Research ArticleNeuroscience

Learning-dependent chromatin remodeling highlights noncoding regulatory regions linked to autism

See allHide authors and affiliations

Science Signaling  16 Jan 2018:
Vol. 11, Issue 513, eaan6500
DOI: 10.1126/scisignal.aan6500
  • Fig. 1 Learning increases chromatin accessibility in the mouse hippocampus.

    (A) Outline of experimental design to measure changes in chromatin accessibility after contextual fear conditioning (FC) versus time of day matched controls (HC) using sonication of cross-linked chromatin followed by sequencing (Sono-seq) (B) The effect of learning in chromatin accessibility at transcriptional start sites (TSS) genome-wide. The plot shows normalized read counts relative to genomic position around TSS for all promoters in the mouse genome (mm9). Average per condition is shown (n = 4 samples per group). Average sequencing depth per sample: 100 million 100-bp single-end reads. The number of learning-regulated regions detected at false discovery rate (FDR) < 0.05 requiring four replicates to reproduce the peak using DEScan is shown (see figs. S1 and S2 and table S1).

  • Fig. 2 Learning recapitulates development at the epigenetic level and is linked to changes in gene expression.

    (A) Overlap of learning-regulated regions relative to regions of accessible chromatin during development (ATAC-seq), CpG islands, alternative promoters, and histone marks associated with active (H3K4me2/3, H3K9ac, and H3K27ac) or inactive, repressed (H3K27me3) transcription. Permutation testing was performed to evaluate statistical significance. The number of overlaps observed for each set with the learning-regulated regions was compared to overlaps with 10,000 random samplings of the same number of regions from a background set of all Sono-seq peaks present in two or more samples. The P value corresponds to the proportion of permuted samples with more overlaps than the observed value; ***P < 0.001, where no permutation showed greater overlap than the observed value. Violin plots display the full distribution of permuted set overlaps with each genomic region data set, whereas circles indicate the observed value of overlaps with learning-regulated promoters. All ChIP-seq and ATAC-seq data obtained from ENCODE are derived from mouse forebrain at E12.5. CpG island and alternative promoter (altEvents) annotations obtained from the UCSC genome browser. (B) Overlap of learning-regulated regions with active histone marks in the forebrain, midbrain, and hindbrain at E12.5 relative to the overlap in the same tissues at postnatal day 0 (P0). (C) Enrichment of learning-regulated promoters relative to gene expression measured using RNA-seq after FC and retrieval of memory 24 hours after (RT), assessed as the number of overlaps between the learning-regulated regions and each set of differentially expressed or spliced set of genes (FDR < 0.05) compared to overlaps with 10,000 random samplings of genes expressed in the hippocampus. P value determined as described in (A). Violin plots display the full distribution of permuted set overlaps with each genomic region data set, whereas circles indicate the observed value of overlaps.

  • Fig. 3 Learning increases accessibility to Shank3 promoter 6, highlighting a regulatory region linked to ASD.

    (A) Chromatin accessibility at the Shank3 locus. Sono-seq signals (mapped reads) from representative samples are shown as mapped to the UCSC genome browser (mm9) for both fear-conditioned (FC) and control (HC) animals. Dark blue track shows UCSC gene model for Shank3, with filled rectangles indicating the 22 exons. The light blue region highlights a region of significant enrichment compared to controls as detected by DEScan (n = 4 per group, FDR < 0.05). Green boxes represent CpG islands. The promoters of Shank3 are shown for reference, adapted from (46). (B) Differential expression (by qPCR) of major Shank3 isoforms after fear conditioning in an independent cohort of animals, normalized to Hprt expression and quantified using the ΔΔCt method. Data are mean ± SE from n = 7 animals per condition. Only P < 0.1 is shown; by two-sided t tests. Primer pairs used to quantify Shank3 isoform expression were designed to span Shank3a exons 2 to 10, Shank3b exons 3 to 6, Shank3c exons 12 to 14, Shank3d exons 13 to 15, Shank3e exons 17 to 19, and Shank3f exon 22. (C) Association of genetic variation in promoter 6 of SHANK3 (SNP rs6010065) with ASD (blue, n = 422 patients) versus TDCs (red, n = 182 individuals) as assessed by the χ2 test per genotype relative to the sum of all other genotypes. SNP rs6010065 does not overlap SHANK3 exon 21. Further details on genotyping of patients are in fig. S3.

