SEARCH

SEARCH BY CITATION

Keywords:

  • peptide vaccine;
  • peripheral blood;
  • biomarker;
  • microarray;
  • granulocyte;
  • interleukin 6

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. FUNDING SOURCES
  9. REFERENCES

BACKGROUND:

Because only a subset of patients show clinical responses to peptide-based cancer vaccination, it is critical to identify biomarkers for selecting patients who would most likely benefit from this treatment.

METHODS:

The authors characterized the gene expression profiles in peripheral blood of vaccinated patients to identify biomarkers to predict patient prognosis. Peripheral blood was obtained from advanced castration-resistant prostate cancer patients, who survived for >900 days (long-term survivors, n = 20) or died within 300 days (short-term survivors, n = 20) after treatment with personalized peptide vaccination. Gene expression profiles in prevaccination and postvaccination peripheral blood mononuclear cells (PBMCs) were assessed by DNA microarray.

RESULTS:

There were no statistically significant differences in the clinical or pathological features between the 2 groups. Microarray analysis of prevaccination PBMCs identified 19 genes that were differentially expressed between the short-term and long-term survivors. Among the 15 up-regulated genes in the short-term survivors, 13 genes, which were also differentially expressed in postvaccination PBMCs, were associated with gene signatures of granulocytes. When a set of 4 differentially expressed genes were selected as the best combination to determine patient survival, prognosis was correctly predicted in 12 of 13 patients in a validation set (accuracy, 92%).

CONCLUSIONS:

These results suggested that abnormal granulocytes present in the PBMC faction may contribute to poor prognosis in advanced prostate cancer patients receiving personalized peptide vaccination. Gene expression profiling in peripheral blood might thus be informative for devising better therapeutic strategies by predicting patient prognosis after cancer vaccines. Cancer 2012;118: 3208–21. © 2011 American Cancer Society.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. FUNDING SOURCES
  9. REFERENCES

Together with the progressive increase of basic knowledge in tumor immunology, the field of cancer vaccines has dramatically moved forward.1-5 However, because only a limited number of patients show clinically beneficial responses to cancer vaccination, it would be critical to identify clinical and/or biological markers useful for selecting patients who would most likely benefit from this treatment.5-8 Recently, polymorphisms of several genes, including CCR5, interferon (IFN)-γ, interleukin (IL)-6, and cytotoxic T lymphocyte antigen 4 (CTLA-4), have been reported to be associated with clinical responses in nonspecific immunotherapies, such as IL-2, IFN-α, Bacille Calmette-Guérin, and anti-CTLA-4 antibody therapies.9-12 In addition, levels of serum cytokines or growth factors, including IL-1β, IL-1α, IL-6, tumor necrosis factor (TNF)-α, CCL3, CCL4, and vascular endothelial growth factor (VEGF), have also been shown to be correlated with clinical responses in nonspecific cytokine therapies.13, 14 However, because no reliable markers are currently in widespread use for predicting clinical outcomes in specific immunotherapies, novel biomarkers remain to be identified.

Recently, high-throughput technologies have been developed as a novel approach to discovering biomarkers. In particular, DNA microarray technology is among the most widely recognized and extensively studied to identify new biomarkers for individualized therapies.15-20 For example, gene expression profiles examined on a genome-wide scale in tumor tissues have been reported to clearly reflect clinical outcomes and/or responses to treatments in cancer patients.15-17 In addition, expression array data of peripheral blood have also been shown to afford a comprehensive view of the patients' immune status in a variety of fields, including organ transplantation and autoimmune diseases.18-20 However, there is little information available regarding gene expression profiles in peripheral blood of patients receiving cancer vaccines.

We have developed personalized peptide vaccination as a novel modality for cancer treatment, in which vaccine antigens are selected on the basis of pre-existing immune responses against vaccine antigens.5, 21-24 For example, our results in a recent small randomized clinical trial showed a potential clinical benefit of personalized peptide vaccination in advanced castration-resistant prostate cancer patients.22 However, for further development of this approach, novel predictive biomarkers for selecting suitable patients with better clinical responses remain to be identified. Sipuleucel-T (Provenge; Dendreon Corporation, Seattle, Wash), an autologous active cellular immunotherapy product designed to stimulate a T-cell immune response against human prostatic acid phosphatase, was first approved for castration-resistant prostate cancer patients by the US Food and Drug Administration in 2010.3 In this immunotherapy, CD54 up-regulation, a measure of the product's potency, has been reported to be correlated with patient overall survival.25 However, this surrogate marker may be applicable only for dendritic cell-based immunotherapies. In the current study, we performed a gene expression profiling in peripheral blood samples of castration-resistant prostate cancer patients, who showed good or poor prognosis after personalized peptide vaccination, to identify promising biomarkers that are predictive of patient prognosis after treatment. Although it is likely that tumor tissues may have more informative gene signatures than peripheral blood mononuclear cells (PBMCs), they are usually difficult to obtain in patients with advanced castration-resistant prostate cancer. Therefore, given the ease of sampling and the ability to perform analyses at multiple time points, we used PBMCs for gene expression profiling in the current study. Our results suggested that the gene expression profiles in prevaccination PBMCs would be informative for devising better therapeutic strategies by predicting the subpopulation of castration-resistant prostate cancer patients who would most likely benefit from cancer vaccines.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. FUNDING SOURCES
  9. REFERENCES

