Epithelial ovarian cancer (EOC) is one of the most deadly carcinomas in females. Immune systems can recognize EOCs; however, a defect of human leukocyte antigen (HLA) class I expression is known to be a major mechanism for escape from immune systems, resulting in poor prognosis. The purpose of this study is to identify novel correlations between immunologic responses and other clinical factors. We investigated the expression of immunologic components in 122 cases of EOCs for which surgical operations were performed between 2001 and 2011. We immunohistochemically stained EOC specimens using an anti-pan HLA class I monoclonal antibody (EMR8-5) and anti-CD3, -CD4, and -CD8 antibodies, and we analyzed correlations between immunologic parameters and clinical factors. In multivariate analysis that used the Cox proportional hazards model, independent prognostic factors for overall survival in advanced EOCs included low expression level of HLA class I [risk ratio (RR), 1.97; 95% confidence interval (CI), 1.01–3.83; P = 0.046] and loss of intraepithelial cytotoxic T lymphocyte (CTL) infiltration (RR, 2.11; 95% CI, 1.06–4.20; P = 0.033). Interestingly, almost all platinum-resistant cases showed a significantly low rate of intraepithelial CTL infiltration in the χ2 test (positive vs. negative: 9.0% vs. 97.7%; P < 0.001). Results from a logistic regression model revealed that low CTL infiltration rate was an independent factor of platinum resistance in multivariate analysis (OR, 3.77; 95% CI, 1.08–13.12; P = 0.037). Platinum-resistant EOCs show poor immunologic responses. The immune escape system of EOCs may be one of the mechanisms of platinum resistance. Cancer Immunol Res; 2(12); 1220–9. ©2014 AACR.
Epithelial ovarian cancer (EOC) is the sixth most common female cancer worldwide and is the leading cause of death in gynecologic malignancies. EOC was comparatively uncommon in Japan, but the number of patients with EOC has been increasing, according to the rise of average body mass index (BMI; ref. 1). EOC is difficult to diagnose and most cases are found in an advanced stage often with peritoneal dissemination or distant metastasis. Prognosis has been improved by platinum-based chemotherapy, which is used as a first-line chemotherapy (2); however, patients who are resistant to platinum agents have a very poor prognosis with a median survival period of only 6 months (3). Several mechanisms have been reported to be involved in platinum resistance, including intracellular drug accumulation and/or increased drug efflux, drug inactivation by increased levels of cellular thiols, alterations in drug target, processing of drug-induced damage by increased nucleotide excision-repair activity, and decreased mismatch-repair activity and evasion apoptosis (4–6). Despite the revelation of mechanisms, it is still not clear how platinum resistance in EOC can be overcome.
An interesting study revealed that intratumoral infiltration of CD3+ T cells was a prognostic factor for patients with EOC (7). Results from that study suggested that immunologic reaction for EOC has a significant impact on disease control, and cancer immunotherapy has emerged as another treatment modality for EOCs. CD8+ cytotoxic T lymphocytes (CTL) recognize an antigenic peptide presented by human leukocyte antigen (HLA) class I. Various immunotherapy trials for EOC have been conducted and reviewed by Sabbatini and Odunsi (8). Prognosis of patients with clear-cell adenocarcinomas was improved by EOC immunotherapy using grypcan-3–derived peptide (9). Furthermore, a phase II trial of anti–PD-1 antibody therapy is ongoing since 2011 (10, 11). However, there are several mechanisms by which cancer cells escape detection and destruction by the immune system (12–14). HLA class I loss, a condition in which CTLs cannot recognize cancer cells, is a major mechanism of immune escape. Downregulation or loss of HLA class I molecules in EOC has been reported (15–18), and the mechanism of immune escape is a critical factor in immunotherapy.
Recently, we have generated a monoclonal antibody (mAb) against HLA class I molecules (clone: EMR8-5; ref. 19). EMR8-5 is reactive for HLA-A, HLA-B, and HLA-Cw, whereas clone HC10, a commonly used HLA class I antibody, is reactive for only HLA-B and HLA-Cw (19). Therefore, EMR8-5 is more suitable for evaluation of HLA class I expression. In this study, we examined the expression profiles of HLA class I molecules immunohistochemically using clone EMR8-5 in representative subtypes of EOC, and we analyzed their correlations with infiltration of intratumoral CD3+, CD4+, and CD8+ T cells and various clinical characteristics.
