Molecular classification and biomarkers of outcome with immunotherapy in extensive-stage small-cell lung cancer: analyses of the CASPIAN phase 3 study.

BACKGROUND: We explored potential predictive biomarkers of immunotherapy response in patients with extensive-stage small-cell lung cancer (ES-SCLC) treated with durvalumab (D) + tremelimumab (T) + etoposide-platinum (EP), D + EP, or EP in the randomized phase 3 CASPIAN trial. METHODS: 805 treatment-naïve patients with ES-SCLC were randomized (1:1:1) to receive D + T + EP, D + EP, or EP. The primary endpoint was overall survival (OS). Patients were required to provide an archived tumor tissue block (or ≥ 15 newly cut unstained slides) at screening, if these samples existed. After assessment for programmed cell death ligand-1 expression and tissue tumor mutational burden, residual tissue was used for additional molecular profiling including by RNA sequencing and immunohistochemistry. RESULTS: In 182 patients with transcriptional molecular subtyping, OS with D ± T + EP was numerically highest in the SCLC-inflamed subtype (n = 10, median 24.0 months). Patients derived benefit from immunotherapy across subtypes; thus, additional biomarkers were investigated. OS benefit with D ± T + EP versus EP was greater with high versus low CD8A expression/CD8 cell density by immunohistochemistry, but with no additional benefit with D + T + EP versus D + EP. OS benefit with D + T + EP versus D + EP was associated with high expression of CD4 (median 25.9 vs. 11.4 months) and antigen-presenting and processing machinery (25.9 vs. 14.6 months) and MHC I and II (23.6 vs. 17.3 months) gene signatures, and with higher MHC I expression by immunohistochemistry. CONCLUSIONS: These findings demonstrate the tumor microenvironment is important in mediating better outcomes with D ± T + EP in ES-SCLC, with canonical immune markers associated with hypothesized immunotherapy mechanisms of action defining patient subsets that respond to D ± T. TRIAL REGISTRATION: ClinicalTrials.gov, NCT03043872.

Durvalumab ± Tremelimumab + Platinum-Etoposide in Extensive-Stage Small Cell Lung Cancer (CASPIAN): Outcomes by PD-L1 Expression and Tissue Tumor Mutational Burden.

PURPOSE: In the CASPIAN trial, first-line durvalumab plus platinum-etoposide (EP) significantly improved overall survival (OS) versus EP alone in extensive-stage small cell lung cancer (ES-SCLC). We report exploratory analyses of CASPIAN outcomes by programmed cell death ligand-1 (PD-L1) expression and tissue tumor mutational burden (tTMB). EXPERIMENTAL DESIGN: Patients were randomized (1:1:1) to durvalumab (1,500 mg) plus EP, durvalumab plus tremelimumab (75 mg) plus EP, or EP alone. Treatment effects in PD-L1 and tTMB subgroups were estimated using an unstratified Cox proportional hazards model. RESULTS: The PD-L1 and tTMB biomarker-evaluable populations (BEP) comprised 54.4% (438/805) and 35.2% (283/805) of the intention-to-treat population, respectively. PD-L1 prevalence was low: 5.7%, 25.8%, and 28.3% had PD-L1 expression on ≥1% tumor cells (TC), ≥1% immune cells (IC), and ≥1% TCs or ICs, respectively. OS benefit with durvalumab plus EP versus EP was similar across PD-L1 subgroups, with HRs all falling within the 95% confidence interval (CI) for the PD-L1 BEP (0.47‒0.79). OS benefit with durvalumab plus tremelimumab plus EP versus EP was greater in PD-L1 ≥1% versus <1% subgroups, although CIs overlapped. There was no evidence of an interaction between tTMB and treatment effect on OS (durvalumab plus EP vs. EP, P = 0.916; durvalumab plus tremelimumab plus EP vs. EP, P = 0.672). CONCLUSIONS: OS benefit with first-line durvalumab plus EP in patients with ES-SCLC was observed regardless of PD-L1 or tTMB status. PD-L1 expression may prove to be a useful biomarker for combined treatment with PD-(L)1 and CTLA-4 inhibition, although this requires confirmation with an independent dataset. See related commentary by Rolfo and Russo, p. 652.

