Spatial Profiling of the Prostate Cancer Tumor Microenvironment Reveals Multiple Differences in Gene Expression and Correlation with Recurrence Risk.

The tumor microenvironment plays a crucial role in both the development and progression of prostate cancer. Furthermore, identifying protein and gene expression differences between different regions is valuable for treatment development. We applied Digital Spatial Profiling multiplex analysis to formalin-fixed paraffin embedded prostatectomy tissue blocks to investigate protein and transcriptome differences between tumor, tumor-adjacent stroma (TAS), CD45+ tumor, and CD45+ TAS tissue. Differential expression of an immunology/oncology protein panel (n = 58) was measured. OX40L and CTLA4 were expressed at higher levels while 22 other proteins, including CD11c, were expressed at lower levels (FDR < 0.2 and p-value < 0.05) in TAS as compared to tumor epithelia. A tissue microarray analysis of 97 patients with 1547 cores found positive correlations between high expression of CD11c and increased time to recurrence in tumor and TAS, and inverse relationships for CTLA4 and OX40L, where higher expression in tumor correlated with lower time to recurrence, but higher time to recurrence in TAS. Spatial transcriptomic analysis using a Cancer Transcriptome Atlas panel (n = 1825 genes) identified 162 genes downregulated and 69 upregulated in TAS versus tumor, 26 downregulated and 6 upregulated in CD45+ TAS versus CD45+ tumor. We utilized CIBERSORTx to estimate the relative immune cell fractions using CD45+ gene expression and found higher average fractions for memory B, naïve B, and T cells in TAS. In summary, the combination of protein expression differences, immune cell fractions, and correlations of protein expression with time to recurrence suggest that closely examining the tumor microenvironment provides valuable data that can improve prognostication and treatment techniques.

Expression of Endogenous Retroviral RNA in Prostate Tumors has Prognostic Value and Shows Differences among Americans of African Versus European/Middle Eastern Ancestry.

Endogenous retroviruses (ERVs) are abundant, repetitive elements dispersed across the human genome and are implicated in various diseases. We investigated two potential roles for ERVs in prostate cancer (PCa). First, the PCa of Black Americans (BA) is diagnosed at an earlier median age and at a more advanced stage than the PCa of White Americans (WA). We used publicly available RNA-seq data from tumor-enriched samples of 27 BA and 65 WA PCa patients in order to identify 12 differentially expressed ERVs (padj < 0.1) and used a tissue microarray of the PCa cores from an independent set of BA and WA patients to validate the differential protein expression of one of these ERVs, ERV3-1 (p = 2.829 × 10-7). Second, we used 57 PCa tumors from patients of all ancestries from one hospital as a training set to identify the ERVs associated with time to biochemical relapse. A 29-ERV prognostic panel was then tested and validated on 35 separate PCa tumors from patients obtained in two different hospitals with a dramatic increase in prognostic power relative to clinical parameters alone (p = 7.4 × 10-11). In summary, ERV RNA expression differences in the prostate tumors of patients of different ancestries may be associated with dissimilarities in the mechanism of cancer progression. In addition, the correlation of expression of certain ERVs in prostate tumors with the risk of biochemical relapse indicates a possible role for ERV expression in cancer progression.

RNA expression differences in prostate tumors and tumor-adjacent stroma between Black and White Americans.

Prostate cancer (PCa) in Black Americans (BA) is diagnosed at an earlier median age and a more advanced stage than PCa in White Americans (WA). Tumor-adjacent stroma (TAS) plays a critical role in tumorigenesis of prostate cancer. We examined RNA expression in both tumor and TAS of BA compared to WA. After evaluating the geographical ancestry of each sample, preliminary analysis of our own RNA-seq data of 7 BA and 7 WA TAS revealed 1706 downregulated and 1844 upregulated genes in BA relative to WA PCa patients (p adj < 0.05). An assessment of published RNA-seq data of clinically matched tumor-enriched tissues from 15 BA and 30 WA patients revealed 932 upregulated and 476 downregulated genes in BA relative to WA (p adj < 0.05). When TAS and tumor epithelial cohorts were compared for the top 2500 statistically significant genes, immune responses were downregulated in BA vs WA TAS, while T cell-exhaustion pathways and the immune checkpoint gene CTLA4 were upregulated in BA vs WA tumors. We found fewer activated dendritic cells in tumor and more CD8 T-cells in TAS of BA versus WA PCa patients. Further characterization of these differences in the immune response of PCa patients of distinct geographical ancestry could help to improve diagnostics, prognostics, and therapy.