Materials and Methods
Patients and specimens
Surgical specimens were obtained from 122 patients with primary EOC treated at Sapporo Medical University Hospital (Sapporo, Japan) during the period from 2001 to 2011. Written informed consent was obtained from each patient according to the guidelines of the Declaration of Helsinki. Patients underwent abdominal hysterectomy, bilateral salpingo-oophorectomy, omentum resection, lymphadenectomy, and resection of metastatic lesion when possible. All H&E-stained slides were reviewed by a pathologist, and the diagnosis was confirmed in accordance with FIGO (International Federation of Gynecology and Obstetrics) stage. Early cases were defined as stages I and II, and advanced cases were defined as stages III and IV. In advanced cases, platinum-based combination agents were administered as adjuvant chemotherapy. Platinum-resistant cases were defined as those with disease progression during treatment with first-line chemotherapy or relapse within 6 months after completion of chemotherapy. Optimally resected cases were defined as cases with complete tumor resection or with residual tumor of less than 1 cm in diameter. Overall survival was documented for all patients, and survival was calculated from the day of the operation until November 31, 2013.
Sections (5 μm in thickness) of formalin-fixed paraffin-embedded tumors were immunostained using mAbs after epitope retrieval by Novocastra epitope retrieval solution pH9. mAb EMR8-5 was used to stain HLA class I molecules (19). For staining of T lymphocytes, we used anti-CD3 (Nichirei; no. 413591), anti-CD4 (Nichirei; no. 413951), anti-CD8 (Dako; no. N1591), and anti-FOXP3 (Abcam; clone SP97) mAbs. EMR8-5 was diluted 1×2,000, anti-FOXP3 antibody was diluted 1×100, and anti-CD3, -CD4, and -CD8 antibodies were already established in working dilution. Subsequent incubation with a secondary biotinylated antibody was performed, and endogenous peroxidase activity was blocked by immersion in 3% peroxidase. Slides were then counterstained with hematoxylin, rinsed, dehydrated through graded alcohols into nonaqueous solution, and coverslipped with mounting medium.
Quantification of HLA class I staining
Membrane immune reactivity levels for HLA class I were categorized as 0, +1, +2, and +3 in accordance with criteria determined by the HLA expression evaluating consortium (Fig. 1A). A score of zero was defined as <10% membrane staining. A score of +1 was defined as 10% to 50% membrane staining or 10% to 90% of the membrane stained weakly. A score of +2 was 50% to 90% membrane staining or >90% of the membrane stained weakly, and a score of +3 was defined as membrane staining in >90% of the tumor cells. Finally, the quantified HLA class I levels were divided into two groups: scores 0 and +1 as HLA class I low group, and scores +2 and +3 as HLA class I high group.
Quantification of intraepithelial T cells
We counted intraepithelial infiltrated CD3+, CD4+, and CD8+ T cells in high-power fields (HPF; ×400) and calculated their averages. On the basis of the average counts of infiltrated lymphocytes, we classified them into two groups: T-cell infiltration–positive group with counts/HPF and T-cell infiltration–negative group with <10 counts/HPF (Fig. 1B).
Statistical analyses were performed with SPSS (version 11 for Windows; SPSS Inc.), and GraphPad Prism (version 4.0 for Windows; GraphPad Software Inc.) was used for plotting Kaplan–Meier curves. Pearson χ2 tests were used to determine the significance of associations between characteristic variables. The Spearman rank correlation coefficient was used to evaluate correlations between HLA class I expression and T-cell infiltration. Survival rates were calculated using the Kaplan–Meier method, and differences between groups were tested using the log-rank test. The Cox proportional hazards model was used for multivariate analysis to determine risk ratio and independent significance of individual factors for prognosis. A logistic regression model was used for multivariate analysis to predict odds ratio of individual factors for platinum resistance. Each multivariate analysis was performed with the stepwise method. In all analyses, P values of <0.05 were considered as statistically significant.
