Research ArticleCELLULAR NEUROSCIENCE

Using 3D epigenomic maps of primary olfactory neuronal cells from living individuals to understand gene regulation

See allHide authors and affiliations

Science Advances  13 Dec 2018:
Vol. 4, no. 12, eaav8550
DOI: 10.1126/sciadv.aav8550
  • Fig. 1 Project overview.

    (A) To develop a comprehensive regulatory map of CNON, we began with nasal biopsies from 63 individuals. Neuroepithelial cells from the biopsied tissues were enriched using specific culture conditions, and the resultant CNON were grown on plastic dishes coated with Matrigel for all assays. (B) Using primary CNON, we performed in situ Hi-C (n = 2) and CTCF ChIP-seq (n = 33) to develop a 3D chromatin map, epigenomic profiling using ChIP-seq for modified histones (n = 56, 47, 4, 22, and 11 for H3K4me3, H3K27ac, H3K4me1, H3K27me3, and H3K9me3, respectively), and NOMe-seq (n = 5) to provide information concerning nucleosome positioning. (C) Using comprehensive epigenomic and chromatin maps, we identified TADs, chromatin interaction loops, and regulatory elements and defined NDRs of open chromatin [including transcription factor (TF) binding sites (TFBSs)]. Last, we identified regulatory elements having variation linked to genotype, gender, smoking, and schizophrenia (SCZ), as well as observed and predicted enhancer-target gene interactions for these olfactory neuroepithelial cells.

  • Fig. 2 Creation of a 3D epigenomic map of active and heterochromatic TADs in CNON.

    (A) A browser snapshot of the H3K4me3, H3K4me1, H3K27ac, CTCF, H3K27me3, and H3K9me3 patterns for a 700-kb region of chromosome (Chr) 19p13.11. All datasets shown are from CNON taken from an individual designated SEP044. A track indicating chromatin state, the RefSeq gene track, a track indicating the identified TADs, and a Hi-C chromatin interaction browser snapshot for the region are also shown. (B) A browser snapshot of the CTCF, H3K27me3, and H3K9me3 patterns for a 2.5-Mb region of chromosome 2 that harbors H3K9me3-marked TAD. The Hi-C interaction data, the RefSeq gene track, a track indicating all TADs within the region, and a track indicating an H3K9me3-marked TAD are shown. (C) The average expression level of genes in H3K9me3-marked TADs (blue) and all other TADs (dark gray). (D) The ChIP fragment depth for CTCF, H3K4me3, H3K9me3, and H3K27me3 datasets, centered on TAD boundaries. Profiles of H3K9me3-marked TADs are shown in blue, and profiles of the other TADs are shown in dark gray.

  • Fig. 3 Identification of target genes of enhancers active in CNON cells.

    (A) The number of regulatory elements identified. (B) The 25 most frequent chromatin interaction categories. Left: For each category, each of the two ends of the interactions is identified by one or more of the five types of regulatory elements described in (A) (none indicates that one or both ends of the interaction did not correspond to a regulatory region). Right: The number of interactions in each category, using a q value cutoff of 1 × 10−12 (see also fig. S2). (C) The chromatin interaction frequencies of promoter-enhancer interactions that have 2, 1, or 0 CTCF anchors. (D) The number of genes in each category of promoter-enhancer interactions and the number of promoter-enhancer predicted pairs (not identified by direct chromatin interaction data) within CTCF-CTCF loops. (E) The expression level of genes in each category from (D). The expression level of all other genes in CNON is shown.

  • Fig. 4 Assessing individual variation in epigenomic profiles of active regulatory elements.

    ChIP-seq was performed from 63 individuals; after quality assessment (see fig. S3), peak sets were retained for 56 H3K4me3 datasets, 33 CTCF datasets, and 47 H3K27ac datasets. (A) The number of peaks shared in the different sets of H3K4me3, CTCF, and H3K27ac ChIP-seq samples. (B) The cumulative number of unique peaks in the merged datasets with increasing numbers of different ChIP-seq samples. Only peaks found in at least 20% of individuals are plotted. (C) A comparison of the size of the SEP044 peaks in the set of H3K4me3, CTCF, and H3K27ac peaks present in every individual ChIP-seq peak sets (100%), in at least 20% of all samples (≥20%), in less than 20% of all samples (<20%), and only in the SEP044 sample. The tag density plot is centered on the middle of the peak (x axis), and the read depth is indicated by y axis.

  • Fig. 5 Motifs enriched in CNON regulatory regions identify neuronal-related TFs.

    (A) An example of a single H3K27ac ChIP-seq peak and the position of the NDRs within the peak. One NDR identifies an NFE2 motif. (B) The average endogenous DNA methylation level [HCG (where H = A, C, or T): black] and the accessibility [GCH (where H = A, C, or T): green] of NDRs at promoters (16,711 H3K4me3 peaks identified in ≥20% of all samples) centered on the TSS, insulators (62,433 CTCF peaks identified in ≥20% of all samples) centered on the CTCF motif, and enhancers (61,419 H3K27ac peaks identified in ≥20% of all samples) centered to the middle of the NDR. (C) Left: The average endogenous DNA methylation level (HCG: black) and the accessibility (GCH: green) for the NDRs without features. Right: A cluster analysis of the NDRs without features based on levels of accessibility. Endogenous DNA methylation for the same NDRs is shown in the middle. (D) A cluster analysis of the top 50 TF binding motifs identified using the 74,323 NDR regions contained within distal CNON H3K27ac peaks.

  • Fig. 6 Characterization of factors that affect individual variation in enhancers.