Landscape of homologous recombination deficiencies in solid tumours: analyses of two independent genomic datasets.

BACKGROUND: DNA repair deficiencies are characteristic of cancer and homologous recombination deficiency (HRD) is the most common. HRD sensitizes tumour cells to PARP inhibitors so it is important to understand the landscape of HRD across different solid tumour types. METHODS: Germline and somatic BRCA mutations in breast and ovarian cancers were evaluated using sequencing data from The Cancer Genome Atlas (TCGA) database. Secondly, a larger independent genomic dataset was analysed to validate the TCGA results and determine the frequency of germline and somatic mutations across 15 different candidate homologous recombination repair (HRR) genes, and their relationship with the genetic events of bi-allelic loss, loss of heterozygosity (LOH) and tumour mutation burden (TMB). RESULTS: Approximately one-third of breast and ovarian cancer BRCA mutations were somatic. These showed a similar degree of bi-allelic loss and clinical outcomes to germline mutations, identifying potentially 50% more patients that may benefit from precision treatments. HRR mutations were present in sizable proportions in all tumour types analysed and were associated with high TMB and LOH scores. We also identified numerous BRCA reversion mutations across all tumour types. CONCLUSIONS: Our results will facilitate future research into the efficacy of precision oncology treatments, including PARP and immune checkpoint inhibitors.

The clear cell sarcoma functional genomic landscape.

Clear cell sarcoma (CCS) is a deadly malignancy affecting adolescents and young adults. It is characterized by reciprocal translocations resulting in expression of the chimeric EWSR1-ATF1 or EWSR1-CREB1 fusion proteins, driving sarcomagenesis. Besides these characteristics, CCS has remained genomically uncharacterized. Copy number analysis of human CCSs showed frequent amplifications of the MITF locus and chromosomes 7 and 8. Few alterations were shared with Ewing sarcoma or desmoplastic, small round cell tumors, which are other EWSR1-rearranged tumors. Exome sequencing in mouse tumors generated by expression of EWSR1-ATF1 from the Rosa26 locus demonstrated no other repeated pathogenic variants. Additionally, we generated a new CCS mouse by Cre-loxP-induced chromosomal translocation between Ewsr1 and Atf1, resulting in copy number loss of chromosome 6 and chromosome 15 instability, including amplification of a portion syntenic to human chromosome 8, surrounding Myc. Additional experiments in the Rosa26 conditional model demonstrated that Mitf or Myc can contribute to sarcomagenesis. Copy number observations in human tumors and genetic experiments in mice rendered, for the first time to our knowledge, a functional landscape of the CCS genome. These data advance efforts to understand the biology of CCS using innovative models that will eventually allow us to validate preclinical therapies necessary to achieve longer and better survival for young patients with this disease.

STAT3 Antisense Oligonucleotide Remodels the Suppressive Tumor Microenvironment to Enhance Immune Activation in Combination with Anti-PD-L1.