Differences in immunologic parameters
The clinicopathologic characteristics of the patients are summarized in Table 1. There were significant differences in HLA class I expression and intraepithelial T-cell infiltration among the histologic subtypes. The clear-cell adenocarcinoma histologic subtype is significantly correlated with lower expression level of HLA class I molecules (HLA class I low; P = 0.007) and lower infiltration rates of CD3+ T cells (CD3+ T-cell negative; P = 0.002) and CD8+ T cells (CD8+ T-cell negative; P < 0.001). The serous adenocarcinoma and endometrioid adenocarcinoma histologic subtypes showed significant correlations with higher infiltration rate of T cells. Interestingly, most CD8+ T cell–negative cases showed a significant correlation with platinum resistance (97.7%; P < 0.001). As shown in Fig. 1C, HLA class I expression and intratumoral T-cell infiltration showed a positive correlation, but it was very weak. CD3+ and CD8+ T-cell infiltrations showed a stronger correlation than did CD4+ T-cell infiltration (ρ = 0.34, 0.34; CD3+, 0.34; CD8+ vs. 0.22; CD4+).
Factors correlated with poor survival rates
Log-rank analyses were performed according to FIGO stages and histologic subtypes (Fig. 2). These analyses revealed that low HLA class I expression level was correlated with poorer prognosis in total cases (P = 0.004), advanced cases (P = 0.032), and early cases (P = 0.035). Low HLA class I was correlated with poorer prognosis in serous adenocarcinoma cases (P = 0.045), endometrioid adenocarcinoma cases (P = 0.039), and mucinous adenocarcinoma cases (P = 0.025), but not in clear-cell adenocarcinoma cases (P = 0.41). CD3+ T-cell positivity was not correlated with prognosis in total cases, advanced cases, early cases, and also in all histologic subtype cases. CD4+ T-cell positivity was correlated with better prognosis in total cases (P = 0.033), but not in advanced cases (P = 0.094) or early cases (P = 0.252). CD4+ T-cell positivity was not correlated with any histologic subtypes (serous, P = 0.799; clear cell, P = 0.206; endometrioid, P = 0.065; mucinous, P = 0.655). Because all CD4+ T cell–positive cases with clear-cell, endometrioid, and mucinous subtypes were alive during that time, we reanalyzed the prognosis by nonserous (clear-cell, endometrioid, and mucinous) subcategory, and found that CD4+ T-cell positivity was correlated with better prognosis in nonserous histologic subtype (P = 0.017), but not in serous subtype (P = 0.799; Fig. 3A). CD8+ T-cell positivity was not correlated with better prognosis in total cases (P = 0.077) or early cases (P = 0.122), but it was correlated with better prognosis in advanced cases (P = 0.004). CD8+ T-cell positivity was not correlated with better prognosis all in histologic subtypes; however, it was correlated with better prognosis in advanced clear-cell adenocarcinoma cases (P = 0.009; Fig. 3B). A previous study had suggested that the presence of regulatory T cells (Treg) was correlated with poorer prognosis in ovarian cancer cases (20); thus, we further investigated the CD4+ T cell–positive cases using anti-FOXP3 antibody (Supplementary Fig. S1). High Treg infiltration (Treg high) was not correlated with the prognoses (P = 0.051); however, high CD8+ T cell:Treg ratio showed tendency to be associated with better prognosis (P = 0.089). Interestingly, patients with high CD8+ T cells:Treg ratios were alive.
To assess whether the immunologic status was an independent marker of prognosis, the relative influence of its expression and other clinical characteristic variables were analyzed by multivariate analysis using the Cox proportional hazard model (Table 2). The cases were divided into three groups as shown in Table 2: total cases, early cases (FIGO stage I/II), and advanced cases (FIGO stage III/IV). HLA class I low was a significant prognostic factor [risk ratio (RR), 2.18; 95% confidence interval (CI), 1.16–4.12; P = 0.016] in total cases. In early-stage cases, factor of final model was restricted only by HLA class I expression, and it did not reach statistical significance in the univariate Cox proportional hazard model (RR, 7.86; 95% CI, 0.82–76.13; P = 0.074). However, using log-rank analysis, the early-stage cases also reached statistically significant difference (P = 0.035). HLA class I low (RR, 1.97; 95% CI, 1.01–3.83; P = 0.046) and CD8+ T-cell positivity (RR, 2.11; 95% CI, 1.06–4.20; P = 0.033) were significant prognosis factors in advanced cases. Clear-cell adenocarcinoma histologic subtype was an independent prognostic factor in advanced cases (RR, 2.38; 95% CI, 1.03–5.48; P = 0.042).