Patients

This is a retrospective analysis with peripheral blood samples from a subset of 164 patients with metastatic castration-resistant prostate cancer, who were positive for human leukocyte antigen (HLA)-A24 or HLA-A2 and enrolled in phase 1, 1-2, and 2 clinical trials for personalized peptide vaccination between February 2001 and April 2008.22, 24 These studies were approved by the ethics review committee at the participating hospitals in Japan (Kurume University Hospital, Kinki University Hospital, Okayama University Hospital, and Nara Medical University Hospital). Before enrollment in the studies, the history of all patients was studied, and physical examination, assessment of performance status, complete blood counts, serum biochemistry profiles, serum prostate-specific antigen (PSA) levels, chest radiograph, electrocardiogram, bone scan, and computerized tomography scans of the abdomen and pelvis were performed. Patients with a progression of disease (PD) after androgen ablation and second-line hormone therapy were enrolled. PD was defined by at least 1 of the following 3 criteria: 1) 2 consecutive 25% increases in PSA levels at least 2 weeks apart, 2) an increase of >25% in bidimensionally measurable soft tissue metastases, or 3) appearance of new foci on radionuclide bone scans. Other eligibility criteria included Eastern Cooperative Oncology Group performance status of 0 or 1, age of 18 years or more, normal hematologic, hepatic, and renal functions, and negative results on serologic tests for hepatitis B and hepatitis C. Patients with evidence of serious illness, an active secondary malignancy that occurred within 5 years before entry, or autoimmune diseases were excluded from the studies. After full explanation of the protocol, written informed consent was obtained from all patients before enrollment.

The right peptides for vaccination to individual patients were selected in consideration of the pre-existing host immunity before vaccination, assessed by titers of immunoglobulin (Ig)G specific to each of the 26 different vaccine candidates, as reported previously.5, 21-24 Peptides selected based on the results of peptide-specific IgG titers (3 or 4 peptides/vaccination; 3 mg/each peptide) were subcutaneously administrated with incomplete Freund adjuvant (Montanide ISA51; Seppic, Paris, France) once per week for 6 consecutive weeks. After the first cycle of 6 vaccinations, antigen peptides, which were reselected according to the titers of peptide-specific IgG at every cycle of 6 vaccinations, were administered every 2 weeks while patients were allowed to continue the vaccinations.

Among the 164 patients enrolled, the patients who survived for >900 days (long-term survivors, n = 20) or who died within 300 days (short-term survivors, n = 20) were selected for analyses of gene expression profiles in PBMCs and soluble factors in plasma. The short-term and long-term survivors were defined in reference to a randomized, nonblinded, multinational phase 3 study of docetaxel-based regimens, TAX327, which involved 1006 men with castration-resistant prostate cancer,26, 27 because the disease conditions of castration-resistant prostate cancer patients in the TAX327 study were similar to those in the current study. Because each patient subgroup in the TAX327 study showed a median survival of 16.3 to 19.2 months,28 we selected the patients who survived for >30 months (900 days) and who died within 10 months (300 days) as the long-term and short-term survivors, respectively, in the current study.

Blood Samples

PBMCs and plasma were used for measurement of gene expression profiles and soluble factors, respectively. Because this was a retrospective study with limited availability of patient samples, PBMCs or plasma from the patients were not equally available for each assay. Prevaccination PBMCs were analyzed by DNA microarray in all of the 40 selected patients (long-term survivors, n = 20; short-term survivors, n = 20). However, postvaccination PBMCs, which were obtained after the completion of 1 cycle of 6 vaccinations, were analyzed by DNA microarray in only a subset of the patients (long-term survivors, n = 16; short-term survivors, n = 14), because of failure in the completion of 1 cycle of vaccinations or the poor quality of purified RNA. Among these 30 postvaccination PBMCs, only 24 (long-term survivors, n = 12; short-term survivors, n = 12) were used for the quantitative real-time polymerase chain reaction (qRT-PCR) assay. Prevaccination plasma samples for soluble factor measurements were used from 36 patients (long-term survivors, n = 18; short-term survivors, n = 18).

The prevaccination PBMCs from all 40 patients were used as a training set to generate a gene classifier to predict patient prognosis. In addition, prevaccination PBMCs from 13 new independent cancer patients, who survived for >600 days (n = 6) or who died within 300 days (n = 7) after personalized peptide vaccination, were used in a validation test.

RNA Isolation From PBMCs

PBMCs were prepared from 20 mL of peripheral blood by density gradient centrifugation using Ficoll-Paque (GE Healthcare Life Sciences, Uppsala, Sweden). All samples were cryopreserved until RNA extraction. Total RNA was isolated using TRIZOL LS reagent (Invitrogen, Carlsbad, Calif) and purified using RNeasy Mini Kit (Qiagen, Valencia, Calif), according to the manufacturer's instructions. Quality and integrity of the purified total RNA were confirmed using an Agilent 2100 bioanalyzer (Agilent Technologies, Palo Alto, Calif) and Nanodrop ND-1000 (Thermo Fisher Scientific, Wilmington, Del).