PURPOSE: Danvatirsen is a therapeutic antisense oligonucleotide (ASO) that selectively targets STAT3 and has shown clinical activity in two phase I clinical studies. We interrogated the clinical mechanism of action using danvatirsen-treated patient samples and conducted back-translational studies to further elucidate its immunomodulatory mechanism of action. EXPERIMENTAL DESIGN: Paired biopsies and blood samples from danvatirsen-treated patients were evaluated using immunohistochemistry and gene-expression analysis. To gain mechanistic insight, we used mass cytometry, flow cytometry, and immunofluorescence analysis of CT26 tumors treated with a mouse surrogate STAT3 ASO, and human immune cells were treated in vitro with danvatirsen. RESULTS: Within the tumors of treated patients, danvatirsen uptake was observed mainly in cells of the tumor microenvironment (TME). Gene expression analysis comparing baseline and on-treatment tumor samples showed increased expression of proinflammatory genes. In mouse models, STAT3 ASO demonstrated partial tumor growth inhibition and enhanced the antitumor activity when combined with anti-PD-L1. Immune profiling revealed reduced STAT3 protein in immune and stromal cells, and decreased suppressive cytokines correlating with increased proinflammatory macrophages and cytokine production. These changes led to enhanced T-cell abundance and function in combination with anti-PD-L1. CONCLUSIONS: STAT3 ASO treatment reverses a suppressive TME and promotes proinflammatory gene expression changes in patients' tumors and mouse models. Preclinical data provide evidence that ASO-mediated inhibition of STAT3 in the immune compartment is sufficient to remodel the TME and enhance the activity of checkpoint blockade without direct STAT3 inhibition in tumor cells. Collectively, these data provide a rationale for testing this combination in the clinic.

Establishing guidelines to harmonize tumor mutational burden (TMB): in silico assessment of variation in TMB quantification across diagnostic platforms: phase I of the Friends of Cancer Research TMB Harmonization Project.

BACKGROUND: Tumor mutational burden (TMB), defined as the number of somatic mutations per megabase of interrogated genomic sequence, demonstrates predictive biomarker potential for the identification of patients with cancer most likely to respond to immune checkpoint inhibitors. TMB is optimally calculated by whole exome sequencing (WES), but next-generation sequencing targeted panels provide TMB estimates in a time-effective and cost-effective manner. However, differences in panel size and gene coverage, in addition to the underlying bioinformatics pipelines, are known drivers of variability in TMB estimates across laboratories. By directly comparing panel-based TMB estimates from participating laboratories, this study aims to characterize the theoretical variability of panel-based TMB estimates, and provides guidelines on TMB reporting, analytic validation requirements and reference standard alignment in order to maintain consistency of TMB estimation across platforms. METHODS: Eleven laboratories used WES data from The Cancer Genome Atlas Multi-Center Mutation calling in Multiple Cancers (MC3) samples and calculated TMB from the subset of the exome restricted to the genes covered by their targeted panel using their own bioinformatics pipeline (panel TMB). A reference TMB value was calculated from the entire exome using a uniform bioinformatics pipeline all members agreed on (WES TMB). Linear regression analyses were performed to investigate the relationship between WES and panel TMB for all 32 cancer types combined and separately. Variability in panel TMB values at various WES TMB values was also quantified using 95% prediction limits. RESULTS: Study results demonstrated that variability within and between panel TMB values increases as the WES TMB values increase. For each panel, prediction limits based on linear regression analyses that modeled panel TMB as a function of WES TMB were calculated and found to approximately capture the intended 95% of observed panel TMB values. Certain cancer types, such as uterine, bladder and colon cancers exhibited greater variability in panel TMB values, compared with lung and head and neck cancers. CONCLUSIONS: Increasing uptake of TMB as a predictive biomarker in the clinic creates an urgent need to bring stakeholders together to agree on the harmonization of key aspects of panel-based TMB estimation, such as the standardization of TMB reporting, standardization of analytical validation studies and the alignment of panel-based TMB values with a reference standard. These harmonization efforts should improve consistency and reliability of panel TMB estimates and aid in clinical decision-making.

Multifactorial Deep Learning Reveals Pan-Cancer Genomic Tumor Clusters with Distinct Immunogenomic Landscape and Response to Immunotherapy.

