Learning representations of chromatin contacts using a recurrent neural network identifies genomic drivers of conformation.

Despite the availability of chromatin conformation capture experiments, discerning the relationship between the 1D genome and 3D conformation remains a challenge, which limits our understanding of their affect on gene expression and disease. We propose Hi-C-LSTM, a method that produces low-dimensional latent representations that summarize intra-chromosomal Hi-C contacts via a recurrent long short-term memory neural network model. We find that these representations contain all the information needed to recreate the observed Hi-C matrix with high accuracy, outperforming existing methods. These representations enable the identification of a variety of conformation-defining genomic elements, including nuclear compartments and conformation-related transcription factors. They furthermore enable in-silico perturbation experiments that measure the influence of cis-regulatory elements on conformation.

Characterization of Adaptive-like γδ T Cells in Ugandan Infants during Primary Cytomegalovirus Infection.

Gamma-delta (γδ) T cells are unconventional T cells that help control cytomegalovirus (CMV) infection in adults. γδ T cells develop early in gestation, and a fetal public γδ T cell receptor (TCR) clonotype is detected in congenital CMV infections. However, age-dependent γδ T cell responses to primary CMV infection are not well-understood. Flow cytometry and TCR sequencing was used to comprehensively characterize γδ T cell responses to CMV infection in a cohort of 32 infants followed prospectively from birth. Peripheral blood γδ T cell frequencies increased during infancy, and were higher among CMV-infected infants relative to uninfected. Clustering analyses revealed associations between CMV infection and activation marker expression on adaptive-like Vδ1 and Vδ3, but not innate-like Vγ9Vδ2 γδ T cell subsets. Frequencies of NKG2C+CD57+ γδ T cells were temporally associated with the quantity of CMV shed in saliva by infants with primary infection. The public γδ TCR clonotype was only detected in CMV-infected infants <120 days old and at lower frequencies than previously described in fetal infections. Our findings support the notion that CMV infection drives age-dependent expansions of specific γδ T cell populations, and provide insight for novel strategies to prevent CMV transmission and disease.

Deep learning of immune cell differentiation.

Although we know many sequence-specific transcription factors (TFs), how the DNA sequence of cis-regulatory elements is decoded and orchestrated on the genome scale to determine immune cell differentiation is beyond our grasp. Leveraging a granular atlas of chromatin accessibility across 81 immune cell types, we asked if a convolutional neural network (CNN) could learn to infer cell type-specific chromatin accessibility solely from regulatory DNA sequences. With a tailored architecture and an ensemble approach to CNN parameter interpretation, we show that our trained network ("AI-TAC") does so by rediscovering ab initio the binding motifs for known regulators and some unknown ones. Motifs whose importance is learned virtually as functionally important overlap strikingly well with positions determined by chromatin immunoprecipitation for several TFs. AI-TAC establishes a hierarchy of TFs and their interactions that drives lineage specification and also identifies stage-specific interactions, like Pax5/Ebf1 vs. Pax5/Prdm1, or the role of different NF-κB dimers in different cell types. AI-TAC assigns Spi1/Cebp and Pax5/Ebf1 as the drivers necessary for myeloid and B lineage fates, respectively, but no factors seemed as dominantly required for T cell differentiation, which may represent a fall-back pathway. Mouse-trained AI-TAC can parse human DNA, revealing a strikingly similar ranking of influential TFs and providing additional support that AI-TAC is a generalizable regulatory sequence decoder. Thus, deep learning can reveal the regulatory syntax predictive of the full differentiative complexity of the immune system.

Single-Cell Transcriptome Profiling of Mouse and hESC-Derived Pancreatic Progenitors.