Transcriptome Analysis of Ovarian and Uterine Clear Cell Malignancies.

PURPOSE: Ovarian and uterine clear cell carcinomas (CCCs) are rare but associated with poor prognosis. This study explored RNA transcription patterns characteristic of these tumors. EXPERIMENTAL DESIGN: RNA sequencing (RNA-seq) of 11 ovarian CCCs and five uterine CCCs was performed and compared to publicly available data from high grade serous ovarian cancers (HGSOCs). Ingenuity Pathway Analyses were performed. CIBERSORT analyses estimated relative fractions of 22 immune cell types in each RNA-seq sample. Sequencing data was correlated with PD-L1 immunohistochemical expression. RESULTS: RNA-seq revealed 1,613 downregulated and 1,212 upregulated genes (corrected p < 0.05, |FC |≥10) in ovarian CCC versus HGSOC. Two subgroups were identified in the ovarian CCC, characterized by ethnicity and expression differences in ARID1A. There were 3,252 differentially expressed genes between PD-L1+/- ovarian CCCs, revealing immune response, cell death, and DNA repair networks, negatively correlated with PD-L1 expression, whereas cellular proliferation networks positively correlated with expression. In clear cell ovarian versus clear cell uterine cancer, 1,607 genes were significantly upregulated, and 109 genes were significantly downregulated (corrected p < 0.05, |FC|≥10). Comparative pathway analysis of late and early stage ovarian CCCs revealed unique metabolic and PTEN pathways, whereas uterine CCCs had unique Wnt/Ca+, estrogen receptor, and CCR5 signaling. CIBERSORT analysis revealed that activated mast cells and regulatory T cell populations were relatively enriched in uterine CCCs. The PD-L1+ ovarian CCCs had enriched resting NK cells and memory B cell populations, while PD-L1- had enriched CD8 T-cells, monocytes, eosinophils, and activated dendritic cells. CONCLUSIONS: Unique transcriptional expression profiles distinguish clear cell uterine and ovarian cancers from each other and from other more common histologic subtypes. These insights may aid in devising novel therapeutics.

Transcriptional Profiling of Age-Associated Gene Expression Changes in Human Circulatory CD1c+ Myeloid Dendritic Cell Subset.

Immune dysfunction is a hallmark of aging and is thought to be responsible for the age-associated diseases. Dendritic cells (DCs) of the immune system function as initiators and regulators of the immune responses. Recent studies have highlighted the division of labor between various DC subsets. CD1c+ DC subset has emerged as a major inducer of CD4 T cell response. There is a scarcity of information regarding the age-associated changes in the functions of DC subsets in the elderly. Here, we investigated the changes in transcriptional profile of CD1c+ DC subset from healthy aged and young individuals using RNA sequencing. Our results suggest that majority of the genes in DCs are upregulated with age. Glucose transport, GPCR, and potassium channel genes are all upregulated in DCs from aged as compared to young indicating an enhanced activation state of DCs from aged individuals. The expression of histones, small nucleolar RNA H/ACA box (SNORA) and small nucleolar RNA C/D/box (SNORD), and long non-coding RNA (lncRNA) is also substantially upregulated in the DCs from aged. In contrast, the antigen-presenting and energy generating pathways are downregulated. In summary, DCs from aged subjects display an activated state coupled with reduced antigen presentation which may be responsible for age-associate immune dysfunction.

Assessing Researcher Needs for a Virtual Biobank.

INTRODUCTION: Biosamples and associated clinical data accelerate translational and clinical research discoveries. A lack of high quality biosamples both stalls projects and limits research advances. In this study, we targeted a wide audience of University of California (UC) biobanking stakeholders who were either involved with the collection or the utilization of biosamples to assess the scope of their biobanking activities and their interest in virtual biobanking or cooperating in the formation of the UC-wide biorepository. MATERIALS AND METHODS: Each institutional review board from the five UC medical campuses' provided a dataset of potential researchers involved with biobanking. Once identified, a brief six item web-based questionnaire was administered electronically to these researchers. RESULTS: Most survey participants (80%) responded "yes" (n = 348) that they were actively collecting biosamples for research, and 68% of participants indicated they would either definitely (30%, n = 131) or maybe (38%, n = 166) request biosample materials now or within the next year. An equal proportion of participants responded yes (42% or n = 184) and maybe (42% or n = 182) when asked if they would voluntarily contribute specimens to a UC-wide virtual biobank. DISCUSSION: The results presented above show high levels of support among UC biobanking stakeholders for both requesting material from and contributing material to a UC-wide virtual biobank. In addition, a considerable number of individual researchers on our five UC medical campuses are conducting biospecimen research (i.e., well over n = 435 respondents).