PURPOSE: Tumor genomic features have been of particular interest because of their potential impact on the tumor immune microenvironment and response to immunotherapy. Due to the substantial heterogeneity, an integrative approach incorporating diverse molecular features is needed to characterize immunologic features underlying primary resistance to immunotherapy and for the establishment of novel predictive biomarkers. EXPERIMENTAL DESIGN: We developed a pan-cancer deep machine learning model integrating tumor mutation burden, microsatellite instability, and somatic copy-number alterations to classify tumors of different types into different genomic clusters, and assessed the immune microenvironment in each genomic cluster and the association of each genomic cluster with response to immunotherapy. RESULTS: Our model grouped 8,646 tumors of 29 cancer types from The Cancer Genome Atlas into four genomic clusters. Analysis of RNA-sequencing data revealed distinct immune microenvironment in tumors of each genomic class. Furthermore, applying this model to tumors from two melanoma immunotherapy clinical cohorts demonstrated that patients with melanoma of different genomic classes achieved different benefit from immunotherapy. Interestingly, tumors in cluster 4 demonstrated a cold immune microenvironment and lack of benefit from immunotherapy despite high microsatellite instability burden. CONCLUSIONS: Our study provides a proof for principle that deep learning modeling may have the potential to discover intrinsic statistical cross-modality correlations of multifactorial input data to dissect the molecular mechanisms underlying primary resistance to immunotherapy, which likely involves multiple factors from both the tumor and host at different molecular levels.

MIRMMR: binary classification of microsatellite instability using methylation and mutations.

SUMMARY: MIRMMR predicts microsatellite instability status in cancer samples using methylation and mutation information, in contrast to existing methods that rely on observed microsatellites. Additionally, MIRMMR highlights those genetic alterations contributing to microsatellite instability. AVAILABILITY AND IMPLEMENTATION: Source code is freely available at https://github.com/ding-lab/MIRMMR under the MIT license, implemented in R and supported on Unix/OS X operating systems. CONTACT: smfoltz@wustl.edu or lding@wustl.edu. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Divergent viral presentation among human tumors and adjacent normal tissues.

We applied a newly developed bioinformatics system called VirusScan to investigate the viral basis of 6,813 human tumors and 559 adjacent normal samples across 23 cancer types and identified 505 virus positive samples with distinctive, organ system- and cancer type-specific distributions. We found that herpes viruses (e.g., subtypes HHV4, HHV5, and HHV6) that are highly prevalent across cancers of the digestive tract showed significantly higher abundances in tumor versus adjacent normal samples, supporting their association with these cancers. We also found three HPV16-positive samples in brain lower grade glioma (LGG). Further, recurrent HBV integration at the KMT2B locus is present in three liver tumors, but absent in their matched adjacent normal samples, indicating that viral integration induced host driver genetic alterations are required on top of viral oncogene expression for initiation and progression of liver hepatocellular carcinoma. Notably, viral integrations were found in many genes, including novel recurrent HPV integrations at PTPN13 in cervical cancer. Finally, we observed a set of HHV4 and HBV variants strongly associated with ethnic groups, likely due to viral sequence evolution under environmental influences. These findings provide important new insights into viral roles of tumor initiation and progression and potential new therapeutic targets.

Systematic discovery of complex insertions and deletions in human cancers.

Complex insertions and deletions (indels) are formed by simultaneously deleting and inserting DNA fragments of different sizes at a common genomic location. Here we present a systematic analysis of somatic complex indels in the coding sequences of samples from over 8,000 cancer cases using Pindel-C. We discovered 285 complex indels in cancer-associated genes (such as PIK3R1, TP53, ARID1A, GATA3 and KMT2D) in approximately 3.5% of cases analyzed; nearly all instances of complex indels were overlooked (81.1%) or misannotated (17.6%) in previous reports of 2,199 samples. In-frame complex indels are enriched in PIK3R1 and EGFR, whereas frameshifts are prevalent in VHL, GATA3, TP53, ARID1A, PTEN and ATRX. Furthermore, complex indels display strong tissue specificity (such as VHL in kidney cancer samples and GATA3 in breast cancer samples). Finally, structural analyses support findings of previously missed, but potentially druggable, mutations in the EGFR, MET and KIT oncogenes. This study indicates the critical importance of improving complex indel discovery and interpretation in medical research.

Patterns and functional implications of rare germline variants across 12 cancer types.

