Predictors of Response to Autologous Dendritic Cell Therapy in Glioblastoma Multiforme

Background Glioblastoma (GBM) is the most common and lethal primary malignant glioma in adults. Dendritic cell (DC) vaccines have demonstrated promising results in GBM clinical trials. However, some patients do not respond well to DC therapy, with survival rates similar to those of conventional therapy. We retrospectively analyzed clinical and laboratory data to evaluate the factors affecting vaccine treatment. Methods Forty-seven patients with de novo GBM were enrolled at China Medical University Hospital between 2005 and 2010 and divided into two subgroups. One subgroup of 27 patients received postsurgical adjuvant immunotherapy with autologous dendritic cell/tumor antigen vaccine (ADCTA) in conjunction with conventional treatment of concomitant chemoradiotherapy (CCRT) with temozolomide. The other 20 patients received only postsurgical conventional treatment without immunotherapy. Immunohistochemistry for CD45, CD4, CD8, programed death ligand 1 (PD-L1), and programed death 1 (PD-1) was performed on sections of surgical tumor specimens and peripheral blood mononuclear cells (PBMCs). Pearson’s correlation, Cox proportional hazard model, and Kaplan–Meier analyses were performed to examine the correlations between the prognostic factors and survival rates. Results Younger age (<57 years), gross total resection, and CCRT and PD-1+ lymphocyte counts were significant prognostic factors of overall survival (OS) and progression-free survival (PFS) in the ADCTA group. Sex, CD45+ lymphocyte count, CD4+ or CD8+ lymphocyte count, tumor PD-L1 expression, isocitrate dehydrogenase 1 mutation, and O6 methylguanine-DNA methyltransferase promoter methylation status were not significant factors in both groups. In the ADCTA group, patients with tumor-infiltrating lymphocytes (TILs) with a lower PD-1+/CD8+ ratio (≤0.21) had longer OS and PFS (median OS 60.97 months, P < 0.001 and PFS 11.2 months, P < 0.008) compared to those with higher PD-1+/CD8+ ratio (>0.21) (median OS 20.07 months, P < 0.001 and PFS 4.43 months, P < 0.008). Similar results were observed in patients’ PBMCs; lymphocyte counts with lower PD-1+/CD8+ ratio (≤0.197) had longer OS and PFS. There was a significant correlation of PD-1+/CD8+ ratio between TILs and PBMCs (Pearson’s correlation R2 = 0.6002, P < 0.001). By contrast, CD4−, CD8−, but PD-1+, CD45+ tumor-infiltrating lymphocytes have no impact on OS and PFS (P = 0.073 and P = 0.249, respectively). Conclusion For patients receiving DC vaccine adjuvant therapy, better outcomes are predicted in patients with younger age, with TILs or PBMCs with lower PD-1+/CD8+ ratio, with gross tumor resection, and receiving CCRT.