Factors correlating with platinum resistance
A multivariate logistic regression model was used for further analysis. Cases used for the analysis were restricted to advanced cases with adjuvant platinum-based chemotherapy. Those cases that achieved optimal debulking were excluded. As shown in Table 3, CD8+ T-cell positivity was correlated with platinum resistance in the study population independent of other clinical characteristic factors (OR, 3.77; 95% CI, 1.08–13.12; P = 0.046). Platinum resistance was not correlated with HLA class I expression in the primary tumor site. Because cancer-specific CTLs are primed at regional lymph nodes, we performed immunohistochemical staining of lymph node specimens in lymph node metastasis–positive cases. The expression of HLA class I molecules in lymph nodes was not correlated with the prognosis or platinum resistance (Supplementary Fig. S2 and Supplementary Table S1). Because the differences in chemotherapeutic regimens may have affected these results, we investigated the chemotherapeutic regimens in detail. Most cases (91 of 97 cases) were treated with carboplatin and taxane (paclitaxel, n = 69; docetaxel, n = 22). There is no difference in TC (paclitaxel and carboplatin) and DC (docetaxel and carboplatin) regimens (Supplementary Fig. S3 and Supplementary Table S2).
Endometrioid histologic subtype showed a significant correlation with good response to platinum agents (OR, 0.22; 95% CI, 0.05–0.97; P = 0.046). Clear-cell adenocarcinoma, known to be a platinum-resistant subtype, was excluded from the final model by the stepwise method. This result might have been caused by an insufficient number of advanced clear-cell adenocarcinoma cases.
CTLs recognize 8 to 10 amino acid peptide fragments presented by HLA molecules, and play essential role in tumor eradication. However, several mechanisms have been described for human tumor cells that escape from detection and/or destruction by CTLs. (12) Downregulation of HLA class I molecules is one of the major mechanisms that enable tumor cell escape from CTLs, and it is found frequently in solid tumors, including malignant melanoma, breast cancer, stomach cancer, colon cancer, and bladder cancer. The establishment of a novel anti-pan HLA class I mAb (EMR8-5) revealed correlations between HLA class I expression and clinical factors. (21–26).
In this study, we have characterized HLA class I expression as a fine prognosis marker in EOC cases. Several previous reports have shown a correlation between HLA class I downregulation and poor prognosis in EOC. Vitale and colleagues (15) immunostained 51 cases of EOC using anti-HLA class I (HC-10), and anti-TAP1 and -TAP2 antibodies, but there were no statistical correlations between these components and prognosis. Rolland and colleagues (16) report combination staining using HC-10 and β2-microglobulin antibody for 339 cases EOC and suggest that the HC-10+/β2-m+ phenotype is a significant independent prognostic factor. Han and colleagues (17) immunostained components of the antigen-processing machinery (APM), including TAP1, TAP2, tapasin, HLA class I (HC-10), β2-microglobulin, CD3+ T cells, and CD8+ T cells. They report that the downregulation of APM components, a lack of intratumoral T-cell infiltrates, and the suboptimal cytoreduction were independent prognostic factors in multivariate analysis. In all previous studies, only the staining of HLA class I heavy chain with HC-10 was not an independent prognostic marker of EOC. CTLs recognize HLA class I molecules, including HLA-A, HLA-B, and HLA-Cw, and most antigenic peptides are presented by HLA-A (27). However, HC-10 can recognize only HLA-B and HLA-Cw and not HLA-A; EMR8-5 recognizes HLA-A as well as HLA-B and HLA-Cw (19). Aptsiauri and colleagues (28) categorized HLA loss in human tumor cells into seven phenotypes: total loss, haplotype loss, allelic loss, compound loss unresponsiveness to IFNγ, and aberrant expression of HLA-E with low expression of HLA class I. HC-10 detects total loss and compound phenotypes that lack HLA-B and HLA-Cw as HLA-negative cases, whereas EMR8-5 detects only total loss as HLA-negative cases. Therefore, the differences between our results using EMR8-5 and previous studies using HC-10 might depend on the detection of HLA-A. Our