Human embryonic stem cells (hESCs) are a potential unlimited source of insulin-producing ß cells for diabetes treatment. A greater understanding of how ß cells form during embryonic development will improve current hESC differentiation protocols. All pancreatic endocrine cells, including ß cells, are derived from Neurog3-expressing endocrine progenitors. This study characterizes the single-cell transcriptomes of 6,905 mouse embryonic day (E) 15.5 and 6,626 E18.5 pancreatic cells isolated from Neurog3-Cre; Rosa26mT/mG embryos, allowing for enrichment of endocrine progenitors (yellow; tdTomato + EGFP) and endocrine cells (green; EGFP). Using a NEUROG3-2A-eGFP CyT49 hESC reporter line (N5-5), 4,462 hESC-derived GFP+ cells were sequenced. Differential expression analysis revealed enrichment of markers that are consistent with progenitor, endocrine, or previously undescribed cell-state populations. This study characterizes the single-cell transcriptomes of mouse and hESC-derived endocrine progenitors and serves as a resource (https://lynnlab.shinyapps.io/embryonic_pancreas) for improving the formation of functional ß-like cells from hESCs.

Concurrent CIC mutations, IDH mutations, and 1p/19q loss distinguish oligodendrogliomas from other cancers.

Oligodendroglioma is characterized by unique clinical, pathological, and genetic features. Recurrent losses of chromosomes 1p and 19q are strongly associated with this brain cancer but knowledge of the identity and function of the genes affected by these alterations is limited. We performed exome sequencing on a discovery set of 16 oligodendrogliomas with 1p/19q co-deletion to identify new molecular features at base-pair resolution. As anticipated, there was a high rate of IDH mutations: all cases had mutations in either IDH1 (14/16) or IDH2 (2/16). In addition, we discovered somatic mutations and insertions/deletions in the CIC gene on chromosome 19q13.2 in 13/16 tumours. These discovery set mutations were validated by deep sequencing of 13 additional tumours, which revealed seven others with CIC mutations, thus bringing the overall mutation rate in oligodendrogliomas in this study to 20/29 (69%). In contrast, deep sequencing of astrocytomas and oligoastrocytomas without 1p/19q loss revealed that CIC alterations were otherwise rare (1/60; 2%). Of the 21 non-synonymous somatic mutations in 20 CIC-mutant oligodendrogliomas, nine were in exon 5 within an annotated DNA-interacting domain and three were in exon 20 within an annotated protein-interacting domain. The remaining nine were found in other exons and frequently included truncations. CIC mutations were highly associated with oligodendroglioma histology, 1p/19q co-deletion, and IDH1/2 mutation (p < 0.001). Although we observed no differences in the clinical outcomes of CIC mutant versus wild-type tumours, in a background of 1p/19q co-deletion, hemizygous CIC mutations are likely important. We hypothesize that the mutant CIC on the single retained 19q allele is linked to the pathogenesis of oligodendrogliomas with IDH mutation. Our detailed study of genetic aberrations in oligodendroglioma suggests a functional interaction between CIC mutation, IDH1/2 mutation, and 1p/19q co-deletion.

System-level analysis of neuroblastoma tumor-initiating cells implicates AURKB as a novel drug target for neuroblastoma.

PURPOSE: Neuroblastoma (NB) is an aggressive tumor of the developing peripheral nervous system that remains difficult to cure in the advanced stages. The poor prognosis for high-risk NB patients is associated with common disease recurrences that fail to respond to available therapies. NB tumor-initiating cells (TICs), isolated from metastases and primary tumors, may escape treatment and contribute to tumor relapse. New therapies that target the TICs may therefore prevent or treat tumor recurrences. EXPERIMENTAL DESIGN: We undertook a system-level characterization of NB TICs to identify potential drug targets against recurrent NB. We used next-generation RNA sequencing and/or human exon arrays to profile the transcriptomes of 11 NB TIC lines from six NB patients, revealing genes that are highly expressed in the TICs compared with normal neural crest-like cells and unrelated cancer tissues. We used gel-free two-dimensional liquid chromatography coupled to shotgun tandem mass spectrometry to confirm the presence of proteins corresponding to the most abundant TIC-enriched transcripts, thereby providing validation to the gene expression result. RESULTS: Our study revealed that genes in the BRCA1 signaling pathway are frequently misexpressed in NB TICs and implicated Aurora B kinase as a potential drug target for NB therapy. Treatment with a selective AURKB inhibitor was cytotoxic to NB TICs but not to the normal neural crest-like cells. CONCLUSION: This work provides the first high-resolution system-level analysis of the transcriptomes of 11 primary human NB TICs and identifies a set of candidate NB TIC-enriched transcripts for further development as therapeutic targets.