inTrODUcTiOn Glioma is a commonly occurring form of brain tumor, and highgrade gliomas are the most common malignant tumors of the central nervous system (1,2). Glioblastoma multiforme (GBM) is the most lethal, with a mortality rate of 88% within 3 years (3). With standard treatments, GBM prognosis remains poor, with a median overall survival (OS) of 14.6 months for newly diagnosed GBM treated with temozolomide (TMZ) and a median OS of 7.4 months for recurrent GBM (4,5).
Even with advances in surgical procedures, radio-therapeutic technologies, and discovery of new chemotherapeutic agents such as bevacizumab (Avastin) (6,7), patients with GBM still have a dismal prognosis, which is approximately 19 months for median OS (8), although some reports showed 2-year survival rates of 40-50% (9,10) to date. Survival rates of patients with glioblastoma are relatively low, and the failure of treatments is mostly due to recurrence. GBM has a low survival rate attributed to unique treatment limitations such as a high cell proliferation, invasive infiltration, tumor location, and poor understanding of the tumor pathophysiology (11).
In recent years, cancer immunotherapy has been pursued by exploitation of dendritic cells (DCs), which are "professional" antigen processing and presenting cells utilized to induce specific antitumor responses (12). DCs are very potent antigen-presenting cells that play a key role in the initiation of the immune response, and are considered a promising tool for immunotherapy (13,14). DC-based therapy may provide a way for cytotoxic T lymphocytes (CTLs), natural killer cells, and cytokines to directly or indirectly kill tumor cells. Active immunotherapy using DCs induces an antigen-specific T-cell response to tumor antigens, and recent reports have shown the feasibility, safety, and bioactivity of autologous DC vaccines for GBM, even with recurrent tumors (15,16).
Clinical trials for GBM with DC vaccine treatment may increase OS time from control group of 13.1 months to DC vaccine group of 15.7 to 35.9 months, according to different trials (12). Since 2001, various research groups have attempted to use DC-based immunotherapy (also called DC vaccines in some studies) in treatment of malignant gliomas (14,(17)(18)(19), and have reported induction of glioma-specific antitumor immune responses and apparent survival benefits for some patients. The median OS in trials recruiting patients with newly diagnosed GBM varied between 16.0 and 38.4 months, whereas for recurrent GBM, it ranged between 9.6 and 35.9 months (20)(21)(22).
To evaluate the efficacy of DC treatment, two investigatorinitiated trials have been conducted in Taiwan, and the results have been published as two peer-review journal articles (23,24). In Cho's series, the 2-year survival rate was 47.2%, and the 3-year survival rate was 26.6%.
Similar to other studies, the results of our early experience have revealed a significant benefit for patients with high-grade glioblastoma; however, survival in DC vaccine-treated GBM patients varied (25). Recently, studies of immune checkpoints have provided many dramatic breakthroughs in tumor immune therapy, and checkpoint blockade has shown effectiveness in lung and breast tumors or others (26)(27)(28)(29)(30)(31)(32)(33). The tumor microenvironment is complex. Immune checkpoint inhibitors such as programed death 1 protein (PD-1) and programmed death ligand 1 (PD-L1) tremendously influence the immune thera peutic outcome in many types of tumor including GBM (34)(35)(36)(37). Many clinical trials of checkpoint blockade are under way (35,38,39). Studies in various cancer models have suggested that immunological checkpoint mechanisms such as the PD-1/PD-L1 pathway may contribute to self-tolerance and induce CD8 + T-cell exhaustion in the tumor microenvironment (28,(40)(41)(42)(43). Recently, the concept of PD-1/PD-L1 in regulating the response to glioma with DC vaccine treatment was also evaluated and proved by preclinical evidence in vitro and in vivo in animal studies (44,45). Therefore, we retrospectively analyzed clinical data and paraffin blocks from our previous study for improving the effectiveness of autologous DC treatment of GBM.

Patients
This was a retrospective review of 47 samples from patients in a previous clinical study (24) between November 2005 and FigUre 1 | Treatment schema and vaccine preparation. Clinical schematic diagram (a). Subjects with primary GBM will be consent for operation and concomitant chemoradiotherapy (CCRT). Subjects assigned to the ADCTA group will be designated to receive dendritic cell (DC) vaccination ten times following the clinical trial schedule after operation. V: visits to hospital, numbers following indicates times of visit. DC vaccine manufacturing protocol (B). In China Medical University Hospital, the DC vaccine is produced in laboratories that meet the requirements of Good Tissue Practices and Good Manufacturing Practices. The final product is used in a clinical trial of autologous DC therapy for GBM patients between years 2005 and 2010.
April 2010 following a new diagnosis of histologically confirmed glioblastoma multiforme (GBM, WHO grade 4 astrocytoma). Patients were between the ages of 14 and 70 years at diagnosis. Inclusion criteria included a Karnofsky performance score (KPS) of at least 70 before surgery and adequate hematologic, renal, and hepatic function [hemoglobin ≥8 g/dL, platelets, ≥100,000/μL, white blood cell count >2,000/μL, absolute neutrophil count ≥1,000/μL, serum blood urea nitrogen <25 mg/dL, serum creatinine <1.8 mg/dL, creatinine clearance >50 mL/min, both serum ALT and serum AST ≤ 3 × the upper limit of normal (ULN), alkaline phosphatase (AP) ≤ 3 × ULN, serum total bilirubin < mg/dL, and prothrombin time and partial thromboplastin time ≤ 1.5 × ULN]. This study was carried out in accordance with the recommendations of ethics guidelines of the institutional hospital with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The ethics committee at China Medical University Hospital (Taiwan) approved the study protocol (approval no. CMUH106-REC1-098).