DNA Microarray Analysis

RNA amplification, labeling, and hybridization on HumanWG-6 v3.0 Expression BeadChip (Illumina Bead Array; Illumina, San Diego, Calif) were performed according to the manufacturer's instructions. Microarray data were extracted using BeadStudio v3.0 software (Illumina) and were then preprocessed and normalized using a variance-stabilizing transformation and robust spline normalization, as implemented in the lumi Bioconductor package. To filter low confidence probes that might increase the false-positive rates in subsequent statistical analyses, probes that did not reach a detection level with a P value <.05 in 70% of all samples were discarded. Accordingly, of the 48,803 probes on the chips, 16,449 remained above the reliable detection level. To assess the differential gene expression between the long-term and short-term survivors, we used the fold-change ranking, together with the P values, using the Linear Models for Microarray Data (Limma) Bioconductor package.29 To determine the fold-change in the gene expression of the samples from the long-term survivors versus those from the short-term survivors, we calculated the fold-change values using the following formula: log2 fold-change = log2(SS/SL), where SL represented the assay range for a target gene in the samples from the long-term survivors and SS represented that from the short-term survivors. Because the gene chip used in the current study (Illumina HumanWG-6 v3.0 Expression BeadChip) contained 48,803 probes, which corresponded to 25,409 annotated genes, some genes had multiple different probes on the gene chip. Therefore, the genes with multiple probes might be repeatedly detected by different probes and identified at multiple times in the list of differentially expressed genes.

qRT-PCR

After the total RNA (200 ng) from postvaccination PBMCs (long-term survivors, n = 12; short-term survivors, n = 12) was reverse-transcribed into the first-strand cDNA with PrimeScript RT reagent kit (Takara Bio, Shiga, Japan), qRT-PCR was performed with a SYBR Premix Ex Taq II kit (Takara Bio) by using a Thermal Cycler Dice Real Time System (Takara Bio). The data were evaluated by the ddCT method. The number of copies of the housekeeping gene glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was measured in each cDNA sample as an internal control. The expression of each gene was normalized to that of GAPDH. The sequences of the primers for qRT-PCR were as follows: defensin alpha 1 (DEFA1): forward, 5′-CGGACATCCCAGAAGTGGT TG-3′, reverse, 5′-CCCTGGTAGATGCAGGTTCCA TA-3′; defensin alpha 4 (DEFA4): forward, 5′-CACTC CAGGCAAGAGGTGATGA-3′, reverse, 5′-GAGGCA GTTCCCAACACGAAGT-3′; myeloperoxidase (MPO): forward, 5′-CTGCATCATCGGTACCCAGTTC-3′, reverse, 5′-GATGCCTGTGTTGTCGCAGA-3′; carcinoembryonic antigen-related cell adhesion molecule 8 (CEACAM8): forward, 5′-TGGCACATTCCAGCAA TACACA-3′, reverse, 5′-ATCATGATGCTGACAGT GGCTCTA-3′; GAPDH: forward, 5′-GCACCGTCA AGGCTGAGAAC-3′, reverse, 5′-TGGTGAAGACGC CAGTGGA-3′.

Measurement of Soluble Factors in Plasma

To detect the plasma levels of cytokines, chemokines, and growth factors before vaccination (long-term survivors, n = 18; short-term survivors, n = 18), a bead-based multiplex assay (xMAP; Luminex, Austin, Tex) was used. For this assay, multiple soluble factors were measured in duplicate 100 μL aliquots of plasma by using the Luminex 200 system according to the manufacturer's instructions. The analyte kit used for the measurement of the levels of multiple cytokines, chemokines, and growth factors, including IL-1Rα, IL-1β, IL-2, IL-2R, IL-4, IL-5, IL-6, IL-7, IL-8, IL-10, IL-12, IL-13, IL-15, IL-17, IFN-α, IFN-γ, TNF-α, granulocyte colony-stimulating factor (G-CSF), granulocyte-macrophage colony-stimulating factor (GM-CSF), interferon-inducible protein (IP)-10, RANTES, Eotaxin, macrophage inflammatory protein (MIP)-1α, MIP-1β, monocyte chemoattractant protein (MCP)-1, monokine induced by interferon-gamma (MIG), VEGF, endothelial growth factor (EGF), human growth factor (HGF), and basic fibroblast growth factor (FGF), was obtained from Invitrogen (Human 30-Plex).

Statistical Analysis

Mann-Whitney and Fisher exact tests were used for statistical analyses of clinical and pathological features of the patients. Overall survival was estimated by the Kaplan-Meier method and log-rank test. Mann-Whitney test was used to compare the plasma levels of cytokines, chemokines, and growth factors, and the gene expression levels in PBMCs assessed by qRT-PCR. All tests were 2-sided, and the differences with P values <.05 were considered statistically significant. In identification of differentially expressed genes in PBMCs, the data were assessed by the fold-change ranking, together with a nonstringent P value cutoff.29 From the differentially expressed genes, the genes critical for accurate classification of the short-term and long-term survivors were selected by stepwise discriminant analysis method. The classification performance of the selected genes was validated in an independent test set (n = 13) by determining sensitivity, specificity, positive predictive value, negative predictive value, and accuracy. All statistical analyses were conducted using SAS version 9.1 (SAS Institute, Cary, NC).

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. FUNDING SOURCES
  9. REFERENCES

Patients

We selected advanced castration-resistant prostate cancer patients who received personalized peptide vaccination and who survived for >900 days (long-term survivors, n = 20) or died within 300 days (short-term survivors, n = 20) for the DNA microarray analysis in PBMCs. For personalized peptide vaccination, different combinations of 4 peptides were selected based on the results of peptide-specific IgG titers in all patients, except for 1 patient receiving 3 peptides in the short-term survivors. Numbers of vaccinations were significantly different between the long-term (median, 50; range, 10-124) and short-term (median, 8; range, 3-14) survivors (P < .001). PSA doubling time calculated by the log-slope method in the long-term and short-term survivors after personalized peptide vaccination was negative in 10 (50%) of 20 and in 4 (20%) of 20 patients, respectively. In the remaining patients positive for PSA doubling time, the long-term survivors (median, 13.6; range 1.6-92.9; n = 10) had a longer PSA doubling time (P = .006) than the short-term survivors (median, 2.1; range, 0.7-79.0; n = 16).