Large-scale cancer sequencing data enable discovery of rare germline cancer susceptibility variants. Here we systematically analyse 4,034 cases from The Cancer Genome Atlas cancer cases representing 12 cancer types. We find that the frequency of rare germline truncations in 114 cancer-susceptibility-associated genes varies widely, from 4% (acute myeloid leukaemia (AML)) to 19% (ovarian cancer), with a notably high frequency of 11% in stomach cancer. Burden testing identifies 13 cancer genes with significant enrichment of rare truncations, some associated with specific cancers (for example, RAD51C, PALB2 and MSH6 in AML, stomach and endometrial cancers, respectively). Significant, tumour-specific loss of heterozygosity occurs in nine genes (ATM, BAP1, BRCA1/2, BRIP1, FANCM, PALB2 and RAD51C/D). Moreover, our homology-directed repair assay of 68 BRCA1 rare missense variants supports the utility of allelic enrichment analysis for characterizing variants of unknown significance. The scale of this analysis and the somatic-germline integration enable the detection of rare variants that may affect individual susceptibility to tumour development, a critical step toward precision medicine.

SPATA4 Counteracts Etoposide-Induced Apoptosis via Modulating Bcl-2 Family Proteins in HeLa Cells.

Spermatogenesis associated 4 (SPATA4) is a testis-specific gene first cloned by our laboratory, and plays an important role in maintaining the physiological function of germ cells. Accumulated evidence suggests that SPATA4 might be associated with apoptosis. Here we established HeLa cells that stably expressed SPATA4 to investigate the function of SPATA4 in apoptosis. SPATA4 protected HeLa cells from etoposide-induced apoptosis through the mitochondrial apoptotic pathway, in the way that SPATA4 suppressed decrease of the mitochondrial membrane potential, the release of cytochrome c, and subsequent activation of caspase-9 and -3. We further demonstrated that SPATA4 upregulated anti-apoptotic members of Bcl-2 family proteins, Bcl-2, and downregulated the pro-apoptotic member of Bcl-2 family proteins, Bax. Knockdown of SPATA4 in HeLa/SPATA4 cells could partially rescue expression levels of bcl-2 and bax. In conclusion, SPATA4 protects HeLa cells against etoposide-induced apoptosis through the mitochondrial apoptotic pathway. Our findings provide further evidence that SPATA4 plays a role in regulating apoptosis.

Age-related mutations associated with clonal hematopoietic expansion and malignancies.

Several genetic alterations characteristic of leukemia and lymphoma have been detected in the blood of individuals without apparent hematological malignancies. The Cancer Genome Atlas (TCGA) provides a unique resource for comprehensive discovery of mutations and genes in blood that may contribute to the clonal expansion of hematopoietic stem/progenitor cells. Here, we analyzed blood-derived sequence data from 2,728 individuals from TCGA and discovered 77 blood-specific mutations in cancer-associated genes, the majority being associated with advanced age. Remarkably, 83% of these mutations were from 19 leukemia and/or lymphoma-associated genes, and nine were recurrently mutated (DNMT3A, TET2, JAK2, ASXL1, TP53, GNAS, PPM1D, BCORL1 and SF3B1). We identified 14 additional mutations in a very small fraction of blood cells, possibly representing the earliest stages of clonal expansion in hematopoietic stem cells. Comparison of these findings to mutations in hematological malignancies identified several recurrently mutated genes that may be disease initiators. Our analyses show that the blood cells of more than 2% of individuals (5-6% of people older than 70 years) contain mutations that may represent premalignant events that cause clonal hematopoietic expansion.

Integrated analysis of germline and somatic variants in ovarian cancer.