study Design
The primary objective was to examine the initial tumor specimen or peripheral blood mononuclear cell (PBMC) expression of CD45, CD4, CD8, PD-L1, and PD-1 in GBM patients who received conventional therapy, compared with those who received conventional therapy with adjuvant autologous dendritic cell tumor antigen (ADCTA) vaccine. The conventional treatment was defined as tumor resection or biopsy (non-resectable) and subsequent concomitant chemoradiotherapy (CCRT) with TMZ, according to the guidelines suggested by Stupp et al. (46) (we defined this as the reference group). The add-on study design included an ADCTA vaccine treatment period, a posttreatment tracking period, and a retrospective pathological analysis ( Figure 1A). The ADCTA vaccine therapy began 1-2 months post-surgery in conjunction with concomitant CCRT and TMZ. The vaccination protocol for this 10-injection course was four times every 2 weeks followed by monthly six times for a course of 8 months. For patients who were too weak or for other reasons unable to complete the full 10 injections, a minimum of 4 injections was required; otherwise, the patient was excluded from the study. In the reference group, patients underwent surgery followed by concomitant CCRT with TMZ only.

Preparation of Dc-Based Vaccine
In the ADCTA group, the protocol for DC-based vaccine preparation was based on 2011 and 2012 reports by Chang et al. and Cho et al. (23,24), and the schematic diagram is represented in Figure 1B.
In brief, fresh tumor tissues removed in the operating room were collected and chopped into small pieces, and then processed by the Brain tumor dissociation kits (130-095-942; MACS ® , Miltenyi Biotec, Germany). After cell culture, at least 100-150 million tumor cells were generated to provide sufficient tumor antigens. Cells were irradiated with 20 Gy and then lysed by quick freezing and thawing to produce tumor antigens. At least 2 mg of total protein from the cell lysates were collected from the supernatant after centrifugation for use as the source of tumor antigens. We collected the patient's PBMC (5.0 × 10 9 cells/mL) from peripheral blood through leukapheresis. Generally, the leukapheresis was performed in 1 month after subjects had an operation. The monocytes were enriched by 2-h attachment method on the plastic dish at 37°C. The isolated monocytes were cultured in CTS™ AIM V™ Medium (Invitrogen, Carlsbad, CA, USA) with various cytokine proteins for immature DC differentiation. After 7 days, prepared tumor antigens were added into the culture medium for manufacturing DC-based vaccine. Finally, we collected and washed the DC-based vaccines and then divided the cells among 12 tubes, which contain 2-5 × 10 7 DCs cells. All DC vaccines were stored in the liquid nitrogen tank. Before used, the DC vaccine was thawed and the cells were washed with 4°C normal saline twice and added 1 mL of saline.
Sequential tissue sections (2 µm) of the GBM tumor sections and cytoblocks isolated from each patient were attached to adhesive glass slides and baked at 70°C for 20 min prior to IHC staining of CD45, CD4, CD8, PD-1, and PD-L1, using the Leica BOND-MAX system (Leica Biosystems, Nussloch, Germany). Tissue sections were dewaxed three times in xylene for 1 min each and rehydrated through graded ethanol for 15 s each, followed by being washed three times with PBS for 15 s each. Antigen retrieval was performed For PD-L1, Novolink Polymer Detection System (Leica, Newcastle, United Kingdom) was applied after primary antibody incubation. In brief, washed slides were incubated with Novolink Polymer for 30 min, and peroxidase activity was developed with diaminobenzidine (DAB) working solution for 5 min. Finally, the slides were counterstained with hematoxylin. An isotype control was also performed for PD-L1 on each specimen.
In GBM tissue sections stained for PD-L1, tumor cells displayed both patchy/diffuse fibrillary and geographic membraneous staining. We scored PD-L1 expression referring to the methods provided by Berghoff et al. (48) and Nduom et al. (49). When PD-L1 staining was identified with intermediate to strong staining intensity (whether fibrillary or membraneous staining pattern) ≥5% of GBM cells, we defined it as PD-L1 expression, otherwise it was defined as non-expression.
For CD45, CD4, CD8, and PD-1, 25 different high-power fields (HPFs, 400× magnification) containing the most abundant tumor infiltrating lymphocytes (TILs) and 25 random HPFs for PBMCs were counted and summed by two board certified neuropathologists (Chia-Ing Jan and Horng-Jyh Harn) who had been blinded to the patients' treatment group and clinical history. The count for each patient's TILs and PBMCs was determined by the two pathologists, and the level of agreement between the pathologists for the manual counting of TILs and PBMCs was compared using kappa statistics.