Identification of Differentially Expressed Genes in Postvaccination PBMCs

We first analyzed postvaccination PBMCs by using DNA microarray analysis (HumanWG-6 v3.0 Expression BeadChip; 48,803 probes corresponding to 25,409 genes in total) to determine the genes that were differentially expressed between the long-term and short-term survivors. As shown in Table 1, there were no statistically significant differences in the clinical or pathological features except for the number of vaccinations (P < .001) and overall survival (log-rank test, P < .001) between the long-term (n = 16) and short-term (n = 14) survivors in whom postvaccination PBMCs were analyzed. Figure 1A shows a volcano plot that graphs the log2 fold-change on the x-axis versus the statistical significance (negative log10 P value) on the y-axis. When the data were assessed by fold-change ranking (log2 fold-change <−1.0 or >1.0) together with P values (P < .01), expressions of 42 probes, corresponding to 38 genes, were significantly altered between the 2 groups; 1 gene was down-regulated, whereas the remaining 37 were up-regulated in the short-term survivors (Table 2). Notably, 20 of the 37 up-regulated genes are known to be preferentially expressed in granulocytes. For example, many of them, including defensins (DEFA1, DEFA3, DEFA4), ELA2, CTSG, CAMP, and MPO, are reportedly localized within the granules in granulocytes and related to defense responses. In addition, other granulocyte-related molecules, such as matrix metalloproteinase 9 (MMP9) and arginase-1 (ARG1), are known to play important roles in tumor promotion and immune suppression.30, 31 The differential gene expression detected by the microarray analysis was further confirmed by qRT-PCR for some of the identified genes, including DEFA1, DEFA4, CEACAM8, and MPO (Fig. 2).

thumbnail image

Figure 1. Volcano plots present the microarray data in prevaccination and postvaccination peripheral blood mononuclear cells (PBMCs). The plot graphs the fold-change (FC; log2[short/long]) on the x-axis versus statistical significance (minus log10 P value) on the y-axis in PBMCs (A) after and (B) before the peptide vaccines.

Download figure to PowerPoint

thumbnail image

Figure 2. Differential gene expression was assessed by quantitative real-time polymerase chain reaction (qRT-PCR). The gene expression of (A) DEFA1, (B) DEFA4, (C) CEACAM8, and (D) MPO were measured by qRT-PCR in postvaccination peripheral blood mononuclear cells of the short-term (n = 12) and long-term (n = 12) survivors. The expression of each gene was normalized to that of GAPDH. The expression ratios of each gene are shown. Box plots show median and interquartile range (IQR). The whiskers (vertical bars) are the lowest value within 1.5 × IQR of the lower quartile and the highest value within 1.5 × IQR of the upper quartile. Data not included between the whiskers were plotted as outliers with dots. Two-sided P values were calculated with Mann-Whitney test.

Download figure to PowerPoint

Table 1. Patient Characteristics (Postvaccination Analysis)
CharacteristicShort-Term Survivors, n = 14Long-Term Survivors, n = 16P
  1. Abbreviations: CI, confidence interval; ECOG, Eastern Cooperative Oncology Group; HLA, human leukocyte antigen; PSA, prostate-specific antigen.

Age, y   
 Median (range)62 (50-81)71.5 (54-78).109
ECOG performance status, No. [%]   
 013 [93]16 [100].467
 11 [7]0 [0] 
HLA typing, No. [%]   
 A2410 [71]9 [56].709
 A23 [21]6 [38] 
 A24 and A21 [7]1 [6] 
PSA, ng/mL   
 Median (range)79 (2-222)34.5 (2-330).308
Gleason score, No. [%]   
 73 [21]5 [31].714
 86 [43]8 [50] 
 95 [36]3 [19] 
Site of metastasis, No. [%]   
 None2 [14]2 [13].888
 Bone only10 [71]13 [81] 
 Bone and lymph nodes1 [7]0 [0] 
 Other organs1 [7]1 [6] 
Number of vaccinations   
 Median (range)9 (5-14)52.5 (10-124)<.001
Survival time, d   
 Median {95% CI}248.5 {176-277}1482 {1120-1764}<.001
Table 2. Differentially Expressed Genes in Postvaccination Peripheral Blood Mononuclear Cells
Gene SymbolGene NameFold-ChangeaPbExpressioncBefore and Afterd
  • a

    log2 (short/long).

  • b

    Limma P value.

  • c

    Preferential expression in granulocyte (G) and erythroid cells (E).

  • d

    Commonly identified in both prevaccination and postvaccination peripheral blood mononuclear cells (#).

  • e

    Identified by multiple different probes on the gene chip.