We report the first large-scale exome-wide analysis of the combined germline-somatic landscape in ovarian cancer. Here we analyse germline and somatic alterations in 429 ovarian carcinoma cases and 557 controls. We identify 3,635 high confidence, rare truncation and 22,953 missense variants with predicted functional impact. We find germline truncation variants and large deletions across Fanconi pathway genes in 20% of cases. Enrichment of rare truncations is shown in BRCA1, BRCA2 and PALB2. In addition, we observe germline truncation variants in genes not previously associated with ovarian cancer susceptibility (NF1, MAP3K4, CDKN2B and MLL3). Evidence for loss of heterozygosity was found in 100 and 76% of cases with germline BRCA1 and BRCA2 truncations, respectively. Germline-somatic interaction analysis combined with extensive bioinformatics annotation identifies 222 candidate functional germline truncation and missense variants, including two pathogenic BRCA1 and 1 TP53 deleterious variants. Finally, integrated analyses of germline and somatic variants identify significantly altered pathways, including the Fanconi, MAPK and MLL pathways.

High frequency strand slippage mutations in CTCF in MSI-positive endometrial cancers.

Tumors with defective mismatch repair acquire large numbers of strand slippage mutations including frameshifts in coding sequence repeats. We identified a mutational hotspot, p.T204fs, in the insulator-binding protein (CTCF) in MSI-positive endometrial cancers. Although CTCF was described as a significantly mutated gene by the endometrial cancer TCGA, the A7 track variants leading to T204 frameshifts were not reported. Reanalysis of TCGA data using Pindel revealed frequent T204fs mutations, confirming CTCF is an MSI target gene and revealed the same frameshifts in tumors with intact mismatch repair. We show that T204fs transcripts are subject to nonsense-mediated decay and as such, T204fs mutations are unlikely to act as dominant negatives. The spectrum and pattern of mutations observed is consistent with CTCF acting as a haploinsufficient tumor suppressor.

MSIsensor: microsatellite instability detection using paired tumor-normal sequence data.

MOTIVATION: Microsatellite instability (MSI) is an important indicator of larger genome instability and has been linked to many genetic diseases, including Lynch syndrome. MSI status is also an independent prognostic factor for favorable survival in multiple cancer types, such as colorectal and endometrial. It also informs the choice of chemotherapeutic agents. However, the current PCR-electrophoresis-based detection procedure is laborious and time-consuming, often requiring visual inspection to categorize samples. We developed MSIsensor, a C++ program for automatically detecting somatic microsatellite changes. It computes length distributions of microsatellites per site in paired tumor and normal sequence data, subsequently using these to statistically compare observed distributions in both samples. Comprehensive testing indicates MSIsensor is an efficient and effective tool for deriving MSI status from standard tumor-normal paired sequence data. AVAILABILITY AND IMPLEMENTATION: https://github.com/ding-lab/msisensor

Mutational landscape and significance across 12 major cancer types.

The Cancer Genome Atlas (TCGA) has used the latest sequencing and analysis methods to identify somatic variants across thousands of tumours. Here we present data and analytical results for point mutations and small insertions/deletions from 3,281 tumours across 12 tumour types as part of the TCGA Pan-Cancer effort. We illustrate the distributions of mutation frequencies, types and contexts across tumour types, and establish their links to tissues of origin, environmental/carcinogen influences, and DNA repair defects. Using the integrated data sets, we identified 127 significantly mutated genes from well-known (for example, mitogen-activated protein kinase, phosphatidylinositol-3-OH kinase, Wnt/ß-catenin and receptor tyrosine kinase signalling pathways, and cell cycle control) and emerging (for example, histone, histone modification, splicing, metabolism and proteolysis) cellular processes in cancer. The average number of mutations in these significantly mutated genes varies across tumour types; most tumours have two to six, indicating that the number of driver mutations required during oncogenesis is relatively small. Mutations in transcriptional factors/regulators show tissue specificity, whereas histone modifiers are often mutated across several cancer types. Clinical association analysis identifies genes having a significant effect on survival, and investigations of mutations with respect to clonal/subclonal architecture delineate their temporal orders during tumorigenesis. Taken together, these results lay the groundwork for developing new diagnostics and individualizing cancer treatment.

Estimating absolute methylation levels at single-CpG resolution from methylation enrichment and restriction enzyme sequencing methods.