Flavokawain A induces deNEDDylation and Skp2 degradation leading to inhibition of tumorigenesis and cancer progression in the TRAMP transgenic mouse model.

S phase kinase-associated protein 2 (Skp2) has been shown to be required for spontaneous tumor development that occurs in the retinoblastoma protein (pRb) deficient mice. Here we have demonstrated that flavokawain A (FKA), a novel chalcone from the kava plant, selectively inhibited the growth of pRb deficient cell lines and resulted in a proteasome-dependent and ubiquitination-mediated Skp2 degradation. Degradation of Skp2 by FKA was found to be involved in a functional Cullin1, but independent of Cdh1 expression. Further studies have demonstrated that FKA docked into the ATP binding pocket of the precursor cell-expressed developmentally down-regulated 8 (NEDD8)-activating enzyme (NAE) complex, inhibited NEDD8 conjugations to both Cullin1 and Ubc12 in PC3 cells and Ubc12 NEDDylation in an in vitro assay. Finally, dietary feeding of the autochthonous transgenic adenocarcinoma of the mouse prostate (TRAMP) mice with FKA inhibited the formation of high-grade prostatic intra-epithelial neoplasia lesions (HG-PIN) and prostate adenocarcinomas, reduced the tumor burden and completely abolished distant organ metastasis. Immunohistochemistry studies revealed that dietary FKA feeding resulted in marked anti-proliferative and apoptotic effects via down-regulation of Skp2 and NEDD8 and up-regulation of p27/Kip1 in the prostate of TRAMP mice. Our findings therefore provide evidence that FKA is a promising NEDDylation inhibitor for targeting Skp2 degradation in prostate cancer prevention and treatment.

Six stroma-based RNA markers diagnostic for prostate cancer in European-Americans validated at the RNA and protein levels in patients in China.

We previously analyzed human prostate tissue containing stroma near to tumor and from cancer-negative tissues of volunteers. Over 100 candidate gene expression differences were identified and used to develop a classifier that could detect nearby tumor with an accuracy of 97% (sensitivity = 98% and specificity = 88%) based on 364 independent test cases from primarily European American cases. These stroma-based gene signatures have the potential to identify cancer patients among those with negative biopsies. In this study, we used prostate tissues from Chinese cases to validate six of these markers (CAV1, COL4A2, HSPB1, ITGB3, MAP1A and MCAM). In validation by real-time PCR, four genes (COL4A2, HSPB1, ITGB3, and MAP1A) demonstrated significantly lower expression in tumor-adjacent stroma compared to normal stroma (p value ≤ 0.05). Next, we tested whether these expression differences could be extended to the protein level. In IHC assays, all six selected proteins showed lower expression in tumor-adjacent stroma compared to the normal stroma, of which COL4A2, HSPB1 and ITGB3 showed significant differences (p value ≤ 0.05). These results suggest that biomarkers for diagnosing prostate cancer based on tumor microenvironment may be applicable across multiple racial groups.

Overexpression of periostin in stroma positively associated with aggressive prostate cancer.

BACKGROUND: Periostin is an important extracellular matrix protein involved in cell development and adhesion. Previously, we identified periostin to be up-regulated in aggressive prostate cancer (CaP) using quantitative glycoproteomics and mass spectrometry. The expression of periostin was further evaluated in primary radical prostatectomy (RP) prostate tumors and adjacent non-tumorous prostate tissues using immunohistochemistry (IHC). Our IHC results revealed a low background periostin levels in the adjacent non-tumorous prostate tissues, but overexpressed periostin levels in the peritumoral stroma of primary CaP tumors. METHODS: In this study, periostin expression in CaP was further examined on multiple tissue microarrays (TMAs), which were conducted in four laboratories. To achieve consistent staining, all TMAs were stained with same protocol and scored by same image computation tool to determine the total periostin staining intensities. The TMAs were further scored by pathologists to characterize the stromal staining and epithelial staining. RESULTS: The periostin staining was observed mainly in peritumoral stromal cells and in some cases in tumor epithelial cells though the stronger staining was found in peritumoral stromal cells. Both periostin stromal staining and epithelial staining can differentiate BPH from CaP including low grade CaP (Gleason score ≤6), with significant p-value of 2.2e-16 and 0.001, respectively. Periostin epithelial staining differentiated PIN from low grade CaP (Gleason score ≤6) (p=0.001), while periostin stromal staining differentiated low grade Cap (Gleason score ≤6) from high grade Cap (Gleason score ≤6) (p=1.7e-05). In addition, a positive correlation between total periostin staining and Gleason score was observed (r=0.87, p=0.002). CONCLUSIONS: The results showed that periostin staining was positively correlated with increasing Gleason score and the aggressiveness of prostate disease.

