glycoprotein; Fc receptor); CI, confidence interval; G-CSF, granulocyte-colony–stimulating factor; HR, hazard ratio; IFN-α, interferon-α; IL-12P40, interleukin 12B (natural killer cell stimulatory factor 2, cytotoxic lymphocyte maturation factor 2, p40); PAI-1, plasminogen activator inhibitor-1; PD-1, programmed death 1; PFS, progression-free survival; sFASL, soluble FAS (tumor necrosis receptor superfamily, member 6) ligand; TGF-β, transforming growth factor-β.

  • a

    These biomarkers achieved statistical significance in univariate and multivariate analyses for an association between biomarkers and PFS.

Univariate analysis for PFS   
 G-CSF0.3950.158-0.987.042
 IL-12P400.8730.785-0.970.013
 PAI-10.9960.992-1.000.033
 sFASL ligand0.8420.717-0.990.033
 TGF-β0.7330.548-0.981.037
 PD-1+ CD40.0000.000-0.007.025
 PD-1+ CD80.0000.000-0.011.009
 CD64+/PD-L1+0.0000.000-0.001< .001
 IFN-α0.8880.772-1.021.088
Multivariate analysis for PFS using only significant biomarkers from the univariate analysis   
 PAI-11.061.00-1.12.061
 sFASL ligand0.010.00-0.94.047
 TGF-β0.010.00-1.05.053
 PD-1+ CD80.000.00-7.76.052
 IFN-α99.581.00-9911.75.050
Multivariate analysis for PFS using both significant biomarkers from univariate analysis and significant clinical covariates   
 sFASL ligand0.600.38-0.95.031
 IFN-α1.551.03-2.34.038

Receiver Operating Characteristic Analysis and Optimal Cutpoint Determination

ROC analysis estimated sFASL level of 15.001 pg/mL and; IFN-α level of 2.7105 pg/mL as the optimal values for stratifying patients to predict recurrence (Fig. 2), whereas the OCD method estimated the optimal values as 15.11 pg/mL for sFASL and 5.17 pg/mL for IFN-α. There was a significant difference in PFS between patient groups separated by the optimal sFASL cutpoint value of 15.11 pg/mL (log-rank P = .0009) (Fig. 3A) but a marginally significant difference with the values for IFN-α (2.7105 pg/mL [log-rank P = .0885] and 5.17pg/mL [P = .0817]) (Fig. 3B,C).

Figure 2.

This is a summary of receiver operating characteristic (ROC) analyses for optimal sensitivity and specificity of soluble FAS (tumor necrosis receptor superfamily, member 6) ligand (sFASL) and interferon-α (IFN-α).

Figure 3.

These are Kaplan-Meier curves of progression-free survival (PFS) for patients with differentiated thyroid cancer. Note that the analyzed data included a single patient who was diagnosed with thyroid cancer in childhood and was without disease recurrence for more than 40 years. (A) PFS is stratified according to the optimal cutoff value of 15.11 pg/mL established for soluble FAS (tumor necrosis receptor superfamily, member 6) ligand (sFASL) using the optimal cutpoint determination (OCD) method. The median PFS was 4.7 years (95% confidence interval [CI], 0.999-7.001) for patients with sFASL levels <15.11 pg/mL versus 46.4 years for patients with sFASL levels ≥15.11 pg/mL. (B) PFS is stratified using the optimal value of interferon-α (IFN-α) established by receiver operating characteristic (ROC) analysis. The median PFS was 6.8 years (95% CI, from 0.594 years to not reached) for patients with IFN-α levels <2.7105 pg/mL versus 46.4 years (95% CI, 4.701-46.379 years) for patients with IFN-α levels ≥2.7105 pg/mL. (C) PFS is stratified using the value established with the OCD method for IFN-α. The median PFS was 6.8 years (95% CI, from 1.227 years to not reached) for patients with IFN-α levels <5.17 pg/mL versus 46.4 years (95% CI, 4.999-46.379 years) for patients with IFN-α levels ≥5.17 pg/mL.