    (A) Top: A browser snapshot of an H3K27ac peak that shows peak strength variation as a correlation with genotype of rs1757069. Bottom: A correlation between the peak height and the making or breaking of a SOX3 motif by rs1757069. (B) Top: A browser snapshot of a cluster of H3K27ac peaks that show peak strength variation as a correlation with gender. Bottom: The number of H3K27ac peaks on each chromosome that are differentially enriched in male or female CNON samples. (C and D) A browser snapshot of an H3K27ac peak that shows peak strength variation as a correlation with smoking (C) or SCZ (D). (E) A heatmap representing unsupervised clustering of differentially enriched H3K27ac sites between SCZ and CON. Cluster 1 is higher in SCZ, and cluster 2 is higher in CON. Individual sample metadata and ChIP-seq data information are on the top of the heatmap, whereas the H3K27ac sites linked to genotype, gender, and smoking that are among the H3K27ac differentially associated with disease state are indicated on the left.

  • Fig. 7 A workflow for identifying SCZ risk-related variants using CNON epigenomic and Hi-C data.

    Upon overlapping CNON active regulatory elements found from 63 individuals with SCZ GWAS risk-associated high LD variants, 3403 high LD variants were identified. Among those, 2891 high LD variants are located either within the same TAD or CTCF-CTCF interaction loop as the index variant; these were used for further analysis. The high LD regulatory variants located in an NDR (n = 215) were used to identify TF motifs changed by the variants, whereas the other 2676 high LD regulatory variants were examined for correlation with ChIP-seq signal strength; examples of altered motifs and peak strength between risk and nonrisk alleles are shown. OR, odds ratio.

  • Fig. 8 Finding target genes of enhancers using 3D epigenomic datasets from many individuals.

    (A) Schematic overview of the process used to identify target genes of CNON enhancers. In situ Hi-C datasets from two individuals (SEP044 and SEP045) were used to reveal chromatin interactions and TADs. Using ChIP-seq peaks from the same individuals, 5623 target genes were identified by direct promoter-enhancer chromatin interactions, and an additional 12,120 target genes of enhancers were predicted. In addition, using H3K4me3 and H3K27ac ChIP-seq data found from many individuals, additional enhancer target genes were predicted on the basis of colocalization within the same TADs. Last, the subsets of target genes of enhancers linked to gender, smoking, and SCZ, as well as 5115 target genes that show a correlation between gene expression and H3K27ac peak strength, was identified. (B) Gene ontology analyses for the target genes of enhancers linked to gender or smoking, differentially present in SCZ samples, or identified by SCZ GWAS variants. UDP, uridine diphosphate.

Supplementary Materials

  • Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/4/12/eaav8550/DC1

    Supplementary Materials and Methods

    Fig. S1. CNON Hi-C datasets.

    Fig. S2. Chromatin interactions with different q value cutoffs.

    Fig. S3. H3K27ac ChIP-seq clustering results.

    Fig. S4. NOMe-seq depth and number of NDRs.

    Table S1. Information about the CNON ChIP-seq, NOMe-seq, and Hi-C datasets.

    Table S2. CNON TADs identified by Hi-C.

    Table S3. CNON ChIP-seq peaks classified as promoters, enhancers, insulators, repressed, or heterochromatin regions.

    Table S4. CNON chromatin interactions identified by Hi-C.

    Table S5. Epigenomic classification of CNON Hi-C chromatin interactions.

    Table S6. Predicted target genes of enhancers using CNON Hi-C.

    Table S7. Coordinates of promoter H3K4me3, non-promoter H3K27ac, and CTCF peaks found in many individuals of CNON datasets.

    Table S8. Identification and classification of NDRs identified by CNON NOMe-seq.

    Table S9. Enhancers linked to genotype, gender, smoking status, and schizophrenia diagnosis.

    Table S10. Classification and characterization of SCZ risk-associated variants.

    Table S11. Predicted target genes of enhancers linked to individual variations.

    References (5567)

  • Supplementary Materials

    The PDF file includes:

    • Supplementary Materials and Methods
    • Fig. S1. CNON Hi-C datasets.
    • Fig. S2. Chromatin interactions with different q value cutoffs.
    • Fig. S3. H3K27ac ChIP-seq clustering results.
    • Fig. S4. NOMe-seq depth and number of NDRs.
    • Legends for tables S1 to S11
    • References (5567)

    Download PDF

    Other Supplementary Material for this manuscript includes the following:

    • Table S1 (Microsoft Excel format). Information about the CNON ChIP-seq, NOMe-seq, and Hi-C datasets.
    • Table S2 (Microsoft Excel format). CNON TADs identified by Hi-C.
    • Table S3 (Microsoft Excel format). CNON ChIP-seq peaks classified as promoters, enhancers, insulators, repressed, or heterochromatin regions.
    • Table S4 (Microsoft Excel format). CNON chromatin interactions identified by Hi-C.
    • Table S5 (Microsoft Excel format). Epigenomic classification of CNON Hi-C chromatin interactions.
    • Table S6 (Microsoft Excel format). Predicted target genes of enhancers using CNON Hi-C.
    • Table S7 (Microsoft Excel format). Coordinates of promoter H3K4me3, non-promoter H3K27ac, and CTCF peaks found in many individuals of CNON datasets.
    • Table S8 (Microsoft Excel format). Identification and classification of NDRs identified by CNON NOMe-seq.
    • Table S9 (Microsoft Excel format). Enhancers linked to genotype, gender, smoking status, and schizophrenia diagnosis.
    • Table S10 (Microsoft Excel format). Classification and characterization of SCZ risk-associated variants.
    • Table S11 (Microsoft Excel format). Predicted target genes of enhancers linked to individual variations.

    Files in this Data Supplement:

Stay Connected to Science Advances

Navigate This Article