Associations of prostate cancer risk variants with disease aggressiveness: results of the NCI-SPORE Genetics Working Group analysis of 18,343 cases.

Genetic studies have identified single nucleotide polymorphisms (SNPs) associated with the risk of prostate cancer (PC). It remains unclear whether such genetic variants are associated with disease aggressiveness. The NCI-SPORE Genetics Working Group retrospectively collected clinicopathologic information and genotype data for 36 SNPs which at the time had been validated to be associated with PC risk from 25,674 cases with PC. Cases were grouped according to race, Gleason score (Gleason ≤ 6, 7, ≥ 8) and aggressiveness (non-aggressive, intermediate, and aggressive disease). Statistical analyses were used to compare the frequency of the SNPs between different disease cohorts. After adjusting for multiple testing, only PC-risk SNP rs2735839 (G) was significantly and inversely associated with aggressive (OR = 0.77; 95 % CI 0.69-0.87) and high-grade disease (OR = 0.77; 95 % CI 0.68-0.86) in European men. Similar associations with aggressive (OR = 0.72; 95 % CI 0.58-0.89) and high-grade disease (OR = 0.69; 95 % CI 0.54-0.87) were documented in African-American subjects. The G allele of rs2735839 was associated with disease aggressiveness even at low PSA levels (<4.0 ng/mL) in both European and African-American men. Our results provide further support that a PC-risk SNP rs2735839 near the KLK3 gene on chromosome 19q13 may be associated with aggressive and high-grade PC. Future prospectively designed, case-case GWAS are needed to identify additional SNPs associated with PC aggressiveness.

The identification of trans-associations between prostate cancer GWAS SNPs and RNA expression differences in tumor-adjacent stroma.

Here we tested the hypothesis that SNPs associated with prostate cancer risk, might differentially affect RNA expression in prostate cancer stroma. The most significant 35 SNP loci were selected from Genome Wide Association (GWA) studies of ~40,000 patients. We also selected 4030 transcripts previously associated with prostate cancer diagnosis and prognosis. eQTL analysis was carried out by a modified BAYES method to analyze the associations between the risk variants and expressed transcripts jointly in a single model. We observed 47 significant associations between eight risk variants and the expression patterns of 46 genes. This is the first study to identify associations between multiple SNPs and multiple in trans gene expression differences in cancer stroma. Potentially, a combination of SNPs and associated expression differences in prostate stroma may increase the power of risk assessment for individuals, and for cancer progression.

A class of genes in the HER2 regulon that is poised for transcription in breast cancer cell lines and expressed in human breast tumors.

HER2-positive breast cancer accounts for 25% of all cases and has a poor prognosis. Although progress has been made in understanding signal transduction, little is known of how HER2 achieves gene regulation. We performed whole genome expression analysis on a HER2⁺ and HER2⁻ breast cancer cell lines and compared these results to expression in 812 primary tumors stratified by their HER2 expression level. Chip-on-chip with anti-RNA polymerase II was compared among breast cancer cell lines to identify genes that are potentially activated by HER2. The expression levels of these HER2-dependent POL II binding genes were determined for the 812 HER2+/- breast cancer tissues. Genes differentially expressed between HER2+/- cell lines were generally regulated in the same direction as in breast cancer tissues. We identified genes that had POLII binding in HER2⁺ cell lines, but without significant gene expression. Of 737 such genes "poised" for expression in cell lines, 113 genes were significantly differentially expressed in breast tumors in a HER2-dependent manner. Pathway analysis of these 113 genes revealed that a large group of genes were associated with stem cell and progenitor cell control as indicated by networks centered on NANOG, SOX2, OCT3/4. HER2 directs POL II binding to a large number of genes in breast cancer cells. A "poised" class of genes in HER2⁺ cell lines with POLII binding and low RNA expression but is differentially expressed in primary tumors, strongly suggests a role of the microenvironment and further suggests a role for stem cells proliferation in HER2-regulated breast cancer tissue.