DISCUSSION

The potential role of host-immune function has been demonstrated previously in thyroid cancer outcomes.6-9 To our knowledge, however, the current study is the first to evaluate a comprehensive panel of immune-related biomarkers in patients with thyroid cancer. The comparison between patients and controls revealed higher levels of leptin and ICAM-1 in patients. Our findings are consistent with those of previous studies, which demonstrated significantly higher levels of leptin in patients with thyroid cancer relative to a control group and a significant drop in serum leptin levels after surgical resection of thyroid cancer.21, 22 It is interesting to note that high leptin levels reportedly portend an aggressive disease course in breast and prostate cancer,23-27 and an altered leptin signaling network, in concert with the phosphatidylinositol 3-kinase/protein kinase B (PI3K/Akt) signaling pathway, reportedly plays a role in thyroid cancer development.28, 29 However, we did not establish a significant difference in leptin levels between patients with or without disease recurrence, perhaps because of the small number of patients.

To identify novel immune-related biomarkers that may predict thyroid cancer recurrence, we adopted a stepwise comparison of the biomarkers of interest in patients with and without disease. This stringent approach enabled us to identify the most robust biomarker. We identified a significant association of low sFASL levels and high IFN-α levels with thyroid cancer recurrence. The finding of low sFASL levels in patients with recurrent thyroid cancer is consistent with the work by Hoffman et al, who also observed lower levels of sFASL in patients with active head and neck cancer relative to patients without disease recurrence or normal controls.10 However, other investigators have reported the contrary finding of high sFASL levels correlating with advanced disease.30-33 The biology and function of sFASL remains controversial, and the blood level of sFASL only partly reflects a dynamic balance between its aberrant production by cancer cells and clearance from the circulation through the scavenger role of aberrant FAS receptor. Signaling abnormalities and increased apoptosis of effector lymphocytes occur in patients with cancer, leading to immune evasion by cancer cells.11, 12, 34 This phenomenon was demonstrated previously in patients with lymphoid and gastrointestinal malignancies, but it has not yet been observed in thyroid cancer.35 It results from apoptosis of tumor-infiltrating lymphocytes induced by membrane FASL on tumor cells and of peripheral circulating lymphocytes by sFASL away from the tumor sites.11, 12, 34 Such a model would predict an association between higher sFASL levels and a greater risk of cancer recurrence. Although the biologic impact of sFASL in thyroid cancer remains to be fully elucidated, prior studies demonstrated that FAS and FASL are expressed in thyroid cancer, while normal and neoplastic cells of thyroid lineage are intrinsically resistant to FAS-mediated apoptosis.36 Indeed, homotrimerization of the FAS receptor by an exogenous agonistic antibody induced a paradoxical cellular proliferation rather than apoptosis.37, 38 It is interesting to note that the same thyroid cell lines were susceptible to apoptosis induced by TRAIL, a related member of the TNF superfamily.37, 38 In the alternative, because activated lymphocytes constitute a major source of sFASL in circulation, it is plausible that sFASL in patients with thyroid cancer arises from the proteolytic cleavage of expressed FASL on activated immune effector cells. This model would suggest that, contrary to the observation in some cancer types, patients with thyroid cancer who have high levels of sFASL have activated immune-competent cells and an effective immune surveillance network that delays or prevents disease recurrence. Consistent with this explanation, elevated sFASL was detected in patients with hyperactive immune function, such as hyperthyroid Graves disease and idiopathic myocarditis,39-41 clinical conditions associated with a good prognosis in patients with thyroid cancer.

Our findings set the table for potential important future clinical applications, such as disease monitoring or prognostication after surgery, as well as novel, testable hypotheses in patients with thyroid cancer. Our discovery goal was aided by the ability to interrogate a large panel of serum-based cytokines from a small volume of blood using the Luminex assay platform. Nonetheless, we recognize some important limitations of the current study, including the modest sample size and the finding that we only performed a single-time-point assessment at various periods after surgery. Despite these limitations, our results provide a useful benchmark on which to build much larger prospective studies to further characterize and validate the predictive and prognostic importance of sFASL in thyroid cancer.

Acknowledgements

We thank Anthea Hammond, PhD, for her editorial suggestions and proof reading the article.

FUNDING SOURCES

This work was supported by National Institutes of Health grants 5P50CA128613 (principal investigator, Dong M. Shin, MD), 1K23CA164015 (principal investigator, Taofeek K. Owonikoko, MD, PhD), and P01 CA116676 (principal investigator, Fadlo Khuri, MD) and by a Georgia Cancer Coalition Distinguished Cancer Scholar Award (principal investigator, Taofeek K. Owonikoko, MD, PhD).

CONFLICT OF INTEREST DISCLOSURES

The potential use of sFASL as a biomarker in thyroid cancer patients is covered by a provisional patent held by Emory University.

Ancillary