results suggest that EMR8-5 is a better mAb for the evaluation of HLA class I molecules. In addition to HLA class I molecules, APM machinery was evaluated in several previous studies. Tumor-associated antigens or misfolded defective ribosomal products (DRiP) are degraded into polypeptide fragments by proteasomes in the cytosol. The polypeptide fragments are translocated into the endoplasmic reticulum (ER) by transporters, TAP1 and TAP2, and loaded onto HLA class I molecules by peptide-loading complexes that are composed from Tapasin, ERp57, and calreticulin. The ER aminopeptidase-1 (ERAP1) generates proper-length peptides (29) for presentation by the HLA class I complex. A previous study reported that downregulation of APM was a poorer prognostic factor (17). A lack of APM molecules may lead to reduced expression of MHC class I molecules on the cell surface because there are less antigenic peptide produced for presentation (29). Thus, the reduced HLA class I expression might result from a lack of APM molecule, and is correlated with poorer prognosis. In addition, a lack of some APM molecules has been reported to affect the repertoire of antigenic peptides (30, 31), which, in turn, might be related to decreased recognition by CTLs, resulting in poorer prognosis.
In our study, downregulation of HLA class I was a significant prognostic factor in each disease stage, but infiltration of CD8+ T cells showed a significant difference only in advanced stages of EOC. The latter may be caused by the differing degree of tumor-debulking achievement by surgery. Complete resection of a tumor can be achieved easily when performed at early stages compared with advanced stages. Even if complete resection is achieved macroscopically, the possibility of microscopic metastasis may be markedly different between EOC of early and advanced stages. Therefore, in early stages of EOC, influence of the immune-escape mechanism may be less than that in advanced stages. Moreover, advanced EOC that showed antigen-specific T-cell immunity preoperatively may have an advantage over those with poor immune potential. In some cases, we observed dissociation of T-cell infiltration and HLA class I expression. Because T cells need to be activated to infiltrate, the activation status of T cells might be one mechanism usurped by tumor cells for immune escape. As described previously, tumor cells produce immune-suppressive factors, including VEGFA, TGFβ, and IL10, to suppress the maturation of DCs, which is associated with failure of T-cell activation and abrogation of T-cell infiltration (32). Disruption of T-cell homing is another possible mechanism of tumor cell immune escape. Chemokines, including CCL2, play essential roles in tumor infiltration (33). Thus, the deregulation of chemokine expression and disruption of chemokine signaling might also prevent T-cell infiltration.
Zhang and colleagues (7) report that intratumoral CD3+ T-cell infiltration improved the survival of patients with EOC. A significant correlation between CD3+ T cells and prognosis was not found in our study (RR, 0.52; 95% CI, 0.44–15.20; P = 0.52; Table 2). Sato and colleagues (20) also report that CD3+ T-cell infiltration was not a significant prognostic factor. In the article, they discussed the potential causes of contrary results, and showed that one possibility for the difference may result from the different chemotherapeutic regimens used in the studies. In the study by Zhang and colleagues, patients were treated with a combination of platinum and/or cyclophosphamide and/or doxorubicin between 1991 and 1995, and they were treated with platinum plus paclitaxel between 1995 and 1999 as were patients in this study. All first-line adjuvant chemotherapy was a combination of taxane and platinum agents in our study. Recently, the most frequently used and effective chemotherapeutic regimens for EOC are shifting from the conventional paclitaxel and carboplatin regimen to a set of dose-dense regimens (34), which controls platinum resistance through a novel immune mechanism (35). The authors detected therapeutic effect of the dose-dense chemotherapy that relied on the preservation of treatment-mediated promotion of tumor-specific immunity. Therefore, novel analyses may be needed for novel regimens.
