An epigenetic signature for monoallelic olfactory receptor expression

Abstract

Constitutive heterochromatin is traditionally viewed as the static form of heterochromatin that silences pericentromeric and telomeric repeats in a cell cycle- and differentiation-independent manner. Here, we show that, in the mouse olfactory epithelium, olfactory receptor (OR) genes are marked in a highly dynamic fashion with the molecular hallmarks of constitutive heterochromatin, H3K9me3 and H4K20me3. The cell type and developmentally dependent deposition of these marks along the OR clusters are, most likely, reversed during the process of OR choice to allow for monogenic and monoallelic OR expression. In contrast to the current view of OR choice, our data suggest that OR silencing takes place before OR expression, indicating that it is not the product of an OR-elicited feedback signal. Our findings suggest that chromatin-mediated silencing lays a molecular foundation upon which singular and stochastic selection for gene expression can be applied.

Figure 1. Genome-wide mapping of H3K9me3 and H4K20me3 reveal a tissue-dependent heterochromatinization of the ORs in the MOE

ChIP-on-chip experiments with antibodies against H3K9me3 and H4K20me3 using native chromatin preparations from the MOE and liver. The log2 ratio of IP/input was calculated and used for the construction of the heatmaps presented here. (A) Positional heatmaps of chromosomes 2, 7 and 9 are shown. Each row represents one gene in 1kb windows from -5kb to +5kb of the TSS. Four states are shown as adjacent columns: liver-H3K9me3, OE-H3K9me3, liver-H4K20me3 and OE-H4K20me3. Arrows indicate OR, V1R and V2R clusters found on these chromosomes. (B) Ranked heatmap illustrating 1,000 randomly selected genes (approx. 1 in 15 genes). Each row represents one gene in 200 bp windows from -1kb to +1kb of the translation start site. The ORs (blue lines) make up the vast majority of the genes that are positive for both modifications and are placed at the top of the heatmap. The red lines represent VRs and FPRs. (C) Ranked heatmap, constructed as the previous one but showing the top 1,000 genes positive for H3K9me3 and H4K20me3 in the MOE. Most of these genes are ORs that, also, rank the highest, as depicted by the blue lines next to the heatmap. (See also Figure S1).

Figure 2. OR clusters in the MOE are coated by tissue-specific heterochromatic blocks of H3K9me3 and H4K20me3

Ma2C analysis of our ChIP-on-chip data viewed on the UCSC genome browser. (A) Part of the biggest OR cluster located on chromosome 2, which contains ~240 genes and spans a 5-MB region. The thin blue (H3K9me3) or red (H4K20me3) bars represent significant peaks (FDR ≤5%) identified in the MOE by MA2C using standard parameters (window=0.5 kb, min number of probes= 5, max gap=0.25 kb); the thick blue or red bars represent the blocks identified with modified parameters (window=10 kb, min number of probes= 20, max gap= 1kb). In the liver, there are only a few, sporadic H3K9me3 peaks and blocks (purple). (B) Results from H3K9me3 ChIP-qPCR analysis using native chromatin preparations from MOE and liver. The Ptprj gene (marked by red rectangle in 2A) stands at the border of the OR cluster, which coincides with the border of the heterochromatic block. Its most proximal -to the OR cluster- intron is enriched for H3K9me3 and H4K20me3 and its most distal intron, located 43kb downstream, is free of these modifications. Zfp560 serves as positive control. (C) Same as B but for H4K20me3. (D) Part of an OR cluster on chromosome 11 is interrupted by a small group of non-OR. Genes that are marked by green rectangle are transcriptionally active in the MOE and genes marked by red rectangles do not have detectable transcripts. (E) Zoomed-in picture of the cluster, which shows that genes Btnl9 and Flt4, which are transcriptionally inactive, are partly methylated. Two sets of primers for each of these genes, one at the beginning (most proximal to the neighboring OR gene) and one at the end of the gene (most distal from the neighboring OR) were used in ChIP-qPCR. (F) Results from H3K9me3 ChIP-qPCR analysis using native chromatin preparations from the MOE. OR genes tested, as well as Zfp354c, and part of the Btnl9 and Flt4 genes were enriched. In contrast, the active genes Mgat1 and Mapk9, Zfp879 and the distal part of Btnl9 and Flt4 were devoid of modifications. (G) Same as F but for H4K20me3. (See also Figure S2). All above experiments were performed in two biological replicates with similar results. Values are the mean of triplicate qPCR. Error bars indicate the S.E.M.

Figure 3. The ORs acquire a highly compacted chromatin structure in the MOE

(A) DNase I accessibility assay with nuclei from both MOE and liver. Nuclei were treated with DNase I, DNA was isolated at various time points (2 to 40 min) and equal amounts were used for qPCR. The amount of DNA measured at each interval was expressed as a fraction of the DNA present at 2 min of enzyme treatment and was plotted over time. We assayed several ORs as well as genes that are active or inactive in the MOE or liver, and their mean is shown here (see Fig. S3A, B for detailed analysis of all genes). Representative data from one experiment are shown here. (B) MNase digested chromatin was submitted to ultracentrifugation through a sucrose gradient. The largest or most compacted chromatin fragments are collected in the fractions with the highest sucrose concentration. For a particular fragment size the most compact chromatin is collected in more concentrated fractions. (C) Fractions from MOE and liver analyzed by agarose gel electrophoresis and Southern blot with a degenerate OR probe. Arrows mark the low molecular weight OR sequences that appear in the bottom fractions of the chromatin from the MOE. Input lanes represent DNA extracted from chromatin that was not loaded in the gradient. (D) Selected fractions from the same experiment analyzed with the use of a ribosomal probe. (See also Figure S3).