immunofluorescence (iF)
Slides were pretreated as described above for IHC. After blocking with 10% fetal bovine serum (GeneDirex, Anaheim, CA, USA) for 1 h, sections were incubated with a rabbit monoclonal antibody against human CD8 (clone SP16 Thermo Fisher, Waltham, MA, USA) for 45 min at 25°C temperature. Tissue sections were washed three times with TBS-Tween 20 at 60 rpm on a shaker for 5 min each and then incubated with Alexa Fluor 488-conjugated goat anti-rabbit secondary antibody (Thermo Fisher, Waltham, MA, USA) for 45 min at 25°C followed by three PBS-Tween 20 washes at 60 rpm on a shaker for 5 min each. PD-1 staining was performed similarly except that a mouse monoclonal mouse antibody against human PD-1 (clone NAT105, Abcam, Cambridge, United Kingdom) at 1:100 dilution and an Alexa Fluor 647-conjugated goat anti-mouse secondary antibody (Thermo Fisher, Waltham, MA, USA) at 1:500 dilution were used. Finally, sections were processed using ProLong Gold Antifade Mountant with DAPI (Thermo Fisher, Waltham, MA, USA) for DAPI nuclear stain and mounting. IF images were captured using a Leica TCS SP8 X white light laser confocal microscope.
isocitrate Dehydrogenase 1 (iDh1) Mutational analysis by ihc and Dna sequencing For IDH1 IHC, the staining procedures were similar to those for PD-L1; the primary monoclonal antibody IDH1-R132H (clone H09, Dianova, Hamburg, Germany) was diluted at 1:100. Isotype control was also performed. Percentage of positive-stained tumor cells (cytosol + nuclear stain) and staining intensity were scored by the two pathologists. The staining intensity and percentage of positive-stained cells were then semi-quantitated into a three-tiered system (diffuse positive, focal positive, and negative) according to previous literature (50)(51)(52). For DNA sequencing, DNA was extracted from FFPE tumor specimens by scraping of tumor sections of three 5-μm-thick unstained paraffin slides. First, excess paraffin and unwanted tissue was removed via trimming with a sterile blade. Then, the remaining tissue on the slides was scraped off with a new sterile blade to remove the tissue section from the slide and transfer the tissue and DNA extraction solution using the DNeasy Blood & Tissue Kit (Qiagen, Hilden, Germany). Mutations in exon 4 of IDH1 were determined by direct sequencing in all cases. Forward and reverse primers used included 5′CGGTCTTCAGAGAAGCCATT3′ and 5′GCAAAATCACATTATTGCCAAC3′, respectively. PCR amplification was performed in a 10-mL reaction mixture containing 50 ng of tumor DNA, 2.5 µL of 10 × Dream Taq buffer, 2.5 µL of 2 mM dNTPs, each forward and reverse primer at 10 µM, and 1.25 µL of Dream Taq DNA Polymerase (Agilent Technologies, Inc., Santa Clara, CA, USA). The initial denaturation was performed at 95°C for 10 min. This was followed by 37 cycles of amplification consisting of denaturation at 95°C for 30 s, annealing at 56°C for 30 s, and extension at 72°C for 30 s.