When Anti-Aging Studies Meet Cancer Chemoprevention: Can Anti-Aging Agent Kill Two Birds with One Blow?

Recent evidence has strongly supported that the rate of aging is controlled, at least to some extent, by evolutionarily conserved nutrient sensing pathways (e.g. the insulin/IGF-1-signaling, mTOR, AMPK, and sirtuins) from worms to humans. These pathways are also commonly involved in carcinogenesis and cancer metabolism. Agents (e.g. metformin, resveratrol, and Rhodiola) that target these nutrient sensing pathways often have both anti-aging and anti-cancer efficacy. These agents not only reprogram energy metabolism of malignant cells, but also target normal postmitotic cells by suppressing their conversion into senescent cells, which confers systematic metabolism benefits. These agents are fundamentally different from chemotherapy (e.g. paclitaxel and doxorubicin) or radiation therapy that causes molecular damage (e.g. DNA and protein damages) and thereby no selection resistance may be expected. By reviewing molecular mechanisms of action, epidemiological evidence, experimental data in tumor models, and early clinical study results, this review provides information supporting the promising use of agents with both anti-aging and anti-cancer efficacy for cancer chemoprevention.

Role of the adjacent stroma cells in prostate cancer development and progression: synergy between TGF-ß and IGF signaling.

This review postulates the role of transforming growth factor-beta (TGF-ß) and insulin-like growth factor (IGF-I/IGF-II) signaling in stromal cells during prostate carcinogenesis and progression. It is known that stromal cells have a reciprocal relationship to the adjacent epithelial cells in the maintenance of structural and functional integrity of the prostate. An interaction between TGF-ß and IGF signaling occupies a central part in this stromal-epithelial interaction. An increase in TGF-ß and IGF signaling will set off the imbalance of this relationship and will lead to cancer development. A continuous input from TGF-ß and IGF in the tumor microenvironment will result in cancer progression. Understanding of these events can help prevention, diagnosis, and therapy of prostate cancer.

Wnt signaling in castration-resistant prostate cancer: implications for therapy.

Increasing evidence has indicated that Wnt signaling plays complex roles in castration resistant prostate cancer (CRPC). Although not all data were consistent, ß-catenin nuclear localization and its co-localization with androgen receptor (AR) were more frequently observed in CRPC compared to hormone naïve prostate cancer. This direct interaction between AR and ß-catenin seemed to elicit a specific expression of a set of target genes in low androgen conditions in CRPC. Paracrine Wnt signaling also was shown to aid resistance to chemotherapy and androgen deprivation therapy. Results from the next generation sequencing studies (i.e. RNA-seq and whole exosome sequcing) of CRPC specimens have identified the Wnt pathway as one of the top signaling pathways with significant genomic alterations in CRPC, whereas, Wnt pathway alterations were virtually absent in hormone naïve primary prostate cancer. Furthermore, Wnt signaling has been suggested to play an important role in cancer stem cell functions in prostate cancer recurrence and resistance to androgen deprivation therapy. Therefore, in this review we have summarized existing knowledge regarding potential roles of Wnt signaling in CRPC and underline Wnt signaling as a potential therapeutic target for CRPC. Further understanding of Wnt signaling in castration resistance may eventually contribute new insights into possible treatment options for this incurable disease.

The transcription factor EGR1 localizes to the nucleolus and is linked to suppression of ribosomal precursor synthesis.

EGR1 is an immediate early gene with a wide range of activities as transcription factor, spanning from regulation of cell growth to differentiation. Numerous studies show that EGR1 either promotes the proliferation of stimulated cells or suppresses the tumorigenic growth of transformed cells. Upon interaction with ARF, EGR1 is sumoylated and acquires the ability to bind to specific targets such as PTEN and in turn to regulate cell growth. ARF is mainly localized to the periphery of nucleolus where is able to negatively regulate ribosome biogenesis. Since EGR1 colocalizes with ARF under IGF-1 stimulation we asked the question of whether EGR1 also relocate to the nucleolus to interact with ARF. Here we show that EGR1 colocalizes with nucleolar markers such as fibrillarin and B23 in the presence of ARF. Western analysis of nucleolar extracts from HeLa cells was used to confirm the presence of EGR1 in the nucleolus mainly as the 100 kDa sumoylated form. We also show that the level of the ribosomal RNA precursor 47S is inversely correlated to the level of EGR1 transcripts. The EGR1 iseffective to regulate the synthesis of the 47S rRNA precursor. Then we demonstrated that EGR1 binds to the Upstream Binding Factor (UBF) leading us to hypothesize that the regulating activity of EGR1 is mediated by its interaction within the transcriptional complex of RNA polymerase I. These results confirm the presence of EGR1 in the nucleolus and point to a role for EGR1 in the control of nucleolar metabolism.