Figure 4. ChIP-qPCR assays for H3K9me3 and H4K20me3 in sorted cell populations from the MOE

(A) Section of the MOE from an adult OMP-IRES-GFP mouse. Mature OSNs are GFP⁺. (B) GFP⁺ cells (mature OSNs) were isolated with FACS from OMP-IRES-GFP mice and were used for ChIP-qPCR experiments. Golf, Tbp and Omp are active genes in these cells and used as negative controls. Zfp560 and major satellite repeats are used as positive controls. Olfr690 is a type I OR. (C) Immunostaining of MOE section with SUS4 antibody that specifically labels the sustentacular cells. (D) ChIP-qPCR with isolated sustentacular cells. Cbr is an active gene and is used as a negative control. (E) Immunostaining of MOE section with an antibody against ICAM (PE-ICAM) that specifically labels the HBCs. (F) ChIP-qPCR experiments with isolated HBCs. (G) ChIP-qPCR with immature neurons and progenitors from the MOE isolated by collecting OMP^-, ICAM^-, iLR^- and Sus4^- cells (quadruple negative). (H) RNA isolated from combined OMP-GFP⁺, sustentacular and basal cells or quadruple negative cells was used in qRT-PCR reactions with primers for different ORs. Actin was used as endogenous control. (See also Figure S4). All above experiments were performed in, at least, two biological replicates with similar results. Values shown here are the mean of triplicate qPCRs. Error bars represent the SEM.

Figure 5. Expression and ChIP-qPCR analysis of Ngn1⁺ cells

(A,B) Sections of the MOE from an adult Ngn1-GFP mouse stained with antibodies against ORs MOR28 and M50 (A), or M71 (B) (all in red). GBCs and immature neurons are the GFP⁺ cells (green). (C) Expression levels of OR transcripts, as determined by Illumina mRNA-seq, are quantified by normalized RPKM (reads per kilobase of exon model per million mapped reads). RPKM increases with radius from the center of the figure, clamped at a maximum of 1. Each radial bar represents the level of expression of a single OR. For each OR gene, red indicates the expression level in Ngn1⁺ neurons and blue indicates the expression level in OMP⁺ neurons. ORs are sorted first by chromosome, indicated by the number or letter exterior to each wedge of the figure, and then by increasing gene start position within each chromosome. The two classes of ORs (Class I and II) are demarcated by background shading. See supplemental Excel file for detailed results. (D) Boxplot representation of all ORs gene expression in OMP⁺ and Ngn1⁺ cells showing that there is an ~8-fold difference between the two cell types. Per gene expression was calculated in log2(RPKM) units. (E) ChIP-qPCR analysis for H3K9me3 and H4K20me3 with isolated Ngn1-GFP⁺ cells. (F) ICAM⁺, Ngn1⁺ and OMP⁺ cells were sorted from the MOE of adult mice, their nuclei were extracted, digested with DNase I and analyzed by agarose gel electrophoresis and Southern blot with a degenerate OR or a ribosomal probe. (See also Figure S5).

Figure 6. The active OR allele is not enriched for H3K9me3 or H4K20me3, but it is marked with H3K4me3

Heterozygote P2-IRES-GFP and MOR28-IRES-GFP mice were used to isolate GFP⁺ and GFP^- cells by FACS. ChIP experiments were performed in these cells with antibodies against H3K9me3 and H4K20me3, or H3K4me3. (A) The location of the primers used in this experiment is depicted here. Primers for the GFP sequence were used to specifically monitor the active allele, while the ORWT primers specifically amplified the inactive allele. (B,C) GFP is hypomethylated on H3K9 (B) in the GFP⁺ cells, where it is transcribed, but not in the GFP^- cells (C), where this P2 allele is inactive. The inactive allele, amplified specifically by the p2WT primers, shows high enrichment for H3K9me3 in both GFP⁺ and GFP^- populations. Omp and Tbp are used as negative and Zfp560 and repeats (major satellite) as positive controls (D,E) As above, but the GFP⁺ cells from MOR28-IRES-GFP heterozygous mice were subject to a second round of FACS to yield a >95% pure population and then used for H3K9me3 ChIPs. (F) Similar results were obtained for the H4K20me3 ChIPs with P2-GFP sorted cells. (G) We repeated the same ChIP-qPCR experiment with an antibody against H3K4me3. There is significant enrichment for H3K4me3 throughout the P2 gene, but not on the neighboring P3 gene or a distant OR (Olfr177) in the GFP⁺ cells. As expected, there was no H3K4me3 on the P2 gene, or any other OR gene, in the GFP^- cells. (See also Figure S6). Values are the mean of triplicate qPCR. Error bars represent the SEM.

Figure 7. Tissue-specific OR modifications are associated with OR-like transgene expression

(A) Graphic representation of the Olfr459 locus and the OMP-LacZ insertion site located 55 kb away. Positions marked A, B, and C depict assayed regions in the qPCR analysis below. (B) ChIP-qPCRs with chromatin from the MOE of wild type mouse show that the Olfr459 is enriched for H3K9me3 and H4K20me3 and both modifications appear to extend to the insertion site. (C) ChIP-qPCR analysis of the MOE and liver from OMP-LacZ positive animals. Both H3K9me3 and H4K20me3 show MOE-specific deposition on Olfr459, the OMP-LacZ transgene, and the regions proximal to these loci. (D) X-gal stains of lateral whole mounts of the nasal cavities from hemizygote and homozygote OMP-LacZ animals. N, number of biological replicates. *p < 10^-4, Student's t-test. (See also Figure S7).