O6 Methylguanine-Dna Methyltransferase (MgMT) Promoter Methylation by Methylation-specific Pcr (MsP)
The O6 MGMT promoter methylation status of the GBM tumor was determined using MSP analysis. The DNA extraction method was the same as that for IDH1. The DNA extracted from tumor tissues was treated with sodium bisulfite using the EpiTect Fast DNA Bisulfite Kit (Qiagen GmbH, Hilden, Germany). Primer sequences utilized for the methylated forward and reverse primers included 5′GTTTTTAGAACGTTTTGCGTTTCGAC3′ and 5′CACCGTCCCGAA AAAAAACTCCG3′, respectively, and the unmethylated forward and reverse primers used included 5′TGTGTTTTTAGAATGTTTTGTGTTTTGAT3′ and 5′CTA CCACCATCCCAAAAAAAAACT CCA3′, respectively. DNA extracted from the colon cancer SW48 cell line (obtained from American Type Culture Collection), which carries a methylated MGMT promoter, was used as a positive control for MGMT promoter methylation. Genomic DNA extracted from peripheral blood leukocytes of normal patients was used as an unmethylated control sample. De-ionized water was used as a double-negative control. A further control reaction without any template DNA was performed together with each PCR experiment.

Flow cytometry analysis of PBMcs
Patient PBMCs were isolated by apheresis as described previously (23). Briefly, 1 × 10 5 PBMCs were collected and stained with antibodies against PD-1 (clone EH12.1 BD Biosciences, San Jose, CA, USA), CD8 (clone RPA-T8, BD Biosciences), CD3 (clone UCHT1, BD Biosciences) or their isotype antibodies, and then analyzed on a Becton Dickinson FACSCanto. The percpCy5.5-labeled CD3 + T cells were identified after prior gating from single and living cell according to size and granularity; then these CD3 + T cells were analyzed by dot-plot with CD8 and PD-1. The cells of PE-Cy7 high CD8 + and/or FITC high PD-1 + were gated less than 1% as isotype control staining. The same gating strategy was applied in another PBMC samples to identify the populations of CD8 and PD-1 expressing T cells. All of the results were analyzed using FACSDiva software.

statistical analysis
The primary end point was OS calculated from the date of the first surgery confirming GBM to the date of death. Secondary end points were progression-free survival (PFS) calculated from the date of the first surgery confirming GBM to the date of GBM recurrence confirmed by either pathology or MRI report. All statistical analyses were performed using SPSS version 20.0. Cox proportional hazards models were used to calculate hazard ratios (HRs) of recurrence or death according to the number of CD45 + , CD4 + , CD8 + , and PD-1 TILs and PD-L1 of GBM. Kaplan-Meier survival analysis was used to determine the distribution of OS and PFS time, and differences were analyzed using the log-rank test.

Patient characteristics and subtyping of gBM by iDh1 Mutation and MgMT Methylation
There was a total of 47 patients enrolled, of which 27 patients received ADCTA added to conventional therapy (ADCTA group) and 20 patients received conventional therapy only (reference group). Demographic and clinical data of the total patients are shown in Table 1. The mean age was 51.8 years. Detailed individual patient data of the ADCTA and reference groups including sex, age, extent of tumor resection, receiving TMZ or not, receiving radiation therapy/CCRT or not, receiving salvage gamma knife (GKS) for recurrent/residual tumor or not, Karnofsky performance score (KPS), and survival and recurrence (shown  Quantitative evaluation of cD45 + , cD4 + , cD8 + , and PD-1 + Tils in gBM Tissue sections and PBMc cytoblocks A high level of agreement was observed between the two pathologists for the counting of CD45, CD4, CD8, and PD-1 lymphocytes (κ = 0.9998, P < 0.001 for CD45 + , κ = 0.9988, P < 0.001 for CD4 + , κ = 0.9997, P < 0.001 for CD8 + , and κ = 0.9999, P < 0.001 for PD-1 + ).    For all patient samples, the total count (expressed as the median and SD) of CD45 + , CD4 + , CD8 + , and PD-1 + lymphocytes and the ratio of PD-1 + lymphocytes to CD8 + lymphocytes (PD-1 + /CD8 + ratio) in both groups are shown in Table 4. The median CD45 + , CD4 + , CD8 + , and PD-1 + lymphocyte count and median PD-1 + /CD8 + ratio were used for statistical analysis. In tumor-infiltrating lymphocytes, the median CD45 + count was 1,782 in ADCTA group, whereas in the reference group, it was 704 per 25 HPF. The median CD4 + count was 262 in ADCTA group, compared to the reference group it was 130 per 25 HPF. The median CD8 + count was 241 in ADCTA group, and in the reference group it was 141.5 per 25 HPF. The median PD-1 + count was 26 in ADCTA group, whereas in the reference group,  4 | The mean, median, and SD values of total count of CD45 + , CD4 + , CD8 + , and PD-1 + lymphocytes and the ratio of PD-1 + lymphocytes in CD8 + lymphocytes (PD-1 + /CD8 + ratio) for patients in this study, in ADCTA group, and reference group.