LTBLymphotoxin beta−1.03<.001  
OLR1Oxidized low-density lipoprotein receptor 11.04.004  
CEACAM1Carcinoembryonic antigen-related cell adhesion molecule 11.07<.001G 
ARG1Arginase, liver1.10<.001G 
MYL4Myosin, light chain 4, alkali; atrial, embryonic1.14.007  
ALAS2Delta-aminolevulinate, synthase 21.20.009E 
SLPISecretory leukocyte peptidase inhibitor1.22<.001G 
SELENBP1Selenium-binding protein 11.22.008  
SNCAeSynuclein, alpha1.25.008  
AZU1Azurocidin 11.25<.001G#
HMGXB4HMG box domain containing 41.27.001  
RNASE3Ribonuclease, RNase A family, 31.28.001G#
HBQ1Hemoglobin, theta 11.31.001E 
MMP9Matrix metallopeptidase 91.32<.001G 
GYPEGlycophorin E1.36<.001E 
SNCAeSynuclein, alpha1.39.005  
EPB42Erythrocyte membrane protein band 4.21.45.003E 
HPHaptoglobin1.50<.001E 
IFIT1LInterferon-induced protein with tetratricopeptide repeats 1-like1.51.003  
CD24CD24 molecule1.55<.001G 
BPIBactericidal/permeability-increasing protein1.64<.001G 
CEACAM6Carcinoembryonic antigen-related cell adhesion molecule 61.72<.001G#
PGLYRP1Peptidoglycan recognition protein 11.80<.001G#
MPOMyeloperoxidase1.94<.001G#
OLFM4Olfactomedin 42.01<.001  
HBMHemoglobin, mu2.05.002E 
ALAS2Delta-aminolevulinate, synthase 22.11.005E 
CEACAM8Carcinoembryonic antigen-related cell adhesion molecule 82.13<.001G#
ERAFErythroid-associated factor2.29.002E 
CA1Carbonic anhydrase I2.31<.001G 
HBDHemoglobin, delta2.37.002E 
LCN2Lipocalin 22.40<.001G#
CTSGCathepsin G2.40<.001G#
DEFA1eDefensin, alpha 12.40<.001G#
CAMPCathelicidin antimicrobial peptide2.41<.001G#
ELA2Elastase 2, neutrophil2.44<.001G#
DEFA4Defensin, alpha 4, corticostatin2.53<.001G#
DEFA3Defensin, alpha 3, neutrophil-specific2.65<.001G#
DEFA1eDefensin, alpha 12.65<.001G#
DEFA1eDefensin, alpha 12.67<.001G#
DEFA1eDefensin, alpha 12.68<.001G#
DEFA1eDefensin, alpha 12.87<.001G#

Identification of Differentially Expressed Genes in Prevaccination PBMCs

We next investigated the differentially expressed genes in prevaccination PBMCs from the long-term and short-term survivors. There were no statistically significant differences in the clinical or pathological features except for the number of vaccinations (P < .001) and overall survival (log-rank test, P < .001) between the long-term (n = 20) and short-term (n = 20) survivors in whom prevaccination PBMCs were analyzed (Table 3). As shown in the volcano plot, both fold-change and Limma P values in prevaccination samples were substantially lower than those in the postvaccination samples (Fig. 1B). Indeed, when the data were assessed with the same criteria as those for the postvaccination samples (log2 fold-change <−1.0 or >1.0 and P < .01), only 5 genes (5 probes) were identified as being differentially expressed (data not shown). However, when a less stringent criterion (log2 fold-change <−0.6 or >0.6 and P < .05) was used, 19 genes (23 probes) were identified; among these, 4 genes were down-regulated, whereas 15 were up-regulated in the short-term survivors (Table 4). Notably, of the 15 up-regulated genes, 13 genes, all of which were commonly identified in both prevaccination and postvaccination PBMCs, were associated with gene signatures of granulocytes.

Table 3. Patient Characteristics (Prevaccination Analysis)
CharacteristicShort-Term Survivors, n = 20Long-Term Survivors, n = 20P
  1. Abbreviations: CI, confidence interval; ECOG, Eastern Cooperative Oncology Group; HLA, human leukocyte antigen; PSA, prostate-specific antigen.

Age, y   
 Median (range);62 (50-81)71 (54-78).058
ECOG performance status, No. [%] 
 017 [85]20 [100].231
 13 [15]0 [0] 
HLA typing, No. [%]
 A2413 [65]12 [60]1.000
 A25 [25]6 [30] 
 A24 and A22 [10]2 [10] 
PSA, ng/mL   
 Median (range)73.5 (2-296)34.5 (2-330).239
Gleason score, No. [%]
 74 [20]5 [25].710
 88 [40]10 [50] 
 98 [40]5 [25] 
Site of metastasis, No. [%]
 None2 [10]3 [15]1.000
 Bone only14 [70]14 [70] 
 Bone and lymph nodes3 [15]2 [10] 
 Other organs1 [5]1 [5] 
Number of vaccinations   
 Median (range)8 (3-14)50 (10-124)<.001
Survival time, d   
 Median {95% CI}196 {135-273}1482 {1120-1764}<.001
Table 4. Differentially Expressed Genes in Prevaccination Peripheral Blood Mononuclear Cells
Gene SymbolGene NameFold-ChangeaPbExpressioncBefore and Afterd
  • a

    Log2 (short/long).

  • b

    Limma P value.

  • c

    Preferential expression in granulocyte (G).

  • d

    Commonly identified in both prevaccination and postvaccination peripheral blood mononuclear cells (#).

  • e

    Identified by multiple different probes on the gene chip.

PRKAR1AProtein kinase, cAMP-dependent, regulatory, type I, alpha−0.82.049  
LRRN3Leucine-rich repeat neuronal 3−0.61.008  
PCDH17Protocadherin 17−0.60.002  
TTNTitin−0.60.008  
LAIR2Leukocyte-associated immunoglobulin-like receptor 20.60.032  
RNASE3Ribonuclease, RNase A family, 30.63.020G#
CEACAM6Carcinoembryonic antigen-related cell adhesion molecule 60.65.010G#
AZU1Azurocidin 10.66.006G#
HIST1H4CHistone cluster 1, H4c0.71.025  
PGLYRP1Peptidoglycan recognition protein 10.72.007G#
CEACAM8Carcinoembryonic antigen-related cell adhesion molecule 80.78.015G#
LCN2Lipocalin 21.00.005G#
MPOMyeloperoxidase1.04.001G#
CAMPCathelicidin antimicrobial peptide1.09.007G#
DEFA1eDefensin, alpha 11.17.031G#
DEFA1eDefensin, alpha 11.20.018G#
DEFA1eDefensin, alpha 11.26.018G#
DEFA3Defensin, alpha 3, neutrophil-specific1.27.017G#
DEFA1eDefensin, alpha 11.27.020G#
DEFA1eDefensin, alpha 11.30.015G#
CTSGCathepsin G1.32.003G#
DEFA4Defensin, alpha 4, corticostatin1.33.002G#
ELA2Elastase 2, neutrophil1.36.002G#