Tregs are characterized as CD4+CD25+FOXP3+ T cells that suppress the immune system and the antitumoral immune response of CD8+ T cells (36, 37). Sato and colleagues (20) report that only CD4+ T-cell infiltration was not a poor prognostic factor of EOC but that the CD8:CD4 ratio and CD8:Treg (CD25+FOXP3+) ratio were significant prognostic factors. In this study, there was no significant difference between the CD4+ T cell–positive and –negative groups with serous histology as reported by Sato and colleagues. However, in the nonserous histologic subtype, CD4+ T-cell positivity was a significant factor of good prognosis in the log-rank test, and all patients in the CD4+ T cell–positive group were alive in the observation period. Furthermore, the high CD8+:Treg ratio showed tendency to be associated with better prognosis than that of low CD8+:Treg ratio cases (P = 0.089). The result did not reach statistically significance, but the data suggest the importance of CD8+ T cell:Treg ratio, and our results support those of the previous study.
Shehata and colleagues (18) analyzed the correlation of HLA class I expression and platinum resistance. They showed that HLA class I expression was not a prognostic factor in the platinum-sensitive group but was a prognostic factor in the platinum-resistant group. We performed the same analysis with our data, and found that HLA class I low expression level was not a significant prognostic factor in platinum-resistant cases. In multivariate analysis using a logistic regression model, intraepithelial CD8+ T-cell infiltration was an independent risk factor of platinum resistance in our cases. Endometrioid adenocarcinoma showed an approximately 2-fold lower risk for platinum resistance than other histologic subtypes. Clear-cell adenocarcinoma, which is generally a platinum-resistant subtype, showed no significant difference in this analysis. This may be due to an insufficient number of cases. Clinically, some chemotherapeutic agents show the possibility of overcoming platinum resistance. Gemcitabine, which is commonly used in second-line to third-line chemotherapy for EOC, has been shown to decrease DNA repair in vitro (38). Safra and colleagues (39) report that response rates to the combination of gemcitabine and carboplatin were the same in platinum-sensitive patients and platinum-resistant patients with recurrent EOC (43.2% vs. 39.1%). Bevacizumab, the first reported effective molecular target agent of ovarian cancer, is expected to be effective for platinum-resistant EOC, and several studies using this agent have been conducted or are ongoing (40). However, its effect is controversial at this time. Our data showed the possibility of a correlation between platinum resistance mechanisms and immune-escape systems. The molecular mechanisms are still elusive; however, there are several hypotheses. Because platinum causes cancer cell death by either apoptosis or necrosis, antigen-presenting cells might acquire tumor-associated antigens expressed in dead cancer cells, resulting in an antitumor immune response. Thus, the clinical response to platinum might be partially due to a secondary immune response, and good immunologic responders might be related to platinum-sensitive cases. Further immunologic analyses are needed.
In summary, our data showed that HLA class I expression detected by EMR8-5 might be a good prognostic marker of EOC and that CD8+ T-cell infiltration in tumors might be a predictive marker of platinum treatment.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: Y. Hirohashi, T. Torigoe, T. Asano, T. Kuroda, T. Saito, N. Sato
Development of methodology: T. Torigoe, T. Asano, T. Kuroda, T. Saito
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T. Mariya, T. Torigoe, T. Kuroda, T. Saito
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T. Mariya, Y. Hirohashi, T. Torigoe, T. Kuroda, K. Yasuda, T. Sonoda, T. Saito
Writing, review, and/or revision of the manuscript: T. Mariya, Y. Hirohashi, T. Saito
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T. Saito
Study supervision: M. Mizuuchi, T. Saito, N. Sato
This study was supported by a Grant-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology of Japan (to N. Sato), program for developing the supporting system for upgrading education and research from the Ministry of Education, Culture, Sports, Science and Technology of Japan (to N. Sato) and Takeda Science Foundation (to Y. Hirohashi), Sagawa Foundation for Promotion of Cancer Research (to Y. Hirohashi), Suharakinenzaidan Co., Ltd. (to Y. Hirohashi) and Kobayashi foundation for cancer research (to Y. Hirohashi). This study was supported, in part, by Grants-in-Aid for Regional R&D Proposal-Based Program from Northern Advancement Center for Science & Technology of Hokkaido Japan (to Y. Hirohashi and T. Torigoe).
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.
Note: Supplementary data for this article are available at Cancer Immunology Research Online (http://cancerimmunolres.aacrjournals.org/).
- Received May 22, 2014.
- Revision received September 1, 2014.
- Accepted September 23, 2014.
- ©2014 American Association for Cancer Research.