PD-1-Presenting lymphocytes and gBM clinical Prognostic Features
To explore the relationship between CD45, CD4, CD8, PD-1presenting lymphocytes and GBM clinical prognostic features, we dichotomized the CD45 + , CD4 + , CD8 + , PD-1 + TIL counts and the PD-1 + /CD8 + ratio into high and low using either the median count or the median ratio as cutoff points, as appropriate. The HR for each variable was calculated using the Cox proportion hazards model (Tables 5 and 6).
In both groups, high CD45, high CD4 count, high CD8 count, gender, receiving radiotherapy alone, TMZ therapy alone, salvage gamma knife (GKS) treatment, or tumor cells expressing IDH1 mutation, and MGMT methylation were all not significant prognostic factors under HR evaluation.
Prognostic significance of cD45, cD4, cD8, and PD-1-Presenting Tils in gBM Tissue sections The OS time of patients in the ADCTA group was significantly longer than that of the reference group. The median survival time in the ADCTA group was 31.0 months, whereas it was 16.0 months for the reference group (P < 0.001, Figure 4).
The prognostic effect of CD45 + , CD4 + , CD8 + , and PD-1 + lymphocyte counts were not of statistical significance in both ADCTA group and reference group.

Positive correlation Between PBMcs, Tils, and Patient survival in the aDcTa group
To determine the correlation between patient survival in the ADCTA group and the PD-1 + /CD8 + ratio in TILs and PBMCs, we performed a chi squared test using data from the 18 (out of 27) ADCTA patients who had peripheral blood available for cytoblock preparation of PBMCs. There was a strong positive correlation between the PD-1 + /CD8 + ratio found in TILs and that found in the PBMCs (r = 0.775; R 2 = 0.6002; P < 0.001; Figure 7).

co-expression of PD-1 Occurs in cD8 + lymphocytes in Tils and PBMcs
On the basis of the significant association observed between a high PD-1 + /CD8 + ratio in TILs/PBMCs and shorter patient OS or DFS in the ADCTA treatment group, we hypothesized that levels of PD-1 expression in cytotoxic CD8 + lymphocytes may determine the therapeutic efficacy of ADCTA therapy.
To determine whether CD8 + lymphocytes from patients with a high PD-1 + /CD8 + ratio did indeed co-express PD-1 and CD8, we performed dual PD-1 and CD8 IF staining of GBM sections and PBMC cytoblocks followed by confocal microscopy. GBM tissue sections and PBMC cytoblocks with both high and low PD-1 + / CD8 + ratios were selected for IF staining. As shown in Figure 9A, patient no. CM10 with a high PD-1 + /CD8 + ratio showed abundant CD8 + and PD-1 + co-staining in lymphocytes in GBM tissue sections. Furthermore, co-localization of CD8 + and PD-1 + was observed in the merged image, confirmed by reconstruction of single slices of xz and yz planes in the z-axis stacked image ( Figure 9A, left low panel). A similar phenotype was observed in the PBMC cytoblock sections from the same patient no. (CM10), in which many CD8 + lymphocytes co-localized with PD-1 + lymphocytes ( Figure 9B left upper and low panels). By contrast, GBM tissue sections ( Figure 9C left upper and low panels, no. CM 27) and PBMC cytoblock sections ( Figure 9D upper and low panels, no. CM27) having a low PD-1 + /CD8 + ratio presented abundant CD8 + lymphocytes but, as expected, a low number of PD-1 + lymphocytes, and no co-localization was observed.
Possible Positive association among Frozen PBMcs' Dual expression PD-1 + / cD8 + cells and Their PD-1 + /cD8 + ratio in PBMc cytoblocks, and Tils in aDcTa group In order to observe the associations among frozen PBMCs, PBMC cytoblocks, and TILs, we performed flow cytometry analysis of PD-1 and CD8 in frozen PBMCs in ADCTA group from the patients with available frozen PBMCs for flow cytometry analysis. Representative results of high and low PD-1 + /CD8 + ratio under flow cytometry were demonstrated. One patient (Figure 10A CD8 + ratio, there seems to be a trend in correlation between flow cytometry evaluation, PBMC cytoblock, and TILs by IHC (flow cytometry versus PBMC cytoblock P = 0.057 and flow cytometry versus TILs P = 0.068, data not shown). Our data suggest that there could be positive correlation between flow cytometry of PD-1 + /CD8 + lymophocytes and PMBC cytoblocks PD-1 + /CD8 + lymphocytes, and flow cytometry of PD-1 + /CD8 + lymophocytes and TIL PD-1 + /CD8 + lymophocytes. However, more studies are required to confirm this finding.