Changes in the Gene Expression Profiles in PBMCs After Personalized Peptide Vaccination

To investigate how personalized peptide vaccination affected the gene expression profiles in PBMCs, we further compared them between before and after personalized peptide vaccination in the long-term (n = 16) and short-term survivors (n = 14). The changes were assessed by fold-change ranking (log2 fold-change <−1.0 or >1.0) together with P values (P < .01). In the long-term survivors, only 1 gene, titin (TTN), was down-regulated (log2 fold-change = −1.04, P < .001) after personalized peptide vaccination, whereas no genes were up-regulated. In contrast, as shown in Table 5, 41 genes (47 probes) were up-regulated after personalized peptide vaccination, whereas no genes were down-regulated in the short-term survivors. Notably, many of the 41 up-regulated genes in the short-term survivors were also identified as being differentially expressed in pre- and/or postvaccination PBMCs.

Table 5. Upregulated Genes After Vaccination in Peripheral Blood Mononuclear Cells From the Short-Term Survivors
Gene SymbolGene NameFold-ChangeaPbExpressioncBefore and Afterd
  • a

    log2 (postvaccination/prevaccination).

  • b

    Limma P value.

  • c

    Preferential expression in granulocytes (G) and erythroid cells (E).

  • d

    Identified as differentially expressed genes in prevaccination and/or postvaccination peripheral blood mononuclear cells.

  • e

    Identified by multiple different probes on the gene chip.

RNASE2Ribonuclease, RNase A family, 21.02<.001  
SLC4A1Solute carrier family 4, anion exchanger, member 11.06.008E 
HEMGNHemogen (HEMGN), transcript variant 21.08.001E 
CEACAM1Carcinoembryonic antigen-related cell adhesion molecule 11.09<.001GAfter
S100PS100 calcium-binding protein P1.09.001  
ALS2Amyotrophic lateral sclerosis 21.09.001  
ARG1Arginase, liver1.10<.001GAfter
SLPISecretory leukocyte peptidase inhibitor1.12<.001GAfter
OLR1Oxidized low-density lipoprotein (lectin-like) receptor 11.14<.001 After
RETNResistin1.15.005  
HBQ1Hemoglobin, theta 11.16.007EAfter
ALAS2eDelta-aminolevulinate, synthase 21.19.004EAfter
MMP9Matrix metallopeptidase 91.22<.001GAfter
RNASE3Ribonuclease, RNase A family, 31.24<.001GBefore, after
HMGXB4HMG box domain containing 41.24.003 After
SELENBP1Selenium-binding protein 11.24.003 After
GYPEGlycophorin E1.36.001EAfter
BPIBactericidal/permeability-increasing protein1.36<.001GAfter
TCN1Transcobalamin I1.38<.001G 
ORM1Orosomucoid 11.38<.001  
CEACAM6Carcinoembryonic antigen-related cell adhesion molecule 61.40<.001GBefore, after
SNCAeSynuclein, alpha1.40.001 After
MPOMyeloperoxidase1.44.002GBefore, after
SNCAeSynuclein, alpha1.44<.001 After
HPHaptoglobin1.46<.001EAfter
CD24CD24 molecule1.48<.001GAfter
IFIT1LInterferon-induced protein with tetratricopeptide repeats 1-like1.55.003 After
EPB42Erythrocyte membrane protein band 4.21.56.002EAfter
CTSGCathepsin G1.56.004GBefore, after
ELA2Elastase 2, neutrophil1.74.002GBefore, after
PGLYRP1Peptidoglycan recognition protein 11.77<.001GBefore, after
DEFA1eDefensin, alpha 11.79<.001GBefore, after
CEACAM8Carcinoembryonic antigen-related cell adhesion molecule 81.80<.001GBefore, after
HBMHemoglobin, mu1.86.005EAfter
DEFA4Defensin, alpha 4, corticostatin1.91<.001GBefore, after
ALAS2eDelta-aminolevulinate, synthase 21.94.005EAfter
CAMPCathelicidin antimicrobial peptide2.03<.001GBefore, after
LCN2Lipocalin 22.04<.001GBefore, after
OLFM4Olfactomedin 42.05<.001 After
DEFA3Defensin, alpha 3, neutrophil-specific2.12<.001GBefore, after
DEFA1eDefensin, alpha 12.12<.001GBefore, after
DEFA1eDefensin, alpha 12.16<.001GBefore, after
DEFA1eDefensin, alpha 12.25<.001GBefore, after
ERAFErythroid associated factor2.29.002EAfter
CA1Carbonic anhydrase I2.45<.001GAfter
HBDHemoglobin, delta2.48.001EAfter
DEFA1eDefensin, alpha 12.73<.001GBefore, after

Selection of a Gene Classifier for Predicting Patient Prognosis After Personalized Peptide Vaccination

One of the most important applications of microarray-based gene expression data is the ability to predict clinical endpoints after treatments.18-20 Thus, we examined whether the gene expression profile obtained by DNA microarray analysis of prevaccination PBMCs would be useful for predicting patient prognosis after personalized peptide vaccination. When a stepwise discriminant analysis method was used to choose a gene set from the 23 probes differentially expressed in the prevaccination PBMCs, a combination of 4 genes, LRRN3, PCDH17, HIST1H4C, and PGLYRP1, gave the best prediction of short-term survivors, with a sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of 85%, 75%, 77%, 83%, and 80%, respectively (Table 6). Importantly, when this 4-gene classifier was used in 13 new independent cancer patients as a validation test, prognosis was correctly predicted in 12 of the 13 patients with a sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of 100%, 83%, 88%, 100%, and 92%, respectively, for the prediction of short-term survival (Table 6).