Recent advancements in sequencing-based DNA methylation profiling methods provide an unprecedented opportunity to map complete DNA methylomes. These include whole-genome bisulfite sequencing (WGBS, MethylC-seq, or BS-seq), reduced-representation bisulfite sequencing (RRBS), and enrichment-based methods such as MeDIP-seq, MBD-seq, and MRE-seq. These methods yield largely comparable results but differ significantly in extent of genomic CpG coverage, resolution, quantitative accuracy, and cost, at least while using current algorithms to interrogate the data. None of these existing methods provides single-CpG resolution, comprehensive genome-wide coverage, and cost feasibility for a typical laboratory. We introduce methylCRF, a novel conditional random fields-based algorithm that integrates methylated DNA immunoprecipitation (MeDIP-seq) and methylation-sensitive restriction enzyme (MRE-seq) sequencing data to predict DNA methylation levels at single-CpG resolution. Our method is a combined computational and experimental strategy to produce DNA methylomes of all 28 million CpGs in the human genome for a fraction (<10%) of the cost of whole-genome bisulfite sequencing methods. methylCRF was benchmarked for accuracy against Infinium arrays, RRBS, WGBS sequencing, and locus-specific bisulfite sequencing performed on the same human embryonic stem cell line. methylCRF transformation of MeDIP-seq/MRE-seq was equivalent to a biological replicate of WGBS in quantification, coverage, and resolution. We used conventional bisulfite conversion, PCR, cloning, and sequencing to validate loci where our predictions do not agree with whole-genome bisulfite data, and in 11 out of 12 cases, methylCRF predictions of methylation level agree better with validated results than does whole-genome bisulfite sequencing. Therefore, methylCRF transformation of MeDIP-seq/MRE-seq data provides an accurate, inexpensive, and widely accessible strategy to create full DNA methylomes.

DNA hypomethylation within specific transposable element families associates with tissue-specific enhancer landscape.

Transposable element (TE)-derived sequences comprise half of the human genome and DNA methylome and are presumed to be densely methylated and inactive. Examination of genome-wide DNA methylation status within 928 TE subfamilies in human embryonic and adult tissues identified unexpected tissue-specific and subfamily-specific hypomethylation signatures. Genes proximal to tissue-specific hypomethylated TE sequences were enriched for functions important for the relevant tissue type, and their expression correlated strongly with hypomethylation within the TEs. When hypomethylated, these TE sequences gained tissue-specific enhancer marks, including monomethylation of histone H3 at lysine 4 (H3K4me1) and occupancy by p300, and a majority exhibited enhancer activity in reporter gene assays. Many such TEs also harbored binding sites for transcription factors that are important for tissue-specific functions and showed evidence of evolutionary selection. These data suggest that sequences derived from TEs may be responsible for wiring tissue type-specific regulatory networks and may have acquired tissue-specific epigenetic regulation.

Comparative epigenomic annotation of regulatory DNA.

Despite the explosive growth of genomic data, functional annotation of regulatory sequences remains difficult. Here, we introduce "comparative epigenomics"-interspecies comparison of DNA and histone modifications-as an approach for annotation of the regulatory genome. We measured in human, mouse, and pig pluripotent stem cells the genomic distributions of cytosine methylation, H2A.Z, H3K4me1/2/3, H3K9me3, H3K27me3, H3K27ac, H3K36me3, transcribed RNAs, and P300, TAF1, OCT4, and NANOG binding. We observed that epigenomic conservation was strong in both rapidly evolving and slowly evolving DNA sequences, but not in neutrally evolving sequences. In contrast, evolutionary changes of the epigenome and the transcriptome exhibited a linear correlation. We suggest that the conserved colocalization of different epigenomic marks can be used to discover regulatory sequences. Indeed, seven pairs of epigenomic marks identified exhibited regulatory functions during differentiation of embryonic stem cells into mesendoderm cells. Thus, comparative epigenomics reveals regulatory features of the genome that cannot be discerned from sequence comparisons alone.