DiscUssiOn
In this study of glioblastoma patients treated with adjuvant ADCTA immunotherapy compared to conventional adjuvant therapy, the treatment responsiveness in ADCTA group was found to be strongly associated with a low TIL PD-1 + /CD8 + ratio within the glioblastoma tumor. Other predictors of treatment res ponse are younger age (<57 years), gross total tumor resection, complete CCRT, and PD-1 + lymphocyte counts. However, CD45 + , CD4 + , and CD8 + lymphocyte count, and CD4 − /CD8 − /PD-1 + / CD45 + immunophenotype lymphocytes does not seem to effect prognosis. Tumor expression of PD-L1 in this study also failed to predict the effectiveness of adjuvant ADCTA therapy.
From the melanoma experience (53), PD-1 + /CD8 + TILs are known to express CTLA4 and Ki67 markers and lack expression of CD127, a phenotypic characteristic of exhausted T-cells. Moreover, effector cytokine production (IL2 and interferongamma) is also impaired PD-1 + /CD8 + cells (54). In GL261 mice, DC vaccination promotes an antitumor, infiltrating T cell response but vaccination is less effective in intracranial GBM. Treatment with both DC vaccination and PD-1 + blockade resulted in long-term survival, whereas neither agent alone provided a survival benefit in animals with larger, established tumors (45).
To further elucidate the relationship between the PD-1/PD-L1 axis and patient prognosis, initially we also examined the expression of the PD-1 ligand PD-L1, and demonstrated that tumor PD-L1 expression was not significantly associated with prognosis in both ADCTA and reference groups. Although Nduom et al. (49) indicated that expression of PD-L1 in GBM tumor cells is associated with worse OS (49); we could not corroborate this finding in our study. The major reason could be that PD-L1 is present not only in GBM cells but also in tumor-infiltrating myeloid cells (TIMs) such as macrophages and T-regulatory cells. Some reports demonstrated that PD-L1 expression in the GBM microenvironment is contributed by TIM rather than by the tumor cells themselves. This means that patients with scarce PD-L1 + GBM tumor cells but with high TIM numbers surrounding the tumor environment would still have strong immune inhibition (55)(56)(57). It may also be due to the complex expression and frequent mutations of PD-L1 in glioblastoma cells (58).
Many clinical trials using DCs to treat GBM have reported an increase in CD8 + TILs after vaccination (59). However, in the complicated tumor microenvironment, GBM tumor cells may inhibit the immune response by the PD-1/PD-L1 pathway. The high PD-1 + /CD8 + TIL ratio causes TILs to be exhausted and limits the efficacy of immunotherapy (60)(61)(62)(63).
In this study, the median OS and PFS in the ADCTA group was 31 and 16 months without stratification of patients, respectively. But after separating the patients with a low PD-1 + / CD8 + ratio, TILs extended the median OS to 61 months and PFS to 11.2 months versus high PD-1 + /CD8 + ratio median OS 20.7 months and PFS 4.43 months. This drastic difference was not seen in the reference group. It could be because GBM tumor cells have strong immune system inhibition effecting immune cell proliferation and function (64). In the reference group, there are no so called "tumor antigen activated" cytotoxic T-cells to effectively kill GBM tumor cells, but in ADCTA group, there are many of these cells. So if these "tumor antigen activated" cytotoxic T-cells exhibit anergy, it will surely effect prognosis. Our results also suggest that in the PBMC of ADCTA group, although high counts of PD1+, CD4+, or CD8+ T-cells and high