Table 6. Selection of a Gene Classifier for Predicting Short-Term Survival
Training/TestSensitivity (%)Specificity (%)Positive Predictive Value (%)Negative Predictive Value (%)Accuracy (%)
Training, n = 4017/20 (85)15/20 (75)17/22 (77)15/18 (83)32/40 (80)
Test, n = 137/7 (100)5/6 (83)7/8 (88)5/5 (100)12/13 (92)

Increase in the Prevaccination Plasma IL-6 Levels in the Patients With Poor Prognosis

Expression of cytokines, chemokines, and growth factors, which may result from proinflammatory and/or anti-inflammatory tumor microenvironments, gives a broad picture of the immunological status of cancer patients.32-35 We therefore examined the levels of these soluble factors using a bead-based multiplex assay with prevaccination plasma samples from the long-term and short-term survivors. As shown in Figure 3, the plasma levels of proinflammatory cytokine IL-6 were significantly higher in the short-term survivors than in the long-term survivors (P = .009). However, the plasma levels of other cytokines, chemokines, or growth factors, including IL-1Rα, IL-1β, IL-2, IL-2R, IL-4, IL-5, IL-7, IL-8, IL-10, IL-12, IL-13, IL-15, IL-17, IFN-α, IFN-γ, TNF-α, G-CSF, GM-CSF, IP-10, RANTES, Eotaxin, MIP-1α, MIP-1β, MCP-1, MIG, VEGF, EGF, HGF, and basic FGF, were not significantly different between the 2 groups (data not shown).

thumbnail image

Figure 3. Increase in plasma interleukin (IL)-6 levels in the short-term survivors is shown. The levels of IL-6 assessed by bead-based multiplex assay in prevaccination plasma were compared between the short-term (n = 18) and long-term (n = 18) survivors. Box plots show median and interquartile range (IQR). The whiskers (vertical bars) are the lowest value within 1.5 × IQR of the lower quartile and the highest value within 1.5 × IQR of the upper quartile. Data not included between the whiskers were plotted as outliers with dots. Two-sided P value was calculated with Mann-Whitney test.

Download figure to PowerPoint

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. FUNDING SOURCES
  9. REFERENCES

The identification of biomarkers to predict clinical responses to treatment is a challenging but important issue for the development of individualized therapies.5-8 Although recent advances in high-throughput microarray technology have allowed gene expression profiling for subclassifications of patients in a variety of fields, including organ transplantation and autoimmune diseases,18-20 little information is available regarding gene expression profiles in peripheral blood of patients treated with immunotherapies. In the current study, to identify promising biomarkers that are predictive of patient prognosis after personalized peptide vaccination, we examined gene expression profiles in PBMCs from 40 advanced castration-resistant prostate cancer patients who showed good or poor prognosis after personalized peptide vaccination. Our DNA microarray analysis in PBMCs identified distinctive genes that were differentially expressed between the long-term and short-term survivors. Interestingly, a statistical prediction model provided a 4-gene classifier that was able to predict patient prognosis with an accuracy of 92% in a validation test, suggesting that the identification of suitable patients for cancer vaccines may be possible with the profiling of a modest number of genes in peripheral blood samples. Because there were no significant differences in the other clinical and pathological features of the patients enrolled in the current study, except for the number of vaccinations and overall survival, our findings seem to be quite informative for the further development of cancer vaccines.

In the current study, 4 genes, LRRN3, PCDH17, HIST1H4C, and PGLYRP1, were selected as the best combination for prediction of patient prognosis. LRRN3 gene encodes a highly conserved transmembrane protein with multiple leucine-rich repeats, which is abundantly expressed in the developing and adult central nervous system. Polymorphisms in this gene were reported to be associated with autism spectrum disorder susceptibility.36 PCDH17 is 1 of the cadherin superfamily genes and is expressed predominantly in the nervous system. This molecule was reported to be a tumor suppressor gene candidate in squamous cell carcinomas.37 HIST1H4C gene encodes a member of the histone H4 family, which forms the nucleosome structure of the chromosomal fiber, and may play a central role in transcription regulation, DNA repair and replication, and chromosomal stability.38 PGLYRP1 gene encodes a pattern recognition receptor related to innate immunity against bacteria, which is expressed primarily in the granules of granulocytes.39 Although this information is available from the literature, little is known about the roles of these molecules in immune responses to cancer vaccines. Further studies remain to be done to elucidate them.