Generation of "virtual" control groups for single arm prostate cancer adjuvant trials.

It is difficult to construct a control group for trials of adjuvant therapy (Rx) of prostate cancer after radical prostatectomy (RP) due to ethical issues and patient acceptance. We utilized 8 curve-fitting models to estimate the time to 60%, 65%, … 95% chance of progression free survival (PFS) based on the data derived from Kattan post-RP nomogram. The 8 models were systematically applied to a training set of 153 post-RP cases without adjuvant Rx to develop 8 subsets of cases (reference case sets) whose observed PFS times were most accurately predicted by each model. To prepare a virtual control group for a single-arm adjuvant Rx trial, we first select the optimal model for the trial cases based on the minimum weighted Euclidean distance between the trial case set and the reference case set in terms of clinical features, and then compare the virtual PFS times calculated by the optimum model with the observed PFSs of the trial cases by the logrank test. The method was validated using an independent dataset of 155 post-RP patients without adjuvant Rx. We then applied the method to patients on a Phase II trial of adjuvant chemo-hormonal Rx post RP, which indicated that the adjuvant Rx is highly effective in prolonging PFS after RP in patients at high risk for prostate cancer recurrence. The method can accurately generate control groups for single-arm, post-RP adjuvant Rx trials for prostate cancer, facilitating development of new therapeutic strategies.

Expression differences between African American and Caucasian prostate cancer tissue reveals that stroma is the site of aggressive changes.

In prostate cancer, race/ethnicity is the highest risk factor after adjusting for age. African Americans have more aggressive tumors at every clinical stage of the disease, resulting in poorer prognosis and increased mortality. A major barrier to identifying crucial gene activity differences is heterogeneity, including tissue composition variation intrinsic to the histology of prostate cancer. We hypothesized that differences in gene expression in specific tissue types would reveal mechanisms involved in the racial disparities of prostate cancer. We examined 17 pairs of arrays for AAs and Caucasians that were formed by closely matching the samples based on the known tissue type composition of the tumors. Using pair-wise t-test we found significantly altered gene expression between AAs and CAs. Independently, we performed multiple linear regression analyses to associate gene expression with race considering variation in percent tumor and stroma tissue. The majority of differentially expressed genes were associated with tumor-adjacent stroma rather than tumor tissue. Extracellular matrix, integrin family and signaling mediators of the epithelial-to-mesenchymal transition (EMT) pathways were all downregulated in stroma of AAs. Using MetaCore (GeneGo) analysis, we observed that 35% of significant (p < 10(-3)) pathways identified EMT and 25% identified immune response pathways especially for interleukins-2, -4, -5, -6, -7, -10, -13, -15 and -22 as the major changes. Our studies reveal that altered immune and EMT processes in tumor-adjacent stroma may be responsible for the aggressive nature of prostate cancer in AAs.

A gradient boosting algorithm for survival analysis via direct optimization of concordance index.

Survival analysis focuses on modeling and predicting the time to an event of interest. Many statistical models have been proposed for survival analysis. They often impose strong assumptions on hazard functions, which describe how the risk of an event changes over time depending on covariates associated with each individual. In particular, the prevalent proportional hazards model assumes that covariates are multiplicatively related to the hazard. Here we propose a nonparametric model for survival analysis that does not explicitly assume particular forms of hazard functions. Our nonparametric model utilizes an ensemble of regression trees to determine how the hazard function varies according to the associated covariates. The ensemble model is trained using a gradient boosting method to optimize a smoothed approximation of the concordance index, which is one of the most widely used metrics in survival model performance evaluation. We implemented our model in a software package called GBMCI (gradient boosting machine for concordance index) and benchmarked the performance of our model against other popular survival models with a large-scale breast cancer prognosis dataset. Our experiment shows that GBMCI consistently outperforms other methods based on a number of covariate settings. GBMCI is implemented in R and is freely available online.