counts of PD-1+, but CD4 − and CD8 − lymphoid cells (including CD4 and CD8 double-negative T-cells) are identified, not all of these cells can reach the GBM tumor site and effect the ability of tumor killing by T-cells. Tumor microenvironment plays a critical role here. In order for circulating T-cells to reach the tumor microenvironment, they have to pass through the partially permissive anatomic blood-brain barrier, the desmoplastic reaction of produced by the fibroblasts surrounding the tumor cells. In addition, tumor necrosis hinders T-cell trafficking; pericytes and endothelial cells in tumor-associated angiogenesis also inhibit the circulating T-cells from reaching the tumor site.
Based on the findings of the current study, we have proposed a model to illustrate our hypothesis for predicting the effectiveness of ADCTA treatment in GBM patients. In the systemic circulation, PD-1 + or PD-1 − cytotoxic T cells in the blood will arrive at post DC vaccine injection regional lymph nodes for antigen activation. After the cytotoxic T cells reaches the tumor microenvironment, the efficiency of killing tumor cells by tumor antigen-activated cytotoxic T (CD8 + ) cells depends on the proportion of PD-1 + cytotoxic T (CD8 + ) cells, as shown in Figure 11. If patients have a low PD-1 + /CD8 + ratio, they may have a better outcome due to reduced exhausted cytotoxic (CD8 + ) cells.
Obtaining better GBM control is critical issue. Young age, gross complete tumor resection, neuronavigation with 5 Ala, Gradel (BCNU waffle), irradiation, CCRT, IDH-1 mutation, methylated MGMT, TMZ, Avastin, and immunotherapy (2,4,6,65,66) have all been studied and may improve the tumor control. In addition, a vaccine preparation with whole tumor lysate under tumor irradiation may also improve vaccine efficacy. Furthermore, many combinatorial therapies such as immunotherapy with checkpoint blockade or with antiangiogenic therapy and cytotoxic therapy may optimize immunogenicity and target tumor adaptive immunosuppressive factors. Recent studies show an increase of the 2-year OS to 40-50%.
Programed death 1 is now a key focus in many cancer controls. In our retrospective study, the low PD-1 + /CD8 + ratio may have been similar to a PD-1 blockade, resulting in less immunosuppression and improved efficacy of DC vaccination. However, PD-1 alone is insufficient for GBM control (45); DCs may be necessary to stimulate a large amount of CD8 + cells to attack tumor cells. Therefore, using combinatorial therapy with DCs and PD-1 blockade may provide better outcomes in patients with high immunosuppression (67,68).
In conclusion, our study results suggest that PD-1 + /CD8 + ratio is a critical factor affecting both the OS and PFS of GBM patients receiving ADCTA therapy. Other important, statistically significant factors include: age, gross total tumor removal, receiving complete CCRT, and PD-1 lymphocyte count. The TIL or PBMC PD-1 + /CD8 + ratio provides a simple and feasible method of determining whether GBM patients are suitable for ADCTA adjuvant therapy. Patients who still wish to receive ADCTA immunotherapy but have a high TIL or PBMC PD-1 + / CD8 + ratio may benefit from a combination therapy with anti-PD-1 and/or anti-PD-L1 monoclonal antibodies or adoptive T-cell therapy (69).

eThics sTaTeMenT
This study was carried out in accordance with the recommendations of ethics guidelines of the institutional hospital with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The ethics committee at China Medical University Hospital (Taiwan) approved the study protocol (approval no. CMUH106-REC1-098).