One of the most striking features of the differentially expressed genes is that many of the up-regulated genes in both prevaccination and postvaccination PBMCs from the short-term survivors were associated with gene signatures of granulocytes. This may possibly be reflected by the different frequencies of granulocytes in the PBMC fraction purified from peripheral whole blood on density gradient centrifugation using Ficoll-Paque. In healthy donors, normal granulocytes are usually separated from the PBMC fraction on Ficoll-Paque density gradient. However, patients with various types of cancers have been reported to show increased numbers of activated granulocytes in their peripheral blood, which are purified in the PBMC fraction.40-42 Recently, these abnormal granulocytes have been defined as granulocytic myeloid-derived suppressor cells, which express higher levels of inhibitory molecules, such as ARG1 and inducible nitric oxide synthase,41, 42 and impair the immunological functions of T cells and other immune cells.43-45 In addition, several studies have recently shown the critical roles for neutrophils, a main subset of granulocytes, in tumorigenesis.46 Neutrophils have a significant impact on the tumor microenvironment by producing cytokines, chemokines, and other products, such as reactive oxygen species and proteinases, which regulate inflammatory cell activation/recruitment, tumor cell proliferation, angiogenesis, and metastasis.47-49 For example, recent clinical studies have revealed that the presence of neutrophils in tumors was significantly associated with poor outcomes.50, 51 Unfortunately, because of the limited availability of blood samples, we have not fully characterized the granulocytes that were purified in the PBMC fraction, but it is highly possible that abnormal granulocytes in peripheral blood inhibit beneficial immune responses and lead to poor prognosis after peptide vaccines. The current study might provide a novel treatment approach capable of enhancing the clinical efficacy of cancer vaccines. Recently, chemotherapeutic drugs, such as gemcitabine and 5-fluorouracil, have been shown to selectively eliminate myeloid-derived suppressor cells in mice.52, 53 In addition, targeting of VEGF-mediated signaling using a tyrosine kinase inhibitor, sunitinib, has been reported to block expansion of CD15+CD14 granulocytic myeloid-derived suppressor cells in patients with renal cell cancers.54 It would thus be possible that accompanying treatments with such chemotherapeutic or molecularly targeted drugs before providing cancer vaccines suppress the gene signatures related to poor prognosis and improve patient outcomes after personalized peptide vaccination.

In addition to the granulocyte-related genes, other interesting genes were also differentially expressed between the long-term and short-term survivors. For example, leukocyte-associated immunoglobulin-like receptor 2 (LAIR2), a member of the immunoglobulin superfamily, was down-regulated in the prevaccination PBMCs of short-term survivors. Although not well studied, this molecule has been suggested to function as a proinflammatory mediator by suppressing the homologous immune inhibitor, leukocyte-associated immunoglobulin-like receptor 1 (LAIR-1), which is present on several types of mononuclear leukocytes.55 In addition, another noticeable finding is that several erythroid-specific genes, such as hemoglobin families (HBQ1, HBM, HBD), ALAS2, GYPE, EPB42, HP, and ERAF, were up-regulated in the postvaccination PBMCs of short-term survivors. The precise roles of these differentially expressed genes in immune responses to cancer vaccines need to be determined.

Interestingly, when the gene expression profiles in PBMCs were compared between before and after personalized peptide vaccination, many of the differentially expressed genes in prevaccination and/or postvaccination PBMCs, including granulocyte-related and erythroid-related genes, were up-regulated after personalized peptide vaccination in the short-term survivors, but not in the long-term survivors. This finding may be explained by the possibility that induction of granulocyte and erythroid gene signatures may be prevented by personalized peptide vaccination in the long-term survivors.

It should also be noted that the levels of the proinflammatory cytokine IL-6 in prevaccination plasma were significantly elevated in the short-term survivors. IL-6 is a multifunctional cytokine that regulates various aspects of immune responses, acute phase reactions, and hematopoiesis. In particular, IL-6 has been reported to be deeply involved in inflammation associated with cancer development and progression.34 There have been many studies describing the correlation between IL-6 levels and prognosis in various types of cancers, including prostate cancer.56-59 Interestingly, IL-6 has been also shown to rapidly generate myeloid-derived suppressor cells from precursors that are present in murine and human bone marrow or PBMCs, in the presence of other cytokines such as GM-CSF,60, 61 although in the current study, the expression levels of plasma IL-6 were not well correlated with expressions of granulocyte-related genes in the microarray analysis (data not shown). Although the role of IL-6 in the immune responses to cancer vaccines still remains to be clarified, it is possible that the blockage of IL-6 signaling would be beneficial for enhancing the therapeutic efficacy of cancer vaccines.

To the best of our knowledge, this is the first study to characterize gene expression profiles in peripheral blood and thereby identify biomarkers for predicting clinical outcomes after peptide vaccines. Our findings suggest that the widely available gene expression profiling in peripheral blood may permit future development of molecular-based personalized immunotherapies through discrimination between patients with good and poor prognoses. Although our experimental approaches were not novel, the ability to predict patient prognosis on the basis of relatively simple assays with easily available peripheral blood samples would be of importance. It may be possible that the current study would provide important information for defining eligibility and/or exclusion criteria for personalized peptide vaccination in castration-resistant prostate cancer patients. Nevertheless, because this is a retrospective study with a limited number of patients, all of whom received personalized peptide vaccination, clinical utility of the identified gene signatures and gene classifier needs to be confirmed in future larger-scale, prospective trials conducted in defined patient populations receiving or not receiving personalized peptide vaccination. In addition, the gene expression profiles identified in the current study remain to be verified by using other, independent methods for mRNA and/or protein quantification.

Acknowledgements

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. FUNDING SOURCES
  9. REFERENCES

We thank Drs. Hiromitsu Araki and Kaori Yasuda (Cell Innovator, Inc.) for their helpful discussion and technical help.

FUNDING SOURCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. FUNDING SOURCES
  9. REFERENCES

This study was supported by the grant, Regional Innovation Cluster Program of the Ministry of Education, Culture, Sports, Science, and Technology of Japan (to K.I.).

CONFLICT OF INTEREST DISCLOSURES

The authors made no disclosures.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. FUNDING SOURCES
  9. REFERENCES