  • Table 1 Learning-regulated regions are disproportionately associated with known ASD and cancer candidate genes.

    ASD candidate genes were obtained from the Simons Foundation Autism Research Initiative (SFARI) gene database. Cancer candidate genes were obtained from http://cancer.sanger.ac.uk/census. Statistical significance was calculated using χ2 tests comparing observed versus expected frequencies. ASD genes associated with learning-regulated promoters can be found in table S2.

    Gene setsDisease genes/genes
    associated with learning-
    regulated regions
    (observed)
    Disease genes/all genes
    (expected)
    Fold enrichmentP value
    ASD genes (all SFARI)127 of 2279880 of 53,1103.4<0.001
    ASD genes (SFARI score ≤ 4)81 of 2279519 of 53,1103.6<0.001
    ASD genes (syndromic)20 of 227998 of 53,1104.8<0.001
    Cancer genes105 of 2279573 of 53,1104.3<0.001

Supplementary Materials

  • www.sciencesignaling.org/cgi/content/full/11/513/eaan6500/DC1

    Fig. S1. DEScan is a tool to perform differential analysis of data obtained from epigenetic HTS experiments with multiple biological replicates.

    Fig. S2. Normalization affects detection of differences in chromatin accessibility after learning.

    Fig. S3. SNP rs6010065 is differentially associated with ASD or intelligence quotient depending on genetic background.

    Table S1. Detailed annotation of learning-regulated promoters.

    Table S2. Details of which learning-regulated regions are associated with known ASD risk genes.

    File S1. DEScan user manual.

  • Supplementary Materials for:

    Learning-dependent chromatin remodeling highlights noncoding regulatory regions linked to autism

    John N. Koberstein, Shane G. Poplawski, Mathieu E. Wimmer, Giulia Porcari, Charlly Kao, Bruce Gomes, Davide Risso, Hakon Hakonarson, Nancy R. Zhang, Robert T. Schultz, Ted Abel, Lucia Peixoto*

    *Corresponding author. Email: lucia.peixoto{at}wsu.edu

    This PDF file includes:

    • Fig. S1. DEScan is a tool to perform differential analysis of data obtained from epigenetic HTS experiments with multiple biological replicates.
    • Fig. S2. Normalization affects detection of differences in chromatin accessibility after learning.
    • Fig. S3. SNP rs6010065 is differentially associated with ASD or intelligence quotient depending on genetic background.
    • Legends for tables S1 and S2
    • Legend for file S1

    [Download PDF]

    Technical Details

    Format: Adobe Acrobat PDF

    Size: 1.24 MB

    Other Supplementary Material for this manuscript includes the following:

    • Table S1 (Microsoft Excel format). Detailed annotation of learning-regulated promoters.
    • Table S2 (Microsoft Excel format). Details of which learning-regulated regions are associated with known ASD risk genes.
    • File S1 (.pdf format). DEScan user manual.

    [Download Tables S1 and S2]

    [Download File S1]


    Citation: J. N. Koberstein, S. G. Poplawski, M. E. Wimmer, G. Porcari, C. Kao, B. Gomes, D. Risso, H. Hakonarson, N. R. Zhang, R. T. Schultz, T. Abel, L. Peixoto, Learning-dependent chromatin remodeling highlights noncoding regulatory regions linked to autism. Sci. Signal. 11, eaan6500 (2018).

    © 2018 American Association for the Advancement of Science

Stay Connected to Science Signaling

Navigate This Article